xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision d7a3d85e)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/file.h>
42 #include <linux/string.h>
43 #include <linux/ratelimit.h>
44 #include <linux/printk.h>
45 #include <linux/slab.h>
46 #include <linux/sunrpc/clnt.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58 
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69 
70 #include "nfs4trace.h"
71 
72 #define NFSDBG_FACILITY		NFSDBG_PROC
73 
74 #define NFS4_POLL_RETRY_MIN	(HZ/10)
75 #define NFS4_POLL_RETRY_MAX	(15*HZ)
76 
77 struct nfs4_opendata;
78 static int _nfs4_proc_open(struct nfs4_opendata *data);
79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
83 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
84 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
85 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
86 			    struct nfs_fattr *fattr, struct iattr *sattr,
87 			    struct nfs4_state *state, struct nfs4_label *ilabel,
88 			    struct nfs4_label *olabel);
89 #ifdef CONFIG_NFS_V4_1
90 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
91 		struct rpc_cred *);
92 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
93 		struct rpc_cred *);
94 #endif
95 
96 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
97 static inline struct nfs4_label *
98 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
99 	struct iattr *sattr, struct nfs4_label *label)
100 {
101 	int err;
102 
103 	if (label == NULL)
104 		return NULL;
105 
106 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
107 		return NULL;
108 
109 	err = security_dentry_init_security(dentry, sattr->ia_mode,
110 				&dentry->d_name, (void **)&label->label, &label->len);
111 	if (err == 0)
112 		return label;
113 
114 	return NULL;
115 }
116 static inline void
117 nfs4_label_release_security(struct nfs4_label *label)
118 {
119 	if (label)
120 		security_release_secctx(label->label, label->len);
121 }
122 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
123 {
124 	if (label)
125 		return server->attr_bitmask;
126 
127 	return server->attr_bitmask_nl;
128 }
129 #else
130 static inline struct nfs4_label *
131 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
132 	struct iattr *sattr, struct nfs4_label *l)
133 { return NULL; }
134 static inline void
135 nfs4_label_release_security(struct nfs4_label *label)
136 { return; }
137 static inline u32 *
138 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
139 { return server->attr_bitmask; }
140 #endif
141 
142 /* Prevent leaks of NFSv4 errors into userland */
143 static int nfs4_map_errors(int err)
144 {
145 	if (err >= -1000)
146 		return err;
147 	switch (err) {
148 	case -NFS4ERR_RESOURCE:
149 	case -NFS4ERR_LAYOUTTRYLATER:
150 	case -NFS4ERR_RECALLCONFLICT:
151 		return -EREMOTEIO;
152 	case -NFS4ERR_WRONGSEC:
153 	case -NFS4ERR_WRONG_CRED:
154 		return -EPERM;
155 	case -NFS4ERR_BADOWNER:
156 	case -NFS4ERR_BADNAME:
157 		return -EINVAL;
158 	case -NFS4ERR_SHARE_DENIED:
159 		return -EACCES;
160 	case -NFS4ERR_MINOR_VERS_MISMATCH:
161 		return -EPROTONOSUPPORT;
162 	case -NFS4ERR_FILE_OPEN:
163 		return -EBUSY;
164 	default:
165 		dprintk("%s could not handle NFSv4 error %d\n",
166 				__func__, -err);
167 		break;
168 	}
169 	return -EIO;
170 }
171 
172 /*
173  * This is our standard bitmap for GETATTR requests.
174  */
175 const u32 nfs4_fattr_bitmap[3] = {
176 	FATTR4_WORD0_TYPE
177 	| FATTR4_WORD0_CHANGE
178 	| FATTR4_WORD0_SIZE
179 	| FATTR4_WORD0_FSID
180 	| FATTR4_WORD0_FILEID,
181 	FATTR4_WORD1_MODE
182 	| FATTR4_WORD1_NUMLINKS
183 	| FATTR4_WORD1_OWNER
184 	| FATTR4_WORD1_OWNER_GROUP
185 	| FATTR4_WORD1_RAWDEV
186 	| FATTR4_WORD1_SPACE_USED
187 	| FATTR4_WORD1_TIME_ACCESS
188 	| FATTR4_WORD1_TIME_METADATA
189 	| FATTR4_WORD1_TIME_MODIFY
190 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
192 	FATTR4_WORD2_SECURITY_LABEL
193 #endif
194 };
195 
196 static const u32 nfs4_pnfs_open_bitmap[3] = {
197 	FATTR4_WORD0_TYPE
198 	| FATTR4_WORD0_CHANGE
199 	| FATTR4_WORD0_SIZE
200 	| FATTR4_WORD0_FSID
201 	| FATTR4_WORD0_FILEID,
202 	FATTR4_WORD1_MODE
203 	| FATTR4_WORD1_NUMLINKS
204 	| FATTR4_WORD1_OWNER
205 	| FATTR4_WORD1_OWNER_GROUP
206 	| FATTR4_WORD1_RAWDEV
207 	| FATTR4_WORD1_SPACE_USED
208 	| FATTR4_WORD1_TIME_ACCESS
209 	| FATTR4_WORD1_TIME_METADATA
210 	| FATTR4_WORD1_TIME_MODIFY,
211 	FATTR4_WORD2_MDSTHRESHOLD
212 };
213 
214 static const u32 nfs4_open_noattr_bitmap[3] = {
215 	FATTR4_WORD0_TYPE
216 	| FATTR4_WORD0_CHANGE
217 	| FATTR4_WORD0_FILEID,
218 };
219 
220 const u32 nfs4_statfs_bitmap[3] = {
221 	FATTR4_WORD0_FILES_AVAIL
222 	| FATTR4_WORD0_FILES_FREE
223 	| FATTR4_WORD0_FILES_TOTAL,
224 	FATTR4_WORD1_SPACE_AVAIL
225 	| FATTR4_WORD1_SPACE_FREE
226 	| FATTR4_WORD1_SPACE_TOTAL
227 };
228 
229 const u32 nfs4_pathconf_bitmap[3] = {
230 	FATTR4_WORD0_MAXLINK
231 	| FATTR4_WORD0_MAXNAME,
232 	0
233 };
234 
235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
236 			| FATTR4_WORD0_MAXREAD
237 			| FATTR4_WORD0_MAXWRITE
238 			| FATTR4_WORD0_LEASE_TIME,
239 			FATTR4_WORD1_TIME_DELTA
240 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
241 			FATTR4_WORD2_LAYOUT_BLKSIZE
242 };
243 
244 const u32 nfs4_fs_locations_bitmap[3] = {
245 	FATTR4_WORD0_TYPE
246 	| FATTR4_WORD0_CHANGE
247 	| FATTR4_WORD0_SIZE
248 	| FATTR4_WORD0_FSID
249 	| FATTR4_WORD0_FILEID
250 	| FATTR4_WORD0_FS_LOCATIONS,
251 	FATTR4_WORD1_MODE
252 	| FATTR4_WORD1_NUMLINKS
253 	| FATTR4_WORD1_OWNER
254 	| FATTR4_WORD1_OWNER_GROUP
255 	| FATTR4_WORD1_RAWDEV
256 	| FATTR4_WORD1_SPACE_USED
257 	| FATTR4_WORD1_TIME_ACCESS
258 	| FATTR4_WORD1_TIME_METADATA
259 	| FATTR4_WORD1_TIME_MODIFY
260 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
261 };
262 
263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
264 		struct nfs4_readdir_arg *readdir)
265 {
266 	__be32 *start, *p;
267 
268 	if (cookie > 2) {
269 		readdir->cookie = cookie;
270 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
271 		return;
272 	}
273 
274 	readdir->cookie = 0;
275 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
276 	if (cookie == 2)
277 		return;
278 
279 	/*
280 	 * NFSv4 servers do not return entries for '.' and '..'
281 	 * Therefore, we fake these entries here.  We let '.'
282 	 * have cookie 0 and '..' have cookie 1.  Note that
283 	 * when talking to the server, we always send cookie 0
284 	 * instead of 1 or 2.
285 	 */
286 	start = p = kmap_atomic(*readdir->pages);
287 
288 	if (cookie == 0) {
289 		*p++ = xdr_one;                                  /* next */
290 		*p++ = xdr_zero;                   /* cookie, first word */
291 		*p++ = xdr_one;                   /* cookie, second word */
292 		*p++ = xdr_one;                             /* entry len */
293 		memcpy(p, ".\0\0\0", 4);                        /* entry */
294 		p++;
295 		*p++ = xdr_one;                         /* bitmap length */
296 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
297 		*p++ = htonl(8);              /* attribute buffer length */
298 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
299 	}
300 
301 	*p++ = xdr_one;                                  /* next */
302 	*p++ = xdr_zero;                   /* cookie, first word */
303 	*p++ = xdr_two;                   /* cookie, second word */
304 	*p++ = xdr_two;                             /* entry len */
305 	memcpy(p, "..\0\0", 4);                         /* entry */
306 	p++;
307 	*p++ = xdr_one;                         /* bitmap length */
308 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
309 	*p++ = htonl(8);              /* attribute buffer length */
310 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
311 
312 	readdir->pgbase = (char *)p - (char *)start;
313 	readdir->count -= readdir->pgbase;
314 	kunmap_atomic(start);
315 }
316 
317 static long nfs4_update_delay(long *timeout)
318 {
319 	long ret;
320 	if (!timeout)
321 		return NFS4_POLL_RETRY_MAX;
322 	if (*timeout <= 0)
323 		*timeout = NFS4_POLL_RETRY_MIN;
324 	if (*timeout > NFS4_POLL_RETRY_MAX)
325 		*timeout = NFS4_POLL_RETRY_MAX;
326 	ret = *timeout;
327 	*timeout <<= 1;
328 	return ret;
329 }
330 
331 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
332 {
333 	int res = 0;
334 
335 	might_sleep();
336 
337 	freezable_schedule_timeout_killable_unsafe(
338 		nfs4_update_delay(timeout));
339 	if (fatal_signal_pending(current))
340 		res = -ERESTARTSYS;
341 	return res;
342 }
343 
344 /* This is the error handling routine for processes that are allowed
345  * to sleep.
346  */
347 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
348 {
349 	struct nfs_client *clp = server->nfs_client;
350 	struct nfs4_state *state = exception->state;
351 	struct inode *inode = exception->inode;
352 	int ret = errorcode;
353 
354 	exception->retry = 0;
355 	switch(errorcode) {
356 		case 0:
357 			return 0;
358 		case -NFS4ERR_OPENMODE:
359 			if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
360 				nfs4_inode_return_delegation(inode);
361 				exception->retry = 1;
362 				return 0;
363 			}
364 			if (state == NULL)
365 				break;
366 			ret = nfs4_schedule_stateid_recovery(server, state);
367 			if (ret < 0)
368 				break;
369 			goto wait_on_recovery;
370 		case -NFS4ERR_DELEG_REVOKED:
371 		case -NFS4ERR_ADMIN_REVOKED:
372 		case -NFS4ERR_BAD_STATEID:
373 			if (state == NULL)
374 				break;
375 			ret = nfs4_schedule_stateid_recovery(server, state);
376 			if (ret < 0)
377 				break;
378 			goto wait_on_recovery;
379 		case -NFS4ERR_EXPIRED:
380 			if (state != NULL) {
381 				ret = nfs4_schedule_stateid_recovery(server, state);
382 				if (ret < 0)
383 					break;
384 			}
385 		case -NFS4ERR_STALE_STATEID:
386 		case -NFS4ERR_STALE_CLIENTID:
387 			nfs4_schedule_lease_recovery(clp);
388 			goto wait_on_recovery;
389 		case -NFS4ERR_MOVED:
390 			ret = nfs4_schedule_migration_recovery(server);
391 			if (ret < 0)
392 				break;
393 			goto wait_on_recovery;
394 		case -NFS4ERR_LEASE_MOVED:
395 			nfs4_schedule_lease_moved_recovery(clp);
396 			goto wait_on_recovery;
397 #if defined(CONFIG_NFS_V4_1)
398 		case -NFS4ERR_BADSESSION:
399 		case -NFS4ERR_BADSLOT:
400 		case -NFS4ERR_BAD_HIGH_SLOT:
401 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
402 		case -NFS4ERR_DEADSESSION:
403 		case -NFS4ERR_SEQ_FALSE_RETRY:
404 		case -NFS4ERR_SEQ_MISORDERED:
405 			dprintk("%s ERROR: %d Reset session\n", __func__,
406 				errorcode);
407 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
408 			goto wait_on_recovery;
409 #endif /* defined(CONFIG_NFS_V4_1) */
410 		case -NFS4ERR_FILE_OPEN:
411 			if (exception->timeout > HZ) {
412 				/* We have retried a decent amount, time to
413 				 * fail
414 				 */
415 				ret = -EBUSY;
416 				break;
417 			}
418 		case -NFS4ERR_GRACE:
419 		case -NFS4ERR_DELAY:
420 			ret = nfs4_delay(server->client, &exception->timeout);
421 			if (ret != 0)
422 				break;
423 		case -NFS4ERR_RETRY_UNCACHED_REP:
424 		case -NFS4ERR_OLD_STATEID:
425 			exception->retry = 1;
426 			break;
427 		case -NFS4ERR_BADOWNER:
428 			/* The following works around a Linux server bug! */
429 		case -NFS4ERR_BADNAME:
430 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
431 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
432 				exception->retry = 1;
433 				printk(KERN_WARNING "NFS: v4 server %s "
434 						"does not accept raw "
435 						"uid/gids. "
436 						"Reenabling the idmapper.\n",
437 						server->nfs_client->cl_hostname);
438 			}
439 	}
440 	/* We failed to handle the error */
441 	return nfs4_map_errors(ret);
442 wait_on_recovery:
443 	ret = nfs4_wait_clnt_recover(clp);
444 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
445 		return -EIO;
446 	if (ret == 0)
447 		exception->retry = 1;
448 	return ret;
449 }
450 
451 /*
452  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
453  * or 'false' otherwise.
454  */
455 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
456 {
457 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
458 
459 	if (flavor == RPC_AUTH_GSS_KRB5I ||
460 	    flavor == RPC_AUTH_GSS_KRB5P)
461 		return true;
462 
463 	return false;
464 }
465 
466 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
467 {
468 	spin_lock(&clp->cl_lock);
469 	if (time_before(clp->cl_last_renewal,timestamp))
470 		clp->cl_last_renewal = timestamp;
471 	spin_unlock(&clp->cl_lock);
472 }
473 
474 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
475 {
476 	do_renew_lease(server->nfs_client, timestamp);
477 }
478 
479 struct nfs4_call_sync_data {
480 	const struct nfs_server *seq_server;
481 	struct nfs4_sequence_args *seq_args;
482 	struct nfs4_sequence_res *seq_res;
483 };
484 
485 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
486 			       struct nfs4_sequence_res *res, int cache_reply)
487 {
488 	args->sa_slot = NULL;
489 	args->sa_cache_this = cache_reply;
490 	args->sa_privileged = 0;
491 
492 	res->sr_slot = NULL;
493 }
494 
495 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
496 {
497 	args->sa_privileged = 1;
498 }
499 
500 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
501 			 struct nfs4_sequence_args *args,
502 			 struct nfs4_sequence_res *res,
503 			 struct rpc_task *task)
504 {
505 	struct nfs4_slot *slot;
506 
507 	/* slot already allocated? */
508 	if (res->sr_slot != NULL)
509 		goto out_start;
510 
511 	spin_lock(&tbl->slot_tbl_lock);
512 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
513 		goto out_sleep;
514 
515 	slot = nfs4_alloc_slot(tbl);
516 	if (IS_ERR(slot)) {
517 		if (slot == ERR_PTR(-ENOMEM))
518 			task->tk_timeout = HZ >> 2;
519 		goto out_sleep;
520 	}
521 	spin_unlock(&tbl->slot_tbl_lock);
522 
523 	args->sa_slot = slot;
524 	res->sr_slot = slot;
525 
526 out_start:
527 	rpc_call_start(task);
528 	return 0;
529 
530 out_sleep:
531 	if (args->sa_privileged)
532 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
533 				NULL, RPC_PRIORITY_PRIVILEGED);
534 	else
535 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
536 	spin_unlock(&tbl->slot_tbl_lock);
537 	return -EAGAIN;
538 }
539 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
540 
541 static int nfs40_sequence_done(struct rpc_task *task,
542 			       struct nfs4_sequence_res *res)
543 {
544 	struct nfs4_slot *slot = res->sr_slot;
545 	struct nfs4_slot_table *tbl;
546 
547 	if (slot == NULL)
548 		goto out;
549 
550 	tbl = slot->table;
551 	spin_lock(&tbl->slot_tbl_lock);
552 	if (!nfs41_wake_and_assign_slot(tbl, slot))
553 		nfs4_free_slot(tbl, slot);
554 	spin_unlock(&tbl->slot_tbl_lock);
555 
556 	res->sr_slot = NULL;
557 out:
558 	return 1;
559 }
560 
561 #if defined(CONFIG_NFS_V4_1)
562 
563 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
564 {
565 	struct nfs4_session *session;
566 	struct nfs4_slot_table *tbl;
567 	struct nfs4_slot *slot = res->sr_slot;
568 	bool send_new_highest_used_slotid = false;
569 
570 	tbl = slot->table;
571 	session = tbl->session;
572 
573 	spin_lock(&tbl->slot_tbl_lock);
574 	/* Be nice to the server: try to ensure that the last transmitted
575 	 * value for highest_user_slotid <= target_highest_slotid
576 	 */
577 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
578 		send_new_highest_used_slotid = true;
579 
580 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
581 		send_new_highest_used_slotid = false;
582 		goto out_unlock;
583 	}
584 	nfs4_free_slot(tbl, slot);
585 
586 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
587 		send_new_highest_used_slotid = false;
588 out_unlock:
589 	spin_unlock(&tbl->slot_tbl_lock);
590 	res->sr_slot = NULL;
591 	if (send_new_highest_used_slotid)
592 		nfs41_server_notify_highest_slotid_update(session->clp);
593 }
594 
595 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
596 {
597 	struct nfs4_session *session;
598 	struct nfs4_slot *slot = res->sr_slot;
599 	struct nfs_client *clp;
600 	bool interrupted = false;
601 	int ret = 1;
602 
603 	if (slot == NULL)
604 		goto out_noaction;
605 	/* don't increment the sequence number if the task wasn't sent */
606 	if (!RPC_WAS_SENT(task))
607 		goto out;
608 
609 	session = slot->table->session;
610 
611 	if (slot->interrupted) {
612 		slot->interrupted = 0;
613 		interrupted = true;
614 	}
615 
616 	trace_nfs4_sequence_done(session, res);
617 	/* Check the SEQUENCE operation status */
618 	switch (res->sr_status) {
619 	case 0:
620 		/* Update the slot's sequence and clientid lease timer */
621 		++slot->seq_nr;
622 		clp = session->clp;
623 		do_renew_lease(clp, res->sr_timestamp);
624 		/* Check sequence flags */
625 		if (res->sr_status_flags != 0)
626 			nfs4_schedule_lease_recovery(clp);
627 		nfs41_update_target_slotid(slot->table, slot, res);
628 		break;
629 	case 1:
630 		/*
631 		 * sr_status remains 1 if an RPC level error occurred.
632 		 * The server may or may not have processed the sequence
633 		 * operation..
634 		 * Mark the slot as having hosted an interrupted RPC call.
635 		 */
636 		slot->interrupted = 1;
637 		goto out;
638 	case -NFS4ERR_DELAY:
639 		/* The server detected a resend of the RPC call and
640 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
641 		 * of RFC5661.
642 		 */
643 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
644 			__func__,
645 			slot->slot_nr,
646 			slot->seq_nr);
647 		goto out_retry;
648 	case -NFS4ERR_BADSLOT:
649 		/*
650 		 * The slot id we used was probably retired. Try again
651 		 * using a different slot id.
652 		 */
653 		goto retry_nowait;
654 	case -NFS4ERR_SEQ_MISORDERED:
655 		/*
656 		 * Was the last operation on this sequence interrupted?
657 		 * If so, retry after bumping the sequence number.
658 		 */
659 		if (interrupted) {
660 			++slot->seq_nr;
661 			goto retry_nowait;
662 		}
663 		/*
664 		 * Could this slot have been previously retired?
665 		 * If so, then the server may be expecting seq_nr = 1!
666 		 */
667 		if (slot->seq_nr != 1) {
668 			slot->seq_nr = 1;
669 			goto retry_nowait;
670 		}
671 		break;
672 	case -NFS4ERR_SEQ_FALSE_RETRY:
673 		++slot->seq_nr;
674 		goto retry_nowait;
675 	default:
676 		/* Just update the slot sequence no. */
677 		++slot->seq_nr;
678 	}
679 out:
680 	/* The session may be reset by one of the error handlers. */
681 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
682 	nfs41_sequence_free_slot(res);
683 out_noaction:
684 	return ret;
685 retry_nowait:
686 	if (rpc_restart_call_prepare(task)) {
687 		task->tk_status = 0;
688 		ret = 0;
689 	}
690 	goto out;
691 out_retry:
692 	if (!rpc_restart_call(task))
693 		goto out;
694 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
695 	return 0;
696 }
697 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
698 
699 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
700 {
701 	if (res->sr_slot == NULL)
702 		return 1;
703 	if (!res->sr_slot->table->session)
704 		return nfs40_sequence_done(task, res);
705 	return nfs41_sequence_done(task, res);
706 }
707 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
708 
709 int nfs41_setup_sequence(struct nfs4_session *session,
710 				struct nfs4_sequence_args *args,
711 				struct nfs4_sequence_res *res,
712 				struct rpc_task *task)
713 {
714 	struct nfs4_slot *slot;
715 	struct nfs4_slot_table *tbl;
716 
717 	dprintk("--> %s\n", __func__);
718 	/* slot already allocated? */
719 	if (res->sr_slot != NULL)
720 		goto out_success;
721 
722 	tbl = &session->fc_slot_table;
723 
724 	task->tk_timeout = 0;
725 
726 	spin_lock(&tbl->slot_tbl_lock);
727 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
728 	    !args->sa_privileged) {
729 		/* The state manager will wait until the slot table is empty */
730 		dprintk("%s session is draining\n", __func__);
731 		goto out_sleep;
732 	}
733 
734 	slot = nfs4_alloc_slot(tbl);
735 	if (IS_ERR(slot)) {
736 		/* If out of memory, try again in 1/4 second */
737 		if (slot == ERR_PTR(-ENOMEM))
738 			task->tk_timeout = HZ >> 2;
739 		dprintk("<-- %s: no free slots\n", __func__);
740 		goto out_sleep;
741 	}
742 	spin_unlock(&tbl->slot_tbl_lock);
743 
744 	args->sa_slot = slot;
745 
746 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
747 			slot->slot_nr, slot->seq_nr);
748 
749 	res->sr_slot = slot;
750 	res->sr_timestamp = jiffies;
751 	res->sr_status_flags = 0;
752 	/*
753 	 * sr_status is only set in decode_sequence, and so will remain
754 	 * set to 1 if an rpc level failure occurs.
755 	 */
756 	res->sr_status = 1;
757 	trace_nfs4_setup_sequence(session, args);
758 out_success:
759 	rpc_call_start(task);
760 	return 0;
761 out_sleep:
762 	/* Privileged tasks are queued with top priority */
763 	if (args->sa_privileged)
764 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
765 				NULL, RPC_PRIORITY_PRIVILEGED);
766 	else
767 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
768 	spin_unlock(&tbl->slot_tbl_lock);
769 	return -EAGAIN;
770 }
771 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
772 
773 static int nfs4_setup_sequence(const struct nfs_server *server,
774 			       struct nfs4_sequence_args *args,
775 			       struct nfs4_sequence_res *res,
776 			       struct rpc_task *task)
777 {
778 	struct nfs4_session *session = nfs4_get_session(server);
779 	int ret = 0;
780 
781 	if (!session)
782 		return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
783 					    args, res, task);
784 
785 	dprintk("--> %s clp %p session %p sr_slot %u\n",
786 		__func__, session->clp, session, res->sr_slot ?
787 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
788 
789 	ret = nfs41_setup_sequence(session, args, res, task);
790 
791 	dprintk("<-- %s status=%d\n", __func__, ret);
792 	return ret;
793 }
794 
795 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
796 {
797 	struct nfs4_call_sync_data *data = calldata;
798 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
799 
800 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
801 
802 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
803 }
804 
805 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
806 {
807 	struct nfs4_call_sync_data *data = calldata;
808 
809 	nfs41_sequence_done(task, data->seq_res);
810 }
811 
812 static const struct rpc_call_ops nfs41_call_sync_ops = {
813 	.rpc_call_prepare = nfs41_call_sync_prepare,
814 	.rpc_call_done = nfs41_call_sync_done,
815 };
816 
817 #else	/* !CONFIG_NFS_V4_1 */
818 
819 static int nfs4_setup_sequence(const struct nfs_server *server,
820 			       struct nfs4_sequence_args *args,
821 			       struct nfs4_sequence_res *res,
822 			       struct rpc_task *task)
823 {
824 	return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
825 				    args, res, task);
826 }
827 
828 int nfs4_sequence_done(struct rpc_task *task,
829 		       struct nfs4_sequence_res *res)
830 {
831 	return nfs40_sequence_done(task, res);
832 }
833 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
834 
835 #endif	/* !CONFIG_NFS_V4_1 */
836 
837 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
838 {
839 	struct nfs4_call_sync_data *data = calldata;
840 	nfs4_setup_sequence(data->seq_server,
841 				data->seq_args, data->seq_res, task);
842 }
843 
844 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
845 {
846 	struct nfs4_call_sync_data *data = calldata;
847 	nfs4_sequence_done(task, data->seq_res);
848 }
849 
850 static const struct rpc_call_ops nfs40_call_sync_ops = {
851 	.rpc_call_prepare = nfs40_call_sync_prepare,
852 	.rpc_call_done = nfs40_call_sync_done,
853 };
854 
855 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
856 				   struct nfs_server *server,
857 				   struct rpc_message *msg,
858 				   struct nfs4_sequence_args *args,
859 				   struct nfs4_sequence_res *res)
860 {
861 	int ret;
862 	struct rpc_task *task;
863 	struct nfs_client *clp = server->nfs_client;
864 	struct nfs4_call_sync_data data = {
865 		.seq_server = server,
866 		.seq_args = args,
867 		.seq_res = res,
868 	};
869 	struct rpc_task_setup task_setup = {
870 		.rpc_client = clnt,
871 		.rpc_message = msg,
872 		.callback_ops = clp->cl_mvops->call_sync_ops,
873 		.callback_data = &data
874 	};
875 
876 	task = rpc_run_task(&task_setup);
877 	if (IS_ERR(task))
878 		ret = PTR_ERR(task);
879 	else {
880 		ret = task->tk_status;
881 		rpc_put_task(task);
882 	}
883 	return ret;
884 }
885 
886 int nfs4_call_sync(struct rpc_clnt *clnt,
887 		   struct nfs_server *server,
888 		   struct rpc_message *msg,
889 		   struct nfs4_sequence_args *args,
890 		   struct nfs4_sequence_res *res,
891 		   int cache_reply)
892 {
893 	nfs4_init_sequence(args, res, cache_reply);
894 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
895 }
896 
897 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
898 {
899 	struct nfs_inode *nfsi = NFS_I(dir);
900 
901 	spin_lock(&dir->i_lock);
902 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
903 	if (!cinfo->atomic || cinfo->before != dir->i_version)
904 		nfs_force_lookup_revalidate(dir);
905 	dir->i_version = cinfo->after;
906 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
907 	nfs_fscache_invalidate(dir);
908 	spin_unlock(&dir->i_lock);
909 }
910 
911 struct nfs4_opendata {
912 	struct kref kref;
913 	struct nfs_openargs o_arg;
914 	struct nfs_openres o_res;
915 	struct nfs_open_confirmargs c_arg;
916 	struct nfs_open_confirmres c_res;
917 	struct nfs4_string owner_name;
918 	struct nfs4_string group_name;
919 	struct nfs_fattr f_attr;
920 	struct nfs4_label *f_label;
921 	struct dentry *dir;
922 	struct dentry *dentry;
923 	struct nfs4_state_owner *owner;
924 	struct nfs4_state *state;
925 	struct iattr attrs;
926 	unsigned long timestamp;
927 	unsigned int rpc_done : 1;
928 	unsigned int file_created : 1;
929 	unsigned int is_recover : 1;
930 	int rpc_status;
931 	int cancelled;
932 };
933 
934 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
935 		int err, struct nfs4_exception *exception)
936 {
937 	if (err != -EINVAL)
938 		return false;
939 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
940 		return false;
941 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
942 	exception->retry = 1;
943 	return true;
944 }
945 
946 static u32
947 nfs4_map_atomic_open_share(struct nfs_server *server,
948 		fmode_t fmode, int openflags)
949 {
950 	u32 res = 0;
951 
952 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
953 	case FMODE_READ:
954 		res = NFS4_SHARE_ACCESS_READ;
955 		break;
956 	case FMODE_WRITE:
957 		res = NFS4_SHARE_ACCESS_WRITE;
958 		break;
959 	case FMODE_READ|FMODE_WRITE:
960 		res = NFS4_SHARE_ACCESS_BOTH;
961 	}
962 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
963 		goto out;
964 	/* Want no delegation if we're using O_DIRECT */
965 	if (openflags & O_DIRECT)
966 		res |= NFS4_SHARE_WANT_NO_DELEG;
967 out:
968 	return res;
969 }
970 
971 static enum open_claim_type4
972 nfs4_map_atomic_open_claim(struct nfs_server *server,
973 		enum open_claim_type4 claim)
974 {
975 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
976 		return claim;
977 	switch (claim) {
978 	default:
979 		return claim;
980 	case NFS4_OPEN_CLAIM_FH:
981 		return NFS4_OPEN_CLAIM_NULL;
982 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
983 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
984 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
985 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
986 	}
987 }
988 
989 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
990 {
991 	p->o_res.f_attr = &p->f_attr;
992 	p->o_res.f_label = p->f_label;
993 	p->o_res.seqid = p->o_arg.seqid;
994 	p->c_res.seqid = p->c_arg.seqid;
995 	p->o_res.server = p->o_arg.server;
996 	p->o_res.access_request = p->o_arg.access;
997 	nfs_fattr_init(&p->f_attr);
998 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
999 }
1000 
1001 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1002 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1003 		const struct iattr *attrs,
1004 		struct nfs4_label *label,
1005 		enum open_claim_type4 claim,
1006 		gfp_t gfp_mask)
1007 {
1008 	struct dentry *parent = dget_parent(dentry);
1009 	struct inode *dir = d_inode(parent);
1010 	struct nfs_server *server = NFS_SERVER(dir);
1011 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1012 	struct nfs4_opendata *p;
1013 
1014 	p = kzalloc(sizeof(*p), gfp_mask);
1015 	if (p == NULL)
1016 		goto err;
1017 
1018 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1019 	if (IS_ERR(p->f_label))
1020 		goto err_free_p;
1021 
1022 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1023 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1024 	if (IS_ERR(p->o_arg.seqid))
1025 		goto err_free_label;
1026 	nfs_sb_active(dentry->d_sb);
1027 	p->dentry = dget(dentry);
1028 	p->dir = parent;
1029 	p->owner = sp;
1030 	atomic_inc(&sp->so_count);
1031 	p->o_arg.open_flags = flags;
1032 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1033 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1034 			fmode, flags);
1035 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1036 	 * will return permission denied for all bits until close */
1037 	if (!(flags & O_EXCL)) {
1038 		/* ask server to check for all possible rights as results
1039 		 * are cached */
1040 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1041 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1042 	}
1043 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1044 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1045 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1046 	p->o_arg.name = &dentry->d_name;
1047 	p->o_arg.server = server;
1048 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1049 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1050 	p->o_arg.label = label;
1051 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1052 	switch (p->o_arg.claim) {
1053 	case NFS4_OPEN_CLAIM_NULL:
1054 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1055 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1056 		p->o_arg.fh = NFS_FH(dir);
1057 		break;
1058 	case NFS4_OPEN_CLAIM_PREVIOUS:
1059 	case NFS4_OPEN_CLAIM_FH:
1060 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1061 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1062 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1063 	}
1064 	if (attrs != NULL && attrs->ia_valid != 0) {
1065 		__u32 verf[2];
1066 
1067 		p->o_arg.u.attrs = &p->attrs;
1068 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1069 
1070 		verf[0] = jiffies;
1071 		verf[1] = current->pid;
1072 		memcpy(p->o_arg.u.verifier.data, verf,
1073 				sizeof(p->o_arg.u.verifier.data));
1074 	}
1075 	p->c_arg.fh = &p->o_res.fh;
1076 	p->c_arg.stateid = &p->o_res.stateid;
1077 	p->c_arg.seqid = p->o_arg.seqid;
1078 	nfs4_init_opendata_res(p);
1079 	kref_init(&p->kref);
1080 	return p;
1081 
1082 err_free_label:
1083 	nfs4_label_free(p->f_label);
1084 err_free_p:
1085 	kfree(p);
1086 err:
1087 	dput(parent);
1088 	return NULL;
1089 }
1090 
1091 static void nfs4_opendata_free(struct kref *kref)
1092 {
1093 	struct nfs4_opendata *p = container_of(kref,
1094 			struct nfs4_opendata, kref);
1095 	struct super_block *sb = p->dentry->d_sb;
1096 
1097 	nfs_free_seqid(p->o_arg.seqid);
1098 	if (p->state != NULL)
1099 		nfs4_put_open_state(p->state);
1100 	nfs4_put_state_owner(p->owner);
1101 
1102 	nfs4_label_free(p->f_label);
1103 
1104 	dput(p->dir);
1105 	dput(p->dentry);
1106 	nfs_sb_deactive(sb);
1107 	nfs_fattr_free_names(&p->f_attr);
1108 	kfree(p->f_attr.mdsthreshold);
1109 	kfree(p);
1110 }
1111 
1112 static void nfs4_opendata_put(struct nfs4_opendata *p)
1113 {
1114 	if (p != NULL)
1115 		kref_put(&p->kref, nfs4_opendata_free);
1116 }
1117 
1118 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1119 {
1120 	int ret;
1121 
1122 	ret = rpc_wait_for_completion_task(task);
1123 	return ret;
1124 }
1125 
1126 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1127 {
1128 	int ret = 0;
1129 
1130 	if (open_mode & (O_EXCL|O_TRUNC))
1131 		goto out;
1132 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1133 		case FMODE_READ:
1134 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1135 				&& state->n_rdonly != 0;
1136 			break;
1137 		case FMODE_WRITE:
1138 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1139 				&& state->n_wronly != 0;
1140 			break;
1141 		case FMODE_READ|FMODE_WRITE:
1142 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1143 				&& state->n_rdwr != 0;
1144 	}
1145 out:
1146 	return ret;
1147 }
1148 
1149 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1150 {
1151 	if (delegation == NULL)
1152 		return 0;
1153 	if ((delegation->type & fmode) != fmode)
1154 		return 0;
1155 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1156 		return 0;
1157 	nfs_mark_delegation_referenced(delegation);
1158 	return 1;
1159 }
1160 
1161 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1162 {
1163 	switch (fmode) {
1164 		case FMODE_WRITE:
1165 			state->n_wronly++;
1166 			break;
1167 		case FMODE_READ:
1168 			state->n_rdonly++;
1169 			break;
1170 		case FMODE_READ|FMODE_WRITE:
1171 			state->n_rdwr++;
1172 	}
1173 	nfs4_state_set_mode_locked(state, state->state | fmode);
1174 }
1175 
1176 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1177 {
1178 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1179 	bool need_recover = false;
1180 
1181 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1182 		need_recover = true;
1183 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1184 		need_recover = true;
1185 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1186 		need_recover = true;
1187 	if (need_recover)
1188 		nfs4_state_mark_reclaim_nograce(clp, state);
1189 }
1190 
1191 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1192 		nfs4_stateid *stateid)
1193 {
1194 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1195 		return true;
1196 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1197 		nfs_test_and_clear_all_open_stateid(state);
1198 		return true;
1199 	}
1200 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1201 		return true;
1202 	return false;
1203 }
1204 
1205 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1206 {
1207 	if (state->n_wronly)
1208 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1209 	if (state->n_rdonly)
1210 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1211 	if (state->n_rdwr)
1212 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1213 }
1214 
1215 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1216 		nfs4_stateid *stateid, fmode_t fmode)
1217 {
1218 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1219 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1220 	case FMODE_WRITE:
1221 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1222 		break;
1223 	case FMODE_READ:
1224 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1225 		break;
1226 	case 0:
1227 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1228 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1229 		clear_bit(NFS_OPEN_STATE, &state->flags);
1230 	}
1231 	if (stateid == NULL)
1232 		return;
1233 	/* Handle races with OPEN */
1234 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1235 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1236 		nfs_resync_open_stateid_locked(state);
1237 		return;
1238 	}
1239 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1240 		nfs4_stateid_copy(&state->stateid, stateid);
1241 	nfs4_stateid_copy(&state->open_stateid, stateid);
1242 }
1243 
1244 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1245 {
1246 	write_seqlock(&state->seqlock);
1247 	nfs_clear_open_stateid_locked(state, stateid, fmode);
1248 	write_sequnlock(&state->seqlock);
1249 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1250 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1251 }
1252 
1253 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1254 {
1255 	switch (fmode) {
1256 		case FMODE_READ:
1257 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1258 			break;
1259 		case FMODE_WRITE:
1260 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1261 			break;
1262 		case FMODE_READ|FMODE_WRITE:
1263 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1264 	}
1265 	if (!nfs_need_update_open_stateid(state, stateid))
1266 		return;
1267 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1268 		nfs4_stateid_copy(&state->stateid, stateid);
1269 	nfs4_stateid_copy(&state->open_stateid, stateid);
1270 }
1271 
1272 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1273 {
1274 	/*
1275 	 * Protect the call to nfs4_state_set_mode_locked and
1276 	 * serialise the stateid update
1277 	 */
1278 	write_seqlock(&state->seqlock);
1279 	if (deleg_stateid != NULL) {
1280 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1281 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1282 	}
1283 	if (open_stateid != NULL)
1284 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1285 	write_sequnlock(&state->seqlock);
1286 	spin_lock(&state->owner->so_lock);
1287 	update_open_stateflags(state, fmode);
1288 	spin_unlock(&state->owner->so_lock);
1289 }
1290 
1291 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1292 {
1293 	struct nfs_inode *nfsi = NFS_I(state->inode);
1294 	struct nfs_delegation *deleg_cur;
1295 	int ret = 0;
1296 
1297 	fmode &= (FMODE_READ|FMODE_WRITE);
1298 
1299 	rcu_read_lock();
1300 	deleg_cur = rcu_dereference(nfsi->delegation);
1301 	if (deleg_cur == NULL)
1302 		goto no_delegation;
1303 
1304 	spin_lock(&deleg_cur->lock);
1305 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1306 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1307 	    (deleg_cur->type & fmode) != fmode)
1308 		goto no_delegation_unlock;
1309 
1310 	if (delegation == NULL)
1311 		delegation = &deleg_cur->stateid;
1312 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1313 		goto no_delegation_unlock;
1314 
1315 	nfs_mark_delegation_referenced(deleg_cur);
1316 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1317 	ret = 1;
1318 no_delegation_unlock:
1319 	spin_unlock(&deleg_cur->lock);
1320 no_delegation:
1321 	rcu_read_unlock();
1322 
1323 	if (!ret && open_stateid != NULL) {
1324 		__update_open_stateid(state, open_stateid, NULL, fmode);
1325 		ret = 1;
1326 	}
1327 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1328 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1329 
1330 	return ret;
1331 }
1332 
1333 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1334 		const nfs4_stateid *stateid)
1335 {
1336 	struct nfs4_state *state = lsp->ls_state;
1337 	bool ret = false;
1338 
1339 	spin_lock(&state->state_lock);
1340 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1341 		goto out_noupdate;
1342 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1343 		goto out_noupdate;
1344 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1345 	ret = true;
1346 out_noupdate:
1347 	spin_unlock(&state->state_lock);
1348 	return ret;
1349 }
1350 
1351 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1352 {
1353 	struct nfs_delegation *delegation;
1354 
1355 	rcu_read_lock();
1356 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1357 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1358 		rcu_read_unlock();
1359 		return;
1360 	}
1361 	rcu_read_unlock();
1362 	nfs4_inode_return_delegation(inode);
1363 }
1364 
1365 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1366 {
1367 	struct nfs4_state *state = opendata->state;
1368 	struct nfs_inode *nfsi = NFS_I(state->inode);
1369 	struct nfs_delegation *delegation;
1370 	int open_mode = opendata->o_arg.open_flags;
1371 	fmode_t fmode = opendata->o_arg.fmode;
1372 	nfs4_stateid stateid;
1373 	int ret = -EAGAIN;
1374 
1375 	for (;;) {
1376 		spin_lock(&state->owner->so_lock);
1377 		if (can_open_cached(state, fmode, open_mode)) {
1378 			update_open_stateflags(state, fmode);
1379 			spin_unlock(&state->owner->so_lock);
1380 			goto out_return_state;
1381 		}
1382 		spin_unlock(&state->owner->so_lock);
1383 		rcu_read_lock();
1384 		delegation = rcu_dereference(nfsi->delegation);
1385 		if (!can_open_delegated(delegation, fmode)) {
1386 			rcu_read_unlock();
1387 			break;
1388 		}
1389 		/* Save the delegation */
1390 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1391 		rcu_read_unlock();
1392 		nfs_release_seqid(opendata->o_arg.seqid);
1393 		if (!opendata->is_recover) {
1394 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1395 			if (ret != 0)
1396 				goto out;
1397 		}
1398 		ret = -EAGAIN;
1399 
1400 		/* Try to update the stateid using the delegation */
1401 		if (update_open_stateid(state, NULL, &stateid, fmode))
1402 			goto out_return_state;
1403 	}
1404 out:
1405 	return ERR_PTR(ret);
1406 out_return_state:
1407 	atomic_inc(&state->count);
1408 	return state;
1409 }
1410 
1411 static void
1412 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1413 {
1414 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1415 	struct nfs_delegation *delegation;
1416 	int delegation_flags = 0;
1417 
1418 	rcu_read_lock();
1419 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1420 	if (delegation)
1421 		delegation_flags = delegation->flags;
1422 	rcu_read_unlock();
1423 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1424 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1425 				   "returning a delegation for "
1426 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1427 				   clp->cl_hostname);
1428 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1429 		nfs_inode_set_delegation(state->inode,
1430 					 data->owner->so_cred,
1431 					 &data->o_res);
1432 	else
1433 		nfs_inode_reclaim_delegation(state->inode,
1434 					     data->owner->so_cred,
1435 					     &data->o_res);
1436 }
1437 
1438 /*
1439  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1440  * and update the nfs4_state.
