xref: /linux/fs/nfs/nfs4proc.c (revision 827634added7f38b7d724cab1dccdb2b004c13c3)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57 
58 #include "nfs4_fs.h"
59 #include "delegation.h"
60 #include "internal.h"
61 #include "iostat.h"
62 #include "callback.h"
63 #include "pnfs.h"
64 #include "netns.h"
65 #include "nfs4idmap.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 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
190 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
191 	FATTR4_WORD2_SECURITY_LABEL
192 #endif
193 };
194 
195 static const u32 nfs4_pnfs_open_bitmap[3] = {
196 	FATTR4_WORD0_TYPE
197 	| FATTR4_WORD0_CHANGE
198 	| FATTR4_WORD0_SIZE
199 	| FATTR4_WORD0_FSID
200 	| FATTR4_WORD0_FILEID,
201 	FATTR4_WORD1_MODE
202 	| FATTR4_WORD1_NUMLINKS
203 	| FATTR4_WORD1_OWNER
204 	| FATTR4_WORD1_OWNER_GROUP
205 	| FATTR4_WORD1_RAWDEV
206 	| FATTR4_WORD1_SPACE_USED
207 	| FATTR4_WORD1_TIME_ACCESS
208 	| FATTR4_WORD1_TIME_METADATA
209 	| FATTR4_WORD1_TIME_MODIFY,
210 	FATTR4_WORD2_MDSTHRESHOLD
211 };
212 
213 static const u32 nfs4_open_noattr_bitmap[3] = {
214 	FATTR4_WORD0_TYPE
215 	| FATTR4_WORD0_CHANGE
216 	| FATTR4_WORD0_FILEID,
217 };
218 
219 const u32 nfs4_statfs_bitmap[3] = {
220 	FATTR4_WORD0_FILES_AVAIL
221 	| FATTR4_WORD0_FILES_FREE
222 	| FATTR4_WORD0_FILES_TOTAL,
223 	FATTR4_WORD1_SPACE_AVAIL
224 	| FATTR4_WORD1_SPACE_FREE
225 	| FATTR4_WORD1_SPACE_TOTAL
226 };
227 
228 const u32 nfs4_pathconf_bitmap[3] = {
229 	FATTR4_WORD0_MAXLINK
230 	| FATTR4_WORD0_MAXNAME,
231 	0
232 };
233 
234 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
235 			| FATTR4_WORD0_MAXREAD
236 			| FATTR4_WORD0_MAXWRITE
237 			| FATTR4_WORD0_LEASE_TIME,
238 			FATTR4_WORD1_TIME_DELTA
239 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
240 			FATTR4_WORD2_LAYOUT_BLKSIZE
241 };
242 
243 const u32 nfs4_fs_locations_bitmap[3] = {
244 	FATTR4_WORD0_TYPE
245 	| FATTR4_WORD0_CHANGE
246 	| FATTR4_WORD0_SIZE
247 	| FATTR4_WORD0_FSID
248 	| FATTR4_WORD0_FILEID
249 	| FATTR4_WORD0_FS_LOCATIONS,
250 	FATTR4_WORD1_MODE
251 	| FATTR4_WORD1_NUMLINKS
252 	| FATTR4_WORD1_OWNER
253 	| FATTR4_WORD1_OWNER_GROUP
254 	| FATTR4_WORD1_RAWDEV
255 	| FATTR4_WORD1_SPACE_USED
256 	| FATTR4_WORD1_TIME_ACCESS
257 	| FATTR4_WORD1_TIME_METADATA
258 	| FATTR4_WORD1_TIME_MODIFY
259 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
260 };
261 
262 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
263 		struct nfs4_readdir_arg *readdir)
264 {
265 	__be32 *start, *p;
266 
267 	if (cookie > 2) {
268 		readdir->cookie = cookie;
269 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
270 		return;
271 	}
272 
273 	readdir->cookie = 0;
274 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
275 	if (cookie == 2)
276 		return;
277 
278 	/*
279 	 * NFSv4 servers do not return entries for '.' and '..'
280 	 * Therefore, we fake these entries here.  We let '.'
281 	 * have cookie 0 and '..' have cookie 1.  Note that
282 	 * when talking to the server, we always send cookie 0
283 	 * instead of 1 or 2.
284 	 */
285 	start = p = kmap_atomic(*readdir->pages);
286 
287 	if (cookie == 0) {
288 		*p++ = xdr_one;                                  /* next */
289 		*p++ = xdr_zero;                   /* cookie, first word */
290 		*p++ = xdr_one;                   /* cookie, second word */
291 		*p++ = xdr_one;                             /* entry len */
292 		memcpy(p, ".\0\0\0", 4);                        /* entry */
293 		p++;
294 		*p++ = xdr_one;                         /* bitmap length */
295 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
296 		*p++ = htonl(8);              /* attribute buffer length */
297 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
298 	}
299 
300 	*p++ = xdr_one;                                  /* next */
301 	*p++ = xdr_zero;                   /* cookie, first word */
302 	*p++ = xdr_two;                   /* cookie, second word */
303 	*p++ = xdr_two;                             /* entry len */
304 	memcpy(p, "..\0\0", 4);                         /* entry */
305 	p++;
306 	*p++ = xdr_one;                         /* bitmap length */
307 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
308 	*p++ = htonl(8);              /* attribute buffer length */
309 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
310 
311 	readdir->pgbase = (char *)p - (char *)start;
312 	readdir->count -= readdir->pgbase;
313 	kunmap_atomic(start);
314 }
315 
316 static long nfs4_update_delay(long *timeout)
317 {
318 	long ret;
319 	if (!timeout)
320 		return NFS4_POLL_RETRY_MAX;
321 	if (*timeout <= 0)
322 		*timeout = NFS4_POLL_RETRY_MIN;
323 	if (*timeout > NFS4_POLL_RETRY_MAX)
324 		*timeout = NFS4_POLL_RETRY_MAX;
325 	ret = *timeout;
326 	*timeout <<= 1;
327 	return ret;
328 }
329 
330 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
331 {
332 	int res = 0;
333 
334 	might_sleep();
335 
336 	freezable_schedule_timeout_killable_unsafe(
337 		nfs4_update_delay(timeout));
338 	if (fatal_signal_pending(current))
339 		res = -ERESTARTSYS;
340 	return res;
341 }
342 
343 /* This is the error handling routine for processes that are allowed
344  * to sleep.
345  */
346 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
347 {
348 	struct nfs_client *clp = server->nfs_client;
349 	struct nfs4_state *state = exception->state;
350 	struct inode *inode = exception->inode;
351 	int ret = errorcode;
352 
353 	exception->retry = 0;
354 	switch(errorcode) {
355 		case 0:
356 			return 0;
357 		case -NFS4ERR_OPENMODE:
358 			if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
359 				nfs4_inode_return_delegation(inode);
360 				exception->retry = 1;
361 				return 0;
362 			}
363 			if (state == NULL)
364 				break;
365 			ret = nfs4_schedule_stateid_recovery(server, state);
366 			if (ret < 0)
367 				break;
368 			goto wait_on_recovery;
369 		case -NFS4ERR_DELEG_REVOKED:
370 		case -NFS4ERR_ADMIN_REVOKED:
371 		case -NFS4ERR_BAD_STATEID:
372 			if (state == NULL)
373 				break;
374 			ret = nfs4_schedule_stateid_recovery(server, state);
375 			if (ret < 0)
376 				break;
377 			goto wait_on_recovery;
378 		case -NFS4ERR_EXPIRED:
379 			if (state != NULL) {
380 				ret = nfs4_schedule_stateid_recovery(server, state);
381 				if (ret < 0)
382 					break;
383 			}
384 		case -NFS4ERR_STALE_STATEID:
385 		case -NFS4ERR_STALE_CLIENTID:
386 			nfs4_schedule_lease_recovery(clp);
387 			goto wait_on_recovery;
388 		case -NFS4ERR_MOVED:
389 			ret = nfs4_schedule_migration_recovery(server);
390 			if (ret < 0)
391 				break;
392 			goto wait_on_recovery;
393 		case -NFS4ERR_LEASE_MOVED:
394 			nfs4_schedule_lease_moved_recovery(clp);
395 			goto wait_on_recovery;
396 #if defined(CONFIG_NFS_V4_1)
397 		case -NFS4ERR_BADSESSION:
398 		case -NFS4ERR_BADSLOT:
399 		case -NFS4ERR_BAD_HIGH_SLOT:
400 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
401 		case -NFS4ERR_DEADSESSION:
402 		case -NFS4ERR_SEQ_FALSE_RETRY:
403 		case -NFS4ERR_SEQ_MISORDERED:
404 			dprintk("%s ERROR: %d Reset session\n", __func__,
405 				errorcode);
406 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
407 			goto wait_on_recovery;
408 #endif /* defined(CONFIG_NFS_V4_1) */
409 		case -NFS4ERR_FILE_OPEN:
410 			if (exception->timeout > HZ) {
411 				/* We have retried a decent amount, time to
412 				 * fail
413 				 */
414 				ret = -EBUSY;
415 				break;
416 			}
417 		case -NFS4ERR_GRACE:
418 		case -NFS4ERR_DELAY:
419 			ret = nfs4_delay(server->client, &exception->timeout);
420 			if (ret != 0)
421 				break;
422 		case -NFS4ERR_RETRY_UNCACHED_REP:
423 		case -NFS4ERR_OLD_STATEID:
424 			exception->retry = 1;
425 			break;
426 		case -NFS4ERR_BADOWNER:
427 			/* The following works around a Linux server bug! */
428 		case -NFS4ERR_BADNAME:
429 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
430 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
431 				exception->retry = 1;
432 				printk(KERN_WARNING "NFS: v4 server %s "
433 						"does not accept raw "
434 						"uid/gids. "
435 						"Reenabling the idmapper.\n",
436 						server->nfs_client->cl_hostname);
437 			}
438 	}
439 	/* We failed to handle the error */
440 	return nfs4_map_errors(ret);
441 wait_on_recovery:
442 	ret = nfs4_wait_clnt_recover(clp);
443 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
444 		return -EIO;
445 	if (ret == 0)
446 		exception->retry = 1;
447 	return ret;
448 }
449 
450 /*
451  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
452  * or 'false' otherwise.
453  */
454 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
455 {
456 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
457 
458 	if (flavor == RPC_AUTH_GSS_KRB5I ||
459 	    flavor == RPC_AUTH_GSS_KRB5P)
460 		return true;
461 
462 	return false;
463 }
464 
465 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
466 {
467 	spin_lock(&clp->cl_lock);
468 	if (time_before(clp->cl_last_renewal,timestamp))
469 		clp->cl_last_renewal = timestamp;
470 	spin_unlock(&clp->cl_lock);
471 }
472 
473 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
474 {
475 	do_renew_lease(server->nfs_client, timestamp);
476 }
477 
478 struct nfs4_call_sync_data {
479 	const struct nfs_server *seq_server;
480 	struct nfs4_sequence_args *seq_args;
481 	struct nfs4_sequence_res *seq_res;
482 };
483 
484 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
485 			       struct nfs4_sequence_res *res, int cache_reply)
486 {
487 	args->sa_slot = NULL;
488 	args->sa_cache_this = cache_reply;
489 	args->sa_privileged = 0;
490 
491 	res->sr_slot = NULL;
492 }
493 
494 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
495 {
496 	args->sa_privileged = 1;
497 }
498 
499 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
500 			 struct nfs4_sequence_args *args,
501 			 struct nfs4_sequence_res *res,
502 			 struct rpc_task *task)
503 {
504 	struct nfs4_slot *slot;
505 
506 	/* slot already allocated? */
507 	if (res->sr_slot != NULL)
508 		goto out_start;
509 
510 	spin_lock(&tbl->slot_tbl_lock);
511 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
512 		goto out_sleep;
513 
514 	slot = nfs4_alloc_slot(tbl);
515 	if (IS_ERR(slot)) {
516 		if (slot == ERR_PTR(-ENOMEM))
517 			task->tk_timeout = HZ >> 2;
518 		goto out_sleep;
519 	}
520 	spin_unlock(&tbl->slot_tbl_lock);
521 
522 	args->sa_slot = slot;
523 	res->sr_slot = slot;
524 
525 out_start:
526 	rpc_call_start(task);
527 	return 0;
528 
529 out_sleep:
530 	if (args->sa_privileged)
531 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
532 				NULL, RPC_PRIORITY_PRIVILEGED);
533 	else
534 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
535 	spin_unlock(&tbl->slot_tbl_lock);
536 	return -EAGAIN;
537 }
538 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
539 
540 static int nfs40_sequence_done(struct rpc_task *task,
541 			       struct nfs4_sequence_res *res)
542 {
543 	struct nfs4_slot *slot = res->sr_slot;
544 	struct nfs4_slot_table *tbl;
545 
546 	if (slot == NULL)
547 		goto out;
548 
549 	tbl = slot->table;
550 	spin_lock(&tbl->slot_tbl_lock);
551 	if (!nfs41_wake_and_assign_slot(tbl, slot))
552 		nfs4_free_slot(tbl, slot);
553 	spin_unlock(&tbl->slot_tbl_lock);
554 
555 	res->sr_slot = NULL;
556 out:
557 	return 1;
558 }
559 
560 #if defined(CONFIG_NFS_V4_1)
561 
562 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
563 {
564 	struct nfs4_session *session;
565 	struct nfs4_slot_table *tbl;
566 	struct nfs4_slot *slot = res->sr_slot;
567 	bool send_new_highest_used_slotid = false;
568 
569 	tbl = slot->table;
570 	session = tbl->session;
571 
572 	spin_lock(&tbl->slot_tbl_lock);
573 	/* Be nice to the server: try to ensure that the last transmitted
574 	 * value for highest_user_slotid <= target_highest_slotid
575 	 */
576 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
577 		send_new_highest_used_slotid = true;
578 
579 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
580 		send_new_highest_used_slotid = false;
581 		goto out_unlock;
582 	}
583 	nfs4_free_slot(tbl, slot);
584 
585 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
586 		send_new_highest_used_slotid = false;
587 out_unlock:
588 	spin_unlock(&tbl->slot_tbl_lock);
589 	res->sr_slot = NULL;
590 	if (send_new_highest_used_slotid)
591 		nfs41_server_notify_highest_slotid_update(session->clp);
592 }
593 
594 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
595 {
596 	struct nfs4_session *session;
597 	struct nfs4_slot *slot = res->sr_slot;
598 	struct nfs_client *clp;
599 	bool interrupted = false;
600 	int ret = 1;
601 
602 	if (slot == NULL)
603 		goto out_noaction;
604 	/* don't increment the sequence number if the task wasn't sent */
605 	if (!RPC_WAS_SENT(task))
606 		goto out;
607 
608 	session = slot->table->session;
609 
610 	if (slot->interrupted) {
611 		slot->interrupted = 0;
612 		interrupted = true;
613 	}
614 
615 	trace_nfs4_sequence_done(session, res);
616 	/* Check the SEQUENCE operation status */
617 	switch (res->sr_status) {
618 	case 0:
619 		/* Update the slot's sequence and clientid lease timer */
620 		++slot->seq_nr;
621 		clp = session->clp;
622 		do_renew_lease(clp, res->sr_timestamp);
623 		/* Check sequence flags */
624 		if (res->sr_status_flags != 0)
625 			nfs4_schedule_lease_recovery(clp);
626 		nfs41_update_target_slotid(slot->table, slot, res);
627 		break;
628 	case 1:
629 		/*
630 		 * sr_status remains 1 if an RPC level error occurred.
631 		 * The server may or may not have processed the sequence
632 		 * operation..
633 		 * Mark the slot as having hosted an interrupted RPC call.
634 		 */
635 		slot->interrupted = 1;
636 		goto out;
637 	case -NFS4ERR_DELAY:
638 		/* The server detected a resend of the RPC call and
639 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
640 		 * of RFC5661.
641 		 */
642 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
643 			__func__,
644 			slot->slot_nr,
645 			slot->seq_nr);
646 		goto out_retry;
647 	case -NFS4ERR_BADSLOT:
648 		/*
649 		 * The slot id we used was probably retired. Try again
650 		 * using a different slot id.
651 		 */
652 		goto retry_nowait;
653 	case -NFS4ERR_SEQ_MISORDERED:
654 		/*
655 		 * Was the last operation on this sequence interrupted?
656 		 * If so, retry after bumping the sequence number.
657 		 */
658 		if (interrupted) {
659 			++slot->seq_nr;
660 			goto retry_nowait;
661 		}
662 		/*
663 		 * Could this slot have been previously retired?
664 		 * If so, then the server may be expecting seq_nr = 1!
665 		 */
666 		if (slot->seq_nr != 1) {
667 			slot->seq_nr = 1;
668 			goto retry_nowait;
669 		}
670 		break;
671 	case -NFS4ERR_SEQ_FALSE_RETRY:
672 		++slot->seq_nr;
673 		goto retry_nowait;
674 	default:
675 		/* Just update the slot sequence no. */
676 		++slot->seq_nr;
677 	}
678 out:
679 	/* The session may be reset by one of the error handlers. */
680 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
681 	nfs41_sequence_free_slot(res);
682 out_noaction:
683 	return ret;
684 retry_nowait:
685 	if (rpc_restart_call_prepare(task)) {
686 		task->tk_status = 0;
687 		ret = 0;
688 	}
689 	goto out;
690 out_retry:
691 	if (!rpc_restart_call(task))
692 		goto out;
693 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
694 	return 0;
695 }
696 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
697 
698 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
699 {
700 	if (res->sr_slot == NULL)
701 		return 1;
702 	if (!res->sr_slot->table->session)
703 		return nfs40_sequence_done(task, res);
704 	return nfs41_sequence_done(task, res);
705 }
706 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
707 
708 int nfs41_setup_sequence(struct nfs4_session *session,
709 				struct nfs4_sequence_args *args,
710 				struct nfs4_sequence_res *res,
711 				struct rpc_task *task)
712 {
713 	struct nfs4_slot *slot;
714 	struct nfs4_slot_table *tbl;
715 
716 	dprintk("--> %s\n", __func__);
717 	/* slot already allocated? */
718 	if (res->sr_slot != NULL)
719 		goto out_success;
720 
721 	tbl = &session->fc_slot_table;
722 
723 	task->tk_timeout = 0;
724 
725 	spin_lock(&tbl->slot_tbl_lock);
726 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
727 	    !args->sa_privileged) {
728 		/* The state manager will wait until the slot table is empty */
729 		dprintk("%s session is draining\n", __func__);
730 		goto out_sleep;
731 	}
732 
733 	slot = nfs4_alloc_slot(tbl);
734 	if (IS_ERR(slot)) {
735 		/* If out of memory, try again in 1/4 second */
736 		if (slot == ERR_PTR(-ENOMEM))
737 			task->tk_timeout = HZ >> 2;
738 		dprintk("<-- %s: no free slots\n", __func__);
739 		goto out_sleep;
740 	}
741 	spin_unlock(&tbl->slot_tbl_lock);
742 
743 	args->sa_slot = slot;
744 
745 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
746 			slot->slot_nr, slot->seq_nr);
747 
748 	res->sr_slot = slot;
749 	res->sr_timestamp = jiffies;
750 	res->sr_status_flags = 0;
751 	/*
752 	 * sr_status is only set in decode_sequence, and so will remain
753 	 * set to 1 if an rpc level failure occurs.
754 	 */
755 	res->sr_status = 1;
756 	trace_nfs4_setup_sequence(session, args);
757 out_success:
758 	rpc_call_start(task);
759 	return 0;
760 out_sleep:
761 	/* Privileged tasks are queued with top priority */
762 	if (args->sa_privileged)
763 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
764 				NULL, RPC_PRIORITY_PRIVILEGED);
765 	else
766 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
767 	spin_unlock(&tbl->slot_tbl_lock);
768 	return -EAGAIN;
769 }
770 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
771 
772 static int nfs4_setup_sequence(const struct nfs_server *server,
773 			       struct nfs4_sequence_args *args,
774 			       struct nfs4_sequence_res *res,
775 			       struct rpc_task *task)
776 {
777 	struct nfs4_session *session = nfs4_get_session(server);
778 	int ret = 0;
779 
780 	if (!session)
781 		return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
782 					    args, res, task);
783 
784 	dprintk("--> %s clp %p session %p sr_slot %u\n",
785 		__func__, session->clp, session, res->sr_slot ?
786 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
787 
788 	ret = nfs41_setup_sequence(session, args, res, task);
789 
790 	dprintk("<-- %s status=%d\n", __func__, ret);
791 	return ret;
792 }
793 
794 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
795 {
796 	struct nfs4_call_sync_data *data = calldata;
797 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
798 
799 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
800 
801 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
802 }
803 
804 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
805 {
806 	struct nfs4_call_sync_data *data = calldata;
807 
808 	nfs41_sequence_done(task, data->seq_res);
809 }
810 
811 static const struct rpc_call_ops nfs41_call_sync_ops = {
812 	.rpc_call_prepare = nfs41_call_sync_prepare,
813 	.rpc_call_done = nfs41_call_sync_done,
814 };
815 
816 #else	/* !CONFIG_NFS_V4_1 */
817 
818 static int nfs4_setup_sequence(const struct nfs_server *server,
819 			       struct nfs4_sequence_args *args,
820 			       struct nfs4_sequence_res *res,
821 			       struct rpc_task *task)
822 {
823 	return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
824 				    args, res, task);
825 }
826 
827 int nfs4_sequence_done(struct rpc_task *task,
828 		       struct nfs4_sequence_res *res)
829 {
830 	return nfs40_sequence_done(task, res);
831 }
832 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
833 
834 #endif	/* !CONFIG_NFS_V4_1 */
835 
836 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
837 {
838 	struct nfs4_call_sync_data *data = calldata;
839 	nfs4_setup_sequence(data->seq_server,
840 				data->seq_args, data->seq_res, task);
841 }
842 
843 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
844 {
845 	struct nfs4_call_sync_data *data = calldata;
846 	nfs4_sequence_done(task, data->seq_res);
847 }
848 
849 static const struct rpc_call_ops nfs40_call_sync_ops = {
850 	.rpc_call_prepare = nfs40_call_sync_prepare,
851 	.rpc_call_done = nfs40_call_sync_done,
852 };
853 
854 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
855 				   struct nfs_server *server,
856 				   struct rpc_message *msg,
857 				   struct nfs4_sequence_args *args,
858 				   struct nfs4_sequence_res *res)
859 {
860 	int ret;
861 	struct rpc_task *task;
862 	struct nfs_client *clp = server->nfs_client;
863 	struct nfs4_call_sync_data data = {
864 		.seq_server = server,
865 		.seq_args = args,
866 		.seq_res = res,
867 	};
868 	struct rpc_task_setup task_setup = {
869 		.rpc_client = clnt,
870 		.rpc_message = msg,
871 		.callback_ops = clp->cl_mvops->call_sync_ops,
872 		.callback_data = &data
873 	};
874 
875 	task = rpc_run_task(&task_setup);
876 	if (IS_ERR(task))
877 		ret = PTR_ERR(task);
878 	else {
879 		ret = task->tk_status;
880 		rpc_put_task(task);
881 	}
882 	return ret;
883 }
884 
885 int nfs4_call_sync(struct rpc_clnt *clnt,
886 		   struct nfs_server *server,
887 		   struct rpc_message *msg,
888 		   struct nfs4_sequence_args *args,
889 		   struct nfs4_sequence_res *res,
890 		   int cache_reply)
891 {
892 	nfs4_init_sequence(args, res, cache_reply);
893 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
894 }
895 
896 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
897 {
898 	struct nfs_inode *nfsi = NFS_I(dir);
899 
900 	spin_lock(&dir->i_lock);
901 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
902 	if (!cinfo->atomic || cinfo->before != dir->i_version)
903 		nfs_force_lookup_revalidate(dir);
904 	dir->i_version = cinfo->after;
905 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
906 	nfs_fscache_invalidate(dir);
907 	spin_unlock(&dir->i_lock);
908 }
909 
910 struct nfs4_opendata {
911 	struct kref kref;
912 	struct nfs_openargs o_arg;
913 	struct nfs_openres o_res;
914 	struct nfs_open_confirmargs c_arg;
915 	struct nfs_open_confirmres c_res;
916 	struct nfs4_string owner_name;
917 	struct nfs4_string group_name;
918 	struct nfs_fattr f_attr;
919 	struct nfs4_label *f_label;
920 	struct dentry *dir;
921 	struct dentry *dentry;
922 	struct nfs4_state_owner *owner;
923 	struct nfs4_state *state;
924 	struct iattr attrs;
925 	unsigned long timestamp;
926 	unsigned int rpc_done : 1;
927 	unsigned int file_created : 1;
928 	unsigned int is_recover : 1;
929 	int rpc_status;
930 	int cancelled;
931 };
932 
933 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
934 		int err, struct nfs4_exception *exception)
935 {
936 	if (err != -EINVAL)
937 		return false;
938 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
939 		return false;
940 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
941 	exception->retry = 1;
942 	return true;
943 }
944 
945 static u32
946 nfs4_map_atomic_open_share(struct nfs_server *server,
947 		fmode_t fmode, int openflags)
948 {
949 	u32 res = 0;
950 
951 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
952 	case FMODE_READ:
953 		res = NFS4_SHARE_ACCESS_READ;
954 		break;
955 	case FMODE_WRITE:
956 		res = NFS4_SHARE_ACCESS_WRITE;
957 		break;
958 	case FMODE_READ|FMODE_WRITE:
959 		res = NFS4_SHARE_ACCESS_BOTH;
960 	}
961 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
962 		goto out;
963 	/* Want no delegation if we're using O_DIRECT */
964 	if (openflags & O_DIRECT)
965 		res |= NFS4_SHARE_WANT_NO_DELEG;
966 out:
967 	return res;
968 }
969 
970 static enum open_claim_type4
971 nfs4_map_atomic_open_claim(struct nfs_server *server,
972 		enum open_claim_type4 claim)
973 {
974 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
975 		return claim;
976 	switch (claim) {
977 	default:
978 		return claim;
979 	case NFS4_OPEN_CLAIM_FH:
980 		return NFS4_OPEN_CLAIM_NULL;
981 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
982 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
983 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
984 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
985 	}
986 }
987 
988 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
989 {
990 	p->o_res.f_attr = &p->f_attr;
991 	p->o_res.f_label = p->f_label;
992 	p->o_res.seqid = p->o_arg.seqid;
993 	p->c_res.seqid = p->c_arg.seqid;
994 	p->o_res.server = p->o_arg.server;
995 	p->o_res.access_request = p->o_arg.access;
996 	nfs_fattr_init(&p->f_attr);
997 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
998 }
999 
1000 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1001 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1002 		const struct iattr *attrs,
1003 		struct nfs4_label *label,
1004 		enum open_claim_type4 claim,
1005 		gfp_t gfp_mask)
1006 {
1007 	struct dentry *parent = dget_parent(dentry);
1008 	struct inode *dir = d_inode(parent);
1009 	struct nfs_server *server = NFS_SERVER(dir);
1010 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1011 	struct nfs4_opendata *p;
1012 
1013 	p = kzalloc(sizeof(*p), gfp_mask);
1014 	if (p == NULL)
1015 		goto err;
1016 
1017 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1018 	if (IS_ERR(p->f_label))
1019 		goto err_free_p;
1020 
1021 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1022 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1023 	if (IS_ERR(p->o_arg.seqid))
1024 		goto err_free_label;
1025 	nfs_sb_active(dentry->d_sb);
1026 	p->dentry = dget(dentry);
1027 	p->dir = parent;
1028 	p->owner = sp;
1029 	atomic_inc(&sp->so_count);
1030 	p->o_arg.open_flags = flags;
1031 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1032 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1033 			fmode, flags);
1034 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1035 	 * will return permission denied for all bits until close */
1036 	if (!(flags & O_EXCL)) {
1037 		/* ask server to check for all possible rights as results
1038 		 * are cached */
1039 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1040 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1041 	}
1042 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1043 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1044 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1045 	p->o_arg.name = &dentry->d_name;
1046 	p->o_arg.server = server;
1047 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1048 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1049 	p->o_arg.label = label;
1050 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1051 	switch (p->o_arg.claim) {
1052 	case NFS4_OPEN_CLAIM_NULL:
1053 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1054 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1055 		p->o_arg.fh = NFS_FH(dir);
1056 		break;
1057 	case NFS4_OPEN_CLAIM_PREVIOUS:
1058 	case NFS4_OPEN_CLAIM_FH:
1059 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1060 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1061 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1062 	}
1063 	if (attrs != NULL && attrs->ia_valid != 0) {
1064 		__u32 verf[2];
1065 
1066 		p->o_arg.u.attrs = &p->attrs;
1067 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1068 
1069 		verf[0] = jiffies;
1070 		verf[1] = current->pid;
1071 		memcpy(p->o_arg.u.verifier.data, verf,
1072 				sizeof(p->o_arg.u.verifier.data));
1073 	}
1074 	p->c_arg.fh = &p->o_res.fh;
1075 	p->c_arg.stateid = &p->o_res.stateid;
1076 	p->c_arg.seqid = p->o_arg.seqid;
1077 	nfs4_init_opendata_res(p);
1078 	kref_init(&p->kref);
1079 	return p;
1080 
1081 err_free_label:
1082 	nfs4_label_free(p->f_label);
1083 err_free_p:
1084 	kfree(p);
1085 err:
1086 	dput(parent);
1087 	return NULL;
1088 }
1089 
1090 static void nfs4_opendata_free(struct kref *kref)
1091 {
1092 	struct nfs4_opendata *p = container_of(kref,
1093 			struct nfs4_opendata, kref);
1094 	struct super_block *sb = p->dentry->d_sb;
1095 
1096 	nfs_free_seqid(p->o_arg.seqid);
1097 	if (p->state != NULL)
1098 		nfs4_put_open_state(p->state);
1099 	nfs4_put_state_owner(p->owner);
1100 
1101 	nfs4_label_free(p->f_label);
1102 
1103 	dput(p->dir);
1104 	dput(p->dentry);
1105 	nfs_sb_deactive(sb);
1106 	nfs_fattr_free_names(&p->f_attr);
1107 	kfree(p->f_attr.mdsthreshold);
1108 	kfree(p);
1109 }
1110 
1111 static void nfs4_opendata_put(struct nfs4_opendata *p)
1112 {
1113 	if (p != NULL)
1114 		kref_put(&p->kref, nfs4_opendata_free);
1115 }
1116 
1117 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1118 {
1119 	int ret;
1120 
1121 	ret = rpc_wait_for_completion_task(task);
1122 	return ret;
1123 }
1124 
1125 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1126 {
1127 	int ret = 0;
1128 
1129 	if (open_mode & (O_EXCL|O_TRUNC))
1130 		goto out;
1131 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1132 		case FMODE_READ:
1133 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1134 				&& state->n_rdonly != 0;
1135 			break;
1136 		case FMODE_WRITE:
1137 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1138 				&& state->n_wronly != 0;
1139 			break;
1140 		case FMODE_READ|FMODE_WRITE:
1141 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1142 				&& state->n_rdwr != 0;
1143 	}
1144 out:
1145 	return ret;
1146 }
1147 
1148 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1149 {
1150 	if (delegation == NULL)
1151 		return 0;
1152 	if ((delegation->type & fmode) != fmode)
1153 		return 0;
1154 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1155 		return 0;
1156 	nfs_mark_delegation_referenced(delegation);
1157 	return 1;
1158 }
1159 
1160 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1161 {
1162 	switch (fmode) {
1163 		case FMODE_WRITE:
1164 			state->n_wronly++;
1165 			break;
1166 		case FMODE_READ:
1167 			state->n_rdonly++;
1168 			break;
1169 		case FMODE_READ|FMODE_WRITE:
1170 			state->n_rdwr++;
1171 	}
1172 	nfs4_state_set_mode_locked(state, state->state | fmode);
1173 }
1174 
1175 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1176 {
1177 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1178 	bool need_recover = false;
1179 
1180 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1181 		need_recover = true;
1182 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1183 		need_recover = true;
1184 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1185 		need_recover = true;
1186 	if (need_recover)
1187 		nfs4_state_mark_reclaim_nograce(clp, state);
1188 }
1189 
1190 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1191 		nfs4_stateid *stateid)
1192 {
1193 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1194 		return true;
1195 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1196 		nfs_test_and_clear_all_open_stateid(state);
1197 		return true;
1198 	}
1199 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1200 		return true;
1201 	return false;
1202 }
1203 
1204 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1205 {
1206 	if (state->n_wronly)
1207 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1208 	if (state->n_rdonly)
1209 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1210 	if (state->n_rdwr)
1211 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1212 }
1213 
1214 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1215 		nfs4_stateid *stateid, fmode_t fmode)
1216 {
1217 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1218 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1219 	case FMODE_WRITE:
1220 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1221 		break;
1222 	case FMODE_READ:
1223 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1224 		break;
1225 	case 0:
1226 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1227 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1228 		clear_bit(NFS_OPEN_STATE, &state->flags);
1229 	}
1230 	if (stateid == NULL)
1231 		return;
1232 	/* Handle races with OPEN */
1233 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1234 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1235 		nfs_resync_open_stateid_locked(state);
1236 		return;
1237 	}
1238 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1239 		nfs4_stateid_copy(&state->stateid, stateid);
1240 	nfs4_stateid_copy(&state->open_stateid, stateid);
1241 }
1242 
1243 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1244 {
1245 	write_seqlock(&state->seqlock);
1246 	nfs_clear_open_stateid_locked(state, stateid, fmode);
1247 	write_sequnlock(&state->seqlock);
1248 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1249 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1250 }
1251 
1252 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1253 {
1254 	switch (fmode) {
1255 		case FMODE_READ:
1256 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1257 			break;
1258 		case FMODE_WRITE:
1259 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1260 			break;
1261 		case FMODE_READ|FMODE_WRITE:
1262 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1263 	}
1264 	if (!nfs_need_update_open_stateid(state, stateid))
1265 		return;
1266 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1267 		nfs4_stateid_copy(&state->stateid, stateid);
1268 	nfs4_stateid_copy(&state->open_stateid, stateid);
1269 }
1270 
1271 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1272 {
1273 	/*
1274 	 * Protect the call to nfs4_state_set_mode_locked and
1275 	 * serialise the stateid update
1276 	 */
1277 	write_seqlock(&state->seqlock);
1278 	if (deleg_stateid != NULL) {
1279 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1280 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1281 	}
1282 	if (open_stateid != NULL)
1283 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1284 	write_sequnlock(&state->seqlock);
1285 	spin_lock(&state->owner->so_lock);
1286 	update_open_stateflags(state, fmode);
1287 	spin_unlock(&state->owner->so_lock);
1288 }
1289 
1290 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1291 {
1292 	struct nfs_inode *nfsi = NFS_I(state->inode);
1293 	struct nfs_delegation *deleg_cur;
1294 	int ret = 0;
1295 
1296 	fmode &= (FMODE_READ|FMODE_WRITE);
1297 
1298 	rcu_read_lock();
1299 	deleg_cur = rcu_dereference(nfsi->delegation);
1300 	if (deleg_cur == NULL)
1301 		goto no_delegation;
1302 
1303 	spin_lock(&deleg_cur->lock);
1304 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1305 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1306 	    (deleg_cur->type & fmode) != fmode)
1307 		goto no_delegation_unlock;
1308 
1309 	if (delegation == NULL)
1310 		delegation = &deleg_cur->stateid;
1311 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1312 		goto no_delegation_unlock;
1313 
1314 	nfs_mark_delegation_referenced(deleg_cur);
1315 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1316 	ret = 1;
1317 no_delegation_unlock:
1318 	spin_unlock(&deleg_cur->lock);
1319 no_delegation:
1320 	rcu_read_unlock();
1321 
1322 	if (!ret && open_stateid != NULL) {
1323 		__update_open_stateid(state, open_stateid, NULL, fmode);
1324 		ret = 1;
1325 	}
1326 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1327 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1328 
1329 	return ret;
1330 }
1331 
1332 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1333 		const nfs4_stateid *stateid)
1334 {
1335 	struct nfs4_state *state = lsp->ls_state;
1336 	bool ret = false;
1337 
1338 	spin_lock(&state->state_lock);
1339 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1340 		goto out_noupdate;
1341 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1342 		goto out_noupdate;
1343 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1344 	ret = true;
1345 out_noupdate:
1346 	spin_unlock(&state->state_lock);
1347 	return ret;
1348 }
1349 
1350 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1351 {
1352 	struct nfs_delegation *delegation;
1353 
1354 	rcu_read_lock();
1355 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1356 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1357 		rcu_read_unlock();
1358 		return;
1359 	}
1360 	rcu_read_unlock();
1361 	nfs4_inode_return_delegation(inode);
1362 }
1363 
1364 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1365 {
1366 	struct nfs4_state *state = opendata->state;
1367 	struct nfs_inode *nfsi = NFS_I(state->inode);
1368 	struct nfs_delegation *delegation;
1369 	int open_mode = opendata->o_arg.open_flags;
1370 	fmode_t fmode = opendata->o_arg.fmode;
1371 	nfs4_stateid stateid;
1372 	int ret = -EAGAIN;
1373 
1374 	for (;;) {
1375 		spin_lock(&state->owner->so_lock);
1376 		if (can_open_cached(state, fmode, open_mode)) {
1377 			update_open_stateflags(state, fmode);
1378 			spin_unlock(&state->owner->so_lock);
1379 			goto out_return_state;
1380 		}
1381 		spin_unlock(&state->owner->so_lock);
1382 		rcu_read_lock();
1383 		delegation = rcu_dereference(nfsi->delegation);
1384 		if (!can_open_delegated(delegation, fmode)) {
1385 			rcu_read_unlock();
1386 			break;
1387 		}
1388 		/* Save the delegation */
1389 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1390 		rcu_read_unlock();
1391 		nfs_release_seqid(opendata->o_arg.seqid);
1392 		if (!opendata->is_recover) {
1393 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1394 			if (ret != 0)
1395 				goto out;
1396 		}
1397 		ret = -EAGAIN;
1398 
1399 		/* Try to update the stateid using the delegation */
1400 		if (update_open_stateid(state, NULL, &stateid, fmode))
1401 			goto out_return_state;
1402 	}
1403 out:
1404 	return ERR_PTR(ret);
1405 out_return_state:
1406 	atomic_inc(&state->count);
1407 	return state;
1408 }
1409 
1410 static void
1411 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1412 {
1413 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1414 	struct nfs_delegation *delegation;
1415 	int delegation_flags = 0;
1416 
1417 	rcu_read_lock();
1418 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1419 	if (delegation)
1420 		delegation_flags = delegation->flags;
1421 	rcu_read_unlock();
1422 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1423 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1424 				   "returning a delegation for "
1425 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1426 				   clp->cl_hostname);
1427 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1428 		nfs_inode_set_delegation(state->inode,
1429 					 data->owner->so_cred,
1430 					 &data->o_res);
1431 	else
1432 		nfs_inode_reclaim_delegation(state->inode,
1433 					     data->owner->so_cred,
1434 					     &data->o_res);
1435 }
1436 
1437 /*
1438  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1439  * and update the nfs4_state.
