xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 40c64c26a43494ba9982fd1b87dc54e3819566fc)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57 
58 #include "nfs4_fs.h"
59 #include "delegation.h"
60 #include "internal.h"
61 #include "iostat.h"
62 #include "callback.h"
63 #include "pnfs.h"
64 #include "netns.h"
65 #include "nfs4idmap.h"
66 #include "nfs4session.h"
67 #include "fscache.h"
68 
69 #include "nfs4trace.h"
70 
71 #define NFSDBG_FACILITY		NFSDBG_PROC
72 
73 #define NFS4_POLL_RETRY_MIN	(HZ/10)
74 #define NFS4_POLL_RETRY_MAX	(15*HZ)
75 
76 struct nfs4_opendata;
77 static int _nfs4_proc_open(struct nfs4_opendata *data);
78 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
79 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
80 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
85 			    struct nfs_fattr *fattr, struct iattr *sattr,
86 			    struct nfs4_state *state, struct nfs4_label *ilabel,
87 			    struct nfs4_label *olabel);
88 #ifdef CONFIG_NFS_V4_1
89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
90 		struct rpc_cred *);
91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
92 		struct rpc_cred *);
93 #endif
94 
95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
96 static inline struct nfs4_label *
97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
98 	struct iattr *sattr, struct nfs4_label *label)
99 {
100 	int err;
101 
102 	if (label == NULL)
103 		return NULL;
104 
105 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
106 		return NULL;
107 
108 	err = security_dentry_init_security(dentry, sattr->ia_mode,
109 				&dentry->d_name, (void **)&label->label, &label->len);
110 	if (err == 0)
111 		return label;
112 
113 	return NULL;
114 }
115 static inline void
116 nfs4_label_release_security(struct nfs4_label *label)
117 {
118 	if (label)
119 		security_release_secctx(label->label, label->len);
120 }
121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
122 {
123 	if (label)
124 		return server->attr_bitmask;
125 
126 	return server->attr_bitmask_nl;
127 }
128 #else
129 static inline struct nfs4_label *
130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
131 	struct iattr *sattr, struct nfs4_label *l)
132 { return NULL; }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 { return; }
136 static inline u32 *
137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 { return server->attr_bitmask; }
139 #endif
140 
141 /* Prevent leaks of NFSv4 errors into userland */
142 static int nfs4_map_errors(int err)
143 {
144 	if (err >= -1000)
145 		return err;
146 	switch (err) {
147 	case -NFS4ERR_RESOURCE:
148 	case -NFS4ERR_LAYOUTTRYLATER:
149 	case -NFS4ERR_RECALLCONFLICT:
150 		return -EREMOTEIO;
151 	case -NFS4ERR_WRONGSEC:
152 	case -NFS4ERR_WRONG_CRED:
153 		return -EPERM;
154 	case -NFS4ERR_BADOWNER:
155 	case -NFS4ERR_BADNAME:
156 		return -EINVAL;
157 	case -NFS4ERR_SHARE_DENIED:
158 		return -EACCES;
159 	case -NFS4ERR_MINOR_VERS_MISMATCH:
160 		return -EPROTONOSUPPORT;
161 	case -NFS4ERR_FILE_OPEN:
162 		return -EBUSY;
163 	default:
164 		dprintk("%s could not handle NFSv4 error %d\n",
165 				__func__, -err);
166 		break;
167 	}
168 	return -EIO;
169 }
170 
171 /*
172  * This is our standard bitmap for GETATTR requests.
173  */
174 const u32 nfs4_fattr_bitmap[3] = {
175 	FATTR4_WORD0_TYPE
176 	| FATTR4_WORD0_CHANGE
177 	| FATTR4_WORD0_SIZE
178 	| FATTR4_WORD0_FSID
179 	| FATTR4_WORD0_FILEID,
180 	FATTR4_WORD1_MODE
181 	| FATTR4_WORD1_NUMLINKS
182 	| FATTR4_WORD1_OWNER
183 	| FATTR4_WORD1_OWNER_GROUP
184 	| FATTR4_WORD1_RAWDEV
185 	| FATTR4_WORD1_SPACE_USED
186 	| FATTR4_WORD1_TIME_ACCESS
187 	| FATTR4_WORD1_TIME_METADATA
188 	| FATTR4_WORD1_TIME_MODIFY
189 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
190 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
191 	FATTR4_WORD2_SECURITY_LABEL
192 #endif
193 };
194 
195 static const u32 nfs4_pnfs_open_bitmap[3] = {
196 	FATTR4_WORD0_TYPE
197 	| FATTR4_WORD0_CHANGE
198 	| FATTR4_WORD0_SIZE
199 	| FATTR4_WORD0_FSID
200 	| FATTR4_WORD0_FILEID,
201 	FATTR4_WORD1_MODE
202 	| FATTR4_WORD1_NUMLINKS
203 	| FATTR4_WORD1_OWNER
204 	| FATTR4_WORD1_OWNER_GROUP
205 	| FATTR4_WORD1_RAWDEV
206 	| FATTR4_WORD1_SPACE_USED
207 	| FATTR4_WORD1_TIME_ACCESS
208 	| FATTR4_WORD1_TIME_METADATA
209 	| FATTR4_WORD1_TIME_MODIFY,
210 	FATTR4_WORD2_MDSTHRESHOLD
211 };
212 
213 static const u32 nfs4_open_noattr_bitmap[3] = {
214 	FATTR4_WORD0_TYPE
215 	| FATTR4_WORD0_CHANGE
216 	| FATTR4_WORD0_FILEID,
217 };
218 
219 const u32 nfs4_statfs_bitmap[3] = {
220 	FATTR4_WORD0_FILES_AVAIL
221 	| FATTR4_WORD0_FILES_FREE
222 	| FATTR4_WORD0_FILES_TOTAL,
223 	FATTR4_WORD1_SPACE_AVAIL
224 	| FATTR4_WORD1_SPACE_FREE
225 	| FATTR4_WORD1_SPACE_TOTAL
226 };
227 
228 const u32 nfs4_pathconf_bitmap[3] = {
229 	FATTR4_WORD0_MAXLINK
230 	| FATTR4_WORD0_MAXNAME,
231 	0
232 };
233 
234 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
235 			| FATTR4_WORD0_MAXREAD
236 			| FATTR4_WORD0_MAXWRITE
237 			| FATTR4_WORD0_LEASE_TIME,
238 			FATTR4_WORD1_TIME_DELTA
239 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
240 			FATTR4_WORD2_LAYOUT_BLKSIZE
241 };
242 
243 const u32 nfs4_fs_locations_bitmap[3] = {
244 	FATTR4_WORD0_TYPE
245 	| FATTR4_WORD0_CHANGE
246 	| FATTR4_WORD0_SIZE
247 	| FATTR4_WORD0_FSID
248 	| FATTR4_WORD0_FILEID
249 	| FATTR4_WORD0_FS_LOCATIONS,
250 	FATTR4_WORD1_MODE
251 	| FATTR4_WORD1_NUMLINKS
252 	| FATTR4_WORD1_OWNER
253 	| FATTR4_WORD1_OWNER_GROUP
254 	| FATTR4_WORD1_RAWDEV
255 	| FATTR4_WORD1_SPACE_USED
256 	| FATTR4_WORD1_TIME_ACCESS
257 	| FATTR4_WORD1_TIME_METADATA
258 	| FATTR4_WORD1_TIME_MODIFY
259 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
260 };
261 
262 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
263 		struct nfs4_readdir_arg *readdir)
264 {
265 	__be32 *start, *p;
266 
267 	if (cookie > 2) {
268 		readdir->cookie = cookie;
269 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
270 		return;
271 	}
272 
273 	readdir->cookie = 0;
274 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
275 	if (cookie == 2)
276 		return;
277 
278 	/*
279 	 * NFSv4 servers do not return entries for '.' and '..'
280 	 * Therefore, we fake these entries here.  We let '.'
281 	 * have cookie 0 and '..' have cookie 1.  Note that
282 	 * when talking to the server, we always send cookie 0
283 	 * instead of 1 or 2.
284 	 */
285 	start = p = kmap_atomic(*readdir->pages);
286 
287 	if (cookie == 0) {
288 		*p++ = xdr_one;                                  /* next */
289 		*p++ = xdr_zero;                   /* cookie, first word */
290 		*p++ = xdr_one;                   /* cookie, second word */
291 		*p++ = xdr_one;                             /* entry len */
292 		memcpy(p, ".\0\0\0", 4);                        /* entry */
293 		p++;
294 		*p++ = xdr_one;                         /* bitmap length */
295 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
296 		*p++ = htonl(8);              /* attribute buffer length */
297 		p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
298 	}
299 
300 	*p++ = xdr_one;                                  /* next */
301 	*p++ = xdr_zero;                   /* cookie, first word */
302 	*p++ = xdr_two;                   /* cookie, second word */
303 	*p++ = xdr_two;                             /* entry len */
304 	memcpy(p, "..\0\0", 4);                         /* entry */
305 	p++;
306 	*p++ = xdr_one;                         /* bitmap length */
307 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
308 	*p++ = htonl(8);              /* attribute buffer length */
309 	p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
310 
311 	readdir->pgbase = (char *)p - (char *)start;
312 	readdir->count -= readdir->pgbase;
313 	kunmap_atomic(start);
314 }
315 
316 static long nfs4_update_delay(long *timeout)
317 {
318 	long ret;
319 	if (!timeout)
320 		return NFS4_POLL_RETRY_MAX;
321 	if (*timeout <= 0)
322 		*timeout = NFS4_POLL_RETRY_MIN;
323 	if (*timeout > NFS4_POLL_RETRY_MAX)
324 		*timeout = NFS4_POLL_RETRY_MAX;
325 	ret = *timeout;
326 	*timeout <<= 1;
327 	return ret;
328 }
329 
330 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
331 {
332 	int res = 0;
333 
334 	might_sleep();
335 
336 	freezable_schedule_timeout_killable_unsafe(
337 		nfs4_update_delay(timeout));
338 	if (fatal_signal_pending(current))
339 		res = -ERESTARTSYS;
340 	return res;
341 }
342 
343 /* This is the error handling routine for processes that are allowed
344  * to sleep.
345  */
346 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
347 {
348 	struct nfs_client *clp = server->nfs_client;
349 	struct nfs4_state *state = exception->state;
350 	struct inode *inode = exception->inode;
351 	int ret = errorcode;
352 
353 	exception->retry = 0;
354 	switch(errorcode) {
355 		case 0:
356 			return 0;
357 		case -NFS4ERR_OPENMODE:
358 			if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
359 				nfs4_inode_return_delegation(inode);
360 				exception->retry = 1;
361 				return 0;
362 			}
363 			if (state == NULL)
364 				break;
365 			ret = nfs4_schedule_stateid_recovery(server, state);
366 			if (ret < 0)
367 				break;
368 			goto wait_on_recovery;
369 		case -NFS4ERR_DELEG_REVOKED:
370 		case -NFS4ERR_ADMIN_REVOKED:
371 		case -NFS4ERR_BAD_STATEID:
372 			if (state == NULL)
373 				break;
374 			ret = nfs4_schedule_stateid_recovery(server, state);
375 			if (ret < 0)
376 				break;
377 			goto wait_on_recovery;
378 		case -NFS4ERR_EXPIRED:
379 			if (state != NULL) {
380 				ret = nfs4_schedule_stateid_recovery(server, state);
381 				if (ret < 0)
382 					break;
383 			}
384 		case -NFS4ERR_STALE_STATEID:
385 		case -NFS4ERR_STALE_CLIENTID:
386 			nfs4_schedule_lease_recovery(clp);
387 			goto wait_on_recovery;
388 		case -NFS4ERR_MOVED:
389 			ret = nfs4_schedule_migration_recovery(server);
390 			if (ret < 0)
391 				break;
392 			goto wait_on_recovery;
393 		case -NFS4ERR_LEASE_MOVED:
394 			nfs4_schedule_lease_moved_recovery(clp);
395 			goto wait_on_recovery;
396 #if defined(CONFIG_NFS_V4_1)
397 		case -NFS4ERR_BADSESSION:
398 		case -NFS4ERR_BADSLOT:
399 		case -NFS4ERR_BAD_HIGH_SLOT:
400 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
401 		case -NFS4ERR_DEADSESSION:
402 		case -NFS4ERR_SEQ_FALSE_RETRY:
403 		case -NFS4ERR_SEQ_MISORDERED:
404 			dprintk("%s ERROR: %d Reset session\n", __func__,
405 				errorcode);
406 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
407 			goto wait_on_recovery;
408 #endif /* defined(CONFIG_NFS_V4_1) */
409 		case -NFS4ERR_FILE_OPEN:
410 			if (exception->timeout > HZ) {
411 				/* We have retried a decent amount, time to
412 				 * fail
413 				 */
414 				ret = -EBUSY;
415 				break;
416 			}
417 		case -NFS4ERR_GRACE:
418 		case -NFS4ERR_DELAY:
419 			ret = nfs4_delay(server->client, &exception->timeout);
420 			if (ret != 0)
421 				break;
422 		case -NFS4ERR_RETRY_UNCACHED_REP:
423 		case -NFS4ERR_OLD_STATEID:
424 			exception->retry = 1;
425 			break;
426 		case -NFS4ERR_BADOWNER:
427 			/* The following works around a Linux server bug! */
428 		case -NFS4ERR_BADNAME:
429 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
430 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
431 				exception->retry = 1;
432 				printk(KERN_WARNING "NFS: v4 server %s "
433 						"does not accept raw "
434 						"uid/gids. "
435 						"Reenabling the idmapper.\n",
436 						server->nfs_client->cl_hostname);
437 			}
438 	}
439 	/* We failed to handle the error */
440 	return nfs4_map_errors(ret);
441 wait_on_recovery:
442 	ret = nfs4_wait_clnt_recover(clp);
443 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
444 		return -EIO;
445 	if (ret == 0)
446 		exception->retry = 1;
447 	return ret;
448 }
449 
450 /*
451  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
452  * or 'false' otherwise.