1441  */
1442 static struct nfs4_state *
1443 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1444 {
1445 	struct inode *inode = data->state->inode;
1446 	struct nfs4_state *state = data->state;
1447 	int ret;
1448 
1449 	if (!data->rpc_done) {
1450 		if (data->rpc_status) {
1451 			ret = data->rpc_status;
1452 			goto err;
1453 		}
1454 		/* cached opens have already been processed */
1455 		goto update;
1456 	}
1457 
1458 	ret = nfs_refresh_inode(inode, &data->f_attr);
1459 	if (ret)
1460 		goto err;
1461 
1462 	if (data->o_res.delegation_type != 0)
1463 		nfs4_opendata_check_deleg(data, state);
1464 update:
1465 	update_open_stateid(state, &data->o_res.stateid, NULL,
1466 			    data->o_arg.fmode);
1467 	atomic_inc(&state->count);
1468 
1469 	return state;
1470 err:
1471 	return ERR_PTR(ret);
1472 
1473 }
1474 
1475 static struct nfs4_state *
1476 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1477 {
1478 	struct inode *inode;
1479 	struct nfs4_state *state = NULL;
1480 	int ret;
1481 
1482 	if (!data->rpc_done) {
1483 		state = nfs4_try_open_cached(data);
1484 		goto out;
1485 	}
1486 
1487 	ret = -EAGAIN;
1488 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1489 		goto err;
1490 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1491 	ret = PTR_ERR(inode);
1492 	if (IS_ERR(inode))
1493 		goto err;
1494 	ret = -ENOMEM;
1495 	state = nfs4_get_open_state(inode, data->owner);
1496 	if (state == NULL)
1497 		goto err_put_inode;
1498 	if (data->o_res.delegation_type != 0)
1499 		nfs4_opendata_check_deleg(data, state);
1500 	update_open_stateid(state, &data->o_res.stateid, NULL,
1501 			data->o_arg.fmode);
1502 	iput(inode);
1503 out:
1504 	nfs_release_seqid(data->o_arg.seqid);
1505 	return state;
1506 err_put_inode:
1507 	iput(inode);
1508 err:
1509 	return ERR_PTR(ret);
1510 }
1511 
1512 static struct nfs4_state *
1513 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1514 {
1515 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1516 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1517 	return _nfs4_opendata_to_nfs4_state(data);
1518 }
1519 
1520 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1521 {
1522 	struct nfs_inode *nfsi = NFS_I(state->inode);
1523 	struct nfs_open_context *ctx;
1524 
1525 	spin_lock(&state->inode->i_lock);
1526 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1527 		if (ctx->state != state)
1528 			continue;
1529 		get_nfs_open_context(ctx);
1530 		spin_unlock(&state->inode->i_lock);
1531 		return ctx;
1532 	}
1533 	spin_unlock(&state->inode->i_lock);
1534 	return ERR_PTR(-ENOENT);
1535 }
1536 
1537 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1538 		struct nfs4_state *state, enum open_claim_type4 claim)
1539 {
1540 	struct nfs4_opendata *opendata;
1541 
1542 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1543 			NULL, NULL, claim, GFP_NOFS);
1544 	if (opendata == NULL)
1545 		return ERR_PTR(-ENOMEM);
1546 	opendata->state = state;
1547 	atomic_inc(&state->count);
1548 	return opendata;
1549 }
1550 
1551 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1552 {
1553 	struct nfs4_state *newstate;
1554 	int ret;
1555 
1556 	opendata->o_arg.open_flags = 0;
1557 	opendata->o_arg.fmode = fmode;
1558 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1559 			NFS_SB(opendata->dentry->d_sb),
1560 			fmode, 0);
1561 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1562 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1563 	nfs4_init_opendata_res(opendata);
1564 	ret = _nfs4_recover_proc_open(opendata);
1565 	if (ret != 0)
1566 		return ret;
1567 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1568 	if (IS_ERR(newstate))
1569 		return PTR_ERR(newstate);
1570 	nfs4_close_state(newstate, fmode);
1571 	*res = newstate;
1572 	return 0;
1573 }
1574 
1575 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1576 {
1577 	struct nfs4_state *newstate;
1578 	int ret;
1579 
1580 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1581 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1582 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1583 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1584 	/* memory barrier prior to reading state->n_* */
1585 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1586 	clear_bit(NFS_OPEN_STATE, &state->flags);
1587 	smp_rmb();
1588 	if (state->n_rdwr != 0) {
1589 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1590 		if (ret != 0)
1591 			return ret;
1592 		if (newstate != state)
1593 			return -ESTALE;
1594 	}
1595 	if (state->n_wronly != 0) {
1596 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1597 		if (ret != 0)
1598 			return ret;
1599 		if (newstate != state)
1600 			return -ESTALE;
1601 	}
1602 	if (state->n_rdonly != 0) {
1603 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1604 		if (ret != 0)
1605 			return ret;
1606 		if (newstate != state)
1607 			return -ESTALE;
1608 	}
1609 	/*
1610 	 * We may have performed cached opens for all three recoveries.
1611 	 * Check if we need to update the current stateid.
1612 	 */
1613 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1614 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1615 		write_seqlock(&state->seqlock);
1616 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1617 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1618 		write_sequnlock(&state->seqlock);
1619 	}
1620 	return 0;
1621 }
1622 
1623 /*
1624  * OPEN_RECLAIM:
1625  * 	reclaim state on the server after a reboot.
1626  */
1627 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1628 {
1629 	struct nfs_delegation *delegation;
1630 	struct nfs4_opendata *opendata;
1631 	fmode_t delegation_type = 0;
1632 	int status;
1633 
1634 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1635 			NFS4_OPEN_CLAIM_PREVIOUS);
1636 	if (IS_ERR(opendata))
1637 		return PTR_ERR(opendata);
1638 	rcu_read_lock();
1639 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1640 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1641 		delegation_type = delegation->type;
1642 	rcu_read_unlock();
1643 	opendata->o_arg.u.delegation_type = delegation_type;
1644 	status = nfs4_open_recover(opendata, state);
1645 	nfs4_opendata_put(opendata);
1646 	return status;
1647 }
1648 
1649 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1650 {
1651 	struct nfs_server *server = NFS_SERVER(state->inode);
1652 	struct nfs4_exception exception = { };
1653 	int err;
1654 	do {
1655 		err = _nfs4_do_open_reclaim(ctx, state);
1656 		trace_nfs4_open_reclaim(ctx, 0, err);
1657 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1658 			continue;
1659 		if (err != -NFS4ERR_DELAY)
1660 			break;
1661 		nfs4_handle_exception(server, err, &exception);
1662 	} while (exception.retry);
1663 	return err;
1664 }
1665 
1666 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1667 {
1668 	struct nfs_open_context *ctx;
1669 	int ret;
1670 
1671 	ctx = nfs4_state_find_open_context(state);
1672 	if (IS_ERR(ctx))
1673 		return -EAGAIN;
1674 	ret = nfs4_do_open_reclaim(ctx, state);
1675 	put_nfs_open_context(ctx);
1676 	return ret;
1677 }
1678 
1679 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1680 {
1681 	switch (err) {
1682 		default:
1683 			printk(KERN_ERR "NFS: %s: unhandled error "
1684 					"%d.\n", __func__, err);
1685 		case 0:
1686 		case -ENOENT:
1687 		case -ESTALE:
1688 			break;
1689 		case -NFS4ERR_BADSESSION:
1690 		case -NFS4ERR_BADSLOT:
1691 		case -NFS4ERR_BAD_HIGH_SLOT:
1692 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1693 		case -NFS4ERR_DEADSESSION:
1694 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1695 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1696 			return -EAGAIN;
1697 		case -NFS4ERR_STALE_CLIENTID:
1698 		case -NFS4ERR_STALE_STATEID:
1699 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1700 		case -NFS4ERR_EXPIRED:
1701 			/* Don't recall a delegation if it was lost */
1702 			nfs4_schedule_lease_recovery(server->nfs_client);
1703 			return -EAGAIN;
1704 		case -NFS4ERR_MOVED:
1705 			nfs4_schedule_migration_recovery(server);
1706 			return -EAGAIN;
1707 		case -NFS4ERR_LEASE_MOVED:
1708 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1709 			return -EAGAIN;
1710 		case -NFS4ERR_DELEG_REVOKED:
1711 		case -NFS4ERR_ADMIN_REVOKED:
1712 		case -NFS4ERR_BAD_STATEID:
1713 		case -NFS4ERR_OPENMODE:
1714 			nfs_inode_find_state_and_recover(state->inode,
1715 					stateid);
1716 			nfs4_schedule_stateid_recovery(server, state);
1717 			return -EAGAIN;
1718 		case -NFS4ERR_DELAY:
1719 		case -NFS4ERR_GRACE:
1720 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1721 			ssleep(1);
1722 			return -EAGAIN;
1723 		case -ENOMEM:
1724 		case -NFS4ERR_DENIED:
1725 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1726 			return 0;
1727 	}
1728 	return err;
1729 }
1730 
1731 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1732 {
1733 	struct nfs_server *server = NFS_SERVER(state->inode);
1734 	struct nfs4_opendata *opendata;
1735 	int err;
1736 
1737 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1738 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1739 	if (IS_ERR(opendata))
1740 		return PTR_ERR(opendata);
1741 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1742 	err = nfs4_open_recover(opendata, state);
1743 	nfs4_opendata_put(opendata);
1744 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1745 }
1746 
1747 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1748 {
1749 	struct nfs4_opendata *data = calldata;
1750 
1751 	nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1752 			     &data->c_arg.seq_args, &data->c_res.seq_res, task);
1753 }
1754 
1755 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1756 {
1757 	struct nfs4_opendata *data = calldata;
1758 
1759 	nfs40_sequence_done(task, &data->c_res.seq_res);
1760 
1761 	data->rpc_status = task->tk_status;
1762 	if (data->rpc_status == 0) {
1763 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1764 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1765 		renew_lease(data->o_res.server, data->timestamp);
1766 		data->rpc_done = 1;
1767 	}
1768 }
1769 
1770 static void nfs4_open_confirm_release(void *calldata)
1771 {
1772 	struct nfs4_opendata *data = calldata;
1773 	struct nfs4_state *state = NULL;
1774 
1775 	/* If this request hasn't been cancelled, do nothing */
1776 	if (data->cancelled == 0)
1777 		goto out_free;
1778 	/* In case of error, no cleanup! */
1779 	if (!data->rpc_done)
1780 		goto out_free;
1781 	state = nfs4_opendata_to_nfs4_state(data);
1782 	if (!IS_ERR(state))
1783 		nfs4_close_state(state, data->o_arg.fmode);
1784 out_free:
1785 	nfs4_opendata_put(data);
1786 }
1787 
1788 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1789 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1790 	.rpc_call_done = nfs4_open_confirm_done,
1791 	.rpc_release = nfs4_open_confirm_release,
1792 };
1793 
1794 /*
1795  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1796  */
1797 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1798 {
1799 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1800 	struct rpc_task *task;
1801 	struct  rpc_message msg = {
1802 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1803 		.rpc_argp = &data->c_arg,
1804 		.rpc_resp = &data->c_res,
1805 		.rpc_cred = data->owner->so_cred,
1806 	};
1807 	struct rpc_task_setup task_setup_data = {
1808 		.rpc_client = server->client,
1809 		.rpc_message = &msg,
1810 		.callback_ops = &nfs4_open_confirm_ops,
1811 		.callback_data = data,
1812 		.workqueue = nfsiod_workqueue,
1813 		.flags = RPC_TASK_ASYNC,
1814 	};
1815 	int status;
1816 
1817 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1818 	kref_get(&data->kref);
1819 	data->rpc_done = 0;
1820 	data->rpc_status = 0;
1821 	data->timestamp = jiffies;
1822 	task = rpc_run_task(&task_setup_data);
1823 	if (IS_ERR(task))
1824 		return PTR_ERR(task);
1825 	status = nfs4_wait_for_completion_rpc_task(task);
1826 	if (status != 0) {
1827 		data->cancelled = 1;
1828 		smp_wmb();
1829 	} else
1830 		status = data->rpc_status;
1831 	rpc_put_task(task);
1832 	return status;
1833 }
1834 
1835 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1836 {
1837 	struct nfs4_opendata *data = calldata;
1838 	struct nfs4_state_owner *sp = data->owner;
1839 	struct nfs_client *clp = sp->so_server->nfs_client;
1840 
1841 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1842 		goto out_wait;
1843 	/*
1844 	 * Check if we still need to send an OPEN call, or if we can use
1845 	 * a delegation instead.
1846 	 */
1847 	if (data->state != NULL) {
1848 		struct nfs_delegation *delegation;
1849 
1850 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1851 			goto out_no_action;
1852 		rcu_read_lock();
1853 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1854 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1855 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1856 		    can_open_delegated(delegation, data->o_arg.fmode))
1857 			goto unlock_no_action;
1858 		rcu_read_unlock();
1859 	}
1860 	/* Update client id. */
1861 	data->o_arg.clientid = clp->cl_clientid;
1862 	switch (data->o_arg.claim) {
1863 	case NFS4_OPEN_CLAIM_PREVIOUS:
1864 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1865 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1866 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1867 	case NFS4_OPEN_CLAIM_FH:
1868 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1869 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1870 	}
1871 	data->timestamp = jiffies;
1872 	if (nfs4_setup_sequence(data->o_arg.server,
1873 				&data->o_arg.seq_args,
1874 				&data->o_res.seq_res,
1875 				task) != 0)
1876 		nfs_release_seqid(data->o_arg.seqid);
1877 
1878 	/* Set the create mode (note dependency on the session type) */
1879 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1880 	if (data->o_arg.open_flags & O_EXCL) {
1881 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1882 		if (nfs4_has_persistent_session(clp))
1883 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1884 		else if (clp->cl_mvops->minor_version > 0)
1885 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1886 	}
1887 	return;
1888 unlock_no_action:
1889 	rcu_read_unlock();
1890 out_no_action:
1891 	task->tk_action = NULL;
1892 out_wait:
1893 	nfs4_sequence_done(task, &data->o_res.seq_res);
1894 }
1895 
1896 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1897 {
1898 	struct nfs4_opendata *data = calldata;
1899 
1900 	data->rpc_status = task->tk_status;
1901 
1902 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1903 		return;
1904 
1905 	if (task->tk_status == 0) {
1906 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1907 			switch (data->o_res.f_attr->mode & S_IFMT) {
1908 			case S_IFREG:
1909 				break;
1910 			case S_IFLNK:
1911 				data->rpc_status = -ELOOP;
1912 				break;
1913 			case S_IFDIR:
1914 				data->rpc_status = -EISDIR;
1915 				break;
1916 			default:
1917 				data->rpc_status = -ENOTDIR;
1918 			}
1919 		}
1920 		renew_lease(data->o_res.server, data->timestamp);
1921 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1922 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1923 	}
1924 	data->rpc_done = 1;
1925 }
1926 
1927 static void nfs4_open_release(void *calldata)
1928 {
1929 	struct nfs4_opendata *data = calldata;
1930 	struct nfs4_state *state = NULL;
1931 
1932 	/* If this request hasn't been cancelled, do nothing */
1933 	if (data->cancelled == 0)
1934 		goto out_free;
1935 	/* In case of error, no cleanup! */
1936 	if (data->rpc_status != 0 || !data->rpc_done)
1937 		goto out_free;
1938 	/* In case we need an open_confirm, no cleanup! */
1939 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1940 		goto out_free;
1941 	state = nfs4_opendata_to_nfs4_state(data);
1942 	if (!IS_ERR(state))
1943 		nfs4_close_state(state, data->o_arg.fmode);
1944 out_free:
1945 	nfs4_opendata_put(data);
1946 }
1947 
1948 static const struct rpc_call_ops nfs4_open_ops = {
1949 	.rpc_call_prepare = nfs4_open_prepare,
1950 	.rpc_call_done = nfs4_open_done,
1951 	.rpc_release = nfs4_open_release,
1952 };
1953 
1954 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1955 {
1956 	struct inode *dir = d_inode(data->dir);
1957 	struct nfs_server *server = NFS_SERVER(dir);
1958 	struct nfs_openargs *o_arg = &data->o_arg;
1959 	struct nfs_openres *o_res = &data->o_res;
1960 	struct rpc_task *task;
1961 	struct rpc_message msg = {
1962 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1963 		.rpc_argp = o_arg,
1964 		.rpc_resp = o_res,
1965 		.rpc_cred = data->owner->so_cred,
1966 	};
1967 	struct rpc_task_setup task_setup_data = {
1968 		.rpc_client = server->client,
1969 		.rpc_message = &msg,
1970 		.callback_ops = &nfs4_open_ops,
1971 		.callback_data = data,
1972 		.workqueue = nfsiod_workqueue,
1973 		.flags = RPC_TASK_ASYNC,
1974 	};
1975 	int status;
1976 
1977 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1978 	kref_get(&data->kref);
1979 	data->rpc_done = 0;
1980 	data->rpc_status = 0;
1981 	data->cancelled = 0;
1982 	data->is_recover = 0;
1983 	if (isrecover) {
1984 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1985 		data->is_recover = 1;
1986 	}
1987 	task = rpc_run_task(&task_setup_data);
1988         if (IS_ERR(task))
1989                 return PTR_ERR(task);
1990         status = nfs4_wait_for_completion_rpc_task(task);
1991         if (status != 0) {
1992                 data->cancelled = 1;
1993                 smp_wmb();
1994         } else
1995                 status = data->rpc_status;
1996         rpc_put_task(task);
1997 
1998 	return status;
1999 }
2000 
2001 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2002 {
2003 	struct inode *dir = d_inode(data->dir);
2004 	struct nfs_openres *o_res = &data->o_res;
2005         int status;
2006 
2007 	status = nfs4_run_open_task(data, 1);
2008 	if (status != 0 || !data->rpc_done)
2009 		return status;
2010 
2011 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2012 
2013 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2014 		status = _nfs4_proc_open_confirm(data);
2015 		if (status != 0)
2016 			return status;
2017 	}
2018 
2019 	return status;
2020 }
2021 
2022 /*
2023  * Additional permission checks in order to distinguish between an
2024  * open for read, and an open for execute. This works around the
2025  * fact that NFSv4 OPEN treats read and execute permissions as being
2026  * the same.
2027  * Note that in the non-execute case, we want to turn off permission
2028  * checking if we just created a new file (POSIX open() semantics).
2029  */
2030 static int nfs4_opendata_access(struct rpc_cred *cred,
2031 				struct nfs4_opendata *opendata,
2032 				struct nfs4_state *state, fmode_t fmode,
2033 				int openflags)
2034 {
2035 	struct nfs_access_entry cache;
2036 	u32 mask;
2037 
2038 	/* access call failed or for some reason the server doesn't
2039 	 * support any access modes -- defer access call until later */
2040 	if (opendata->o_res.access_supported == 0)
2041 		return 0;
2042 
2043 	mask = 0;
2044 	/*
2045 	 * Use openflags to check for exec, because fmode won't
2046 	 * always have FMODE_EXEC set when file open for exec.
2047 	 */
2048 	if (openflags & __FMODE_EXEC) {
2049 		/* ONLY check for exec rights */
2050 		mask = MAY_EXEC;
2051 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2052 		mask = MAY_READ;
2053 
2054 	cache.cred = cred;
2055 	cache.jiffies = jiffies;
2056 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2057 	nfs_access_add_cache(state->inode, &cache);
2058 
2059 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2060 		return 0;
2061 
2062 	/* even though OPEN succeeded, access is denied. Close the file */
2063 	nfs4_close_state(state, fmode);
2064 	return -EACCES;
2065 }
2066 
2067 /*
2068  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2069  */
2070 static int _nfs4_proc_open(struct nfs4_opendata *data)
2071 {
2072 	struct inode *dir = d_inode(data->dir);
2073 	struct nfs_server *server = NFS_SERVER(dir);
2074 	struct nfs_openargs *o_arg = &data->o_arg;
2075 	struct nfs_openres *o_res = &data->o_res;
2076 	int status;
2077 
2078 	status = nfs4_run_open_task(data, 0);
2079 	if (!data->rpc_done)
2080 		return status;
2081 	if (status != 0) {
2082 		if (status == -NFS4ERR_BADNAME &&
2083 				!(o_arg->open_flags & O_CREAT))
2084 			return -ENOENT;
2085 		return status;
2086 	}
2087 
2088 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2089 
2090 	if (o_arg->open_flags & O_CREAT) {
2091 		update_changeattr(dir, &o_res->cinfo);
2092 		if (o_arg->open_flags & O_EXCL)
2093 			data->file_created = 1;
2094 		else if (o_res->cinfo.before != o_res->cinfo.after)
2095 			data->file_created = 1;
2096 	}
2097 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2098 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2099 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2100 		status = _nfs4_proc_open_confirm(data);
2101 		if (status != 0)
2102 			return status;
2103 	}
2104 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2105 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2106 	return 0;
2107 }
2108 
2109 static int nfs4_recover_expired_lease(struct nfs_server *server)
2110 {
2111 	return nfs4_client_recover_expired_lease(server->nfs_client);
2112 }
2113 
2114 /*
2115  * OPEN_EXPIRED:
2116  * 	reclaim state on the server after a network partition.
2117  * 	Assumes caller holds the appropriate lock
2118  */
2119 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2120 {
2121 	struct nfs4_opendata *opendata;
2122 	int ret;
2123 
2124 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2125 			NFS4_OPEN_CLAIM_FH);
2126 	if (IS_ERR(opendata))
2127 		return PTR_ERR(opendata);
2128 	ret = nfs4_open_recover(opendata, state);
2129 	if (ret == -ESTALE)
2130 		d_drop(ctx->dentry);
2131 	nfs4_opendata_put(opendata);
2132 	return ret;
2133 }
2134 
2135 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2136 {
2137 	struct nfs_server *server = NFS_SERVER(state->inode);
2138 	struct nfs4_exception exception = { };
2139 	int err;
2140 
2141 	do {
2142 		err = _nfs4_open_expired(ctx, state);
2143 		trace_nfs4_open_expired(ctx, 0, err);
2144 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2145 			continue;
2146 		switch (err) {
2147 		default:
2148 			goto out;
2149 		case -NFS4ERR_GRACE:
2150 		case -NFS4ERR_DELAY:
2151 			nfs4_handle_exception(server, err, &exception);
2152 			err = 0;
2153 		}
2154 	} while (exception.retry);
2155 out:
2156 	return err;
2157 }
2158 
2159 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2160 {
2161 	struct nfs_open_context *ctx;
2162 	int ret;
2163 
2164 	ctx = nfs4_state_find_open_context(state);
2165 	if (IS_ERR(ctx))
2166 		return -EAGAIN;
2167 	ret = nfs4_do_open_expired(ctx, state);
2168 	put_nfs_open_context(ctx);
2169 	return ret;
2170 }
2171 
2172 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2173 {
2174 	nfs_remove_bad_delegation(state->inode);
2175 	write_seqlock(&state->seqlock);
2176 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2177 	write_sequnlock(&state->seqlock);
2178 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2179 }
2180 
2181 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2182 {
2183 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2184 		nfs_finish_clear_delegation_stateid(state);
2185 }
2186 
2187 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2188 {
2189 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2190 	nfs40_clear_delegation_stateid(state);
2191 	return nfs4_open_expired(sp, state);
2192 }
2193 
2194 #if defined(CONFIG_NFS_V4_1)
2195 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2196 {
2197 	struct nfs_server *server = NFS_SERVER(state->inode);
2198 	nfs4_stateid stateid;
2199 	struct nfs_delegation *delegation;
2200 	struct rpc_cred *cred;
2201 	int status;
2202 
2203 	/* Get the delegation credential for use by test/free_stateid */
2204 	rcu_read_lock();
2205 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2206 	if (delegation == NULL) {
2207 		rcu_read_unlock();
2208 		return;
2209 	}
2210 
2211 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2212 	cred = get_rpccred(delegation->cred);
2213 	rcu_read_unlock();
2214 	status = nfs41_test_stateid(server, &stateid, cred);
2215 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2216 
2217 	if (status != NFS_OK) {
2218 		/* Free the stateid unless the server explicitly
2219 		 * informs us the stateid is unrecognized. */
2220 		if (status != -NFS4ERR_BAD_STATEID)
2221 			nfs41_free_stateid(server, &stateid, cred);
2222 		nfs_finish_clear_delegation_stateid(state);
2223 	}
2224 
2225 	put_rpccred(cred);
2226 }
2227 
2228 /**
2229  * nfs41_check_open_stateid - possibly free an open stateid
2230  *
2231  * @state: NFSv4 state for an inode
2232  *
2233  * Returns NFS_OK if recovery for this stateid is now finished.