1440  */
1441 static struct nfs4_state *
1442 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1443 {
1444 	struct inode *inode = data->state->inode;
1445 	struct nfs4_state *state = data->state;
1446 	int ret;
1447 
1448 	if (!data->rpc_done) {
1449 		if (data->rpc_status) {
1450 			ret = data->rpc_status;
1451 			goto err;
1452 		}
1453 		/* cached opens have already been processed */
1454 		goto update;
1455 	}
1456 
1457 	ret = nfs_refresh_inode(inode, &data->f_attr);
1458 	if (ret)
1459 		goto err;
1460 
1461 	if (data->o_res.delegation_type != 0)
1462 		nfs4_opendata_check_deleg(data, state);
1463 update:
1464 	update_open_stateid(state, &data->o_res.stateid, NULL,
1465 			    data->o_arg.fmode);
1466 	atomic_inc(&state->count);
1467 
1468 	return state;
1469 err:
1470 	return ERR_PTR(ret);
1471 
1472 }
1473 
1474 static struct nfs4_state *
1475 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1476 {
1477 	struct inode *inode;
1478 	struct nfs4_state *state = NULL;
1479 	int ret;
1480 
1481 	if (!data->rpc_done) {
1482 		state = nfs4_try_open_cached(data);
1483 		goto out;
1484 	}
1485 
1486 	ret = -EAGAIN;
1487 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1488 		goto err;
1489 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1490 	ret = PTR_ERR(inode);
1491 	if (IS_ERR(inode))
1492 		goto err;
1493 	ret = -ENOMEM;
1494 	state = nfs4_get_open_state(inode, data->owner);
1495 	if (state == NULL)
1496 		goto err_put_inode;
1497 	if (data->o_res.delegation_type != 0)
1498 		nfs4_opendata_check_deleg(data, state);
1499 	update_open_stateid(state, &data->o_res.stateid, NULL,
1500 			data->o_arg.fmode);
1501 	iput(inode);
1502 out:
1503 	nfs_release_seqid(data->o_arg.seqid);
1504 	return state;
1505 err_put_inode:
1506 	iput(inode);
1507 err:
1508 	return ERR_PTR(ret);
1509 }
1510 
1511 static struct nfs4_state *
1512 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1513 {
1514 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1515 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1516 	return _nfs4_opendata_to_nfs4_state(data);
1517 }
1518 
1519 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1520 {
1521 	struct nfs_inode *nfsi = NFS_I(state->inode);
1522 	struct nfs_open_context *ctx;
1523 
1524 	spin_lock(&state->inode->i_lock);
1525 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1526 		if (ctx->state != state)
1527 			continue;
1528 		get_nfs_open_context(ctx);
1529 		spin_unlock(&state->inode->i_lock);
1530 		return ctx;
1531 	}
1532 	spin_unlock(&state->inode->i_lock);
1533 	return ERR_PTR(-ENOENT);
1534 }
1535 
1536 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1537 		struct nfs4_state *state, enum open_claim_type4 claim)
1538 {
1539 	struct nfs4_opendata *opendata;
1540 
1541 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1542 			NULL, NULL, claim, GFP_NOFS);
1543 	if (opendata == NULL)
1544 		return ERR_PTR(-ENOMEM);
1545 	opendata->state = state;
1546 	atomic_inc(&state->count);
1547 	return opendata;
1548 }
1549 
1550 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1551 {
1552 	struct nfs4_state *newstate;
1553 	int ret;
1554 
1555 	opendata->o_arg.open_flags = 0;
1556 	opendata->o_arg.fmode = fmode;
1557 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1558 			NFS_SB(opendata->dentry->d_sb),
1559 			fmode, 0);
1560 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1561 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1562 	nfs4_init_opendata_res(opendata);
1563 	ret = _nfs4_recover_proc_open(opendata);
1564 	if (ret != 0)
1565 		return ret;
1566 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1567 	if (IS_ERR(newstate))
1568 		return PTR_ERR(newstate);
1569 	nfs4_close_state(newstate, fmode);
1570 	*res = newstate;
1571 	return 0;
1572 }
1573 
1574 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1575 {
1576 	struct nfs4_state *newstate;
1577 	int ret;
1578 
1579 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1580 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1581 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1582 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1583 	/* memory barrier prior to reading state->n_* */
1584 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1585 	clear_bit(NFS_OPEN_STATE, &state->flags);
1586 	smp_rmb();
1587 	if (state->n_rdwr != 0) {
1588 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1589 		if (ret != 0)
1590 			return ret;
1591 		if (newstate != state)
1592 			return -ESTALE;
1593 	}
1594 	if (state->n_wronly != 0) {
1595 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1596 		if (ret != 0)
1597 			return ret;
1598 		if (newstate != state)
1599 			return -ESTALE;
1600 	}
1601 	if (state->n_rdonly != 0) {
1602 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1603 		if (ret != 0)
1604 			return ret;
1605 		if (newstate != state)
1606 			return -ESTALE;
1607 	}
1608 	/*
1609 	 * We may have performed cached opens for all three recoveries.
1610 	 * Check if we need to update the current stateid.
1611 	 */
1612 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1613 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1614 		write_seqlock(&state->seqlock);
1615 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1616 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1617 		write_sequnlock(&state->seqlock);
1618 	}
1619 	return 0;
1620 }
1621 
1622 /*
1623  * OPEN_RECLAIM:
1624  * 	reclaim state on the server after a reboot.
1625  */
1626 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1627 {
1628 	struct nfs_delegation *delegation;
1629 	struct nfs4_opendata *opendata;
1630 	fmode_t delegation_type = 0;
1631 	int status;
1632 
1633 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1634 			NFS4_OPEN_CLAIM_PREVIOUS);
1635 	if (IS_ERR(opendata))
1636 		return PTR_ERR(opendata);
1637 	rcu_read_lock();
1638 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1639 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1640 		delegation_type = delegation->type;
1641 	rcu_read_unlock();
1642 	opendata->o_arg.u.delegation_type = delegation_type;
1643 	status = nfs4_open_recover(opendata, state);
1644 	nfs4_opendata_put(opendata);
1645 	return status;
1646 }
1647 
1648 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1649 {
1650 	struct nfs_server *server = NFS_SERVER(state->inode);
1651 	struct nfs4_exception exception = { };
1652 	int err;
1653 	do {
1654 		err = _nfs4_do_open_reclaim(ctx, state);
1655 		trace_nfs4_open_reclaim(ctx, 0, err);
1656 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1657 			continue;
1658 		if (err != -NFS4ERR_DELAY)
1659 			break;
1660 		nfs4_handle_exception(server, err, &exception);
1661 	} while (exception.retry);
1662 	return err;
1663 }
1664 
1665 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1666 {
1667 	struct nfs_open_context *ctx;
1668 	int ret;
1669 
1670 	ctx = nfs4_state_find_open_context(state);
1671 	if (IS_ERR(ctx))
1672 		return -EAGAIN;
1673 	ret = nfs4_do_open_reclaim(ctx, state);
1674 	put_nfs_open_context(ctx);
1675 	return ret;
1676 }
1677 
1678 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1679 {
1680 	switch (err) {
1681 		default:
1682 			printk(KERN_ERR "NFS: %s: unhandled error "
1683 					"%d.\n", __func__, err);
1684 		case 0:
1685 		case -ENOENT:
1686 		case -ESTALE:
1687 			break;
1688 		case -NFS4ERR_BADSESSION:
1689 		case -NFS4ERR_BADSLOT:
1690 		case -NFS4ERR_BAD_HIGH_SLOT:
1691 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1692 		case -NFS4ERR_DEADSESSION:
1693 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1694 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1695 			return -EAGAIN;
1696 		case -NFS4ERR_STALE_CLIENTID:
1697 		case -NFS4ERR_STALE_STATEID:
1698 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1699 		case -NFS4ERR_EXPIRED:
1700 			/* Don't recall a delegation if it was lost */
1701 			nfs4_schedule_lease_recovery(server->nfs_client);
1702 			return -EAGAIN;
1703 		case -NFS4ERR_MOVED:
1704 			nfs4_schedule_migration_recovery(server);
1705 			return -EAGAIN;
1706 		case -NFS4ERR_LEASE_MOVED:
1707 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1708 			return -EAGAIN;
1709 		case -NFS4ERR_DELEG_REVOKED:
1710 		case -NFS4ERR_ADMIN_REVOKED:
1711 		case -NFS4ERR_BAD_STATEID:
1712 		case -NFS4ERR_OPENMODE:
1713 			nfs_inode_find_state_and_recover(state->inode,
1714 					stateid);
1715 			nfs4_schedule_stateid_recovery(server, state);
1716 			return -EAGAIN;
1717 		case -NFS4ERR_DELAY:
1718 		case -NFS4ERR_GRACE:
1719 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1720 			ssleep(1);
1721 			return -EAGAIN;
1722 		case -ENOMEM:
1723 		case -NFS4ERR_DENIED:
1724 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1725 			return 0;
1726 	}
1727 	return err;
1728 }
1729 
1730 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1731 {
1732 	struct nfs_server *server = NFS_SERVER(state->inode);
1733 	struct nfs4_opendata *opendata;
1734 	int err;
1735 
1736 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1737 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1738 	if (IS_ERR(opendata))
1739 		return PTR_ERR(opendata);
1740 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1741 	err = nfs4_open_recover(opendata, state);
1742 	nfs4_opendata_put(opendata);
1743 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1744 }
1745 
1746 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1747 {
1748 	struct nfs4_opendata *data = calldata;
1749 
1750 	nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1751 			     &data->c_arg.seq_args, &data->c_res.seq_res, task);
1752 }
1753 
1754 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1755 {
1756 	struct nfs4_opendata *data = calldata;
1757 
1758 	nfs40_sequence_done(task, &data->c_res.seq_res);
1759 
1760 	data->rpc_status = task->tk_status;
1761 	if (data->rpc_status == 0) {
1762 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1763 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1764 		renew_lease(data->o_res.server, data->timestamp);
1765 		data->rpc_done = 1;
1766 	}
1767 }
1768 
1769 static void nfs4_open_confirm_release(void *calldata)
1770 {
1771 	struct nfs4_opendata *data = calldata;
1772 	struct nfs4_state *state = NULL;
1773 
1774 	/* If this request hasn't been cancelled, do nothing */
1775 	if (data->cancelled == 0)
1776 		goto out_free;
1777 	/* In case of error, no cleanup! */
1778 	if (!data->rpc_done)
1779 		goto out_free;
1780 	state = nfs4_opendata_to_nfs4_state(data);
1781 	if (!IS_ERR(state))
1782 		nfs4_close_state(state, data->o_arg.fmode);
1783 out_free:
1784 	nfs4_opendata_put(data);
1785 }
1786 
1787 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1788 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1789 	.rpc_call_done = nfs4_open_confirm_done,
1790 	.rpc_release = nfs4_open_confirm_release,
1791 };
1792 
1793 /*
1794  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1795  */
1796 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1797 {
1798 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1799 	struct rpc_task *task;
1800 	struct  rpc_message msg = {
1801 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1802 		.rpc_argp = &data->c_arg,
1803 		.rpc_resp = &data->c_res,
1804 		.rpc_cred = data->owner->so_cred,
1805 	};
1806 	struct rpc_task_setup task_setup_data = {
1807 		.rpc_client = server->client,
1808 		.rpc_message = &msg,
1809 		.callback_ops = &nfs4_open_confirm_ops,
1810 		.callback_data = data,
1811 		.workqueue = nfsiod_workqueue,
1812 		.flags = RPC_TASK_ASYNC,
1813 	};
1814 	int status;
1815 
1816 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1817 	kref_get(&data->kref);
1818 	data->rpc_done = 0;
1819 	data->rpc_status = 0;
1820 	data->timestamp = jiffies;
1821 	task = rpc_run_task(&task_setup_data);
1822 	if (IS_ERR(task))
1823 		return PTR_ERR(task);
1824 	status = nfs4_wait_for_completion_rpc_task(task);
1825 	if (status != 0) {
1826 		data->cancelled = 1;
1827 		smp_wmb();
1828 	} else
1829 		status = data->rpc_status;
1830 	rpc_put_task(task);
1831 	return status;
1832 }
1833 
1834 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1835 {
1836 	struct nfs4_opendata *data = calldata;
1837 	struct nfs4_state_owner *sp = data->owner;
1838 	struct nfs_client *clp = sp->so_server->nfs_client;
1839 
1840 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1841 		goto out_wait;
1842 	/*
1843 	 * Check if we still need to send an OPEN call, or if we can use
1844 	 * a delegation instead.
1845 	 */
1846 	if (data->state != NULL) {
1847 		struct nfs_delegation *delegation;
1848 
1849 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1850 			goto out_no_action;
1851 		rcu_read_lock();
1852 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1853 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1854 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1855 		    can_open_delegated(delegation, data->o_arg.fmode))
1856 			goto unlock_no_action;
1857 		rcu_read_unlock();
1858 	}
1859 	/* Update client id. */
1860 	data->o_arg.clientid = clp->cl_clientid;
1861 	switch (data->o_arg.claim) {
1862 	case NFS4_OPEN_CLAIM_PREVIOUS:
1863 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1864 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1865 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1866 	case NFS4_OPEN_CLAIM_FH:
1867 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1868 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1869 	}
1870 	data->timestamp = jiffies;
1871 	if (nfs4_setup_sequence(data->o_arg.server,
1872 				&data->o_arg.seq_args,
1873 				&data->o_res.seq_res,
1874 				task) != 0)
1875 		nfs_release_seqid(data->o_arg.seqid);
1876 
1877 	/* Set the create mode (note dependency on the session type) */
1878 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1879 	if (data->o_arg.open_flags & O_EXCL) {
1880 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1881 		if (nfs4_has_persistent_session(clp))
1882 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1883 		else if (clp->cl_mvops->minor_version > 0)
1884 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1885 	}
1886 	return;
1887 unlock_no_action:
1888 	rcu_read_unlock();
1889 out_no_action:
1890 	task->tk_action = NULL;
1891 out_wait:
1892 	nfs4_sequence_done(task, &data->o_res.seq_res);
1893 }
1894 
1895 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1896 {
1897 	struct nfs4_opendata *data = calldata;
1898 
1899 	data->rpc_status = task->tk_status;
1900 
1901 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1902 		return;
1903 
1904 	if (task->tk_status == 0) {
1905 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1906 			switch (data->o_res.f_attr->mode & S_IFMT) {
1907 			case S_IFREG:
1908 				break;
1909 			case S_IFLNK:
1910 				data->rpc_status = -ELOOP;
1911 				break;
1912 			case S_IFDIR:
1913 				data->rpc_status = -EISDIR;
1914 				break;
1915 			default:
1916 				data->rpc_status = -ENOTDIR;
1917 			}
1918 		}
1919 		renew_lease(data->o_res.server, data->timestamp);
1920 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1921 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1922 	}
1923 	data->rpc_done = 1;
1924 }
1925 
1926 static void nfs4_open_release(void *calldata)
1927 {
1928 	struct nfs4_opendata *data = calldata;
1929 	struct nfs4_state *state = NULL;
1930 
1931 	/* If this request hasn't been cancelled, do nothing */
1932 	if (data->cancelled == 0)
1933 		goto out_free;
1934 	/* In case of error, no cleanup! */
1935 	if (data->rpc_status != 0 || !data->rpc_done)
1936 		goto out_free;
1937 	/* In case we need an open_confirm, no cleanup! */
1938 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1939 		goto out_free;
1940 	state = nfs4_opendata_to_nfs4_state(data);
1941 	if (!IS_ERR(state))
1942 		nfs4_close_state(state, data->o_arg.fmode);
1943 out_free:
1944 	nfs4_opendata_put(data);
1945 }
1946 
1947 static const struct rpc_call_ops nfs4_open_ops = {
1948 	.rpc_call_prepare = nfs4_open_prepare,
1949 	.rpc_call_done = nfs4_open_done,
1950 	.rpc_release = nfs4_open_release,
1951 };
1952 
1953 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1954 {
1955 	struct inode *dir = d_inode(data->dir);
1956 	struct nfs_server *server = NFS_SERVER(dir);
1957 	struct nfs_openargs *o_arg = &data->o_arg;
1958 	struct nfs_openres *o_res = &data->o_res;
1959 	struct rpc_task *task;
1960 	struct rpc_message msg = {
1961 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1962 		.rpc_argp = o_arg,
1963 		.rpc_resp = o_res,
1964 		.rpc_cred = data->owner->so_cred,
1965 	};
1966 	struct rpc_task_setup task_setup_data = {
1967 		.rpc_client = server->client,
1968 		.rpc_message = &msg,
1969 		.callback_ops = &nfs4_open_ops,
1970 		.callback_data = data,
1971 		.workqueue = nfsiod_workqueue,
1972 		.flags = RPC_TASK_ASYNC,
1973 	};
1974 	int status;
1975 
1976 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1977 	kref_get(&data->kref);
1978 	data->rpc_done = 0;
1979 	data->rpc_status = 0;
1980 	data->cancelled = 0;
1981 	data->is_recover = 0;
1982 	if (isrecover) {
1983 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1984 		data->is_recover = 1;
1985 	}
1986 	task = rpc_run_task(&task_setup_data);
1987         if (IS_ERR(task))
1988                 return PTR_ERR(task);
1989         status = nfs4_wait_for_completion_rpc_task(task);
1990         if (status != 0) {
1991                 data->cancelled = 1;
1992                 smp_wmb();
1993         } else
1994                 status = data->rpc_status;
1995         rpc_put_task(task);
1996 
1997 	return status;
1998 }
1999 
2000 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2001 {
2002 	struct inode *dir = d_inode(data->dir);
2003 	struct nfs_openres *o_res = &data->o_res;
2004         int status;
2005 
2006 	status = nfs4_run_open_task(data, 1);
2007 	if (status != 0 || !data->rpc_done)
2008 		return status;
2009 
2010 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2011 
2012 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2013 		status = _nfs4_proc_open_confirm(data);
2014 		if (status != 0)
2015 			return status;
2016 	}
2017 
2018 	return status;
2019 }
2020 
2021 /*
2022  * Additional permission checks in order to distinguish between an
2023  * open for read, and an open for execute. This works around the
2024  * fact that NFSv4 OPEN treats read and execute permissions as being
2025  * the same.
2026  * Note that in the non-execute case, we want to turn off permission
2027  * checking if we just created a new file (POSIX open() semantics).
2028  */
2029 static int nfs4_opendata_access(struct rpc_cred *cred,
2030 				struct nfs4_opendata *opendata,
2031 				struct nfs4_state *state, fmode_t fmode,
2032 				int openflags)
2033 {
2034 	struct nfs_access_entry cache;
2035 	u32 mask;
2036 
2037 	/* access call failed or for some reason the server doesn't
2038 	 * support any access modes -- defer access call until later */
2039 	if (opendata->o_res.access_supported == 0)
2040 		return 0;
2041 
2042 	mask = 0;
2043 	/*
2044 	 * Use openflags to check for exec, because fmode won't
2045 	 * always have FMODE_EXEC set when file open for exec.
2046 	 */
2047 	if (openflags & __FMODE_EXEC) {
2048 		/* ONLY check for exec rights */
2049 		mask = MAY_EXEC;
2050 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2051 		mask = MAY_READ;
2052 
2053 	cache.cred = cred;
2054 	cache.jiffies = jiffies;
2055 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2056 	nfs_access_add_cache(state->inode, &cache);
2057 
2058 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2059 		return 0;
2060 
2061 	/* even though OPEN succeeded, access is denied. Close the file */
2062 	nfs4_close_state(state, fmode);
2063 	return -EACCES;
2064 }
2065 
2066 /*
2067  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2068  */
2069 static int _nfs4_proc_open(struct nfs4_opendata *data)
2070 {
2071 	struct inode *dir = d_inode(data->dir);
2072 	struct nfs_server *server = NFS_SERVER(dir);
2073 	struct nfs_openargs *o_arg = &data->o_arg;
2074 	struct nfs_openres *o_res = &data->o_res;
2075 	int status;
2076 
2077 	status = nfs4_run_open_task(data, 0);
2078 	if (!data->rpc_done)
2079 		return status;
2080 	if (status != 0) {
2081 		if (status == -NFS4ERR_BADNAME &&
2082 				!(o_arg->open_flags & O_CREAT))
2083 			return -ENOENT;
2084 		return status;
2085 	}
2086 
2087 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2088 
2089 	if (o_arg->open_flags & O_CREAT) {
2090 		update_changeattr(dir, &o_res->cinfo);
2091 		if (o_arg->open_flags & O_EXCL)
2092 			data->file_created = 1;
2093 		else if (o_res->cinfo.before != o_res->cinfo.after)
2094 			data->file_created = 1;
2095 	}
2096 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2097 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2098 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2099 		status = _nfs4_proc_open_confirm(data);
2100 		if (status != 0)
2101 			return status;
2102 	}
2103 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2104 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2105 	return 0;
2106 }
2107 
2108 static int nfs4_recover_expired_lease(struct nfs_server *server)
2109 {
2110 	return nfs4_client_recover_expired_lease(server->nfs_client);
2111 }
2112 
2113 /*
2114  * OPEN_EXPIRED:
2115  * 	reclaim state on the server after a network partition.
2116  * 	Assumes caller holds the appropriate lock
2117  */
2118 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2119 {
2120 	struct nfs4_opendata *opendata;
2121 	int ret;
2122 
2123 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2124 			NFS4_OPEN_CLAIM_FH);
2125 	if (IS_ERR(opendata))
2126 		return PTR_ERR(opendata);
2127 	ret = nfs4_open_recover(opendata, state);
2128 	if (ret == -ESTALE)
2129 		d_drop(ctx->dentry);
2130 	nfs4_opendata_put(opendata);
2131 	return ret;
2132 }
2133 
2134 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2135 {
2136 	struct nfs_server *server = NFS_SERVER(state->inode);
2137 	struct nfs4_exception exception = { };
2138 	int err;
2139 
2140 	do {
2141 		err = _nfs4_open_expired(ctx, state);
2142 		trace_nfs4_open_expired(ctx, 0, err);
2143 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2144 			continue;
2145 		switch (err) {
2146 		default:
2147 			goto out;
2148 		case -NFS4ERR_GRACE:
2149 		case -NFS4ERR_DELAY:
2150 			nfs4_handle_exception(server, err, &exception);
2151 			err = 0;
2152 		}
2153 	} while (exception.retry);
2154 out:
2155 	return err;
2156 }
2157 
2158 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2159 {
2160 	struct nfs_open_context *ctx;
2161 	int ret;
2162 
2163 	ctx = nfs4_state_find_open_context(state);
2164 	if (IS_ERR(ctx))
2165 		return -EAGAIN;
2166 	ret = nfs4_do_open_expired(ctx, state);
2167 	put_nfs_open_context(ctx);
2168 	return ret;
2169 }
2170 
2171 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2172 {
2173 	nfs_remove_bad_delegation(state->inode);
2174 	write_seqlock(&state->seqlock);
2175 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2176 	write_sequnlock(&state->seqlock);
2177 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2178 }
2179 
2180 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2181 {
2182 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2183 		nfs_finish_clear_delegation_stateid(state);
2184 }
2185 
2186 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2187 {
2188 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2189 	nfs40_clear_delegation_stateid(state);
2190 	return nfs4_open_expired(sp, state);
2191 }
2192 
2193 #if defined(CONFIG_NFS_V4_1)
2194 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2195 {
2196 	struct nfs_server *server = NFS_SERVER(state->inode);
2197 	nfs4_stateid stateid;
2198 	struct nfs_delegation *delegation;
2199 	struct rpc_cred *cred;
2200 	int status;
2201 
2202 	/* Get the delegation credential for use by test/free_stateid */
2203 	rcu_read_lock();
2204 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2205 	if (delegation == NULL) {
2206 		rcu_read_unlock();
2207 		return;
2208 	}
2209 
2210 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2211 	cred = get_rpccred(delegation->cred);
2212 	rcu_read_unlock();
2213 	status = nfs41_test_stateid(server, &stateid, cred);
2214 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2215 
2216 	if (status != NFS_OK) {
2217 		/* Free the stateid unless the server explicitly
2218 		 * informs us the stateid is unrecognized. */
2219 		if (status != -NFS4ERR_BAD_STATEID)
2220 			nfs41_free_stateid(server, &stateid, cred);
2221 		nfs_finish_clear_delegation_stateid(state);
2222 	}
2223 
2224 	put_rpccred(cred);
2225 }
2226 
2227 /**
2228  * nfs41_check_open_stateid - possibly free an open stateid
2229  *
2230  * @state: NFSv4 state for an inode
2231  *
2232  * Returns NFS_OK if recovery for this stateid is now finished.
2233  * Otherwise a negative NFS4ERR value is returned.
2234  */
2235 static int nfs41_check_open_stateid(struct nfs4_state *state)
2236 {
2237 	struct nfs_server *server = NFS_SERVER(state->inode);
2238 	nfs4_stateid *stateid = &state->open_stateid;
2239 	struct rpc_cred *cred = state->owner->so_cred;
2240 	int status;
2241 
2242 	/* If a state reset has been done, test_stateid is unneeded */
2243 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2244 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2245 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2246 		return -NFS4ERR_BAD_STATEID;
2247 
2248 	status = nfs41_test_stateid(server, stateid, cred);
2249 	trace_nfs4_test_open_stateid(state, NULL, status);
2250 	if (status != NFS_OK) {
2251 		/* Free the stateid unless the server explicitly
2252 		 * informs us the stateid is unrecognized. */
2253 		if (status != -NFS4ERR_BAD_STATEID)
2254 			nfs41_free_stateid(server, stateid, cred);
2255 
2256 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2257 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2258 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2259 		clear_bit(NFS_OPEN_STATE, &state->flags);
2260 	}
2261 	return status;
2262 }
2263 
2264 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2265 {
2266 	int status;
2267 
2268 	nfs41_check_delegation_stateid(state);
2269 	status = nfs41_check_open_stateid(state);
2270 	if (status != NFS_OK)
2271 		status = nfs4_open_expired(sp, state);
2272 	return status;
2273 }
2274 #endif
2275 
2276 /*
2277  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2278  * fields corresponding to attributes that were used to store the verifier.