453  */
454 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
455 {
456 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
457 
458 	if (flavor == RPC_AUTH_GSS_KRB5I ||
459 	    flavor == RPC_AUTH_GSS_KRB5P)
460 		return true;
461 
462 	return false;
463 }
464 
465 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
466 {
467 	spin_lock(&clp->cl_lock);
468 	if (time_before(clp->cl_last_renewal,timestamp))
469 		clp->cl_last_renewal = timestamp;
470 	spin_unlock(&clp->cl_lock);
471 }
472 
473 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
474 {
475 	do_renew_lease(server->nfs_client, timestamp);
476 }
477 
478 struct nfs4_call_sync_data {
479 	const struct nfs_server *seq_server;
480 	struct nfs4_sequence_args *seq_args;
481 	struct nfs4_sequence_res *seq_res;
482 };
483 
484 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
485 			       struct nfs4_sequence_res *res, int cache_reply)
486 {
487 	args->sa_slot = NULL;
488 	args->sa_cache_this = cache_reply;
489 	args->sa_privileged = 0;
490 
491 	res->sr_slot = NULL;
492 }
493 
494 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
495 {
496 	args->sa_privileged = 1;
497 }
498 
499 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
500 			 struct nfs4_sequence_args *args,
501 			 struct nfs4_sequence_res *res,
502 			 struct rpc_task *task)
503 {
504 	struct nfs4_slot *slot;
505 
506 	/* slot already allocated? */
507 	if (res->sr_slot != NULL)
508 		goto out_start;
509 
510 	spin_lock(&tbl->slot_tbl_lock);
511 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
512 		goto out_sleep;
513 
514 	slot = nfs4_alloc_slot(tbl);
515 	if (IS_ERR(slot)) {
516 		if (slot == ERR_PTR(-ENOMEM))
517 			task->tk_timeout = HZ >> 2;
518 		goto out_sleep;
519 	}
520 	spin_unlock(&tbl->slot_tbl_lock);
521 
522 	args->sa_slot = slot;
523 	res->sr_slot = slot;
524 
525 out_start:
526 	rpc_call_start(task);
527 	return 0;
528 
529 out_sleep:
530 	if (args->sa_privileged)
531 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
532 				NULL, RPC_PRIORITY_PRIVILEGED);
533 	else
534 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
535 	spin_unlock(&tbl->slot_tbl_lock);
536 	return -EAGAIN;
537 }
538 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
539 
540 static int nfs40_sequence_done(struct rpc_task *task,
541 			       struct nfs4_sequence_res *res)
542 {
543 	struct nfs4_slot *slot = res->sr_slot;
544 	struct nfs4_slot_table *tbl;
545 
546 	if (slot == NULL)
547 		goto out;
548 
549 	tbl = slot->table;
550 	spin_lock(&tbl->slot_tbl_lock);
551 	if (!nfs41_wake_and_assign_slot(tbl, slot))
552 		nfs4_free_slot(tbl, slot);
553 	spin_unlock(&tbl->slot_tbl_lock);
554 
555 	res->sr_slot = NULL;
556 out:
557 	return 1;
558 }
559 
560 #if defined(CONFIG_NFS_V4_1)
561 
562 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
563 {
564 	struct nfs4_session *session;
565 	struct nfs4_slot_table *tbl;
566 	struct nfs4_slot *slot = res->sr_slot;
567 	bool send_new_highest_used_slotid = false;
568 
569 	tbl = slot->table;
570 	session = tbl->session;
571 
572 	spin_lock(&tbl->slot_tbl_lock);
573 	/* Be nice to the server: try to ensure that the last transmitted
574 	 * value for highest_user_slotid <= target_highest_slotid
575 	 */
576 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
577 		send_new_highest_used_slotid = true;
578 
579 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
580 		send_new_highest_used_slotid = false;
581 		goto out_unlock;
582 	}
583 	nfs4_free_slot(tbl, slot);
584 
585 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
586 		send_new_highest_used_slotid = false;
587 out_unlock:
588 	spin_unlock(&tbl->slot_tbl_lock);
589 	res->sr_slot = NULL;
590 	if (send_new_highest_used_slotid)
591 		nfs41_server_notify_highest_slotid_update(session->clp);
592 }
593 
594 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
595 {
596 	struct nfs4_session *session;
597 	struct nfs4_slot *slot = res->sr_slot;
598 	struct nfs_client *clp;
599 	bool interrupted = false;
600 	int ret = 1;
601 
602 	if (slot == NULL)
603 		goto out_noaction;
604 	/* don't increment the sequence number if the task wasn't sent */
605 	if (!RPC_WAS_SENT(task))
606 		goto out;
607 
608 	session = slot->table->session;
609 
610 	if (slot->interrupted) {
611 		slot->interrupted = 0;
612 		interrupted = true;
613 	}
614 
615 	trace_nfs4_sequence_done(session, res);
616 	/* Check the SEQUENCE operation status */
617 	switch (res->sr_status) {
618 	case 0:
619 		/* Update the slot's sequence and clientid lease timer */
620 		++slot->seq_nr;
621 		clp = session->clp;
622 		do_renew_lease(clp, res->sr_timestamp);
623 		/* Check sequence flags */
624 		if (res->sr_status_flags != 0)
625 			nfs4_schedule_lease_recovery(clp);
626 		nfs41_update_target_slotid(slot->table, slot, res);
627 		break;
628 	case 1:
629 		/*
630 		 * sr_status remains 1 if an RPC level error occurred.
631 		 * The server may or may not have processed the sequence
632 		 * operation..
633 		 * Mark the slot as having hosted an interrupted RPC call.
634 		 */
635 		slot->interrupted = 1;
636 		goto out;
637 	case -NFS4ERR_DELAY:
638 		/* The server detected a resend of the RPC call and
639 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
640 		 * of RFC5661.
641 		 */
642 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
643 			__func__,
644 			slot->slot_nr,
645 			slot->seq_nr);
646 		goto out_retry;
647 	case -NFS4ERR_BADSLOT:
648 		/*
649 		 * The slot id we used was probably retired. Try again
650 		 * using a different slot id.
651 		 */
652 		goto retry_nowait;
653 	case -NFS4ERR_SEQ_MISORDERED:
654 		/*
655 		 * Was the last operation on this sequence interrupted?
656 		 * If so, retry after bumping the sequence number.
657 		 */
658 		if (interrupted) {
659 			++slot->seq_nr;
660 			goto retry_nowait;
661 		}
662 		/*
663 		 * Could this slot have been previously retired?
664 		 * If so, then the server may be expecting seq_nr = 1!
665 		 */
666 		if (slot->seq_nr != 1) {
667 			slot->seq_nr = 1;
668 			goto retry_nowait;
669 		}
670 		break;
671 	case -NFS4ERR_SEQ_FALSE_RETRY:
672 		++slot->seq_nr;
673 		goto retry_nowait;
674 	default:
675 		/* Just update the slot sequence no. */
676 		++slot->seq_nr;
677 	}
678 out:
679 	/* The session may be reset by one of the error handlers. */
680 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
681 	nfs41_sequence_free_slot(res);
682 out_noaction:
683 	return ret;
684 retry_nowait:
685 	if (rpc_restart_call_prepare(task)) {
686 		task->tk_status = 0;
687 		ret = 0;
688 	}
689 	goto out;
690 out_retry:
691 	if (!rpc_restart_call(task))
692 		goto out;
693 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
694 	return 0;
695 }
696 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
697 
698 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
699 {
700 	if (res->sr_slot == NULL)
701 		return 1;
702 	if (!res->sr_slot->table->session)
703 		return nfs40_sequence_done(task, res);
704 	return nfs41_sequence_done(task, res);
705 }
706 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
707 
708 int nfs41_setup_sequence(struct nfs4_session *session,
709 				struct nfs4_sequence_args *args,
710 				struct nfs4_sequence_res *res,
711 				struct rpc_task *task)
712 {
713 	struct nfs4_slot *slot;
714 	struct nfs4_slot_table *tbl;
715 
716 	dprintk("--> %s\n", __func__);
717 	/* slot already allocated? */
718 	if (res->sr_slot != NULL)
719 		goto out_success;
720 
721 	tbl = &session->fc_slot_table;
722 
723 	task->tk_timeout = 0;
724 
725 	spin_lock(&tbl->slot_tbl_lock);
726 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
727 	    !args->sa_privileged) {
728 		/* The state manager will wait until the slot table is empty */
729 		dprintk("%s session is draining\n", __func__);
730 		goto out_sleep;
731 	}
732 
733 	slot = nfs4_alloc_slot(tbl);
734 	if (IS_ERR(slot)) {
735 		/* If out of memory, try again in 1/4 second */
736 		if (slot == ERR_PTR(-ENOMEM))
737 			task->tk_timeout = HZ >> 2;
738 		dprintk("<-- %s: no free slots\n", __func__);
739 		goto out_sleep;
740 	}
741 	spin_unlock(&tbl->slot_tbl_lock);
742 
743 	args->sa_slot = slot;
744 
745 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
746 			slot->slot_nr, slot->seq_nr);
747 
748 	res->sr_slot = slot;
749 	res->sr_timestamp = jiffies;
750 	res->sr_status_flags = 0;
751 	/*
752 	 * sr_status is only set in decode_sequence, and so will remain
753 	 * set to 1 if an rpc level failure occurs.
754 	 */
755 	res->sr_status = 1;
756 	trace_nfs4_setup_sequence(session, args);
757 out_success:
758 	rpc_call_start(task);
759 	return 0;
760 out_sleep:
761 	/* Privileged tasks are queued with top priority */
762 	if (args->sa_privileged)
763 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
764 				NULL, RPC_PRIORITY_PRIVILEGED);
765 	else
766 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
767 	spin_unlock(&tbl->slot_tbl_lock);
768 	return -EAGAIN;
769 }
770 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
771 
772 static int nfs4_setup_sequence(const struct nfs_server *server,
773 			       struct nfs4_sequence_args *args,
774 			       struct nfs4_sequence_res *res,
775 			       struct rpc_task *task)
776 {
777 	struct nfs4_session *session = nfs4_get_session(server);
778 	int ret = 0;
779 
780 	if (!session)
781 		return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
782 					    args, res, task);
783 
784 	dprintk("--> %s clp %p session %p sr_slot %u\n",
785 		__func__, session->clp, session, res->sr_slot ?
786 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
787 
788 	ret = nfs41_setup_sequence(session, args, res, task);
789 
790 	dprintk("<-- %s status=%d\n", __func__, ret);
791 	return ret;
792 }
793 
794 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
795 {
796 	struct nfs4_call_sync_data *data = calldata;
797 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
798 
799 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
800 
801 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
802 }
803 
804 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
805 {
806 	struct nfs4_call_sync_data *data = calldata;
807 
808 	nfs41_sequence_done(task, data->seq_res);
809 }
810 
811 static const struct rpc_call_ops nfs41_call_sync_ops = {
812 	.rpc_call_prepare = nfs41_call_sync_prepare,
813 	.rpc_call_done = nfs41_call_sync_done,
814 };
815 
816 #else	/* !CONFIG_NFS_V4_1 */
817 
818 static int nfs4_setup_sequence(const struct nfs_server *server,
819 			       struct nfs4_sequence_args *args,
820 			       struct nfs4_sequence_res *res,
821 			       struct rpc_task *task)
822 {
823 	return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
824 				    args, res, task);
825 }
826 
827 int nfs4_sequence_done(struct rpc_task *task,
828 		       struct nfs4_sequence_res *res)
829 {
830 	return nfs40_sequence_done(task, res);
831 }
832 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
833 
834 #endif	/* !CONFIG_NFS_V4_1 */
835 
836 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
837 {
838 	struct nfs4_call_sync_data *data = calldata;
839 	nfs4_setup_sequence(data->seq_server,
840 				data->seq_args, data->seq_res, task);
841 }
842 
843 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
844 {
845 	struct nfs4_call_sync_data *data = calldata;
846 	nfs4_sequence_done(task, data->seq_res);
847 }
848 
849 static const struct rpc_call_ops nfs40_call_sync_ops = {
850 	.rpc_call_prepare = nfs40_call_sync_prepare,
851 	.rpc_call_done = nfs40_call_sync_done,
852 };
853 
854 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
855 				   struct nfs_server *server,
856 				   struct rpc_message *msg,
857 				   struct nfs4_sequence_args *args,
858 				   struct nfs4_sequence_res *res)
859 {
860 	int ret;
861 	struct rpc_task *task;
862 	struct nfs_client *clp = server->nfs_client;
863 	struct nfs4_call_sync_data data = {
864 		.seq_server = server,
865 		.seq_args = args,
866 		.seq_res = res,
867 	};
868 	struct rpc_task_setup task_setup = {
869 		.rpc_client = clnt,
870 		.rpc_message = msg,
871 		.callback_ops = clp->cl_mvops->call_sync_ops,
872 		.callback_data = &data
873 	};
874 
875 	task = rpc_run_task(&task_setup);
876 	if (IS_ERR(task))
877 		ret = PTR_ERR(task);
878 	else {
879 		ret = task->tk_status;
880 		rpc_put_task(task);
881 	}
882 	return ret;
883 }
884 
885 int nfs4_call_sync(struct rpc_clnt *clnt,
886 		   struct nfs_server *server,
887 		   struct rpc_message *msg,
888 		   struct nfs4_sequence_args *args,
889 		   struct nfs4_sequence_res *res,
890 		   int cache_reply)
891 {
892 	nfs4_init_sequence(args, res, cache_reply);
893 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
894 }
895 
896 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
897 {
898 	struct nfs_inode *nfsi = NFS_I(dir);
899 
900 	spin_lock(&dir->i_lock);
901 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
902 	if (!cinfo->atomic || cinfo->before != dir->i_version)
903 		nfs_force_lookup_revalidate(dir);
904 	dir->i_version = cinfo->after;
905 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
906 	nfs_fscache_invalidate(dir);
907 	spin_unlock(&dir->i_lock);
908 }
909 
910 struct nfs4_opendata {
911 	struct kref kref;
912 	struct nfs_openargs o_arg;
913 	struct nfs_openres o_res;
914 	struct nfs_open_confirmargs c_arg;
915 	struct nfs_open_confirmres c_res;
916 	struct nfs4_string owner_name;
917 	struct nfs4_string group_name;
918 	struct nfs_fattr f_attr;
919 	struct nfs4_label *f_label;
920 	struct dentry *dir;
921 	struct dentry *dentry;
922 	struct nfs4_state_owner *owner;
923 	struct nfs4_state *state;
924 	struct iattr attrs;
925 	unsigned long timestamp;
926 	unsigned int rpc_done : 1;
927 	unsigned int file_created : 1;
928 	unsigned int is_recover : 1;
929 	int rpc_status;
930 	int cancelled;
931 };
932 
933 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
934 		int err, struct nfs4_exception *exception)
935 {
936 	if (err != -EINVAL)
937 		return false;
938 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
939 		return false;
940 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
941 	exception->retry = 1;
942 	return true;
943 }
944 
945 static u32
946 nfs4_map_atomic_open_share(struct nfs_server *server,
947 		fmode_t fmode, int openflags)
948 {
949 	u32 res = 0;
950 
951 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
952 	case FMODE_READ:
953 		res = NFS4_SHARE_ACCESS_READ;
954 		break;
955 	case FMODE_WRITE:
956 		res = NFS4_SHARE_ACCESS_WRITE;
957 		break;
958 	case FMODE_READ|FMODE_WRITE:
959 		res = NFS4_SHARE_ACCESS_BOTH;
960 	}
961 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
962 		goto out;
963 	/* Want no delegation if we're using O_DIRECT */
964 	if (openflags & O_DIRECT)
965 		res |= NFS4_SHARE_WANT_NO_DELEG;
966 out:
967 	return res;
968 }
969 
970 static enum open_claim_type4
971 nfs4_map_atomic_open_claim(struct nfs_server *server,