2234  * Otherwise a negative NFS4ERR value is returned.
2235  */
2236 static int nfs41_check_open_stateid(struct nfs4_state *state)
2237 {
2238 	struct nfs_server *server = NFS_SERVER(state->inode);
2239 	nfs4_stateid *stateid = &state->open_stateid;
2240 	struct rpc_cred *cred = state->owner->so_cred;
2241 	int status;
2242 
2243 	/* If a state reset has been done, test_stateid is unneeded */
2244 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2245 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2246 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2247 		return -NFS4ERR_BAD_STATEID;
2248 
2249 	status = nfs41_test_stateid(server, stateid, cred);
2250 	trace_nfs4_test_open_stateid(state, NULL, status);
2251 	if (status != NFS_OK) {
2252 		/* Free the stateid unless the server explicitly
2253 		 * informs us the stateid is unrecognized. */
2254 		if (status != -NFS4ERR_BAD_STATEID)
2255 			nfs41_free_stateid(server, stateid, cred);
2256 
2257 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2258 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2259 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2260 		clear_bit(NFS_OPEN_STATE, &state->flags);
2261 	}
2262 	return status;
2263 }
2264 
2265 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2266 {
2267 	int status;
2268 
2269 	nfs41_check_delegation_stateid(state);
2270 	status = nfs41_check_open_stateid(state);
2271 	if (status != NFS_OK)
2272 		status = nfs4_open_expired(sp, state);
2273 	return status;
2274 }
2275 #endif
2276 
2277 /*
2278  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2279  * fields corresponding to attributes that were used to store the verifier.
2280  * Make sure we clobber those fields in the later setattr call
2281  */
2282 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2283 {
2284 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2285 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2286 		sattr->ia_valid |= ATTR_ATIME;
2287 
2288 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2289 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2290 		sattr->ia_valid |= ATTR_MTIME;
2291 }
2292 
2293 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2294 		fmode_t fmode,
2295 		int flags,
2296 		struct nfs_open_context *ctx)
2297 {
2298 	struct nfs4_state_owner *sp = opendata->owner;
2299 	struct nfs_server *server = sp->so_server;
2300 	struct dentry *dentry;
2301 	struct nfs4_state *state;
2302 	unsigned int seq;
2303 	int ret;
2304 
2305 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2306 
2307 	ret = _nfs4_proc_open(opendata);
2308 	if (ret != 0)
2309 		goto out;
2310 
2311 	state = nfs4_opendata_to_nfs4_state(opendata);
2312 	ret = PTR_ERR(state);
2313 	if (IS_ERR(state))
2314 		goto out;
2315 	if (server->caps & NFS_CAP_POSIX_LOCK)
2316 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2317 
2318 	dentry = opendata->dentry;
2319 	if (d_really_is_negative(dentry)) {
2320 		/* FIXME: Is this d_drop() ever needed? */
2321 		d_drop(dentry);
2322 		dentry = d_add_unique(dentry, igrab(state->inode));
2323 		if (dentry == NULL) {
2324 			dentry = opendata->dentry;
2325 		} else if (dentry != ctx->dentry) {
2326 			dput(ctx->dentry);
2327 			ctx->dentry = dget(dentry);
2328 		}
2329 		nfs_set_verifier(dentry,
2330 				nfs_save_change_attribute(d_inode(opendata->dir)));
2331 	}
2332 
2333 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2334 	if (ret != 0)
2335 		goto out;
2336 
2337 	ctx->state = state;
2338 	if (d_inode(dentry) == state->inode) {
2339 		nfs_inode_attach_open_context(ctx);
2340 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2341 			nfs4_schedule_stateid_recovery(server, state);
2342 	}
2343 out:
2344 	return ret;
2345 }
2346 
2347 /*
2348  * Returns a referenced nfs4_state
2349  */
2350 static int _nfs4_do_open(struct inode *dir,
2351 			struct nfs_open_context *ctx,
2352 			int flags,
2353 			struct iattr *sattr,
2354 			struct nfs4_label *label,
2355 			int *opened)
2356 {
2357 	struct nfs4_state_owner  *sp;
2358 	struct nfs4_state     *state = NULL;
2359 	struct nfs_server       *server = NFS_SERVER(dir);
2360 	struct nfs4_opendata *opendata;
2361 	struct dentry *dentry = ctx->dentry;
2362 	struct rpc_cred *cred = ctx->cred;
2363 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2364 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2365 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2366 	struct nfs4_label *olabel = NULL;
2367 	int status;
2368 
2369 	/* Protect against reboot recovery conflicts */
2370 	status = -ENOMEM;
2371 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2372 	if (sp == NULL) {
2373 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2374 		goto out_err;
2375 	}
2376 	status = nfs4_recover_expired_lease(server);
2377 	if (status != 0)
2378 		goto err_put_state_owner;
2379 	if (d_really_is_positive(dentry))
2380 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2381 	status = -ENOMEM;
2382 	if (d_really_is_positive(dentry))
2383 		claim = NFS4_OPEN_CLAIM_FH;
2384 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2385 			label, claim, GFP_KERNEL);
2386 	if (opendata == NULL)
2387 		goto err_put_state_owner;
2388 
2389 	if (label) {
2390 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2391 		if (IS_ERR(olabel)) {
2392 			status = PTR_ERR(olabel);
2393 			goto err_opendata_put;
2394 		}
2395 	}
2396 
2397 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2398 		if (!opendata->f_attr.mdsthreshold) {
2399 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2400 			if (!opendata->f_attr.mdsthreshold)
2401 				goto err_free_label;
2402 		}
2403 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2404 	}
2405 	if (d_really_is_positive(dentry))
2406 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2407 
2408 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2409 	if (status != 0)
2410 		goto err_free_label;
2411 	state = ctx->state;
2412 
2413 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2414 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2415 		nfs4_exclusive_attrset(opendata, sattr);
2416 
2417 		nfs_fattr_init(opendata->o_res.f_attr);
2418 		status = nfs4_do_setattr(state->inode, cred,
2419 				opendata->o_res.f_attr, sattr,
2420 				state, label, olabel);
2421 		if (status == 0) {
2422 			nfs_setattr_update_inode(state->inode, sattr,
2423 					opendata->o_res.f_attr);
2424 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2425 		}
2426 	}
2427 	if (opendata->file_created)
2428 		*opened |= FILE_CREATED;
2429 
2430 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2431 		*ctx_th = opendata->f_attr.mdsthreshold;
2432 		opendata->f_attr.mdsthreshold = NULL;
2433 	}
2434 
2435 	nfs4_label_free(olabel);
2436 
2437 	nfs4_opendata_put(opendata);
2438 	nfs4_put_state_owner(sp);
2439 	return 0;
2440 err_free_label:
2441 	nfs4_label_free(olabel);
2442 err_opendata_put:
2443 	nfs4_opendata_put(opendata);
2444 err_put_state_owner:
2445 	nfs4_put_state_owner(sp);
2446 out_err:
2447 	return status;
2448 }
2449 
2450 
2451 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2452 					struct nfs_open_context *ctx,
2453 					int flags,
2454 					struct iattr *sattr,
2455 					struct nfs4_label *label,
2456 					int *opened)
2457 {
2458 	struct nfs_server *server = NFS_SERVER(dir);
2459 	struct nfs4_exception exception = { };
2460 	struct nfs4_state *res;
2461 	int status;
2462 
2463 	do {
2464 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2465 		res = ctx->state;
2466 		trace_nfs4_open_file(ctx, flags, status);
2467 		if (status == 0)
2468 			break;
2469 		/* NOTE: BAD_SEQID means the server and client disagree about the
2470 		 * book-keeping w.r.t. state-changing operations
2471 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2472 		 * It is actually a sign of a bug on the client or on the server.
2473 		 *
2474 		 * If we receive a BAD_SEQID error in the particular case of
2475 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2476 		 * have unhashed the old state_owner for us, and that we can
2477 		 * therefore safely retry using a new one. We should still warn
2478 		 * the user though...
2479 		 */
2480 		if (status == -NFS4ERR_BAD_SEQID) {
2481 			pr_warn_ratelimited("NFS: v4 server %s "
2482 					" returned a bad sequence-id error!\n",
2483 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2484 			exception.retry = 1;
2485 			continue;
2486 		}
2487 		/*
2488 		 * BAD_STATEID on OPEN means that the server cancelled our
2489 		 * state before it received the OPEN_CONFIRM.
2490 		 * Recover by retrying the request as per the discussion
2491 		 * on Page 181 of RFC3530.
2492 		 */
2493 		if (status == -NFS4ERR_BAD_STATEID) {
2494 			exception.retry = 1;
2495 			continue;
2496 		}
2497 		if (status == -EAGAIN) {
2498 			/* We must have found a delegation */
2499 			exception.retry = 1;
2500 			continue;
2501 		}
2502 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2503 			continue;
2504 		res = ERR_PTR(nfs4_handle_exception(server,
2505 					status, &exception));
2506 	} while (exception.retry);
2507 	return res;
2508 }
2509 
2510 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2511 			    struct nfs_fattr *fattr, struct iattr *sattr,
2512 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2513 			    struct nfs4_label *olabel)
2514 {
2515 	struct nfs_server *server = NFS_SERVER(inode);
2516         struct nfs_setattrargs  arg = {
2517                 .fh             = NFS_FH(inode),
2518                 .iap            = sattr,
2519 		.server		= server,
2520 		.bitmask = server->attr_bitmask,
2521 		.label		= ilabel,
2522         };
2523         struct nfs_setattrres  res = {
2524 		.fattr		= fattr,
2525 		.label		= olabel,
2526 		.server		= server,
2527         };
2528         struct rpc_message msg = {
2529 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2530 		.rpc_argp	= &arg,
2531 		.rpc_resp	= &res,
2532 		.rpc_cred	= cred,
2533         };
2534 	unsigned long timestamp = jiffies;
2535 	fmode_t fmode;
2536 	bool truncate;
2537 	int status;
2538 
2539 	arg.bitmask = nfs4_bitmask(server, ilabel);
2540 	if (ilabel)
2541 		arg.bitmask = nfs4_bitmask(server, olabel);
2542 
2543 	nfs_fattr_init(fattr);
2544 
2545 	/* Servers should only apply open mode checks for file size changes */
2546 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2547 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2548 
2549 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2550 		/* Use that stateid */
2551 	} else if (truncate && state != NULL) {
2552 		struct nfs_lockowner lockowner = {
2553 			.l_owner = current->files,
2554 			.l_pid = current->tgid,
2555 		};
2556 		if (!nfs4_valid_open_stateid(state))
2557 			return -EBADF;
2558 		if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2559 				&lockowner) == -EIO)
2560 			return -EBADF;
2561 	} else
2562 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2563 
2564 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2565 	if (status == 0 && state != NULL)
2566 		renew_lease(server, timestamp);
2567 	return status;
2568 }
2569 
2570 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2571 			   struct nfs_fattr *fattr, struct iattr *sattr,
2572 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2573 			   struct nfs4_label *olabel)
2574 {
2575 	struct nfs_server *server = NFS_SERVER(inode);
2576 	struct nfs4_exception exception = {
2577 		.state = state,
2578 		.inode = inode,
2579 	};
2580 	int err;
2581 	do {
2582 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2583 		trace_nfs4_setattr(inode, err);
2584 		switch (err) {
2585 		case -NFS4ERR_OPENMODE:
2586 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2587 				pr_warn_once("NFSv4: server %s is incorrectly "
2588 						"applying open mode checks to "
2589 						"a SETATTR that is not "
2590 						"changing file size.\n",
2591 						server->nfs_client->cl_hostname);
2592 			}
2593 			if (state && !(state->state & FMODE_WRITE)) {
2594 				err = -EBADF;
2595 				if (sattr->ia_valid & ATTR_OPEN)
2596 					err = -EACCES;
2597 				goto out;
2598 			}
2599 		}
2600 		err = nfs4_handle_exception(server, err, &exception);
2601 	} while (exception.retry);
2602 out:
2603 	return err;
2604 }
2605 
2606 struct nfs4_closedata {
2607 	struct inode *inode;
2608 	struct nfs4_state *state;
2609 	struct nfs_closeargs arg;
2610 	struct nfs_closeres res;
2611 	struct nfs_fattr fattr;
2612 	unsigned long timestamp;
2613 	bool roc;
2614 	u32 roc_barrier;
2615 };
2616 
2617 static void nfs4_free_closedata(void *data)
2618 {
2619 	struct nfs4_closedata *calldata = data;
2620 	struct nfs4_state_owner *sp = calldata->state->owner;
2621 	struct super_block *sb = calldata->state->inode->i_sb;
2622 
2623 	if (calldata->roc)
2624 		pnfs_roc_release(calldata->state->inode);
2625 	nfs4_put_open_state(calldata->state);
2626 	nfs_free_seqid(calldata->arg.seqid);
2627 	nfs4_put_state_owner(sp);
2628 	nfs_sb_deactive(sb);
2629 	kfree(calldata);
2630 }
2631 
2632 static void nfs4_close_done(struct rpc_task *task, void *data)
2633 {
2634 	struct nfs4_closedata *calldata = data;
2635 	struct nfs4_state *state = calldata->state;
2636 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2637 	nfs4_stateid *res_stateid = NULL;
2638 
2639 	dprintk("%s: begin!\n", __func__);
2640 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2641 		return;
2642 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2643         /* hmm. we are done with the inode, and in the process of freeing
2644 	 * the state_owner. we keep this around to process errors
2645 	 */
2646 	switch (task->tk_status) {
2647 		case 0:
2648 			res_stateid = &calldata->res.stateid;
2649 			if (calldata->arg.fmode == 0 && calldata->roc)
2650 				pnfs_roc_set_barrier(state->inode,
2651 						     calldata->roc_barrier);
2652 			renew_lease(server, calldata->timestamp);
2653 			break;
2654 		case -NFS4ERR_ADMIN_REVOKED:
2655 		case -NFS4ERR_STALE_STATEID:
2656 		case -NFS4ERR_OLD_STATEID:
2657 		case -NFS4ERR_BAD_STATEID:
2658 		case -NFS4ERR_EXPIRED:
2659 			if (!nfs4_stateid_match(&calldata->arg.stateid,
2660 						&state->open_stateid)) {
2661 				rpc_restart_call_prepare(task);
2662 				goto out_release;
2663 			}
2664 			if (calldata->arg.fmode == 0)
2665 				break;
2666 		default:
2667 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2668 				rpc_restart_call_prepare(task);
2669 				goto out_release;
2670 			}
2671 	}
2672 	nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2673 out_release:
2674 	nfs_release_seqid(calldata->arg.seqid);
2675 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2676 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2677 }
2678 
2679 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2680 {
2681 	struct nfs4_closedata *calldata = data;
2682 	struct nfs4_state *state = calldata->state;
2683 	struct inode *inode = calldata->inode;
2684 	bool is_rdonly, is_wronly, is_rdwr;
2685 	int call_close = 0;
2686 
2687 	dprintk("%s: begin!\n", __func__);
2688 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2689 		goto out_wait;
2690 
2691 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2692 	spin_lock(&state->owner->so_lock);
2693 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2694 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2695 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2696 	nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2697 	/* Calculate the change in open mode */
2698 	calldata->arg.fmode = 0;
2699 	if (state->n_rdwr == 0) {
2700 		if (state->n_rdonly == 0)
2701 			call_close |= is_rdonly;
2702 		else if (is_rdonly)
2703 			calldata->arg.fmode |= FMODE_READ;
2704 		if (state->n_wronly == 0)
2705 			call_close |= is_wronly;
2706 		else if (is_wronly)
2707 			calldata->arg.fmode |= FMODE_WRITE;
2708 	} else if (is_rdwr)
2709 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2710 
2711 	if (calldata->arg.fmode == 0)
2712 		call_close |= is_rdwr;
2713 
2714 	if (!nfs4_valid_open_stateid(state))
2715 		call_close = 0;
2716 	spin_unlock(&state->owner->so_lock);
2717 
2718 	if (!call_close) {
2719 		/* Note: exit _without_ calling nfs4_close_done */
2720 		goto out_no_action;
2721 	}
2722 
2723 	if (calldata->arg.fmode == 0) {
2724 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2725 		if (calldata->roc &&
2726 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2727 			nfs_release_seqid(calldata->arg.seqid);
2728 			goto out_wait;
2729 		    }
2730 	}
2731 	calldata->arg.share_access =
2732 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
2733 				calldata->arg.fmode, 0);
2734 
2735 	nfs_fattr_init(calldata->res.fattr);
2736 	calldata->timestamp = jiffies;
2737 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2738 				&calldata->arg.seq_args,
2739 				&calldata->res.seq_res,
2740 				task) != 0)
2741 		nfs_release_seqid(calldata->arg.seqid);
2742 	dprintk("%s: done!\n", __func__);
2743 	return;
2744 out_no_action:
2745 	task->tk_action = NULL;
2746 out_wait:
2747 	nfs4_sequence_done(task, &calldata->res.seq_res);
2748 }
2749 
2750 static const struct rpc_call_ops nfs4_close_ops = {
2751 	.rpc_call_prepare = nfs4_close_prepare,
2752 	.rpc_call_done = nfs4_close_done,
2753 	.rpc_release = nfs4_free_closedata,
2754 };
2755 
2756 static bool nfs4_roc(struct inode *inode)
2757 {
2758 	if (!nfs_have_layout(inode))
2759 		return false;
2760 	return pnfs_roc(inode);
2761 }
2762 
2763 /*
2764  * It is possible for data to be read/written from a mem-mapped file
2765  * after the sys_close call (which hits the vfs layer as a flush).
2766  * This means that we can't safely call nfsv4 close on a file until
2767  * the inode is cleared. This in turn means that we are not good
2768  * NFSv4 citizens - we do not indicate to the server to update the file's
2769  * share state even when we are done with one of the three share
2770  * stateid's in the inode.
2771  *
2772  * NOTE: Caller must be holding the sp->so_owner semaphore!
2773  */
2774 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2775 {
2776 	struct nfs_server *server = NFS_SERVER(state->inode);
2777 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2778 	struct nfs4_closedata *calldata;
2779 	struct nfs4_state_owner *sp = state->owner;
2780 	struct rpc_task *task;
2781 	struct rpc_message msg = {
2782 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2783 		.rpc_cred = state->owner->so_cred,
2784 	};
2785 	struct rpc_task_setup task_setup_data = {
2786 		.rpc_client = server->client,
2787 		.rpc_message = &msg,
2788 		.callback_ops = &nfs4_close_ops,
2789 		.workqueue = nfsiod_workqueue,
2790 		.flags = RPC_TASK_ASYNC,
2791 	};
2792 	int status = -ENOMEM;
2793 
2794 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2795 		&task_setup_data.rpc_client, &msg);
2796 
2797 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2798 	if (calldata == NULL)
2799 		goto out;
2800 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2801 	calldata->inode = state->inode;
2802 	calldata->state = state;
2803 	calldata->arg.fh = NFS_FH(state->inode);
2804 	/* Serialization for the sequence id */
2805 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2806 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2807 	if (IS_ERR(calldata->arg.seqid))
2808 		goto out_free_calldata;
2809 	calldata->arg.fmode = 0;
2810 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2811 	calldata->res.fattr = &calldata->fattr;
2812 	calldata->res.seqid = calldata->arg.seqid;
2813 	calldata->res.server = server;
2814 	calldata->roc = nfs4_roc(state->inode);
2815 	nfs_sb_active(calldata->inode->i_sb);
2816 
2817 	msg.rpc_argp = &calldata->arg;
2818 	msg.rpc_resp = &calldata->res;
2819 	task_setup_data.callback_data = calldata;
2820 	task = rpc_run_task(&task_setup_data);
2821 	if (IS_ERR(task))
2822 		return PTR_ERR(task);
2823 	status = 0;
2824 	if (wait)
2825 		status = rpc_wait_for_completion_task(task);
2826 	rpc_put_task(task);
2827 	return status;
2828 out_free_calldata:
2829 	kfree(calldata);
2830 out:
2831 	nfs4_put_open_state(state);
2832 	nfs4_put_state_owner(sp);
2833 	return status;
2834 }
2835 
2836 static struct inode *
2837 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2838 		int open_flags, struct iattr *attr, int *opened)
2839 {
2840 	struct nfs4_state *state;
2841 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2842 
2843 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2844 
2845 	/* Protect against concurrent sillydeletes */
2846 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2847 
2848 	nfs4_label_release_security(label);
2849 
2850 	if (IS_ERR(state))
2851 		return ERR_CAST(state);
2852 	return state->inode;
2853 }
2854 
2855 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2856 {
2857 	if (ctx->state == NULL)
2858 		return;
2859 	if (is_sync)
2860 		nfs4_close_sync(ctx->state, ctx->mode);
2861 	else
2862 		nfs4_close_state(ctx->state, ctx->mode);
2863 }
2864 
2865 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2866 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2867 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2868 
2869 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2870 {
2871 	struct nfs4_server_caps_arg args = {
2872 		.fhandle = fhandle,
2873 	};
2874 	struct nfs4_server_caps_res res = {};
2875 	struct rpc_message msg = {
2876 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2877 		.rpc_argp = &args,
2878 		.rpc_resp = &res,
2879 	};
2880 	int status;
2881 
2882 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2883 	if (status == 0) {
2884 		/* Sanity check the server answers */
2885 		switch (server->nfs_client->cl_minorversion) {
2886 		case 0:
2887 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2888 			res.attr_bitmask[2] = 0;
2889 			break;
2890 		case 1:
2891 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2892 			break;
2893 		case 2:
2894 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2895 		}
2896 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2897 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2898 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2899 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2900 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2901 				NFS_CAP_CTIME|NFS_CAP_MTIME|
2902 				NFS_CAP_SECURITY_LABEL);
2903 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2904 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2905 			server->caps |= NFS_CAP_ACLS;
2906 		if (res.has_links != 0)
2907 			server->caps |= NFS_CAP_HARDLINKS;
2908 		if (res.has_symlinks != 0)
2909 			server->caps |= NFS_CAP_SYMLINKS;
2910 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2911 			server->caps |= NFS_CAP_FILEID;
2912 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2913 			server->caps |= NFS_CAP_MODE;
2914 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2915 			server->caps |= NFS_CAP_NLINK;
2916 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2917 			server->caps |= NFS_CAP_OWNER;
2918 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2919 			server->caps |= NFS_CAP_OWNER_GROUP;
2920 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2921 			server->caps |= NFS_CAP_ATIME;
2922 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2923 			server->caps |= NFS_CAP_CTIME;
2924 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2925 			server->caps |= NFS_CAP_MTIME;
2926 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2927 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2928 			server->caps |= NFS_CAP_SECURITY_LABEL;
2929 #endif
2930 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2931 				sizeof(server->attr_bitmask));
2932 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2933 
2934 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2935 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2936 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2937 		server->cache_consistency_bitmask[2] = 0;
2938 		server->acl_bitmask = res.acl_bitmask;
2939 		server->fh_expire_type = res.fh_expire_type;
2940 	}
2941 
2942 	return status;
2943 }
2944 
2945 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2946 {
2947 	struct nfs4_exception exception = { };
2948 	int err;
2949 	do {
2950 		err = nfs4_handle_exception(server,
2951 				_nfs4_server_capabilities(server, fhandle),
2952 				&exception);
2953 	} while (exception.retry);
2954 	return err;
2955 }
2956 
2957 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2958 		struct nfs_fsinfo *info)
2959 {
2960 	u32 bitmask[3];
2961 	struct nfs4_lookup_root_arg args = {
2962 		.bitmask = bitmask,
2963 	};
2964 	struct nfs4_lookup_res res = {
2965 		.server = server,
2966 		.fattr = info->fattr,
2967 		.fh = fhandle,
2968 	};
2969 	struct rpc_message msg = {
2970 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2971 		.rpc_argp = &args,
2972 		.rpc_resp = &res,
2973 	};
2974 
2975 	bitmask[0] = nfs4_fattr_bitmap[0];
2976 	bitmask[1] = nfs4_fattr_bitmap[1];
2977 	/*
2978 	 * Process the label in the upcoming getfattr
2979 	 */
2980 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2981 
2982 	nfs_fattr_init(info->fattr);
2983 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2984 }
2985 
2986 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2987 		struct nfs_fsinfo *info)
2988 {
2989 	struct nfs4_exception exception = { };
2990 	int err;
2991 	do {
2992 		err = _nfs4_lookup_root(server, fhandle, info);
2993 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2994 		switch (err) {
2995 		case 0:
2996 		case -NFS4ERR_WRONGSEC:
2997 			goto out;
2998 		default:
2999 			err = nfs4_handle_exception(server, err, &exception);
3000 		}
3001 	} while (exception.retry);
3002 out:
3003 	return err;
3004 }
3005 
3006 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3007 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3008 {
3009 	struct rpc_auth_create_args auth_args = {
3010 		.pseudoflavor = flavor,
3011 	};
3012 	struct rpc_auth *auth;
3013 	int ret;
3014 
3015 	auth = rpcauth_create(&auth_args, server->client);
3016 	if (IS_ERR(auth)) {
3017 		ret = -EACCES;
3018 		goto out;
3019 	}
3020 	ret = nfs4_lookup_root(server, fhandle, info);
3021 out:
3022 	return ret;
3023 }
3024 
3025 /*
3026  * Retry pseudoroot lookup with various security flavors.  We do this when:
3027  *
3028  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3029  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3030  *
3031  * Returns zero on success, or a negative NFS4ERR value, or a
3032  * negative errno value.
3033  */
3034 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3035 			      struct nfs_fsinfo *info)
3036 {
3037 	/* Per 3530bis 15.33.5 */
3038 	static const rpc_authflavor_t flav_array[] = {
3039 		RPC_AUTH_GSS_KRB5P,
3040 		RPC_AUTH_GSS_KRB5I,
3041 		RPC_AUTH_GSS_KRB5,
3042 		RPC_AUTH_UNIX,			/* courtesy */
3043 		RPC_AUTH_NULL,
3044 	};
3045 	int status = -EPERM;
3046 	size_t i;
3047 
3048 	if (server->auth_info.flavor_len > 0) {
3049 		/* try each flavor specified by user */
3050 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3051 			status = nfs4_lookup_root_sec(server, fhandle, info,
3052 						server->auth_info.flavors[i]);
3053 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3054 				continue;
3055 			break;
3056 		}
3057 	} else {
3058 		/* no flavors specified by user, try default list */
3059 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3060 			status = nfs4_lookup_root_sec(server, fhandle, info,
3061 						      flav_array[i]);
3062 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3063 				continue;
3064 			break;
3065 		}
3066 	}
3067 
3068 	/*
3069 	 * -EACCESS could mean that the user doesn't have correct permissions
3070 	 * to access the mount.  It could also mean that we tried to mount
3071 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3072 	 * existing mount programs don't handle -EACCES very well so it should
3073 	 * be mapped to -EPERM instead.
3074 	 */
3075 	if (status == -EACCES)
3076 		status = -EPERM;
3077 	return status;
3078 }
3079 
3080 static int nfs4_do_find_root_sec(struct nfs_server *server,
3081 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3082 {
3083 	int mv = server->nfs_client->cl_minorversion;
3084 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3085 }
3086 
3087 /**
3088  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3089  * @server: initialized nfs_server handle
3090  * @fhandle: we fill in the pseudo-fs root file handle
3091  * @info: we fill in an FSINFO struct
3092  * @auth_probe: probe the auth flavours
3093  *
3094  * Returns zero on success, or a negative errno.
3095  */
3096 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3097 			 struct nfs_fsinfo *info,
3098 			 bool auth_probe)
3099 {
3100 	int status = 0;
3101 
3102 	if (!auth_probe)
3103 		status = nfs4_lookup_root(server, fhandle, info);
3104 
3105 	if (auth_probe || status == NFS4ERR_WRONGSEC)
3106 		status = nfs4_do_find_root_sec(server, fhandle, info);
3107 
3108 	if (status == 0)
3109 		status = nfs4_server_capabilities(server, fhandle);
3110 	if (status == 0)
3111 		status = nfs4_do_fsinfo(server, fhandle, info);
3112 
3113 	return nfs4_map_errors(status);
3114 }
3115 
3116 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3117 			      struct nfs_fsinfo *info)
3118 {
3119 	int error;
3120 	struct nfs_fattr *fattr = info->fattr;
3121 	struct nfs4_label *label = NULL;
3122 
3123 	error = nfs4_server_capabilities(server, mntfh);
3124 	if (error < 0) {
3125 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3126 		return error;
3127 	}
3128 
3129 	label = nfs4_label_alloc(server, GFP_KERNEL);
3130 	if (IS_ERR(label))
3131 		return PTR_ERR(label);
3132 
3133 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3134 	if (error < 0) {
3135 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3136 		goto err_free_label;
3137 	}
3138 
3139 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3140 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3141 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3142 
3143 err_free_label:
3144 	nfs4_label_free(label);
3145 
3146 	return error;
3147 }
3148 
3149 /*
3150  * Get locations and (maybe) other attributes of a referral.
3151  * Note that we'll actually follow the referral later when
3152  * we detect fsid mismatch in inode revalidation
3153  */
3154 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3155 			     const struct qstr *name, struct nfs_fattr *fattr,
3156 			     struct nfs_fh *fhandle)
3157 {
3158 	int status = -ENOMEM;
3159 	struct page *page = NULL;
3160 	struct nfs4_fs_locations *locations = NULL;
3161 
3162 	page = alloc_page(GFP_KERNEL);
3163 	if (page == NULL)
3164 		goto out;
3165 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3166 	if (locations == NULL)
3167 		goto out;
3168 
3169 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3170 	if (status != 0)
3171 		goto out;
3172 
3173 	/*
3174 	 * If the fsid didn't change, this is a migration event, not a
3175 	 * referral.  Cause us to drop into the exception handler, which
3176 	 * will kick off migration recovery.
3177 	 */
3178 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3179 		dprintk("%s: server did not return a different fsid for"
3180 			" a referral at %s\n", __func__, name->name);
3181 		status = -NFS4ERR_MOVED;
3182 		goto out;
3183 	}
3184 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3185 	nfs_fixup_referral_attributes(&locations->fattr);
3186 
3187 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3188 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3189 	memset(fhandle, 0, sizeof(struct nfs_fh));
3190 out:
3191 	if (page)
3192 		__free_page(page);
3193 	kfree(locations);
3194 	return status;
3195 }
3196 
3197 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3198 				struct nfs_fattr *fattr, struct nfs4_label *label)
3199 {
3200 	struct nfs4_getattr_arg args = {
3201 		.fh = fhandle,
3202 		.bitmask = server->attr_bitmask,
3203 	};
3204 	struct nfs4_getattr_res res = {
3205 		.fattr = fattr,
3206 		.label = label,
3207 		.server = server,
3208 	};
3209 	struct rpc_message msg = {
3210 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3211 		.rpc_argp = &args,
3212 		.rpc_resp = &res,
3213 	};
3214 
3215 	args.bitmask = nfs4_bitmask(server, label);
3216 
3217 	nfs_fattr_init(fattr);
3218 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3219 }
3220 
3221 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3222 				struct nfs_fattr *fattr, struct nfs4_label *label)
3223 {
3224 	struct nfs4_exception exception = { };
3225 	int err;
3226 	do {
3227 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3228 		trace_nfs4_getattr(server, fhandle, fattr, err);
3229 		err = nfs4_handle_exception(server, err,
3230 				&exception);
3231 	} while (exception.retry);
3232 	return err;
3233 }
3234 
3235 /*
3236  * The file is not closed if it is opened due to the a request to change
3237  * the size of the file. The open call will not be needed once the
3238  * VFS layer lookup-intents are implemented.
3239  *
3240  * Close is called when the inode is destroyed.
3241  * If we haven't opened the file for O_WRONLY, we
3242  * need to in the size_change case to obtain a stateid.
3243  *
3244  * Got race?
3245  * Because OPEN is always done by name in nfsv4, it is
3246  * possible that we opened a different file by the same
3247  * name.  We can recognize this race condition, but we
3248  * can't do anything about it besides returning an error.
3249  *
3250  * This will be fixed with VFS changes (lookup-intent).