2279  * Make sure we clobber those fields in the later setattr call
2280  */
2281 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2282 {
2283 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2284 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2285 		sattr->ia_valid |= ATTR_ATIME;
2286 
2287 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2288 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2289 		sattr->ia_valid |= ATTR_MTIME;
2290 }
2291 
2292 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2293 		fmode_t fmode,
2294 		int flags,
2295 		struct nfs_open_context *ctx)
2296 {
2297 	struct nfs4_state_owner *sp = opendata->owner;
2298 	struct nfs_server *server = sp->so_server;
2299 	struct dentry *dentry;
2300 	struct nfs4_state *state;
2301 	unsigned int seq;
2302 	int ret;
2303 
2304 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2305 
2306 	ret = _nfs4_proc_open(opendata);
2307 	if (ret != 0)
2308 		goto out;
2309 
2310 	state = nfs4_opendata_to_nfs4_state(opendata);
2311 	ret = PTR_ERR(state);
2312 	if (IS_ERR(state))
2313 		goto out;
2314 	if (server->caps & NFS_CAP_POSIX_LOCK)
2315 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2316 
2317 	dentry = opendata->dentry;
2318 	if (d_really_is_negative(dentry)) {
2319 		/* FIXME: Is this d_drop() ever needed? */
2320 		d_drop(dentry);
2321 		dentry = d_add_unique(dentry, igrab(state->inode));
2322 		if (dentry == NULL) {
2323 			dentry = opendata->dentry;
2324 		} else if (dentry != ctx->dentry) {
2325 			dput(ctx->dentry);
2326 			ctx->dentry = dget(dentry);
2327 		}
2328 		nfs_set_verifier(dentry,
2329 				nfs_save_change_attribute(d_inode(opendata->dir)));
2330 	}
2331 
2332 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2333 	if (ret != 0)
2334 		goto out;
2335 
2336 	ctx->state = state;
2337 	if (d_inode(dentry) == state->inode) {
2338 		nfs_inode_attach_open_context(ctx);
2339 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2340 			nfs4_schedule_stateid_recovery(server, state);
2341 	}
2342 out:
2343 	return ret;
2344 }
2345 
2346 /*
2347  * Returns a referenced nfs4_state
2348  */
2349 static int _nfs4_do_open(struct inode *dir,
2350 			struct nfs_open_context *ctx,
2351 			int flags,
2352 			struct iattr *sattr,
2353 			struct nfs4_label *label,
2354 			int *opened)
2355 {
2356 	struct nfs4_state_owner  *sp;
2357 	struct nfs4_state     *state = NULL;
2358 	struct nfs_server       *server = NFS_SERVER(dir);
2359 	struct nfs4_opendata *opendata;
2360 	struct dentry *dentry = ctx->dentry;
2361 	struct rpc_cred *cred = ctx->cred;
2362 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2363 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2364 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2365 	struct nfs4_label *olabel = NULL;
2366 	int status;
2367 
2368 	/* Protect against reboot recovery conflicts */
2369 	status = -ENOMEM;
2370 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2371 	if (sp == NULL) {
2372 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2373 		goto out_err;
2374 	}
2375 	status = nfs4_recover_expired_lease(server);
2376 	if (status != 0)
2377 		goto err_put_state_owner;
2378 	if (d_really_is_positive(dentry))
2379 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2380 	status = -ENOMEM;
2381 	if (d_really_is_positive(dentry))
2382 		claim = NFS4_OPEN_CLAIM_FH;
2383 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2384 			label, claim, GFP_KERNEL);
2385 	if (opendata == NULL)
2386 		goto err_put_state_owner;
2387 
2388 	if (label) {
2389 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2390 		if (IS_ERR(olabel)) {
2391 			status = PTR_ERR(olabel);
2392 			goto err_opendata_put;
2393 		}
2394 	}
2395 
2396 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2397 		if (!opendata->f_attr.mdsthreshold) {
2398 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2399 			if (!opendata->f_attr.mdsthreshold)
2400 				goto err_free_label;
2401 		}
2402 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2403 	}
2404 	if (d_really_is_positive(dentry))
2405 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2406 
2407 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2408 	if (status != 0)
2409 		goto err_free_label;
2410 	state = ctx->state;
2411 
2412 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2413 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2414 		nfs4_exclusive_attrset(opendata, sattr);
2415 
2416 		nfs_fattr_init(opendata->o_res.f_attr);
2417 		status = nfs4_do_setattr(state->inode, cred,
2418 				opendata->o_res.f_attr, sattr,
2419 				state, label, olabel);
2420 		if (status == 0) {
2421 			nfs_setattr_update_inode(state->inode, sattr,
2422 					opendata->o_res.f_attr);
2423 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2424 		}
2425 	}
2426 	if (opendata->file_created)
2427 		*opened |= FILE_CREATED;
2428 
2429 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2430 		*ctx_th = opendata->f_attr.mdsthreshold;
2431 		opendata->f_attr.mdsthreshold = NULL;
2432 	}
2433 
2434 	nfs4_label_free(olabel);
2435 
2436 	nfs4_opendata_put(opendata);
2437 	nfs4_put_state_owner(sp);
2438 	return 0;
2439 err_free_label:
2440 	nfs4_label_free(olabel);
2441 err_opendata_put:
2442 	nfs4_opendata_put(opendata);
2443 err_put_state_owner:
2444 	nfs4_put_state_owner(sp);
2445 out_err:
2446 	return status;
2447 }
2448 
2449 
2450 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2451 					struct nfs_open_context *ctx,
2452 					int flags,
2453 					struct iattr *sattr,
2454 					struct nfs4_label *label,
2455 					int *opened)
2456 {
2457 	struct nfs_server *server = NFS_SERVER(dir);
2458 	struct nfs4_exception exception = { };
2459 	struct nfs4_state *res;
2460 	int status;
2461 
2462 	do {
2463 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2464 		res = ctx->state;
2465 		trace_nfs4_open_file(ctx, flags, status);
2466 		if (status == 0)
2467 			break;
2468 		/* NOTE: BAD_SEQID means the server and client disagree about the
2469 		 * book-keeping w.r.t. state-changing operations
2470 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2471 		 * It is actually a sign of a bug on the client or on the server.
2472 		 *
2473 		 * If we receive a BAD_SEQID error in the particular case of
2474 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2475 		 * have unhashed the old state_owner for us, and that we can
2476 		 * therefore safely retry using a new one. We should still warn
2477 		 * the user though...
2478 		 */
2479 		if (status == -NFS4ERR_BAD_SEQID) {
2480 			pr_warn_ratelimited("NFS: v4 server %s "
2481 					" returned a bad sequence-id error!\n",
2482 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2483 			exception.retry = 1;
2484 			continue;
2485 		}
2486 		/*
2487 		 * BAD_STATEID on OPEN means that the server cancelled our
2488 		 * state before it received the OPEN_CONFIRM.
2489 		 * Recover by retrying the request as per the discussion
2490 		 * on Page 181 of RFC3530.
2491 		 */
2492 		if (status == -NFS4ERR_BAD_STATEID) {
2493 			exception.retry = 1;
2494 			continue;
2495 		}
2496 		if (status == -EAGAIN) {
2497 			/* We must have found a delegation */
2498 			exception.retry = 1;
2499 			continue;
2500 		}
2501 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2502 			continue;
2503 		res = ERR_PTR(nfs4_handle_exception(server,
2504 					status, &exception));
2505 	} while (exception.retry);
2506 	return res;
2507 }
2508 
2509 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2510 			    struct nfs_fattr *fattr, struct iattr *sattr,
2511 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2512 			    struct nfs4_label *olabel)
2513 {
2514 	struct nfs_server *server = NFS_SERVER(inode);
2515         struct nfs_setattrargs  arg = {
2516                 .fh             = NFS_FH(inode),
2517                 .iap            = sattr,
2518 		.server		= server,
2519 		.bitmask = server->attr_bitmask,
2520 		.label		= ilabel,
2521         };
2522         struct nfs_setattrres  res = {
2523 		.fattr		= fattr,
2524 		.label		= olabel,
2525 		.server		= server,
2526         };
2527         struct rpc_message msg = {
2528 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2529 		.rpc_argp	= &arg,
2530 		.rpc_resp	= &res,
2531 		.rpc_cred	= cred,
2532         };
2533 	unsigned long timestamp = jiffies;
2534 	fmode_t fmode;
2535 	bool truncate;
2536 	int status;
2537 
2538 	arg.bitmask = nfs4_bitmask(server, ilabel);
2539 	if (ilabel)
2540 		arg.bitmask = nfs4_bitmask(server, olabel);
2541 
2542 	nfs_fattr_init(fattr);
2543 
2544 	/* Servers should only apply open mode checks for file size changes */
2545 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2546 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2547 
2548 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2549 		/* Use that stateid */
2550 	} else if (truncate && state != NULL) {
2551 		struct nfs_lockowner lockowner = {
2552 			.l_owner = current->files,
2553 			.l_pid = current->tgid,
2554 		};
2555 		if (!nfs4_valid_open_stateid(state))
2556 			return -EBADF;
2557 		if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2558 				&lockowner) == -EIO)
2559 			return -EBADF;
2560 	} else
2561 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2562 
2563 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2564 	if (status == 0 && state != NULL)
2565 		renew_lease(server, timestamp);
2566 	return status;
2567 }
2568 
2569 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2570 			   struct nfs_fattr *fattr, struct iattr *sattr,
2571 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2572 			   struct nfs4_label *olabel)
2573 {
2574 	struct nfs_server *server = NFS_SERVER(inode);
2575 	struct nfs4_exception exception = {
2576 		.state = state,
2577 		.inode = inode,
2578 	};
2579 	int err;
2580 	do {
2581 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2582 		trace_nfs4_setattr(inode, err);
2583 		switch (err) {
2584 		case -NFS4ERR_OPENMODE:
2585 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2586 				pr_warn_once("NFSv4: server %s is incorrectly "
2587 						"applying open mode checks to "
2588 						"a SETATTR that is not "
2589 						"changing file size.\n",
2590 						server->nfs_client->cl_hostname);
2591 			}
2592 			if (state && !(state->state & FMODE_WRITE)) {
2593 				err = -EBADF;
2594 				if (sattr->ia_valid & ATTR_OPEN)
2595 					err = -EACCES;
2596 				goto out;
2597 			}
2598 		}
2599 		err = nfs4_handle_exception(server, err, &exception);
2600 	} while (exception.retry);
2601 out:
2602 	return err;
2603 }
2604 
2605 struct nfs4_closedata {
2606 	struct inode *inode;
2607 	struct nfs4_state *state;
2608 	struct nfs_closeargs arg;
2609 	struct nfs_closeres res;
2610 	struct nfs_fattr fattr;
2611 	unsigned long timestamp;
2612 	bool roc;
2613 	u32 roc_barrier;
2614 };
2615 
2616 static void nfs4_free_closedata(void *data)
2617 {
2618 	struct nfs4_closedata *calldata = data;
2619 	struct nfs4_state_owner *sp = calldata->state->owner;
2620 	struct super_block *sb = calldata->state->inode->i_sb;
2621 
2622 	if (calldata->roc)
2623 		pnfs_roc_release(calldata->state->inode);
2624 	nfs4_put_open_state(calldata->state);
2625 	nfs_free_seqid(calldata->arg.seqid);
2626 	nfs4_put_state_owner(sp);
2627 	nfs_sb_deactive(sb);
2628 	kfree(calldata);
2629 }
2630 
2631 static void nfs4_close_done(struct rpc_task *task, void *data)
2632 {
2633 	struct nfs4_closedata *calldata = data;
2634 	struct nfs4_state *state = calldata->state;
2635 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2636 	nfs4_stateid *res_stateid = NULL;
2637 
2638 	dprintk("%s: begin!\n", __func__);
2639 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2640 		return;
2641 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2642         /* hmm. we are done with the inode, and in the process of freeing
2643 	 * the state_owner. we keep this around to process errors
2644 	 */
2645 	switch (task->tk_status) {
2646 		case 0:
2647 			res_stateid = &calldata->res.stateid;
2648 			if (calldata->arg.fmode == 0 && calldata->roc)
2649 				pnfs_roc_set_barrier(state->inode,
2650 						     calldata->roc_barrier);
2651 			renew_lease(server, calldata->timestamp);
2652 			break;
2653 		case -NFS4ERR_ADMIN_REVOKED:
2654 		case -NFS4ERR_STALE_STATEID:
2655 		case -NFS4ERR_OLD_STATEID:
2656 		case -NFS4ERR_BAD_STATEID:
2657 		case -NFS4ERR_EXPIRED:
2658 			if (!nfs4_stateid_match(&calldata->arg.stateid,
2659 						&state->open_stateid)) {
2660 				rpc_restart_call_prepare(task);
2661 				goto out_release;
2662 			}
2663 			if (calldata->arg.fmode == 0)
2664 				break;
2665 		default:
2666 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2667 				rpc_restart_call_prepare(task);
2668 				goto out_release;
2669 			}
2670 	}
2671 	nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2672 out_release:
2673 	nfs_release_seqid(calldata->arg.seqid);
2674 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2675 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2676 }
2677 
2678 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2679 {
2680 	struct nfs4_closedata *calldata = data;
2681 	struct nfs4_state *state = calldata->state;
2682 	struct inode *inode = calldata->inode;
2683 	bool is_rdonly, is_wronly, is_rdwr;
2684 	int call_close = 0;
2685 
2686 	dprintk("%s: begin!\n", __func__);
2687 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2688 		goto out_wait;
2689 
2690 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2691 	spin_lock(&state->owner->so_lock);
2692 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2693 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2694 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2695 	nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2696 	/* Calculate the change in open mode */
2697 	calldata->arg.fmode = 0;
2698 	if (state->n_rdwr == 0) {
2699 		if (state->n_rdonly == 0)
2700 			call_close |= is_rdonly;
2701 		else if (is_rdonly)
2702 			calldata->arg.fmode |= FMODE_READ;
2703 		if (state->n_wronly == 0)
2704 			call_close |= is_wronly;
2705 		else if (is_wronly)
2706 			calldata->arg.fmode |= FMODE_WRITE;
2707 	} else if (is_rdwr)
2708 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2709 
2710 	if (calldata->arg.fmode == 0)
2711 		call_close |= is_rdwr;
2712 
2713 	if (!nfs4_valid_open_stateid(state))
2714 		call_close = 0;
2715 	spin_unlock(&state->owner->so_lock);
2716 
2717 	if (!call_close) {
2718 		/* Note: exit _without_ calling nfs4_close_done */
2719 		goto out_no_action;
2720 	}
2721 
2722 	if (calldata->arg.fmode == 0) {
2723 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2724 		if (calldata->roc &&
2725 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2726 			nfs_release_seqid(calldata->arg.seqid);
2727 			goto out_wait;
2728 		    }
2729 	}
2730 	calldata->arg.share_access =
2731 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
2732 				calldata->arg.fmode, 0);
2733 
2734 	nfs_fattr_init(calldata->res.fattr);
2735 	calldata->timestamp = jiffies;
2736 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2737 				&calldata->arg.seq_args,
2738 				&calldata->res.seq_res,
2739 				task) != 0)
2740 		nfs_release_seqid(calldata->arg.seqid);
2741 	dprintk("%s: done!\n", __func__);
2742 	return;
2743 out_no_action:
2744 	task->tk_action = NULL;
2745 out_wait:
2746 	nfs4_sequence_done(task, &calldata->res.seq_res);
2747 }
2748 
2749 static const struct rpc_call_ops nfs4_close_ops = {
2750 	.rpc_call_prepare = nfs4_close_prepare,
2751 	.rpc_call_done = nfs4_close_done,
2752 	.rpc_release = nfs4_free_closedata,
2753 };
2754 
2755 static bool nfs4_roc(struct inode *inode)
2756 {
2757 	if (!nfs_have_layout(inode))
2758 		return false;
2759 	return pnfs_roc(inode);
2760 }
2761 
2762 /*
2763  * It is possible for data to be read/written from a mem-mapped file
2764  * after the sys_close call (which hits the vfs layer as a flush).
2765  * This means that we can't safely call nfsv4 close on a file until
2766  * the inode is cleared. This in turn means that we are not good
2767  * NFSv4 citizens - we do not indicate to the server to update the file's
2768  * share state even when we are done with one of the three share
2769  * stateid's in the inode.
2770  *
2771  * NOTE: Caller must be holding the sp->so_owner semaphore!
2772  */
2773 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2774 {
2775 	struct nfs_server *server = NFS_SERVER(state->inode);
2776 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2777 	struct nfs4_closedata *calldata;
2778 	struct nfs4_state_owner *sp = state->owner;
2779 	struct rpc_task *task;
2780 	struct rpc_message msg = {
2781 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2782 		.rpc_cred = state->owner->so_cred,
2783 	};
2784 	struct rpc_task_setup task_setup_data = {
2785 		.rpc_client = server->client,
2786 		.rpc_message = &msg,
2787 		.callback_ops = &nfs4_close_ops,
2788 		.workqueue = nfsiod_workqueue,
2789 		.flags = RPC_TASK_ASYNC,
2790 	};
2791 	int status = -ENOMEM;
2792 
2793 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2794 		&task_setup_data.rpc_client, &msg);
2795 
2796 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2797 	if (calldata == NULL)
2798 		goto out;
2799 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2800 	calldata->inode = state->inode;
2801 	calldata->state = state;
2802 	calldata->arg.fh = NFS_FH(state->inode);
2803 	/* Serialization for the sequence id */
2804 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2805 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2806 	if (IS_ERR(calldata->arg.seqid))
2807 		goto out_free_calldata;
2808 	calldata->arg.fmode = 0;
2809 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2810 	calldata->res.fattr = &calldata->fattr;
2811 	calldata->res.seqid = calldata->arg.seqid;
2812 	calldata->res.server = server;
2813 	calldata->roc = nfs4_roc(state->inode);
2814 	nfs_sb_active(calldata->inode->i_sb);
2815 
2816 	msg.rpc_argp = &calldata->arg;
2817 	msg.rpc_resp = &calldata->res;
2818 	task_setup_data.callback_data = calldata;
2819 	task = rpc_run_task(&task_setup_data);
2820 	if (IS_ERR(task))
2821 		return PTR_ERR(task);
2822 	status = 0;
2823 	if (wait)
2824 		status = rpc_wait_for_completion_task(task);
2825 	rpc_put_task(task);
2826 	return status;
2827 out_free_calldata:
2828 	kfree(calldata);
2829 out:
2830 	nfs4_put_open_state(state);
2831 	nfs4_put_state_owner(sp);
2832 	return status;
2833 }
2834 
2835 static struct inode *
2836 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2837 		int open_flags, struct iattr *attr, int *opened)
2838 {
2839 	struct nfs4_state *state;
2840 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2841 
2842 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2843 
2844 	/* Protect against concurrent sillydeletes */
2845 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2846 
2847 	nfs4_label_release_security(label);
2848 
2849 	if (IS_ERR(state))
2850 		return ERR_CAST(state);
2851 	return state->inode;
2852 }
2853 
2854 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2855 {
2856 	if (ctx->state == NULL)
2857 		return;
2858 	if (is_sync)
2859 		nfs4_close_sync(ctx->state, ctx->mode);
2860 	else
2861 		nfs4_close_state(ctx->state, ctx->mode);
2862 }
2863 
2864 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2865 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2866 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2867 
2868 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2869 {
2870 	struct nfs4_server_caps_arg args = {
2871 		.fhandle = fhandle,
2872 	};
2873 	struct nfs4_server_caps_res res = {};
2874 	struct rpc_message msg = {
2875 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2876 		.rpc_argp = &args,
2877 		.rpc_resp = &res,
2878 	};
2879 	int status;
2880 
2881 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2882 	if (status == 0) {
2883 		/* Sanity check the server answers */
2884 		switch (server->nfs_client->cl_minorversion) {
2885 		case 0:
2886 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2887 			res.attr_bitmask[2] = 0;
2888 			break;
2889 		case 1:
2890 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2891 			break;
2892 		case 2:
2893 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2894 		}
2895 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2896 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2897 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2898 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2899 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2900 				NFS_CAP_CTIME|NFS_CAP_MTIME|
2901 				NFS_CAP_SECURITY_LABEL);
2902 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2903 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2904 			server->caps |= NFS_CAP_ACLS;
2905 		if (res.has_links != 0)
2906 			server->caps |= NFS_CAP_HARDLINKS;
2907 		if (res.has_symlinks != 0)
2908 			server->caps |= NFS_CAP_SYMLINKS;
2909 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2910 			server->caps |= NFS_CAP_FILEID;
2911 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2912 			server->caps |= NFS_CAP_MODE;
2913 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2914 			server->caps |= NFS_CAP_NLINK;
2915 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2916 			server->caps |= NFS_CAP_OWNER;
2917 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2918 			server->caps |= NFS_CAP_OWNER_GROUP;
2919 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2920 			server->caps |= NFS_CAP_ATIME;
2921 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2922 			server->caps |= NFS_CAP_CTIME;
2923 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2924 			server->caps |= NFS_CAP_MTIME;
2925 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2926 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2927 			server->caps |= NFS_CAP_SECURITY_LABEL;
2928 #endif
2929 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2930 				sizeof(server->attr_bitmask));
2931 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2932 
2933 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2934 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2935 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2936 		server->cache_consistency_bitmask[2] = 0;
2937 		server->acl_bitmask = res.acl_bitmask;
2938 		server->fh_expire_type = res.fh_expire_type;
2939 	}
2940 
2941 	return status;
2942 }
2943 
2944 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2945 {
2946 	struct nfs4_exception exception = { };
2947 	int err;
2948 	do {
2949 		err = nfs4_handle_exception(server,
2950 				_nfs4_server_capabilities(server, fhandle),
2951 				&exception);
2952 	} while (exception.retry);
2953 	return err;
2954 }
2955 
2956 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2957 		struct nfs_fsinfo *info)
2958 {
2959 	u32 bitmask[3];
2960 	struct nfs4_lookup_root_arg args = {
2961 		.bitmask = bitmask,
2962 	};
2963 	struct nfs4_lookup_res res = {
2964 		.server = server,
2965 		.fattr = info->fattr,
2966 		.fh = fhandle,
2967 	};
2968 	struct rpc_message msg = {
2969 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2970 		.rpc_argp = &args,
2971 		.rpc_resp = &res,
2972 	};
2973 
2974 	bitmask[0] = nfs4_fattr_bitmap[0];
2975 	bitmask[1] = nfs4_fattr_bitmap[1];
2976 	/*
2977 	 * Process the label in the upcoming getfattr
2978 	 */
2979 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2980 
2981 	nfs_fattr_init(info->fattr);
2982 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2983 }
2984 
2985 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2986 		struct nfs_fsinfo *info)
2987 {
2988 	struct nfs4_exception exception = { };
2989 	int err;
2990 	do {
2991 		err = _nfs4_lookup_root(server, fhandle, info);
2992 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2993 		switch (err) {
2994 		case 0:
2995 		case -NFS4ERR_WRONGSEC:
2996 			goto out;
2997 		default:
2998 			err = nfs4_handle_exception(server, err, &exception);
2999 		}
3000 	} while (exception.retry);
3001 out:
3002 	return err;
3003 }
3004 
3005 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3006 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3007 {
3008 	struct rpc_auth_create_args auth_args = {
3009 		.pseudoflavor = flavor,
3010 	};
3011 	struct rpc_auth *auth;
3012 	int ret;
3013 
3014 	auth = rpcauth_create(&auth_args, server->client);
3015 	if (IS_ERR(auth)) {
3016 		ret = -EACCES;
3017 		goto out;
3018 	}
3019 	ret = nfs4_lookup_root(server, fhandle, info);
3020 out:
3021 	return ret;
3022 }
3023 
3024 /*
3025  * Retry pseudoroot lookup with various security flavors.  We do this when:
3026  *
3027  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3028  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3029  *
3030  * Returns zero on success, or a negative NFS4ERR value, or a
3031  * negative errno value.
3032  */
3033 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3034 			      struct nfs_fsinfo *info)
3035 {
3036 	/* Per 3530bis 15.33.5 */
3037 	static const rpc_authflavor_t flav_array[] = {
3038 		RPC_AUTH_GSS_KRB5P,
3039 		RPC_AUTH_GSS_KRB5I,
3040 		RPC_AUTH_GSS_KRB5,
3041 		RPC_AUTH_UNIX,			/* courtesy */
3042 		RPC_AUTH_NULL,
3043 	};
3044 	int status = -EPERM;
3045 	size_t i;
3046 
3047 	if (server->auth_info.flavor_len > 0) {
3048 		/* try each flavor specified by user */
3049 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3050 			status = nfs4_lookup_root_sec(server, fhandle, info,
3051 						server->auth_info.flavors[i]);
3052 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3053 				continue;
3054 			break;
3055 		}
3056 	} else {
3057 		/* no flavors specified by user, try default list */
3058 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3059 			status = nfs4_lookup_root_sec(server, fhandle, info,
3060 						      flav_array[i]);
3061 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3062 				continue;
3063 			break;
3064 		}
3065 	}
3066 
3067 	/*
3068 	 * -EACCESS could mean that the user doesn't have correct permissions
3069 	 * to access the mount.  It could also mean that we tried to mount
3070 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3071 	 * existing mount programs don't handle -EACCES very well so it should
3072 	 * be mapped to -EPERM instead.
3073 	 */
3074 	if (status == -EACCES)
3075 		status = -EPERM;
3076 	return status;
3077 }
3078 
3079 static int nfs4_do_find_root_sec(struct nfs_server *server,
3080 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3081 {
3082 	int mv = server->nfs_client->cl_minorversion;
3083 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3084 }
3085 
3086 /**
3087  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3088  * @server: initialized nfs_server handle
3089  * @fhandle: we fill in the pseudo-fs root file handle
3090  * @info: we fill in an FSINFO struct
3091  * @auth_probe: probe the auth flavours
3092  *
3093  * Returns zero on success, or a negative errno.
3094  */
3095 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3096 			 struct nfs_fsinfo *info,
3097 			 bool auth_probe)
3098 {
3099 	int status = 0;
3100 
3101 	if (!auth_probe)
3102 		status = nfs4_lookup_root(server, fhandle, info);
3103 
3104 	if (auth_probe || status == NFS4ERR_WRONGSEC)
3105 		status = nfs4_do_find_root_sec(server, fhandle, info);
3106 
3107 	if (status == 0)
3108 		status = nfs4_server_capabilities(server, fhandle);
3109 	if (status == 0)
3110 		status = nfs4_do_fsinfo(server, fhandle, info);
3111 
3112 	return nfs4_map_errors(status);
3113 }
3114 
3115 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3116 			      struct nfs_fsinfo *info)
3117 {
3118 	int error;
3119 	struct nfs_fattr *fattr = info->fattr;
3120 	struct nfs4_label *label = NULL;
3121 
3122 	error = nfs4_server_capabilities(server, mntfh);
3123 	if (error < 0) {
3124 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3125 		return error;
3126 	}
3127 
3128 	label = nfs4_label_alloc(server, GFP_KERNEL);
3129 	if (IS_ERR(label))
3130 		return PTR_ERR(label);
3131 
3132 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3133 	if (error < 0) {
3134 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3135 		goto err_free_label;
3136 	}
3137 
3138 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3139 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3140 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3141 
3142 err_free_label:
3143 	nfs4_label_free(label);
3144 
3145 	return error;
3146 }
3147 
3148 /*
3149  * Get locations and (maybe) other attributes of a referral.
3150  * Note that we'll actually follow the referral later when
3151  * we detect fsid mismatch in inode revalidation
3152  */
3153 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3154 			     const struct qstr *name, struct nfs_fattr *fattr,
3155 			     struct nfs_fh *fhandle)
3156 {
3157 	int status = -ENOMEM;
3158 	struct page *page = NULL;
3159 	struct nfs4_fs_locations *locations = NULL;
3160 
3161 	page = alloc_page(GFP_KERNEL);
3162 	if (page == NULL)
3163 		goto out;
3164 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3165 	if (locations == NULL)
3166 		goto out;
3167 
3168 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3169 	if (status != 0)
3170 		goto out;
3171 
3172 	/*
3173 	 * If the fsid didn't change, this is a migration event, not a
3174 	 * referral.  Cause us to drop into the exception handler, which
3175 	 * will kick off migration recovery.
3176 	 */
3177 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3178 		dprintk("%s: server did not return a different fsid for"
3179 			" a referral at %s\n", __func__, name->name);
3180 		status = -NFS4ERR_MOVED;
3181 		goto out;
3182 	}
3183 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3184 	nfs_fixup_referral_attributes(&locations->fattr);
3185 
3186 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3187 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3188 	memset(fhandle, 0, sizeof(struct nfs_fh));
3189 out:
3190 	if (page)
3191 		__free_page(page);
3192 	kfree(locations);
3193 	return status;
3194 }
3195 
3196 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3197 				struct nfs_fattr *fattr, struct nfs4_label *label)
3198 {
3199 	struct nfs4_getattr_arg args = {
3200 		.fh = fhandle,
3201 		.bitmask = server->attr_bitmask,
3202 	};
3203 	struct nfs4_getattr_res res = {
3204 		.fattr = fattr,
3205 		.label = label,
3206 		.server = server,
3207 	};
3208 	struct rpc_message msg = {
3209 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3210 		.rpc_argp = &args,
3211 		.rpc_resp = &res,
3212 	};
3213 
3214 	args.bitmask = nfs4_bitmask(server, label);
3215 
3216 	nfs_fattr_init(fattr);
3217 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3218 }
3219 
3220 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3221 				struct nfs_fattr *fattr, struct nfs4_label *label)
3222 {
3223 	struct nfs4_exception exception = { };
3224 	int err;
3225 	do {
3226 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3227 		trace_nfs4_getattr(server, fhandle, fattr, err);
3228 		err = nfs4_handle_exception(server, err,
3229 				&exception);
3230 	} while (exception.retry);
3231 	return err;
3232 }
3233 
3234 /*
3235  * The file is not closed if it is opened due to the a request to change
3236  * the size of the file. The open call will not be needed once the
3237  * VFS layer lookup-intents are implemented.
3238  *
3239  * Close is called when the inode is destroyed.
3240  * If we haven't opened the file for O_WRONLY, we
3241  * need to in the size_change case to obtain a stateid.
3242  *
3243  * Got race?
3244  * Because OPEN is always done by name in nfsv4, it is
3245  * possible that we opened a different file by the same
3246  * name.  We can recognize this race condition, but we
3247  * can't do anything about it besides returning an error.
3248  *
3249  * This will be fixed with VFS changes (lookup-intent).
3250  */
3251 static int
3252 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3253 		  struct iattr *sattr)
3254 {
3255 	struct inode *inode = d_inode(dentry);
3256 	struct rpc_cred *cred = NULL;
3257 	struct nfs4_state *state = NULL;
3258 	struct nfs4_label *label = NULL;
3259 	int status;
3260 
3261 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3262 	    sattr->ia_valid & ATTR_SIZE &&
3263 	    sattr->ia_size < i_size_read(inode))
3264 		pnfs_commit_and_return_layout(inode);
3265 
3266 	nfs_fattr_init(fattr);
3267 
3268 	/* Deal with open(O_TRUNC) */
3269 	if (sattr->ia_valid & ATTR_OPEN)
3270 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3271 
3272 	/* Optimization: if the end result is no change, don't RPC */
3273 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3274 		return 0;
3275 
3276 	/* Search for an existing open(O_WRITE) file */
3277 	if (sattr->ia_valid & ATTR_FILE) {
3278 		struct nfs_open_context *ctx;
3279 
3280 		ctx = nfs_file_open_context(sattr->ia_file);
3281 		if (ctx) {
3282 			cred = ctx->cred;
3283 			state = ctx->state;
3284 		}
3285 	}
3286 
3287 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3288 	if (IS_ERR(label))
3289 		return PTR_ERR(label);
3290 
3291 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3292 	if (status == 0) {
3293 		nfs_setattr_update_inode(inode, sattr, fattr);
3294 		nfs_setsecurity(inode, fattr, label);
3295 	}
3296 	nfs4_label_free(label);
3297 	return status;
3298 }
3299 
3300 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3301 		const struct qstr *name, struct nfs_fh *fhandle,
3302 		struct nfs_fattr *fattr, struct nfs4_label *label)
3303 {
3304 	struct nfs_server *server = NFS_SERVER(dir);
3305 	int		       status;
3306 	struct nfs4_lookup_arg args = {
3307 		.bitmask = server->attr_bitmask,
3308 		.dir_fh = NFS_FH(dir),
3309 		.name = name,
3310 	};
3311 	struct nfs4_lookup_res res = {
3312 		.server = server,
3313 		.fattr = fattr,
3314 		.label = label,
3315 		.fh = fhandle,
3316 	};
3317 	struct rpc_message msg = {
3318 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3319 		.rpc_argp = &args,
3320 		.rpc_resp = &res,
3321 	};
3322 
3323 	args.bitmask = nfs4_bitmask(server, label);
3324 
3325 	nfs_fattr_init(fattr);
3326 
3327 	dprintk("NFS call  lookup %s\n", name->name);
3328 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3329 	dprintk("NFS reply lookup: %d\n", status);
3330 	return status;
3331 }
3332 
3333 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3334 {
3335 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3336 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3337 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3338 	fattr->nlink = 2;
3339 }
3340 
3341 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3342 				   struct qstr *name, struct nfs_fh *fhandle,
3343 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3344 {
3345 	struct nfs4_exception exception = { };
3346 	struct rpc_clnt *client = *clnt;
3347 	int err;
3348 	do {
3349 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3350 		trace_nfs4_lookup(dir, name, err);
3351 		switch (err) {
3352 		case -NFS4ERR_BADNAME:
3353 			err = -ENOENT;
3354 			goto out;
3355 		case -NFS4ERR_MOVED:
3356 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3357 			goto out;
3358 		case -NFS4ERR_WRONGSEC:
3359 			err = -EPERM;
3360 			if (client != *clnt)
3361 				goto out;
3362 			client = nfs4_negotiate_security(client, dir, name);
3363 			if (IS_ERR(client))
3364 				return PTR_ERR(client);
3365 
3366 			exception.retry = 1;
3367 			break;
3368 		default:
3369 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3370 		}
3371 	} while (exception.retry);
3372 
3373 out:
3374 	if (err == 0)
3375 		*clnt = client;
3376 	else if (client != *clnt)
3377 		rpc_shutdown_client(client);
3378 
3379 	return err;
3380 }
3381 
3382 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3383 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3384 			    struct nfs4_label *label)
3385 {
3386 	int status;
3387 	struct rpc_clnt *client = NFS_CLIENT(dir);
3388 
3389 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3390 	if (client != NFS_CLIENT(dir)) {
3391 		rpc_shutdown_client(client);
3392 		nfs_fixup_secinfo_attributes(fattr);
3393 	}
3394 	return status;
3395 }
3396 
3397 struct rpc_clnt *
3398 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3399 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3400 {
3401 	struct rpc_clnt *client = NFS_CLIENT(dir);
3402 	int status;
3403 
3404 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3405 	if (status < 0)
3406 		return ERR_PTR(status);
3407 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3408 }
3409 
3410 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3411 {
3412 	struct nfs_server *server = NFS_SERVER(inode);
3413 	struct nfs4_accessargs args = {
3414 		.fh = NFS_FH(inode),
3415 		.bitmask = server->cache_consistency_bitmask,
3416 	};
3417 	struct nfs4_accessres res = {
3418 		.server = server,
3419 	};
3420 	struct rpc_message msg = {
3421 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3422 		.rpc_argp = &args,
3423 		.rpc_resp = &res,
3424 		.rpc_cred = entry->cred,
3425 	};
3426 	int mode = entry->mask;
3427 	int status = 0;
3428 
3429 	/*
3430 	 * Determine which access bits we want to ask for...