972 		enum open_claim_type4 claim)
973 {
974 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
975 		return claim;
976 	switch (claim) {
977 	default:
978 		return claim;
979 	case NFS4_OPEN_CLAIM_FH:
980 		return NFS4_OPEN_CLAIM_NULL;
981 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
982 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
983 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
984 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
985 	}
986 }
987 
988 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
989 {
990 	p->o_res.f_attr = &p->f_attr;
991 	p->o_res.f_label = p->f_label;
992 	p->o_res.seqid = p->o_arg.seqid;
993 	p->c_res.seqid = p->c_arg.seqid;
994 	p->o_res.server = p->o_arg.server;
995 	p->o_res.access_request = p->o_arg.access;
996 	nfs_fattr_init(&p->f_attr);
997 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
998 }
999 
1000 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1001 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1002 		const struct iattr *attrs,
1003 		struct nfs4_label *label,
1004 		enum open_claim_type4 claim,
1005 		gfp_t gfp_mask)
1006 {
1007 	struct dentry *parent = dget_parent(dentry);
1008 	struct inode *dir = parent->d_inode;
1009 	struct nfs_server *server = NFS_SERVER(dir);
1010 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1011 	struct nfs4_opendata *p;
1012 
1013 	p = kzalloc(sizeof(*p), gfp_mask);
1014 	if (p == NULL)
1015 		goto err;
1016 
1017 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1018 	if (IS_ERR(p->f_label))
1019 		goto err_free_p;
1020 
1021 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1022 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1023 	if (IS_ERR(p->o_arg.seqid))
1024 		goto err_free_label;
1025 	nfs_sb_active(dentry->d_sb);
1026 	p->dentry = dget(dentry);
1027 	p->dir = parent;
1028 	p->owner = sp;
1029 	atomic_inc(&sp->so_count);
1030 	p->o_arg.open_flags = flags;
1031 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1032 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1033 			fmode, flags);
1034 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1035 	 * will return permission denied for all bits until close */
1036 	if (!(flags & O_EXCL)) {
1037 		/* ask server to check for all possible rights as results
1038 		 * are cached */
1039 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1040 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1041 	}
1042 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1043 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1044 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1045 	p->o_arg.name = &dentry->d_name;
1046 	p->o_arg.server = server;
1047 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1048 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1049 	p->o_arg.label = label;
1050 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1051 	switch (p->o_arg.claim) {
1052 	case NFS4_OPEN_CLAIM_NULL:
1053 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1054 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1055 		p->o_arg.fh = NFS_FH(dir);
1056 		break;
1057 	case NFS4_OPEN_CLAIM_PREVIOUS:
1058 	case NFS4_OPEN_CLAIM_FH:
1059 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1060 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1061 		p->o_arg.fh = NFS_FH(dentry->d_inode);
1062 	}
1063 	if (attrs != NULL && attrs->ia_valid != 0) {
1064 		__u32 verf[2];
1065 
1066 		p->o_arg.u.attrs = &p->attrs;
1067 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1068 
1069 		verf[0] = jiffies;
1070 		verf[1] = current->pid;
1071 		memcpy(p->o_arg.u.verifier.data, verf,
1072 				sizeof(p->o_arg.u.verifier.data));
1073 	}
1074 	p->c_arg.fh = &p->o_res.fh;
1075 	p->c_arg.stateid = &p->o_res.stateid;
1076 	p->c_arg.seqid = p->o_arg.seqid;
1077 	nfs4_init_opendata_res(p);
1078 	kref_init(&p->kref);
1079 	return p;
1080 
1081 err_free_label:
1082 	nfs4_label_free(p->f_label);
1083 err_free_p:
1084 	kfree(p);
1085 err:
1086 	dput(parent);
1087 	return NULL;
1088 }
1089 
1090 static void nfs4_opendata_free(struct kref *kref)
1091 {
1092 	struct nfs4_opendata *p = container_of(kref,
1093 			struct nfs4_opendata, kref);
1094 	struct super_block *sb = p->dentry->d_sb;
1095 
1096 	nfs_free_seqid(p->o_arg.seqid);
1097 	if (p->state != NULL)
1098 		nfs4_put_open_state(p->state);
1099 	nfs4_put_state_owner(p->owner);
1100 
1101 	nfs4_label_free(p->f_label);
1102 
1103 	dput(p->dir);
1104 	dput(p->dentry);
1105 	nfs_sb_deactive(sb);
1106 	nfs_fattr_free_names(&p->f_attr);
1107 	kfree(p->f_attr.mdsthreshold);
1108 	kfree(p);
1109 }
1110 
1111 static void nfs4_opendata_put(struct nfs4_opendata *p)
1112 {
1113 	if (p != NULL)
1114 		kref_put(&p->kref, nfs4_opendata_free);
1115 }
1116 
1117 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1118 {
1119 	int ret;
1120 
1121 	ret = rpc_wait_for_completion_task(task);
1122 	return ret;
1123 }
1124 
1125 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1126 {
1127 	int ret = 0;
1128 
1129 	if (open_mode & (O_EXCL|O_TRUNC))
1130 		goto out;
1131 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1132 		case FMODE_READ:
1133 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1134 				&& state->n_rdonly != 0;
1135 			break;
1136 		case FMODE_WRITE:
1137 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1138 				&& state->n_wronly != 0;
1139 			break;
1140 		case FMODE_READ|FMODE_WRITE:
1141 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1142 				&& state->n_rdwr != 0;
1143 	}
1144 out:
1145 	return ret;
1146 }
1147 
1148 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1149 {
1150 	if (delegation == NULL)
1151 		return 0;
1152 	if ((delegation->type & fmode) != fmode)
1153 		return 0;
1154 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1155 		return 0;
1156 	nfs_mark_delegation_referenced(delegation);
1157 	return 1;
1158 }
1159 
1160 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1161 {
1162 	switch (fmode) {
1163 		case FMODE_WRITE:
1164 			state->n_wronly++;
1165 			break;
1166 		case FMODE_READ:
1167 			state->n_rdonly++;
1168 			break;
1169 		case FMODE_READ|FMODE_WRITE:
1170 			state->n_rdwr++;
1171 	}
1172 	nfs4_state_set_mode_locked(state, state->state | fmode);
1173 }
1174 
1175 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1176 {
1177 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1178 	bool need_recover = false;
1179 
1180 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1181 		need_recover = true;
1182 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1183 		need_recover = true;
1184 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1185 		need_recover = true;
1186 	if (need_recover)
1187 		nfs4_state_mark_reclaim_nograce(clp, state);
1188 }
1189 
1190 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1191 		nfs4_stateid *stateid)
1192 {
1193 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1194 		return true;
1195 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1196 		nfs_test_and_clear_all_open_stateid(state);
1197 		return true;
1198 	}
1199 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1200 		return true;
1201 	return false;
1202 }
1203 
1204 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1205 {
1206 	if (state->n_wronly)
1207 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1208 	if (state->n_rdonly)
1209 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1210 	if (state->n_rdwr)
1211 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1212 }
1213 
1214 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1215 		nfs4_stateid *stateid, fmode_t fmode)
1216 {
1217 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1218 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1219 	case FMODE_WRITE:
1220 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1221 		break;
1222 	case FMODE_READ:
1223 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1224 		break;
1225 	case 0:
1226 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1227 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1228 		clear_bit(NFS_OPEN_STATE, &state->flags);
1229 	}
1230 	if (stateid == NULL)
1231 		return;
1232 	/* Handle races with OPEN */
1233 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1234 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1235 		nfs_resync_open_stateid_locked(state);
1236 		return;
1237 	}
1238 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1239 		nfs4_stateid_copy(&state->stateid, stateid);
1240 	nfs4_stateid_copy(&state->open_stateid, stateid);
1241 }
1242 
1243 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1244 {
1245 	write_seqlock(&state->seqlock);
1246 	nfs_clear_open_stateid_locked(state, stateid, fmode);
1247 	write_sequnlock(&state->seqlock);
1248 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1249 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1250 }
1251 
1252 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1253 {
1254 	switch (fmode) {
1255 		case FMODE_READ:
1256 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1257 			break;
1258 		case FMODE_WRITE:
1259 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1260 			break;
1261 		case FMODE_READ|FMODE_WRITE:
1262 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1263 	}
1264 	if (!nfs_need_update_open_stateid(state, stateid))
1265 		return;
1266 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1267 		nfs4_stateid_copy(&state->stateid, stateid);
1268 	nfs4_stateid_copy(&state->open_stateid, stateid);
1269 }
1270 
1271 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1272 {
1273 	/*
1274 	 * Protect the call to nfs4_state_set_mode_locked and
1275 	 * serialise the stateid update
1276 	 */
1277 	write_seqlock(&state->seqlock);
1278 	if (deleg_stateid != NULL) {
1279 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1280 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1281 	}
1282 	if (open_stateid != NULL)
1283 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1284 	write_sequnlock(&state->seqlock);
1285 	spin_lock(&state->owner->so_lock);
1286 	update_open_stateflags(state, fmode);
1287 	spin_unlock(&state->owner->so_lock);
1288 }
1289 
1290 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1291 {
1292 	struct nfs_inode *nfsi = NFS_I(state->inode);
1293 	struct nfs_delegation *deleg_cur;
1294 	int ret = 0;
1295 
1296 	fmode &= (FMODE_READ|FMODE_WRITE);
1297 
1298 	rcu_read_lock();
1299 	deleg_cur = rcu_dereference(nfsi->delegation);
1300 	if (deleg_cur == NULL)
1301 		goto no_delegation;
1302 
1303 	spin_lock(&deleg_cur->lock);
1304 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1305 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1306 	    (deleg_cur->type & fmode) != fmode)
1307 		goto no_delegation_unlock;
1308 
1309 	if (delegation == NULL)
1310 		delegation = &deleg_cur->stateid;
1311 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1312 		goto no_delegation_unlock;
1313 
1314 	nfs_mark_delegation_referenced(deleg_cur);
1315 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1316 	ret = 1;
1317 no_delegation_unlock:
1318 	spin_unlock(&deleg_cur->lock);
1319 no_delegation:
1320 	rcu_read_unlock();
1321 
1322 	if (!ret && open_stateid != NULL) {
1323 		__update_open_stateid(state, open_stateid, NULL, fmode);
1324 		ret = 1;
1325 	}
1326 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1327 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1328 
1329 	return ret;
1330 }
1331 
1332 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1333 		const nfs4_stateid *stateid)
1334 {
1335 	struct nfs4_state *state = lsp->ls_state;
1336 	bool ret = false;
1337 
1338 	spin_lock(&state->state_lock);
1339 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1340 		goto out_noupdate;
1341 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1342 		goto out_noupdate;
1343 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1344 	ret = true;
1345 out_noupdate:
1346 	spin_unlock(&state->state_lock);
1347 	return ret;
1348 }
1349 
1350 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1351 {
1352 	struct nfs_delegation *delegation;
1353 
1354 	rcu_read_lock();
1355 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1356 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1357 		rcu_read_unlock();
1358 		return;
1359 	}
1360 	rcu_read_unlock();
1361 	nfs4_inode_return_delegation(inode);
1362 }
1363 
1364 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1365 {
1366 	struct nfs4_state *state = opendata->state;
1367 	struct nfs_inode *nfsi = NFS_I(state->inode);
1368 	struct nfs_delegation *delegation;
1369 	int open_mode = opendata->o_arg.open_flags;
1370 	fmode_t fmode = opendata->o_arg.fmode;
1371 	nfs4_stateid stateid;
1372 	int ret = -EAGAIN;
1373 
1374 	for (;;) {
1375 		spin_lock(&state->owner->so_lock);
1376 		if (can_open_cached(state, fmode, open_mode)) {
1377 			update_open_stateflags(state, fmode);
1378 			spin_unlock(&state->owner->so_lock);
1379 			goto out_return_state;
1380 		}
1381 		spin_unlock(&state->owner->so_lock);
1382 		rcu_read_lock();
1383 		delegation = rcu_dereference(nfsi->delegation);
1384 		if (!can_open_delegated(delegation, fmode)) {
1385 			rcu_read_unlock();
1386 			break;
1387 		}
1388 		/* Save the delegation */
1389 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1390 		rcu_read_unlock();
1391 		nfs_release_seqid(opendata->o_arg.seqid);
1392 		if (!opendata->is_recover) {
1393 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1394 			if (ret != 0)
1395 				goto out;
1396 		}
1397 		ret = -EAGAIN;
1398 
1399 		/* Try to update the stateid using the delegation */
1400 		if (update_open_stateid(state, NULL, &stateid, fmode))
1401 			goto out_return_state;
1402 	}
1403 out:
1404 	return ERR_PTR(ret);
1405 out_return_state:
1406 	atomic_inc(&state->count);
1407 	return state;
1408 }
1409 
1410 static void
1411 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1412 {
1413 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1414 	struct nfs_delegation *delegation;
1415 	int delegation_flags = 0;
1416 
1417 	rcu_read_lock();
1418 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1419 	if (delegation)
1420 		delegation_flags = delegation->flags;
1421 	rcu_read_unlock();
1422 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1423 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1424 				   "returning a delegation for "
1425 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1426 				   clp->cl_hostname);
1427 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1428 		nfs_inode_set_delegation(state->inode,
1429 					 data->owner->so_cred,
1430 					 &data->o_res);
1431 	else
1432 		nfs_inode_reclaim_delegation(state->inode,
1433 					     data->owner->so_cred,
1434 					     &data->o_res);
1435 }
1436 
1437 /*
1438  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1439  * and update the nfs4_state.