3251  */
3252 static int
3253 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3254 		  struct iattr *sattr)
3255 {
3256 	struct inode *inode = d_inode(dentry);
3257 	struct rpc_cred *cred = NULL;
3258 	struct nfs4_state *state = NULL;
3259 	struct nfs4_label *label = NULL;
3260 	int status;
3261 
3262 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3263 	    sattr->ia_valid & ATTR_SIZE &&
3264 	    sattr->ia_size < i_size_read(inode))
3265 		pnfs_commit_and_return_layout(inode);
3266 
3267 	nfs_fattr_init(fattr);
3268 
3269 	/* Deal with open(O_TRUNC) */
3270 	if (sattr->ia_valid & ATTR_OPEN)
3271 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3272 
3273 	/* Optimization: if the end result is no change, don't RPC */
3274 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3275 		return 0;
3276 
3277 	/* Search for an existing open(O_WRITE) file */
3278 	if (sattr->ia_valid & ATTR_FILE) {
3279 		struct nfs_open_context *ctx;
3280 
3281 		ctx = nfs_file_open_context(sattr->ia_file);
3282 		if (ctx) {
3283 			cred = ctx->cred;
3284 			state = ctx->state;
3285 		}
3286 	}
3287 
3288 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3289 	if (IS_ERR(label))
3290 		return PTR_ERR(label);
3291 
3292 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3293 	if (status == 0) {
3294 		nfs_setattr_update_inode(inode, sattr, fattr);
3295 		nfs_setsecurity(inode, fattr, label);
3296 	}
3297 	nfs4_label_free(label);
3298 	return status;
3299 }
3300 
3301 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3302 		const struct qstr *name, struct nfs_fh *fhandle,
3303 		struct nfs_fattr *fattr, struct nfs4_label *label)
3304 {
3305 	struct nfs_server *server = NFS_SERVER(dir);
3306 	int		       status;
3307 	struct nfs4_lookup_arg args = {
3308 		.bitmask = server->attr_bitmask,
3309 		.dir_fh = NFS_FH(dir),
3310 		.name = name,
3311 	};
3312 	struct nfs4_lookup_res res = {
3313 		.server = server,
3314 		.fattr = fattr,
3315 		.label = label,
3316 		.fh = fhandle,
3317 	};
3318 	struct rpc_message msg = {
3319 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3320 		.rpc_argp = &args,
3321 		.rpc_resp = &res,
3322 	};
3323 
3324 	args.bitmask = nfs4_bitmask(server, label);
3325 
3326 	nfs_fattr_init(fattr);
3327 
3328 	dprintk("NFS call  lookup %s\n", name->name);
3329 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3330 	dprintk("NFS reply lookup: %d\n", status);
3331 	return status;
3332 }
3333 
3334 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3335 {
3336 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3337 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3338 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3339 	fattr->nlink = 2;
3340 }
3341 
3342 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3343 				   struct qstr *name, struct nfs_fh *fhandle,
3344 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3345 {
3346 	struct nfs4_exception exception = { };
3347 	struct rpc_clnt *client = *clnt;
3348 	int err;
3349 	do {
3350 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3351 		trace_nfs4_lookup(dir, name, err);
3352 		switch (err) {
3353 		case -NFS4ERR_BADNAME:
3354 			err = -ENOENT;
3355 			goto out;
3356 		case -NFS4ERR_MOVED:
3357 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3358 			goto out;
3359 		case -NFS4ERR_WRONGSEC:
3360 			err = -EPERM;
3361 			if (client != *clnt)
3362 				goto out;
3363 			client = nfs4_negotiate_security(client, dir, name);
3364 			if (IS_ERR(client))
3365 				return PTR_ERR(client);
3366 
3367 			exception.retry = 1;
3368 			break;
3369 		default:
3370 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3371 		}
3372 	} while (exception.retry);
3373 
3374 out:
3375 	if (err == 0)
3376 		*clnt = client;
3377 	else if (client != *clnt)
3378 		rpc_shutdown_client(client);
3379 
3380 	return err;
3381 }
3382 
3383 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3384 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3385 			    struct nfs4_label *label)
3386 {
3387 	int status;
3388 	struct rpc_clnt *client = NFS_CLIENT(dir);
3389 
3390 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3391 	if (client != NFS_CLIENT(dir)) {
3392 		rpc_shutdown_client(client);
3393 		nfs_fixup_secinfo_attributes(fattr);
3394 	}
3395 	return status;
3396 }
3397 
3398 struct rpc_clnt *
3399 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3400 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3401 {
3402 	struct rpc_clnt *client = NFS_CLIENT(dir);
3403 	int status;
3404 
3405 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3406 	if (status < 0)
3407 		return ERR_PTR(status);
3408 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3409 }
3410 
3411 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3412 {
3413 	struct nfs_server *server = NFS_SERVER(inode);
3414 	struct nfs4_accessargs args = {
3415 		.fh = NFS_FH(inode),
3416 		.bitmask = server->cache_consistency_bitmask,
3417 	};
3418 	struct nfs4_accessres res = {
3419 		.server = server,
3420 	};
3421 	struct rpc_message msg = {
3422 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3423 		.rpc_argp = &args,
3424 		.rpc_resp = &res,
3425 		.rpc_cred = entry->cred,
3426 	};
3427 	int mode = entry->mask;
3428 	int status = 0;
3429 
3430 	/*
3431 	 * Determine which access bits we want to ask for...
3432 	 */
3433 	if (mode & MAY_READ)
3434 		args.access |= NFS4_ACCESS_READ;
3435 	if (S_ISDIR(inode->i_mode)) {
3436 		if (mode & MAY_WRITE)
3437 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3438 		if (mode & MAY_EXEC)
3439 			args.access |= NFS4_ACCESS_LOOKUP;
3440 	} else {
3441 		if (mode & MAY_WRITE)
3442 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3443 		if (mode & MAY_EXEC)
3444 			args.access |= NFS4_ACCESS_EXECUTE;
3445 	}
3446 
3447 	res.fattr = nfs_alloc_fattr();
3448 	if (res.fattr == NULL)
3449 		return -ENOMEM;
3450 
3451 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3452 	if (!status) {
3453 		nfs_access_set_mask(entry, res.access);
3454 		nfs_refresh_inode(inode, res.fattr);
3455 	}
3456 	nfs_free_fattr(res.fattr);
3457 	return status;
3458 }
3459 
3460 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3461 {
3462 	struct nfs4_exception exception = { };
3463 	int err;
3464 	do {
3465 		err = _nfs4_proc_access(inode, entry);
3466 		trace_nfs4_access(inode, err);
3467 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3468 				&exception);
3469 	} while (exception.retry);
3470 	return err;
3471 }
3472 
3473 /*
3474  * TODO: For the time being, we don't try to get any attributes
3475  * along with any of the zero-copy operations READ, READDIR,
3476  * READLINK, WRITE.
3477  *
3478  * In the case of the first three, we want to put the GETATTR
3479  * after the read-type operation -- this is because it is hard
3480  * to predict the length of a GETATTR response in v4, and thus
3481  * align the READ data correctly.  This means that the GETATTR
3482  * may end up partially falling into the page cache, and we should
3483  * shift it into the 'tail' of the xdr_buf before processing.
3484  * To do this efficiently, we need to know the total length
3485  * of data received, which doesn't seem to be available outside
3486  * of the RPC layer.
3487  *
3488  * In the case of WRITE, we also want to put the GETATTR after
3489  * the operation -- in this case because we want to make sure
3490  * we get the post-operation mtime and size.
3491  *
3492  * Both of these changes to the XDR layer would in fact be quite
3493  * minor, but I decided to leave them for a subsequent patch.
3494  */
3495 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3496 		unsigned int pgbase, unsigned int pglen)
3497 {
3498 	struct nfs4_readlink args = {
3499 		.fh       = NFS_FH(inode),
3500 		.pgbase	  = pgbase,
3501 		.pglen    = pglen,
3502 		.pages    = &page,
3503 	};
3504 	struct nfs4_readlink_res res;
3505 	struct rpc_message msg = {
3506 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3507 		.rpc_argp = &args,
3508 		.rpc_resp = &res,
3509 	};
3510 
3511 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3512 }
3513 
3514 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3515 		unsigned int pgbase, unsigned int pglen)
3516 {
3517 	struct nfs4_exception exception = { };
3518 	int err;
3519 	do {
3520 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3521 		trace_nfs4_readlink(inode, err);
3522 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3523 				&exception);
3524 	} while (exception.retry);
3525 	return err;
3526 }
3527 
3528 /*
3529  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3530  */
3531 static int
3532 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3533 		 int flags)
3534 {
3535 	struct nfs4_label l, *ilabel = NULL;
3536 	struct nfs_open_context *ctx;
3537 	struct nfs4_state *state;
3538 	int opened = 0;
3539 	int status = 0;
3540 
3541 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3542 	if (IS_ERR(ctx))
3543 		return PTR_ERR(ctx);
3544 
3545 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3546 
3547 	sattr->ia_mode &= ~current_umask();
3548 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3549 	if (IS_ERR(state)) {
3550 		status = PTR_ERR(state);
3551 		goto out;
3552 	}
3553 out:
3554 	nfs4_label_release_security(ilabel);
3555 	put_nfs_open_context(ctx);
3556 	return status;
3557 }
3558 
3559 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3560 {
3561 	struct nfs_server *server = NFS_SERVER(dir);
3562 	struct nfs_removeargs args = {
3563 		.fh = NFS_FH(dir),
3564 		.name = *name,
3565 	};
3566 	struct nfs_removeres res = {
3567 		.server = server,
3568 	};
3569 	struct rpc_message msg = {
3570 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3571 		.rpc_argp = &args,
3572 		.rpc_resp = &res,
3573 	};
3574 	int status;
3575 
3576 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3577 	if (status == 0)
3578 		update_changeattr(dir, &res.cinfo);
3579 	return status;
3580 }
3581 
3582 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3583 {
3584 	struct nfs4_exception exception = { };
3585 	int err;
3586 	do {
3587 		err = _nfs4_proc_remove(dir, name);
3588 		trace_nfs4_remove(dir, name, err);
3589 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3590 				&exception);
3591 	} while (exception.retry);
3592 	return err;
3593 }
3594 
3595 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3596 {
3597 	struct nfs_server *server = NFS_SERVER(dir);
3598 	struct nfs_removeargs *args = msg->rpc_argp;
3599 	struct nfs_removeres *res = msg->rpc_resp;
3600 
3601 	res->server = server;
3602 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3603 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3604 
3605 	nfs_fattr_init(res->dir_attr);
3606 }
3607 
3608 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3609 {
3610 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3611 			&data->args.seq_args,
3612 			&data->res.seq_res,
3613 			task);
3614 }
3615 
3616 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3617 {
3618 	struct nfs_unlinkdata *data = task->tk_calldata;
3619 	struct nfs_removeres *res = &data->res;
3620 
3621 	if (!nfs4_sequence_done(task, &res->seq_res))
3622 		return 0;
3623 	if (nfs4_async_handle_error(task, res->server, NULL,
3624 				    &data->timeout) == -EAGAIN)
3625 		return 0;
3626 	update_changeattr(dir, &res->cinfo);
3627 	return 1;
3628 }
3629 
3630 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3631 {
3632 	struct nfs_server *server = NFS_SERVER(dir);
3633 	struct nfs_renameargs *arg = msg->rpc_argp;
3634 	struct nfs_renameres *res = msg->rpc_resp;
3635 
3636 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3637 	res->server = server;
3638 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3639 }
3640 
3641 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3642 {
3643 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3644 			&data->args.seq_args,
3645 			&data->res.seq_res,
3646 			task);
3647 }
3648 
3649 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3650 				 struct inode *new_dir)
3651 {
3652 	struct nfs_renamedata *data = task->tk_calldata;
3653 	struct nfs_renameres *res = &data->res;
3654 
3655 	if (!nfs4_sequence_done(task, &res->seq_res))
3656 		return 0;
3657 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3658 		return 0;
3659 
3660 	update_changeattr(old_dir, &res->old_cinfo);
3661 	update_changeattr(new_dir, &res->new_cinfo);
3662 	return 1;
3663 }
3664 
3665 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3666 {
3667 	struct nfs_server *server = NFS_SERVER(inode);
3668 	struct nfs4_link_arg arg = {
3669 		.fh     = NFS_FH(inode),
3670 		.dir_fh = NFS_FH(dir),
3671 		.name   = name,
3672 		.bitmask = server->attr_bitmask,
3673 	};
3674 	struct nfs4_link_res res = {
3675 		.server = server,
3676 		.label = NULL,
3677 	};
3678 	struct rpc_message msg = {
3679 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3680 		.rpc_argp = &arg,
3681 		.rpc_resp = &res,
3682 	};
3683 	int status = -ENOMEM;
3684 
3685 	res.fattr = nfs_alloc_fattr();
3686 	if (res.fattr == NULL)
3687 		goto out;
3688 
3689 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3690 	if (IS_ERR(res.label)) {
3691 		status = PTR_ERR(res.label);
3692 		goto out;
3693 	}
3694 	arg.bitmask = nfs4_bitmask(server, res.label);
3695 
3696 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3697 	if (!status) {
3698 		update_changeattr(dir, &res.cinfo);
3699 		status = nfs_post_op_update_inode(inode, res.fattr);
3700 		if (!status)
3701 			nfs_setsecurity(inode, res.fattr, res.label);
3702 	}
3703 
3704 
3705 	nfs4_label_free(res.label);
3706 
3707 out:
3708 	nfs_free_fattr(res.fattr);
3709 	return status;
3710 }
3711 
3712 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3713 {
3714 	struct nfs4_exception exception = { };
3715 	int err;
3716 	do {
3717 		err = nfs4_handle_exception(NFS_SERVER(inode),
3718 				_nfs4_proc_link(inode, dir, name),
3719 				&exception);
3720 	} while (exception.retry);
3721 	return err;
3722 }
3723 
3724 struct nfs4_createdata {
3725 	struct rpc_message msg;
3726 	struct nfs4_create_arg arg;
3727 	struct nfs4_create_res res;
3728 	struct nfs_fh fh;
3729 	struct nfs_fattr fattr;
3730 	struct nfs4_label *label;
3731 };
3732 
3733 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3734 		struct qstr *name, struct iattr *sattr, u32 ftype)
3735 {
3736 	struct nfs4_createdata *data;
3737 
3738 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3739 	if (data != NULL) {
3740 		struct nfs_server *server = NFS_SERVER(dir);
3741 
3742 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3743 		if (IS_ERR(data->label))
3744 			goto out_free;
3745 
3746 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3747 		data->msg.rpc_argp = &data->arg;
3748 		data->msg.rpc_resp = &data->res;
3749 		data->arg.dir_fh = NFS_FH(dir);
3750 		data->arg.server = server;
3751 		data->arg.name = name;
3752 		data->arg.attrs = sattr;
3753 		data->arg.ftype = ftype;
3754 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3755 		data->res.server = server;
3756 		data->res.fh = &data->fh;
3757 		data->res.fattr = &data->fattr;
3758 		data->res.label = data->label;
3759 		nfs_fattr_init(data->res.fattr);
3760 	}
3761 	return data;
3762 out_free:
3763 	kfree(data);
3764 	return NULL;
3765 }
3766 
3767 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3768 {
3769 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3770 				    &data->arg.seq_args, &data->res.seq_res, 1);
3771 	if (status == 0) {
3772 		update_changeattr(dir, &data->res.dir_cinfo);
3773 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3774 	}
3775 	return status;
3776 }
3777 
3778 static void nfs4_free_createdata(struct nfs4_createdata *data)
3779 {
3780 	nfs4_label_free(data->label);
3781 	kfree(data);
3782 }
3783 
3784 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3785 		struct page *page, unsigned int len, struct iattr *sattr,
3786 		struct nfs4_label *label)
3787 {
3788 	struct nfs4_createdata *data;
3789 	int status = -ENAMETOOLONG;
3790 
3791 	if (len > NFS4_MAXPATHLEN)
3792 		goto out;
3793 
3794 	status = -ENOMEM;
3795 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3796 	if (data == NULL)
3797 		goto out;
3798 
3799 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3800 	data->arg.u.symlink.pages = &page;
3801 	data->arg.u.symlink.len = len;
3802 	data->arg.label = label;
3803 
3804 	status = nfs4_do_create(dir, dentry, data);
3805 
3806 	nfs4_free_createdata(data);
3807 out:
3808 	return status;
3809 }
3810 
3811 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3812 		struct page *page, unsigned int len, struct iattr *sattr)
3813 {
3814 	struct nfs4_exception exception = { };
3815 	struct nfs4_label l, *label = NULL;
3816 	int err;
3817 
3818 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3819 
3820 	do {
3821 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3822 		trace_nfs4_symlink(dir, &dentry->d_name, err);
3823 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3824 				&exception);
3825 	} while (exception.retry);
3826 
3827 	nfs4_label_release_security(label);
3828 	return err;
3829 }
3830 
3831 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3832 		struct iattr *sattr, struct nfs4_label *label)
3833 {
3834 	struct nfs4_createdata *data;
3835 	int status = -ENOMEM;
3836 
3837 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3838 	if (data == NULL)
3839 		goto out;
3840 
3841 	data->arg.label = label;
3842 	status = nfs4_do_create(dir, dentry, data);
3843 
3844 	nfs4_free_createdata(data);
3845 out:
3846 	return status;
3847 }
3848 
3849 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3850 		struct iattr *sattr)
3851 {
3852 	struct nfs4_exception exception = { };
3853 	struct nfs4_label l, *label = NULL;
3854 	int err;
3855 
3856 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3857 
3858 	sattr->ia_mode &= ~current_umask();
3859 	do {
3860 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3861 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
3862 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3863 				&exception);
3864 	} while (exception.retry);
3865 	nfs4_label_release_security(label);
3866 
3867 	return err;
3868 }
3869 
3870 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3871 		u64 cookie, struct page **pages, unsigned int count, int plus)
3872 {
3873 	struct inode		*dir = d_inode(dentry);
3874 	struct nfs4_readdir_arg args = {
3875 		.fh = NFS_FH(dir),
3876 		.pages = pages,
3877 		.pgbase = 0,
3878 		.count = count,
3879 		.bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3880 		.plus = plus,
3881 	};
3882 	struct nfs4_readdir_res res;
3883 	struct rpc_message msg = {
3884 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3885 		.rpc_argp = &args,
3886 		.rpc_resp = &res,
3887 		.rpc_cred = cred,
3888 	};
3889 	int			status;
3890 
3891 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3892 			dentry,
3893 			(unsigned long long)cookie);
3894 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3895 	res.pgbase = args.pgbase;
3896 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3897 	if (status >= 0) {
3898 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3899 		status += args.pgbase;
3900 	}
3901 
3902 	nfs_invalidate_atime(dir);
3903 
3904 	dprintk("%s: returns %d\n", __func__, status);
3905 	return status;
3906 }
3907 
3908 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3909 		u64 cookie, struct page **pages, unsigned int count, int plus)
3910 {
3911 	struct nfs4_exception exception = { };
3912 	int err;
3913 	do {
3914 		err = _nfs4_proc_readdir(dentry, cred, cookie,
3915 				pages, count, plus);
3916 		trace_nfs4_readdir(d_inode(dentry), err);
3917 		err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3918 				&exception);
3919 	} while (exception.retry);
3920 	return err;
3921 }
3922 
3923 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3924 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3925 {
3926 	struct nfs4_createdata *data;
3927 	int mode = sattr->ia_mode;
3928 	int status = -ENOMEM;
3929 
3930 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3931 	if (data == NULL)
3932 		goto out;
3933 
3934 	if (S_ISFIFO(mode))
3935 		data->arg.ftype = NF4FIFO;
3936 	else if (S_ISBLK(mode)) {
3937 		data->arg.ftype = NF4BLK;
3938 		data->arg.u.device.specdata1 = MAJOR(rdev);
3939 		data->arg.u.device.specdata2 = MINOR(rdev);
3940 	}
3941 	else if (S_ISCHR(mode)) {
3942 		data->arg.ftype = NF4CHR;
3943 		data->arg.u.device.specdata1 = MAJOR(rdev);
3944 		data->arg.u.device.specdata2 = MINOR(rdev);
3945 	} else if (!S_ISSOCK(mode)) {
3946 		status = -EINVAL;
3947 		goto out_free;
3948 	}
3949 
3950 	data->arg.label = label;
3951 	status = nfs4_do_create(dir, dentry, data);
3952 out_free:
3953 	nfs4_free_createdata(data);
3954 out:
3955 	return status;
3956 }
3957 
3958 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3959 		struct iattr *sattr, dev_t rdev)
3960 {
3961 	struct nfs4_exception exception = { };
3962 	struct nfs4_label l, *label = NULL;
3963 	int err;
3964 
3965 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3966 
3967 	sattr->ia_mode &= ~current_umask();
3968 	do {
3969 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3970 		trace_nfs4_mknod(dir, &dentry->d_name, err);
3971 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3972 				&exception);
3973 	} while (exception.retry);
3974 
3975 	nfs4_label_release_security(label);
3976 
3977 	return err;
3978 }
3979 
3980 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3981 		 struct nfs_fsstat *fsstat)
3982 {
3983 	struct nfs4_statfs_arg args = {
3984 		.fh = fhandle,
3985 		.bitmask = server->attr_bitmask,
3986 	};
3987 	struct nfs4_statfs_res res = {
3988 		.fsstat = fsstat,
3989 	};
3990 	struct rpc_message msg = {
3991 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3992 		.rpc_argp = &args,
3993 		.rpc_resp = &res,
3994 	};
3995 
3996 	nfs_fattr_init(fsstat->fattr);
3997 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3998 }
3999 
4000 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4001 {
4002 	struct nfs4_exception exception = { };
4003 	int err;
4004 	do {
4005 		err = nfs4_handle_exception(server,
4006 				_nfs4_proc_statfs(server, fhandle, fsstat),
4007 				&exception);
4008 	} while (exception.retry);
4009 	return err;
4010 }
4011 
4012 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4013 		struct nfs_fsinfo *fsinfo)
4014 {
4015 	struct nfs4_fsinfo_arg args = {
4016 		.fh = fhandle,
4017 		.bitmask = server->attr_bitmask,
4018 	};
4019 	struct nfs4_fsinfo_res res = {
4020 		.fsinfo = fsinfo,
4021 	};
4022 	struct rpc_message msg = {
4023 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4024 		.rpc_argp = &args,
4025 		.rpc_resp = &res,
4026 	};
4027 
4028 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4029 }
4030 
4031 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4032 {
4033 	struct nfs4_exception exception = { };
4034 	unsigned long now = jiffies;
4035 	int err;
4036 
4037 	do {
4038 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4039 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4040 		if (err == 0) {
4041 			struct nfs_client *clp = server->nfs_client;
4042 
4043 			spin_lock(&clp->cl_lock);
4044 			clp->cl_lease_time = fsinfo->lease_time * HZ;
4045 			clp->cl_last_renewal = now;
4046 			spin_unlock(&clp->cl_lock);
4047 			break;
4048 		}
4049 		err = nfs4_handle_exception(server, err, &exception);
4050 	} while (exception.retry);
4051 	return err;
4052 }
4053 
4054 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4055 {
4056 	int error;
4057 
4058 	nfs_fattr_init(fsinfo->fattr);
4059 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4060 	if (error == 0) {
4061 		/* block layout checks this! */
4062 		server->pnfs_blksize = fsinfo->blksize;
4063 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4064 	}
4065 
4066 	return error;
4067 }
4068 
4069 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4070 		struct nfs_pathconf *pathconf)
4071 {
4072 	struct nfs4_pathconf_arg args = {
4073 		.fh = fhandle,
4074 		.bitmask = server->attr_bitmask,
4075 	};
4076 	struct nfs4_pathconf_res res = {
4077 		.pathconf = pathconf,
4078 	};
4079 	struct rpc_message msg = {
4080 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4081 		.rpc_argp = &args,
4082 		.rpc_resp = &res,
4083 	};
4084 
4085 	/* None of the pathconf attributes are mandatory to implement */
4086 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4087 		memset(pathconf, 0, sizeof(*pathconf));
4088 		return 0;
4089 	}
4090 
4091 	nfs_fattr_init(pathconf->fattr);
4092 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4093 }
4094 
4095 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4096 		struct nfs_pathconf *pathconf)
4097 {
4098 	struct nfs4_exception exception = { };
4099 	int err;
4100 
4101 	do {
4102 		err = nfs4_handle_exception(server,
4103 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4104 				&exception);
4105 	} while (exception.retry);
4106 	return err;
4107 }
4108 
4109 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4110 		const struct nfs_open_context *ctx,
4111 		const struct nfs_lock_context *l_ctx,
4112 		fmode_t fmode)
4113 {
4114 	const struct nfs_lockowner *lockowner = NULL;
4115 
4116 	if (l_ctx != NULL)
4117 		lockowner = &l_ctx->lockowner;
4118 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4119 }
4120 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4121 
4122 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4123 		const struct nfs_open_context *ctx,
4124 		const struct nfs_lock_context *l_ctx,
4125 		fmode_t fmode)
4126 {
4127 	nfs4_stateid current_stateid;
4128 
4129 	/* If the current stateid represents a lost lock, then exit */
4130 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4131 		return true;
4132 	return nfs4_stateid_match(stateid, &current_stateid);
4133 }
4134 
4135 static bool nfs4_error_stateid_expired(int err)
4136 {
4137 	switch (err) {
4138 	case -NFS4ERR_DELEG_REVOKED:
4139 	case -NFS4ERR_ADMIN_REVOKED:
4140 	case -NFS4ERR_BAD_STATEID:
4141 	case -NFS4ERR_STALE_STATEID:
4142 	case -NFS4ERR_OLD_STATEID:
4143 	case -NFS4ERR_OPENMODE:
4144 	case -NFS4ERR_EXPIRED:
4145 		return true;
4146 	}
4147 	return false;
4148 }
4149 
4150 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4151 {
4152 	nfs_invalidate_atime(hdr->inode);
4153 }
4154 
4155 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4156 {
4157 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4158 
4159 	trace_nfs4_read(hdr, task->tk_status);
4160 	if (nfs4_async_handle_error(task, server,
4161 				    hdr->args.context->state,
4162 				    NULL) == -EAGAIN) {
4163 		rpc_restart_call_prepare(task);
4164 		return -EAGAIN;
4165 	}
4166 
4167 	__nfs4_read_done_cb(hdr);
4168 	if (task->tk_status > 0)
4169 		renew_lease(server, hdr->timestamp);
4170 	return 0;
4171 }
4172 
4173 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4174 		struct nfs_pgio_args *args)
4175 {
4176 
4177 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4178 		nfs4_stateid_is_current(&args->stateid,
4179 				args->context,
4180 				args->lock_context,
4181 				FMODE_READ))
4182 		return false;
4183 	rpc_restart_call_prepare(task);
4184 	return true;
4185 }
4186 
4187 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4188 {
4189 
4190 	dprintk("--> %s\n", __func__);
4191 
4192 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4193 		return -EAGAIN;
4194 	if (nfs4_read_stateid_changed(task, &hdr->args))
4195 		return -EAGAIN;
4196 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4197 				    nfs4_read_done_cb(task, hdr);
4198 }
4199 
4200 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4201 				 struct rpc_message *msg)
4202 {
4203 	hdr->timestamp   = jiffies;
4204 	hdr->pgio_done_cb = nfs4_read_done_cb;
4205 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4206 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4207 }
4208 
4209 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4210 				      struct nfs_pgio_header *hdr)
4211 {
4212 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4213 			&hdr->args.seq_args,
4214 			&hdr->res.seq_res,
4215 			task))
4216 		return 0;
4217 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4218 				hdr->args.lock_context,
4219 				hdr->rw_ops->rw_mode) == -EIO)
4220 		return -EIO;
4221 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4222 		return -EIO;
4223 	return 0;
4224 }
4225 
4226 static int nfs4_write_done_cb(struct rpc_task *task,
4227 			      struct nfs_pgio_header *hdr)
4228 {
4229 	struct inode *inode = hdr->inode;
4230 
4231 	trace_nfs4_write(hdr, task->tk_status);
4232 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4233 				    hdr->args.context->state,
4234 				    NULL) == -EAGAIN) {
4235 		rpc_restart_call_prepare(task);
4236 		return -EAGAIN;
4237 	}
4238 	if (task->tk_status >= 0) {
4239 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4240 		nfs_writeback_update_inode(hdr);
4241 	}
4242 	return 0;
4243 }
4244 
4245 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4246 		struct nfs_pgio_args *args)
4247 {
4248 
4249 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4250 		nfs4_stateid_is_current(&args->stateid,
4251 				args->context,
4252 				args->lock_context,
4253 				FMODE_WRITE))
4254 		return false;
4255 	rpc_restart_call_prepare(task);
4256 	return true;
4257 }
4258 
4259 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4260 {
4261 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4262 		return -EAGAIN;
4263 	if (nfs4_write_stateid_changed(task, &hdr->args))
4264 		return -EAGAIN;
4265 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4266 		nfs4_write_done_cb(task, hdr);
4267 }
4268 
4269 static
4270 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4271 {
4272 	/* Don't request attributes for pNFS or O_DIRECT writes */
4273 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4274 		return false;
4275 	/* Otherwise, request attributes if and only if we don't hold
4276 	 * a delegation
4277 	 */
4278 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4279 }
4280 
4281 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4282 				  struct rpc_message *msg)
4283 {
4284 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4285 
4286 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4287 		hdr->args.bitmask = NULL;
4288 		hdr->res.fattr = NULL;
4289 	} else
4290 		hdr->args.bitmask = server->cache_consistency_bitmask;
4291 
4292 	if (!hdr->pgio_done_cb)
4293 		hdr->pgio_done_cb = nfs4_write_done_cb;
4294 	hdr->res.server = server;
4295 	hdr->timestamp   = jiffies;
4296 
4297 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4298 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4299 }
4300 
4301 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4302 {
4303 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4304 			&data->args.seq_args,
4305 			&data->res.seq_res,
4306 			task);
4307 }
4308 
4309 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4310 {
4311 	struct inode *inode = data->inode;
4312 
4313 	trace_nfs4_commit(data, task->tk_status);
4314 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4315 				    NULL, NULL) == -EAGAIN) {
4316 		rpc_restart_call_prepare(task);
4317 		return -EAGAIN;
4318 	}
4319 	return 0;
4320 }
4321 
4322 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4323 {
4324 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4325 		return -EAGAIN;
4326 	return data->commit_done_cb(task, data);
4327 }
4328 
4329 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4330 {
4331 	struct nfs_server *server = NFS_SERVER(data->inode);
4332 
4333 	if (data->commit_done_cb == NULL)
4334 		data->commit_done_cb = nfs4_commit_done_cb;
4335 	data->res.server = server;
4336 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4337 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4338 }
4339 
4340 struct nfs4_renewdata {
4341 	struct nfs_client	*client;
4342 	unsigned long		timestamp;
4343 };
4344 
4345 /*
4346  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4347  * standalone procedure for queueing an asynchronous RENEW.
4348  */
4349 static void nfs4_renew_release(void *calldata)
4350 {
4351 	struct nfs4_renewdata *data = calldata;
4352 	struct nfs_client *clp = data->client;
4353 
4354 	if (atomic_read(&clp->cl_count) > 1)
4355 		nfs4_schedule_state_renewal(clp);
4356 	nfs_put_client(clp);
4357 	kfree(data);
4358 }
4359 
4360 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4361 {
4362 	struct nfs4_renewdata *data = calldata;
4363 	struct nfs_client *clp = data->client;
4364 	unsigned long timestamp = data->timestamp;
4365 
4366 	trace_nfs4_renew_async(clp, task->tk_status);
4367 	switch (task->tk_status) {
4368 	case 0:
4369 		break;
4370 	case -NFS4ERR_LEASE_MOVED:
4371 		nfs4_schedule_lease_moved_recovery(clp);
4372 		break;
4373 	default:
4374 		/* Unless we're shutting down, schedule state recovery! */
4375 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4376 			return;
4377 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4378 			nfs4_schedule_lease_recovery(clp);
4379 			return;
4380 		}
4381 		nfs4_schedule_path_down_recovery(clp);
4382 	}
4383 	do_renew_lease(clp, timestamp);
4384 }
4385 
4386 static const struct rpc_call_ops nfs4_renew_ops = {
4387 	.rpc_call_done = nfs4_renew_done,
4388 	.rpc_release = nfs4_renew_release,
4389 };
4390 
4391 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4392 {
4393 	struct rpc_message msg = {
4394 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4395 		.rpc_argp	= clp,
4396 		.rpc_cred	= cred,
4397 	};
4398 	struct nfs4_renewdata *data;
4399 
4400 	if (renew_flags == 0)
4401 		return 0;
4402 	if (!atomic_inc_not_zero(&clp->cl_count))
4403 		return -EIO;
4404 	data = kmalloc(sizeof(*data), GFP_NOFS);
4405 	if (data == NULL)
4406 		return -ENOMEM;
4407 	data->client = clp;
4408 	data->timestamp = jiffies;
4409 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4410 			&nfs4_renew_ops, data);
4411 }
4412 
4413 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4414 {
4415 	struct rpc_message msg = {
4416 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4417 		.rpc_argp	= clp,
4418 		.rpc_cred	= cred,
4419 	};
4420 	unsigned long now = jiffies;
4421 	int status;
4422 
4423 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4424 	if (status < 0)
4425 		return status;
4426 	do_renew_lease(clp, now);
4427 	return 0;
4428 }
4429 
4430 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4431 {
4432 	return server->caps & NFS_CAP_ACLS;
4433 }
4434 
4435 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4436  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4437  * the stack.