3431 	 */
3432 	if (mode & MAY_READ)
3433 		args.access |= NFS4_ACCESS_READ;
3434 	if (S_ISDIR(inode->i_mode)) {
3435 		if (mode & MAY_WRITE)
3436 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3437 		if (mode & MAY_EXEC)
3438 			args.access |= NFS4_ACCESS_LOOKUP;
3439 	} else {
3440 		if (mode & MAY_WRITE)
3441 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3442 		if (mode & MAY_EXEC)
3443 			args.access |= NFS4_ACCESS_EXECUTE;
3444 	}
3445 
3446 	res.fattr = nfs_alloc_fattr();
3447 	if (res.fattr == NULL)
3448 		return -ENOMEM;
3449 
3450 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3451 	if (!status) {
3452 		nfs_access_set_mask(entry, res.access);
3453 		nfs_refresh_inode(inode, res.fattr);
3454 	}
3455 	nfs_free_fattr(res.fattr);
3456 	return status;
3457 }
3458 
3459 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3460 {
3461 	struct nfs4_exception exception = { };
3462 	int err;
3463 	do {
3464 		err = _nfs4_proc_access(inode, entry);
3465 		trace_nfs4_access(inode, err);
3466 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3467 				&exception);
3468 	} while (exception.retry);
3469 	return err;
3470 }
3471 
3472 /*
3473  * TODO: For the time being, we don't try to get any attributes
3474  * along with any of the zero-copy operations READ, READDIR,
3475  * READLINK, WRITE.
3476  *
3477  * In the case of the first three, we want to put the GETATTR
3478  * after the read-type operation -- this is because it is hard
3479  * to predict the length of a GETATTR response in v4, and thus
3480  * align the READ data correctly.  This means that the GETATTR
3481  * may end up partially falling into the page cache, and we should
3482  * shift it into the 'tail' of the xdr_buf before processing.
3483  * To do this efficiently, we need to know the total length
3484  * of data received, which doesn't seem to be available outside
3485  * of the RPC layer.
3486  *
3487  * In the case of WRITE, we also want to put the GETATTR after
3488  * the operation -- in this case because we want to make sure
3489  * we get the post-operation mtime and size.
3490  *
3491  * Both of these changes to the XDR layer would in fact be quite
3492  * minor, but I decided to leave them for a subsequent patch.
3493  */
3494 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3495 		unsigned int pgbase, unsigned int pglen)
3496 {
3497 	struct nfs4_readlink args = {
3498 		.fh       = NFS_FH(inode),
3499 		.pgbase	  = pgbase,
3500 		.pglen    = pglen,
3501 		.pages    = &page,
3502 	};
3503 	struct nfs4_readlink_res res;
3504 	struct rpc_message msg = {
3505 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3506 		.rpc_argp = &args,
3507 		.rpc_resp = &res,
3508 	};
3509 
3510 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3511 }
3512 
3513 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3514 		unsigned int pgbase, unsigned int pglen)
3515 {
3516 	struct nfs4_exception exception = { };
3517 	int err;
3518 	do {
3519 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3520 		trace_nfs4_readlink(inode, err);
3521 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3522 				&exception);
3523 	} while (exception.retry);
3524 	return err;
3525 }
3526 
3527 /*
3528  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3529  */
3530 static int
3531 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3532 		 int flags)
3533 {
3534 	struct nfs4_label l, *ilabel = NULL;
3535 	struct nfs_open_context *ctx;
3536 	struct nfs4_state *state;
3537 	int opened = 0;
3538 	int status = 0;
3539 
3540 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3541 	if (IS_ERR(ctx))
3542 		return PTR_ERR(ctx);
3543 
3544 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3545 
3546 	sattr->ia_mode &= ~current_umask();
3547 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3548 	if (IS_ERR(state)) {
3549 		status = PTR_ERR(state);
3550 		goto out;
3551 	}
3552 out:
3553 	nfs4_label_release_security(ilabel);
3554 	put_nfs_open_context(ctx);
3555 	return status;
3556 }
3557 
3558 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3559 {
3560 	struct nfs_server *server = NFS_SERVER(dir);
3561 	struct nfs_removeargs args = {
3562 		.fh = NFS_FH(dir),
3563 		.name = *name,
3564 	};
3565 	struct nfs_removeres res = {
3566 		.server = server,
3567 	};
3568 	struct rpc_message msg = {
3569 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3570 		.rpc_argp = &args,
3571 		.rpc_resp = &res,
3572 	};
3573 	int status;
3574 
3575 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3576 	if (status == 0)
3577 		update_changeattr(dir, &res.cinfo);
3578 	return status;
3579 }
3580 
3581 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3582 {
3583 	struct nfs4_exception exception = { };
3584 	int err;
3585 	do {
3586 		err = _nfs4_proc_remove(dir, name);
3587 		trace_nfs4_remove(dir, name, err);
3588 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3589 				&exception);
3590 	} while (exception.retry);
3591 	return err;
3592 }
3593 
3594 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3595 {
3596 	struct nfs_server *server = NFS_SERVER(dir);
3597 	struct nfs_removeargs *args = msg->rpc_argp;
3598 	struct nfs_removeres *res = msg->rpc_resp;
3599 
3600 	res->server = server;
3601 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3602 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3603 
3604 	nfs_fattr_init(res->dir_attr);
3605 }
3606 
3607 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3608 {
3609 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3610 			&data->args.seq_args,
3611 			&data->res.seq_res,
3612 			task);
3613 }
3614 
3615 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3616 {
3617 	struct nfs_unlinkdata *data = task->tk_calldata;
3618 	struct nfs_removeres *res = &data->res;
3619 
3620 	if (!nfs4_sequence_done(task, &res->seq_res))
3621 		return 0;
3622 	if (nfs4_async_handle_error(task, res->server, NULL,
3623 				    &data->timeout) == -EAGAIN)
3624 		return 0;
3625 	update_changeattr(dir, &res->cinfo);
3626 	return 1;
3627 }
3628 
3629 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3630 {
3631 	struct nfs_server *server = NFS_SERVER(dir);
3632 	struct nfs_renameargs *arg = msg->rpc_argp;
3633 	struct nfs_renameres *res = msg->rpc_resp;
3634 
3635 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3636 	res->server = server;
3637 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3638 }
3639 
3640 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3641 {
3642 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3643 			&data->args.seq_args,
3644 			&data->res.seq_res,
3645 			task);
3646 }
3647 
3648 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3649 				 struct inode *new_dir)
3650 {
3651 	struct nfs_renamedata *data = task->tk_calldata;
3652 	struct nfs_renameres *res = &data->res;
3653 
3654 	if (!nfs4_sequence_done(task, &res->seq_res))
3655 		return 0;
3656 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3657 		return 0;
3658 
3659 	update_changeattr(old_dir, &res->old_cinfo);
3660 	update_changeattr(new_dir, &res->new_cinfo);
3661 	return 1;
3662 }
3663 
3664 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3665 {
3666 	struct nfs_server *server = NFS_SERVER(inode);
3667 	struct nfs4_link_arg arg = {
3668 		.fh     = NFS_FH(inode),
3669 		.dir_fh = NFS_FH(dir),
3670 		.name   = name,
3671 		.bitmask = server->attr_bitmask,
3672 	};
3673 	struct nfs4_link_res res = {
3674 		.server = server,
3675 		.label = NULL,
3676 	};
3677 	struct rpc_message msg = {
3678 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3679 		.rpc_argp = &arg,
3680 		.rpc_resp = &res,
3681 	};
3682 	int status = -ENOMEM;
3683 
3684 	res.fattr = nfs_alloc_fattr();
3685 	if (res.fattr == NULL)
3686 		goto out;
3687 
3688 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3689 	if (IS_ERR(res.label)) {
3690 		status = PTR_ERR(res.label);
3691 		goto out;
3692 	}
3693 	arg.bitmask = nfs4_bitmask(server, res.label);
3694 
3695 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3696 	if (!status) {
3697 		update_changeattr(dir, &res.cinfo);
3698 		status = nfs_post_op_update_inode(inode, res.fattr);
3699 		if (!status)
3700 			nfs_setsecurity(inode, res.fattr, res.label);
3701 	}
3702 
3703 
3704 	nfs4_label_free(res.label);
3705 
3706 out:
3707 	nfs_free_fattr(res.fattr);
3708 	return status;
3709 }
3710 
3711 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3712 {
3713 	struct nfs4_exception exception = { };
3714 	int err;
3715 	do {
3716 		err = nfs4_handle_exception(NFS_SERVER(inode),
3717 				_nfs4_proc_link(inode, dir, name),
3718 				&exception);
3719 	} while (exception.retry);
3720 	return err;
3721 }
3722 
3723 struct nfs4_createdata {
3724 	struct rpc_message msg;
3725 	struct nfs4_create_arg arg;
3726 	struct nfs4_create_res res;
3727 	struct nfs_fh fh;
3728 	struct nfs_fattr fattr;
3729 	struct nfs4_label *label;
3730 };
3731 
3732 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3733 		struct qstr *name, struct iattr *sattr, u32 ftype)
3734 {
3735 	struct nfs4_createdata *data;
3736 
3737 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3738 	if (data != NULL) {
3739 		struct nfs_server *server = NFS_SERVER(dir);
3740 
3741 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3742 		if (IS_ERR(data->label))
3743 			goto out_free;
3744 
3745 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3746 		data->msg.rpc_argp = &data->arg;
3747 		data->msg.rpc_resp = &data->res;
3748 		data->arg.dir_fh = NFS_FH(dir);
3749 		data->arg.server = server;
3750 		data->arg.name = name;
3751 		data->arg.attrs = sattr;
3752 		data->arg.ftype = ftype;
3753 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3754 		data->res.server = server;
3755 		data->res.fh = &data->fh;
3756 		data->res.fattr = &data->fattr;
3757 		data->res.label = data->label;
3758 		nfs_fattr_init(data->res.fattr);
3759 	}
3760 	return data;
3761 out_free:
3762 	kfree(data);
3763 	return NULL;
3764 }
3765 
3766 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3767 {
3768 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3769 				    &data->arg.seq_args, &data->res.seq_res, 1);
3770 	if (status == 0) {
3771 		update_changeattr(dir, &data->res.dir_cinfo);
3772 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3773 	}
3774 	return status;
3775 }
3776 
3777 static void nfs4_free_createdata(struct nfs4_createdata *data)
3778 {
3779 	nfs4_label_free(data->label);
3780 	kfree(data);
3781 }
3782 
3783 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3784 		struct page *page, unsigned int len, struct iattr *sattr,
3785 		struct nfs4_label *label)
3786 {
3787 	struct nfs4_createdata *data;
3788 	int status = -ENAMETOOLONG;
3789 
3790 	if (len > NFS4_MAXPATHLEN)
3791 		goto out;
3792 
3793 	status = -ENOMEM;
3794 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3795 	if (data == NULL)
3796 		goto out;
3797 
3798 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3799 	data->arg.u.symlink.pages = &page;
3800 	data->arg.u.symlink.len = len;
3801 	data->arg.label = label;
3802 
3803 	status = nfs4_do_create(dir, dentry, data);
3804 
3805 	nfs4_free_createdata(data);
3806 out:
3807 	return status;
3808 }
3809 
3810 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3811 		struct page *page, unsigned int len, struct iattr *sattr)
3812 {
3813 	struct nfs4_exception exception = { };
3814 	struct nfs4_label l, *label = NULL;
3815 	int err;
3816 
3817 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3818 
3819 	do {
3820 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3821 		trace_nfs4_symlink(dir, &dentry->d_name, err);
3822 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3823 				&exception);
3824 	} while (exception.retry);
3825 
3826 	nfs4_label_release_security(label);
3827 	return err;
3828 }
3829 
3830 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3831 		struct iattr *sattr, struct nfs4_label *label)
3832 {
3833 	struct nfs4_createdata *data;
3834 	int status = -ENOMEM;
3835 
3836 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3837 	if (data == NULL)
3838 		goto out;
3839 
3840 	data->arg.label = label;
3841 	status = nfs4_do_create(dir, dentry, data);
3842 
3843 	nfs4_free_createdata(data);
3844 out:
3845 	return status;
3846 }
3847 
3848 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3849 		struct iattr *sattr)
3850 {
3851 	struct nfs4_exception exception = { };
3852 	struct nfs4_label l, *label = NULL;
3853 	int err;
3854 
3855 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3856 
3857 	sattr->ia_mode &= ~current_umask();
3858 	do {
3859 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3860 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
3861 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3862 				&exception);
3863 	} while (exception.retry);
3864 	nfs4_label_release_security(label);
3865 
3866 	return err;
3867 }
3868 
3869 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3870 		u64 cookie, struct page **pages, unsigned int count, int plus)
3871 {
3872 	struct inode		*dir = d_inode(dentry);
3873 	struct nfs4_readdir_arg args = {
3874 		.fh = NFS_FH(dir),
3875 		.pages = pages,
3876 		.pgbase = 0,
3877 		.count = count,
3878 		.bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3879 		.plus = plus,
3880 	};
3881 	struct nfs4_readdir_res res;
3882 	struct rpc_message msg = {
3883 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3884 		.rpc_argp = &args,
3885 		.rpc_resp = &res,
3886 		.rpc_cred = cred,
3887 	};
3888 	int			status;
3889 
3890 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3891 			dentry,
3892 			(unsigned long long)cookie);
3893 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3894 	res.pgbase = args.pgbase;
3895 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3896 	if (status >= 0) {
3897 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3898 		status += args.pgbase;
3899 	}
3900 
3901 	nfs_invalidate_atime(dir);
3902 
3903 	dprintk("%s: returns %d\n", __func__, status);
3904 	return status;
3905 }
3906 
3907 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3908 		u64 cookie, struct page **pages, unsigned int count, int plus)
3909 {
3910 	struct nfs4_exception exception = { };
3911 	int err;
3912 	do {
3913 		err = _nfs4_proc_readdir(dentry, cred, cookie,
3914 				pages, count, plus);
3915 		trace_nfs4_readdir(d_inode(dentry), err);
3916 		err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3917 				&exception);
3918 	} while (exception.retry);
3919 	return err;
3920 }
3921 
3922 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3923 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3924 {
3925 	struct nfs4_createdata *data;
3926 	int mode = sattr->ia_mode;
3927 	int status = -ENOMEM;
3928 
3929 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3930 	if (data == NULL)
3931 		goto out;
3932 
3933 	if (S_ISFIFO(mode))
3934 		data->arg.ftype = NF4FIFO;
3935 	else if (S_ISBLK(mode)) {
3936 		data->arg.ftype = NF4BLK;
3937 		data->arg.u.device.specdata1 = MAJOR(rdev);
3938 		data->arg.u.device.specdata2 = MINOR(rdev);
3939 	}
3940 	else if (S_ISCHR(mode)) {
3941 		data->arg.ftype = NF4CHR;
3942 		data->arg.u.device.specdata1 = MAJOR(rdev);
3943 		data->arg.u.device.specdata2 = MINOR(rdev);
3944 	} else if (!S_ISSOCK(mode)) {
3945 		status = -EINVAL;
3946 		goto out_free;
3947 	}
3948 
3949 	data->arg.label = label;
3950 	status = nfs4_do_create(dir, dentry, data);
3951 out_free:
3952 	nfs4_free_createdata(data);
3953 out:
3954 	return status;
3955 }
3956 
3957 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3958 		struct iattr *sattr, dev_t rdev)
3959 {
3960 	struct nfs4_exception exception = { };
3961 	struct nfs4_label l, *label = NULL;
3962 	int err;
3963 
3964 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3965 
3966 	sattr->ia_mode &= ~current_umask();
3967 	do {
3968 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3969 		trace_nfs4_mknod(dir, &dentry->d_name, err);
3970 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3971 				&exception);
3972 	} while (exception.retry);
3973 
3974 	nfs4_label_release_security(label);
3975 
3976 	return err;
3977 }
3978 
3979 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3980 		 struct nfs_fsstat *fsstat)
3981 {
3982 	struct nfs4_statfs_arg args = {
3983 		.fh = fhandle,
3984 		.bitmask = server->attr_bitmask,
3985 	};
3986 	struct nfs4_statfs_res res = {
3987 		.fsstat = fsstat,
3988 	};
3989 	struct rpc_message msg = {
3990 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3991 		.rpc_argp = &args,
3992 		.rpc_resp = &res,
3993 	};
3994 
3995 	nfs_fattr_init(fsstat->fattr);
3996 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3997 }
3998 
3999 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4000 {
4001 	struct nfs4_exception exception = { };
4002 	int err;
4003 	do {
4004 		err = nfs4_handle_exception(server,
4005 				_nfs4_proc_statfs(server, fhandle, fsstat),
4006 				&exception);
4007 	} while (exception.retry);
4008 	return err;
4009 }
4010 
4011 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4012 		struct nfs_fsinfo *fsinfo)
4013 {
4014 	struct nfs4_fsinfo_arg args = {
4015 		.fh = fhandle,
4016 		.bitmask = server->attr_bitmask,
4017 	};
4018 	struct nfs4_fsinfo_res res = {
4019 		.fsinfo = fsinfo,
4020 	};
4021 	struct rpc_message msg = {
4022 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4023 		.rpc_argp = &args,
4024 		.rpc_resp = &res,
4025 	};
4026 
4027 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4028 }
4029 
4030 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4031 {
4032 	struct nfs4_exception exception = { };
4033 	unsigned long now = jiffies;
4034 	int err;
4035 
4036 	do {
4037 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4038 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4039 		if (err == 0) {
4040 			struct nfs_client *clp = server->nfs_client;
4041 
4042 			spin_lock(&clp->cl_lock);
4043 			clp->cl_lease_time = fsinfo->lease_time * HZ;
4044 			clp->cl_last_renewal = now;
4045 			spin_unlock(&clp->cl_lock);
4046 			break;
4047 		}
4048 		err = nfs4_handle_exception(server, err, &exception);
4049 	} while (exception.retry);
4050 	return err;
4051 }
4052 
4053 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4054 {
4055 	int error;
4056 
4057 	nfs_fattr_init(fsinfo->fattr);
4058 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4059 	if (error == 0) {
4060 		/* block layout checks this! */
4061 		server->pnfs_blksize = fsinfo->blksize;
4062 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4063 	}
4064 
4065 	return error;
4066 }
4067 
4068 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4069 		struct nfs_pathconf *pathconf)
4070 {
4071 	struct nfs4_pathconf_arg args = {
4072 		.fh = fhandle,
4073 		.bitmask = server->attr_bitmask,
4074 	};
4075 	struct nfs4_pathconf_res res = {
4076 		.pathconf = pathconf,
4077 	};
4078 	struct rpc_message msg = {
4079 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4080 		.rpc_argp = &args,
4081 		.rpc_resp = &res,
4082 	};
4083 
4084 	/* None of the pathconf attributes are mandatory to implement */
4085 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4086 		memset(pathconf, 0, sizeof(*pathconf));
4087 		return 0;
4088 	}
4089 
4090 	nfs_fattr_init(pathconf->fattr);
4091 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4092 }
4093 
4094 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4095 		struct nfs_pathconf *pathconf)
4096 {
4097 	struct nfs4_exception exception = { };
4098 	int err;
4099 
4100 	do {
4101 		err = nfs4_handle_exception(server,
4102 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4103 				&exception);
4104 	} while (exception.retry);
4105 	return err;
4106 }
4107 
4108 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4109 		const struct nfs_open_context *ctx,
4110 		const struct nfs_lock_context *l_ctx,
4111 		fmode_t fmode)
4112 {
4113 	const struct nfs_lockowner *lockowner = NULL;
4114 
4115 	if (l_ctx != NULL)
4116 		lockowner = &l_ctx->lockowner;
4117 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4118 }
4119 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4120 
4121 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4122 		const struct nfs_open_context *ctx,
4123 		const struct nfs_lock_context *l_ctx,
4124 		fmode_t fmode)
4125 {
4126 	nfs4_stateid current_stateid;
4127 
4128 	/* If the current stateid represents a lost lock, then exit */
4129 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4130 		return true;
4131 	return nfs4_stateid_match(stateid, &current_stateid);
4132 }
4133 
4134 static bool nfs4_error_stateid_expired(int err)
4135 {
4136 	switch (err) {
4137 	case -NFS4ERR_DELEG_REVOKED:
4138 	case -NFS4ERR_ADMIN_REVOKED:
4139 	case -NFS4ERR_BAD_STATEID:
4140 	case -NFS4ERR_STALE_STATEID:
4141 	case -NFS4ERR_OLD_STATEID:
4142 	case -NFS4ERR_OPENMODE:
4143 	case -NFS4ERR_EXPIRED:
4144 		return true;
4145 	}
4146 	return false;
4147 }
4148 
4149 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4150 {
4151 	nfs_invalidate_atime(hdr->inode);
4152 }
4153 
4154 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4155 {
4156 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4157 
4158 	trace_nfs4_read(hdr, task->tk_status);
4159 	if (nfs4_async_handle_error(task, server,
4160 				    hdr->args.context->state,
4161 				    NULL) == -EAGAIN) {
4162 		rpc_restart_call_prepare(task);
4163 		return -EAGAIN;
4164 	}
4165 
4166 	__nfs4_read_done_cb(hdr);
4167 	if (task->tk_status > 0)
4168 		renew_lease(server, hdr->timestamp);
4169 	return 0;
4170 }
4171 
4172 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4173 		struct nfs_pgio_args *args)
4174 {
4175 
4176 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4177 		nfs4_stateid_is_current(&args->stateid,
4178 				args->context,
4179 				args->lock_context,
4180 				FMODE_READ))
4181 		return false;
4182 	rpc_restart_call_prepare(task);
4183 	return true;
4184 }
4185 
4186 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4187 {
4188 
4189 	dprintk("--> %s\n", __func__);
4190 
4191 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4192 		return -EAGAIN;
4193 	if (nfs4_read_stateid_changed(task, &hdr->args))
4194 		return -EAGAIN;
4195 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4196 				    nfs4_read_done_cb(task, hdr);
4197 }
4198 
4199 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4200 				 struct rpc_message *msg)
4201 {
4202 	hdr->timestamp   = jiffies;
4203 	hdr->pgio_done_cb = nfs4_read_done_cb;
4204 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4205 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4206 }
4207 
4208 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4209 				      struct nfs_pgio_header *hdr)
4210 {
4211 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4212 			&hdr->args.seq_args,
4213 			&hdr->res.seq_res,
4214 			task))
4215 		return 0;
4216 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4217 				hdr->args.lock_context,
4218 				hdr->rw_ops->rw_mode) == -EIO)
4219 		return -EIO;
4220 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4221 		return -EIO;
4222 	return 0;
4223 }
4224 
4225 static int nfs4_write_done_cb(struct rpc_task *task,
4226 			      struct nfs_pgio_header *hdr)
4227 {
4228 	struct inode *inode = hdr->inode;
4229 
4230 	trace_nfs4_write(hdr, task->tk_status);
4231 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4232 				    hdr->args.context->state,
4233 				    NULL) == -EAGAIN) {
4234 		rpc_restart_call_prepare(task);
4235 		return -EAGAIN;
4236 	}
4237 	if (task->tk_status >= 0) {
4238 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4239 		nfs_writeback_update_inode(hdr);
4240 	}
4241 	return 0;
4242 }
4243 
4244 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4245 		struct nfs_pgio_args *args)
4246 {
4247 
4248 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4249 		nfs4_stateid_is_current(&args->stateid,
4250 				args->context,
4251 				args->lock_context,
4252 				FMODE_WRITE))
4253 		return false;
4254 	rpc_restart_call_prepare(task);
4255 	return true;
4256 }
4257 
4258 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4259 {
4260 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4261 		return -EAGAIN;
4262 	if (nfs4_write_stateid_changed(task, &hdr->args))
4263 		return -EAGAIN;
4264 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4265 		nfs4_write_done_cb(task, hdr);
4266 }
4267 
4268 static
4269 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4270 {
4271 	/* Don't request attributes for pNFS or O_DIRECT writes */
4272 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4273 		return false;
4274 	/* Otherwise, request attributes if and only if we don't hold
4275 	 * a delegation
4276 	 */
4277 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4278 }
4279 
4280 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4281 				  struct rpc_message *msg)
4282 {
4283 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4284 
4285 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4286 		hdr->args.bitmask = NULL;
4287 		hdr->res.fattr = NULL;
4288 	} else
4289 		hdr->args.bitmask = server->cache_consistency_bitmask;
4290 
4291 	if (!hdr->pgio_done_cb)
4292 		hdr->pgio_done_cb = nfs4_write_done_cb;
4293 	hdr->res.server = server;
4294 	hdr->timestamp   = jiffies;
4295 
4296 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4297 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4298 }
4299 
4300 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4301 {
4302 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4303 			&data->args.seq_args,
4304 			&data->res.seq_res,
4305 			task);
4306 }
4307 
4308 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4309 {
4310 	struct inode *inode = data->inode;
4311 
4312 	trace_nfs4_commit(data, task->tk_status);
4313 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4314 				    NULL, NULL) == -EAGAIN) {
4315 		rpc_restart_call_prepare(task);
4316 		return -EAGAIN;
4317 	}
4318 	return 0;
4319 }
4320 
4321 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4322 {
4323 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4324 		return -EAGAIN;
4325 	return data->commit_done_cb(task, data);
4326 }
4327 
4328 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4329 {
4330 	struct nfs_server *server = NFS_SERVER(data->inode);
4331 
4332 	if (data->commit_done_cb == NULL)
4333 		data->commit_done_cb = nfs4_commit_done_cb;
4334 	data->res.server = server;
4335 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4336 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4337 }
4338 
4339 struct nfs4_renewdata {
4340 	struct nfs_client	*client;
4341 	unsigned long		timestamp;
4342 };
4343 
4344 /*
4345  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4346  * standalone procedure for queueing an asynchronous RENEW.
4347  */
4348 static void nfs4_renew_release(void *calldata)
4349 {
4350 	struct nfs4_renewdata *data = calldata;
4351 	struct nfs_client *clp = data->client;
4352 
4353 	if (atomic_read(&clp->cl_count) > 1)
4354 		nfs4_schedule_state_renewal(clp);
4355 	nfs_put_client(clp);
4356 	kfree(data);
4357 }
4358 
4359 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4360 {
4361 	struct nfs4_renewdata *data = calldata;
4362 	struct nfs_client *clp = data->client;
4363 	unsigned long timestamp = data->timestamp;
4364 
4365 	trace_nfs4_renew_async(clp, task->tk_status);
4366 	switch (task->tk_status) {
4367 	case 0:
4368 		break;
4369 	case -NFS4ERR_LEASE_MOVED:
4370 		nfs4_schedule_lease_moved_recovery(clp);
4371 		break;
4372 	default:
4373 		/* Unless we're shutting down, schedule state recovery! */
4374 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4375 			return;
4376 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4377 			nfs4_schedule_lease_recovery(clp);
4378 			return;
4379 		}
4380 		nfs4_schedule_path_down_recovery(clp);
4381 	}
4382 	do_renew_lease(clp, timestamp);
4383 }
4384 
4385 static const struct rpc_call_ops nfs4_renew_ops = {
4386 	.rpc_call_done = nfs4_renew_done,
4387 	.rpc_release = nfs4_renew_release,
4388 };
4389 
4390 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4391 {
4392 	struct rpc_message msg = {
4393 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4394 		.rpc_argp	= clp,
4395 		.rpc_cred	= cred,
4396 	};
4397 	struct nfs4_renewdata *data;
4398 
4399 	if (renew_flags == 0)
4400 		return 0;
4401 	if (!atomic_inc_not_zero(&clp->cl_count))
4402 		return -EIO;
4403 	data = kmalloc(sizeof(*data), GFP_NOFS);
4404 	if (data == NULL)
4405 		return -ENOMEM;
4406 	data->client = clp;
4407 	data->timestamp = jiffies;
4408 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4409 			&nfs4_renew_ops, data);
4410 }
4411 
4412 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4413 {
4414 	struct rpc_message msg = {
4415 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4416 		.rpc_argp	= clp,
4417 		.rpc_cred	= cred,
4418 	};
4419 	unsigned long now = jiffies;
4420 	int status;
4421 
4422 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4423 	if (status < 0)
4424 		return status;
4425 	do_renew_lease(clp, now);
4426 	return 0;
4427 }
4428 
4429 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4430 {
4431 	return server->caps & NFS_CAP_ACLS;
4432 }
4433 
4434 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4435  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4436  * the stack.
4437  */
4438 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4439 
4440 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4441 		struct page **pages, unsigned int *pgbase)
4442 {
4443 	struct page *newpage, **spages;
4444 	int rc = 0;
4445 	size_t len;
4446 	spages = pages;
4447 
4448 	do {
4449 		len = min_t(size_t, PAGE_SIZE, buflen);
4450 		newpage = alloc_page(GFP_KERNEL);
4451 
4452 		if (newpage == NULL)
4453 			goto unwind;
4454 		memcpy(page_address(newpage), buf, len);
4455                 buf += len;
4456                 buflen -= len;
4457 		*pages++ = newpage;
4458 		rc++;
4459 	} while (buflen != 0);
4460 
4461 	return rc;
4462 
4463 unwind:
4464 	for(; rc > 0; rc--)
4465 		__free_page(spages[rc-1]);
4466 	return -ENOMEM;
4467 }
4468 
4469 struct nfs4_cached_acl {
4470 	int cached;
4471 	size_t len;
4472 	char data[0];
4473 };
4474 
4475 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4476 {
4477 	struct nfs_inode *nfsi = NFS_I(inode);
4478 
4479 	spin_lock(&inode->i_lock);
4480 	kfree(nfsi->nfs4_acl);
4481 	nfsi->nfs4_acl = acl;
4482 	spin_unlock(&inode->i_lock);
4483 }
4484 
4485 static void nfs4_zap_acl_attr(struct inode *inode)
4486 {
4487 	nfs4_set_cached_acl(inode, NULL);
4488 }
4489 
4490 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4491 {
4492 	struct nfs_inode *nfsi = NFS_I(inode);
4493 	struct nfs4_cached_acl *acl;
4494 	int ret = -ENOENT;
4495 
4496 	spin_lock(&inode->i_lock);
4497 	acl = nfsi->nfs4_acl;
4498 	if (acl == NULL)
4499 		goto out;
4500 	if (buf == NULL) /* user is just asking for length */
4501 		goto out_len;
4502 	if (acl->cached == 0)
4503 		goto out;
4504 	ret = -ERANGE; /* see getxattr(2) man page */
4505 	if (acl->len > buflen)
4506 		goto out;
4507 	memcpy(buf, acl->data, acl->len);
4508 out_len:
4509 	ret = acl->len;
4510 out:
4511 	spin_unlock(&inode->i_lock);
4512 	return ret;
4513 }
4514 
4515 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4516 {
4517 	struct nfs4_cached_acl *acl;
4518 	size_t buflen = sizeof(*acl) + acl_len;
4519 
4520 	if (buflen <= PAGE_SIZE) {
4521 		acl = kmalloc(buflen, GFP_KERNEL);
4522 		if (acl == NULL)
4523 			goto out;
4524 		acl->cached = 1;
4525 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4526 	} else {
4527 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4528 		if (acl == NULL)
4529 			goto out;
4530 		acl->cached = 0;
4531 	}
4532 	acl->len = acl_len;
4533 out:
4534 	nfs4_set_cached_acl(inode, acl);
4535 }
4536 
4537 /*
4538  * The getxattr API returns the required buffer length when called with a
4539  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4540  * the required buf.  On a NULL buf, we send a page of data to the server
4541  * guessing that the ACL request can be serviced by a page. If so, we cache
4542  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4543  * the cache. If not so, we throw away the page, and cache the required
4544  * length. The next getxattr call will then produce another round trip to
4545  * the server, this time with the input buf of the required size.