1440  */
1441 static struct nfs4_state *
1442 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1443 {
1444 	struct inode *inode = data->state->inode;
1445 	struct nfs4_state *state = data->state;
1446 	int ret;
1447 
1448 	if (!data->rpc_done) {
1449 		if (data->rpc_status) {
1450 			ret = data->rpc_status;
1451 			goto err;
1452 		}
1453 		/* cached opens have already been processed */
1454 		goto update;
1455 	}
1456 
1457 	ret = nfs_refresh_inode(inode, &data->f_attr);
1458 	if (ret)
1459 		goto err;
1460 
1461 	if (data->o_res.delegation_type != 0)
1462 		nfs4_opendata_check_deleg(data, state);
1463 update:
1464 	update_open_stateid(state, &data->o_res.stateid, NULL,
1465 			    data->o_arg.fmode);
1466 	atomic_inc(&state->count);
1467 
1468 	return state;
1469 err:
1470 	return ERR_PTR(ret);
1471 
1472 }
1473 
1474 static struct nfs4_state *
1475 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1476 {
1477 	struct inode *inode;
1478 	struct nfs4_state *state = NULL;
1479 	int ret;
1480 
1481 	if (!data->rpc_done) {
1482 		state = nfs4_try_open_cached(data);
1483 		goto out;
1484 	}
1485 
1486 	ret = -EAGAIN;
1487 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1488 		goto err;
1489 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1490 	ret = PTR_ERR(inode);
1491 	if (IS_ERR(inode))
1492 		goto err;
1493 	ret = -ENOMEM;
1494 	state = nfs4_get_open_state(inode, data->owner);
1495 	if (state == NULL)
1496 		goto err_put_inode;
1497 	if (data->o_res.delegation_type != 0)
1498 		nfs4_opendata_check_deleg(data, state);
1499 	update_open_stateid(state, &data->o_res.stateid, NULL,
1500 			data->o_arg.fmode);
1501 	iput(inode);
1502 out:
1503 	nfs_release_seqid(data->o_arg.seqid);
1504 	return state;
1505 err_put_inode:
1506 	iput(inode);
1507 err:
1508 	return ERR_PTR(ret);
1509 }
1510 
1511 static struct nfs4_state *
1512 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1513 {
1514 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1515 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1516 	return _nfs4_opendata_to_nfs4_state(data);
1517 }
1518 
1519 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1520 {
1521 	struct nfs_inode *nfsi = NFS_I(state->inode);
1522 	struct nfs_open_context *ctx;
1523 
1524 	spin_lock(&state->inode->i_lock);
1525 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1526 		if (ctx->state != state)
1527 			continue;
1528 		get_nfs_open_context(ctx);
1529 		spin_unlock(&state->inode->i_lock);
1530 		return ctx;
1531 	}
1532 	spin_unlock(&state->inode->i_lock);
1533 	return ERR_PTR(-ENOENT);
1534 }
1535 
1536 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1537 		struct nfs4_state *state, enum open_claim_type4 claim)
1538 {
1539 	struct nfs4_opendata *opendata;
1540 
1541 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1542 			NULL, NULL, claim, GFP_NOFS);
1543 	if (opendata == NULL)
1544 		return ERR_PTR(-ENOMEM);
1545 	opendata->state = state;
1546 	atomic_inc(&state->count);
1547 	return opendata;
1548 }
1549 
1550 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1551 {
1552 	struct nfs4_state *newstate;
1553 	int ret;
1554 
1555 	opendata->o_arg.open_flags = 0;
1556 	opendata->o_arg.fmode = fmode;
1557 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1558 			NFS_SB(opendata->dentry->d_sb),
1559 			fmode, 0);
1560 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1561 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1562 	nfs4_init_opendata_res(opendata);
1563 	ret = _nfs4_recover_proc_open(opendata);
1564 	if (ret != 0)
1565 		return ret;
1566 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1567 	if (IS_ERR(newstate))
1568 		return PTR_ERR(newstate);
1569 	nfs4_close_state(newstate, fmode);
1570 	*res = newstate;
1571 	return 0;
1572 }
1573 
1574 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1575 {
1576 	struct nfs4_state *newstate;
1577 	int ret;
1578 
1579 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1580 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1581 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1582 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1583 	/* memory barrier prior to reading state->n_* */
1584 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1585 	clear_bit(NFS_OPEN_STATE, &state->flags);
1586 	smp_rmb();
1587 	if (state->n_rdwr != 0) {
1588 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1589 		if (ret != 0)
1590 			return ret;
1591 		if (newstate != state)
1592 			return -ESTALE;
1593 	}
1594 	if (state->n_wronly != 0) {
1595 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1596 		if (ret != 0)
1597 			return ret;
1598 		if (newstate != state)
1599 			return -ESTALE;
1600 	}
1601 	if (state->n_rdonly != 0) {
1602 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1603 		if (ret != 0)
1604 			return ret;
1605 		if (newstate != state)
1606 			return -ESTALE;
1607 	}
1608 	/*
1609 	 * We may have performed cached opens for all three recoveries.
1610 	 * Check if we need to update the current stateid.
1611 	 */
1612 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1613 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1614 		write_seqlock(&state->seqlock);
1615 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1616 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1617 		write_sequnlock(&state->seqlock);
1618 	}
1619 	return 0;
1620 }
1621 
1622 /*
1623  * OPEN_RECLAIM:
1624  * 	reclaim state on the server after a reboot.
1625  */
1626 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1627 {
1628 	struct nfs_delegation *delegation;
1629 	struct nfs4_opendata *opendata;
1630 	fmode_t delegation_type = 0;
1631 	int status;
1632 
1633 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1634 			NFS4_OPEN_CLAIM_PREVIOUS);
1635 	if (IS_ERR(opendata))
1636 		return PTR_ERR(opendata);
1637 	rcu_read_lock();
1638 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1639 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1640 		delegation_type = delegation->type;
1641 	rcu_read_unlock();
1642 	opendata->o_arg.u.delegation_type = delegation_type;
1643 	status = nfs4_open_recover(opendata, state);
1644 	nfs4_opendata_put(opendata);
1645 	return status;
1646 }
1647 
1648 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1649 {
1650 	struct nfs_server *server = NFS_SERVER(state->inode);
1651 	struct nfs4_exception exception = { };
1652 	int err;
1653 	do {
1654 		err = _nfs4_do_open_reclaim(ctx, state);
1655 		trace_nfs4_open_reclaim(ctx, 0, err);
1656 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1657 			continue;
1658 		if (err != -NFS4ERR_DELAY)
1659 			break;
1660 		nfs4_handle_exception(server, err, &exception);
1661 	} while (exception.retry);
1662 	return err;
1663 }
1664 
1665 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1666 {
1667 	struct nfs_open_context *ctx;
1668 	int ret;
1669 
1670 	ctx = nfs4_state_find_open_context(state);
1671 	if (IS_ERR(ctx))
1672 		return -EAGAIN;
1673 	ret = nfs4_do_open_reclaim(ctx, state);
1674 	put_nfs_open_context(ctx);
1675 	return ret;
1676 }
1677 
1678 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1679 {
1680 	switch (err) {
1681 		default:
1682 			printk(KERN_ERR "NFS: %s: unhandled error "
1683 					"%d.\n", __func__, err);
1684 		case 0:
1685 		case -ENOENT:
1686 		case -ESTALE:
1687 			break;
1688 		case -NFS4ERR_BADSESSION:
1689 		case -NFS4ERR_BADSLOT:
1690 		case -NFS4ERR_BAD_HIGH_SLOT:
1691 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1692 		case -NFS4ERR_DEADSESSION:
1693 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1694 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1695 			return -EAGAIN;
1696 		case -NFS4ERR_STALE_CLIENTID:
1697 		case -NFS4ERR_STALE_STATEID:
1698 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1699 		case -NFS4ERR_EXPIRED:
1700 			/* Don't recall a delegation if it was lost */
1701 			nfs4_schedule_lease_recovery(server->nfs_client);
1702 			return -EAGAIN;
1703 		case -NFS4ERR_MOVED:
1704 			nfs4_schedule_migration_recovery(server);
1705 			return -EAGAIN;
1706 		case -NFS4ERR_LEASE_MOVED:
1707 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1708 			return -EAGAIN;
1709 		case -NFS4ERR_DELEG_REVOKED:
1710 		case -NFS4ERR_ADMIN_REVOKED:
1711 		case -NFS4ERR_BAD_STATEID:
1712 		case -NFS4ERR_OPENMODE:
1713 			nfs_inode_find_state_and_recover(state->inode,
1714 					stateid);
1715 			nfs4_schedule_stateid_recovery(server, state);
1716 			return -EAGAIN;
1717 		case -NFS4ERR_DELAY:
1718 		case -NFS4ERR_GRACE:
1719 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1720 			ssleep(1);
1721 			return -EAGAIN;
1722 		case -ENOMEM:
1723 		case -NFS4ERR_DENIED:
1724 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1725 			return 0;
1726 	}
1727 	return err;
1728 }
1729 
1730 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1731 {
1732 	struct nfs_server *server = NFS_SERVER(state->inode);
1733 	struct nfs4_opendata *opendata;
1734 	int err;
1735 
1736 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1737 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1738 	if (IS_ERR(opendata))
1739 		return PTR_ERR(opendata);
1740 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1741 	err = nfs4_open_recover(opendata, state);
1742 	nfs4_opendata_put(opendata);
1743 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1744 }
1745 
1746 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1747 {
1748 	struct nfs4_opendata *data = calldata;
1749 
1750 	nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1751 			     &data->c_arg.seq_args, &data->c_res.seq_res, task);
1752 }
1753 
1754 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1755 {
1756 	struct nfs4_opendata *data = calldata;
1757 
1758 	nfs40_sequence_done(task, &data->c_res.seq_res);
1759 
1760 	data->rpc_status = task->tk_status;
1761 	if (data->rpc_status == 0) {
1762 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1763 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1764 		renew_lease(data->o_res.server, data->timestamp);
1765 		data->rpc_done = 1;
1766 	}
1767 }
1768 
1769 static void nfs4_open_confirm_release(void *calldata)
1770 {
1771 	struct nfs4_opendata *data = calldata;
1772 	struct nfs4_state *state = NULL;
1773 
1774 	/* If this request hasn't been cancelled, do nothing */
1775 	if (data->cancelled == 0)
1776 		goto out_free;
1777 	/* In case of error, no cleanup! */
1778 	if (!data->rpc_done)
1779 		goto out_free;
1780 	state = nfs4_opendata_to_nfs4_state(data);
1781 	if (!IS_ERR(state))
1782 		nfs4_close_state(state, data->o_arg.fmode);
1783 out_free:
1784 	nfs4_opendata_put(data);
1785 }
1786 
1787 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1788 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1789 	.rpc_call_done = nfs4_open_confirm_done,
1790 	.rpc_release = nfs4_open_confirm_release,
1791 };
1792 
1793 /*
1794  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1795  */
1796 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1797 {
1798 	struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1799 	struct rpc_task *task;
1800 	struct  rpc_message msg = {
1801 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1802 		.rpc_argp = &data->c_arg,
1803 		.rpc_resp = &data->c_res,
1804 		.rpc_cred = data->owner->so_cred,
1805 	};
1806 	struct rpc_task_setup task_setup_data = {
1807 		.rpc_client = server->client,
1808 		.rpc_message = &msg,
1809 		.callback_ops = &nfs4_open_confirm_ops,
1810 		.callback_data = data,
1811 		.workqueue = nfsiod_workqueue,
1812 		.flags = RPC_TASK_ASYNC,
1813 	};
1814 	int status;
1815 
1816 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1817 	kref_get(&data->kref);
1818 	data->rpc_done = 0;
1819 	data->rpc_status = 0;
1820 	data->timestamp = jiffies;
1821 	task = rpc_run_task(&task_setup_data);
1822 	if (IS_ERR(task))
1823 		return PTR_ERR(task);
1824 	status = nfs4_wait_for_completion_rpc_task(task);
1825 	if (status != 0) {
1826 		data->cancelled = 1;
1827 		smp_wmb();
1828 	} else
1829 		status = data->rpc_status;
1830 	rpc_put_task(task);
1831 	return status;
1832 }
1833 
1834 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1835 {
1836 	struct nfs4_opendata *data = calldata;
1837 	struct nfs4_state_owner *sp = data->owner;
1838 	struct nfs_client *clp = sp->so_server->nfs_client;
1839 
1840 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1841 		goto out_wait;
1842 	/*
1843 	 * Check if we still need to send an OPEN call, or if we can use
1844 	 * a delegation instead.