4438  */
4439 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4440 
4441 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4442 		struct page **pages, unsigned int *pgbase)
4443 {
4444 	struct page *newpage, **spages;
4445 	int rc = 0;
4446 	size_t len;
4447 	spages = pages;
4448 
4449 	do {
4450 		len = min_t(size_t, PAGE_SIZE, buflen);
4451 		newpage = alloc_page(GFP_KERNEL);
4452 
4453 		if (newpage == NULL)
4454 			goto unwind;
4455 		memcpy(page_address(newpage), buf, len);
4456                 buf += len;
4457                 buflen -= len;
4458 		*pages++ = newpage;
4459 		rc++;
4460 	} while (buflen != 0);
4461 
4462 	return rc;
4463 
4464 unwind:
4465 	for(; rc > 0; rc--)
4466 		__free_page(spages[rc-1]);
4467 	return -ENOMEM;
4468 }
4469 
4470 struct nfs4_cached_acl {
4471 	int cached;
4472 	size_t len;
4473 	char data[0];
4474 };
4475 
4476 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4477 {
4478 	struct nfs_inode *nfsi = NFS_I(inode);
4479 
4480 	spin_lock(&inode->i_lock);
4481 	kfree(nfsi->nfs4_acl);
4482 	nfsi->nfs4_acl = acl;
4483 	spin_unlock(&inode->i_lock);
4484 }
4485 
4486 static void nfs4_zap_acl_attr(struct inode *inode)
4487 {
4488 	nfs4_set_cached_acl(inode, NULL);
4489 }
4490 
4491 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4492 {
4493 	struct nfs_inode *nfsi = NFS_I(inode);
4494 	struct nfs4_cached_acl *acl;
4495 	int ret = -ENOENT;
4496 
4497 	spin_lock(&inode->i_lock);
4498 	acl = nfsi->nfs4_acl;
4499 	if (acl == NULL)
4500 		goto out;
4501 	if (buf == NULL) /* user is just asking for length */
4502 		goto out_len;
4503 	if (acl->cached == 0)
4504 		goto out;
4505 	ret = -ERANGE; /* see getxattr(2) man page */
4506 	if (acl->len > buflen)
4507 		goto out;
4508 	memcpy(buf, acl->data, acl->len);
4509 out_len:
4510 	ret = acl->len;
4511 out:
4512 	spin_unlock(&inode->i_lock);
4513 	return ret;
4514 }
4515 
4516 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4517 {
4518 	struct nfs4_cached_acl *acl;
4519 	size_t buflen = sizeof(*acl) + acl_len;
4520 
4521 	if (buflen <= PAGE_SIZE) {
4522 		acl = kmalloc(buflen, GFP_KERNEL);
4523 		if (acl == NULL)
4524 			goto out;
4525 		acl->cached = 1;
4526 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4527 	} else {
4528 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4529 		if (acl == NULL)
4530 			goto out;
4531 		acl->cached = 0;
4532 	}
4533 	acl->len = acl_len;
4534 out:
4535 	nfs4_set_cached_acl(inode, acl);
4536 }
4537 
4538 /*
4539  * The getxattr API returns the required buffer length when called with a
4540  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4541  * the required buf.  On a NULL buf, we send a page of data to the server
4542  * guessing that the ACL request can be serviced by a page. If so, we cache
4543  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4544  * the cache. If not so, we throw away the page, and cache the required
4545  * length. The next getxattr call will then produce another round trip to
4546  * the server, this time with the input buf of the required size.
4547  */
4548 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4549 {
4550 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4551 	struct nfs_getaclargs args = {
4552 		.fh = NFS_FH(inode),
4553 		.acl_pages = pages,
4554 		.acl_len = buflen,
4555 	};
4556 	struct nfs_getaclres res = {
4557 		.acl_len = buflen,
4558 	};
4559 	struct rpc_message msg = {
4560 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4561 		.rpc_argp = &args,
4562 		.rpc_resp = &res,
4563 	};
4564 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4565 	int ret = -ENOMEM, i;
4566 
4567 	/* As long as we're doing a round trip to the server anyway,
4568 	 * let's be prepared for a page of acl data. */
4569 	if (npages == 0)
4570 		npages = 1;
4571 	if (npages > ARRAY_SIZE(pages))
4572 		return -ERANGE;
4573 
4574 	for (i = 0; i < npages; i++) {
4575 		pages[i] = alloc_page(GFP_KERNEL);
4576 		if (!pages[i])
4577 			goto out_free;
4578 	}
4579 
4580 	/* for decoding across pages */
4581 	res.acl_scratch = alloc_page(GFP_KERNEL);
4582 	if (!res.acl_scratch)
4583 		goto out_free;
4584 
4585 	args.acl_len = npages * PAGE_SIZE;
4586 	args.acl_pgbase = 0;
4587 
4588 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4589 		__func__, buf, buflen, npages, args.acl_len);
4590 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4591 			     &msg, &args.seq_args, &res.seq_res, 0);
4592 	if (ret)
4593 		goto out_free;
4594 
4595 	/* Handle the case where the passed-in buffer is too short */
4596 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4597 		/* Did the user only issue a request for the acl length? */
4598 		if (buf == NULL)
4599 			goto out_ok;
4600 		ret = -ERANGE;
4601 		goto out_free;
4602 	}
4603 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4604 	if (buf) {
4605 		if (res.acl_len > buflen) {
4606 			ret = -ERANGE;
4607 			goto out_free;
4608 		}
4609 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4610 	}
4611 out_ok:
4612 	ret = res.acl_len;
4613 out_free:
4614 	for (i = 0; i < npages; i++)
4615 		if (pages[i])
4616 			__free_page(pages[i]);
4617 	if (res.acl_scratch)
4618 		__free_page(res.acl_scratch);
4619 	return ret;
4620 }
4621 
4622 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4623 {
4624 	struct nfs4_exception exception = { };
4625 	ssize_t ret;
4626 	do {
4627 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4628 		trace_nfs4_get_acl(inode, ret);
4629 		if (ret >= 0)
4630 			break;
4631 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4632 	} while (exception.retry);
4633 	return ret;
4634 }
4635 
4636 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4637 {
4638 	struct nfs_server *server = NFS_SERVER(inode);
4639 	int ret;
4640 
4641 	if (!nfs4_server_supports_acls(server))
4642 		return -EOPNOTSUPP;
4643 	ret = nfs_revalidate_inode(server, inode);
4644 	if (ret < 0)
4645 		return ret;
4646 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4647 		nfs_zap_acl_cache(inode);
4648 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4649 	if (ret != -ENOENT)
4650 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4651 		 * but no cached acl data, just the acl length */
4652 		return ret;
4653 	return nfs4_get_acl_uncached(inode, buf, buflen);
4654 }
4655 
4656 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4657 {
4658 	struct nfs_server *server = NFS_SERVER(inode);
4659 	struct page *pages[NFS4ACL_MAXPAGES];
4660 	struct nfs_setaclargs arg = {
4661 		.fh		= NFS_FH(inode),
4662 		.acl_pages	= pages,
4663 		.acl_len	= buflen,
4664 	};
4665 	struct nfs_setaclres res;
4666 	struct rpc_message msg = {
4667 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4668 		.rpc_argp	= &arg,
4669 		.rpc_resp	= &res,
4670 	};
4671 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4672 	int ret, i;
4673 
4674 	if (!nfs4_server_supports_acls(server))
4675 		return -EOPNOTSUPP;
4676 	if (npages > ARRAY_SIZE(pages))
4677 		return -ERANGE;
4678 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4679 	if (i < 0)
4680 		return i;
4681 	nfs4_inode_return_delegation(inode);
4682 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4683 
4684 	/*
4685 	 * Free each page after tx, so the only ref left is
4686 	 * held by the network stack
4687 	 */
4688 	for (; i > 0; i--)
4689 		put_page(pages[i-1]);
4690 
4691 	/*
4692 	 * Acl update can result in inode attribute update.
4693 	 * so mark the attribute cache invalid.
4694 	 */
4695 	spin_lock(&inode->i_lock);
4696 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4697 	spin_unlock(&inode->i_lock);
4698 	nfs_access_zap_cache(inode);
4699 	nfs_zap_acl_cache(inode);
4700 	return ret;
4701 }
4702 
4703 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4704 {
4705 	struct nfs4_exception exception = { };
4706 	int err;
4707 	do {
4708 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4709 		trace_nfs4_set_acl(inode, err);
4710 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4711 				&exception);
4712 	} while (exception.retry);
4713 	return err;
4714 }
4715 
4716 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4717 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4718 					size_t buflen)
4719 {
4720 	struct nfs_server *server = NFS_SERVER(inode);
4721 	struct nfs_fattr fattr;
4722 	struct nfs4_label label = {0, 0, buflen, buf};
4723 
4724 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4725 	struct nfs4_getattr_arg arg = {
4726 		.fh		= NFS_FH(inode),
4727 		.bitmask	= bitmask,
4728 	};
4729 	struct nfs4_getattr_res res = {
4730 		.fattr		= &fattr,
4731 		.label		= &label,
4732 		.server		= server,
4733 	};
4734 	struct rpc_message msg = {
4735 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4736 		.rpc_argp	= &arg,
4737 		.rpc_resp	= &res,
4738 	};
4739 	int ret;
4740 
4741 	nfs_fattr_init(&fattr);
4742 
4743 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4744 	if (ret)
4745 		return ret;
4746 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4747 		return -ENOENT;
4748 	if (buflen < label.len)
4749 		return -ERANGE;
4750 	return 0;
4751 }
4752 
4753 static int nfs4_get_security_label(struct inode *inode, void *buf,
4754 					size_t buflen)
4755 {
4756 	struct nfs4_exception exception = { };
4757 	int err;
4758 
4759 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4760 		return -EOPNOTSUPP;
4761 
4762 	do {
4763 		err = _nfs4_get_security_label(inode, buf, buflen);
4764 		trace_nfs4_get_security_label(inode, err);
4765 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4766 				&exception);
4767 	} while (exception.retry);
4768 	return err;
4769 }
4770 
4771 static int _nfs4_do_set_security_label(struct inode *inode,
4772 		struct nfs4_label *ilabel,
4773 		struct nfs_fattr *fattr,
4774 		struct nfs4_label *olabel)
4775 {
4776 
4777 	struct iattr sattr = {0};
4778 	struct nfs_server *server = NFS_SERVER(inode);
4779 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4780 	struct nfs_setattrargs arg = {
4781 		.fh             = NFS_FH(inode),
4782 		.iap            = &sattr,
4783 		.server		= server,
4784 		.bitmask	= bitmask,
4785 		.label		= ilabel,
4786 	};
4787 	struct nfs_setattrres res = {
4788 		.fattr		= fattr,
4789 		.label		= olabel,
4790 		.server		= server,
4791 	};
4792 	struct rpc_message msg = {
4793 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4794 		.rpc_argp       = &arg,
4795 		.rpc_resp       = &res,
4796 	};
4797 	int status;
4798 
4799 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4800 
4801 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4802 	if (status)
4803 		dprintk("%s failed: %d\n", __func__, status);
4804 
4805 	return status;
4806 }
4807 
4808 static int nfs4_do_set_security_label(struct inode *inode,
4809 		struct nfs4_label *ilabel,
4810 		struct nfs_fattr *fattr,
4811 		struct nfs4_label *olabel)
4812 {
4813 	struct nfs4_exception exception = { };
4814 	int err;
4815 
4816 	do {
4817 		err = _nfs4_do_set_security_label(inode, ilabel,
4818 				fattr, olabel);
4819 		trace_nfs4_set_security_label(inode, err);
4820 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4821 				&exception);
4822 	} while (exception.retry);
4823 	return err;
4824 }
4825 
4826 static int
4827 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4828 {
4829 	struct nfs4_label ilabel, *olabel = NULL;
4830 	struct nfs_fattr fattr;
4831 	struct rpc_cred *cred;
4832 	struct inode *inode = d_inode(dentry);
4833 	int status;
4834 
4835 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4836 		return -EOPNOTSUPP;
4837 
4838 	nfs_fattr_init(&fattr);
4839 
4840 	ilabel.pi = 0;
4841 	ilabel.lfs = 0;
4842 	ilabel.label = (char *)buf;
4843 	ilabel.len = buflen;
4844 
4845 	cred = rpc_lookup_cred();
4846 	if (IS_ERR(cred))
4847 		return PTR_ERR(cred);
4848 
4849 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4850 	if (IS_ERR(olabel)) {
4851 		status = -PTR_ERR(olabel);
4852 		goto out;
4853 	}
4854 
4855 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4856 	if (status == 0)
4857 		nfs_setsecurity(inode, &fattr, olabel);
4858 
4859 	nfs4_label_free(olabel);
4860 out:
4861 	put_rpccred(cred);
4862 	return status;
4863 }
4864 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4865 
4866 
4867 static int
4868 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4869 			struct nfs4_state *state, long *timeout)
4870 {
4871 	struct nfs_client *clp = server->nfs_client;
4872 
4873 	if (task->tk_status >= 0)
4874 		return 0;
4875 	switch(task->tk_status) {
4876 		case -NFS4ERR_DELEG_REVOKED:
4877 		case -NFS4ERR_ADMIN_REVOKED:
4878 		case -NFS4ERR_BAD_STATEID:
4879 		case -NFS4ERR_OPENMODE:
4880 			if (state == NULL)
4881 				break;
4882 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4883 				goto recovery_failed;
4884 			goto wait_on_recovery;
4885 		case -NFS4ERR_EXPIRED:
4886 			if (state != NULL) {
4887 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4888 					goto recovery_failed;
4889 			}
4890 		case -NFS4ERR_STALE_STATEID:
4891 		case -NFS4ERR_STALE_CLIENTID:
4892 			nfs4_schedule_lease_recovery(clp);
4893 			goto wait_on_recovery;
4894 		case -NFS4ERR_MOVED:
4895 			if (nfs4_schedule_migration_recovery(server) < 0)
4896 				goto recovery_failed;
4897 			goto wait_on_recovery;
4898 		case -NFS4ERR_LEASE_MOVED:
4899 			nfs4_schedule_lease_moved_recovery(clp);
4900 			goto wait_on_recovery;
4901 #if defined(CONFIG_NFS_V4_1)
4902 		case -NFS4ERR_BADSESSION:
4903 		case -NFS4ERR_BADSLOT:
4904 		case -NFS4ERR_BAD_HIGH_SLOT:
4905 		case -NFS4ERR_DEADSESSION:
4906 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4907 		case -NFS4ERR_SEQ_FALSE_RETRY:
4908 		case -NFS4ERR_SEQ_MISORDERED:
4909 			dprintk("%s ERROR %d, Reset session\n", __func__,
4910 				task->tk_status);
4911 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4912 			goto wait_on_recovery;
4913 #endif /* CONFIG_NFS_V4_1 */
4914 		case -NFS4ERR_DELAY:
4915 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4916 			rpc_delay(task, nfs4_update_delay(timeout));
4917 			goto restart_call;
4918 		case -NFS4ERR_GRACE:
4919 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4920 		case -NFS4ERR_RETRY_UNCACHED_REP:
4921 		case -NFS4ERR_OLD_STATEID:
4922 			goto restart_call;
4923 	}
4924 	task->tk_status = nfs4_map_errors(task->tk_status);
4925 	return 0;
4926 recovery_failed:
4927 	task->tk_status = -EIO;
4928 	return 0;
4929 wait_on_recovery:
4930 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4931 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4932 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4933 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4934 		goto recovery_failed;
4935 restart_call:
4936 	task->tk_status = 0;
4937 	return -EAGAIN;
4938 }
4939 
4940 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4941 				    nfs4_verifier *bootverf)
4942 {
4943 	__be32 verf[2];
4944 
4945 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4946 		/* An impossible timestamp guarantees this value
4947 		 * will never match a generated boot time. */
4948 		verf[0] = 0;
4949 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4950 	} else {
4951 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4952 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4953 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4954 	}
4955 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4956 }
4957 
4958 static unsigned int
4959 nfs4_init_nonuniform_client_string(struct nfs_client *clp,
4960 				   char *buf, size_t len)
4961 {
4962 	unsigned int result;
4963 
4964 	if (clp->cl_owner_id != NULL)
4965 		return strlcpy(buf, clp->cl_owner_id, len);
4966 
4967 	rcu_read_lock();
4968 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4969 				clp->cl_ipaddr,
4970 				rpc_peeraddr2str(clp->cl_rpcclient,
4971 							RPC_DISPLAY_ADDR),
4972 				rpc_peeraddr2str(clp->cl_rpcclient,
4973 							RPC_DISPLAY_PROTO));
4974 	rcu_read_unlock();
4975 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4976 	return result;
4977 }
4978 
4979 static unsigned int
4980 nfs4_init_uniform_client_string(struct nfs_client *clp,
4981 				char *buf, size_t len)
4982 {
4983 	const char *nodename = clp->cl_rpcclient->cl_nodename;
4984 	unsigned int result;
4985 
4986 	if (clp->cl_owner_id != NULL)
4987 		return strlcpy(buf, clp->cl_owner_id, len);
4988 
4989 	if (nfs4_client_id_uniquifier[0] != '\0')
4990 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4991 				clp->rpc_ops->version,
4992 				clp->cl_minorversion,
4993 				nfs4_client_id_uniquifier,
4994 				nodename);
4995 	else
4996 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s",
4997 				clp->rpc_ops->version, clp->cl_minorversion,
4998 				nodename);
4999 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
5000 	return result;
5001 }
5002 
5003 /*
5004  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5005  * services.  Advertise one based on the address family of the
5006  * clientaddr.
5007  */
5008 static unsigned int
5009 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5010 {
5011 	if (strchr(clp->cl_ipaddr, ':') != NULL)
5012 		return scnprintf(buf, len, "tcp6");
5013 	else
5014 		return scnprintf(buf, len, "tcp");
5015 }
5016 
5017 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5018 {
5019 	struct nfs4_setclientid *sc = calldata;
5020 
5021 	if (task->tk_status == 0)
5022 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5023 }
5024 
5025 static const struct rpc_call_ops nfs4_setclientid_ops = {
5026 	.rpc_call_done = nfs4_setclientid_done,
5027 };
5028 
5029 /**
5030  * nfs4_proc_setclientid - Negotiate client ID
5031  * @clp: state data structure
5032  * @program: RPC program for NFSv4 callback service
5033  * @port: IP port number for NFS4 callback service
5034  * @cred: RPC credential to use for this call
5035  * @res: where to place the result
5036  *
5037  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5038  */
5039 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5040 		unsigned short port, struct rpc_cred *cred,
5041 		struct nfs4_setclientid_res *res)
5042 {
5043 	nfs4_verifier sc_verifier;
5044 	struct nfs4_setclientid setclientid = {
5045 		.sc_verifier = &sc_verifier,
5046 		.sc_prog = program,
5047 		.sc_cb_ident = clp->cl_cb_ident,
5048 	};
5049 	struct rpc_message msg = {
5050 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5051 		.rpc_argp = &setclientid,
5052 		.rpc_resp = res,
5053 		.rpc_cred = cred,
5054 	};
5055 	struct rpc_task *task;
5056 	struct rpc_task_setup task_setup_data = {
5057 		.rpc_client = clp->cl_rpcclient,
5058 		.rpc_message = &msg,
5059 		.callback_ops = &nfs4_setclientid_ops,
5060 		.callback_data = &setclientid,
5061 		.flags = RPC_TASK_TIMEOUT,
5062 	};
5063 	int status;
5064 
5065 	/* nfs_client_id4 */
5066 	nfs4_init_boot_verifier(clp, &sc_verifier);
5067 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5068 		setclientid.sc_name_len =
5069 				nfs4_init_uniform_client_string(clp,
5070 						setclientid.sc_name,
5071 						sizeof(setclientid.sc_name));
5072 	else
5073 		setclientid.sc_name_len =
5074 				nfs4_init_nonuniform_client_string(clp,
5075 						setclientid.sc_name,
5076 						sizeof(setclientid.sc_name));
5077 	/* cb_client4 */
5078 	setclientid.sc_netid_len =
5079 				nfs4_init_callback_netid(clp,
5080 						setclientid.sc_netid,
5081 						sizeof(setclientid.sc_netid));
5082 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5083 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5084 				clp->cl_ipaddr, port >> 8, port & 255);
5085 
5086 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5087 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5088 		setclientid.sc_name_len, setclientid.sc_name);
5089 	task = rpc_run_task(&task_setup_data);
5090 	if (IS_ERR(task)) {
5091 		status = PTR_ERR(task);
5092 		goto out;
5093 	}
5094 	status = task->tk_status;
5095 	if (setclientid.sc_cred) {
5096 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5097 		put_rpccred(setclientid.sc_cred);
5098 	}
5099 	rpc_put_task(task);
5100 out:
5101 	trace_nfs4_setclientid(clp, status);
5102 	dprintk("NFS reply setclientid: %d\n", status);
5103 	return status;
5104 }
5105 
5106 /**
5107  * nfs4_proc_setclientid_confirm - Confirm client ID
5108  * @clp: state data structure
5109  * @res: result of a previous SETCLIENTID
5110  * @cred: RPC credential to use for this call
5111  *
5112  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5113  */
5114 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5115 		struct nfs4_setclientid_res *arg,
5116 		struct rpc_cred *cred)
5117 {
5118 	struct rpc_message msg = {
5119 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5120 		.rpc_argp = arg,
5121 		.rpc_cred = cred,
5122 	};
5123 	int status;
5124 
5125 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5126 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5127 		clp->cl_clientid);
5128 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5129 	trace_nfs4_setclientid_confirm(clp, status);
5130 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5131 	return status;
5132 }
5133 
5134 struct nfs4_delegreturndata {
5135 	struct nfs4_delegreturnargs args;
5136 	struct nfs4_delegreturnres res;
5137 	struct nfs_fh fh;
5138 	nfs4_stateid stateid;
5139 	unsigned long timestamp;
5140 	struct nfs_fattr fattr;
5141 	int rpc_status;
5142 	struct inode *inode;
5143 	bool roc;
5144 	u32 roc_barrier;
5145 };
5146 
5147 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5148 {
5149 	struct nfs4_delegreturndata *data = calldata;
5150 
5151 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5152 		return;
5153 
5154 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5155 	switch (task->tk_status) {
5156 	case 0:
5157 		renew_lease(data->res.server, data->timestamp);
5158 	case -NFS4ERR_ADMIN_REVOKED:
5159 	case -NFS4ERR_DELEG_REVOKED:
5160 	case -NFS4ERR_BAD_STATEID:
5161 	case -NFS4ERR_OLD_STATEID:
5162 	case -NFS4ERR_STALE_STATEID:
5163 	case -NFS4ERR_EXPIRED:
5164 		task->tk_status = 0;
5165 		if (data->roc)
5166 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5167 		break;
5168 	default:
5169 		if (nfs4_async_handle_error(task, data->res.server,
5170 					    NULL, NULL) == -EAGAIN) {
5171 			rpc_restart_call_prepare(task);
5172 			return;
5173 		}
5174 	}
5175 	data->rpc_status = task->tk_status;
5176 }
5177 
5178 static void nfs4_delegreturn_release(void *calldata)
5179 {
5180 	struct nfs4_delegreturndata *data = calldata;
5181 	struct inode *inode = data->inode;
5182 
5183 	if (inode) {
5184 		if (data->roc)
5185 			pnfs_roc_release(inode);
5186 		nfs_iput_and_deactive(inode);
5187 	}
5188 	kfree(calldata);
5189 }
5190 
5191 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5192 {
5193 	struct nfs4_delegreturndata *d_data;
5194 
5195 	d_data = (struct nfs4_delegreturndata *)data;
5196 
5197 	if (d_data->roc &&
5198 	    pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5199 		return;
5200 
5201 	nfs4_setup_sequence(d_data->res.server,
5202 			&d_data->args.seq_args,
5203 			&d_data->res.seq_res,
5204 			task);
5205 }
5206 
5207 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5208 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5209 	.rpc_call_done = nfs4_delegreturn_done,
5210 	.rpc_release = nfs4_delegreturn_release,
5211 };
5212 
5213 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5214 {
5215 	struct nfs4_delegreturndata *data;
5216 	struct nfs_server *server = NFS_SERVER(inode);
5217 	struct rpc_task *task;
5218 	struct rpc_message msg = {
5219 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5220 		.rpc_cred = cred,
5221 	};
5222 	struct rpc_task_setup task_setup_data = {
5223 		.rpc_client = server->client,
5224 		.rpc_message = &msg,
5225 		.callback_ops = &nfs4_delegreturn_ops,
5226 		.flags = RPC_TASK_ASYNC,
5227 	};
5228 	int status = 0;
5229 
5230 	data = kzalloc(sizeof(*data), GFP_NOFS);
5231 	if (data == NULL)
5232 		return -ENOMEM;
5233 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5234 	data->args.fhandle = &data->fh;
5235 	data->args.stateid = &data->stateid;
5236 	data->args.bitmask = server->cache_consistency_bitmask;
5237 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5238 	nfs4_stateid_copy(&data->stateid, stateid);
5239 	data->res.fattr = &data->fattr;
5240 	data->res.server = server;
5241 	nfs_fattr_init(data->res.fattr);
5242 	data->timestamp = jiffies;
5243 	data->rpc_status = 0;
5244 	data->inode = nfs_igrab_and_active(inode);
5245 	if (data->inode)
5246 		data->roc = nfs4_roc(inode);
5247 
5248 	task_setup_data.callback_data = data;
5249 	msg.rpc_argp = &data->args;
5250 	msg.rpc_resp = &data->res;
5251 	task = rpc_run_task(&task_setup_data);
5252 	if (IS_ERR(task))
5253 		return PTR_ERR(task);
5254 	if (!issync)
5255 		goto out;
5256 	status = nfs4_wait_for_completion_rpc_task(task);
5257 	if (status != 0)
5258 		goto out;
5259 	status = data->rpc_status;
5260 	if (status == 0)
5261 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5262 	else
5263 		nfs_refresh_inode(inode, &data->fattr);
5264 out:
5265 	rpc_put_task(task);
5266 	return status;
5267 }
5268 
5269 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5270 {
5271 	struct nfs_server *server = NFS_SERVER(inode);
5272 	struct nfs4_exception exception = { };
5273 	int err;
5274 	do {
5275 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5276 		trace_nfs4_delegreturn(inode, err);
5277 		switch (err) {
5278 			case -NFS4ERR_STALE_STATEID:
5279 			case -NFS4ERR_EXPIRED:
5280 			case 0:
5281 				return 0;
5282 		}
5283 		err = nfs4_handle_exception(server, err, &exception);
5284 	} while (exception.retry);
5285 	return err;
5286 }
5287 
5288 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5289 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5290 
5291 /*
5292  * sleep, with exponential backoff, and retry the LOCK operation.
5293  */
5294 static unsigned long
5295 nfs4_set_lock_task_retry(unsigned long timeout)
5296 {
5297 	freezable_schedule_timeout_killable_unsafe(timeout);
5298 	timeout <<= 1;
5299 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5300 		return NFS4_LOCK_MAXTIMEOUT;
5301 	return timeout;
5302 }
5303 
5304 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5305 {
5306 	struct inode *inode = state->inode;
5307 	struct nfs_server *server = NFS_SERVER(inode);
5308 	struct nfs_client *clp = server->nfs_client;
5309 	struct nfs_lockt_args arg = {
5310 		.fh = NFS_FH(inode),
5311 		.fl = request,
5312 	};
5313 	struct nfs_lockt_res res = {
5314 		.denied = request,
5315 	};
5316 	struct rpc_message msg = {
5317 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5318 		.rpc_argp       = &arg,
5319 		.rpc_resp       = &res,
5320 		.rpc_cred	= state->owner->so_cred,
5321 	};
5322 	struct nfs4_lock_state *lsp;
5323 	int status;
5324 
5325 	arg.lock_owner.clientid = clp->cl_clientid;
5326 	status = nfs4_set_lock_state(state, request);
5327 	if (status != 0)
5328 		goto out;
5329 	lsp = request->fl_u.nfs4_fl.owner;
5330 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5331 	arg.lock_owner.s_dev = server->s_dev;
5332 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5333 	switch (status) {
5334 		case 0:
5335 			request->fl_type = F_UNLCK;
5336 			break;
5337 		case -NFS4ERR_DENIED:
5338 			status = 0;
5339 	}
5340 	request->fl_ops->fl_release_private(request);
5341 	request->fl_ops = NULL;
5342 out:
5343 	return status;
5344 }
5345 
5346 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5347 {
5348 	struct nfs4_exception exception = { };
5349 	int err;
5350 
5351 	do {
5352 		err = _nfs4_proc_getlk(state, cmd, request);
5353 		trace_nfs4_get_lock(request, state, cmd, err);
5354 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5355 				&exception);
5356 	} while (exception.retry);
5357 	return err;
5358 }
5359 
5360 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5361 {
5362 	int res = 0;
5363 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5364 		case FL_POSIX:
5365 			res = posix_lock_file_wait(file, fl);
5366 			break;
5367 		case FL_FLOCK:
5368 			res = flock_lock_file_wait(file, fl);
5369 			break;
5370 		default:
5371 			BUG();
5372 	}
5373 	return res;
5374 }
5375 
5376 struct nfs4_unlockdata {
5377 	struct nfs_locku_args arg;
5378 	struct nfs_locku_res res;
5379 	struct nfs4_lock_state *lsp;
5380 	struct nfs_open_context *ctx;
5381 	struct file_lock fl;
5382 	const struct nfs_server *server;
5383 	unsigned long timestamp;
5384 };
5385 
5386 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5387 		struct nfs_open_context *ctx,
5388 		struct nfs4_lock_state *lsp,
5389 		struct nfs_seqid *seqid)
5390 {
5391 	struct nfs4_unlockdata *p;
5392 	struct inode *inode = lsp->ls_state->inode;
5393 
5394 	p = kzalloc(sizeof(*p), GFP_NOFS);
5395 	if (p == NULL)
5396 		return NULL;
5397 	p->arg.fh = NFS_FH(inode);
5398 	p->arg.fl = &p->fl;
5399 	p->arg.seqid = seqid;
5400 	p->res.seqid = seqid;
5401 	p->lsp = lsp;
5402 	atomic_inc(&lsp->ls_count);
5403 	/* Ensure we don't close file until we're done freeing locks! */
5404 	p->ctx = get_nfs_open_context(ctx);
5405 	memcpy(&p->fl, fl, sizeof(p->fl));
5406 	p->server = NFS_SERVER(inode);
5407 	return p;
5408 }
5409 
5410 static void nfs4_locku_release_calldata(void *data)
5411 {
5412 	struct nfs4_unlockdata *calldata = data;
5413 	nfs_free_seqid(calldata->arg.seqid);
5414 	nfs4_put_lock_state(calldata->lsp);
5415 	put_nfs_open_context(calldata->ctx);
5416 	kfree(calldata);
5417 }
5418 
5419 static void nfs4_locku_done(struct rpc_task *task, void *data)
5420 {
5421 	struct nfs4_unlockdata *calldata = data;
5422 
5423 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5424 		return;
5425 	switch (task->tk_status) {
5426 		case 0:
5427 			renew_lease(calldata->server, calldata->timestamp);
5428 			do_vfs_lock(calldata->fl.fl_file, &calldata->fl);
5429 			if (nfs4_update_lock_stateid(calldata->lsp,
5430 					&calldata->res.stateid))
5431 				break;
5432 		case -NFS4ERR_BAD_STATEID:
5433 		case -NFS4ERR_OLD_STATEID:
5434 		case -NFS4ERR_STALE_STATEID:
5435 		case -NFS4ERR_EXPIRED:
5436 			if (!nfs4_stateid_match(&calldata->arg.stateid,
5437 						&calldata->lsp->ls_stateid))
5438 				rpc_restart_call_prepare(task);
5439 			break;
5440 		default:
5441 			if (nfs4_async_handle_error(task, calldata->server,
5442 						    NULL, NULL) == -EAGAIN)
5443 				rpc_restart_call_prepare(task);
5444 	}
5445 	nfs_release_seqid(calldata->arg.seqid);
5446 }
5447 
5448 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5449 {
5450 	struct nfs4_unlockdata *calldata = data;
5451 
5452 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5453 		goto out_wait;
5454 	nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5455 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5456 		/* Note: exit _without_ running nfs4_locku_done */
5457 		goto out_no_action;
5458 	}
5459 	calldata->timestamp = jiffies;
5460 	if (nfs4_setup_sequence(calldata->server,
5461 				&calldata->arg.seq_args,
5462 				&calldata->res.seq_res,
5463 				task) != 0)
5464 		nfs_release_seqid(calldata->arg.seqid);
5465 	return;
5466 out_no_action:
5467 	task->tk_action = NULL;
5468 out_wait:
5469 	nfs4_sequence_done(task, &calldata->res.seq_res);
5470 }
5471 
5472 static const struct rpc_call_ops nfs4_locku_ops = {
5473 	.rpc_call_prepare = nfs4_locku_prepare,
5474 	.rpc_call_done = nfs4_locku_done,
5475 	.rpc_release = nfs4_locku_release_calldata,
5476 };
5477 
5478 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5479 		struct nfs_open_context *ctx,
5480 		struct nfs4_lock_state *lsp,
5481 		struct nfs_seqid *seqid)
5482 {
5483 	struct nfs4_unlockdata *data;
5484 	struct rpc_message msg = {
5485 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5486 		.rpc_cred = ctx->cred,
5487 	};
5488 	struct rpc_task_setup task_setup_data = {
5489 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5490 		.rpc_message = &msg,
5491 		.callback_ops = &nfs4_locku_ops,
5492 		.workqueue = nfsiod_workqueue,
5493 		.flags = RPC_TASK_ASYNC,
5494 	};
5495 
5496 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5497 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5498 
5499 	/* Ensure this is an unlock - when canceling a lock, the
5500 	 * canceled lock is passed in, and it won't be an unlock.