4546  */
4547 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4548 {
4549 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4550 	struct nfs_getaclargs args = {
4551 		.fh = NFS_FH(inode),
4552 		.acl_pages = pages,
4553 		.acl_len = buflen,
4554 	};
4555 	struct nfs_getaclres res = {
4556 		.acl_len = buflen,
4557 	};
4558 	struct rpc_message msg = {
4559 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4560 		.rpc_argp = &args,
4561 		.rpc_resp = &res,
4562 	};
4563 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4564 	int ret = -ENOMEM, i;
4565 
4566 	/* As long as we're doing a round trip to the server anyway,
4567 	 * let's be prepared for a page of acl data. */
4568 	if (npages == 0)
4569 		npages = 1;
4570 	if (npages > ARRAY_SIZE(pages))
4571 		return -ERANGE;
4572 
4573 	for (i = 0; i < npages; i++) {
4574 		pages[i] = alloc_page(GFP_KERNEL);
4575 		if (!pages[i])
4576 			goto out_free;
4577 	}
4578 
4579 	/* for decoding across pages */
4580 	res.acl_scratch = alloc_page(GFP_KERNEL);
4581 	if (!res.acl_scratch)
4582 		goto out_free;
4583 
4584 	args.acl_len = npages * PAGE_SIZE;
4585 	args.acl_pgbase = 0;
4586 
4587 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4588 		__func__, buf, buflen, npages, args.acl_len);
4589 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4590 			     &msg, &args.seq_args, &res.seq_res, 0);
4591 	if (ret)
4592 		goto out_free;
4593 
4594 	/* Handle the case where the passed-in buffer is too short */
4595 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4596 		/* Did the user only issue a request for the acl length? */
4597 		if (buf == NULL)
4598 			goto out_ok;
4599 		ret = -ERANGE;
4600 		goto out_free;
4601 	}
4602 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4603 	if (buf) {
4604 		if (res.acl_len > buflen) {
4605 			ret = -ERANGE;
4606 			goto out_free;
4607 		}
4608 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4609 	}
4610 out_ok:
4611 	ret = res.acl_len;
4612 out_free:
4613 	for (i = 0; i < npages; i++)
4614 		if (pages[i])
4615 			__free_page(pages[i]);
4616 	if (res.acl_scratch)
4617 		__free_page(res.acl_scratch);
4618 	return ret;
4619 }
4620 
4621 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4622 {
4623 	struct nfs4_exception exception = { };
4624 	ssize_t ret;
4625 	do {
4626 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4627 		trace_nfs4_get_acl(inode, ret);
4628 		if (ret >= 0)
4629 			break;
4630 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4631 	} while (exception.retry);
4632 	return ret;
4633 }
4634 
4635 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4636 {
4637 	struct nfs_server *server = NFS_SERVER(inode);
4638 	int ret;
4639 
4640 	if (!nfs4_server_supports_acls(server))
4641 		return -EOPNOTSUPP;
4642 	ret = nfs_revalidate_inode(server, inode);
4643 	if (ret < 0)
4644 		return ret;
4645 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4646 		nfs_zap_acl_cache(inode);
4647 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4648 	if (ret != -ENOENT)
4649 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4650 		 * but no cached acl data, just the acl length */
4651 		return ret;
4652 	return nfs4_get_acl_uncached(inode, buf, buflen);
4653 }
4654 
4655 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4656 {
4657 	struct nfs_server *server = NFS_SERVER(inode);
4658 	struct page *pages[NFS4ACL_MAXPAGES];
4659 	struct nfs_setaclargs arg = {
4660 		.fh		= NFS_FH(inode),
4661 		.acl_pages	= pages,
4662 		.acl_len	= buflen,
4663 	};
4664 	struct nfs_setaclres res;
4665 	struct rpc_message msg = {
4666 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4667 		.rpc_argp	= &arg,
4668 		.rpc_resp	= &res,
4669 	};
4670 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4671 	int ret, i;
4672 
4673 	if (!nfs4_server_supports_acls(server))
4674 		return -EOPNOTSUPP;
4675 	if (npages > ARRAY_SIZE(pages))
4676 		return -ERANGE;
4677 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4678 	if (i < 0)
4679 		return i;
4680 	nfs4_inode_return_delegation(inode);
4681 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4682 
4683 	/*
4684 	 * Free each page after tx, so the only ref left is
4685 	 * held by the network stack
4686 	 */
4687 	for (; i > 0; i--)
4688 		put_page(pages[i-1]);
4689 
4690 	/*
4691 	 * Acl update can result in inode attribute update.
4692 	 * so mark the attribute cache invalid.
4693 	 */
4694 	spin_lock(&inode->i_lock);
4695 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4696 	spin_unlock(&inode->i_lock);
4697 	nfs_access_zap_cache(inode);
4698 	nfs_zap_acl_cache(inode);
4699 	return ret;
4700 }
4701 
4702 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4703 {
4704 	struct nfs4_exception exception = { };
4705 	int err;
4706 	do {
4707 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4708 		trace_nfs4_set_acl(inode, err);
4709 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4710 				&exception);
4711 	} while (exception.retry);
4712 	return err;
4713 }
4714 
4715 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4716 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4717 					size_t buflen)
4718 {
4719 	struct nfs_server *server = NFS_SERVER(inode);
4720 	struct nfs_fattr fattr;
4721 	struct nfs4_label label = {0, 0, buflen, buf};
4722 
4723 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4724 	struct nfs4_getattr_arg arg = {
4725 		.fh		= NFS_FH(inode),
4726 		.bitmask	= bitmask,
4727 	};
4728 	struct nfs4_getattr_res res = {
4729 		.fattr		= &fattr,
4730 		.label		= &label,
4731 		.server		= server,
4732 	};
4733 	struct rpc_message msg = {
4734 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4735 		.rpc_argp	= &arg,
4736 		.rpc_resp	= &res,
4737 	};
4738 	int ret;
4739 
4740 	nfs_fattr_init(&fattr);
4741 
4742 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4743 	if (ret)
4744 		return ret;
4745 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4746 		return -ENOENT;
4747 	if (buflen < label.len)
4748 		return -ERANGE;
4749 	return 0;
4750 }
4751 
4752 static int nfs4_get_security_label(struct inode *inode, void *buf,
4753 					size_t buflen)
4754 {
4755 	struct nfs4_exception exception = { };
4756 	int err;
4757 
4758 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4759 		return -EOPNOTSUPP;
4760 
4761 	do {
4762 		err = _nfs4_get_security_label(inode, buf, buflen);
4763 		trace_nfs4_get_security_label(inode, err);
4764 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4765 				&exception);
4766 	} while (exception.retry);
4767 	return err;
4768 }
4769 
4770 static int _nfs4_do_set_security_label(struct inode *inode,
4771 		struct nfs4_label *ilabel,
4772 		struct nfs_fattr *fattr,
4773 		struct nfs4_label *olabel)
4774 {
4775 
4776 	struct iattr sattr = {0};
4777 	struct nfs_server *server = NFS_SERVER(inode);
4778 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4779 	struct nfs_setattrargs arg = {
4780 		.fh             = NFS_FH(inode),
4781 		.iap            = &sattr,
4782 		.server		= server,
4783 		.bitmask	= bitmask,
4784 		.label		= ilabel,
4785 	};
4786 	struct nfs_setattrres res = {
4787 		.fattr		= fattr,
4788 		.label		= olabel,
4789 		.server		= server,
4790 	};
4791 	struct rpc_message msg = {
4792 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4793 		.rpc_argp       = &arg,
4794 		.rpc_resp       = &res,
4795 	};
4796 	int status;
4797 
4798 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4799 
4800 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4801 	if (status)
4802 		dprintk("%s failed: %d\n", __func__, status);
4803 
4804 	return status;
4805 }
4806 
4807 static int nfs4_do_set_security_label(struct inode *inode,
4808 		struct nfs4_label *ilabel,
4809 		struct nfs_fattr *fattr,
4810 		struct nfs4_label *olabel)
4811 {
4812 	struct nfs4_exception exception = { };
4813 	int err;
4814 
4815 	do {
4816 		err = _nfs4_do_set_security_label(inode, ilabel,
4817 				fattr, olabel);
4818 		trace_nfs4_set_security_label(inode, err);
4819 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4820 				&exception);
4821 	} while (exception.retry);
4822 	return err;
4823 }
4824 
4825 static int
4826 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4827 {
4828 	struct nfs4_label ilabel, *olabel = NULL;
4829 	struct nfs_fattr fattr;
4830 	struct rpc_cred *cred;
4831 	struct inode *inode = d_inode(dentry);
4832 	int status;
4833 
4834 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4835 		return -EOPNOTSUPP;
4836 
4837 	nfs_fattr_init(&fattr);
4838 
4839 	ilabel.pi = 0;
4840 	ilabel.lfs = 0;
4841 	ilabel.label = (char *)buf;
4842 	ilabel.len = buflen;
4843 
4844 	cred = rpc_lookup_cred();
4845 	if (IS_ERR(cred))
4846 		return PTR_ERR(cred);
4847 
4848 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4849 	if (IS_ERR(olabel)) {
4850 		status = -PTR_ERR(olabel);
4851 		goto out;
4852 	}
4853 
4854 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4855 	if (status == 0)
4856 		nfs_setsecurity(inode, &fattr, olabel);
4857 
4858 	nfs4_label_free(olabel);
4859 out:
4860 	put_rpccred(cred);
4861 	return status;
4862 }
4863 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4864 
4865 
4866 static int
4867 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4868 			struct nfs4_state *state, long *timeout)
4869 {
4870 	struct nfs_client *clp = server->nfs_client;
4871 
4872 	if (task->tk_status >= 0)
4873 		return 0;
4874 	switch(task->tk_status) {
4875 		case -NFS4ERR_DELEG_REVOKED:
4876 		case -NFS4ERR_ADMIN_REVOKED:
4877 		case -NFS4ERR_BAD_STATEID:
4878 		case -NFS4ERR_OPENMODE:
4879 			if (state == NULL)
4880 				break;
4881 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4882 				goto recovery_failed;
4883 			goto wait_on_recovery;
4884 		case -NFS4ERR_EXPIRED:
4885 			if (state != NULL) {
4886 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4887 					goto recovery_failed;
4888 			}
4889 		case -NFS4ERR_STALE_STATEID:
4890 		case -NFS4ERR_STALE_CLIENTID:
4891 			nfs4_schedule_lease_recovery(clp);
4892 			goto wait_on_recovery;
4893 		case -NFS4ERR_MOVED:
4894 			if (nfs4_schedule_migration_recovery(server) < 0)
4895 				goto recovery_failed;
4896 			goto wait_on_recovery;
4897 		case -NFS4ERR_LEASE_MOVED:
4898 			nfs4_schedule_lease_moved_recovery(clp);
4899 			goto wait_on_recovery;
4900 #if defined(CONFIG_NFS_V4_1)
4901 		case -NFS4ERR_BADSESSION:
4902 		case -NFS4ERR_BADSLOT:
4903 		case -NFS4ERR_BAD_HIGH_SLOT:
4904 		case -NFS4ERR_DEADSESSION:
4905 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4906 		case -NFS4ERR_SEQ_FALSE_RETRY:
4907 		case -NFS4ERR_SEQ_MISORDERED:
4908 			dprintk("%s ERROR %d, Reset session\n", __func__,
4909 				task->tk_status);
4910 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4911 			goto wait_on_recovery;
4912 #endif /* CONFIG_NFS_V4_1 */
4913 		case -NFS4ERR_DELAY:
4914 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4915 			rpc_delay(task, nfs4_update_delay(timeout));
4916 			goto restart_call;
4917 		case -NFS4ERR_GRACE:
4918 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4919 		case -NFS4ERR_RETRY_UNCACHED_REP:
4920 		case -NFS4ERR_OLD_STATEID:
4921 			goto restart_call;
4922 	}
4923 	task->tk_status = nfs4_map_errors(task->tk_status);
4924 	return 0;
4925 recovery_failed:
4926 	task->tk_status = -EIO;
4927 	return 0;
4928 wait_on_recovery:
4929 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4930 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4931 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4932 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4933 		goto recovery_failed;
4934 restart_call:
4935 	task->tk_status = 0;
4936 	return -EAGAIN;
4937 }
4938 
4939 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4940 				    nfs4_verifier *bootverf)
4941 {
4942 	__be32 verf[2];
4943 
4944 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4945 		/* An impossible timestamp guarantees this value
4946 		 * will never match a generated boot time. */
4947 		verf[0] = 0;
4948 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4949 	} else {
4950 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4951 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4952 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4953 	}
4954 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4955 }
4956 
4957 static unsigned int
4958 nfs4_init_nonuniform_client_string(struct nfs_client *clp,
4959 				   char *buf, size_t len)
4960 {
4961 	unsigned int result;
4962 
4963 	if (clp->cl_owner_id != NULL)
4964 		return strlcpy(buf, clp->cl_owner_id, len);
4965 
4966 	rcu_read_lock();
4967 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4968 				clp->cl_ipaddr,
4969 				rpc_peeraddr2str(clp->cl_rpcclient,
4970 							RPC_DISPLAY_ADDR),
4971 				rpc_peeraddr2str(clp->cl_rpcclient,
4972 							RPC_DISPLAY_PROTO));
4973 	rcu_read_unlock();
4974 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4975 	return result;
4976 }
4977 
4978 static unsigned int
4979 nfs4_init_uniform_client_string(struct nfs_client *clp,
4980 				char *buf, size_t len)
4981 {
4982 	const char *nodename = clp->cl_rpcclient->cl_nodename;
4983 	unsigned int result;
4984 
4985 	if (clp->cl_owner_id != NULL)
4986 		return strlcpy(buf, clp->cl_owner_id, len);
4987 
4988 	if (nfs4_client_id_uniquifier[0] != '\0')
4989 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4990 				clp->rpc_ops->version,
4991 				clp->cl_minorversion,
4992 				nfs4_client_id_uniquifier,
4993 				nodename);
4994 	else
4995 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s",
4996 				clp->rpc_ops->version, clp->cl_minorversion,
4997 				nodename);
4998 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4999 	return result;
5000 }
5001 
5002 /*
5003  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5004  * services.  Advertise one based on the address family of the
5005  * clientaddr.
5006  */
5007 static unsigned int
5008 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5009 {
5010 	if (strchr(clp->cl_ipaddr, ':') != NULL)
5011 		return scnprintf(buf, len, "tcp6");
5012 	else
5013 		return scnprintf(buf, len, "tcp");
5014 }
5015 
5016 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5017 {
5018 	struct nfs4_setclientid *sc = calldata;
5019 
5020 	if (task->tk_status == 0)
5021 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5022 }
5023 
5024 static const struct rpc_call_ops nfs4_setclientid_ops = {
5025 	.rpc_call_done = nfs4_setclientid_done,
5026 };
5027 
5028 /**
5029  * nfs4_proc_setclientid - Negotiate client ID
5030  * @clp: state data structure
5031  * @program: RPC program for NFSv4 callback service
5032  * @port: IP port number for NFS4 callback service
5033  * @cred: RPC credential to use for this call
5034  * @res: where to place the result
5035  *
5036  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5037  */
5038 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5039 		unsigned short port, struct rpc_cred *cred,
5040 		struct nfs4_setclientid_res *res)
5041 {
5042 	nfs4_verifier sc_verifier;
5043 	struct nfs4_setclientid setclientid = {
5044 		.sc_verifier = &sc_verifier,
5045 		.sc_prog = program,
5046 		.sc_cb_ident = clp->cl_cb_ident,
5047 	};
5048 	struct rpc_message msg = {
5049 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5050 		.rpc_argp = &setclientid,
5051 		.rpc_resp = res,
5052 		.rpc_cred = cred,
5053 	};
5054 	struct rpc_task *task;
5055 	struct rpc_task_setup task_setup_data = {
5056 		.rpc_client = clp->cl_rpcclient,
5057 		.rpc_message = &msg,
5058 		.callback_ops = &nfs4_setclientid_ops,
5059 		.callback_data = &setclientid,
5060 		.flags = RPC_TASK_TIMEOUT,
5061 	};
5062 	int status;
5063 
5064 	/* nfs_client_id4 */
5065 	nfs4_init_boot_verifier(clp, &sc_verifier);
5066 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5067 		setclientid.sc_name_len =
5068 				nfs4_init_uniform_client_string(clp,
5069 						setclientid.sc_name,
5070 						sizeof(setclientid.sc_name));
5071 	else
5072 		setclientid.sc_name_len =
5073 				nfs4_init_nonuniform_client_string(clp,
5074 						setclientid.sc_name,
5075 						sizeof(setclientid.sc_name));
5076 	/* cb_client4 */
5077 	setclientid.sc_netid_len =
5078 				nfs4_init_callback_netid(clp,
5079 						setclientid.sc_netid,
5080 						sizeof(setclientid.sc_netid));
5081 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5082 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5083 				clp->cl_ipaddr, port >> 8, port & 255);
5084 
5085 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5086 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5087 		setclientid.sc_name_len, setclientid.sc_name);
5088 	task = rpc_run_task(&task_setup_data);
5089 	if (IS_ERR(task)) {
5090 		status = PTR_ERR(task);
5091 		goto out;
5092 	}
5093 	status = task->tk_status;
5094 	if (setclientid.sc_cred) {
5095 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5096 		put_rpccred(setclientid.sc_cred);
5097 	}
5098 	rpc_put_task(task);
5099 out:
5100 	trace_nfs4_setclientid(clp, status);
5101 	dprintk("NFS reply setclientid: %d\n", status);
5102 	return status;
5103 }
5104 
5105 /**
5106  * nfs4_proc_setclientid_confirm - Confirm client ID
5107  * @clp: state data structure
5108  * @res: result of a previous SETCLIENTID
5109  * @cred: RPC credential to use for this call
5110  *
5111  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5112  */
5113 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5114 		struct nfs4_setclientid_res *arg,
5115 		struct rpc_cred *cred)
5116 {
5117 	struct rpc_message msg = {
5118 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5119 		.rpc_argp = arg,
5120 		.rpc_cred = cred,
5121 	};
5122 	int status;
5123 
5124 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5125 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5126 		clp->cl_clientid);
5127 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5128 	trace_nfs4_setclientid_confirm(clp, status);
5129 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5130 	return status;
5131 }
5132 
5133 struct nfs4_delegreturndata {
5134 	struct nfs4_delegreturnargs args;
5135 	struct nfs4_delegreturnres res;
5136 	struct nfs_fh fh;
5137 	nfs4_stateid stateid;
5138 	unsigned long timestamp;
5139 	struct nfs_fattr fattr;
5140 	int rpc_status;
5141 	struct inode *inode;
5142 	bool roc;
5143 	u32 roc_barrier;
5144 };
5145 
5146 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5147 {
5148 	struct nfs4_delegreturndata *data = calldata;
5149 
5150 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5151 		return;
5152 
5153 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5154 	switch (task->tk_status) {
5155 	case 0:
5156 		renew_lease(data->res.server, data->timestamp);
5157 	case -NFS4ERR_ADMIN_REVOKED:
5158 	case -NFS4ERR_DELEG_REVOKED:
5159 	case -NFS4ERR_BAD_STATEID:
5160 	case -NFS4ERR_OLD_STATEID:
5161 	case -NFS4ERR_STALE_STATEID:
5162 	case -NFS4ERR_EXPIRED:
5163 		task->tk_status = 0;
5164 		if (data->roc)
5165 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5166 		break;
5167 	default:
5168 		if (nfs4_async_handle_error(task, data->res.server,
5169 					    NULL, NULL) == -EAGAIN) {
5170 			rpc_restart_call_prepare(task);
5171 			return;
5172 		}
5173 	}
5174 	data->rpc_status = task->tk_status;
5175 }
5176 
5177 static void nfs4_delegreturn_release(void *calldata)
5178 {
5179 	struct nfs4_delegreturndata *data = calldata;
5180 	struct inode *inode = data->inode;
5181 
5182 	if (inode) {
5183 		if (data->roc)
5184 			pnfs_roc_release(inode);
5185 		nfs_iput_and_deactive(inode);
5186 	}
5187 	kfree(calldata);
5188 }
5189 
5190 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5191 {
5192 	struct nfs4_delegreturndata *d_data;
5193 
5194 	d_data = (struct nfs4_delegreturndata *)data;
5195 
5196 	if (d_data->roc &&
5197 	    pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5198 		return;
5199 
5200 	nfs4_setup_sequence(d_data->res.server,
5201 			&d_data->args.seq_args,
5202 			&d_data->res.seq_res,
5203 			task);
5204 }
5205 
5206 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5207 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5208 	.rpc_call_done = nfs4_delegreturn_done,
5209 	.rpc_release = nfs4_delegreturn_release,
5210 };
5211 
5212 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5213 {
5214 	struct nfs4_delegreturndata *data;
5215 	struct nfs_server *server = NFS_SERVER(inode);
5216 	struct rpc_task *task;
5217 	struct rpc_message msg = {
5218 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5219 		.rpc_cred = cred,
5220 	};
5221 	struct rpc_task_setup task_setup_data = {
5222 		.rpc_client = server->client,
5223 		.rpc_message = &msg,
5224 		.callback_ops = &nfs4_delegreturn_ops,
5225 		.flags = RPC_TASK_ASYNC,
5226 	};
5227 	int status = 0;
5228 
5229 	data = kzalloc(sizeof(*data), GFP_NOFS);
5230 	if (data == NULL)
5231 		return -ENOMEM;
5232 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5233 	data->args.fhandle = &data->fh;
5234 	data->args.stateid = &data->stateid;
5235 	data->args.bitmask = server->cache_consistency_bitmask;
5236 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5237 	nfs4_stateid_copy(&data->stateid, stateid);
5238 	data->res.fattr = &data->fattr;
5239 	data->res.server = server;
5240 	nfs_fattr_init(data->res.fattr);
5241 	data->timestamp = jiffies;
5242 	data->rpc_status = 0;
5243 	data->inode = nfs_igrab_and_active(inode);
5244 	if (data->inode)
5245 		data->roc = nfs4_roc(inode);
5246 
5247 	task_setup_data.callback_data = data;
5248 	msg.rpc_argp = &data->args;
5249 	msg.rpc_resp = &data->res;
5250 	task = rpc_run_task(&task_setup_data);
5251 	if (IS_ERR(task))
5252 		return PTR_ERR(task);
5253 	if (!issync)
5254 		goto out;
5255 	status = nfs4_wait_for_completion_rpc_task(task);
5256 	if (status != 0)
5257 		goto out;
5258 	status = data->rpc_status;
5259 	if (status == 0)
5260 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5261 	else
5262 		nfs_refresh_inode(inode, &data->fattr);
5263 out:
5264 	rpc_put_task(task);
5265 	return status;
5266 }
5267 
5268 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5269 {
5270 	struct nfs_server *server = NFS_SERVER(inode);
5271 	struct nfs4_exception exception = { };
5272 	int err;
5273 	do {
5274 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5275 		trace_nfs4_delegreturn(inode, err);
5276 		switch (err) {
5277 			case -NFS4ERR_STALE_STATEID:
5278 			case -NFS4ERR_EXPIRED:
5279 			case 0:
5280 				return 0;
5281 		}
5282 		err = nfs4_handle_exception(server, err, &exception);
5283 	} while (exception.retry);
5284 	return err;
5285 }
5286 
5287 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5288 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5289 
5290 /*
5291  * sleep, with exponential backoff, and retry the LOCK operation.
5292  */
5293 static unsigned long
5294 nfs4_set_lock_task_retry(unsigned long timeout)
5295 {
5296 	freezable_schedule_timeout_killable_unsafe(timeout);
5297 	timeout <<= 1;
5298 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5299 		return NFS4_LOCK_MAXTIMEOUT;
5300 	return timeout;
5301 }
5302 
5303 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5304 {
5305 	struct inode *inode = state->inode;
5306 	struct nfs_server *server = NFS_SERVER(inode);
5307 	struct nfs_client *clp = server->nfs_client;
5308 	struct nfs_lockt_args arg = {
5309 		.fh = NFS_FH(inode),
5310 		.fl = request,
5311 	};
5312 	struct nfs_lockt_res res = {
5313 		.denied = request,
5314 	};
5315 	struct rpc_message msg = {
5316 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5317 		.rpc_argp       = &arg,
5318 		.rpc_resp       = &res,
5319 		.rpc_cred	= state->owner->so_cred,
5320 	};
5321 	struct nfs4_lock_state *lsp;
5322 	int status;
5323 
5324 	arg.lock_owner.clientid = clp->cl_clientid;
5325 	status = nfs4_set_lock_state(state, request);
5326 	if (status != 0)
5327 		goto out;
5328 	lsp = request->fl_u.nfs4_fl.owner;
5329 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5330 	arg.lock_owner.s_dev = server->s_dev;
5331 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5332 	switch (status) {
5333 		case 0:
5334 			request->fl_type = F_UNLCK;
5335 			break;
5336 		case -NFS4ERR_DENIED:
5337 			status = 0;
5338 	}
5339 	request->fl_ops->fl_release_private(request);
5340 	request->fl_ops = NULL;
5341 out:
5342 	return status;
5343 }
5344 
5345 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5346 {
5347 	struct nfs4_exception exception = { };
5348 	int err;
5349 
5350 	do {
5351 		err = _nfs4_proc_getlk(state, cmd, request);
5352 		trace_nfs4_get_lock(request, state, cmd, err);
5353 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5354 				&exception);
5355 	} while (exception.retry);
5356 	return err;
5357 }
5358 
5359 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5360 {
5361 	int res = 0;
5362 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5363 		case FL_POSIX:
5364 			res = posix_lock_file_wait(file, fl);
5365 			break;
5366 		case FL_FLOCK:
5367 			res = flock_lock_file_wait(file, fl);
5368 			break;
5369 		default:
5370 			BUG();
5371 	}
5372 	return res;
5373 }
5374 
5375 struct nfs4_unlockdata {
5376 	struct nfs_locku_args arg;
5377 	struct nfs_locku_res res;
5378 	struct nfs4_lock_state *lsp;
5379 	struct nfs_open_context *ctx;
5380 	struct file_lock fl;
5381 	const struct nfs_server *server;
5382 	unsigned long timestamp;
5383 };
5384 
5385 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5386 		struct nfs_open_context *ctx,
5387 		struct nfs4_lock_state *lsp,
5388 		struct nfs_seqid *seqid)
5389 {
5390 	struct nfs4_unlockdata *p;
5391 	struct inode *inode = lsp->ls_state->inode;
5392 
5393 	p = kzalloc(sizeof(*p), GFP_NOFS);
5394 	if (p == NULL)
5395 		return NULL;
5396 	p->arg.fh = NFS_FH(inode);
5397 	p->arg.fl = &p->fl;
5398 	p->arg.seqid = seqid;
5399 	p->res.seqid = seqid;
5400 	p->lsp = lsp;
5401 	atomic_inc(&lsp->ls_count);
5402 	/* Ensure we don't close file until we're done freeing locks! */
5403 	p->ctx = get_nfs_open_context(ctx);
5404 	memcpy(&p->fl, fl, sizeof(p->fl));
5405 	p->server = NFS_SERVER(inode);
5406 	return p;
5407 }
5408 
5409 static void nfs4_locku_release_calldata(void *data)
5410 {
5411 	struct nfs4_unlockdata *calldata = data;
5412 	nfs_free_seqid(calldata->arg.seqid);
5413 	nfs4_put_lock_state(calldata->lsp);
5414 	put_nfs_open_context(calldata->ctx);
5415 	kfree(calldata);
5416 }
5417 
5418 static void nfs4_locku_done(struct rpc_task *task, void *data)
5419 {
5420 	struct nfs4_unlockdata *calldata = data;
5421 
5422 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5423 		return;
5424 	switch (task->tk_status) {
5425 		case 0:
5426 			renew_lease(calldata->server, calldata->timestamp);
5427 			do_vfs_lock(calldata->fl.fl_file, &calldata->fl);
5428 			if (nfs4_update_lock_stateid(calldata->lsp,
5429 					&calldata->res.stateid))
5430 				break;
5431 		case -NFS4ERR_BAD_STATEID:
5432 		case -NFS4ERR_OLD_STATEID:
5433 		case -NFS4ERR_STALE_STATEID:
5434 		case -NFS4ERR_EXPIRED:
5435 			if (!nfs4_stateid_match(&calldata->arg.stateid,
5436 						&calldata->lsp->ls_stateid))
5437 				rpc_restart_call_prepare(task);
5438 			break;
5439 		default:
5440 			if (nfs4_async_handle_error(task, calldata->server,
5441 						    NULL, NULL) == -EAGAIN)
5442 				rpc_restart_call_prepare(task);
5443 	}
5444 	nfs_release_seqid(calldata->arg.seqid);
5445 }
5446 
5447 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5448 {
5449 	struct nfs4_unlockdata *calldata = data;
5450 
5451 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5452 		goto out_wait;
5453 	nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5454 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5455 		/* Note: exit _without_ running nfs4_locku_done */
5456 		goto out_no_action;
5457 	}
5458 	calldata->timestamp = jiffies;
5459 	if (nfs4_setup_sequence(calldata->server,
5460 				&calldata->arg.seq_args,
5461 				&calldata->res.seq_res,
5462 				task) != 0)
5463 		nfs_release_seqid(calldata->arg.seqid);
5464 	return;
5465 out_no_action:
5466 	task->tk_action = NULL;
5467 out_wait:
5468 	nfs4_sequence_done(task, &calldata->res.seq_res);
5469 }
5470 
5471 static const struct rpc_call_ops nfs4_locku_ops = {
5472 	.rpc_call_prepare = nfs4_locku_prepare,
5473 	.rpc_call_done = nfs4_locku_done,
5474 	.rpc_release = nfs4_locku_release_calldata,
5475 };
5476 
5477 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5478 		struct nfs_open_context *ctx,
5479 		struct nfs4_lock_state *lsp,
5480 		struct nfs_seqid *seqid)
5481 {
5482 	struct nfs4_unlockdata *data;
5483 	struct rpc_message msg = {
5484 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5485 		.rpc_cred = ctx->cred,
5486 	};
5487 	struct rpc_task_setup task_setup_data = {
5488 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5489 		.rpc_message = &msg,
5490 		.callback_ops = &nfs4_locku_ops,
5491 		.workqueue = nfsiod_workqueue,
5492 		.flags = RPC_TASK_ASYNC,
5493 	};
5494 
5495 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5496 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5497 
5498 	/* Ensure this is an unlock - when canceling a lock, the
5499 	 * canceled lock is passed in, and it won't be an unlock.