1845 	 */
1846 	if (data->state != NULL) {
1847 		struct nfs_delegation *delegation;
1848 
1849 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1850 			goto out_no_action;
1851 		rcu_read_lock();
1852 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1853 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1854 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1855 		    can_open_delegated(delegation, data->o_arg.fmode))
1856 			goto unlock_no_action;
1857 		rcu_read_unlock();
1858 	}
1859 	/* Update client id. */
1860 	data->o_arg.clientid = clp->cl_clientid;
1861 	switch (data->o_arg.claim) {
1862 	case NFS4_OPEN_CLAIM_PREVIOUS:
1863 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1864 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1865 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1866 	case NFS4_OPEN_CLAIM_FH:
1867 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1868 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1869 	}
1870 	data->timestamp = jiffies;
1871 	if (nfs4_setup_sequence(data->o_arg.server,
1872 				&data->o_arg.seq_args,
1873 				&data->o_res.seq_res,
1874 				task) != 0)
1875 		nfs_release_seqid(data->o_arg.seqid);
1876 
1877 	/* Set the create mode (note dependency on the session type) */
1878 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1879 	if (data->o_arg.open_flags & O_EXCL) {
1880 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1881 		if (nfs4_has_persistent_session(clp))
1882 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1883 		else if (clp->cl_mvops->minor_version > 0)
1884 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1885 	}
1886 	return;
1887 unlock_no_action:
1888 	rcu_read_unlock();
1889 out_no_action:
1890 	task->tk_action = NULL;
1891 out_wait:
1892 	nfs4_sequence_done(task, &data->o_res.seq_res);
1893 }
1894 
1895 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1896 {
1897 	struct nfs4_opendata *data = calldata;
1898 
1899 	data->rpc_status = task->tk_status;
1900 
1901 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1902 		return;
1903 
1904 	if (task->tk_status == 0) {
1905 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1906 			switch (data->o_res.f_attr->mode & S_IFMT) {
1907 			case S_IFREG:
1908 				break;
1909 			case S_IFLNK:
1910 				data->rpc_status = -ELOOP;
1911 				break;
1912 			case S_IFDIR:
1913 				data->rpc_status = -EISDIR;
1914 				break;
1915 			default:
1916 				data->rpc_status = -ENOTDIR;
1917 			}
1918 		}
1919 		renew_lease(data->o_res.server, data->timestamp);
1920 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1921 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1922 	}
1923 	data->rpc_done = 1;
1924 }
1925 
1926 static void nfs4_open_release(void *calldata)
1927 {
1928 	struct nfs4_opendata *data = calldata;
1929 	struct nfs4_state *state = NULL;
1930 
1931 	/* If this request hasn't been cancelled, do nothing */
1932 	if (data->cancelled == 0)
1933 		goto out_free;
1934 	/* In case of error, no cleanup! */
1935 	if (data->rpc_status != 0 || !data->rpc_done)
1936 		goto out_free;
1937 	/* In case we need an open_confirm, no cleanup! */
1938 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1939 		goto out_free;
1940 	state = nfs4_opendata_to_nfs4_state(data);
1941 	if (!IS_ERR(state))
1942 		nfs4_close_state(state, data->o_arg.fmode);
1943 out_free:
1944 	nfs4_opendata_put(data);
1945 }
1946 
1947 static const struct rpc_call_ops nfs4_open_ops = {
1948 	.rpc_call_prepare = nfs4_open_prepare,
1949 	.rpc_call_done = nfs4_open_done,
1950 	.rpc_release = nfs4_open_release,
1951 };
1952 
1953 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1954 {
1955 	struct inode *dir = data->dir->d_inode;
1956 	struct nfs_server *server = NFS_SERVER(dir);
1957 	struct nfs_openargs *o_arg = &data->o_arg;
1958 	struct nfs_openres *o_res = &data->o_res;
1959 	struct rpc_task *task;
1960 	struct rpc_message msg = {
1961 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1962 		.rpc_argp = o_arg,
1963 		.rpc_resp = o_res,
1964 		.rpc_cred = data->owner->so_cred,
1965 	};
1966 	struct rpc_task_setup task_setup_data = {
1967 		.rpc_client = server->client,
1968 		.rpc_message = &msg,
1969 		.callback_ops = &nfs4_open_ops,
1970 		.callback_data = data,
1971 		.workqueue = nfsiod_workqueue,
1972 		.flags = RPC_TASK_ASYNC,
1973 	};
1974 	int status;
1975 
1976 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1977 	kref_get(&data->kref);
1978 	data->rpc_done = 0;
1979 	data->rpc_status = 0;
1980 	data->cancelled = 0;
1981 	data->is_recover = 0;
1982 	if (isrecover) {
1983 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1984 		data->is_recover = 1;
1985 	}
1986 	task = rpc_run_task(&task_setup_data);
1987         if (IS_ERR(task))
1988                 return PTR_ERR(task);
1989         status = nfs4_wait_for_completion_rpc_task(task);
1990         if (status != 0) {
1991                 data->cancelled = 1;
1992                 smp_wmb();
1993         } else
1994                 status = data->rpc_status;
1995         rpc_put_task(task);
1996 
1997 	return status;
1998 }
1999 
2000 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2001 {
2002 	struct inode *dir = data->dir->d_inode;
2003 	struct nfs_openres *o_res = &data->o_res;
2004         int status;
2005 
2006 	status = nfs4_run_open_task(data, 1);
2007 	if (status != 0 || !data->rpc_done)
2008 		return status;
2009 
2010 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2011 
2012 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2013 		status = _nfs4_proc_open_confirm(data);
2014 		if (status != 0)
2015 			return status;
2016 	}
2017 
2018 	return status;
2019 }
2020 
2021 /*
2022  * Additional permission checks in order to distinguish between an
2023  * open for read, and an open for execute. This works around the
2024  * fact that NFSv4 OPEN treats read and execute permissions as being
2025  * the same.
2026  * Note that in the non-execute case, we want to turn off permission
2027  * checking if we just created a new file (POSIX open() semantics).
2028  */
2029 static int nfs4_opendata_access(struct rpc_cred *cred,
2030 				struct nfs4_opendata *opendata,
2031 				struct nfs4_state *state, fmode_t fmode,
2032 				int openflags)
2033 {
2034 	struct nfs_access_entry cache;
2035 	u32 mask;
2036 
2037 	/* access call failed or for some reason the server doesn't
2038 	 * support any access modes -- defer access call until later */
2039 	if (opendata->o_res.access_supported == 0)
2040 		return 0;
2041 
2042 	mask = 0;
2043 	/*
2044 	 * Use openflags to check for exec, because fmode won't
2045 	 * always have FMODE_EXEC set when file open for exec.
2046 	 */
2047 	if (openflags & __FMODE_EXEC) {
2048 		/* ONLY check for exec rights */
2049 		mask = MAY_EXEC;
2050 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2051 		mask = MAY_READ;
2052 
2053 	cache.cred = cred;
2054 	cache.jiffies = jiffies;
2055 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2056 	nfs_access_add_cache(state->inode, &cache);
2057 
2058 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2059 		return 0;
2060 
2061 	/* even though OPEN succeeded, access is denied. Close the file */
2062 	nfs4_close_state(state, fmode);
2063 	return -EACCES;
2064 }
2065 
2066 /*
2067  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2068  */
2069 static int _nfs4_proc_open(struct nfs4_opendata *data)
2070 {
2071 	struct inode *dir = data->dir->d_inode;
2072 	struct nfs_server *server = NFS_SERVER(dir);
2073 	struct nfs_openargs *o_arg = &data->o_arg;
2074 	struct nfs_openres *o_res = &data->o_res;
2075 	int status;
2076 
2077 	status = nfs4_run_open_task(data, 0);
2078 	if (!data->rpc_done)
2079 		return status;
2080 	if (status != 0) {
2081 		if (status == -NFS4ERR_BADNAME &&
2082 				!(o_arg->open_flags & O_CREAT))
2083 			return -ENOENT;
2084 		return status;
2085 	}
2086 
2087 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2088 
2089 	if (o_arg->open_flags & O_CREAT) {
2090 		update_changeattr(dir, &o_res->cinfo);
2091 		if (o_arg->open_flags & O_EXCL)
2092 			data->file_created = 1;
2093 		else if (o_res->cinfo.before != o_res->cinfo.after)
2094 			data->file_created = 1;
2095 	}
2096 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2097 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2098 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2099 		status = _nfs4_proc_open_confirm(data);
2100 		if (status != 0)
2101 			return status;
2102 	}
2103 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2104 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2105 	return 0;
2106 }
2107 
2108 static int nfs4_recover_expired_lease(struct nfs_server *server)
2109 {
2110 	return nfs4_client_recover_expired_lease(server->nfs_client);
2111 }
2112 
2113 /*
2114  * OPEN_EXPIRED:
2115  * 	reclaim state on the server after a network partition.
2116  * 	Assumes caller holds the appropriate lock
2117  */
2118 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2119 {
2120 	struct nfs4_opendata *opendata;
2121 	int ret;
2122 
2123 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2124 			NFS4_OPEN_CLAIM_FH);
2125 	if (IS_ERR(opendata))
2126 		return PTR_ERR(opendata);
2127 	ret = nfs4_open_recover(opendata, state);
2128 	if (ret == -ESTALE)
2129 		d_drop(ctx->dentry);
2130 	nfs4_opendata_put(opendata);
2131 	return ret;
2132 }
2133 
2134 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2135 {
2136 	struct nfs_server *server = NFS_SERVER(state->inode);
2137 	struct nfs4_exception exception = { };
2138 	int err;
2139 
2140 	do {
2141 		err = _nfs4_open_expired(ctx, state);
2142 		trace_nfs4_open_expired(ctx, 0, err);
2143 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2144 			continue;
2145 		switch (err) {
2146 		default:
2147 			goto out;
2148 		case -NFS4ERR_GRACE:
2149 		case -NFS4ERR_DELAY:
2150 			nfs4_handle_exception(server, err, &exception);
2151 			err = 0;
2152 		}
2153 	} while (exception.retry);
2154 out:
2155 	return err;
2156 }
2157 
2158 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2159 {
2160 	struct nfs_open_context *ctx;
2161 	int ret;
2162 
2163 	ctx = nfs4_state_find_open_context(state);
2164 	if (IS_ERR(ctx))
2165 		return -EAGAIN;
2166 	ret = nfs4_do_open_expired(ctx, state);
2167 	put_nfs_open_context(ctx);
2168 	return ret;
2169 }
2170 
2171 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2172 {
2173 	nfs_remove_bad_delegation(state->inode);
2174 	write_seqlock(&state->seqlock);
2175 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2176 	write_sequnlock(&state->seqlock);
2177 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2178 }
2179 
2180 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2181 {
2182 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2183 		nfs_finish_clear_delegation_stateid(state);
2184 }
2185 
2186 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2187 {
2188 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2189 	nfs40_clear_delegation_stateid(state);
2190 	return nfs4_open_expired(sp, state);
2191 }
2192 
2193 #if defined(CONFIG_NFS_V4_1)
2194 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2195 {
2196 	struct nfs_server *server = NFS_SERVER(state->inode);
2197 	nfs4_stateid stateid;
2198 	struct nfs_delegation *delegation;
2199 	struct rpc_cred *cred;
2200 	int status;
2201 
2202 	/* Get the delegation credential for use by test/free_stateid */
2203 	rcu_read_lock();
2204 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2205 	if (delegation == NULL) {
2206 		rcu_read_unlock();
2207 		return;
2208 	}
2209 
2210 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2211 	cred = get_rpccred(delegation->cred);
2212 	rcu_read_unlock();
2213 	status = nfs41_test_stateid(server, &stateid, cred);
2214 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2215 
2216 	if (status != NFS_OK) {
2217 		/* Free the stateid unless the server explicitly
2218 		 * informs us the stateid is unrecognized. */
2219 		if (status != -NFS4ERR_BAD_STATEID)
2220 			nfs41_free_stateid(server, &stateid, cred);
2221 		nfs_finish_clear_delegation_stateid(state);
2222 	}
2223 
2224 	put_rpccred(cred);
2225 }
2226 
2227 /**
2228  * nfs41_check_open_stateid - possibly free an open stateid
2229  *
2230  * @state: NFSv4 state for an inode
2231  *
2232  * Returns NFS_OK if recovery for this stateid is now finished.
2233  * Otherwise a negative NFS4ERR value is returned.
2234  */
2235 static int nfs41_check_open_stateid(struct nfs4_state *state)
2236 {
2237 	struct nfs_server *server = NFS_SERVER(state->inode);
2238 	nfs4_stateid *stateid = &state->open_stateid;
2239 	struct rpc_cred *cred = state->owner->so_cred;
2240 	int status;
2241 
2242 	/* If a state reset has been done, test_stateid is unneeded */
2243 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2244 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2245 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2246 		return -NFS4ERR_BAD_STATEID;
2247 
2248 	status = nfs41_test_stateid(server, stateid, cred);
2249 	trace_nfs4_test_open_stateid(state, NULL, status);
2250 	if (status != NFS_OK) {
2251 		/* Free the stateid unless the server explicitly
2252 		 * informs us the stateid is unrecognized. */
2253 		if (status != -NFS4ERR_BAD_STATEID)
2254 			nfs41_free_stateid(server, stateid, cred);
2255 
2256 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2257 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2258 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2259 		clear_bit(NFS_OPEN_STATE, &state->flags);
2260 	}
2261 	return status;
2262 }
2263 
2264 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2265 {
2266 	int status;
2267 
2268 	nfs41_check_delegation_stateid(state);
2269 	status = nfs41_check_open_stateid(state);
2270 	if (status != NFS_OK)
2271 		status = nfs4_open_expired(sp, state);
2272 	return status;
2273 }
2274 #endif
2275 
2276 /*
2277  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2278  * fields corresponding to attributes that were used to store the verifier.
2279  * Make sure we clobber those fields in the later setattr call
2280  */
2281 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2282 {
2283 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2284 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2285 		sattr->ia_valid |= ATTR_ATIME;
2286 
2287 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2288 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2289 		sattr->ia_valid |= ATTR_MTIME;
2290 }
2291 
2292 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2293 		fmode_t fmode,
2294 		int flags,
2295 		struct nfs_open_context *ctx)
2296 {
2297 	struct nfs4_state_owner *sp = opendata->owner;
2298 	struct nfs_server *server = sp->so_server;
2299 	struct dentry *dentry;
2300 	struct nfs4_state *state;
2301 	unsigned int seq;
2302 	int ret;
2303 
2304 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2305 
2306 	ret = _nfs4_proc_open(opendata);
2307 	if (ret != 0)
2308 		goto out;
2309 
2310 	state = nfs4_opendata_to_nfs4_state(opendata);
2311 	ret = PTR_ERR(state);
2312 	if (IS_ERR(state))
2313 		goto out;
2314 	if (server->caps & NFS_CAP_POSIX_LOCK)
2315 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2316 
2317 	dentry = opendata->dentry;
2318 	if (dentry->d_inode == NULL) {
2319 		/* FIXME: Is this d_drop() ever needed? */
2320 		d_drop(dentry);
2321 		dentry = d_add_unique(dentry, igrab(state->inode));
2322 		if (dentry == NULL) {
2323 			dentry = opendata->dentry;
2324 		} else if (dentry != ctx->dentry) {
2325 			dput(ctx->dentry);
2326 			ctx->dentry = dget(dentry);
2327 		}
2328 		nfs_set_verifier(dentry,
2329 				nfs_save_change_attribute(opendata->dir->d_inode));
2330 	}
2331 
2332 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2333 	if (ret != 0)
2334 		goto out;
2335 
2336 	ctx->state = state;
2337 	if (dentry->d_inode == state->inode) {
2338 		nfs_inode_attach_open_context(ctx);
2339 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2340 			nfs4_schedule_stateid_recovery(server, state);
2341 	}
2342 out:
2343 	return ret;
2344 }
2345 
2346 /*
2347  * Returns a referenced nfs4_state
2348  */
2349 static int _nfs4_do_open(struct inode *dir,
2350 			struct nfs_open_context *ctx,
2351 			int flags,
2352 			struct iattr *sattr,
2353 			struct nfs4_label *label,
2354 			int *opened)
2355 {
2356 	struct nfs4_state_owner  *sp;
2357 	struct nfs4_state     *state = NULL;
2358 	struct nfs_server       *server = NFS_SERVER(dir);
2359 	struct nfs4_opendata *opendata;
2360 	struct dentry *dentry = ctx->dentry;
2361 	struct rpc_cred *cred = ctx->cred;
2362 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2363 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2364 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2365 	struct nfs4_label *olabel = NULL;
2366 	int status;
2367 
2368 	/* Protect against reboot recovery conflicts */
2369 	status = -ENOMEM;
2370 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2371 	if (sp == NULL) {
2372 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2373 		goto out_err;
2374 	}
2375 	status = nfs4_recover_expired_lease(server);
2376 	if (status != 0)
2377 		goto err_put_state_owner;
2378 	if (dentry->d_inode != NULL)
2379 		nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2380 	status = -ENOMEM;
2381 	if (dentry->d_inode)
2382 		claim = NFS4_OPEN_CLAIM_FH;
2383 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2384 			label, claim, GFP_KERNEL);
2385 	if (opendata == NULL)
2386 		goto err_put_state_owner;
2387 
2388 	if (label) {
2389 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2390 		if (IS_ERR(olabel)) {
2391 			status = PTR_ERR(olabel);
2392 			goto err_opendata_put;
2393 		}
2394 	}
2395 
2396 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2397 		if (!opendata->f_attr.mdsthreshold) {
2398 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2399 			if (!opendata->f_attr.mdsthreshold)
2400 				goto err_free_label;
2401 		}
2402 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2403 	}
2404 	if (dentry->d_inode != NULL)
2405 		opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2406 
2407 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2408 	if (status != 0)
2409 		goto err_free_label;
2410 	state = ctx->state;
2411 
2412 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2413 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2414 		nfs4_exclusive_attrset(opendata, sattr);
2415 
2416 		nfs_fattr_init(opendata->o_res.f_attr);
2417 		status = nfs4_do_setattr(state->inode, cred,
2418 				opendata->o_res.f_attr, sattr,
2419 				state, label, olabel);
2420 		if (status == 0) {
2421 			nfs_setattr_update_inode(state->inode, sattr,
2422 					opendata->o_res.f_attr);
2423 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2424 		}
2425 	}
2426 	if (opendata->file_created)
2427 		*opened |= FILE_CREATED;
2428 
2429 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2430 		*ctx_th = opendata->f_attr.mdsthreshold;
2431 		opendata->f_attr.mdsthreshold = NULL;
2432 	}
2433 
2434 	nfs4_label_free(olabel);
2435 
2436 	nfs4_opendata_put(opendata);
2437 	nfs4_put_state_owner(sp);
2438 	return 0;
2439 err_free_label:
2440 	nfs4_label_free(olabel);
2441 err_opendata_put:
2442 	nfs4_opendata_put(opendata);
2443 err_put_state_owner:
2444 	nfs4_put_state_owner(sp);
2445 out_err:
2446 	return status;
2447 }
2448 
2449 
2450 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2451 					struct nfs_open_context *ctx,
2452 					int flags,
2453 					struct iattr *sattr,
2454 					struct nfs4_label *label,
2455 					int *opened)
2456 {
2457 	struct nfs_server *server = NFS_SERVER(dir);
2458 	struct nfs4_exception exception = { };
2459 	struct nfs4_state *res;
2460 	int status;
2461 
2462 	do {
2463 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2464 		res = ctx->state;
2465 		trace_nfs4_open_file(ctx, flags, status);
2466 		if (status == 0)
2467 			break;
2468 		/* NOTE: BAD_SEQID means the server and client disagree about the
2469 		 * book-keeping w.r.t. state-changing operations
2470 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2471 		 * It is actually a sign of a bug on the client or on the server.