5501 	 */
5502 	fl->fl_type = F_UNLCK;
5503 
5504 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5505 	if (data == NULL) {
5506 		nfs_free_seqid(seqid);
5507 		return ERR_PTR(-ENOMEM);
5508 	}
5509 
5510 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5511 	msg.rpc_argp = &data->arg;
5512 	msg.rpc_resp = &data->res;
5513 	task_setup_data.callback_data = data;
5514 	return rpc_run_task(&task_setup_data);
5515 }
5516 
5517 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5518 {
5519 	struct inode *inode = state->inode;
5520 	struct nfs4_state_owner *sp = state->owner;
5521 	struct nfs_inode *nfsi = NFS_I(inode);
5522 	struct nfs_seqid *seqid;
5523 	struct nfs4_lock_state *lsp;
5524 	struct rpc_task *task;
5525 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5526 	int status = 0;
5527 	unsigned char fl_flags = request->fl_flags;
5528 
5529 	status = nfs4_set_lock_state(state, request);
5530 	/* Unlock _before_ we do the RPC call */
5531 	request->fl_flags |= FL_EXISTS;
5532 	/* Exclude nfs_delegation_claim_locks() */
5533 	mutex_lock(&sp->so_delegreturn_mutex);
5534 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5535 	down_read(&nfsi->rwsem);
5536 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5537 		up_read(&nfsi->rwsem);
5538 		mutex_unlock(&sp->so_delegreturn_mutex);
5539 		goto out;
5540 	}
5541 	up_read(&nfsi->rwsem);
5542 	mutex_unlock(&sp->so_delegreturn_mutex);
5543 	if (status != 0)
5544 		goto out;
5545 	/* Is this a delegated lock? */
5546 	lsp = request->fl_u.nfs4_fl.owner;
5547 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5548 		goto out;
5549 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5550 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5551 	status = -ENOMEM;
5552 	if (IS_ERR(seqid))
5553 		goto out;
5554 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5555 	status = PTR_ERR(task);
5556 	if (IS_ERR(task))
5557 		goto out;
5558 	status = nfs4_wait_for_completion_rpc_task(task);
5559 	rpc_put_task(task);
5560 out:
5561 	request->fl_flags = fl_flags;
5562 	trace_nfs4_unlock(request, state, F_SETLK, status);
5563 	return status;
5564 }
5565 
5566 struct nfs4_lockdata {
5567 	struct nfs_lock_args arg;
5568 	struct nfs_lock_res res;
5569 	struct nfs4_lock_state *lsp;
5570 	struct nfs_open_context *ctx;
5571 	struct file_lock fl;
5572 	unsigned long timestamp;
5573 	int rpc_status;
5574 	int cancelled;
5575 	struct nfs_server *server;
5576 };
5577 
5578 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5579 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5580 		gfp_t gfp_mask)
5581 {
5582 	struct nfs4_lockdata *p;
5583 	struct inode *inode = lsp->ls_state->inode;
5584 	struct nfs_server *server = NFS_SERVER(inode);
5585 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5586 
5587 	p = kzalloc(sizeof(*p), gfp_mask);
5588 	if (p == NULL)
5589 		return NULL;
5590 
5591 	p->arg.fh = NFS_FH(inode);
5592 	p->arg.fl = &p->fl;
5593 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5594 	if (IS_ERR(p->arg.open_seqid))
5595 		goto out_free;
5596 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5597 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5598 	if (IS_ERR(p->arg.lock_seqid))
5599 		goto out_free_seqid;
5600 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5601 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5602 	p->arg.lock_owner.s_dev = server->s_dev;
5603 	p->res.lock_seqid = p->arg.lock_seqid;
5604 	p->lsp = lsp;
5605 	p->server = server;
5606 	atomic_inc(&lsp->ls_count);
5607 	p->ctx = get_nfs_open_context(ctx);
5608 	get_file(fl->fl_file);
5609 	memcpy(&p->fl, fl, sizeof(p->fl));
5610 	return p;
5611 out_free_seqid:
5612 	nfs_free_seqid(p->arg.open_seqid);
5613 out_free:
5614 	kfree(p);
5615 	return NULL;
5616 }
5617 
5618 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5619 {
5620 	struct nfs4_lockdata *data = calldata;
5621 	struct nfs4_state *state = data->lsp->ls_state;
5622 
5623 	dprintk("%s: begin!\n", __func__);
5624 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5625 		goto out_wait;
5626 	/* Do we need to do an open_to_lock_owner? */
5627 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5628 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5629 			goto out_release_lock_seqid;
5630 		}
5631 		nfs4_stateid_copy(&data->arg.open_stateid,
5632 				&state->open_stateid);
5633 		data->arg.new_lock_owner = 1;
5634 		data->res.open_seqid = data->arg.open_seqid;
5635 	} else {
5636 		data->arg.new_lock_owner = 0;
5637 		nfs4_stateid_copy(&data->arg.lock_stateid,
5638 				&data->lsp->ls_stateid);
5639 	}
5640 	if (!nfs4_valid_open_stateid(state)) {
5641 		data->rpc_status = -EBADF;
5642 		task->tk_action = NULL;
5643 		goto out_release_open_seqid;
5644 	}
5645 	data->timestamp = jiffies;
5646 	if (nfs4_setup_sequence(data->server,
5647 				&data->arg.seq_args,
5648 				&data->res.seq_res,
5649 				task) == 0)
5650 		return;
5651 out_release_open_seqid:
5652 	nfs_release_seqid(data->arg.open_seqid);
5653 out_release_lock_seqid:
5654 	nfs_release_seqid(data->arg.lock_seqid);
5655 out_wait:
5656 	nfs4_sequence_done(task, &data->res.seq_res);
5657 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5658 }
5659 
5660 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5661 {
5662 	struct nfs4_lockdata *data = calldata;
5663 	struct nfs4_lock_state *lsp = data->lsp;
5664 
5665 	dprintk("%s: begin!\n", __func__);
5666 
5667 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5668 		return;
5669 
5670 	data->rpc_status = task->tk_status;
5671 	switch (task->tk_status) {
5672 	case 0:
5673 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5674 				data->timestamp);
5675 		if (data->arg.new_lock) {
5676 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5677 			if (do_vfs_lock(data->fl.fl_file, &data->fl) < 0) {
5678 				rpc_restart_call_prepare(task);
5679 				break;
5680 			}
5681 		}
5682 		if (data->arg.new_lock_owner != 0) {
5683 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
5684 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5685 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5686 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5687 			rpc_restart_call_prepare(task);
5688 		break;
5689 	case -NFS4ERR_BAD_STATEID:
5690 	case -NFS4ERR_OLD_STATEID:
5691 	case -NFS4ERR_STALE_STATEID:
5692 	case -NFS4ERR_EXPIRED:
5693 		if (data->arg.new_lock_owner != 0) {
5694 			if (!nfs4_stateid_match(&data->arg.open_stateid,
5695 						&lsp->ls_state->open_stateid))
5696 				rpc_restart_call_prepare(task);
5697 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5698 						&lsp->ls_stateid))
5699 				rpc_restart_call_prepare(task);
5700 	}
5701 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5702 }
5703 
5704 static void nfs4_lock_release(void *calldata)
5705 {
5706 	struct nfs4_lockdata *data = calldata;
5707 
5708 	dprintk("%s: begin!\n", __func__);
5709 	nfs_free_seqid(data->arg.open_seqid);
5710 	if (data->cancelled != 0) {
5711 		struct rpc_task *task;
5712 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5713 				data->arg.lock_seqid);
5714 		if (!IS_ERR(task))
5715 			rpc_put_task_async(task);
5716 		dprintk("%s: cancelling lock!\n", __func__);
5717 	} else
5718 		nfs_free_seqid(data->arg.lock_seqid);
5719 	nfs4_put_lock_state(data->lsp);
5720 	put_nfs_open_context(data->ctx);
5721 	fput(data->fl.fl_file);
5722 	kfree(data);
5723 	dprintk("%s: done!\n", __func__);
5724 }
5725 
5726 static const struct rpc_call_ops nfs4_lock_ops = {
5727 	.rpc_call_prepare = nfs4_lock_prepare,
5728 	.rpc_call_done = nfs4_lock_done,
5729 	.rpc_release = nfs4_lock_release,
5730 };
5731 
5732 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5733 {
5734 	switch (error) {
5735 	case -NFS4ERR_ADMIN_REVOKED:
5736 	case -NFS4ERR_BAD_STATEID:
5737 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5738 		if (new_lock_owner != 0 ||
5739 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5740 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5741 		break;
5742 	case -NFS4ERR_STALE_STATEID:
5743 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5744 	case -NFS4ERR_EXPIRED:
5745 		nfs4_schedule_lease_recovery(server->nfs_client);
5746 	};
5747 }
5748 
5749 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5750 {
5751 	struct nfs4_lockdata *data;
5752 	struct rpc_task *task;
5753 	struct rpc_message msg = {
5754 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5755 		.rpc_cred = state->owner->so_cred,
5756 	};
5757 	struct rpc_task_setup task_setup_data = {
5758 		.rpc_client = NFS_CLIENT(state->inode),
5759 		.rpc_message = &msg,
5760 		.callback_ops = &nfs4_lock_ops,
5761 		.workqueue = nfsiod_workqueue,
5762 		.flags = RPC_TASK_ASYNC,
5763 	};
5764 	int ret;
5765 
5766 	dprintk("%s: begin!\n", __func__);
5767 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5768 			fl->fl_u.nfs4_fl.owner,
5769 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5770 	if (data == NULL)
5771 		return -ENOMEM;
5772 	if (IS_SETLKW(cmd))
5773 		data->arg.block = 1;
5774 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5775 	msg.rpc_argp = &data->arg;
5776 	msg.rpc_resp = &data->res;
5777 	task_setup_data.callback_data = data;
5778 	if (recovery_type > NFS_LOCK_NEW) {
5779 		if (recovery_type == NFS_LOCK_RECLAIM)
5780 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5781 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5782 	} else
5783 		data->arg.new_lock = 1;
5784 	task = rpc_run_task(&task_setup_data);
5785 	if (IS_ERR(task))
5786 		return PTR_ERR(task);
5787 	ret = nfs4_wait_for_completion_rpc_task(task);
5788 	if (ret == 0) {
5789 		ret = data->rpc_status;
5790 		if (ret)
5791 			nfs4_handle_setlk_error(data->server, data->lsp,
5792 					data->arg.new_lock_owner, ret);
5793 	} else
5794 		data->cancelled = 1;
5795 	rpc_put_task(task);
5796 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5797 	return ret;
5798 }
5799 
5800 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5801 {
5802 	struct nfs_server *server = NFS_SERVER(state->inode);
5803 	struct nfs4_exception exception = {
5804 		.inode = state->inode,
5805 	};
5806 	int err;
5807 
5808 	do {
5809 		/* Cache the lock if possible... */
5810 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5811 			return 0;
5812 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5813 		trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5814 		if (err != -NFS4ERR_DELAY)
5815 			break;
5816 		nfs4_handle_exception(server, err, &exception);
5817 	} while (exception.retry);
5818 	return err;
5819 }
5820 
5821 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5822 {
5823 	struct nfs_server *server = NFS_SERVER(state->inode);
5824 	struct nfs4_exception exception = {
5825 		.inode = state->inode,
5826 	};
5827 	int err;
5828 
5829 	err = nfs4_set_lock_state(state, request);
5830 	if (err != 0)
5831 		return err;
5832 	if (!recover_lost_locks) {
5833 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5834 		return 0;
5835 	}
5836 	do {
5837 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5838 			return 0;
5839 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5840 		trace_nfs4_lock_expired(request, state, F_SETLK, err);
5841 		switch (err) {
5842 		default:
5843 			goto out;
5844 		case -NFS4ERR_GRACE:
5845 		case -NFS4ERR_DELAY:
5846 			nfs4_handle_exception(server, err, &exception);
5847 			err = 0;
5848 		}
5849 	} while (exception.retry);
5850 out:
5851 	return err;
5852 }
5853 
5854 #if defined(CONFIG_NFS_V4_1)
5855 /**
5856  * nfs41_check_expired_locks - possibly free a lock stateid
5857  *
5858  * @state: NFSv4 state for an inode
5859  *
5860  * Returns NFS_OK if recovery for this stateid is now finished.
5861  * Otherwise a negative NFS4ERR value is returned.
5862  */
5863 static int nfs41_check_expired_locks(struct nfs4_state *state)
5864 {
5865 	int status, ret = -NFS4ERR_BAD_STATEID;
5866 	struct nfs4_lock_state *lsp;
5867 	struct nfs_server *server = NFS_SERVER(state->inode);
5868 
5869 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5870 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5871 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5872 
5873 			status = nfs41_test_stateid(server,
5874 					&lsp->ls_stateid,
5875 					cred);
5876 			trace_nfs4_test_lock_stateid(state, lsp, status);
5877 			if (status != NFS_OK) {
5878 				/* Free the stateid unless the server
5879 				 * informs us the stateid is unrecognized. */
5880 				if (status != -NFS4ERR_BAD_STATEID)
5881 					nfs41_free_stateid(server,
5882 							&lsp->ls_stateid,
5883 							cred);
5884 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5885 				ret = status;
5886 			}
5887 		}
5888 	};
5889 
5890 	return ret;
5891 }
5892 
5893 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5894 {
5895 	int status = NFS_OK;
5896 
5897 	if (test_bit(LK_STATE_IN_USE, &state->flags))
5898 		status = nfs41_check_expired_locks(state);
5899 	if (status != NFS_OK)
5900 		status = nfs4_lock_expired(state, request);
5901 	return status;
5902 }
5903 #endif
5904 
5905 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5906 {
5907 	struct nfs_inode *nfsi = NFS_I(state->inode);
5908 	unsigned char fl_flags = request->fl_flags;
5909 	int status = -ENOLCK;
5910 
5911 	if ((fl_flags & FL_POSIX) &&
5912 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5913 		goto out;
5914 	/* Is this a delegated open? */
5915 	status = nfs4_set_lock_state(state, request);
5916 	if (status != 0)
5917 		goto out;
5918 	request->fl_flags |= FL_ACCESS;
5919 	status = do_vfs_lock(request->fl_file, request);
5920 	if (status < 0)
5921 		goto out;
5922 	down_read(&nfsi->rwsem);
5923 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5924 		/* Yes: cache locks! */
5925 		/* ...but avoid races with delegation recall... */
5926 		request->fl_flags = fl_flags & ~FL_SLEEP;
5927 		status = do_vfs_lock(request->fl_file, request);
5928 		up_read(&nfsi->rwsem);
5929 		goto out;
5930 	}
5931 	up_read(&nfsi->rwsem);
5932 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5933 out:
5934 	request->fl_flags = fl_flags;
5935 	return status;
5936 }
5937 
5938 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5939 {
5940 	struct nfs4_exception exception = {
5941 		.state = state,
5942 		.inode = state->inode,
5943 	};
5944 	int err;
5945 
5946 	do {
5947 		err = _nfs4_proc_setlk(state, cmd, request);
5948 		trace_nfs4_set_lock(request, state, cmd, err);
5949 		if (err == -NFS4ERR_DENIED)
5950 			err = -EAGAIN;
5951 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
5952 				err, &exception);
5953 	} while (exception.retry);
5954 	return err;
5955 }
5956 
5957 static int
5958 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5959 {
5960 	struct nfs_open_context *ctx;
5961 	struct nfs4_state *state;
5962 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5963 	int status;
5964 
5965 	/* verify open state */
5966 	ctx = nfs_file_open_context(filp);
5967 	state = ctx->state;
5968 
5969 	if (request->fl_start < 0 || request->fl_end < 0)
5970 		return -EINVAL;
5971 
5972 	if (IS_GETLK(cmd)) {
5973 		if (state != NULL)
5974 			return nfs4_proc_getlk(state, F_GETLK, request);
5975 		return 0;
5976 	}
5977 
5978 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5979 		return -EINVAL;
5980 
5981 	if (request->fl_type == F_UNLCK) {
5982 		if (state != NULL)
5983 			return nfs4_proc_unlck(state, cmd, request);
5984 		return 0;
5985 	}
5986 
5987 	if (state == NULL)
5988 		return -ENOLCK;
5989 	/*
5990 	 * Don't rely on the VFS having checked the file open mode,
5991 	 * since it won't do this for flock() locks.
5992 	 */
5993 	switch (request->fl_type) {
5994 	case F_RDLCK:
5995 		if (!(filp->f_mode & FMODE_READ))
5996 			return -EBADF;
5997 		break;
5998 	case F_WRLCK:
5999 		if (!(filp->f_mode & FMODE_WRITE))
6000 			return -EBADF;
6001 	}
6002 
6003 	do {
6004 		status = nfs4_proc_setlk(state, cmd, request);
6005 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6006 			break;
6007 		timeout = nfs4_set_lock_task_retry(timeout);
6008 		status = -ERESTARTSYS;
6009 		if (signalled())
6010 			break;
6011 	} while(status < 0);
6012 	return status;
6013 }
6014 
6015 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6016 {
6017 	struct nfs_server *server = NFS_SERVER(state->inode);
6018 	int err;
6019 
6020 	err = nfs4_set_lock_state(state, fl);
6021 	if (err != 0)
6022 		return err;
6023 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6024 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6025 }
6026 
6027 struct nfs_release_lockowner_data {
6028 	struct nfs4_lock_state *lsp;
6029 	struct nfs_server *server;
6030 	struct nfs_release_lockowner_args args;
6031 	struct nfs_release_lockowner_res res;
6032 	unsigned long timestamp;
6033 };
6034 
6035 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6036 {
6037 	struct nfs_release_lockowner_data *data = calldata;
6038 	struct nfs_server *server = data->server;
6039 	nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6040 			     &data->args.seq_args, &data->res.seq_res, task);
6041 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6042 	data->timestamp = jiffies;
6043 }
6044 
6045 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6046 {
6047 	struct nfs_release_lockowner_data *data = calldata;
6048 	struct nfs_server *server = data->server;
6049 
6050 	nfs40_sequence_done(task, &data->res.seq_res);
6051 
6052 	switch (task->tk_status) {
6053 	case 0:
6054 		renew_lease(server, data->timestamp);
6055 		break;
6056 	case -NFS4ERR_STALE_CLIENTID:
6057 	case -NFS4ERR_EXPIRED:
6058 		nfs4_schedule_lease_recovery(server->nfs_client);
6059 		break;
6060 	case -NFS4ERR_LEASE_MOVED:
6061 	case -NFS4ERR_DELAY:
6062 		if (nfs4_async_handle_error(task, server,
6063 					    NULL, NULL) == -EAGAIN)
6064 			rpc_restart_call_prepare(task);
6065 	}
6066 }
6067 
6068 static void nfs4_release_lockowner_release(void *calldata)
6069 {
6070 	struct nfs_release_lockowner_data *data = calldata;
6071 	nfs4_free_lock_state(data->server, data->lsp);
6072 	kfree(calldata);
6073 }
6074 
6075 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6076 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6077 	.rpc_call_done = nfs4_release_lockowner_done,
6078 	.rpc_release = nfs4_release_lockowner_release,
6079 };
6080 
6081 static void
6082 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6083 {
6084 	struct nfs_release_lockowner_data *data;
6085 	struct rpc_message msg = {
6086 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6087 	};
6088 
6089 	if (server->nfs_client->cl_mvops->minor_version != 0)
6090 		return;
6091 
6092 	data = kmalloc(sizeof(*data), GFP_NOFS);
6093 	if (!data)
6094 		return;
6095 	data->lsp = lsp;
6096 	data->server = server;
6097 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6098 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6099 	data->args.lock_owner.s_dev = server->s_dev;
6100 
6101 	msg.rpc_argp = &data->args;
6102 	msg.rpc_resp = &data->res;
6103 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6104 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6105 }
6106 
6107 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6108 
6109 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6110 				   const void *buf, size_t buflen,
6111 				   int flags, int type)
6112 {
6113 	if (strcmp(key, "") != 0)
6114 		return -EINVAL;
6115 
6116 	return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6117 }
6118 
6119 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6120 				   void *buf, size_t buflen, int type)
6121 {
6122 	if (strcmp(key, "") != 0)
6123 		return -EINVAL;
6124 
6125 	return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6126 }
6127 
6128 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6129 				       size_t list_len, const char *name,
6130 				       size_t name_len, int type)
6131 {
6132 	size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6133 
6134 	if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6135 		return 0;
6136 
6137 	if (list && len <= list_len)
6138 		memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6139 	return len;
6140 }
6141 
6142 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6143 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6144 {
6145 	return server->caps & NFS_CAP_SECURITY_LABEL;
6146 }
6147 
6148 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6149 				   const void *buf, size_t buflen,
6150 				   int flags, int type)
6151 {
6152 	if (security_ismaclabel(key))
6153 		return nfs4_set_security_label(dentry, buf, buflen);
6154 
6155 	return -EOPNOTSUPP;
6156 }
6157 
6158 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6159 				   void *buf, size_t buflen, int type)
6160 {
6161 	if (security_ismaclabel(key))
6162 		return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6163 	return -EOPNOTSUPP;
6164 }
6165 
6166 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6167 				       size_t list_len, const char *name,
6168 				       size_t name_len, int type)
6169 {
6170 	size_t len = 0;
6171 
6172 	if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6173 		len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6174 		if (list && len <= list_len)
6175 			security_inode_listsecurity(d_inode(dentry), list, len);
6176 	}
6177 	return len;
6178 }
6179 
6180 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6181 	.prefix = XATTR_SECURITY_PREFIX,
6182 	.list	= nfs4_xattr_list_nfs4_label,
6183 	.get	= nfs4_xattr_get_nfs4_label,
6184 	.set	= nfs4_xattr_set_nfs4_label,
6185 };
6186 #endif
6187 
6188 
6189 /*
6190  * nfs_fhget will use either the mounted_on_fileid or the fileid
6191  */
6192 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6193 {
6194 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6195 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6196 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6197 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6198 		return;
6199 
6200 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6201 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6202 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6203 	fattr->nlink = 2;
6204 }
6205 
6206 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6207 				   const struct qstr *name,
6208 				   struct nfs4_fs_locations *fs_locations,
6209 				   struct page *page)
6210 {
6211 	struct nfs_server *server = NFS_SERVER(dir);
6212 	u32 bitmask[3] = {
6213 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6214 	};
6215 	struct nfs4_fs_locations_arg args = {
6216 		.dir_fh = NFS_FH(dir),
6217 		.name = name,
6218 		.page = page,
6219 		.bitmask = bitmask,
6220 	};
6221 	struct nfs4_fs_locations_res res = {
6222 		.fs_locations = fs_locations,
6223 	};
6224 	struct rpc_message msg = {
6225 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6226 		.rpc_argp = &args,
6227 		.rpc_resp = &res,
6228 	};
6229 	int status;
6230 
6231 	dprintk("%s: start\n", __func__);
6232 
6233 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6234 	 * is not supported */
6235 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6236 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6237 	else
6238 		bitmask[0] |= FATTR4_WORD0_FILEID;
6239 
6240 	nfs_fattr_init(&fs_locations->fattr);
6241 	fs_locations->server = server;
6242 	fs_locations->nlocations = 0;
6243 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6244 	dprintk("%s: returned status = %d\n", __func__, status);
6245 	return status;
6246 }
6247 
6248 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6249 			   const struct qstr *name,
6250 			   struct nfs4_fs_locations *fs_locations,
6251 			   struct page *page)
6252 {
6253 	struct nfs4_exception exception = { };
6254 	int err;
6255 	do {
6256 		err = _nfs4_proc_fs_locations(client, dir, name,
6257 				fs_locations, page);
6258 		trace_nfs4_get_fs_locations(dir, name, err);
6259 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6260 				&exception);
6261 	} while (exception.retry);
6262 	return err;
6263 }
6264 
6265 /*
6266  * This operation also signals the server that this client is
6267  * performing migration recovery.  The server can stop returning
6268  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6269  * appended to this compound to identify the client ID which is
6270  * performing recovery.
6271  */
6272 static int _nfs40_proc_get_locations(struct inode *inode,
6273 				     struct nfs4_fs_locations *locations,
6274 				     struct page *page, struct rpc_cred *cred)
6275 {
6276 	struct nfs_server *server = NFS_SERVER(inode);
6277 	struct rpc_clnt *clnt = server->client;
6278 	u32 bitmask[2] = {
6279 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6280 	};
6281 	struct nfs4_fs_locations_arg args = {
6282 		.clientid	= server->nfs_client->cl_clientid,
6283 		.fh		= NFS_FH(inode),
6284 		.page		= page,
6285 		.bitmask	= bitmask,
6286 		.migration	= 1,		/* skip LOOKUP */
6287 		.renew		= 1,		/* append RENEW */
6288 	};
6289 	struct nfs4_fs_locations_res res = {
6290 		.fs_locations	= locations,
6291 		.migration	= 1,
6292 		.renew		= 1,
6293 	};
6294 	struct rpc_message msg = {
6295 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6296 		.rpc_argp	= &args,
6297 		.rpc_resp	= &res,
6298 		.rpc_cred	= cred,
6299 	};
6300 	unsigned long now = jiffies;
6301 	int status;
6302 
6303 	nfs_fattr_init(&locations->fattr);
6304 	locations->server = server;
6305 	locations->nlocations = 0;
6306 
6307 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6308 	nfs4_set_sequence_privileged(&args.seq_args);
6309 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6310 					&args.seq_args, &res.seq_res);
6311 	if (status)
6312 		return status;
6313 
6314 	renew_lease(server, now);
6315 	return 0;
6316 }
6317 
6318 #ifdef CONFIG_NFS_V4_1
6319 
6320 /*
6321  * This operation also signals the server that this client is
6322  * performing migration recovery.  The server can stop asserting
6323  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6324  * performing this operation is identified in the SEQUENCE
6325  * operation in this compound.
6326  *
6327  * When the client supports GETATTR(fs_locations_info), it can
6328  * be plumbed in here.
6329  */
6330 static int _nfs41_proc_get_locations(struct inode *inode,
6331 				     struct nfs4_fs_locations *locations,
6332 				     struct page *page, struct rpc_cred *cred)
6333 {
6334 	struct nfs_server *server = NFS_SERVER(inode);
6335 	struct rpc_clnt *clnt = server->client;
6336 	u32 bitmask[2] = {
6337 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6338 	};
6339 	struct nfs4_fs_locations_arg args = {
6340 		.fh		= NFS_FH(inode),
6341 		.page		= page,
6342 		.bitmask	= bitmask,
6343 		.migration	= 1,		/* skip LOOKUP */
6344 	};
6345 	struct nfs4_fs_locations_res res = {
6346 		.fs_locations	= locations,
6347 		.migration	= 1,
6348 	};
6349 	struct rpc_message msg = {
6350 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6351 		.rpc_argp	= &args,
6352 		.rpc_resp	= &res,
6353 		.rpc_cred	= cred,
6354 	};
6355 	int status;
6356 
6357 	nfs_fattr_init(&locations->fattr);
6358 	locations->server = server;
6359 	locations->nlocations = 0;
6360 
6361 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6362 	nfs4_set_sequence_privileged(&args.seq_args);
6363 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6364 					&args.seq_args, &res.seq_res);
6365 	if (status == NFS4_OK &&
6366 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6367 		status = -NFS4ERR_LEASE_MOVED;
6368 	return status;
6369 }
6370 
6371 #endif	/* CONFIG_NFS_V4_1 */
6372 
6373 /**
6374  * nfs4_proc_get_locations - discover locations for a migrated FSID
6375  * @inode: inode on FSID that is migrating
6376  * @locations: result of query
6377  * @page: buffer
6378  * @cred: credential to use for this operation
6379  *
6380  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6381  * operation failed, or a negative errno if a local error occurred.
6382  *
6383  * On success, "locations" is filled in, but if the server has
6384  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6385  * asserted.
6386  *
6387  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6388  * from this client that require migration recovery.
6389  */
6390 int nfs4_proc_get_locations(struct inode *inode,
6391 			    struct nfs4_fs_locations *locations,
6392 			    struct page *page, struct rpc_cred *cred)
6393 {
6394 	struct nfs_server *server = NFS_SERVER(inode);
6395 	struct nfs_client *clp = server->nfs_client;
6396 	const struct nfs4_mig_recovery_ops *ops =
6397 					clp->cl_mvops->mig_recovery_ops;
6398 	struct nfs4_exception exception = { };
6399 	int status;
6400 
6401 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6402 		(unsigned long long)server->fsid.major,
6403 		(unsigned long long)server->fsid.minor,
6404 		clp->cl_hostname);
6405 	nfs_display_fhandle(NFS_FH(inode), __func__);
6406 
6407 	do {
6408 		status = ops->get_locations(inode, locations, page, cred);
6409 		if (status != -NFS4ERR_DELAY)
6410 			break;
6411 		nfs4_handle_exception(server, status, &exception);
6412 	} while (exception.retry);
6413 	return status;
6414 }
6415 
6416 /*
6417  * This operation also signals the server that this client is
6418  * performing "lease moved" recovery.  The server can stop
6419  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6420  * is appended to this compound to identify the client ID which is
6421  * performing recovery.
6422  */
6423 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6424 {
6425 	struct nfs_server *server = NFS_SERVER(inode);
6426 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6427 	struct rpc_clnt *clnt = server->client;
6428 	struct nfs4_fsid_present_arg args = {
6429 		.fh		= NFS_FH(inode),
6430 		.clientid	= clp->cl_clientid,
6431 		.renew		= 1,		/* append RENEW */
6432 	};
6433 	struct nfs4_fsid_present_res res = {
6434 		.renew		= 1,
6435 	};
6436 	struct rpc_message msg = {
6437 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6438 		.rpc_argp	= &args,
6439 		.rpc_resp	= &res,
6440 		.rpc_cred	= cred,
6441 	};
6442 	unsigned long now = jiffies;
6443 	int status;
6444 
6445 	res.fh = nfs_alloc_fhandle();
6446 	if (res.fh == NULL)
6447 		return -ENOMEM;
6448 
6449 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6450 	nfs4_set_sequence_privileged(&args.seq_args);
6451 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6452 						&args.seq_args, &res.seq_res);
6453 	nfs_free_fhandle(res.fh);
6454 	if (status)
6455 		return status;
6456 
6457 	do_renew_lease(clp, now);
6458 	return 0;
6459 }
6460 
6461 #ifdef CONFIG_NFS_V4_1
6462 
6463 /*
6464  * This operation also signals the server that this client is
6465  * performing "lease moved" recovery.  The server can stop asserting
6466  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6467  * this operation is identified in the SEQUENCE operation in this
6468  * compound.
6469  */
6470 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6471 {
6472 	struct nfs_server *server = NFS_SERVER(inode);
6473 	struct rpc_clnt *clnt = server->client;
6474 	struct nfs4_fsid_present_arg args = {
6475 		.fh		= NFS_FH(inode),
6476 	};
6477 	struct nfs4_fsid_present_res res = {
6478 	};
6479 	struct rpc_message msg = {
6480 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6481 		.rpc_argp	= &args,
6482 		.rpc_resp	= &res,
6483 		.rpc_cred	= cred,
6484 	};
6485 	int status;
6486 
6487 	res.fh = nfs_alloc_fhandle();
6488 	if (res.fh == NULL)
6489 		return -ENOMEM;
6490 
6491 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6492 	nfs4_set_sequence_privileged(&args.seq_args);
6493 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6494 						&args.seq_args, &res.seq_res);
6495 	nfs_free_fhandle(res.fh);
6496 	if (status == NFS4_OK &&
6497 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6498 		status = -NFS4ERR_LEASE_MOVED;
6499 	return status;
6500 }
6501 
6502 #endif	/* CONFIG_NFS_V4_1 */
6503 
6504 /**
6505  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6506  * @inode: inode on FSID to check
6507  * @cred: credential to use for this operation
6508  *
6509  * Server indicates whether the FSID is present, moved, or not
6510  * recognized.  This operation is necessary to clear a LEASE_MOVED
6511  * condition for this client ID.
6512  *
6513  * Returns NFS4_OK if the FSID is present on this server,
6514  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6515  *  NFS4ERR code if some error occurred on the server, or a
6516  *  negative errno if a local failure occurred.
6517  */
6518 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6519 {
6520 	struct nfs_server *server = NFS_SERVER(inode);
6521 	struct nfs_client *clp = server->nfs_client;
6522 	const struct nfs4_mig_recovery_ops *ops =
6523 					clp->cl_mvops->mig_recovery_ops;
6524 	struct nfs4_exception exception = { };
6525 	int status;
6526 
6527 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6528 		(unsigned long long)server->fsid.major,
6529 		(unsigned long long)server->fsid.minor,
6530 		clp->cl_hostname);
6531 	nfs_display_fhandle(NFS_FH(inode), __func__);
6532 
6533 	do {
6534 		status = ops->fsid_present(inode, cred);
6535 		if (status != -NFS4ERR_DELAY)
6536 			break;
6537 		nfs4_handle_exception(server, status, &exception);
6538 	} while (exception.retry);
6539 	return status;
6540 }
6541 
6542 /**
6543  * If 'use_integrity' is true and the state managment nfs_client
6544  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6545  * and the machine credential as per RFC3530bis and RFC5661 Security
6546  * Considerations sections. Otherwise, just use the user cred with the
6547  * filesystem's rpc_client.