5500 	 */
5501 	fl->fl_type = F_UNLCK;
5502 
5503 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5504 	if (data == NULL) {
5505 		nfs_free_seqid(seqid);
5506 		return ERR_PTR(-ENOMEM);
5507 	}
5508 
5509 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5510 	msg.rpc_argp = &data->arg;
5511 	msg.rpc_resp = &data->res;
5512 	task_setup_data.callback_data = data;
5513 	return rpc_run_task(&task_setup_data);
5514 }
5515 
5516 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5517 {
5518 	struct inode *inode = state->inode;
5519 	struct nfs4_state_owner *sp = state->owner;
5520 	struct nfs_inode *nfsi = NFS_I(inode);
5521 	struct nfs_seqid *seqid;
5522 	struct nfs4_lock_state *lsp;
5523 	struct rpc_task *task;
5524 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5525 	int status = 0;
5526 	unsigned char fl_flags = request->fl_flags;
5527 
5528 	status = nfs4_set_lock_state(state, request);
5529 	/* Unlock _before_ we do the RPC call */
5530 	request->fl_flags |= FL_EXISTS;
5531 	/* Exclude nfs_delegation_claim_locks() */
5532 	mutex_lock(&sp->so_delegreturn_mutex);
5533 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5534 	down_read(&nfsi->rwsem);
5535 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5536 		up_read(&nfsi->rwsem);
5537 		mutex_unlock(&sp->so_delegreturn_mutex);
5538 		goto out;
5539 	}
5540 	up_read(&nfsi->rwsem);
5541 	mutex_unlock(&sp->so_delegreturn_mutex);
5542 	if (status != 0)
5543 		goto out;
5544 	/* Is this a delegated lock? */
5545 	lsp = request->fl_u.nfs4_fl.owner;
5546 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5547 		goto out;
5548 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5549 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5550 	status = -ENOMEM;
5551 	if (IS_ERR(seqid))
5552 		goto out;
5553 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5554 	status = PTR_ERR(task);
5555 	if (IS_ERR(task))
5556 		goto out;
5557 	status = nfs4_wait_for_completion_rpc_task(task);
5558 	rpc_put_task(task);
5559 out:
5560 	request->fl_flags = fl_flags;
5561 	trace_nfs4_unlock(request, state, F_SETLK, status);
5562 	return status;
5563 }
5564 
5565 struct nfs4_lockdata {
5566 	struct nfs_lock_args arg;
5567 	struct nfs_lock_res res;
5568 	struct nfs4_lock_state *lsp;
5569 	struct nfs_open_context *ctx;
5570 	struct file_lock fl;
5571 	unsigned long timestamp;
5572 	int rpc_status;
5573 	int cancelled;
5574 	struct nfs_server *server;
5575 };
5576 
5577 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5578 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5579 		gfp_t gfp_mask)
5580 {
5581 	struct nfs4_lockdata *p;
5582 	struct inode *inode = lsp->ls_state->inode;
5583 	struct nfs_server *server = NFS_SERVER(inode);
5584 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5585 
5586 	p = kzalloc(sizeof(*p), gfp_mask);
5587 	if (p == NULL)
5588 		return NULL;
5589 
5590 	p->arg.fh = NFS_FH(inode);
5591 	p->arg.fl = &p->fl;
5592 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5593 	if (IS_ERR(p->arg.open_seqid))
5594 		goto out_free;
5595 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5596 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5597 	if (IS_ERR(p->arg.lock_seqid))
5598 		goto out_free_seqid;
5599 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5600 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5601 	p->arg.lock_owner.s_dev = server->s_dev;
5602 	p->res.lock_seqid = p->arg.lock_seqid;
5603 	p->lsp = lsp;
5604 	p->server = server;
5605 	atomic_inc(&lsp->ls_count);
5606 	p->ctx = get_nfs_open_context(ctx);
5607 	memcpy(&p->fl, fl, sizeof(p->fl));
5608 	return p;
5609 out_free_seqid:
5610 	nfs_free_seqid(p->arg.open_seqid);
5611 out_free:
5612 	kfree(p);
5613 	return NULL;
5614 }
5615 
5616 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5617 {
5618 	struct nfs4_lockdata *data = calldata;
5619 	struct nfs4_state *state = data->lsp->ls_state;
5620 
5621 	dprintk("%s: begin!\n", __func__);
5622 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5623 		goto out_wait;
5624 	/* Do we need to do an open_to_lock_owner? */
5625 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5626 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5627 			goto out_release_lock_seqid;
5628 		}
5629 		nfs4_stateid_copy(&data->arg.open_stateid,
5630 				&state->open_stateid);
5631 		data->arg.new_lock_owner = 1;
5632 		data->res.open_seqid = data->arg.open_seqid;
5633 	} else {
5634 		data->arg.new_lock_owner = 0;
5635 		nfs4_stateid_copy(&data->arg.lock_stateid,
5636 				&data->lsp->ls_stateid);
5637 	}
5638 	if (!nfs4_valid_open_stateid(state)) {
5639 		data->rpc_status = -EBADF;
5640 		task->tk_action = NULL;
5641 		goto out_release_open_seqid;
5642 	}
5643 	data->timestamp = jiffies;
5644 	if (nfs4_setup_sequence(data->server,
5645 				&data->arg.seq_args,
5646 				&data->res.seq_res,
5647 				task) == 0)
5648 		return;
5649 out_release_open_seqid:
5650 	nfs_release_seqid(data->arg.open_seqid);
5651 out_release_lock_seqid:
5652 	nfs_release_seqid(data->arg.lock_seqid);
5653 out_wait:
5654 	nfs4_sequence_done(task, &data->res.seq_res);
5655 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5656 }
5657 
5658 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5659 {
5660 	struct nfs4_lockdata *data = calldata;
5661 	struct nfs4_lock_state *lsp = data->lsp;
5662 
5663 	dprintk("%s: begin!\n", __func__);
5664 
5665 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5666 		return;
5667 
5668 	data->rpc_status = task->tk_status;
5669 	switch (task->tk_status) {
5670 	case 0:
5671 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5672 				data->timestamp);
5673 		if (data->arg.new_lock) {
5674 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5675 			if (do_vfs_lock(data->fl.fl_file, &data->fl) < 0) {
5676 				rpc_restart_call_prepare(task);
5677 				break;
5678 			}
5679 		}
5680 		if (data->arg.new_lock_owner != 0) {
5681 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
5682 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5683 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5684 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5685 			rpc_restart_call_prepare(task);
5686 		break;
5687 	case -NFS4ERR_BAD_STATEID:
5688 	case -NFS4ERR_OLD_STATEID:
5689 	case -NFS4ERR_STALE_STATEID:
5690 	case -NFS4ERR_EXPIRED:
5691 		if (data->arg.new_lock_owner != 0) {
5692 			if (!nfs4_stateid_match(&data->arg.open_stateid,
5693 						&lsp->ls_state->open_stateid))
5694 				rpc_restart_call_prepare(task);
5695 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5696 						&lsp->ls_stateid))
5697 				rpc_restart_call_prepare(task);
5698 	}
5699 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5700 }
5701 
5702 static void nfs4_lock_release(void *calldata)
5703 {
5704 	struct nfs4_lockdata *data = calldata;
5705 
5706 	dprintk("%s: begin!\n", __func__);
5707 	nfs_free_seqid(data->arg.open_seqid);
5708 	if (data->cancelled != 0) {
5709 		struct rpc_task *task;
5710 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5711 				data->arg.lock_seqid);
5712 		if (!IS_ERR(task))
5713 			rpc_put_task_async(task);
5714 		dprintk("%s: cancelling lock!\n", __func__);
5715 	} else
5716 		nfs_free_seqid(data->arg.lock_seqid);
5717 	nfs4_put_lock_state(data->lsp);
5718 	put_nfs_open_context(data->ctx);
5719 	kfree(data);
5720 	dprintk("%s: done!\n", __func__);
5721 }
5722 
5723 static const struct rpc_call_ops nfs4_lock_ops = {
5724 	.rpc_call_prepare = nfs4_lock_prepare,
5725 	.rpc_call_done = nfs4_lock_done,
5726 	.rpc_release = nfs4_lock_release,
5727 };
5728 
5729 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5730 {
5731 	switch (error) {
5732 	case -NFS4ERR_ADMIN_REVOKED:
5733 	case -NFS4ERR_BAD_STATEID:
5734 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5735 		if (new_lock_owner != 0 ||
5736 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5737 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5738 		break;
5739 	case -NFS4ERR_STALE_STATEID:
5740 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5741 	case -NFS4ERR_EXPIRED:
5742 		nfs4_schedule_lease_recovery(server->nfs_client);
5743 	};
5744 }
5745 
5746 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5747 {
5748 	struct nfs4_lockdata *data;
5749 	struct rpc_task *task;
5750 	struct rpc_message msg = {
5751 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5752 		.rpc_cred = state->owner->so_cred,
5753 	};
5754 	struct rpc_task_setup task_setup_data = {
5755 		.rpc_client = NFS_CLIENT(state->inode),
5756 		.rpc_message = &msg,
5757 		.callback_ops = &nfs4_lock_ops,
5758 		.workqueue = nfsiod_workqueue,
5759 		.flags = RPC_TASK_ASYNC,
5760 	};
5761 	int ret;
5762 
5763 	dprintk("%s: begin!\n", __func__);
5764 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5765 			fl->fl_u.nfs4_fl.owner,
5766 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5767 	if (data == NULL)
5768 		return -ENOMEM;
5769 	if (IS_SETLKW(cmd))
5770 		data->arg.block = 1;
5771 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5772 	msg.rpc_argp = &data->arg;
5773 	msg.rpc_resp = &data->res;
5774 	task_setup_data.callback_data = data;
5775 	if (recovery_type > NFS_LOCK_NEW) {
5776 		if (recovery_type == NFS_LOCK_RECLAIM)
5777 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5778 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5779 	} else
5780 		data->arg.new_lock = 1;
5781 	task = rpc_run_task(&task_setup_data);
5782 	if (IS_ERR(task))
5783 		return PTR_ERR(task);
5784 	ret = nfs4_wait_for_completion_rpc_task(task);
5785 	if (ret == 0) {
5786 		ret = data->rpc_status;
5787 		if (ret)
5788 			nfs4_handle_setlk_error(data->server, data->lsp,
5789 					data->arg.new_lock_owner, ret);
5790 	} else
5791 		data->cancelled = 1;
5792 	rpc_put_task(task);
5793 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5794 	return ret;
5795 }
5796 
5797 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5798 {
5799 	struct nfs_server *server = NFS_SERVER(state->inode);
5800 	struct nfs4_exception exception = {
5801 		.inode = state->inode,
5802 	};
5803 	int err;
5804 
5805 	do {
5806 		/* Cache the lock if possible... */
5807 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5808 			return 0;
5809 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5810 		trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5811 		if (err != -NFS4ERR_DELAY)
5812 			break;
5813 		nfs4_handle_exception(server, err, &exception);
5814 	} while (exception.retry);
5815 	return err;
5816 }
5817 
5818 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5819 {
5820 	struct nfs_server *server = NFS_SERVER(state->inode);
5821 	struct nfs4_exception exception = {
5822 		.inode = state->inode,
5823 	};
5824 	int err;
5825 
5826 	err = nfs4_set_lock_state(state, request);
5827 	if (err != 0)
5828 		return err;
5829 	if (!recover_lost_locks) {
5830 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5831 		return 0;
5832 	}
5833 	do {
5834 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5835 			return 0;
5836 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5837 		trace_nfs4_lock_expired(request, state, F_SETLK, err);
5838 		switch (err) {
5839 		default:
5840 			goto out;
5841 		case -NFS4ERR_GRACE:
5842 		case -NFS4ERR_DELAY:
5843 			nfs4_handle_exception(server, err, &exception);
5844 			err = 0;
5845 		}
5846 	} while (exception.retry);
5847 out:
5848 	return err;
5849 }
5850 
5851 #if defined(CONFIG_NFS_V4_1)
5852 /**
5853  * nfs41_check_expired_locks - possibly free a lock stateid
5854  *
5855  * @state: NFSv4 state for an inode
5856  *
5857  * Returns NFS_OK if recovery for this stateid is now finished.
5858  * Otherwise a negative NFS4ERR value is returned.
5859  */
5860 static int nfs41_check_expired_locks(struct nfs4_state *state)
5861 {
5862 	int status, ret = -NFS4ERR_BAD_STATEID;
5863 	struct nfs4_lock_state *lsp;
5864 	struct nfs_server *server = NFS_SERVER(state->inode);
5865 
5866 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5867 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5868 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5869 
5870 			status = nfs41_test_stateid(server,
5871 					&lsp->ls_stateid,
5872 					cred);
5873 			trace_nfs4_test_lock_stateid(state, lsp, status);
5874 			if (status != NFS_OK) {
5875 				/* Free the stateid unless the server
5876 				 * informs us the stateid is unrecognized. */
5877 				if (status != -NFS4ERR_BAD_STATEID)
5878 					nfs41_free_stateid(server,
5879 							&lsp->ls_stateid,
5880 							cred);
5881 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5882 				ret = status;
5883 			}
5884 		}
5885 	};
5886 
5887 	return ret;
5888 }
5889 
5890 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5891 {
5892 	int status = NFS_OK;
5893 
5894 	if (test_bit(LK_STATE_IN_USE, &state->flags))
5895 		status = nfs41_check_expired_locks(state);
5896 	if (status != NFS_OK)
5897 		status = nfs4_lock_expired(state, request);
5898 	return status;
5899 }
5900 #endif
5901 
5902 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5903 {
5904 	struct nfs_inode *nfsi = NFS_I(state->inode);
5905 	unsigned char fl_flags = request->fl_flags;
5906 	int status = -ENOLCK;
5907 
5908 	if ((fl_flags & FL_POSIX) &&
5909 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5910 		goto out;
5911 	/* Is this a delegated open? */
5912 	status = nfs4_set_lock_state(state, request);
5913 	if (status != 0)
5914 		goto out;
5915 	request->fl_flags |= FL_ACCESS;
5916 	status = do_vfs_lock(request->fl_file, request);
5917 	if (status < 0)
5918 		goto out;
5919 	down_read(&nfsi->rwsem);
5920 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5921 		/* Yes: cache locks! */
5922 		/* ...but avoid races with delegation recall... */
5923 		request->fl_flags = fl_flags & ~FL_SLEEP;
5924 		status = do_vfs_lock(request->fl_file, request);
5925 		up_read(&nfsi->rwsem);
5926 		goto out;
5927 	}
5928 	up_read(&nfsi->rwsem);
5929 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5930 out:
5931 	request->fl_flags = fl_flags;
5932 	return status;
5933 }
5934 
5935 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5936 {
5937 	struct nfs4_exception exception = {
5938 		.state = state,
5939 		.inode = state->inode,
5940 	};
5941 	int err;
5942 
5943 	do {
5944 		err = _nfs4_proc_setlk(state, cmd, request);
5945 		trace_nfs4_set_lock(request, state, cmd, err);
5946 		if (err == -NFS4ERR_DENIED)
5947 			err = -EAGAIN;
5948 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
5949 				err, &exception);
5950 	} while (exception.retry);
5951 	return err;
5952 }
5953 
5954 static int
5955 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5956 {
5957 	struct nfs_open_context *ctx;
5958 	struct nfs4_state *state;
5959 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5960 	int status;
5961 
5962 	/* verify open state */
5963 	ctx = nfs_file_open_context(filp);
5964 	state = ctx->state;
5965 
5966 	if (request->fl_start < 0 || request->fl_end < 0)
5967 		return -EINVAL;
5968 
5969 	if (IS_GETLK(cmd)) {
5970 		if (state != NULL)
5971 			return nfs4_proc_getlk(state, F_GETLK, request);
5972 		return 0;
5973 	}
5974 
5975 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5976 		return -EINVAL;
5977 
5978 	if (request->fl_type == F_UNLCK) {
5979 		if (state != NULL)
5980 			return nfs4_proc_unlck(state, cmd, request);
5981 		return 0;
5982 	}
5983 
5984 	if (state == NULL)
5985 		return -ENOLCK;
5986 	/*
5987 	 * Don't rely on the VFS having checked the file open mode,
5988 	 * since it won't do this for flock() locks.
5989 	 */
5990 	switch (request->fl_type) {
5991 	case F_RDLCK:
5992 		if (!(filp->f_mode & FMODE_READ))
5993 			return -EBADF;
5994 		break;
5995 	case F_WRLCK:
5996 		if (!(filp->f_mode & FMODE_WRITE))
5997 			return -EBADF;
5998 	}
5999 
6000 	do {
6001 		status = nfs4_proc_setlk(state, cmd, request);
6002 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6003 			break;
6004 		timeout = nfs4_set_lock_task_retry(timeout);
6005 		status = -ERESTARTSYS;
6006 		if (signalled())
6007 			break;
6008 	} while(status < 0);
6009 	return status;
6010 }
6011 
6012 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6013 {
6014 	struct nfs_server *server = NFS_SERVER(state->inode);
6015 	int err;
6016 
6017 	err = nfs4_set_lock_state(state, fl);
6018 	if (err != 0)
6019 		return err;
6020 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6021 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6022 }
6023 
6024 struct nfs_release_lockowner_data {
6025 	struct nfs4_lock_state *lsp;
6026 	struct nfs_server *server;
6027 	struct nfs_release_lockowner_args args;
6028 	struct nfs_release_lockowner_res res;
6029 	unsigned long timestamp;
6030 };
6031 
6032 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6033 {
6034 	struct nfs_release_lockowner_data *data = calldata;
6035 	struct nfs_server *server = data->server;
6036 	nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6037 			     &data->args.seq_args, &data->res.seq_res, task);
6038 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6039 	data->timestamp = jiffies;
6040 }
6041 
6042 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6043 {
6044 	struct nfs_release_lockowner_data *data = calldata;
6045 	struct nfs_server *server = data->server;
6046 
6047 	nfs40_sequence_done(task, &data->res.seq_res);
6048 
6049 	switch (task->tk_status) {
6050 	case 0:
6051 		renew_lease(server, data->timestamp);
6052 		break;
6053 	case -NFS4ERR_STALE_CLIENTID:
6054 	case -NFS4ERR_EXPIRED:
6055 		nfs4_schedule_lease_recovery(server->nfs_client);
6056 		break;
6057 	case -NFS4ERR_LEASE_MOVED:
6058 	case -NFS4ERR_DELAY:
6059 		if (nfs4_async_handle_error(task, server,
6060 					    NULL, NULL) == -EAGAIN)
6061 			rpc_restart_call_prepare(task);
6062 	}
6063 }
6064 
6065 static void nfs4_release_lockowner_release(void *calldata)
6066 {
6067 	struct nfs_release_lockowner_data *data = calldata;
6068 	nfs4_free_lock_state(data->server, data->lsp);
6069 	kfree(calldata);
6070 }
6071 
6072 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6073 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6074 	.rpc_call_done = nfs4_release_lockowner_done,
6075 	.rpc_release = nfs4_release_lockowner_release,
6076 };
6077 
6078 static void
6079 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6080 {
6081 	struct nfs_release_lockowner_data *data;
6082 	struct rpc_message msg = {
6083 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6084 	};
6085 
6086 	if (server->nfs_client->cl_mvops->minor_version != 0)
6087 		return;
6088 
6089 	data = kmalloc(sizeof(*data), GFP_NOFS);
6090 	if (!data)
6091 		return;
6092 	data->lsp = lsp;
6093 	data->server = server;
6094 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6095 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6096 	data->args.lock_owner.s_dev = server->s_dev;
6097 
6098 	msg.rpc_argp = &data->args;
6099 	msg.rpc_resp = &data->res;
6100 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6101 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6102 }
6103 
6104 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6105 
6106 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6107 				   const void *buf, size_t buflen,
6108 				   int flags, int type)
6109 {
6110 	if (strcmp(key, "") != 0)
6111 		return -EINVAL;
6112 
6113 	return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6114 }
6115 
6116 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6117 				   void *buf, size_t buflen, int type)
6118 {
6119 	if (strcmp(key, "") != 0)
6120 		return -EINVAL;
6121 
6122 	return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6123 }
6124 
6125 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6126 				       size_t list_len, const char *name,
6127 				       size_t name_len, int type)
6128 {
6129 	size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6130 
6131 	if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6132 		return 0;
6133 
6134 	if (list && len <= list_len)
6135 		memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6136 	return len;
6137 }
6138 
6139 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6140 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6141 {
6142 	return server->caps & NFS_CAP_SECURITY_LABEL;
6143 }
6144 
6145 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6146 				   const void *buf, size_t buflen,
6147 				   int flags, int type)
6148 {
6149 	if (security_ismaclabel(key))
6150 		return nfs4_set_security_label(dentry, buf, buflen);
6151 
6152 	return -EOPNOTSUPP;
6153 }
6154 
6155 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6156 				   void *buf, size_t buflen, int type)
6157 {
6158 	if (security_ismaclabel(key))
6159 		return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6160 	return -EOPNOTSUPP;
6161 }
6162 
6163 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6164 				       size_t list_len, const char *name,
6165 				       size_t name_len, int type)
6166 {
6167 	size_t len = 0;
6168 
6169 	if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6170 		len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6171 		if (list && len <= list_len)
6172 			security_inode_listsecurity(d_inode(dentry), list, len);
6173 	}
6174 	return len;
6175 }
6176 
6177 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6178 	.prefix = XATTR_SECURITY_PREFIX,
6179 	.list	= nfs4_xattr_list_nfs4_label,
6180 	.get	= nfs4_xattr_get_nfs4_label,
6181 	.set	= nfs4_xattr_set_nfs4_label,
6182 };
6183 #endif
6184 
6185 
6186 /*
6187  * nfs_fhget will use either the mounted_on_fileid or the fileid
6188  */
6189 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6190 {
6191 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6192 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6193 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6194 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6195 		return;
6196 
6197 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6198 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6199 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6200 	fattr->nlink = 2;
6201 }
6202 
6203 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6204 				   const struct qstr *name,
6205 				   struct nfs4_fs_locations *fs_locations,
6206 				   struct page *page)
6207 {
6208 	struct nfs_server *server = NFS_SERVER(dir);
6209 	u32 bitmask[3] = {
6210 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6211 	};
6212 	struct nfs4_fs_locations_arg args = {
6213 		.dir_fh = NFS_FH(dir),
6214 		.name = name,
6215 		.page = page,
6216 		.bitmask = bitmask,
6217 	};
6218 	struct nfs4_fs_locations_res res = {
6219 		.fs_locations = fs_locations,
6220 	};
6221 	struct rpc_message msg = {
6222 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6223 		.rpc_argp = &args,
6224 		.rpc_resp = &res,
6225 	};
6226 	int status;
6227 
6228 	dprintk("%s: start\n", __func__);
6229 
6230 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6231 	 * is not supported */
6232 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6233 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6234 	else
6235 		bitmask[0] |= FATTR4_WORD0_FILEID;
6236 
6237 	nfs_fattr_init(&fs_locations->fattr);
6238 	fs_locations->server = server;
6239 	fs_locations->nlocations = 0;
6240 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6241 	dprintk("%s: returned status = %d\n", __func__, status);
6242 	return status;
6243 }
6244 
6245 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6246 			   const struct qstr *name,
6247 			   struct nfs4_fs_locations *fs_locations,
6248 			   struct page *page)
6249 {
6250 	struct nfs4_exception exception = { };
6251 	int err;
6252 	do {
6253 		err = _nfs4_proc_fs_locations(client, dir, name,
6254 				fs_locations, page);
6255 		trace_nfs4_get_fs_locations(dir, name, err);
6256 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6257 				&exception);
6258 	} while (exception.retry);
6259 	return err;
6260 }
6261 
6262 /*
6263  * This operation also signals the server that this client is
6264  * performing migration recovery.  The server can stop returning
6265  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6266  * appended to this compound to identify the client ID which is
6267  * performing recovery.
6268  */
6269 static int _nfs40_proc_get_locations(struct inode *inode,
6270 				     struct nfs4_fs_locations *locations,
6271 				     struct page *page, struct rpc_cred *cred)
6272 {
6273 	struct nfs_server *server = NFS_SERVER(inode);
6274 	struct rpc_clnt *clnt = server->client;
6275 	u32 bitmask[2] = {
6276 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6277 	};
6278 	struct nfs4_fs_locations_arg args = {
6279 		.clientid	= server->nfs_client->cl_clientid,
6280 		.fh		= NFS_FH(inode),
6281 		.page		= page,
6282 		.bitmask	= bitmask,
6283 		.migration	= 1,		/* skip LOOKUP */
6284 		.renew		= 1,		/* append RENEW */
6285 	};
6286 	struct nfs4_fs_locations_res res = {
6287 		.fs_locations	= locations,
6288 		.migration	= 1,
6289 		.renew		= 1,
6290 	};
6291 	struct rpc_message msg = {
6292 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6293 		.rpc_argp	= &args,
6294 		.rpc_resp	= &res,
6295 		.rpc_cred	= cred,
6296 	};
6297 	unsigned long now = jiffies;
6298 	int status;
6299 
6300 	nfs_fattr_init(&locations->fattr);
6301 	locations->server = server;
6302 	locations->nlocations = 0;
6303 
6304 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6305 	nfs4_set_sequence_privileged(&args.seq_args);
6306 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6307 					&args.seq_args, &res.seq_res);
6308 	if (status)
6309 		return status;
6310 
6311 	renew_lease(server, now);
6312 	return 0;
6313 }
6314 
6315 #ifdef CONFIG_NFS_V4_1
6316 
6317 /*
6318  * This operation also signals the server that this client is
6319  * performing migration recovery.  The server can stop asserting
6320  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6321  * performing this operation is identified in the SEQUENCE
6322  * operation in this compound.
6323  *
6324  * When the client supports GETATTR(fs_locations_info), it can
6325  * be plumbed in here.
6326  */
6327 static int _nfs41_proc_get_locations(struct inode *inode,
6328 				     struct nfs4_fs_locations *locations,
6329 				     struct page *page, struct rpc_cred *cred)
6330 {
6331 	struct nfs_server *server = NFS_SERVER(inode);
6332 	struct rpc_clnt *clnt = server->client;
6333 	u32 bitmask[2] = {
6334 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6335 	};
6336 	struct nfs4_fs_locations_arg args = {
6337 		.fh		= NFS_FH(inode),
6338 		.page		= page,
6339 		.bitmask	= bitmask,
6340 		.migration	= 1,		/* skip LOOKUP */
6341 	};
6342 	struct nfs4_fs_locations_res res = {
6343 		.fs_locations	= locations,
6344 		.migration	= 1,
6345 	};
6346 	struct rpc_message msg = {
6347 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6348 		.rpc_argp	= &args,
6349 		.rpc_resp	= &res,
6350 		.rpc_cred	= cred,
6351 	};
6352 	int status;
6353 
6354 	nfs_fattr_init(&locations->fattr);
6355 	locations->server = server;
6356 	locations->nlocations = 0;
6357 
6358 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6359 	nfs4_set_sequence_privileged(&args.seq_args);
6360 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6361 					&args.seq_args, &res.seq_res);
6362 	if (status == NFS4_OK &&
6363 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6364 		status = -NFS4ERR_LEASE_MOVED;
6365 	return status;
6366 }
6367 
6368 #endif	/* CONFIG_NFS_V4_1 */
6369 
6370 /**
6371  * nfs4_proc_get_locations - discover locations for a migrated FSID
6372  * @inode: inode on FSID that is migrating
6373  * @locations: result of query
6374  * @page: buffer
6375  * @cred: credential to use for this operation
6376  *
6377  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6378  * operation failed, or a negative errno if a local error occurred.
6379  *
6380  * On success, "locations" is filled in, but if the server has
6381  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6382  * asserted.
6383  *
6384  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6385  * from this client that require migration recovery.
6386  */
6387 int nfs4_proc_get_locations(struct inode *inode,
6388 			    struct nfs4_fs_locations *locations,
6389 			    struct page *page, struct rpc_cred *cred)
6390 {
6391 	struct nfs_server *server = NFS_SERVER(inode);
6392 	struct nfs_client *clp = server->nfs_client;
6393 	const struct nfs4_mig_recovery_ops *ops =
6394 					clp->cl_mvops->mig_recovery_ops;
6395 	struct nfs4_exception exception = { };
6396 	int status;
6397 
6398 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6399 		(unsigned long long)server->fsid.major,
6400 		(unsigned long long)server->fsid.minor,
6401 		clp->cl_hostname);
6402 	nfs_display_fhandle(NFS_FH(inode), __func__);
6403 
6404 	do {
6405 		status = ops->get_locations(inode, locations, page, cred);
6406 		if (status != -NFS4ERR_DELAY)
6407 			break;
6408 		nfs4_handle_exception(server, status, &exception);
6409 	} while (exception.retry);
6410 	return status;
6411 }
6412 
6413 /*
6414  * This operation also signals the server that this client is
6415  * performing "lease moved" recovery.  The server can stop
6416  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6417  * is appended to this compound to identify the client ID which is
6418  * performing recovery.
6419  */
6420 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6421 {
6422 	struct nfs_server *server = NFS_SERVER(inode);
6423 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6424 	struct rpc_clnt *clnt = server->client;
6425 	struct nfs4_fsid_present_arg args = {
6426 		.fh		= NFS_FH(inode),
6427 		.clientid	= clp->cl_clientid,
6428 		.renew		= 1,		/* append RENEW */
6429 	};
6430 	struct nfs4_fsid_present_res res = {
6431 		.renew		= 1,
6432 	};
6433 	struct rpc_message msg = {
6434 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6435 		.rpc_argp	= &args,
6436 		.rpc_resp	= &res,
6437 		.rpc_cred	= cred,
6438 	};
6439 	unsigned long now = jiffies;
6440 	int status;
6441 
6442 	res.fh = nfs_alloc_fhandle();
6443 	if (res.fh == NULL)
6444 		return -ENOMEM;
6445 
6446 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6447 	nfs4_set_sequence_privileged(&args.seq_args);
6448 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6449 						&args.seq_args, &res.seq_res);
6450 	nfs_free_fhandle(res.fh);
6451 	if (status)
6452 		return status;
6453 
6454 	do_renew_lease(clp, now);
6455 	return 0;
6456 }
6457 
6458 #ifdef CONFIG_NFS_V4_1
6459 
6460 /*
6461  * This operation also signals the server that this client is
6462  * performing "lease moved" recovery.  The server can stop asserting
6463  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6464  * this operation is identified in the SEQUENCE operation in this
6465  * compound.
6466  */
6467 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6468 {
6469 	struct nfs_server *server = NFS_SERVER(inode);
6470 	struct rpc_clnt *clnt = server->client;
6471 	struct nfs4_fsid_present_arg args = {
6472 		.fh		= NFS_FH(inode),
6473 	};
6474 	struct nfs4_fsid_present_res res = {
6475 	};
6476 	struct rpc_message msg = {
6477 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6478 		.rpc_argp	= &args,
6479 		.rpc_resp	= &res,
6480 		.rpc_cred	= cred,
6481 	};
6482 	int status;
6483 
6484 	res.fh = nfs_alloc_fhandle();
6485 	if (res.fh == NULL)
6486 		return -ENOMEM;
6487 
6488 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6489 	nfs4_set_sequence_privileged(&args.seq_args);
6490 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6491 						&args.seq_args, &res.seq_res);
6492 	nfs_free_fhandle(res.fh);
6493 	if (status == NFS4_OK &&
6494 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6495 		status = -NFS4ERR_LEASE_MOVED;
6496 	return status;
6497 }
6498 
6499 #endif	/* CONFIG_NFS_V4_1 */
6500 
6501 /**
6502  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6503  * @inode: inode on FSID to check
6504  * @cred: credential to use for this operation
6505  *
6506  * Server indicates whether the FSID is present, moved, or not
6507  * recognized.  This operation is necessary to clear a LEASE_MOVED
6508  * condition for this client ID.
6509  *
6510  * Returns NFS4_OK if the FSID is present on this server,
6511  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6512  *  NFS4ERR code if some error occurred on the server, or a
6513  *  negative errno if a local failure occurred.
6514  */
6515 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6516 {
6517 	struct nfs_server *server = NFS_SERVER(inode);
6518 	struct nfs_client *clp = server->nfs_client;
6519 	const struct nfs4_mig_recovery_ops *ops =
6520 					clp->cl_mvops->mig_recovery_ops;
6521 	struct nfs4_exception exception = { };
6522 	int status;
6523 
6524 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6525 		(unsigned long long)server->fsid.major,
6526 		(unsigned long long)server->fsid.minor,
6527 		clp->cl_hostname);
6528 	nfs_display_fhandle(NFS_FH(inode), __func__);
6529 
6530 	do {
6531 		status = ops->fsid_present(inode, cred);
6532 		if (status != -NFS4ERR_DELAY)
6533 			break;
6534 		nfs4_handle_exception(server, status, &exception);
6535 	} while (exception.retry);
6536 	return status;
6537 }
6538 
6539 /**
6540  * If 'use_integrity' is true and the state managment nfs_client
6541  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6542  * and the machine credential as per RFC3530bis and RFC5661 Security
6543  * Considerations sections. Otherwise, just use the user cred with the
6544  * filesystem's rpc_client.
6545  */
6546 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6547 {
6548 	int status;
6549 	struct nfs4_secinfo_arg args = {
6550 		.dir_fh = NFS_FH(dir),
6551 		.name   = name,
6552 	};
6553 	struct nfs4_secinfo_res res = {
6554 		.flavors     = flavors,
6555 	};
6556 	struct rpc_message msg = {
6557 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6558 		.rpc_argp = &args,
6559 		.rpc_resp = &res,
6560 	};
6561 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6562 	struct rpc_cred *cred = NULL;
6563 
6564 	if (use_integrity) {
6565 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6566 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6567 		msg.rpc_cred = cred;
6568 	}
6569 
6570 	dprintk("NFS call  secinfo %s\n", name->name);
6571 
6572 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6573 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6574 
6575 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6576 				&res.seq_res, 0);
6577 	dprintk("NFS reply  secinfo: %d\n", status);
6578 
6579 	if (cred)
6580 		put_rpccred(cred);
6581 
6582 	return status;
6583 }
6584 
6585 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6586 		      struct nfs4_secinfo_flavors *flavors)
6587 {
6588 	struct nfs4_exception exception = { };
6589 	int err;
6590 	do {
6591 		err = -NFS4ERR_WRONGSEC;
6592 
6593 		/* try to use integrity protection with machine cred */
6594 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6595 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6596 
6597 		/*
6598 		 * if unable to use integrity protection, or SECINFO with
6599 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6600 		 * disallowed by spec, but exists in deployed servers) use
6601 		 * the current filesystem's rpc_client and the user cred.