2472 		 *
2473 		 * If we receive a BAD_SEQID error in the particular case of
2474 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2475 		 * have unhashed the old state_owner for us, and that we can
2476 		 * therefore safely retry using a new one. We should still warn
2477 		 * the user though...
2478 		 */
2479 		if (status == -NFS4ERR_BAD_SEQID) {
2480 			pr_warn_ratelimited("NFS: v4 server %s "
2481 					" returned a bad sequence-id error!\n",
2482 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2483 			exception.retry = 1;
2484 			continue;
2485 		}
2486 		/*
2487 		 * BAD_STATEID on OPEN means that the server cancelled our
2488 		 * state before it received the OPEN_CONFIRM.
2489 		 * Recover by retrying the request as per the discussion
2490 		 * on Page 181 of RFC3530.
2491 		 */
2492 		if (status == -NFS4ERR_BAD_STATEID) {
2493 			exception.retry = 1;
2494 			continue;
2495 		}
2496 		if (status == -EAGAIN) {
2497 			/* We must have found a delegation */
2498 			exception.retry = 1;
2499 			continue;
2500 		}
2501 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2502 			continue;
2503 		res = ERR_PTR(nfs4_handle_exception(server,
2504 					status, &exception));
2505 	} while (exception.retry);
2506 	return res;
2507 }
2508 
2509 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2510 			    struct nfs_fattr *fattr, struct iattr *sattr,
2511 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2512 			    struct nfs4_label *olabel)
2513 {
2514 	struct nfs_server *server = NFS_SERVER(inode);
2515         struct nfs_setattrargs  arg = {
2516                 .fh             = NFS_FH(inode),
2517                 .iap            = sattr,
2518 		.server		= server,
2519 		.bitmask = server->attr_bitmask,
2520 		.label		= ilabel,
2521         };
2522         struct nfs_setattrres  res = {
2523 		.fattr		= fattr,
2524 		.label		= olabel,
2525 		.server		= server,
2526         };
2527         struct rpc_message msg = {
2528 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2529 		.rpc_argp	= &arg,
2530 		.rpc_resp	= &res,
2531 		.rpc_cred	= cred,
2532         };
2533 	unsigned long timestamp = jiffies;
2534 	fmode_t fmode;
2535 	bool truncate;
2536 	int status;
2537 
2538 	arg.bitmask = nfs4_bitmask(server, ilabel);
2539 	if (ilabel)
2540 		arg.bitmask = nfs4_bitmask(server, olabel);
2541 
2542 	nfs_fattr_init(fattr);
2543 
2544 	/* Servers should only apply open mode checks for file size changes */
2545 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2546 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2547 
2548 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2549 		/* Use that stateid */
2550 	} else if (truncate && state != NULL) {
2551 		struct nfs_lockowner lockowner = {
2552 			.l_owner = current->files,
2553 			.l_pid = current->tgid,
2554 		};
2555 		if (!nfs4_valid_open_stateid(state))
2556 			return -EBADF;
2557 		if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2558 				&lockowner) == -EIO)
2559 			return -EBADF;
2560 	} else
2561 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2562 
2563 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2564 	if (status == 0 && state != NULL)
2565 		renew_lease(server, timestamp);
2566 	return status;
2567 }
2568 
2569 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2570 			   struct nfs_fattr *fattr, struct iattr *sattr,
2571 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2572 			   struct nfs4_label *olabel)
2573 {
2574 	struct nfs_server *server = NFS_SERVER(inode);
2575 	struct nfs4_exception exception = {
2576 		.state = state,
2577 		.inode = inode,
2578 	};
2579 	int err;
2580 	do {
2581 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2582 		trace_nfs4_setattr(inode, err);
2583 		switch (err) {
2584 		case -NFS4ERR_OPENMODE:
2585 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2586 				pr_warn_once("NFSv4: server %s is incorrectly "
2587 						"applying open mode checks to "
2588 						"a SETATTR that is not "
2589 						"changing file size.\n",
2590 						server->nfs_client->cl_hostname);
2591 			}
2592 			if (state && !(state->state & FMODE_WRITE)) {
2593 				err = -EBADF;
2594 				if (sattr->ia_valid & ATTR_OPEN)
2595 					err = -EACCES;
2596 				goto out;
2597 			}
2598 		}
2599 		err = nfs4_handle_exception(server, err, &exception);
2600 	} while (exception.retry);
2601 out:
2602 	return err;
2603 }
2604 
2605 struct nfs4_closedata {
2606 	struct inode *inode;
2607 	struct nfs4_state *state;
2608 	struct nfs_closeargs arg;
2609 	struct nfs_closeres res;
2610 	struct nfs_fattr fattr;
2611 	unsigned long timestamp;
2612 	bool roc;
2613 	u32 roc_barrier;
2614 };
2615 
2616 static void nfs4_free_closedata(void *data)
2617 {
2618 	struct nfs4_closedata *calldata = data;
2619 	struct nfs4_state_owner *sp = calldata->state->owner;
2620 	struct super_block *sb = calldata->state->inode->i_sb;
2621 
2622 	if (calldata->roc)
2623 		pnfs_roc_release(calldata->state->inode);
2624 	nfs4_put_open_state(calldata->state);
2625 	nfs_free_seqid(calldata->arg.seqid);
2626 	nfs4_put_state_owner(sp);
2627 	nfs_sb_deactive(sb);
2628 	kfree(calldata);
2629 }
2630 
2631 static void nfs4_close_done(struct rpc_task *task, void *data)
2632 {
2633 	struct nfs4_closedata *calldata = data;
2634 	struct nfs4_state *state = calldata->state;
2635 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2636 	nfs4_stateid *res_stateid = NULL;
2637 
2638 	dprintk("%s: begin!\n", __func__);
2639 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2640 		return;
2641 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2642         /* hmm. we are done with the inode, and in the process of freeing
2643 	 * the state_owner. we keep this around to process errors
2644 	 */
2645 	switch (task->tk_status) {
2646 		case 0:
2647 			res_stateid = &calldata->res.stateid;
2648 			if (calldata->arg.fmode == 0 && calldata->roc)
2649 				pnfs_roc_set_barrier(state->inode,
2650 						     calldata->roc_barrier);
2651 			renew_lease(server, calldata->timestamp);
2652 			break;
2653 		case -NFS4ERR_ADMIN_REVOKED:
2654 		case -NFS4ERR_STALE_STATEID:
2655 		case -NFS4ERR_OLD_STATEID:
2656 		case -NFS4ERR_BAD_STATEID:
2657 		case -NFS4ERR_EXPIRED:
2658 			if (!nfs4_stateid_match(&calldata->arg.stateid,
2659 						&state->open_stateid)) {
2660 				rpc_restart_call_prepare(task);
2661 				goto out_release;
2662 			}
2663 			if (calldata->arg.fmode == 0)
2664 				break;
2665 		default:
2666 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2667 				rpc_restart_call_prepare(task);
2668 				goto out_release;
2669 			}
2670 	}
2671 	nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2672 out_release:
2673 	nfs_release_seqid(calldata->arg.seqid);
2674 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2675 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2676 }
2677 
2678 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2679 {
2680 	struct nfs4_closedata *calldata = data;
2681 	struct nfs4_state *state = calldata->state;
2682 	struct inode *inode = calldata->inode;
2683 	bool is_rdonly, is_wronly, is_rdwr;
2684 	int call_close = 0;
2685 
2686 	dprintk("%s: begin!\n", __func__);
2687 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2688 		goto out_wait;
2689 
2690 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2691 	spin_lock(&state->owner->so_lock);
2692 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2693 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2694 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2695 	nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2696 	/* Calculate the change in open mode */
2697 	calldata->arg.fmode = 0;
2698 	if (state->n_rdwr == 0) {
2699 		if (state->n_rdonly == 0)
2700 			call_close |= is_rdonly;
2701 		else if (is_rdonly)
2702 			calldata->arg.fmode |= FMODE_READ;
2703 		if (state->n_wronly == 0)
2704 			call_close |= is_wronly;
2705 		else if (is_wronly)
2706 			calldata->arg.fmode |= FMODE_WRITE;
2707 	} else if (is_rdwr)
2708 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2709 
2710 	if (calldata->arg.fmode == 0)
2711 		call_close |= is_rdwr;
2712 
2713 	if (!nfs4_valid_open_stateid(state))
2714 		call_close = 0;
2715 	spin_unlock(&state->owner->so_lock);
2716 
2717 	if (!call_close) {
2718 		/* Note: exit _without_ calling nfs4_close_done */
2719 		goto out_no_action;
2720 	}
2721 
2722 	if (calldata->arg.fmode == 0) {
2723 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2724 		if (calldata->roc &&
2725 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2726 			nfs_release_seqid(calldata->arg.seqid);
2727 			goto out_wait;
2728 		    }
2729 	}
2730 	calldata->arg.share_access =
2731 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
2732 				calldata->arg.fmode, 0);
2733 
2734 	nfs_fattr_init(calldata->res.fattr);
2735 	calldata->timestamp = jiffies;
2736 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2737 				&calldata->arg.seq_args,
2738 				&calldata->res.seq_res,
2739 				task) != 0)
2740 		nfs_release_seqid(calldata->arg.seqid);
2741 	dprintk("%s: done!\n", __func__);
2742 	return;
2743 out_no_action:
2744 	task->tk_action = NULL;
2745 out_wait:
2746 	nfs4_sequence_done(task, &calldata->res.seq_res);
2747 }
2748 
2749 static const struct rpc_call_ops nfs4_close_ops = {
2750 	.rpc_call_prepare = nfs4_close_prepare,
2751 	.rpc_call_done = nfs4_close_done,
2752 	.rpc_release = nfs4_free_closedata,
2753 };
2754 
2755 static bool nfs4_roc(struct inode *inode)
2756 {
2757 	if (!nfs_have_layout(inode))
2758 		return false;
2759 	return pnfs_roc(inode);
2760 }
2761 
2762 /*
2763  * It is possible for data to be read/written from a mem-mapped file
2764  * after the sys_close call (which hits the vfs layer as a flush).
2765  * This means that we can't safely call nfsv4 close on a file until
2766  * the inode is cleared. This in turn means that we are not good
2767  * NFSv4 citizens - we do not indicate to the server to update the file's
2768  * share state even when we are done with one of the three share
2769  * stateid's in the inode.
2770  *
2771  * NOTE: Caller must be holding the sp->so_owner semaphore!