6548  */
6549 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6550 {
6551 	int status;
6552 	struct nfs4_secinfo_arg args = {
6553 		.dir_fh = NFS_FH(dir),
6554 		.name   = name,
6555 	};
6556 	struct nfs4_secinfo_res res = {
6557 		.flavors     = flavors,
6558 	};
6559 	struct rpc_message msg = {
6560 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6561 		.rpc_argp = &args,
6562 		.rpc_resp = &res,
6563 	};
6564 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6565 	struct rpc_cred *cred = NULL;
6566 
6567 	if (use_integrity) {
6568 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6569 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6570 		msg.rpc_cred = cred;
6571 	}
6572 
6573 	dprintk("NFS call  secinfo %s\n", name->name);
6574 
6575 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6576 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6577 
6578 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6579 				&res.seq_res, 0);
6580 	dprintk("NFS reply  secinfo: %d\n", status);
6581 
6582 	if (cred)
6583 		put_rpccred(cred);
6584 
6585 	return status;
6586 }
6587 
6588 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6589 		      struct nfs4_secinfo_flavors *flavors)
6590 {
6591 	struct nfs4_exception exception = { };
6592 	int err;
6593 	do {
6594 		err = -NFS4ERR_WRONGSEC;
6595 
6596 		/* try to use integrity protection with machine cred */
6597 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6598 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6599 
6600 		/*
6601 		 * if unable to use integrity protection, or SECINFO with
6602 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6603 		 * disallowed by spec, but exists in deployed servers) use
6604 		 * the current filesystem's rpc_client and the user cred.
6605 		 */
6606 		if (err == -NFS4ERR_WRONGSEC)
6607 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6608 
6609 		trace_nfs4_secinfo(dir, name, err);
6610 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6611 				&exception);
6612 	} while (exception.retry);
6613 	return err;
6614 }
6615 
6616 #ifdef CONFIG_NFS_V4_1
6617 /*
6618  * Check the exchange flags returned by the server for invalid flags, having
6619  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6620  * DS flags set.
6621  */
6622 static int nfs4_check_cl_exchange_flags(u32 flags)
6623 {
6624 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6625 		goto out_inval;
6626 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6627 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6628 		goto out_inval;
6629 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6630 		goto out_inval;
6631 	return NFS_OK;
6632 out_inval:
6633 	return -NFS4ERR_INVAL;
6634 }
6635 
6636 static bool
6637 nfs41_same_server_scope(struct nfs41_server_scope *a,
6638 			struct nfs41_server_scope *b)
6639 {
6640 	if (a->server_scope_sz == b->server_scope_sz &&
6641 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6642 		return true;
6643 
6644 	return false;
6645 }
6646 
6647 /*
6648  * nfs4_proc_bind_conn_to_session()
6649  *
6650  * The 4.1 client currently uses the same TCP connection for the
6651  * fore and backchannel.
6652  */
6653 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6654 {
6655 	int status;
6656 	struct nfs41_bind_conn_to_session_args args = {
6657 		.client = clp,
6658 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
6659 	};
6660 	struct nfs41_bind_conn_to_session_res res;
6661 	struct rpc_message msg = {
6662 		.rpc_proc =
6663 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6664 		.rpc_argp = &args,
6665 		.rpc_resp = &res,
6666 		.rpc_cred = cred,
6667 	};
6668 
6669 	dprintk("--> %s\n", __func__);
6670 
6671 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6672 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6673 		args.dir = NFS4_CDFC4_FORE;
6674 
6675 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6676 	trace_nfs4_bind_conn_to_session(clp, status);
6677 	if (status == 0) {
6678 		if (memcmp(res.sessionid.data,
6679 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6680 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6681 			status = -EIO;
6682 			goto out;
6683 		}
6684 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6685 			dprintk("NFS: %s: Unexpected direction from server\n",
6686 				__func__);
6687 			status = -EIO;
6688 			goto out;
6689 		}
6690 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6691 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6692 				__func__);
6693 			status = -EIO;
6694 			goto out;
6695 		}
6696 	}
6697 out:
6698 	dprintk("<-- %s status= %d\n", __func__, status);
6699 	return status;
6700 }
6701 
6702 /*
6703  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6704  * and operations we'd like to see to enable certain features in the allow map
6705  */
6706 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6707 	.how = SP4_MACH_CRED,
6708 	.enforce.u.words = {
6709 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6710 		      1 << (OP_EXCHANGE_ID - 32) |
6711 		      1 << (OP_CREATE_SESSION - 32) |
6712 		      1 << (OP_DESTROY_SESSION - 32) |
6713 		      1 << (OP_DESTROY_CLIENTID - 32)
6714 	},
6715 	.allow.u.words = {
6716 		[0] = 1 << (OP_CLOSE) |
6717 		      1 << (OP_LOCKU) |
6718 		      1 << (OP_COMMIT),
6719 		[1] = 1 << (OP_SECINFO - 32) |
6720 		      1 << (OP_SECINFO_NO_NAME - 32) |
6721 		      1 << (OP_TEST_STATEID - 32) |
6722 		      1 << (OP_FREE_STATEID - 32) |
6723 		      1 << (OP_WRITE - 32)
6724 	}
6725 };
6726 
6727 /*
6728  * Select the state protection mode for client `clp' given the server results
6729  * from exchange_id in `sp'.
6730  *
6731  * Returns 0 on success, negative errno otherwise.
6732  */
6733 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6734 				 struct nfs41_state_protection *sp)
6735 {
6736 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6737 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6738 		      1 << (OP_EXCHANGE_ID - 32) |
6739 		      1 << (OP_CREATE_SESSION - 32) |
6740 		      1 << (OP_DESTROY_SESSION - 32) |
6741 		      1 << (OP_DESTROY_CLIENTID - 32)
6742 	};
6743 	unsigned int i;
6744 
6745 	if (sp->how == SP4_MACH_CRED) {
6746 		/* Print state protect result */
6747 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6748 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6749 			if (test_bit(i, sp->enforce.u.longs))
6750 				dfprintk(MOUNT, "  enforce op %d\n", i);
6751 			if (test_bit(i, sp->allow.u.longs))
6752 				dfprintk(MOUNT, "  allow op %d\n", i);
6753 		}
6754 
6755 		/* make sure nothing is on enforce list that isn't supported */
6756 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6757 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6758 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6759 				return -EINVAL;
6760 			}
6761 		}
6762 
6763 		/*
6764 		 * Minimal mode - state operations are allowed to use machine
6765 		 * credential.  Note this already happens by default, so the
6766 		 * client doesn't have to do anything more than the negotiation.
6767 		 *
6768 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6769 		 *       we're already using the machine cred for exchange_id
6770 		 *       and will never use a different cred.
6771 		 */
6772 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6773 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6774 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6775 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6776 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6777 			dfprintk(MOUNT, "  minimal mode enabled\n");
6778 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6779 		} else {
6780 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6781 			return -EINVAL;
6782 		}
6783 
6784 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6785 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
6786 			dfprintk(MOUNT, "  cleanup mode enabled\n");
6787 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6788 		}
6789 
6790 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6791 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6792 			dfprintk(MOUNT, "  secinfo mode enabled\n");
6793 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6794 		}
6795 
6796 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6797 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6798 			dfprintk(MOUNT, "  stateid mode enabled\n");
6799 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6800 		}
6801 
6802 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6803 			dfprintk(MOUNT, "  write mode enabled\n");
6804 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6805 		}
6806 
6807 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6808 			dfprintk(MOUNT, "  commit mode enabled\n");
6809 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6810 		}
6811 	}
6812 
6813 	return 0;
6814 }
6815 
6816 /*
6817  * _nfs4_proc_exchange_id()
6818  *
6819  * Wrapper for EXCHANGE_ID operation.
6820  */
6821 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6822 	u32 sp4_how)
6823 {
6824 	nfs4_verifier verifier;
6825 	struct nfs41_exchange_id_args args = {
6826 		.verifier = &verifier,
6827 		.client = clp,
6828 #ifdef CONFIG_NFS_V4_1_MIGRATION
6829 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6830 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
6831 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6832 #else
6833 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6834 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
6835 #endif
6836 	};
6837 	struct nfs41_exchange_id_res res = {
6838 		0
6839 	};
6840 	int status;
6841 	struct rpc_message msg = {
6842 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6843 		.rpc_argp = &args,
6844 		.rpc_resp = &res,
6845 		.rpc_cred = cred,
6846 	};
6847 
6848 	nfs4_init_boot_verifier(clp, &verifier);
6849 	args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6850 							sizeof(args.id));
6851 	dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6852 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6853 		args.id_len, args.id);
6854 
6855 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6856 					GFP_NOFS);
6857 	if (unlikely(res.server_owner == NULL)) {
6858 		status = -ENOMEM;
6859 		goto out;
6860 	}
6861 
6862 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6863 					GFP_NOFS);
6864 	if (unlikely(res.server_scope == NULL)) {
6865 		status = -ENOMEM;
6866 		goto out_server_owner;
6867 	}
6868 
6869 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6870 	if (unlikely(res.impl_id == NULL)) {
6871 		status = -ENOMEM;
6872 		goto out_server_scope;
6873 	}
6874 
6875 	switch (sp4_how) {
6876 	case SP4_NONE:
6877 		args.state_protect.how = SP4_NONE;
6878 		break;
6879 
6880 	case SP4_MACH_CRED:
6881 		args.state_protect = nfs4_sp4_mach_cred_request;
6882 		break;
6883 
6884 	default:
6885 		/* unsupported! */
6886 		WARN_ON_ONCE(1);
6887 		status = -EINVAL;
6888 		goto out_server_scope;
6889 	}
6890 
6891 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6892 	trace_nfs4_exchange_id(clp, status);
6893 	if (status == 0)
6894 		status = nfs4_check_cl_exchange_flags(res.flags);
6895 
6896 	if (status == 0)
6897 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
6898 
6899 	if (status == 0) {
6900 		clp->cl_clientid = res.clientid;
6901 		clp->cl_exchange_flags = res.flags;
6902 		/* Client ID is not confirmed */
6903 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
6904 			clear_bit(NFS4_SESSION_ESTABLISHED,
6905 					&clp->cl_session->session_state);
6906 			clp->cl_seqid = res.seqid;
6907 		}
6908 
6909 		kfree(clp->cl_serverowner);
6910 		clp->cl_serverowner = res.server_owner;
6911 		res.server_owner = NULL;
6912 
6913 		/* use the most recent implementation id */
6914 		kfree(clp->cl_implid);
6915 		clp->cl_implid = res.impl_id;
6916 
6917 		if (clp->cl_serverscope != NULL &&
6918 		    !nfs41_same_server_scope(clp->cl_serverscope,
6919 					     res.server_scope)) {
6920 			dprintk("%s: server_scope mismatch detected\n",
6921 				__func__);
6922 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6923 			kfree(clp->cl_serverscope);
6924 			clp->cl_serverscope = NULL;
6925 		}
6926 
6927 		if (clp->cl_serverscope == NULL) {
6928 			clp->cl_serverscope = res.server_scope;
6929 			goto out;
6930 		}
6931 	} else
6932 		kfree(res.impl_id);
6933 
6934 out_server_owner:
6935 	kfree(res.server_owner);
6936 out_server_scope:
6937 	kfree(res.server_scope);
6938 out:
6939 	if (clp->cl_implid != NULL)
6940 		dprintk("NFS reply exchange_id: Server Implementation ID: "
6941 			"domain: %s, name: %s, date: %llu,%u\n",
6942 			clp->cl_implid->domain, clp->cl_implid->name,
6943 			clp->cl_implid->date.seconds,
6944 			clp->cl_implid->date.nseconds);
6945 	dprintk("NFS reply exchange_id: %d\n", status);
6946 	return status;
6947 }
6948 
6949 /*
6950  * nfs4_proc_exchange_id()
6951  *
6952  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6953  *
6954  * Since the clientid has expired, all compounds using sessions
6955  * associated with the stale clientid will be returning
6956  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6957  * be in some phase of session reset.
6958  *
6959  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6960  */
6961 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6962 {
6963 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6964 	int status;
6965 
6966 	/* try SP4_MACH_CRED if krb5i/p	*/
6967 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
6968 	    authflavor == RPC_AUTH_GSS_KRB5P) {
6969 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6970 		if (!status)
6971 			return 0;
6972 	}
6973 
6974 	/* try SP4_NONE */
6975 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6976 }
6977 
6978 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6979 		struct rpc_cred *cred)
6980 {
6981 	struct rpc_message msg = {
6982 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6983 		.rpc_argp = clp,
6984 		.rpc_cred = cred,
6985 	};
6986 	int status;
6987 
6988 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6989 	trace_nfs4_destroy_clientid(clp, status);
6990 	if (status)
6991 		dprintk("NFS: Got error %d from the server %s on "
6992 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
6993 	return status;
6994 }
6995 
6996 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6997 		struct rpc_cred *cred)
6998 {
6999 	unsigned int loop;
7000 	int ret;
7001 
7002 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7003 		ret = _nfs4_proc_destroy_clientid(clp, cred);
7004 		switch (ret) {
7005 		case -NFS4ERR_DELAY:
7006 		case -NFS4ERR_CLIENTID_BUSY:
7007 			ssleep(1);
7008 			break;
7009 		default:
7010 			return ret;
7011 		}
7012 	}
7013 	return 0;
7014 }
7015 
7016 int nfs4_destroy_clientid(struct nfs_client *clp)
7017 {
7018 	struct rpc_cred *cred;
7019 	int ret = 0;
7020 
7021 	if (clp->cl_mvops->minor_version < 1)
7022 		goto out;
7023 	if (clp->cl_exchange_flags == 0)
7024 		goto out;
7025 	if (clp->cl_preserve_clid)
7026 		goto out;
7027 	cred = nfs4_get_clid_cred(clp);
7028 	ret = nfs4_proc_destroy_clientid(clp, cred);
7029 	if (cred)
7030 		put_rpccred(cred);
7031 	switch (ret) {
7032 	case 0:
7033 	case -NFS4ERR_STALE_CLIENTID:
7034 		clp->cl_exchange_flags = 0;
7035 	}
7036 out:
7037 	return ret;
7038 }
7039 
7040 struct nfs4_get_lease_time_data {
7041 	struct nfs4_get_lease_time_args *args;
7042 	struct nfs4_get_lease_time_res *res;
7043 	struct nfs_client *clp;
7044 };
7045 
7046 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7047 					void *calldata)
7048 {
7049 	struct nfs4_get_lease_time_data *data =
7050 			(struct nfs4_get_lease_time_data *)calldata;
7051 
7052 	dprintk("--> %s\n", __func__);
7053 	/* just setup sequence, do not trigger session recovery
7054 	   since we're invoked within one */
7055 	nfs41_setup_sequence(data->clp->cl_session,
7056 			&data->args->la_seq_args,
7057 			&data->res->lr_seq_res,
7058 			task);
7059 	dprintk("<-- %s\n", __func__);
7060 }
7061 
7062 /*
7063  * Called from nfs4_state_manager thread for session setup, so don't recover
7064  * from sequence operation or clientid errors.
7065  */
7066 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7067 {
7068 	struct nfs4_get_lease_time_data *data =
7069 			(struct nfs4_get_lease_time_data *)calldata;
7070 
7071 	dprintk("--> %s\n", __func__);
7072 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7073 		return;
7074 	switch (task->tk_status) {
7075 	case -NFS4ERR_DELAY:
7076 	case -NFS4ERR_GRACE:
7077 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7078 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
7079 		task->tk_status = 0;
7080 		/* fall through */
7081 	case -NFS4ERR_RETRY_UNCACHED_REP:
7082 		rpc_restart_call_prepare(task);
7083 		return;
7084 	}
7085 	dprintk("<-- %s\n", __func__);
7086 }
7087 
7088 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7089 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
7090 	.rpc_call_done = nfs4_get_lease_time_done,
7091 };
7092 
7093 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7094 {
7095 	struct rpc_task *task;
7096 	struct nfs4_get_lease_time_args args;
7097 	struct nfs4_get_lease_time_res res = {
7098 		.lr_fsinfo = fsinfo,
7099 	};
7100 	struct nfs4_get_lease_time_data data = {
7101 		.args = &args,
7102 		.res = &res,
7103 		.clp = clp,
7104 	};
7105 	struct rpc_message msg = {
7106 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7107 		.rpc_argp = &args,
7108 		.rpc_resp = &res,
7109 	};
7110 	struct rpc_task_setup task_setup = {
7111 		.rpc_client = clp->cl_rpcclient,
7112 		.rpc_message = &msg,
7113 		.callback_ops = &nfs4_get_lease_time_ops,
7114 		.callback_data = &data,
7115 		.flags = RPC_TASK_TIMEOUT,
7116 	};
7117 	int status;
7118 
7119 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7120 	nfs4_set_sequence_privileged(&args.la_seq_args);
7121 	dprintk("--> %s\n", __func__);
7122 	task = rpc_run_task(&task_setup);
7123 
7124 	if (IS_ERR(task))
7125 		status = PTR_ERR(task);
7126 	else {
7127 		status = task->tk_status;
7128 		rpc_put_task(task);
7129 	}
7130 	dprintk("<-- %s return %d\n", __func__, status);
7131 
7132 	return status;
7133 }
7134 
7135 /*
7136  * Initialize the values to be used by the client in CREATE_SESSION
7137  * If nfs4_init_session set the fore channel request and response sizes,
7138  * use them.
7139  *
7140  * Set the back channel max_resp_sz_cached to zero to force the client to
7141  * always set csa_cachethis to FALSE because the current implementation
7142  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7143  */
7144 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7145 {
7146 	unsigned int max_rqst_sz, max_resp_sz;
7147 
7148 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7149 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7150 
7151 	/* Fore channel attributes */
7152 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7153 	args->fc_attrs.max_resp_sz = max_resp_sz;
7154 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7155 	args->fc_attrs.max_reqs = max_session_slots;
7156 
7157 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7158 		"max_ops=%u max_reqs=%u\n",
7159 		__func__,
7160 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7161 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7162 
7163 	/* Back channel attributes */
7164 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7165 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
7166 	args->bc_attrs.max_resp_sz_cached = 0;
7167 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7168 	args->bc_attrs.max_reqs = 1;
7169 
7170 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7171 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7172 		__func__,
7173 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7174 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7175 		args->bc_attrs.max_reqs);
7176 }
7177 
7178 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7179 		struct nfs41_create_session_res *res)
7180 {
7181 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7182 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7183 
7184 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7185 		return -EINVAL;
7186 	/*
7187 	 * Our requested max_ops is the minimum we need; we're not
7188 	 * prepared to break up compounds into smaller pieces than that.
7189 	 * So, no point even trying to continue if the server won't
7190 	 * cooperate:
7191 	 */
7192 	if (rcvd->max_ops < sent->max_ops)
7193 		return -EINVAL;
7194 	if (rcvd->max_reqs == 0)
7195 		return -EINVAL;
7196 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7197 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7198 	return 0;
7199 }
7200 
7201 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7202 		struct nfs41_create_session_res *res)
7203 {
7204 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7205 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7206 
7207 	if (!(res->flags & SESSION4_BACK_CHAN))
7208 		goto out;
7209 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7210 		return -EINVAL;
7211 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7212 		return -EINVAL;
7213 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7214 		return -EINVAL;
7215 	/* These would render the backchannel useless: */
7216 	if (rcvd->max_ops != sent->max_ops)
7217 		return -EINVAL;
7218 	if (rcvd->max_reqs != sent->max_reqs)
7219 		return -EINVAL;
7220 out:
7221 	return 0;
7222 }
7223 
7224 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7225 				     struct nfs41_create_session_res *res)
7226 {
7227 	int ret;
7228 
7229 	ret = nfs4_verify_fore_channel_attrs(args, res);
7230 	if (ret)
7231 		return ret;
7232 	return nfs4_verify_back_channel_attrs(args, res);
7233 }
7234 
7235 static void nfs4_update_session(struct nfs4_session *session,
7236 		struct nfs41_create_session_res *res)
7237 {
7238 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7239 	/* Mark client id and session as being confirmed */
7240 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7241 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7242 	session->flags = res->flags;
7243 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7244 	if (res->flags & SESSION4_BACK_CHAN)
7245 		memcpy(&session->bc_attrs, &res->bc_attrs,
7246 				sizeof(session->bc_attrs));
7247 }
7248 
7249 static int _nfs4_proc_create_session(struct nfs_client *clp,
7250 		struct rpc_cred *cred)
7251 {
7252 	struct nfs4_session *session = clp->cl_session;
7253 	struct nfs41_create_session_args args = {
7254 		.client = clp,
7255 		.clientid = clp->cl_clientid,
7256 		.seqid = clp->cl_seqid,
7257 		.cb_program = NFS4_CALLBACK,
7258 	};
7259 	struct nfs41_create_session_res res;
7260 
7261 	struct rpc_message msg = {
7262 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7263 		.rpc_argp = &args,
7264 		.rpc_resp = &res,
7265 		.rpc_cred = cred,
7266 	};
7267 	int status;
7268 
7269 	nfs4_init_channel_attrs(&args);
7270 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7271 
7272 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7273 	trace_nfs4_create_session(clp, status);
7274 
7275 	if (!status) {
7276 		/* Verify the session's negotiated channel_attrs values */
7277 		status = nfs4_verify_channel_attrs(&args, &res);
7278 		/* Increment the clientid slot sequence id */
7279 		if (clp->cl_seqid == res.seqid)
7280 			clp->cl_seqid++;
7281 		if (status)
7282 			goto out;
7283 		nfs4_update_session(session, &res);
7284 	}
7285 out:
7286 	return status;
7287 }
7288 
7289 /*
7290  * Issues a CREATE_SESSION operation to the server.
7291  * It is the responsibility of the caller to verify the session is
7292  * expired before calling this routine.
7293  */
7294 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7295 {
7296 	int status;
7297 	unsigned *ptr;
7298 	struct nfs4_session *session = clp->cl_session;
7299 
7300 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7301 
7302 	status = _nfs4_proc_create_session(clp, cred);
7303 	if (status)
7304 		goto out;
7305 
7306 	/* Init or reset the session slot tables */
7307 	status = nfs4_setup_session_slot_tables(session);
7308 	dprintk("slot table setup returned %d\n", status);
7309 	if (status)
7310 		goto out;
7311 
7312 	ptr = (unsigned *)&session->sess_id.data[0];
7313 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7314 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7315 out:
7316 	dprintk("<-- %s\n", __func__);
7317 	return status;
7318 }
7319 
7320 /*
7321  * Issue the over-the-wire RPC DESTROY_SESSION.
7322  * The caller must serialize access to this routine.
7323  */
7324 int nfs4_proc_destroy_session(struct nfs4_session *session,
7325 		struct rpc_cred *cred)
7326 {
7327 	struct rpc_message msg = {
7328 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7329 		.rpc_argp = session,
7330 		.rpc_cred = cred,
7331 	};
7332 	int status = 0;
7333 
7334 	dprintk("--> nfs4_proc_destroy_session\n");
7335 
7336 	/* session is still being setup */
7337 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7338 		return 0;
7339 
7340 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7341 	trace_nfs4_destroy_session(session->clp, status);
7342 
7343 	if (status)
7344 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7345 			"Session has been destroyed regardless...\n", status);
7346 
7347 	dprintk("<-- nfs4_proc_destroy_session\n");
7348 	return status;
7349 }
7350 
7351 /*
7352  * Renew the cl_session lease.
7353  */
7354 struct nfs4_sequence_data {
7355 	struct nfs_client *clp;
7356 	struct nfs4_sequence_args args;
7357 	struct nfs4_sequence_res res;
7358 };
7359 
7360 static void nfs41_sequence_release(void *data)
7361 {
7362 	struct nfs4_sequence_data *calldata = data;
7363 	struct nfs_client *clp = calldata->clp;
7364 
7365 	if (atomic_read(&clp->cl_count) > 1)
7366 		nfs4_schedule_state_renewal(clp);
7367 	nfs_put_client(clp);
7368 	kfree(calldata);
7369 }
7370 
7371 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7372 {
7373 	switch(task->tk_status) {
7374 	case -NFS4ERR_DELAY:
7375 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7376 		return -EAGAIN;
7377 	default:
7378 		nfs4_schedule_lease_recovery(clp);
7379 	}
7380 	return 0;
7381 }
7382 
7383 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7384 {
7385 	struct nfs4_sequence_data *calldata = data;
7386 	struct nfs_client *clp = calldata->clp;
7387 
7388 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7389 		return;
7390 
7391 	trace_nfs4_sequence(clp, task->tk_status);
7392 	if (task->tk_status < 0) {
7393 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7394 		if (atomic_read(&clp->cl_count) == 1)
7395 			goto out;
7396 
7397 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7398 			rpc_restart_call_prepare(task);
7399 			return;
7400 		}
7401 	}
7402 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7403 out:
7404 	dprintk("<-- %s\n", __func__);
7405 }
7406 
7407 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7408 {
7409 	struct nfs4_sequence_data *calldata = data;
7410 	struct nfs_client *clp = calldata->clp;
7411 	struct nfs4_sequence_args *args;
7412 	struct nfs4_sequence_res *res;
7413 
7414 	args = task->tk_msg.rpc_argp;
7415 	res = task->tk_msg.rpc_resp;
7416 
7417 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7418 }
7419 
7420 static const struct rpc_call_ops nfs41_sequence_ops = {
7421 	.rpc_call_done = nfs41_sequence_call_done,
7422 	.rpc_call_prepare = nfs41_sequence_prepare,
7423 	.rpc_release = nfs41_sequence_release,
7424 };
7425 
7426 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7427 		struct rpc_cred *cred,
7428 		bool is_privileged)
7429 {
7430 	struct nfs4_sequence_data *calldata;
7431 	struct rpc_message msg = {
7432 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7433 		.rpc_cred = cred,
7434 	};
7435 	struct rpc_task_setup task_setup_data = {
7436 		.rpc_client = clp->cl_rpcclient,
7437 		.rpc_message = &msg,
7438 		.callback_ops = &nfs41_sequence_ops,
7439 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7440 	};
7441 
7442 	if (!atomic_inc_not_zero(&clp->cl_count))
7443 		return ERR_PTR(-EIO);
7444 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7445 	if (calldata == NULL) {
7446 		nfs_put_client(clp);
7447 		return ERR_PTR(-ENOMEM);
7448 	}
7449 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7450 	if (is_privileged)
7451 		nfs4_set_sequence_privileged(&calldata->args);
7452 	msg.rpc_argp = &calldata->args;
7453 	msg.rpc_resp = &calldata->res;
7454 	calldata->clp = clp;
7455 	task_setup_data.callback_data = calldata;
7456 
7457 	return rpc_run_task(&task_setup_data);
7458 }
7459 
7460 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7461 {
7462 	struct rpc_task *task;
7463 	int ret = 0;
7464 
7465 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7466 		return -EAGAIN;
7467 	task = _nfs41_proc_sequence(clp, cred, false);
7468 	if (IS_ERR(task))
7469 		ret = PTR_ERR(task);
7470 	else
7471 		rpc_put_task_async(task);
7472 	dprintk("<-- %s status=%d\n", __func__, ret);
7473 	return ret;
7474 }
7475 
7476 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7477 {
7478 	struct rpc_task *task;
7479 	int ret;
7480 
7481 	task = _nfs41_proc_sequence(clp, cred, true);
7482 	if (IS_ERR(task)) {
7483 		ret = PTR_ERR(task);
7484 		goto out;
7485 	}
7486 	ret = rpc_wait_for_completion_task(task);
7487 	if (!ret) {
7488 		struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7489 
7490 		if (task->tk_status == 0)
7491 			nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7492 		ret = task->tk_status;
7493 	}
7494 	rpc_put_task(task);
7495 out:
7496 	dprintk("<-- %s status=%d\n", __func__, ret);
7497 	return ret;
7498 }
7499 
7500 struct nfs4_reclaim_complete_data {
7501 	struct nfs_client *clp;
7502 	struct nfs41_reclaim_complete_args arg;
7503 	struct nfs41_reclaim_complete_res res;
7504 };
7505 
7506 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7507 {
7508 	struct nfs4_reclaim_complete_data *calldata = data;
7509 
7510 	nfs41_setup_sequence(calldata->clp->cl_session,
7511 			&calldata->arg.seq_args,
7512 			&calldata->res.seq_res,
7513 			task);
7514 }
7515 
7516 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7517 {
7518 	switch(task->tk_status) {
7519 	case 0:
7520 	case -NFS4ERR_COMPLETE_ALREADY:
7521 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7522 		break;
7523 	case -NFS4ERR_DELAY:
7524 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7525 		/* fall through */
7526 	case -NFS4ERR_RETRY_UNCACHED_REP:
7527 		return -EAGAIN;
7528 	default:
7529 		nfs4_schedule_lease_recovery(clp);
7530 	}
7531 	return 0;
7532 }
7533 
7534 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7535 {
7536 	struct nfs4_reclaim_complete_data *calldata = data;
7537 	struct nfs_client *clp = calldata->clp;
7538 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7539 
7540 	dprintk("--> %s\n", __func__);
7541 	if (!nfs41_sequence_done(task, res))
7542 		return;
7543 
7544 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7545 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7546 		rpc_restart_call_prepare(task);
7547 		return;
7548 	}
7549 	dprintk("<-- %s\n", __func__);
7550 }
7551 
7552 static void nfs4_free_reclaim_complete_data(void *data)
7553 {
7554 	struct nfs4_reclaim_complete_data *calldata = data;
7555 
7556 	kfree(calldata);
7557 }
7558 
7559 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7560 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7561 	.rpc_call_done = nfs4_reclaim_complete_done,
7562 	.rpc_release = nfs4_free_reclaim_complete_data,
7563 };
7564 
7565 /*
7566  * Issue a global reclaim complete.
7567  */
7568 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7569 		struct rpc_cred *cred)
7570 {
7571 	struct nfs4_reclaim_complete_data *calldata;
7572 	struct rpc_task *task;
7573 	struct rpc_message msg = {
7574 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7575 		.rpc_cred = cred,
7576 	};
7577 	struct rpc_task_setup task_setup_data = {
7578 		.rpc_client = clp->cl_rpcclient,
7579 		.rpc_message = &msg,
7580 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7581 		.flags = RPC_TASK_ASYNC,
7582 	};
7583 	int status = -ENOMEM;
7584 
7585 	dprintk("--> %s\n", __func__);
7586 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7587 	if (calldata == NULL)
7588 		goto out;
7589 	calldata->clp = clp;
7590 	calldata->arg.one_fs = 0;
7591 
7592 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7593 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7594 	msg.rpc_argp = &calldata->arg;
7595 	msg.rpc_resp = &calldata->res;
7596 	task_setup_data.callback_data = calldata;
7597 	task = rpc_run_task(&task_setup_data);
7598 	if (IS_ERR(task)) {
7599 		status = PTR_ERR(task);
7600 		goto out;
7601 	}
7602 	status = nfs4_wait_for_completion_rpc_task(task);
7603 	if (status == 0)
7604 		status = task->tk_status;
7605 	rpc_put_task(task);
7606 	return 0;
7607 out:
7608 	dprintk("<-- %s status=%d\n", __func__, status);
7609 	return status;
7610 }
7611 
7612 static void
7613 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7614 {
7615 	struct nfs4_layoutget *lgp = calldata;
7616 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7617 	struct nfs4_session *session = nfs4_get_session(server);
7618 
7619 	dprintk("--> %s\n", __func__);
7620 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
7621 	 * right now covering the LAYOUTGET we are about to send.
7622 	 * However, that is not so catastrophic, and there seems
7623 	 * to be no way to prevent it completely.
7624 	 */
7625 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7626 				&lgp->res.seq_res, task))
7627 		return;
7628 	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7629 					  NFS_I(lgp->args.inode)->layout,
7630 					  &lgp->args.range,
7631 					  lgp->args.ctx->state)) {
7632 		rpc_exit(task, NFS4_OK);
7633 	}
7634 }
7635 
7636 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7637 {
7638 	struct nfs4_layoutget *lgp = calldata;
7639 	struct inode *inode = lgp->args.inode;
7640 	struct nfs_server *server = NFS_SERVER(inode);
7641 	struct pnfs_layout_hdr *lo;
7642 	struct nfs4_state *state = NULL;
7643 	unsigned long timeo, now, giveup;
7644 
7645 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7646 
7647 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7648 		goto out;
7649 
7650 	switch (task->tk_status) {
7651 	case 0:
7652 		goto out;
7653 	/*
7654 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7655 	 * (or clients) writing to the same RAID stripe
7656 	 */
7657 	case -NFS4ERR_LAYOUTTRYLATER:
7658 	/*
7659 	 * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7660 	 * existing layout before getting a new one).
7661 	 */
7662 	case -NFS4ERR_RECALLCONFLICT:
7663 		timeo = rpc_get_timeout(task->tk_client);
7664 		giveup = lgp->args.timestamp + timeo;
7665 		now = jiffies;
7666 		if (time_after(giveup, now)) {
7667 			unsigned long delay;
7668 
7669 			/* Delay for:
7670 			 * - Not less then NFS4_POLL_RETRY_MIN.
7671 			 * - One last time a jiffie before we give up
7672 			 * - exponential backoff (time_now minus start_attempt)
7673 			 */
7674 			delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7675 				    min((giveup - now - 1),
7676 					now - lgp->args.timestamp));
7677 
7678 			dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7679 				__func__, delay);
7680 			rpc_delay(task, delay);
7681 			task->tk_status = 0;
7682 			rpc_restart_call_prepare(task);
7683 			goto out; /* Do not call nfs4_async_handle_error() */
7684 		}
7685 		break;
7686 	case -NFS4ERR_EXPIRED:
7687 	case -NFS4ERR_BAD_STATEID:
7688 		spin_lock(&inode->i_lock);
7689 		lo = NFS_I(inode)->layout;
7690 		if (!lo || list_empty(&lo->plh_segs)) {
7691 			spin_unlock(&inode->i_lock);
7692 			/* If the open stateid was bad, then recover it. */
7693 			state = lgp->args.ctx->state;
7694 		} else {
7695 			LIST_HEAD(head);
7696 
7697 			/*
7698 			 * Mark the bad layout state as invalid, then retry
7699 			 * with the current stateid.