6602 		 */
6603 		if (err == -NFS4ERR_WRONGSEC)
6604 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6605 
6606 		trace_nfs4_secinfo(dir, name, err);
6607 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6608 				&exception);
6609 	} while (exception.retry);
6610 	return err;
6611 }
6612 
6613 #ifdef CONFIG_NFS_V4_1
6614 /*
6615  * Check the exchange flags returned by the server for invalid flags, having
6616  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6617  * DS flags set.
6618  */
6619 static int nfs4_check_cl_exchange_flags(u32 flags)
6620 {
6621 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6622 		goto out_inval;
6623 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6624 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6625 		goto out_inval;
6626 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6627 		goto out_inval;
6628 	return NFS_OK;
6629 out_inval:
6630 	return -NFS4ERR_INVAL;
6631 }
6632 
6633 static bool
6634 nfs41_same_server_scope(struct nfs41_server_scope *a,
6635 			struct nfs41_server_scope *b)
6636 {
6637 	if (a->server_scope_sz == b->server_scope_sz &&
6638 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6639 		return true;
6640 
6641 	return false;
6642 }
6643 
6644 /*
6645  * nfs4_proc_bind_conn_to_session()
6646  *
6647  * The 4.1 client currently uses the same TCP connection for the
6648  * fore and backchannel.
6649  */
6650 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6651 {
6652 	int status;
6653 	struct nfs41_bind_conn_to_session_args args = {
6654 		.client = clp,
6655 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
6656 	};
6657 	struct nfs41_bind_conn_to_session_res res;
6658 	struct rpc_message msg = {
6659 		.rpc_proc =
6660 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6661 		.rpc_argp = &args,
6662 		.rpc_resp = &res,
6663 		.rpc_cred = cred,
6664 	};
6665 
6666 	dprintk("--> %s\n", __func__);
6667 
6668 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6669 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6670 		args.dir = NFS4_CDFC4_FORE;
6671 
6672 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6673 	trace_nfs4_bind_conn_to_session(clp, status);
6674 	if (status == 0) {
6675 		if (memcmp(res.sessionid.data,
6676 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6677 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6678 			status = -EIO;
6679 			goto out;
6680 		}
6681 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6682 			dprintk("NFS: %s: Unexpected direction from server\n",
6683 				__func__);
6684 			status = -EIO;
6685 			goto out;
6686 		}
6687 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6688 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6689 				__func__);
6690 			status = -EIO;
6691 			goto out;
6692 		}
6693 	}
6694 out:
6695 	dprintk("<-- %s status= %d\n", __func__, status);
6696 	return status;
6697 }
6698 
6699 /*
6700  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6701  * and operations we'd like to see to enable certain features in the allow map
6702  */
6703 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6704 	.how = SP4_MACH_CRED,
6705 	.enforce.u.words = {
6706 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6707 		      1 << (OP_EXCHANGE_ID - 32) |
6708 		      1 << (OP_CREATE_SESSION - 32) |
6709 		      1 << (OP_DESTROY_SESSION - 32) |
6710 		      1 << (OP_DESTROY_CLIENTID - 32)
6711 	},
6712 	.allow.u.words = {
6713 		[0] = 1 << (OP_CLOSE) |
6714 		      1 << (OP_LOCKU) |
6715 		      1 << (OP_COMMIT),
6716 		[1] = 1 << (OP_SECINFO - 32) |
6717 		      1 << (OP_SECINFO_NO_NAME - 32) |
6718 		      1 << (OP_TEST_STATEID - 32) |
6719 		      1 << (OP_FREE_STATEID - 32) |
6720 		      1 << (OP_WRITE - 32)
6721 	}
6722 };
6723 
6724 /*
6725  * Select the state protection mode for client `clp' given the server results
6726  * from exchange_id in `sp'.
6727  *
6728  * Returns 0 on success, negative errno otherwise.
6729  */
6730 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6731 				 struct nfs41_state_protection *sp)
6732 {
6733 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6734 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6735 		      1 << (OP_EXCHANGE_ID - 32) |
6736 		      1 << (OP_CREATE_SESSION - 32) |
6737 		      1 << (OP_DESTROY_SESSION - 32) |
6738 		      1 << (OP_DESTROY_CLIENTID - 32)
6739 	};
6740 	unsigned int i;
6741 
6742 	if (sp->how == SP4_MACH_CRED) {
6743 		/* Print state protect result */
6744 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6745 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6746 			if (test_bit(i, sp->enforce.u.longs))
6747 				dfprintk(MOUNT, "  enforce op %d\n", i);
6748 			if (test_bit(i, sp->allow.u.longs))
6749 				dfprintk(MOUNT, "  allow op %d\n", i);
6750 		}
6751 
6752 		/* make sure nothing is on enforce list that isn't supported */
6753 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6754 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6755 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6756 				return -EINVAL;
6757 			}
6758 		}
6759 
6760 		/*
6761 		 * Minimal mode - state operations are allowed to use machine
6762 		 * credential.  Note this already happens by default, so the
6763 		 * client doesn't have to do anything more than the negotiation.
6764 		 *
6765 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6766 		 *       we're already using the machine cred for exchange_id
6767 		 *       and will never use a different cred.
6768 		 */
6769 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6770 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6771 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6772 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6773 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6774 			dfprintk(MOUNT, "  minimal mode enabled\n");
6775 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6776 		} else {
6777 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6778 			return -EINVAL;
6779 		}
6780 
6781 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6782 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
6783 			dfprintk(MOUNT, "  cleanup mode enabled\n");
6784 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6785 		}
6786 
6787 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6788 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6789 			dfprintk(MOUNT, "  secinfo mode enabled\n");
6790 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6791 		}
6792 
6793 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6794 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6795 			dfprintk(MOUNT, "  stateid mode enabled\n");
6796 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6797 		}
6798 
6799 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6800 			dfprintk(MOUNT, "  write mode enabled\n");
6801 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6802 		}
6803 
6804 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6805 			dfprintk(MOUNT, "  commit mode enabled\n");
6806 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6807 		}
6808 	}
6809 
6810 	return 0;
6811 }
6812 
6813 /*
6814  * _nfs4_proc_exchange_id()
6815  *
6816  * Wrapper for EXCHANGE_ID operation.
6817  */
6818 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6819 	u32 sp4_how)
6820 {
6821 	nfs4_verifier verifier;
6822 	struct nfs41_exchange_id_args args = {
6823 		.verifier = &verifier,
6824 		.client = clp,
6825 #ifdef CONFIG_NFS_V4_1_MIGRATION
6826 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6827 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
6828 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6829 #else
6830 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6831 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
6832 #endif
6833 	};
6834 	struct nfs41_exchange_id_res res = {
6835 		0
6836 	};
6837 	int status;
6838 	struct rpc_message msg = {
6839 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6840 		.rpc_argp = &args,
6841 		.rpc_resp = &res,
6842 		.rpc_cred = cred,
6843 	};
6844 
6845 	nfs4_init_boot_verifier(clp, &verifier);
6846 	args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6847 							sizeof(args.id));
6848 	dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6849 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6850 		args.id_len, args.id);
6851 
6852 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6853 					GFP_NOFS);
6854 	if (unlikely(res.server_owner == NULL)) {
6855 		status = -ENOMEM;
6856 		goto out;
6857 	}
6858 
6859 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6860 					GFP_NOFS);
6861 	if (unlikely(res.server_scope == NULL)) {
6862 		status = -ENOMEM;
6863 		goto out_server_owner;
6864 	}
6865 
6866 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6867 	if (unlikely(res.impl_id == NULL)) {
6868 		status = -ENOMEM;
6869 		goto out_server_scope;
6870 	}
6871 
6872 	switch (sp4_how) {
6873 	case SP4_NONE:
6874 		args.state_protect.how = SP4_NONE;
6875 		break;
6876 
6877 	case SP4_MACH_CRED:
6878 		args.state_protect = nfs4_sp4_mach_cred_request;
6879 		break;
6880 
6881 	default:
6882 		/* unsupported! */
6883 		WARN_ON_ONCE(1);
6884 		status = -EINVAL;
6885 		goto out_server_scope;
6886 	}
6887 
6888 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6889 	trace_nfs4_exchange_id(clp, status);
6890 	if (status == 0)
6891 		status = nfs4_check_cl_exchange_flags(res.flags);
6892 
6893 	if (status == 0)
6894 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
6895 
6896 	if (status == 0) {
6897 		clp->cl_clientid = res.clientid;
6898 		clp->cl_exchange_flags = res.flags;
6899 		/* Client ID is not confirmed */
6900 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
6901 			clear_bit(NFS4_SESSION_ESTABLISHED,
6902 					&clp->cl_session->session_state);
6903 			clp->cl_seqid = res.seqid;
6904 		}
6905 
6906 		kfree(clp->cl_serverowner);
6907 		clp->cl_serverowner = res.server_owner;
6908 		res.server_owner = NULL;
6909 
6910 		/* use the most recent implementation id */
6911 		kfree(clp->cl_implid);
6912 		clp->cl_implid = res.impl_id;
6913 
6914 		if (clp->cl_serverscope != NULL &&
6915 		    !nfs41_same_server_scope(clp->cl_serverscope,
6916 					     res.server_scope)) {
6917 			dprintk("%s: server_scope mismatch detected\n",
6918 				__func__);
6919 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6920 			kfree(clp->cl_serverscope);
6921 			clp->cl_serverscope = NULL;
6922 		}
6923 
6924 		if (clp->cl_serverscope == NULL) {
6925 			clp->cl_serverscope = res.server_scope;
6926 			goto out;
6927 		}
6928 	} else
6929 		kfree(res.impl_id);
6930 
6931 out_server_owner:
6932 	kfree(res.server_owner);
6933 out_server_scope:
6934 	kfree(res.server_scope);
6935 out:
6936 	if (clp->cl_implid != NULL)
6937 		dprintk("NFS reply exchange_id: Server Implementation ID: "
6938 			"domain: %s, name: %s, date: %llu,%u\n",
6939 			clp->cl_implid->domain, clp->cl_implid->name,
6940 			clp->cl_implid->date.seconds,
6941 			clp->cl_implid->date.nseconds);
6942 	dprintk("NFS reply exchange_id: %d\n", status);
6943 	return status;
6944 }
6945 
6946 /*
6947  * nfs4_proc_exchange_id()
6948  *
6949  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6950  *
6951  * Since the clientid has expired, all compounds using sessions
6952  * associated with the stale clientid will be returning
6953  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6954  * be in some phase of session reset.
6955  *
6956  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6957  */
6958 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6959 {
6960 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6961 	int status;
6962 
6963 	/* try SP4_MACH_CRED if krb5i/p	*/
6964 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
6965 	    authflavor == RPC_AUTH_GSS_KRB5P) {
6966 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6967 		if (!status)
6968 			return 0;
6969 	}
6970 
6971 	/* try SP4_NONE */
6972 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6973 }
6974 
6975 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6976 		struct rpc_cred *cred)
6977 {
6978 	struct rpc_message msg = {
6979 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6980 		.rpc_argp = clp,
6981 		.rpc_cred = cred,
6982 	};
6983 	int status;
6984 
6985 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6986 	trace_nfs4_destroy_clientid(clp, status);
6987 	if (status)
6988 		dprintk("NFS: Got error %d from the server %s on "
6989 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
6990 	return status;
6991 }
6992 
6993 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6994 		struct rpc_cred *cred)
6995 {
6996 	unsigned int loop;
6997 	int ret;
6998 
6999 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7000 		ret = _nfs4_proc_destroy_clientid(clp, cred);
7001 		switch (ret) {
7002 		case -NFS4ERR_DELAY:
7003 		case -NFS4ERR_CLIENTID_BUSY:
7004 			ssleep(1);
7005 			break;
7006 		default:
7007 			return ret;
7008 		}
7009 	}
7010 	return 0;
7011 }
7012 
7013 int nfs4_destroy_clientid(struct nfs_client *clp)
7014 {
7015 	struct rpc_cred *cred;
7016 	int ret = 0;
7017 
7018 	if (clp->cl_mvops->minor_version < 1)
7019 		goto out;
7020 	if (clp->cl_exchange_flags == 0)
7021 		goto out;
7022 	if (clp->cl_preserve_clid)
7023 		goto out;
7024 	cred = nfs4_get_clid_cred(clp);
7025 	ret = nfs4_proc_destroy_clientid(clp, cred);
7026 	if (cred)
7027 		put_rpccred(cred);
7028 	switch (ret) {
7029 	case 0:
7030 	case -NFS4ERR_STALE_CLIENTID:
7031 		clp->cl_exchange_flags = 0;
7032 	}
7033 out:
7034 	return ret;
7035 }
7036 
7037 struct nfs4_get_lease_time_data {
7038 	struct nfs4_get_lease_time_args *args;
7039 	struct nfs4_get_lease_time_res *res;
7040 	struct nfs_client *clp;
7041 };
7042 
7043 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7044 					void *calldata)
7045 {
7046 	struct nfs4_get_lease_time_data *data =
7047 			(struct nfs4_get_lease_time_data *)calldata;
7048 
7049 	dprintk("--> %s\n", __func__);
7050 	/* just setup sequence, do not trigger session recovery
7051 	   since we're invoked within one */
7052 	nfs41_setup_sequence(data->clp->cl_session,
7053 			&data->args->la_seq_args,
7054 			&data->res->lr_seq_res,
7055 			task);
7056 	dprintk("<-- %s\n", __func__);
7057 }
7058 
7059 /*
7060  * Called from nfs4_state_manager thread for session setup, so don't recover
7061  * from sequence operation or clientid errors.
7062  */
7063 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7064 {
7065 	struct nfs4_get_lease_time_data *data =
7066 			(struct nfs4_get_lease_time_data *)calldata;
7067 
7068 	dprintk("--> %s\n", __func__);
7069 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7070 		return;
7071 	switch (task->tk_status) {
7072 	case -NFS4ERR_DELAY:
7073 	case -NFS4ERR_GRACE:
7074 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7075 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
7076 		task->tk_status = 0;
7077 		/* fall through */
7078 	case -NFS4ERR_RETRY_UNCACHED_REP:
7079 		rpc_restart_call_prepare(task);
7080 		return;
7081 	}
7082 	dprintk("<-- %s\n", __func__);
7083 }
7084 
7085 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7086 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
7087 	.rpc_call_done = nfs4_get_lease_time_done,
7088 };
7089 
7090 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7091 {
7092 	struct rpc_task *task;
7093 	struct nfs4_get_lease_time_args args;
7094 	struct nfs4_get_lease_time_res res = {
7095 		.lr_fsinfo = fsinfo,
7096 	};
7097 	struct nfs4_get_lease_time_data data = {
7098 		.args = &args,
7099 		.res = &res,
7100 		.clp = clp,
7101 	};
7102 	struct rpc_message msg = {
7103 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7104 		.rpc_argp = &args,
7105 		.rpc_resp = &res,
7106 	};
7107 	struct rpc_task_setup task_setup = {
7108 		.rpc_client = clp->cl_rpcclient,
7109 		.rpc_message = &msg,
7110 		.callback_ops = &nfs4_get_lease_time_ops,
7111 		.callback_data = &data,
7112 		.flags = RPC_TASK_TIMEOUT,
7113 	};
7114 	int status;
7115 
7116 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7117 	nfs4_set_sequence_privileged(&args.la_seq_args);
7118 	dprintk("--> %s\n", __func__);
7119 	task = rpc_run_task(&task_setup);
7120 
7121 	if (IS_ERR(task))
7122 		status = PTR_ERR(task);
7123 	else {
7124 		status = task->tk_status;
7125 		rpc_put_task(task);
7126 	}
7127 	dprintk("<-- %s return %d\n", __func__, status);
7128 
7129 	return status;
7130 }
7131 
7132 /*
7133  * Initialize the values to be used by the client in CREATE_SESSION
7134  * If nfs4_init_session set the fore channel request and response sizes,
7135  * use them.
7136  *
7137  * Set the back channel max_resp_sz_cached to zero to force the client to
7138  * always set csa_cachethis to FALSE because the current implementation
7139  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7140  */
7141 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7142 {
7143 	unsigned int max_rqst_sz, max_resp_sz;
7144 
7145 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7146 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7147 
7148 	/* Fore channel attributes */
7149 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7150 	args->fc_attrs.max_resp_sz = max_resp_sz;
7151 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7152 	args->fc_attrs.max_reqs = max_session_slots;
7153 
7154 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7155 		"max_ops=%u max_reqs=%u\n",
7156 		__func__,
7157 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7158 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7159 
7160 	/* Back channel attributes */
7161 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7162 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
7163 	args->bc_attrs.max_resp_sz_cached = 0;
7164 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7165 	args->bc_attrs.max_reqs = 1;
7166 
7167 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7168 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7169 		__func__,
7170 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7171 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7172 		args->bc_attrs.max_reqs);
7173 }
7174 
7175 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7176 		struct nfs41_create_session_res *res)
7177 {
7178 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7179 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7180 
7181 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7182 		return -EINVAL;
7183 	/*
7184 	 * Our requested max_ops is the minimum we need; we're not
7185 	 * prepared to break up compounds into smaller pieces than that.
7186 	 * So, no point even trying to continue if the server won't
7187 	 * cooperate:
7188 	 */
7189 	if (rcvd->max_ops < sent->max_ops)
7190 		return -EINVAL;
7191 	if (rcvd->max_reqs == 0)
7192 		return -EINVAL;
7193 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7194 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7195 	return 0;
7196 }
7197 
7198 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7199 		struct nfs41_create_session_res *res)
7200 {
7201 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7202 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7203 
7204 	if (!(res->flags & SESSION4_BACK_CHAN))
7205 		goto out;
7206 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7207 		return -EINVAL;
7208 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7209 		return -EINVAL;
7210 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7211 		return -EINVAL;
7212 	/* These would render the backchannel useless: */
7213 	if (rcvd->max_ops != sent->max_ops)
7214 		return -EINVAL;
7215 	if (rcvd->max_reqs != sent->max_reqs)
7216 		return -EINVAL;
7217 out:
7218 	return 0;
7219 }
7220 
7221 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7222 				     struct nfs41_create_session_res *res)
7223 {
7224 	int ret;
7225 
7226 	ret = nfs4_verify_fore_channel_attrs(args, res);
7227 	if (ret)
7228 		return ret;
7229 	return nfs4_verify_back_channel_attrs(args, res);
7230 }
7231 
7232 static void nfs4_update_session(struct nfs4_session *session,
7233 		struct nfs41_create_session_res *res)
7234 {
7235 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7236 	/* Mark client id and session as being confirmed */
7237 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7238 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7239 	session->flags = res->flags;
7240 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7241 	if (res->flags & SESSION4_BACK_CHAN)
7242 		memcpy(&session->bc_attrs, &res->bc_attrs,
7243 				sizeof(session->bc_attrs));
7244 }
7245 
7246 static int _nfs4_proc_create_session(struct nfs_client *clp,
7247 		struct rpc_cred *cred)
7248 {
7249 	struct nfs4_session *session = clp->cl_session;
7250 	struct nfs41_create_session_args args = {
7251 		.client = clp,
7252 		.clientid = clp->cl_clientid,
7253 		.seqid = clp->cl_seqid,
7254 		.cb_program = NFS4_CALLBACK,
7255 	};
7256 	struct nfs41_create_session_res res;
7257 
7258 	struct rpc_message msg = {
7259 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7260 		.rpc_argp = &args,
7261 		.rpc_resp = &res,
7262 		.rpc_cred = cred,
7263 	};
7264 	int status;
7265 
7266 	nfs4_init_channel_attrs(&args);
7267 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7268 
7269 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7270 	trace_nfs4_create_session(clp, status);
7271 
7272 	if (!status) {
7273 		/* Verify the session's negotiated channel_attrs values */
7274 		status = nfs4_verify_channel_attrs(&args, &res);
7275 		/* Increment the clientid slot sequence id */
7276 		if (clp->cl_seqid == res.seqid)
7277 			clp->cl_seqid++;
7278 		if (status)
7279 			goto out;
7280 		nfs4_update_session(session, &res);
7281 	}
7282 out:
7283 	return status;
7284 }
7285 
7286 /*
7287  * Issues a CREATE_SESSION operation to the server.
7288  * It is the responsibility of the caller to verify the session is
7289  * expired before calling this routine.
7290  */
7291 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7292 {
7293 	int status;
7294 	unsigned *ptr;
7295 	struct nfs4_session *session = clp->cl_session;
7296 
7297 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7298 
7299 	status = _nfs4_proc_create_session(clp, cred);
7300 	if (status)
7301 		goto out;
7302 
7303 	/* Init or reset the session slot tables */
7304 	status = nfs4_setup_session_slot_tables(session);
7305 	dprintk("slot table setup returned %d\n", status);
7306 	if (status)
7307 		goto out;
7308 
7309 	ptr = (unsigned *)&session->sess_id.data[0];
7310 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7311 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7312 out:
7313 	dprintk("<-- %s\n", __func__);
7314 	return status;
7315 }
7316 
7317 /*
7318  * Issue the over-the-wire RPC DESTROY_SESSION.
7319  * The caller must serialize access to this routine.
7320  */
7321 int nfs4_proc_destroy_session(struct nfs4_session *session,
7322 		struct rpc_cred *cred)
7323 {
7324 	struct rpc_message msg = {
7325 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7326 		.rpc_argp = session,
7327 		.rpc_cred = cred,
7328 	};
7329 	int status = 0;
7330 
7331 	dprintk("--> nfs4_proc_destroy_session\n");
7332 
7333 	/* session is still being setup */
7334 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7335 		return 0;
7336 
7337 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7338 	trace_nfs4_destroy_session(session->clp, status);
7339 
7340 	if (status)
7341 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7342 			"Session has been destroyed regardless...\n", status);
7343 
7344 	dprintk("<-- nfs4_proc_destroy_session\n");
7345 	return status;
7346 }
7347 
7348 /*
7349  * Renew the cl_session lease.
7350  */
7351 struct nfs4_sequence_data {
7352 	struct nfs_client *clp;
7353 	struct nfs4_sequence_args args;
7354 	struct nfs4_sequence_res res;
7355 };
7356 
7357 static void nfs41_sequence_release(void *data)
7358 {
7359 	struct nfs4_sequence_data *calldata = data;
7360 	struct nfs_client *clp = calldata->clp;
7361 
7362 	if (atomic_read(&clp->cl_count) > 1)
7363 		nfs4_schedule_state_renewal(clp);
7364 	nfs_put_client(clp);
7365 	kfree(calldata);
7366 }
7367 
7368 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7369 {
7370 	switch(task->tk_status) {
7371 	case -NFS4ERR_DELAY:
7372 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7373 		return -EAGAIN;
7374 	default:
7375 		nfs4_schedule_lease_recovery(clp);
7376 	}
7377 	return 0;
7378 }
7379 
7380 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7381 {
7382 	struct nfs4_sequence_data *calldata = data;
7383 	struct nfs_client *clp = calldata->clp;
7384 
7385 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7386 		return;
7387 
7388 	trace_nfs4_sequence(clp, task->tk_status);
7389 	if (task->tk_status < 0) {
7390 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7391 		if (atomic_read(&clp->cl_count) == 1)
7392 			goto out;
7393 
7394 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7395 			rpc_restart_call_prepare(task);
7396 			return;
7397 		}
7398 	}
7399 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7400 out:
7401 	dprintk("<-- %s\n", __func__);
7402 }
7403 
7404 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7405 {
7406 	struct nfs4_sequence_data *calldata = data;
7407 	struct nfs_client *clp = calldata->clp;
7408 	struct nfs4_sequence_args *args;
7409 	struct nfs4_sequence_res *res;
7410 
7411 	args = task->tk_msg.rpc_argp;
7412 	res = task->tk_msg.rpc_resp;
7413 
7414 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7415 }
7416 
7417 static const struct rpc_call_ops nfs41_sequence_ops = {
7418 	.rpc_call_done = nfs41_sequence_call_done,
7419 	.rpc_call_prepare = nfs41_sequence_prepare,
7420 	.rpc_release = nfs41_sequence_release,
7421 };
7422 
7423 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7424 		struct rpc_cred *cred,
7425 		bool is_privileged)
7426 {
7427 	struct nfs4_sequence_data *calldata;
7428 	struct rpc_message msg = {
7429 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7430 		.rpc_cred = cred,
7431 	};
7432 	struct rpc_task_setup task_setup_data = {
7433 		.rpc_client = clp->cl_rpcclient,
7434 		.rpc_message = &msg,
7435 		.callback_ops = &nfs41_sequence_ops,
7436 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7437 	};
7438 
7439 	if (!atomic_inc_not_zero(&clp->cl_count))
7440 		return ERR_PTR(-EIO);
7441 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7442 	if (calldata == NULL) {
7443 		nfs_put_client(clp);
7444 		return ERR_PTR(-ENOMEM);
7445 	}
7446 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7447 	if (is_privileged)
7448 		nfs4_set_sequence_privileged(&calldata->args);
7449 	msg.rpc_argp = &calldata->args;
7450 	msg.rpc_resp = &calldata->res;
7451 	calldata->clp = clp;
7452 	task_setup_data.callback_data = calldata;
7453 
7454 	return rpc_run_task(&task_setup_data);
7455 }
7456 
7457 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7458 {
7459 	struct rpc_task *task;
7460 	int ret = 0;
7461 
7462 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7463 		return -EAGAIN;
7464 	task = _nfs41_proc_sequence(clp, cred, false);
7465 	if (IS_ERR(task))
7466 		ret = PTR_ERR(task);
7467 	else
7468 		rpc_put_task_async(task);
7469 	dprintk("<-- %s status=%d\n", __func__, ret);
7470 	return ret;
7471 }
7472 
7473 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7474 {
7475 	struct rpc_task *task;
7476 	int ret;
7477 
7478 	task = _nfs41_proc_sequence(clp, cred, true);
7479 	if (IS_ERR(task)) {
7480 		ret = PTR_ERR(task);
7481 		goto out;
7482 	}
7483 	ret = rpc_wait_for_completion_task(task);
7484 	if (!ret) {
7485 		struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7486 
7487 		if (task->tk_status == 0)
7488 			nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7489 		ret = task->tk_status;
7490 	}
7491 	rpc_put_task(task);
7492 out:
7493 	dprintk("<-- %s status=%d\n", __func__, ret);
7494 	return ret;
7495 }
7496 
7497 struct nfs4_reclaim_complete_data {
7498 	struct nfs_client *clp;
7499 	struct nfs41_reclaim_complete_args arg;
7500 	struct nfs41_reclaim_complete_res res;
7501 };
7502 
7503 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7504 {
7505 	struct nfs4_reclaim_complete_data *calldata = data;
7506 
7507 	nfs41_setup_sequence(calldata->clp->cl_session,
7508 			&calldata->arg.seq_args,
7509 			&calldata->res.seq_res,
7510 			task);
7511 }
7512 
7513 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7514 {
7515 	switch(task->tk_status) {
7516 	case 0:
7517 	case -NFS4ERR_COMPLETE_ALREADY:
7518 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7519 		break;
7520 	case -NFS4ERR_DELAY:
7521 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7522 		/* fall through */
7523 	case -NFS4ERR_RETRY_UNCACHED_REP:
7524 		return -EAGAIN;
7525 	default:
7526 		nfs4_schedule_lease_recovery(clp);
7527 	}
7528 	return 0;
7529 }
7530 
7531 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7532 {
7533 	struct nfs4_reclaim_complete_data *calldata = data;
7534 	struct nfs_client *clp = calldata->clp;
7535 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7536 
7537 	dprintk("--> %s\n", __func__);
7538 	if (!nfs41_sequence_done(task, res))
7539 		return;
7540 
7541 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7542 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7543 		rpc_restart_call_prepare(task);
7544 		return;
7545 	}
7546 	dprintk("<-- %s\n", __func__);
7547 }
7548 
7549 static void nfs4_free_reclaim_complete_data(void *data)
7550 {
7551 	struct nfs4_reclaim_complete_data *calldata = data;
7552 
7553 	kfree(calldata);
7554 }
7555 
7556 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7557 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7558 	.rpc_call_done = nfs4_reclaim_complete_done,
7559 	.rpc_release = nfs4_free_reclaim_complete_data,
7560 };
7561 
7562 /*
7563  * Issue a global reclaim complete.
7564  */
7565 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7566 		struct rpc_cred *cred)
7567 {
7568 	struct nfs4_reclaim_complete_data *calldata;
7569 	struct rpc_task *task;
7570 	struct rpc_message msg = {
7571 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7572 		.rpc_cred = cred,
7573 	};
7574 	struct rpc_task_setup task_setup_data = {
7575 		.rpc_client = clp->cl_rpcclient,
7576 		.rpc_message = &msg,
7577 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7578 		.flags = RPC_TASK_ASYNC,
7579 	};
7580 	int status = -ENOMEM;
7581 
7582 	dprintk("--> %s\n", __func__);
7583 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7584 	if (calldata == NULL)
7585 		goto out;
7586 	calldata->clp = clp;
7587 	calldata->arg.one_fs = 0;
7588 
7589 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7590 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7591 	msg.rpc_argp = &calldata->arg;
7592 	msg.rpc_resp = &calldata->res;
7593 	task_setup_data.callback_data = calldata;
7594 	task = rpc_run_task(&task_setup_data);
7595 	if (IS_ERR(task)) {
7596 		status = PTR_ERR(task);
7597 		goto out;
7598 	}
7599 	status = nfs4_wait_for_completion_rpc_task(task);
7600 	if (status == 0)
7601 		status = task->tk_status;
7602 	rpc_put_task(task);
7603 	return 0;
7604 out:
7605 	dprintk("<-- %s status=%d\n", __func__, status);
7606 	return status;
7607 }
7608 
7609 static void
7610 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7611 {
7612 	struct nfs4_layoutget *lgp = calldata;
7613 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7614 	struct nfs4_session *session = nfs4_get_session(server);
7615 
7616 	dprintk("--> %s\n", __func__);
7617 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
7618 	 * right now covering the LAYOUTGET we are about to send.
7619 	 * However, that is not so catastrophic, and there seems
7620 	 * to be no way to prevent it completely.
7621 	 */
7622 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7623 				&lgp->res.seq_res, task))
7624 		return;
7625 	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7626 					  NFS_I(lgp->args.inode)->layout,
7627 					  &lgp->args.range,
7628 					  lgp->args.ctx->state)) {
7629 		rpc_exit(task, NFS4_OK);
7630 	}
7631 }
7632 
7633 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7634 {
7635 	struct nfs4_layoutget *lgp = calldata;
7636 	struct inode *inode = lgp->args.inode;
7637 	struct nfs_server *server = NFS_SERVER(inode);
7638 	struct pnfs_layout_hdr *lo;
7639 	struct nfs4_state *state = NULL;
7640 	unsigned long timeo, now, giveup;
7641 
7642 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7643 
7644 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7645 		goto out;
7646 
7647 	switch (task->tk_status) {
7648 	case 0:
7649 		goto out;
7650 	/*
7651 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7652 	 * (or clients) writing to the same RAID stripe
7653 	 */
7654 	case -NFS4ERR_LAYOUTTRYLATER:
7655 	/*
7656 	 * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7657 	 * existing layout before getting a new one).
7658 	 */
7659 	case -NFS4ERR_RECALLCONFLICT:
7660 		timeo = rpc_get_timeout(task->tk_client);
7661 		giveup = lgp->args.timestamp + timeo;
7662 		now = jiffies;
7663 		if (time_after(giveup, now)) {
7664 			unsigned long delay;
7665 
7666 			/* Delay for:
7667 			 * - Not less then NFS4_POLL_RETRY_MIN.
7668 			 * - One last time a jiffie before we give up
7669 			 * - exponential backoff (time_now minus start_attempt)
7670 			 */
7671 			delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7672 				    min((giveup - now - 1),
7673 					now - lgp->args.timestamp));
7674 
7675 			dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7676 				__func__, delay);
7677 			rpc_delay(task, delay);
7678 			task->tk_status = 0;
7679 			rpc_restart_call_prepare(task);
7680 			goto out; /* Do not call nfs4_async_handle_error() */
7681 		}
7682 		break;
7683 	case -NFS4ERR_EXPIRED:
7684 	case -NFS4ERR_BAD_STATEID:
7685 		spin_lock(&inode->i_lock);
7686 		lo = NFS_I(inode)->layout;
7687 		if (!lo || list_empty(&lo->plh_segs)) {
7688 			spin_unlock(&inode->i_lock);
7689 			/* If the open stateid was bad, then recover it. */
7690 			state = lgp->args.ctx->state;
7691 		} else {
7692 			LIST_HEAD(head);
7693 
7694 			/*
7695 			 * Mark the bad layout state as invalid, then retry
7696 			 * with the current stateid.