2772  */
2773 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2774 {
2775 	struct nfs_server *server = NFS_SERVER(state->inode);
2776 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2777 	struct nfs4_closedata *calldata;
2778 	struct nfs4_state_owner *sp = state->owner;
2779 	struct rpc_task *task;
2780 	struct rpc_message msg = {
2781 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2782 		.rpc_cred = state->owner->so_cred,
2783 	};
2784 	struct rpc_task_setup task_setup_data = {
2785 		.rpc_client = server->client,
2786 		.rpc_message = &msg,
2787 		.callback_ops = &nfs4_close_ops,
2788 		.workqueue = nfsiod_workqueue,
2789 		.flags = RPC_TASK_ASYNC,
2790 	};
2791 	int status = -ENOMEM;
2792 
2793 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2794 		&task_setup_data.rpc_client, &msg);
2795 
2796 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2797 	if (calldata == NULL)
2798 		goto out;
2799 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2800 	calldata->inode = state->inode;
2801 	calldata->state = state;
2802 	calldata->arg.fh = NFS_FH(state->inode);
2803 	/* Serialization for the sequence id */
2804 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2805 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2806 	if (IS_ERR(calldata->arg.seqid))
2807 		goto out_free_calldata;
2808 	calldata->arg.fmode = 0;
2809 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2810 	calldata->res.fattr = &calldata->fattr;
2811 	calldata->res.seqid = calldata->arg.seqid;
2812 	calldata->res.server = server;
2813 	calldata->roc = nfs4_roc(state->inode);
2814 	nfs_sb_active(calldata->inode->i_sb);
2815 
2816 	msg.rpc_argp = &calldata->arg;
2817 	msg.rpc_resp = &calldata->res;
2818 	task_setup_data.callback_data = calldata;
2819 	task = rpc_run_task(&task_setup_data);
2820 	if (IS_ERR(task))
2821 		return PTR_ERR(task);
2822 	status = 0;
2823 	if (wait)
2824 		status = rpc_wait_for_completion_task(task);
2825 	rpc_put_task(task);
2826 	return status;
2827 out_free_calldata:
2828 	kfree(calldata);
2829 out:
2830 	nfs4_put_open_state(state);
2831 	nfs4_put_state_owner(sp);
2832 	return status;
2833 }
2834 
2835 static struct inode *
2836 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2837 		int open_flags, struct iattr *attr, int *opened)
2838 {
2839 	struct nfs4_state *state;
2840 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2841 
2842 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2843 
2844 	/* Protect against concurrent sillydeletes */
2845 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2846 
2847 	nfs4_label_release_security(label);
2848 
2849 	if (IS_ERR(state))
2850 		return ERR_CAST(state);
2851 	return state->inode;
2852 }
2853 
2854 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2855 {
2856 	if (ctx->state == NULL)
2857 		return;
2858 	if (is_sync)
2859 		nfs4_close_sync(ctx->state, ctx->mode);
2860 	else
2861 		nfs4_close_state(ctx->state, ctx->mode);
2862 }
2863 
2864 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2865 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2866 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2867 
2868 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2869 {
2870 	struct nfs4_server_caps_arg args = {
2871 		.fhandle = fhandle,
2872 	};
2873 	struct nfs4_server_caps_res res = {};
2874 	struct rpc_message msg = {
2875 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2876 		.rpc_argp = &args,
2877 		.rpc_resp = &res,
2878 	};
2879 	int status;
2880 
2881 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2882 	if (status == 0) {
2883 		/* Sanity check the server answers */
2884 		switch (server->nfs_client->cl_minorversion) {
2885 		case 0:
2886 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2887 			res.attr_bitmask[2] = 0;
2888 			break;
2889 		case 1:
2890 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2891 			break;
2892 		case 2:
2893 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2894 		}
2895 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2896 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2897 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2898 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2899 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2900 				NFS_CAP_CTIME|NFS_CAP_MTIME|
2901 				NFS_CAP_SECURITY_LABEL);
2902 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2903 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2904 			server->caps |= NFS_CAP_ACLS;
2905 		if (res.has_links != 0)
2906 			server->caps |= NFS_CAP_HARDLINKS;
2907 		if (res.has_symlinks != 0)
2908 			server->caps |= NFS_CAP_SYMLINKS;
2909 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2910 			server->caps |= NFS_CAP_FILEID;
2911 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2912 			server->caps |= NFS_CAP_MODE;
2913 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2914 			server->caps |= NFS_CAP_NLINK;
2915 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2916 			server->caps |= NFS_CAP_OWNER;
2917 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2918 			server->caps |= NFS_CAP_OWNER_GROUP;
2919 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2920 			server->caps |= NFS_CAP_ATIME;
2921 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2922 			server->caps |= NFS_CAP_CTIME;
2923 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2924 			server->caps |= NFS_CAP_MTIME;
2925 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2926 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2927 			server->caps |= NFS_CAP_SECURITY_LABEL;
2928 #endif
2929 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2930 				sizeof(server->attr_bitmask));
2931 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2932 
2933 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2934 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2935 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2936 		server->cache_consistency_bitmask[2] = 0;
2937 		server->acl_bitmask = res.acl_bitmask;
2938 		server->fh_expire_type = res.fh_expire_type;
2939 	}
2940 
2941 	return status;
2942 }
2943 
2944 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2945 {
2946 	struct nfs4_exception exception = { };
2947 	int err;
2948 	do {
2949 		err = nfs4_handle_exception(server,
2950 				_nfs4_server_capabilities(server, fhandle),
2951 				&exception);
2952 	} while (exception.retry);
2953 	return err;
2954 }
2955 
2956 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2957 		struct nfs_fsinfo *info)
2958 {
2959 	u32 bitmask[3];
2960 	struct nfs4_lookup_root_arg args = {
2961 		.bitmask = bitmask,
2962 	};
2963 	struct nfs4_lookup_res res = {
2964 		.server = server,
2965 		.fattr = info->fattr,
2966 		.fh = fhandle,
2967 	};
2968 	struct rpc_message msg = {
2969 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2970 		.rpc_argp = &args,
2971 		.rpc_resp = &res,
2972 	};
2973 
2974 	bitmask[0] = nfs4_fattr_bitmap[0];
2975 	bitmask[1] = nfs4_fattr_bitmap[1];
2976 	/*
2977 	 * Process the label in the upcoming getfattr
2978 	 */
2979 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2980 
2981 	nfs_fattr_init(info->fattr);
2982 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2983 }
2984 
2985 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2986 		struct nfs_fsinfo *info)
2987 {
2988 	struct nfs4_exception exception = { };
2989 	int err;
2990 	do {
2991 		err = _nfs4_lookup_root(server, fhandle, info);
2992 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2993 		switch (err) {
2994 		case 0:
2995 		case -NFS4ERR_WRONGSEC:
2996 			goto out;
2997 		default:
2998 			err = nfs4_handle_exception(server, err, &exception);
2999 		}
3000 	} while (exception.retry);
3001 out:
3002 	return err;
3003 }
3004 
3005 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3006 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3007 {
3008 	struct rpc_auth_create_args auth_args = {
3009 		.pseudoflavor = flavor,
3010 	};
3011 	struct rpc_auth *auth;
3012 	int ret;
3013 
3014 	auth = rpcauth_create(&auth_args, server->client);
3015 	if (IS_ERR(auth)) {
3016 		ret = -EACCES;
3017 		goto out;
3018 	}
3019 	ret = nfs4_lookup_root(server, fhandle, info);
3020 out:
3021 	return ret;
3022 }
3023 
3024 /*
3025  * Retry pseudoroot lookup with various security flavors.  We do this when:
3026  *
3027  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3028  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3029  *
3030  * Returns zero on success, or a negative NFS4ERR value, or a
3031  * negative errno value.
3032  */
3033 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3034 			      struct nfs_fsinfo *info)
3035 {
3036 	/* Per 3530bis 15.33.5 */
3037 	static const rpc_authflavor_t flav_array[] = {
3038 		RPC_AUTH_GSS_KRB5P,
3039 		RPC_AUTH_GSS_KRB5I,
3040 		RPC_AUTH_GSS_KRB5,
3041 		RPC_AUTH_UNIX,			/* courtesy */
3042 		RPC_AUTH_NULL,
3043 	};
3044 	int status = -EPERM;
3045 	size_t i;
3046 
3047 	if (server->auth_info.flavor_len > 0) {
3048 		/* try each flavor specified by user */
3049 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3050 			status = nfs4_lookup_root_sec(server, fhandle, info,
3051 						server->auth_info.flavors[i]);
3052 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3053 				continue;
3054 			break;
3055 		}
3056 	} else {
3057 		/* no flavors specified by user, try default list */
3058 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3059 			status = nfs4_lookup_root_sec(server, fhandle, info,
3060 						      flav_array[i]);
3061 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3062 				continue;
3063 			break;
3064 		}
3065 	}
3066 
3067 	/*
3068 	 * -EACCESS could mean that the user doesn't have correct permissions
3069 	 * to access the mount.  It could also mean that we tried to mount
3070 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3071 	 * existing mount programs don't handle -EACCES very well so it should
3072 	 * be mapped to -EPERM instead.
3073 	 */
3074 	if (status == -EACCES)
3075 		status = -EPERM;
3076 	return status;
3077 }
3078 
3079 static int nfs4_do_find_root_sec(struct nfs_server *server,
3080 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3081 {
3082 	int mv = server->nfs_client->cl_minorversion;
3083 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3084 }
3085 
3086 /**
3087  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3088  * @server: initialized nfs_server handle
3089  * @fhandle: we fill in the pseudo-fs root file handle
3090  * @info: we fill in an FSINFO struct
3091  * @auth_probe: probe the auth flavours
3092  *
3093  * Returns zero on success, or a negative errno.
3094  */
3095 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3096 			 struct nfs_fsinfo *info,
3097 			 bool auth_probe)
3098 {
3099 	int status;
3100 
3101 	switch (auth_probe) {
3102 	case false:
3103 		status = nfs4_lookup_root(server, fhandle, info);
3104 		if (status != -NFS4ERR_WRONGSEC)
3105 			break;
3106 	default:
3107 		status = nfs4_do_find_root_sec(server, fhandle, info);
3108 	}
3109 
3110 	if (status == 0)
3111 		status = nfs4_server_capabilities(server, fhandle);
3112 	if (status == 0)
3113 		status = nfs4_do_fsinfo(server, fhandle, info);
3114 
3115 	return nfs4_map_errors(status);
3116 }
3117 
3118 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3119 			      struct nfs_fsinfo *info)
3120 {
3121 	int error;
3122 	struct nfs_fattr *fattr = info->fattr;
3123 	struct nfs4_label *label = NULL;
3124 
3125 	error = nfs4_server_capabilities(server, mntfh);
3126 	if (error < 0) {
3127 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3128 		return error;
3129 	}
3130 
3131 	label = nfs4_label_alloc(server, GFP_KERNEL);
3132 	if (IS_ERR(label))
3133 		return PTR_ERR(label);
3134 
3135 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3136 	if (error < 0) {
3137 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3138 		goto err_free_label;
3139 	}
3140 
3141 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3142 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3143 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3144 
3145 err_free_label:
3146 	nfs4_label_free(label);
3147 
3148 	return error;
3149 }
3150 
3151 /*
3152  * Get locations and (maybe) other attributes of a referral.
3153  * Note that we'll actually follow the referral later when
3154  * we detect fsid mismatch in inode revalidation
3155  */
3156 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3157 			     const struct qstr *name, struct nfs_fattr *fattr,
3158 			     struct nfs_fh *fhandle)
3159 {
3160 	int status = -ENOMEM;
3161 	struct page *page = NULL;
3162 	struct nfs4_fs_locations *locations = NULL;
3163 
3164 	page = alloc_page(GFP_KERNEL);
3165 	if (page == NULL)
3166 		goto out;
3167 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3168 	if (locations == NULL)
3169 		goto out;
3170 
3171 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3172 	if (status != 0)
3173 		goto out;
3174 
3175 	/*
3176 	 * If the fsid didn't change, this is a migration event, not a
3177 	 * referral.  Cause us to drop into the exception handler, which
3178 	 * will kick off migration recovery.
3179 	 */
3180 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3181 		dprintk("%s: server did not return a different fsid for"
3182 			" a referral at %s\n", __func__, name->name);
3183 		status = -NFS4ERR_MOVED;
3184 		goto out;
3185 	}
3186 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3187 	nfs_fixup_referral_attributes(&locations->fattr);
3188 
3189 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3190 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3191 	memset(fhandle, 0, sizeof(struct nfs_fh));
3192 out:
3193 	if (page)
3194 		__free_page(page);
3195 	kfree(locations);
3196 	return status;
3197 }
3198 
3199 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3200 				struct nfs_fattr *fattr, struct nfs4_label *label)
3201 {
3202 	struct nfs4_getattr_arg args = {
3203 		.fh = fhandle,
3204 		.bitmask = server->attr_bitmask,
3205 	};
3206 	struct nfs4_getattr_res res = {
3207 		.fattr = fattr,
3208 		.label = label,
3209 		.server = server,
3210 	};
3211 	struct rpc_message msg = {
3212 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3213 		.rpc_argp = &args,
3214 		.rpc_resp = &res,
3215 	};
3216 
3217 	args.bitmask = nfs4_bitmask(server, label);
3218 
3219 	nfs_fattr_init(fattr);
3220 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3221 }
3222 
3223 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3224 				struct nfs_fattr *fattr, struct nfs4_label *label)
3225 {
3226 	struct nfs4_exception exception = { };
3227 	int err;
3228 	do {
3229 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3230 		trace_nfs4_getattr(server, fhandle, fattr, err);
3231 		err = nfs4_handle_exception(server, err,
3232 				&exception);
3233 	} while (exception.retry);
3234 	return err;
3235 }
3236 
3237 /*
3238  * The file is not closed if it is opened due to the a request to change
3239  * the size of the file. The open call will not be needed once the
3240  * VFS layer lookup-intents are implemented.
3241  *
3242  * Close is called when the inode is destroyed.
3243  * If we haven't opened the file for O_WRONLY, we
3244  * need to in the size_change case to obtain a stateid.
3245  *
3246  * Got race?
3247  * Because OPEN is always done by name in nfsv4, it is
3248  * possible that we opened a different file by the same
3249  * name.  We can recognize this race condition, but we
3250  * can't do anything about it besides returning an error.
3251  *
3252  * This will be fixed with VFS changes (lookup-intent).