7700 			 */
7701 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7702 			spin_unlock(&inode->i_lock);
7703 			pnfs_free_lseg_list(&head);
7704 
7705 			task->tk_status = 0;
7706 			rpc_restart_call_prepare(task);
7707 		}
7708 	}
7709 	if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7710 		rpc_restart_call_prepare(task);
7711 out:
7712 	dprintk("<-- %s\n", __func__);
7713 }
7714 
7715 static size_t max_response_pages(struct nfs_server *server)
7716 {
7717 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7718 	return nfs_page_array_len(0, max_resp_sz);
7719 }
7720 
7721 static void nfs4_free_pages(struct page **pages, size_t size)
7722 {
7723 	int i;
7724 
7725 	if (!pages)
7726 		return;
7727 
7728 	for (i = 0; i < size; i++) {
7729 		if (!pages[i])
7730 			break;
7731 		__free_page(pages[i]);
7732 	}
7733 	kfree(pages);
7734 }
7735 
7736 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7737 {
7738 	struct page **pages;
7739 	int i;
7740 
7741 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7742 	if (!pages) {
7743 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7744 		return NULL;
7745 	}
7746 
7747 	for (i = 0; i < size; i++) {
7748 		pages[i] = alloc_page(gfp_flags);
7749 		if (!pages[i]) {
7750 			dprintk("%s: failed to allocate page\n", __func__);
7751 			nfs4_free_pages(pages, size);
7752 			return NULL;
7753 		}
7754 	}
7755 
7756 	return pages;
7757 }
7758 
7759 static void nfs4_layoutget_release(void *calldata)
7760 {
7761 	struct nfs4_layoutget *lgp = calldata;
7762 	struct inode *inode = lgp->args.inode;
7763 	struct nfs_server *server = NFS_SERVER(inode);
7764 	size_t max_pages = max_response_pages(server);
7765 
7766 	dprintk("--> %s\n", __func__);
7767 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
7768 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
7769 	put_nfs_open_context(lgp->args.ctx);
7770 	kfree(calldata);
7771 	dprintk("<-- %s\n", __func__);
7772 }
7773 
7774 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7775 	.rpc_call_prepare = nfs4_layoutget_prepare,
7776 	.rpc_call_done = nfs4_layoutget_done,
7777 	.rpc_release = nfs4_layoutget_release,
7778 };
7779 
7780 struct pnfs_layout_segment *
7781 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7782 {
7783 	struct inode *inode = lgp->args.inode;
7784 	struct nfs_server *server = NFS_SERVER(inode);
7785 	size_t max_pages = max_response_pages(server);
7786 	struct rpc_task *task;
7787 	struct rpc_message msg = {
7788 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7789 		.rpc_argp = &lgp->args,
7790 		.rpc_resp = &lgp->res,
7791 		.rpc_cred = lgp->cred,
7792 	};
7793 	struct rpc_task_setup task_setup_data = {
7794 		.rpc_client = server->client,
7795 		.rpc_message = &msg,
7796 		.callback_ops = &nfs4_layoutget_call_ops,
7797 		.callback_data = lgp,
7798 		.flags = RPC_TASK_ASYNC,
7799 	};
7800 	struct pnfs_layout_segment *lseg = NULL;
7801 	int status = 0;
7802 
7803 	dprintk("--> %s\n", __func__);
7804 
7805 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7806 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
7807 
7808 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7809 	if (!lgp->args.layout.pages) {
7810 		nfs4_layoutget_release(lgp);
7811 		return ERR_PTR(-ENOMEM);
7812 	}
7813 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7814 	lgp->args.timestamp = jiffies;
7815 
7816 	lgp->res.layoutp = &lgp->args.layout;
7817 	lgp->res.seq_res.sr_slot = NULL;
7818 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7819 
7820 	task = rpc_run_task(&task_setup_data);
7821 	if (IS_ERR(task))
7822 		return ERR_CAST(task);
7823 	status = nfs4_wait_for_completion_rpc_task(task);
7824 	if (status == 0)
7825 		status = task->tk_status;
7826 	trace_nfs4_layoutget(lgp->args.ctx,
7827 			&lgp->args.range,
7828 			&lgp->res.range,
7829 			status);
7830 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7831 	if (status == 0 && lgp->res.layoutp->len)
7832 		lseg = pnfs_layout_process(lgp);
7833 	rpc_put_task(task);
7834 	dprintk("<-- %s status=%d\n", __func__, status);
7835 	if (status)
7836 		return ERR_PTR(status);
7837 	return lseg;
7838 }
7839 
7840 static void
7841 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7842 {
7843 	struct nfs4_layoutreturn *lrp = calldata;
7844 
7845 	dprintk("--> %s\n", __func__);
7846 	nfs41_setup_sequence(lrp->clp->cl_session,
7847 			&lrp->args.seq_args,
7848 			&lrp->res.seq_res,
7849 			task);
7850 }
7851 
7852 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7853 {
7854 	struct nfs4_layoutreturn *lrp = calldata;
7855 	struct nfs_server *server;
7856 
7857 	dprintk("--> %s\n", __func__);
7858 
7859 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7860 		return;
7861 
7862 	server = NFS_SERVER(lrp->args.inode);
7863 	switch (task->tk_status) {
7864 	default:
7865 		task->tk_status = 0;
7866 	case 0:
7867 		break;
7868 	case -NFS4ERR_DELAY:
7869 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7870 			break;
7871 		rpc_restart_call_prepare(task);
7872 		return;
7873 	}
7874 	dprintk("<-- %s\n", __func__);
7875 }
7876 
7877 static void nfs4_layoutreturn_release(void *calldata)
7878 {
7879 	struct nfs4_layoutreturn *lrp = calldata;
7880 	struct pnfs_layout_hdr *lo = lrp->args.layout;
7881 
7882 	dprintk("--> %s\n", __func__);
7883 	spin_lock(&lo->plh_inode->i_lock);
7884 	if (lrp->res.lrs_present)
7885 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7886 	pnfs_clear_layoutreturn_waitbit(lo);
7887 	clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, &lo->plh_flags);
7888 	rpc_wake_up(&NFS_SERVER(lo->plh_inode)->roc_rpcwaitq);
7889 	lo->plh_block_lgets--;
7890 	spin_unlock(&lo->plh_inode->i_lock);
7891 	pnfs_put_layout_hdr(lrp->args.layout);
7892 	nfs_iput_and_deactive(lrp->inode);
7893 	kfree(calldata);
7894 	dprintk("<-- %s\n", __func__);
7895 }
7896 
7897 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7898 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
7899 	.rpc_call_done = nfs4_layoutreturn_done,
7900 	.rpc_release = nfs4_layoutreturn_release,
7901 };
7902 
7903 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
7904 {
7905 	struct rpc_task *task;
7906 	struct rpc_message msg = {
7907 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7908 		.rpc_argp = &lrp->args,
7909 		.rpc_resp = &lrp->res,
7910 		.rpc_cred = lrp->cred,
7911 	};
7912 	struct rpc_task_setup task_setup_data = {
7913 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
7914 		.rpc_message = &msg,
7915 		.callback_ops = &nfs4_layoutreturn_call_ops,
7916 		.callback_data = lrp,
7917 	};
7918 	int status = 0;
7919 
7920 	dprintk("--> %s\n", __func__);
7921 	if (!sync) {
7922 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
7923 		if (!lrp->inode) {
7924 			nfs4_layoutreturn_release(lrp);
7925 			return -EAGAIN;
7926 		}
7927 		task_setup_data.flags |= RPC_TASK_ASYNC;
7928 	}
7929 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7930 	task = rpc_run_task(&task_setup_data);
7931 	if (IS_ERR(task))
7932 		return PTR_ERR(task);
7933 	if (sync)
7934 		status = task->tk_status;
7935 	trace_nfs4_layoutreturn(lrp->args.inode, status);
7936 	dprintk("<-- %s status=%d\n", __func__, status);
7937 	rpc_put_task(task);
7938 	return status;
7939 }
7940 
7941 static int
7942 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7943 		struct pnfs_device *pdev,
7944 		struct rpc_cred *cred)
7945 {
7946 	struct nfs4_getdeviceinfo_args args = {
7947 		.pdev = pdev,
7948 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
7949 			NOTIFY_DEVICEID4_DELETE,
7950 	};
7951 	struct nfs4_getdeviceinfo_res res = {
7952 		.pdev = pdev,
7953 	};
7954 	struct rpc_message msg = {
7955 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7956 		.rpc_argp = &args,
7957 		.rpc_resp = &res,
7958 		.rpc_cred = cred,
7959 	};
7960 	int status;
7961 
7962 	dprintk("--> %s\n", __func__);
7963 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7964 	if (res.notification & ~args.notify_types)
7965 		dprintk("%s: unsupported notification\n", __func__);
7966 	if (res.notification != args.notify_types)
7967 		pdev->nocache = 1;
7968 
7969 	dprintk("<-- %s status=%d\n", __func__, status);
7970 
7971 	return status;
7972 }
7973 
7974 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7975 		struct pnfs_device *pdev,
7976 		struct rpc_cred *cred)
7977 {
7978 	struct nfs4_exception exception = { };
7979 	int err;
7980 
7981 	do {
7982 		err = nfs4_handle_exception(server,
7983 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
7984 					&exception);
7985 	} while (exception.retry);
7986 	return err;
7987 }
7988 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7989 
7990 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7991 {
7992 	struct nfs4_layoutcommit_data *data = calldata;
7993 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7994 	struct nfs4_session *session = nfs4_get_session(server);
7995 
7996 	nfs41_setup_sequence(session,
7997 			&data->args.seq_args,
7998 			&data->res.seq_res,
7999 			task);
8000 }
8001 
8002 static void
8003 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8004 {
8005 	struct nfs4_layoutcommit_data *data = calldata;
8006 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8007 
8008 	if (!nfs41_sequence_done(task, &data->res.seq_res))
8009 		return;
8010 
8011 	switch (task->tk_status) { /* Just ignore these failures */
8012 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8013 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8014 	case -NFS4ERR_BADLAYOUT:     /* no layout */
8015 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
8016 		task->tk_status = 0;
8017 	case 0:
8018 		break;
8019 	default:
8020 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8021 			rpc_restart_call_prepare(task);
8022 			return;
8023 		}
8024 	}
8025 }
8026 
8027 static void nfs4_layoutcommit_release(void *calldata)
8028 {
8029 	struct nfs4_layoutcommit_data *data = calldata;
8030 
8031 	pnfs_cleanup_layoutcommit(data);
8032 	nfs_post_op_update_inode_force_wcc(data->args.inode,
8033 					   data->res.fattr);
8034 	put_rpccred(data->cred);
8035 	nfs_iput_and_deactive(data->inode);
8036 	kfree(data);
8037 }
8038 
8039 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8040 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
8041 	.rpc_call_done = nfs4_layoutcommit_done,
8042 	.rpc_release = nfs4_layoutcommit_release,
8043 };
8044 
8045 int
8046 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8047 {
8048 	struct rpc_message msg = {
8049 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8050 		.rpc_argp = &data->args,
8051 		.rpc_resp = &data->res,
8052 		.rpc_cred = data->cred,
8053 	};
8054 	struct rpc_task_setup task_setup_data = {
8055 		.task = &data->task,
8056 		.rpc_client = NFS_CLIENT(data->args.inode),
8057 		.rpc_message = &msg,
8058 		.callback_ops = &nfs4_layoutcommit_ops,
8059 		.callback_data = data,
8060 	};
8061 	struct rpc_task *task;
8062 	int status = 0;
8063 
8064 	dprintk("NFS: %4d initiating layoutcommit call. sync %d "
8065 		"lbw: %llu inode %lu\n",
8066 		data->task.tk_pid, sync,
8067 		data->args.lastbytewritten,
8068 		data->args.inode->i_ino);
8069 
8070 	if (!sync) {
8071 		data->inode = nfs_igrab_and_active(data->args.inode);
8072 		if (data->inode == NULL) {
8073 			nfs4_layoutcommit_release(data);
8074 			return -EAGAIN;
8075 		}
8076 		task_setup_data.flags = RPC_TASK_ASYNC;
8077 	}
8078 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8079 	task = rpc_run_task(&task_setup_data);
8080 	if (IS_ERR(task))
8081 		return PTR_ERR(task);
8082 	if (sync)
8083 		status = task->tk_status;
8084 	trace_nfs4_layoutcommit(data->args.inode, status);
8085 	dprintk("%s: status %d\n", __func__, status);
8086 	rpc_put_task(task);
8087 	return status;
8088 }
8089 
8090 /**
8091  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8092  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8093  */
8094 static int
8095 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8096 		    struct nfs_fsinfo *info,
8097 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8098 {
8099 	struct nfs41_secinfo_no_name_args args = {
8100 		.style = SECINFO_STYLE_CURRENT_FH,
8101 	};
8102 	struct nfs4_secinfo_res res = {
8103 		.flavors = flavors,
8104 	};
8105 	struct rpc_message msg = {
8106 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8107 		.rpc_argp = &args,
8108 		.rpc_resp = &res,
8109 	};
8110 	struct rpc_clnt *clnt = server->client;
8111 	struct rpc_cred *cred = NULL;
8112 	int status;
8113 
8114 	if (use_integrity) {
8115 		clnt = server->nfs_client->cl_rpcclient;
8116 		cred = nfs4_get_clid_cred(server->nfs_client);
8117 		msg.rpc_cred = cred;
8118 	}
8119 
8120 	dprintk("--> %s\n", __func__);
8121 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8122 				&res.seq_res, 0);
8123 	dprintk("<-- %s status=%d\n", __func__, status);
8124 
8125 	if (cred)
8126 		put_rpccred(cred);
8127 
8128 	return status;
8129 }
8130 
8131 static int
8132 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8133 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8134 {
8135 	struct nfs4_exception exception = { };
8136 	int err;
8137 	do {
8138 		/* first try using integrity protection */
8139 		err = -NFS4ERR_WRONGSEC;
8140 
8141 		/* try to use integrity protection with machine cred */
8142 		if (_nfs4_is_integrity_protected(server->nfs_client))
8143 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8144 							  flavors, true);
8145 
8146 		/*
8147 		 * if unable to use integrity protection, or SECINFO with
8148 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8149 		 * disallowed by spec, but exists in deployed servers) use
8150 		 * the current filesystem's rpc_client and the user cred.
8151 		 */
8152 		if (err == -NFS4ERR_WRONGSEC)
8153 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8154 							  flavors, false);
8155 
8156 		switch (err) {
8157 		case 0:
8158 		case -NFS4ERR_WRONGSEC:
8159 		case -ENOTSUPP:
8160 			goto out;
8161 		default:
8162 			err = nfs4_handle_exception(server, err, &exception);
8163 		}
8164 	} while (exception.retry);
8165 out:
8166 	return err;
8167 }
8168 
8169 static int
8170 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8171 		    struct nfs_fsinfo *info)
8172 {
8173 	int err;
8174 	struct page *page;
8175 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8176 	struct nfs4_secinfo_flavors *flavors;
8177 	struct nfs4_secinfo4 *secinfo;
8178 	int i;
8179 
8180 	page = alloc_page(GFP_KERNEL);
8181 	if (!page) {
8182 		err = -ENOMEM;
8183 		goto out;
8184 	}
8185 
8186 	flavors = page_address(page);
8187 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8188 
8189 	/*
8190 	 * Fall back on "guess and check" method if
8191 	 * the server doesn't support SECINFO_NO_NAME
8192 	 */
8193 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8194 		err = nfs4_find_root_sec(server, fhandle, info);
8195 		goto out_freepage;
8196 	}
8197 	if (err)
8198 		goto out_freepage;
8199 
8200 	for (i = 0; i < flavors->num_flavors; i++) {
8201 		secinfo = &flavors->flavors[i];
8202 
8203 		switch (secinfo->flavor) {
8204 		case RPC_AUTH_NULL:
8205 		case RPC_AUTH_UNIX:
8206 		case RPC_AUTH_GSS:
8207 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8208 					&secinfo->flavor_info);
8209 			break;
8210 		default:
8211 			flavor = RPC_AUTH_MAXFLAVOR;
8212 			break;
8213 		}
8214 
8215 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8216 			flavor = RPC_AUTH_MAXFLAVOR;
8217 
8218 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8219 			err = nfs4_lookup_root_sec(server, fhandle,
8220 						   info, flavor);
8221 			if (!err)
8222 				break;
8223 		}
8224 	}
8225 
8226 	if (flavor == RPC_AUTH_MAXFLAVOR)
8227 		err = -EPERM;
8228 
8229 out_freepage:
8230 	put_page(page);
8231 	if (err == -EACCES)
8232 		return -EPERM;
8233 out:
8234 	return err;
8235 }
8236 
8237 static int _nfs41_test_stateid(struct nfs_server *server,
8238 		nfs4_stateid *stateid,
8239 		struct rpc_cred *cred)
8240 {
8241 	int status;
8242 	struct nfs41_test_stateid_args args = {
8243 		.stateid = stateid,
8244 	};
8245 	struct nfs41_test_stateid_res res;
8246 	struct rpc_message msg = {
8247 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8248 		.rpc_argp = &args,
8249 		.rpc_resp = &res,
8250 		.rpc_cred = cred,
8251 	};
8252 	struct rpc_clnt *rpc_client = server->client;
8253 
8254 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8255 		&rpc_client, &msg);
8256 
8257 	dprintk("NFS call  test_stateid %p\n", stateid);
8258 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8259 	nfs4_set_sequence_privileged(&args.seq_args);
8260 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8261 			&args.seq_args, &res.seq_res);
8262 	if (status != NFS_OK) {
8263 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8264 		return status;
8265 	}
8266 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8267 	return -res.status;
8268 }
8269 
8270 /**
8271  * nfs41_test_stateid - perform a TEST_STATEID operation
8272  *
8273  * @server: server / transport on which to perform the operation
8274  * @stateid: state ID to test
8275  * @cred: credential
8276  *
8277  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8278  * Otherwise a negative NFS4ERR value is returned if the operation
8279  * failed or the state ID is not currently valid.
8280  */
8281 static int nfs41_test_stateid(struct nfs_server *server,
8282 		nfs4_stateid *stateid,
8283 		struct rpc_cred *cred)
8284 {
8285 	struct nfs4_exception exception = { };
8286 	int err;
8287 	do {
8288 		err = _nfs41_test_stateid(server, stateid, cred);
8289 		if (err != -NFS4ERR_DELAY)
8290 			break;
8291 		nfs4_handle_exception(server, err, &exception);
8292 	} while (exception.retry);
8293 	return err;
8294 }
8295 
8296 struct nfs_free_stateid_data {
8297 	struct nfs_server *server;
8298 	struct nfs41_free_stateid_args args;
8299 	struct nfs41_free_stateid_res res;
8300 };
8301 
8302 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8303 {
8304 	struct nfs_free_stateid_data *data = calldata;
8305 	nfs41_setup_sequence(nfs4_get_session(data->server),
8306 			&data->args.seq_args,
8307 			&data->res.seq_res,
8308 			task);
8309 }
8310 
8311 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8312 {
8313 	struct nfs_free_stateid_data *data = calldata;
8314 
8315 	nfs41_sequence_done(task, &data->res.seq_res);
8316 
8317 	switch (task->tk_status) {
8318 	case -NFS4ERR_DELAY:
8319 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8320 			rpc_restart_call_prepare(task);
8321 	}
8322 }
8323 
8324 static void nfs41_free_stateid_release(void *calldata)
8325 {
8326 	kfree(calldata);
8327 }
8328 
8329 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8330 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8331 	.rpc_call_done = nfs41_free_stateid_done,
8332 	.rpc_release = nfs41_free_stateid_release,
8333 };
8334 
8335 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8336 		nfs4_stateid *stateid,
8337 		struct rpc_cred *cred,
8338 		bool privileged)
8339 {
8340 	struct rpc_message msg = {
8341 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8342 		.rpc_cred = cred,
8343 	};
8344 	struct rpc_task_setup task_setup = {
8345 		.rpc_client = server->client,
8346 		.rpc_message = &msg,
8347 		.callback_ops = &nfs41_free_stateid_ops,
8348 		.flags = RPC_TASK_ASYNC,
8349 	};
8350 	struct nfs_free_stateid_data *data;
8351 
8352 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8353 		&task_setup.rpc_client, &msg);
8354 
8355 	dprintk("NFS call  free_stateid %p\n", stateid);
8356 	data = kmalloc(sizeof(*data), GFP_NOFS);
8357 	if (!data)
8358 		return ERR_PTR(-ENOMEM);
8359 	data->server = server;
8360 	nfs4_stateid_copy(&data->args.stateid, stateid);
8361 
8362 	task_setup.callback_data = data;
8363 
8364 	msg.rpc_argp = &data->args;
8365 	msg.rpc_resp = &data->res;
8366 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8367 	if (privileged)
8368 		nfs4_set_sequence_privileged(&data->args.seq_args);
8369 
8370 	return rpc_run_task(&task_setup);
8371 }
8372 
8373 /**
8374  * nfs41_free_stateid - perform a FREE_STATEID operation
8375  *
8376  * @server: server / transport on which to perform the operation
8377  * @stateid: state ID to release
8378  * @cred: credential
8379  *
8380  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8381  * negative NFS4ERR value is returned.
8382  */
8383 static int nfs41_free_stateid(struct nfs_server *server,
8384 		nfs4_stateid *stateid,
8385 		struct rpc_cred *cred)
8386 {
8387 	struct rpc_task *task;
8388 	int ret;
8389 
8390 	task = _nfs41_free_stateid(server, stateid, cred, true);
8391 	if (IS_ERR(task))
8392 		return PTR_ERR(task);
8393 	ret = rpc_wait_for_completion_task(task);
8394 	if (!ret)
8395 		ret = task->tk_status;
8396 	rpc_put_task(task);
8397 	return ret;
8398 }
8399 
8400 static void
8401 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8402 {
8403 	struct rpc_task *task;
8404 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8405 
8406 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8407 	nfs4_free_lock_state(server, lsp);
8408 	if (IS_ERR(task))
8409 		return;
8410 	rpc_put_task(task);
8411 }
8412 
8413 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8414 		const nfs4_stateid *s2)
8415 {
8416 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8417 		return false;
8418 
8419 	if (s1->seqid == s2->seqid)
8420 		return true;
8421 	if (s1->seqid == 0 || s2->seqid == 0)
8422 		return true;
8423 
8424 	return false;
8425 }
8426 
8427 #endif /* CONFIG_NFS_V4_1 */
8428 
8429 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8430 		const nfs4_stateid *s2)
8431 {
8432 	return nfs4_stateid_match(s1, s2);
8433 }
8434 
8435 
8436 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8437 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8438 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8439 	.recover_open	= nfs4_open_reclaim,
8440 	.recover_lock	= nfs4_lock_reclaim,
8441 	.establish_clid = nfs4_init_clientid,
8442 	.detect_trunking = nfs40_discover_server_trunking,
8443 };
8444 
8445 #if defined(CONFIG_NFS_V4_1)
8446 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8447 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8448 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8449 	.recover_open	= nfs4_open_reclaim,
8450 	.recover_lock	= nfs4_lock_reclaim,
8451 	.establish_clid = nfs41_init_clientid,
8452 	.reclaim_complete = nfs41_proc_reclaim_complete,
8453 	.detect_trunking = nfs41_discover_server_trunking,
8454 };
8455 #endif /* CONFIG_NFS_V4_1 */
8456 
8457 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8458 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8459 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8460 	.recover_open	= nfs40_open_expired,
8461 	.recover_lock	= nfs4_lock_expired,
8462 	.establish_clid = nfs4_init_clientid,
8463 };
8464 
8465 #if defined(CONFIG_NFS_V4_1)
8466 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8467 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8468 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8469 	.recover_open	= nfs41_open_expired,
8470 	.recover_lock	= nfs41_lock_expired,
8471 	.establish_clid = nfs41_init_clientid,
8472 };
8473 #endif /* CONFIG_NFS_V4_1 */
8474 
8475 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8476 	.sched_state_renewal = nfs4_proc_async_renew,
8477 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8478 	.renew_lease = nfs4_proc_renew,
8479 };
8480 
8481 #if defined(CONFIG_NFS_V4_1)
8482 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8483 	.sched_state_renewal = nfs41_proc_async_sequence,
8484 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8485 	.renew_lease = nfs4_proc_sequence,
8486 };
8487 #endif
8488 
8489 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8490 	.get_locations = _nfs40_proc_get_locations,
8491 	.fsid_present = _nfs40_proc_fsid_present,
8492 };
8493 
8494 #if defined(CONFIG_NFS_V4_1)
8495 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8496 	.get_locations = _nfs41_proc_get_locations,
8497 	.fsid_present = _nfs41_proc_fsid_present,
8498 };
8499 #endif	/* CONFIG_NFS_V4_1 */
8500 
8501 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8502 	.minor_version = 0,
8503 	.init_caps = NFS_CAP_READDIRPLUS
8504 		| NFS_CAP_ATOMIC_OPEN
8505 		| NFS_CAP_CHANGE_ATTR
8506 		| NFS_CAP_POSIX_LOCK,
8507 	.init_client = nfs40_init_client,
8508 	.shutdown_client = nfs40_shutdown_client,
8509 	.match_stateid = nfs4_match_stateid,
8510 	.find_root_sec = nfs4_find_root_sec,
8511 	.free_lock_state = nfs4_release_lockowner,
8512 	.alloc_seqid = nfs_alloc_seqid,
8513 	.call_sync_ops = &nfs40_call_sync_ops,
8514 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8515 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8516 	.state_renewal_ops = &nfs40_state_renewal_ops,
8517 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8518 };
8519 
8520 #if defined(CONFIG_NFS_V4_1)
8521 static struct nfs_seqid *
8522 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8523 {
8524 	return NULL;
8525 }
8526 
8527 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8528 	.minor_version = 1,
8529 	.init_caps = NFS_CAP_READDIRPLUS
8530 		| NFS_CAP_ATOMIC_OPEN
8531 		| NFS_CAP_CHANGE_ATTR
8532 		| NFS_CAP_POSIX_LOCK
8533 		| NFS_CAP_STATEID_NFSV41
8534 		| NFS_CAP_ATOMIC_OPEN_V1,
8535 	.init_client = nfs41_init_client,
8536 	.shutdown_client = nfs41_shutdown_client,
8537 	.match_stateid = nfs41_match_stateid,
8538 	.find_root_sec = nfs41_find_root_sec,
8539 	.free_lock_state = nfs41_free_lock_state,
8540 	.alloc_seqid = nfs_alloc_no_seqid,
8541 	.call_sync_ops = &nfs41_call_sync_ops,
8542 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8543 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8544 	.state_renewal_ops = &nfs41_state_renewal_ops,
8545 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8546 };
8547 #endif
8548 
8549 #if defined(CONFIG_NFS_V4_2)
8550 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8551 	.minor_version = 2,
8552 	.init_caps = NFS_CAP_READDIRPLUS
8553 		| NFS_CAP_ATOMIC_OPEN
8554 		| NFS_CAP_CHANGE_ATTR
8555 		| NFS_CAP_POSIX_LOCK
8556 		| NFS_CAP_STATEID_NFSV41
8557 		| NFS_CAP_ATOMIC_OPEN_V1
8558 		| NFS_CAP_ALLOCATE
8559 		| NFS_CAP_DEALLOCATE
8560 		| NFS_CAP_SEEK,
8561 	.init_client = nfs41_init_client,
8562 	.shutdown_client = nfs41_shutdown_client,
8563 	.match_stateid = nfs41_match_stateid,
8564 	.find_root_sec = nfs41_find_root_sec,
8565 	.free_lock_state = nfs41_free_lock_state,
8566 	.call_sync_ops = &nfs41_call_sync_ops,
8567 	.alloc_seqid = nfs_alloc_no_seqid,
8568 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8569 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8570 	.state_renewal_ops = &nfs41_state_renewal_ops,
8571 };
8572 #endif
8573 
8574 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8575 	[0] = &nfs_v4_0_minor_ops,
8576 #if defined(CONFIG_NFS_V4_1)
8577 	[1] = &nfs_v4_1_minor_ops,
8578 #endif
8579 #if defined(CONFIG_NFS_V4_2)
8580 	[2] = &nfs_v4_2_minor_ops,
8581 #endif
8582 };
8583 
8584 static const struct inode_operations nfs4_dir_inode_operations = {
8585 	.create		= nfs_create,
8586 	.lookup		= nfs_lookup,
8587 	.atomic_open	= nfs_atomic_open,
8588 	.link		= nfs_link,
8589 	.unlink		= nfs_unlink,
8590 	.symlink	= nfs_symlink,
8591 	.mkdir		= nfs_mkdir,
8592 	.rmdir		= nfs_rmdir,
8593 	.mknod		= nfs_mknod,
8594 	.rename		= nfs_rename,
8595 	.permission	= nfs_permission,
8596 	.getattr	= nfs_getattr,
8597 	.setattr	= nfs_setattr,
8598 	.getxattr	= generic_getxattr,
8599 	.setxattr	= generic_setxattr,
8600 	.listxattr	= generic_listxattr,
8601 	.removexattr	= generic_removexattr,
8602 };
8603 
8604 static const struct inode_operations nfs4_file_inode_operations = {
8605 	.permission	= nfs_permission,
8606 	.getattr	= nfs_getattr,
8607 	.setattr	= nfs_setattr,
8608 	.getxattr	= generic_getxattr,
8609 	.setxattr	= generic_setxattr,
8610 	.listxattr	= generic_listxattr,
8611 	.removexattr	= generic_removexattr,
8612 };
8613 
8614 const struct nfs_rpc_ops nfs_v4_clientops = {
8615 	.version	= 4,			/* protocol version */
8616 	.dentry_ops	= &nfs4_dentry_operations,
8617 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8618 	.file_inode_ops	= &nfs4_file_inode_operations,
8619 	.file_ops	= &nfs4_file_operations,
8620 	.getroot	= nfs4_proc_get_root,
8621 	.submount	= nfs4_submount,
8622 	.try_mount	= nfs4_try_mount,
8623 	.getattr	= nfs4_proc_getattr,
8624 	.setattr	= nfs4_proc_setattr,
8625 	.lookup		= nfs4_proc_lookup,
8626 	.access		= nfs4_proc_access,
8627 	.readlink	= nfs4_proc_readlink,
8628 	.create		= nfs4_proc_create,
8629 	.remove		= nfs4_proc_remove,
8630 	.unlink_setup	= nfs4_proc_unlink_setup,
8631 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8632 	.unlink_done	= nfs4_proc_unlink_done,
8633 	.rename_setup	= nfs4_proc_rename_setup,
8634 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8635 	.rename_done	= nfs4_proc_rename_done,
8636 	.link		= nfs4_proc_link,
8637 	.symlink	= nfs4_proc_symlink,
8638 	.mkdir		= nfs4_proc_mkdir,
8639 	.rmdir		= nfs4_proc_remove,
8640 	.readdir	= nfs4_proc_readdir,
8641 	.mknod		= nfs4_proc_mknod,
8642 	.statfs		= nfs4_proc_statfs,
8643 	.fsinfo		= nfs4_proc_fsinfo,
8644 	.pathconf	= nfs4_proc_pathconf,
8645 	.set_capabilities = nfs4_server_capabilities,
8646 	.decode_dirent	= nfs4_decode_dirent,
8647 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8648 	.read_setup	= nfs4_proc_read_setup,
8649 	.read_done	= nfs4_read_done,
8650 	.write_setup	= nfs4_proc_write_setup,
8651 	.write_done	= nfs4_write_done,
8652 	.commit_setup	= nfs4_proc_commit_setup,
8653 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8654 	.commit_done	= nfs4_commit_done,
8655 	.lock		= nfs4_proc_lock,
8656 	.clear_acl_cache = nfs4_zap_acl_attr,
8657 	.close_context  = nfs4_close_context,
8658 	.open_context	= nfs4_atomic_open,
8659 	.have_delegation = nfs4_have_delegation,
8660 	.return_delegation = nfs4_inode_return_delegation,
8661 	.alloc_client	= nfs4_alloc_client,
8662 	.init_client	= nfs4_init_client,
8663 	.free_client	= nfs4_free_client,
8664 	.create_server	= nfs4_create_server,
8665 	.clone_server	= nfs_clone_server,
8666 };
8667 
8668 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8669 	.prefix	= XATTR_NAME_NFSV4_ACL,
8670 	.list	= nfs4_xattr_list_nfs4_acl,
8671 	.get	= nfs4_xattr_get_nfs4_acl,
8672 	.set	= nfs4_xattr_set_nfs4_acl,
8673 };
8674 
8675 const struct xattr_handler *nfs4_xattr_handlers[] = {
8676 	&nfs4_xattr_nfs4_acl_handler,
8677 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8678 	&nfs4_xattr_nfs4_label_handler,
8679 #endif
8680 	NULL
8681 };
8682 
8683 /*
8684  * Local variables:
8685  *  c-basic-offset: 8
8686  * End:
8687  */
8688