7697 			 */
7698 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7699 			spin_unlock(&inode->i_lock);
7700 			pnfs_free_lseg_list(&head);
7701 
7702 			task->tk_status = 0;
7703 			rpc_restart_call_prepare(task);
7704 		}
7705 	}
7706 	if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7707 		rpc_restart_call_prepare(task);
7708 out:
7709 	dprintk("<-- %s\n", __func__);
7710 }
7711 
7712 static size_t max_response_pages(struct nfs_server *server)
7713 {
7714 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7715 	return nfs_page_array_len(0, max_resp_sz);
7716 }
7717 
7718 static void nfs4_free_pages(struct page **pages, size_t size)
7719 {
7720 	int i;
7721 
7722 	if (!pages)
7723 		return;
7724 
7725 	for (i = 0; i < size; i++) {
7726 		if (!pages[i])
7727 			break;
7728 		__free_page(pages[i]);
7729 	}
7730 	kfree(pages);
7731 }
7732 
7733 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7734 {
7735 	struct page **pages;
7736 	int i;
7737 
7738 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7739 	if (!pages) {
7740 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7741 		return NULL;
7742 	}
7743 
7744 	for (i = 0; i < size; i++) {
7745 		pages[i] = alloc_page(gfp_flags);
7746 		if (!pages[i]) {
7747 			dprintk("%s: failed to allocate page\n", __func__);
7748 			nfs4_free_pages(pages, size);
7749 			return NULL;
7750 		}
7751 	}
7752 
7753 	return pages;
7754 }
7755 
7756 static void nfs4_layoutget_release(void *calldata)
7757 {
7758 	struct nfs4_layoutget *lgp = calldata;
7759 	struct inode *inode = lgp->args.inode;
7760 	struct nfs_server *server = NFS_SERVER(inode);
7761 	size_t max_pages = max_response_pages(server);
7762 
7763 	dprintk("--> %s\n", __func__);
7764 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
7765 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
7766 	put_nfs_open_context(lgp->args.ctx);
7767 	kfree(calldata);
7768 	dprintk("<-- %s\n", __func__);
7769 }
7770 
7771 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7772 	.rpc_call_prepare = nfs4_layoutget_prepare,
7773 	.rpc_call_done = nfs4_layoutget_done,
7774 	.rpc_release = nfs4_layoutget_release,
7775 };
7776 
7777 struct pnfs_layout_segment *
7778 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7779 {
7780 	struct inode *inode = lgp->args.inode;
7781 	struct nfs_server *server = NFS_SERVER(inode);
7782 	size_t max_pages = max_response_pages(server);
7783 	struct rpc_task *task;
7784 	struct rpc_message msg = {
7785 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7786 		.rpc_argp = &lgp->args,
7787 		.rpc_resp = &lgp->res,
7788 		.rpc_cred = lgp->cred,
7789 	};
7790 	struct rpc_task_setup task_setup_data = {
7791 		.rpc_client = server->client,
7792 		.rpc_message = &msg,
7793 		.callback_ops = &nfs4_layoutget_call_ops,
7794 		.callback_data = lgp,
7795 		.flags = RPC_TASK_ASYNC,
7796 	};
7797 	struct pnfs_layout_segment *lseg = NULL;
7798 	int status = 0;
7799 
7800 	dprintk("--> %s\n", __func__);
7801 
7802 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7803 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
7804 
7805 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7806 	if (!lgp->args.layout.pages) {
7807 		nfs4_layoutget_release(lgp);
7808 		return ERR_PTR(-ENOMEM);
7809 	}
7810 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7811 	lgp->args.timestamp = jiffies;
7812 
7813 	lgp->res.layoutp = &lgp->args.layout;
7814 	lgp->res.seq_res.sr_slot = NULL;
7815 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7816 
7817 	task = rpc_run_task(&task_setup_data);
7818 	if (IS_ERR(task))
7819 		return ERR_CAST(task);
7820 	status = nfs4_wait_for_completion_rpc_task(task);
7821 	if (status == 0)
7822 		status = task->tk_status;
7823 	trace_nfs4_layoutget(lgp->args.ctx,
7824 			&lgp->args.range,
7825 			&lgp->res.range,
7826 			status);
7827 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7828 	if (status == 0 && lgp->res.layoutp->len)
7829 		lseg = pnfs_layout_process(lgp);
7830 	rpc_put_task(task);
7831 	dprintk("<-- %s status=%d\n", __func__, status);
7832 	if (status)
7833 		return ERR_PTR(status);
7834 	return lseg;
7835 }
7836 
7837 static void
7838 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7839 {
7840 	struct nfs4_layoutreturn *lrp = calldata;
7841 
7842 	dprintk("--> %s\n", __func__);
7843 	nfs41_setup_sequence(lrp->clp->cl_session,
7844 			&lrp->args.seq_args,
7845 			&lrp->res.seq_res,
7846 			task);
7847 }
7848 
7849 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7850 {
7851 	struct nfs4_layoutreturn *lrp = calldata;
7852 	struct nfs_server *server;
7853 
7854 	dprintk("--> %s\n", __func__);
7855 
7856 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7857 		return;
7858 
7859 	server = NFS_SERVER(lrp->args.inode);
7860 	switch (task->tk_status) {
7861 	default:
7862 		task->tk_status = 0;
7863 	case 0:
7864 		break;
7865 	case -NFS4ERR_DELAY:
7866 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7867 			break;
7868 		rpc_restart_call_prepare(task);
7869 		return;
7870 	}
7871 	dprintk("<-- %s\n", __func__);
7872 }
7873 
7874 static void nfs4_layoutreturn_release(void *calldata)
7875 {
7876 	struct nfs4_layoutreturn *lrp = calldata;
7877 	struct pnfs_layout_hdr *lo = lrp->args.layout;
7878 
7879 	dprintk("--> %s\n", __func__);
7880 	spin_lock(&lo->plh_inode->i_lock);
7881 	if (lrp->res.lrs_present)
7882 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7883 	pnfs_clear_layoutreturn_waitbit(lo);
7884 	clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, &lo->plh_flags);
7885 	rpc_wake_up(&NFS_SERVER(lo->plh_inode)->roc_rpcwaitq);
7886 	lo->plh_block_lgets--;
7887 	spin_unlock(&lo->plh_inode->i_lock);
7888 	pnfs_put_layout_hdr(lrp->args.layout);
7889 	nfs_iput_and_deactive(lrp->inode);
7890 	kfree(calldata);
7891 	dprintk("<-- %s\n", __func__);
7892 }
7893 
7894 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7895 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
7896 	.rpc_call_done = nfs4_layoutreturn_done,
7897 	.rpc_release = nfs4_layoutreturn_release,
7898 };
7899 
7900 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
7901 {
7902 	struct rpc_task *task;
7903 	struct rpc_message msg = {
7904 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7905 		.rpc_argp = &lrp->args,
7906 		.rpc_resp = &lrp->res,
7907 		.rpc_cred = lrp->cred,
7908 	};
7909 	struct rpc_task_setup task_setup_data = {
7910 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
7911 		.rpc_message = &msg,
7912 		.callback_ops = &nfs4_layoutreturn_call_ops,
7913 		.callback_data = lrp,
7914 	};
7915 	int status = 0;
7916 
7917 	dprintk("--> %s\n", __func__);
7918 	if (!sync) {
7919 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
7920 		if (!lrp->inode) {
7921 			nfs4_layoutreturn_release(lrp);
7922 			return -EAGAIN;
7923 		}
7924 		task_setup_data.flags |= RPC_TASK_ASYNC;
7925 	}
7926 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7927 	task = rpc_run_task(&task_setup_data);
7928 	if (IS_ERR(task))
7929 		return PTR_ERR(task);
7930 	if (sync)
7931 		status = task->tk_status;
7932 	trace_nfs4_layoutreturn(lrp->args.inode, status);
7933 	dprintk("<-- %s status=%d\n", __func__, status);
7934 	rpc_put_task(task);
7935 	return status;
7936 }
7937 
7938 static int
7939 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7940 		struct pnfs_device *pdev,
7941 		struct rpc_cred *cred)
7942 {
7943 	struct nfs4_getdeviceinfo_args args = {
7944 		.pdev = pdev,
7945 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
7946 			NOTIFY_DEVICEID4_DELETE,
7947 	};
7948 	struct nfs4_getdeviceinfo_res res = {
7949 		.pdev = pdev,
7950 	};
7951 	struct rpc_message msg = {
7952 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7953 		.rpc_argp = &args,
7954 		.rpc_resp = &res,
7955 		.rpc_cred = cred,
7956 	};
7957 	int status;
7958 
7959 	dprintk("--> %s\n", __func__);
7960 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7961 	if (res.notification & ~args.notify_types)
7962 		dprintk("%s: unsupported notification\n", __func__);
7963 	if (res.notification != args.notify_types)
7964 		pdev->nocache = 1;
7965 
7966 	dprintk("<-- %s status=%d\n", __func__, status);
7967 
7968 	return status;
7969 }
7970 
7971 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7972 		struct pnfs_device *pdev,
7973 		struct rpc_cred *cred)
7974 {
7975 	struct nfs4_exception exception = { };
7976 	int err;
7977 
7978 	do {
7979 		err = nfs4_handle_exception(server,
7980 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
7981 					&exception);
7982 	} while (exception.retry);
7983 	return err;
7984 }
7985 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7986 
7987 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7988 {
7989 	struct nfs4_layoutcommit_data *data = calldata;
7990 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7991 	struct nfs4_session *session = nfs4_get_session(server);
7992 
7993 	nfs41_setup_sequence(session,
7994 			&data->args.seq_args,
7995 			&data->res.seq_res,
7996 			task);
7997 }
7998 
7999 static void
8000 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8001 {
8002 	struct nfs4_layoutcommit_data *data = calldata;
8003 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8004 
8005 	if (!nfs41_sequence_done(task, &data->res.seq_res))
8006 		return;
8007 
8008 	switch (task->tk_status) { /* Just ignore these failures */
8009 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8010 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8011 	case -NFS4ERR_BADLAYOUT:     /* no layout */
8012 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
8013 		task->tk_status = 0;
8014 	case 0:
8015 		break;
8016 	default:
8017 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8018 			rpc_restart_call_prepare(task);
8019 			return;
8020 		}
8021 	}
8022 }
8023 
8024 static void nfs4_layoutcommit_release(void *calldata)
8025 {
8026 	struct nfs4_layoutcommit_data *data = calldata;
8027 
8028 	pnfs_cleanup_layoutcommit(data);
8029 	nfs_post_op_update_inode_force_wcc(data->args.inode,
8030 					   data->res.fattr);
8031 	put_rpccred(data->cred);
8032 	nfs_iput_and_deactive(data->inode);
8033 	kfree(data);
8034 }
8035 
8036 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8037 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
8038 	.rpc_call_done = nfs4_layoutcommit_done,
8039 	.rpc_release = nfs4_layoutcommit_release,
8040 };
8041 
8042 int
8043 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8044 {
8045 	struct rpc_message msg = {
8046 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8047 		.rpc_argp = &data->args,
8048 		.rpc_resp = &data->res,
8049 		.rpc_cred = data->cred,
8050 	};
8051 	struct rpc_task_setup task_setup_data = {
8052 		.task = &data->task,
8053 		.rpc_client = NFS_CLIENT(data->args.inode),
8054 		.rpc_message = &msg,
8055 		.callback_ops = &nfs4_layoutcommit_ops,
8056 		.callback_data = data,
8057 	};
8058 	struct rpc_task *task;
8059 	int status = 0;
8060 
8061 	dprintk("NFS: %4d initiating layoutcommit call. sync %d "
8062 		"lbw: %llu inode %lu\n",
8063 		data->task.tk_pid, sync,
8064 		data->args.lastbytewritten,
8065 		data->args.inode->i_ino);
8066 
8067 	if (!sync) {
8068 		data->inode = nfs_igrab_and_active(data->args.inode);
8069 		if (data->inode == NULL) {
8070 			nfs4_layoutcommit_release(data);
8071 			return -EAGAIN;
8072 		}
8073 		task_setup_data.flags = RPC_TASK_ASYNC;
8074 	}
8075 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8076 	task = rpc_run_task(&task_setup_data);
8077 	if (IS_ERR(task))
8078 		return PTR_ERR(task);
8079 	if (sync)
8080 		status = task->tk_status;
8081 	trace_nfs4_layoutcommit(data->args.inode, status);
8082 	dprintk("%s: status %d\n", __func__, status);
8083 	rpc_put_task(task);
8084 	return status;
8085 }
8086 
8087 /**
8088  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8089  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8090  */
8091 static int
8092 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8093 		    struct nfs_fsinfo *info,
8094 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8095 {
8096 	struct nfs41_secinfo_no_name_args args = {
8097 		.style = SECINFO_STYLE_CURRENT_FH,
8098 	};
8099 	struct nfs4_secinfo_res res = {
8100 		.flavors = flavors,
8101 	};
8102 	struct rpc_message msg = {
8103 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8104 		.rpc_argp = &args,
8105 		.rpc_resp = &res,
8106 	};
8107 	struct rpc_clnt *clnt = server->client;
8108 	struct rpc_cred *cred = NULL;
8109 	int status;
8110 
8111 	if (use_integrity) {
8112 		clnt = server->nfs_client->cl_rpcclient;
8113 		cred = nfs4_get_clid_cred(server->nfs_client);
8114 		msg.rpc_cred = cred;
8115 	}
8116 
8117 	dprintk("--> %s\n", __func__);
8118 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8119 				&res.seq_res, 0);
8120 	dprintk("<-- %s status=%d\n", __func__, status);
8121 
8122 	if (cred)
8123 		put_rpccred(cred);
8124 
8125 	return status;
8126 }
8127 
8128 static int
8129 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8130 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8131 {
8132 	struct nfs4_exception exception = { };
8133 	int err;
8134 	do {
8135 		/* first try using integrity protection */
8136 		err = -NFS4ERR_WRONGSEC;
8137 
8138 		/* try to use integrity protection with machine cred */
8139 		if (_nfs4_is_integrity_protected(server->nfs_client))
8140 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8141 							  flavors, true);
8142 
8143 		/*
8144 		 * if unable to use integrity protection, or SECINFO with
8145 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8146 		 * disallowed by spec, but exists in deployed servers) use
8147 		 * the current filesystem's rpc_client and the user cred.
8148 		 */
8149 		if (err == -NFS4ERR_WRONGSEC)
8150 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8151 							  flavors, false);
8152 
8153 		switch (err) {
8154 		case 0:
8155 		case -NFS4ERR_WRONGSEC:
8156 		case -ENOTSUPP:
8157 			goto out;
8158 		default:
8159 			err = nfs4_handle_exception(server, err, &exception);
8160 		}
8161 	} while (exception.retry);
8162 out:
8163 	return err;
8164 }
8165 
8166 static int
8167 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8168 		    struct nfs_fsinfo *info)
8169 {
8170 	int err;
8171 	struct page *page;
8172 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8173 	struct nfs4_secinfo_flavors *flavors;
8174 	struct nfs4_secinfo4 *secinfo;
8175 	int i;
8176 
8177 	page = alloc_page(GFP_KERNEL);
8178 	if (!page) {
8179 		err = -ENOMEM;
8180 		goto out;
8181 	}
8182 
8183 	flavors = page_address(page);
8184 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8185 
8186 	/*
8187 	 * Fall back on "guess and check" method if
8188 	 * the server doesn't support SECINFO_NO_NAME
8189 	 */
8190 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8191 		err = nfs4_find_root_sec(server, fhandle, info);
8192 		goto out_freepage;
8193 	}
8194 	if (err)
8195 		goto out_freepage;
8196 
8197 	for (i = 0; i < flavors->num_flavors; i++) {
8198 		secinfo = &flavors->flavors[i];
8199 
8200 		switch (secinfo->flavor) {
8201 		case RPC_AUTH_NULL:
8202 		case RPC_AUTH_UNIX:
8203 		case RPC_AUTH_GSS:
8204 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8205 					&secinfo->flavor_info);
8206 			break;
8207 		default:
8208 			flavor = RPC_AUTH_MAXFLAVOR;
8209 			break;
8210 		}
8211 
8212 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8213 			flavor = RPC_AUTH_MAXFLAVOR;
8214 
8215 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8216 			err = nfs4_lookup_root_sec(server, fhandle,
8217 						   info, flavor);
8218 			if (!err)
8219 				break;
8220 		}
8221 	}
8222 
8223 	if (flavor == RPC_AUTH_MAXFLAVOR)
8224 		err = -EPERM;
8225 
8226 out_freepage:
8227 	put_page(page);
8228 	if (err == -EACCES)
8229 		return -EPERM;
8230 out:
8231 	return err;
8232 }
8233 
8234 static int _nfs41_test_stateid(struct nfs_server *server,
8235 		nfs4_stateid *stateid,
8236 		struct rpc_cred *cred)
8237 {
8238 	int status;
8239 	struct nfs41_test_stateid_args args = {
8240 		.stateid = stateid,
8241 	};
8242 	struct nfs41_test_stateid_res res;
8243 	struct rpc_message msg = {
8244 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8245 		.rpc_argp = &args,
8246 		.rpc_resp = &res,
8247 		.rpc_cred = cred,
8248 	};
8249 	struct rpc_clnt *rpc_client = server->client;
8250 
8251 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8252 		&rpc_client, &msg);
8253 
8254 	dprintk("NFS call  test_stateid %p\n", stateid);
8255 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8256 	nfs4_set_sequence_privileged(&args.seq_args);
8257 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8258 			&args.seq_args, &res.seq_res);
8259 	if (status != NFS_OK) {
8260 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8261 		return status;
8262 	}
8263 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8264 	return -res.status;
8265 }
8266 
8267 /**
8268  * nfs41_test_stateid - perform a TEST_STATEID operation
8269  *
8270  * @server: server / transport on which to perform the operation
8271  * @stateid: state ID to test
8272  * @cred: credential
8273  *
8274  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8275  * Otherwise a negative NFS4ERR value is returned if the operation
8276  * failed or the state ID is not currently valid.
8277  */
8278 static int nfs41_test_stateid(struct nfs_server *server,
8279 		nfs4_stateid *stateid,
8280 		struct rpc_cred *cred)
8281 {
8282 	struct nfs4_exception exception = { };
8283 	int err;
8284 	do {
8285 		err = _nfs41_test_stateid(server, stateid, cred);
8286 		if (err != -NFS4ERR_DELAY)
8287 			break;
8288 		nfs4_handle_exception(server, err, &exception);
8289 	} while (exception.retry);
8290 	return err;
8291 }
8292 
8293 struct nfs_free_stateid_data {
8294 	struct nfs_server *server;
8295 	struct nfs41_free_stateid_args args;
8296 	struct nfs41_free_stateid_res res;
8297 };
8298 
8299 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8300 {
8301 	struct nfs_free_stateid_data *data = calldata;
8302 	nfs41_setup_sequence(nfs4_get_session(data->server),
8303 			&data->args.seq_args,
8304 			&data->res.seq_res,
8305 			task);
8306 }
8307 
8308 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8309 {
8310 	struct nfs_free_stateid_data *data = calldata;
8311 
8312 	nfs41_sequence_done(task, &data->res.seq_res);
8313 
8314 	switch (task->tk_status) {
8315 	case -NFS4ERR_DELAY:
8316 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8317 			rpc_restart_call_prepare(task);
8318 	}
8319 }
8320 
8321 static void nfs41_free_stateid_release(void *calldata)
8322 {
8323 	kfree(calldata);
8324 }
8325 
8326 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8327 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8328 	.rpc_call_done = nfs41_free_stateid_done,
8329 	.rpc_release = nfs41_free_stateid_release,
8330 };
8331 
8332 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8333 		nfs4_stateid *stateid,
8334 		struct rpc_cred *cred,
8335 		bool privileged)
8336 {
8337 	struct rpc_message msg = {
8338 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8339 		.rpc_cred = cred,
8340 	};
8341 	struct rpc_task_setup task_setup = {
8342 		.rpc_client = server->client,
8343 		.rpc_message = &msg,
8344 		.callback_ops = &nfs41_free_stateid_ops,
8345 		.flags = RPC_TASK_ASYNC,
8346 	};
8347 	struct nfs_free_stateid_data *data;
8348 
8349 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8350 		&task_setup.rpc_client, &msg);
8351 
8352 	dprintk("NFS call  free_stateid %p\n", stateid);
8353 	data = kmalloc(sizeof(*data), GFP_NOFS);
8354 	if (!data)
8355 		return ERR_PTR(-ENOMEM);
8356 	data->server = server;
8357 	nfs4_stateid_copy(&data->args.stateid, stateid);
8358 
8359 	task_setup.callback_data = data;
8360 
8361 	msg.rpc_argp = &data->args;
8362 	msg.rpc_resp = &data->res;
8363 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8364 	if (privileged)
8365 		nfs4_set_sequence_privileged(&data->args.seq_args);
8366 
8367 	return rpc_run_task(&task_setup);
8368 }
8369 
8370 /**
8371  * nfs41_free_stateid - perform a FREE_STATEID operation
8372  *
8373  * @server: server / transport on which to perform the operation
8374  * @stateid: state ID to release
8375  * @cred: credential
8376  *
8377  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8378  * negative NFS4ERR value is returned.
8379  */
8380 static int nfs41_free_stateid(struct nfs_server *server,
8381 		nfs4_stateid *stateid,
8382 		struct rpc_cred *cred)
8383 {
8384 	struct rpc_task *task;
8385 	int ret;
8386 
8387 	task = _nfs41_free_stateid(server, stateid, cred, true);
8388 	if (IS_ERR(task))
8389 		return PTR_ERR(task);
8390 	ret = rpc_wait_for_completion_task(task);
8391 	if (!ret)
8392 		ret = task->tk_status;
8393 	rpc_put_task(task);
8394 	return ret;
8395 }
8396 
8397 static void
8398 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8399 {
8400 	struct rpc_task *task;
8401 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8402 
8403 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8404 	nfs4_free_lock_state(server, lsp);
8405 	if (IS_ERR(task))
8406 		return;
8407 	rpc_put_task(task);
8408 }
8409 
8410 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8411 		const nfs4_stateid *s2)
8412 {
8413 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8414 		return false;
8415 
8416 	if (s1->seqid == s2->seqid)
8417 		return true;
8418 	if (s1->seqid == 0 || s2->seqid == 0)
8419 		return true;
8420 
8421 	return false;
8422 }
8423 
8424 #endif /* CONFIG_NFS_V4_1 */
8425 
8426 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8427 		const nfs4_stateid *s2)
8428 {
8429 	return nfs4_stateid_match(s1, s2);
8430 }
8431 
8432 
8433 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8434 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8435 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8436 	.recover_open	= nfs4_open_reclaim,
8437 	.recover_lock	= nfs4_lock_reclaim,
8438 	.establish_clid = nfs4_init_clientid,
8439 	.detect_trunking = nfs40_discover_server_trunking,
8440 };
8441 
8442 #if defined(CONFIG_NFS_V4_1)
8443 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8444 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8445 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8446 	.recover_open	= nfs4_open_reclaim,
8447 	.recover_lock	= nfs4_lock_reclaim,
8448 	.establish_clid = nfs41_init_clientid,
8449 	.reclaim_complete = nfs41_proc_reclaim_complete,
8450 	.detect_trunking = nfs41_discover_server_trunking,
8451 };
8452 #endif /* CONFIG_NFS_V4_1 */
8453 
8454 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8455 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8456 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8457 	.recover_open	= nfs40_open_expired,
8458 	.recover_lock	= nfs4_lock_expired,
8459 	.establish_clid = nfs4_init_clientid,
8460 };
8461 
8462 #if defined(CONFIG_NFS_V4_1)
8463 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8464 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8465 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8466 	.recover_open	= nfs41_open_expired,
8467 	.recover_lock	= nfs41_lock_expired,
8468 	.establish_clid = nfs41_init_clientid,
8469 };
8470 #endif /* CONFIG_NFS_V4_1 */
8471 
8472 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8473 	.sched_state_renewal = nfs4_proc_async_renew,
8474 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8475 	.renew_lease = nfs4_proc_renew,
8476 };
8477 
8478 #if defined(CONFIG_NFS_V4_1)
8479 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8480 	.sched_state_renewal = nfs41_proc_async_sequence,
8481 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8482 	.renew_lease = nfs4_proc_sequence,
8483 };
8484 #endif
8485 
8486 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8487 	.get_locations = _nfs40_proc_get_locations,
8488 	.fsid_present = _nfs40_proc_fsid_present,
8489 };
8490 
8491 #if defined(CONFIG_NFS_V4_1)
8492 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8493 	.get_locations = _nfs41_proc_get_locations,
8494 	.fsid_present = _nfs41_proc_fsid_present,
8495 };
8496 #endif	/* CONFIG_NFS_V4_1 */
8497 
8498 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8499 	.minor_version = 0,
8500 	.init_caps = NFS_CAP_READDIRPLUS
8501 		| NFS_CAP_ATOMIC_OPEN
8502 		| NFS_CAP_CHANGE_ATTR
8503 		| NFS_CAP_POSIX_LOCK,
8504 	.init_client = nfs40_init_client,
8505 	.shutdown_client = nfs40_shutdown_client,
8506 	.match_stateid = nfs4_match_stateid,
8507 	.find_root_sec = nfs4_find_root_sec,
8508 	.free_lock_state = nfs4_release_lockowner,
8509 	.alloc_seqid = nfs_alloc_seqid,
8510 	.call_sync_ops = &nfs40_call_sync_ops,
8511 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8512 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8513 	.state_renewal_ops = &nfs40_state_renewal_ops,
8514 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8515 };
8516 
8517 #if defined(CONFIG_NFS_V4_1)
8518 static struct nfs_seqid *
8519 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8520 {
8521 	return NULL;
8522 }
8523 
8524 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8525 	.minor_version = 1,
8526 	.init_caps = NFS_CAP_READDIRPLUS
8527 		| NFS_CAP_ATOMIC_OPEN
8528 		| NFS_CAP_CHANGE_ATTR
8529 		| NFS_CAP_POSIX_LOCK
8530 		| NFS_CAP_STATEID_NFSV41
8531 		| NFS_CAP_ATOMIC_OPEN_V1,
8532 	.init_client = nfs41_init_client,
8533 	.shutdown_client = nfs41_shutdown_client,
8534 	.match_stateid = nfs41_match_stateid,
8535 	.find_root_sec = nfs41_find_root_sec,
8536 	.free_lock_state = nfs41_free_lock_state,
8537 	.alloc_seqid = nfs_alloc_no_seqid,
8538 	.call_sync_ops = &nfs41_call_sync_ops,
8539 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8540 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8541 	.state_renewal_ops = &nfs41_state_renewal_ops,
8542 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8543 };
8544 #endif
8545 
8546 #if defined(CONFIG_NFS_V4_2)
8547 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8548 	.minor_version = 2,
8549 	.init_caps = NFS_CAP_READDIRPLUS
8550 		| NFS_CAP_ATOMIC_OPEN
8551 		| NFS_CAP_CHANGE_ATTR
8552 		| NFS_CAP_POSIX_LOCK
8553 		| NFS_CAP_STATEID_NFSV41
8554 		| NFS_CAP_ATOMIC_OPEN_V1
8555 		| NFS_CAP_ALLOCATE
8556 		| NFS_CAP_DEALLOCATE
8557 		| NFS_CAP_SEEK,
8558 	.init_client = nfs41_init_client,
8559 	.shutdown_client = nfs41_shutdown_client,
8560 	.match_stateid = nfs41_match_stateid,
8561 	.find_root_sec = nfs41_find_root_sec,
8562 	.free_lock_state = nfs41_free_lock_state,
8563 	.call_sync_ops = &nfs41_call_sync_ops,
8564 	.alloc_seqid = nfs_alloc_no_seqid,
8565 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8566 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8567 	.state_renewal_ops = &nfs41_state_renewal_ops,
8568 };
8569 #endif
8570 
8571 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8572 	[0] = &nfs_v4_0_minor_ops,
8573 #if defined(CONFIG_NFS_V4_1)
8574 	[1] = &nfs_v4_1_minor_ops,
8575 #endif
8576 #if defined(CONFIG_NFS_V4_2)
8577 	[2] = &nfs_v4_2_minor_ops,
8578 #endif
8579 };
8580 
8581 static const struct inode_operations nfs4_dir_inode_operations = {
8582 	.create		= nfs_create,
8583 	.lookup		= nfs_lookup,
8584 	.atomic_open	= nfs_atomic_open,
8585 	.link		= nfs_link,
8586 	.unlink		= nfs_unlink,
8587 	.symlink	= nfs_symlink,
8588 	.mkdir		= nfs_mkdir,
8589 	.rmdir		= nfs_rmdir,
8590 	.mknod		= nfs_mknod,
8591 	.rename		= nfs_rename,
8592 	.permission	= nfs_permission,
8593 	.getattr	= nfs_getattr,
8594 	.setattr	= nfs_setattr,
8595 	.getxattr	= generic_getxattr,
8596 	.setxattr	= generic_setxattr,
8597 	.listxattr	= generic_listxattr,
8598 	.removexattr	= generic_removexattr,
8599 };
8600 
8601 static const struct inode_operations nfs4_file_inode_operations = {
8602 	.permission	= nfs_permission,
8603 	.getattr	= nfs_getattr,
8604 	.setattr	= nfs_setattr,
8605 	.getxattr	= generic_getxattr,
8606 	.setxattr	= generic_setxattr,
8607 	.listxattr	= generic_listxattr,
8608 	.removexattr	= generic_removexattr,
8609 };
8610 
8611 const struct nfs_rpc_ops nfs_v4_clientops = {
8612 	.version	= 4,			/* protocol version */
8613 	.dentry_ops	= &nfs4_dentry_operations,
8614 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8615 	.file_inode_ops	= &nfs4_file_inode_operations,
8616 	.file_ops	= &nfs4_file_operations,
8617 	.getroot	= nfs4_proc_get_root,
8618 	.submount	= nfs4_submount,
8619 	.try_mount	= nfs4_try_mount,
8620 	.getattr	= nfs4_proc_getattr,
8621 	.setattr	= nfs4_proc_setattr,
8622 	.lookup		= nfs4_proc_lookup,
8623 	.access		= nfs4_proc_access,
8624 	.readlink	= nfs4_proc_readlink,
8625 	.create		= nfs4_proc_create,
8626 	.remove		= nfs4_proc_remove,
8627 	.unlink_setup	= nfs4_proc_unlink_setup,
8628 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8629 	.unlink_done	= nfs4_proc_unlink_done,
8630 	.rename_setup	= nfs4_proc_rename_setup,
8631 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8632 	.rename_done	= nfs4_proc_rename_done,
8633 	.link		= nfs4_proc_link,
8634 	.symlink	= nfs4_proc_symlink,
8635 	.mkdir		= nfs4_proc_mkdir,
8636 	.rmdir		= nfs4_proc_remove,
8637 	.readdir	= nfs4_proc_readdir,
8638 	.mknod		= nfs4_proc_mknod,
8639 	.statfs		= nfs4_proc_statfs,
8640 	.fsinfo		= nfs4_proc_fsinfo,
8641 	.pathconf	= nfs4_proc_pathconf,
8642 	.set_capabilities = nfs4_server_capabilities,
8643 	.decode_dirent	= nfs4_decode_dirent,
8644 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8645 	.read_setup	= nfs4_proc_read_setup,
8646 	.read_done	= nfs4_read_done,
8647 	.write_setup	= nfs4_proc_write_setup,
8648 	.write_done	= nfs4_write_done,
8649 	.commit_setup	= nfs4_proc_commit_setup,
8650 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8651 	.commit_done	= nfs4_commit_done,
8652 	.lock		= nfs4_proc_lock,
8653 	.clear_acl_cache = nfs4_zap_acl_attr,
8654 	.close_context  = nfs4_close_context,
8655 	.open_context	= nfs4_atomic_open,
8656 	.have_delegation = nfs4_have_delegation,
8657 	.return_delegation = nfs4_inode_return_delegation,
8658 	.alloc_client	= nfs4_alloc_client,
8659 	.init_client	= nfs4_init_client,
8660 	.free_client	= nfs4_free_client,
8661 	.create_server	= nfs4_create_server,
8662 	.clone_server	= nfs_clone_server,
8663 };
8664 
8665 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8666 	.prefix	= XATTR_NAME_NFSV4_ACL,
8667 	.list	= nfs4_xattr_list_nfs4_acl,
8668 	.get	= nfs4_xattr_get_nfs4_acl,
8669 	.set	= nfs4_xattr_set_nfs4_acl,
8670 };
8671 
8672 const struct xattr_handler *nfs4_xattr_handlers[] = {
8673 	&nfs4_xattr_nfs4_acl_handler,
8674 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8675 	&nfs4_xattr_nfs4_label_handler,
8676 #endif
8677 	NULL
8678 };
8679 
8680 /*
8681  * Local variables:
8682  *  c-basic-offset: 8
8683  * End:
8684  */
8685