3253  */
3254 static int
3255 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3256 		  struct iattr *sattr)
3257 {
3258 	struct inode *inode = dentry->d_inode;
3259 	struct rpc_cred *cred = NULL;
3260 	struct nfs4_state *state = NULL;
3261 	struct nfs4_label *label = NULL;
3262 	int status;
3263 
3264 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3265 	    sattr->ia_valid & ATTR_SIZE &&
3266 	    sattr->ia_size < i_size_read(inode))
3267 		pnfs_commit_and_return_layout(inode);
3268 
3269 	nfs_fattr_init(fattr);
3270 
3271 	/* Deal with open(O_TRUNC) */
3272 	if (sattr->ia_valid & ATTR_OPEN)
3273 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3274 
3275 	/* Optimization: if the end result is no change, don't RPC */
3276 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3277 		return 0;
3278 
3279 	/* Search for an existing open(O_WRITE) file */
3280 	if (sattr->ia_valid & ATTR_FILE) {
3281 		struct nfs_open_context *ctx;
3282 
3283 		ctx = nfs_file_open_context(sattr->ia_file);
3284 		if (ctx) {
3285 			cred = ctx->cred;
3286 			state = ctx->state;
3287 		}
3288 	}
3289 
3290 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3291 	if (IS_ERR(label))
3292 		return PTR_ERR(label);
3293 
3294 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3295 	if (status == 0) {
3296 		nfs_setattr_update_inode(inode, sattr, fattr);
3297 		nfs_setsecurity(inode, fattr, label);
3298 	}
3299 	nfs4_label_free(label);
3300 	return status;
3301 }
3302 
3303 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3304 		const struct qstr *name, struct nfs_fh *fhandle,
3305 		struct nfs_fattr *fattr, struct nfs4_label *label)
3306 {
3307 	struct nfs_server *server = NFS_SERVER(dir);
3308 	int		       status;
3309 	struct nfs4_lookup_arg args = {
3310 		.bitmask = server->attr_bitmask,
3311 		.dir_fh = NFS_FH(dir),
3312 		.name = name,
3313 	};
3314 	struct nfs4_lookup_res res = {
3315 		.server = server,
3316 		.fattr = fattr,
3317 		.label = label,
3318 		.fh = fhandle,
3319 	};
3320 	struct rpc_message msg = {
3321 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3322 		.rpc_argp = &args,
3323 		.rpc_resp = &res,
3324 	};
3325 
3326 	args.bitmask = nfs4_bitmask(server, label);
3327 
3328 	nfs_fattr_init(fattr);
3329 
3330 	dprintk("NFS call  lookup %s\n", name->name);
3331 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3332 	dprintk("NFS reply lookup: %d\n", status);
3333 	return status;
3334 }
3335 
3336 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3337 {
3338 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3339 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3340 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3341 	fattr->nlink = 2;
3342 }
3343 
3344 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3345 				   struct qstr *name, struct nfs_fh *fhandle,
3346 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3347 {
3348 	struct nfs4_exception exception = { };
3349 	struct rpc_clnt *client = *clnt;
3350 	int err;
3351 	do {
3352 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3353 		trace_nfs4_lookup(dir, name, err);
3354 		switch (err) {
3355 		case -NFS4ERR_BADNAME:
3356 			err = -ENOENT;
3357 			goto out;
3358 		case -NFS4ERR_MOVED:
3359 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3360 			goto out;
3361 		case -NFS4ERR_WRONGSEC:
3362 			err = -EPERM;
3363 			if (client != *clnt)
3364 				goto out;
3365 			client = nfs4_negotiate_security(client, dir, name);
3366 			if (IS_ERR(client))
3367 				return PTR_ERR(client);
3368 
3369 			exception.retry = 1;
3370 			break;
3371 		default:
3372 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3373 		}
3374 	} while (exception.retry);
3375 
3376 out:
3377 	if (err == 0)
3378 		*clnt = client;
3379 	else if (client != *clnt)
3380 		rpc_shutdown_client(client);
3381 
3382 	return err;
3383 }
3384 
3385 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3386 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3387 			    struct nfs4_label *label)
3388 {
3389 	int status;
3390 	struct rpc_clnt *client = NFS_CLIENT(dir);
3391 
3392 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3393 	if (client != NFS_CLIENT(dir)) {
3394 		rpc_shutdown_client(client);
3395 		nfs_fixup_secinfo_attributes(fattr);
3396 	}
3397 	return status;
3398 }
3399 
3400 struct rpc_clnt *
3401 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3402 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3403 {
3404 	struct rpc_clnt *client = NFS_CLIENT(dir);
3405 	int status;
3406 
3407 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3408 	if (status < 0)
3409 		return ERR_PTR(status);
3410 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3411 }
3412 
3413 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3414 {
3415 	struct nfs_server *server = NFS_SERVER(inode);
3416 	struct nfs4_accessargs args = {
3417 		.fh = NFS_FH(inode),
3418 		.bitmask = server->cache_consistency_bitmask,
3419 	};
3420 	struct nfs4_accessres res = {
3421 		.server = server,
3422 	};
3423 	struct rpc_message msg = {
3424 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3425 		.rpc_argp = &args,
3426 		.rpc_resp = &res,
3427 		.rpc_cred = entry->cred,
3428 	};
3429 	int mode = entry->mask;
3430 	int status = 0;
3431 
3432 	/*
3433 	 * Determine which access bits we want to ask for...
3434 	 */
3435 	if (mode & MAY_READ)
3436 		args.access |= NFS4_ACCESS_READ;
3437 	if (S_ISDIR(inode->i_mode)) {
3438 		if (mode & MAY_WRITE)
3439 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3440 		if (mode & MAY_EXEC)
3441 			args.access |= NFS4_ACCESS_LOOKUP;
3442 	} else {
3443 		if (mode & MAY_WRITE)
3444 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3445 		if (mode & MAY_EXEC)
3446 			args.access |= NFS4_ACCESS_EXECUTE;
3447 	}
3448 
3449 	res.fattr = nfs_alloc_fattr();
3450 	if (res.fattr == NULL)
3451 		return -ENOMEM;
3452 
3453 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3454 	if (!status) {
3455 		nfs_access_set_mask(entry, res.access);
3456 		nfs_refresh_inode(inode, res.fattr);
3457 	}
3458 	nfs_free_fattr(res.fattr);
3459 	return status;
3460 }
3461 
3462 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3463 {
3464 	struct nfs4_exception exception = { };
3465 	int err;
3466 	do {
3467 		err = _nfs4_proc_access(inode, entry);
3468 		trace_nfs4_access(inode, err);
3469 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3470 				&exception);
3471 	} while (exception.retry);
3472 	return err;
3473 }
3474 
3475 /*
3476  * TODO: For the time being, we don't try to get any attributes
3477  * along with any of the zero-copy operations READ, READDIR,
3478  * READLINK, WRITE.
3479  *
3480  * In the case of the first three, we want to put the GETATTR
3481  * after the read-type operation -- this is because it is hard
3482  * to predict the length of a GETATTR response in v4, and thus
3483  * align the READ data correctly.  This means that the GETATTR
3484  * may end up partially falling into the page cache, and we should
3485  * shift it into the 'tail' of the xdr_buf before processing.
3486  * To do this efficiently, we need to know the total length
3487  * of data received, which doesn't seem to be available outside
3488  * of the RPC layer.
3489  *
3490  * In the case of WRITE, we also want to put the GETATTR after
3491  * the operation -- in this case because we want to make sure
3492  * we get the post-operation mtime and size.
3493  *
3494  * Both of these changes to the XDR layer would in fact be quite
3495  * minor, but I decided to leave them for a subsequent patch.
3496  */
3497 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3498 		unsigned int pgbase, unsigned int pglen)
3499 {
3500 	struct nfs4_readlink args = {
3501 		.fh       = NFS_FH(inode),
3502 		.pgbase	  = pgbase,
3503 		.pglen    = pglen,
3504 		.pages    = &page,
3505 	};
3506 	struct nfs4_readlink_res res;
3507 	struct rpc_message msg = {
3508 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3509 		.rpc_argp = &args,
3510 		.rpc_resp = &res,
3511 	};
3512 
3513 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3514 }
3515 
3516 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3517 		unsigned int pgbase, unsigned int pglen)
3518 {
3519 	struct nfs4_exception exception = { };
3520 	int err;
3521 	do {
3522 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3523 		trace_nfs4_readlink(inode, err);
3524 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3525 				&exception);
3526 	} while (exception.retry);
3527 	return err;
3528 }
3529 
3530 /*
3531  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3532  */
3533 static int
3534 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3535 		 int flags)
3536 {
3537 	struct nfs4_label l, *ilabel = NULL;
3538 	struct nfs_open_context *ctx;
3539 	struct nfs4_state *state;
3540 	int opened = 0;
3541 	int status = 0;
3542 
3543 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3544 	if (IS_ERR(ctx))
3545 		return PTR_ERR(ctx);
3546 
3547 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3548 
3549 	sattr->ia_mode &= ~current_umask();
3550 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3551 	if (IS_ERR(state)) {
3552 		status = PTR_ERR(state);
3553 		goto out;
3554 	}
3555 out:
3556 	nfs4_label_release_security(ilabel);
3557 	put_nfs_open_context(ctx);
3558 	return status;
3559 }
3560 
3561 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3562 {
3563 	struct nfs_server *server = NFS_SERVER(dir);
3564 	struct nfs_removeargs args = {
3565 		.fh = NFS_FH(dir),
3566 		.name = *name,
3567 	};
3568 	struct nfs_removeres res = {
3569 		.server = server,
3570 	};
3571 	struct rpc_message msg = {
3572 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3573 		.rpc_argp = &args,
3574 		.rpc_resp = &res,
3575 	};
3576 	int status;
3577 
3578 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3579 	if (status == 0)
3580 		update_changeattr(dir, &res.cinfo);
3581 	return status;
3582 }
3583 
3584 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3585 {
3586 	struct nfs4_exception exception = { };
3587 	int err;
3588 	do {
3589 		err = _nfs4_proc_remove(dir, name);
3590 		trace_nfs4_remove(dir, name, err);
3591 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3592 				&exception);
3593 	} while (exception.retry);
3594 	return err;
3595 }
3596 
3597 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3598 {
3599 	struct nfs_server *server = NFS_SERVER(dir);
3600 	struct nfs_removeargs *args = msg->rpc_argp;
3601 	struct nfs_removeres *res = msg->rpc_resp;
3602 
3603 	res->server = server;
3604 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3605 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3606 
3607 	nfs_fattr_init(res->dir_attr);
3608 }
3609 
3610 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3611 {
3612 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3613 			&data->args.seq_args,
3614 			&data->res.seq_res,
3615 			task);
3616 }
3617 
3618 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3619 {
3620 	struct nfs_unlinkdata *data = task->tk_calldata;
3621 	struct nfs_removeres *res = &data->res;
3622 
3623 	if (!nfs4_sequence_done(task, &res->seq_res))
3624 		return 0;
3625 	if (nfs4_async_handle_error(task, res->server, NULL,
3626 				    &data->timeout) == -EAGAIN)
3627 		return 0;
3628 	update_changeattr(dir, &res->cinfo);
3629 	return 1;
3630 }
3631 
3632 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3633 {
3634 	struct nfs_server *server = NFS_SERVER(dir);
3635 	struct nfs_renameargs *arg = msg->rpc_argp;
3636 	struct nfs_renameres *res = msg->rpc_resp;
3637 
3638 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3639 	res->server = server;
3640 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3641 }
3642 
3643 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3644 {
3645 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3646 			&data->args.seq_args,
3647 			&data->res.seq_res,
3648 			task);
3649 }
3650 
3651 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3652 				 struct inode *new_dir)
3653 {
3654 	struct nfs_renamedata *data = task->tk_calldata;
3655 	struct nfs_renameres *res = &data->res;
3656 
3657 	if (!nfs4_sequence_done(task, &res->seq_res))
3658 		return 0;
3659 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3660 		return 0;
3661 
3662 	update_changeattr(old_dir, &res->old_cinfo);
3663 	update_changeattr(new_dir, &res->new_cinfo);
3664 	return 1;
3665 }
3666 
3667 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3668 {
3669 	struct nfs_server *server = NFS_SERVER(inode);
3670 	struct nfs4_link_arg arg = {
3671 		.fh     = NFS_FH(inode),
3672 		.dir_fh = NFS_FH(dir),
3673 		.name   = name,
3674 		.bitmask = server->attr_bitmask,
3675 	};
3676 	struct nfs4_link_res res = {
3677 		.server = server,
3678 		.label = NULL,
3679 	};
3680 	struct rpc_message msg = {
3681 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3682 		.rpc_argp = &arg,
3683 		.rpc_resp = &res,
3684 	};
3685 	int status = -ENOMEM;
3686 
3687 	res.fattr = nfs_alloc_fattr();
3688 	if (res.fattr == NULL)
3689 		goto out;
3690 
3691 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3692 	if (IS_ERR(res.label)) {
3693 		status = PTR_ERR(res.label);
3694 		goto out;
3695 	}
3696 	arg.bitmask = nfs4_bitmask(server, res.label);
3697 
3698 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3699 	if (!status) {
3700 		update_changeattr(dir, &res.cinfo);
3701 		status = nfs_post_op_update_inode(inode, res.fattr);
3702 		if (!status)
3703 			nfs_setsecurity(inode, res.fattr, res.label);
3704 	}
3705 
3706 
3707 	nfs4_label_free(res.label);
3708 
3709 out:
3710 	nfs_free_fattr(res.fattr);
3711 	return status;
3712 }
3713 
3714 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3715 {
3716 	struct nfs4_exception exception = { };
3717 	int err;
3718 	do {
3719 		err = nfs4_handle_exception(NFS_SERVER(inode),
3720 				_nfs4_proc_link(inode, dir, name),
3721 				&exception);
3722 	} while (exception.retry);
3723 	return err;
3724 }
3725 
3726 struct nfs4_createdata {
3727 	struct rpc_message msg;
3728 	struct nfs4_create_arg arg;
3729 	struct nfs4_create_res res;
3730 	struct nfs_fh fh;
3731 	struct nfs_fattr fattr;
3732 	struct nfs4_label *label;
3733 };
3734 
3735 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3736 		struct qstr *name, struct iattr *sattr, u32 ftype)
3737 {
3738 	struct nfs4_createdata *data;
3739 
3740 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3741 	if (data != NULL) {
3742 		struct nfs_server *server = NFS_SERVER(dir);
3743 
3744 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3745 		if (IS_ERR(data->label))
3746 			goto out_free;
3747 
3748 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3749 		data->msg.rpc_argp = &data->arg;
3750 		data->msg.rpc_resp = &data->res;
3751 		data->arg.dir_fh = NFS_FH(dir);
3752 		data->arg.server = server;
3753 		data->arg.name = name;
3754 		data->arg.attrs = sattr;
3755 		data->arg.ftype = ftype;
3756 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3757 		data->res.server = server;
3758 		data->res.fh = &data->fh;
3759 		data->res.fattr = &data->fattr;
3760 		data->res.label = data->label;
3761 		nfs_fattr_init(data->res.fattr);
3762 	}
3763 	return data;
3764 out_free:
3765 	kfree(data);
3766 	return NULL;
3767 }
3768 
3769 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3770 {
3771 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3772 				    &data->arg.seq_args, &data->res.seq_res, 1);
3773 	if (status == 0) {
3774 		update_changeattr(dir, &data->res.dir_cinfo);
3775 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3776 	}
3777 	return status;
3778 }
3779 
3780 static void nfs4_free_createdata(struct nfs4_createdata *data)
3781 {
3782 	nfs4_label_free(data->label);
3783 	kfree(data);
3784 }
3785 
3786 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3787 		struct page *page, unsigned int len, struct iattr *sattr,
3788 		struct nfs4_label *label)
3789 {
3790 	struct nfs4_createdata *data;
3791 	int status = -ENAMETOOLONG;
3792 
3793 	if (len > NFS4_MAXPATHLEN)
3794 		goto out;
3795 
3796 	status = -ENOMEM;
3797 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3798 	if (data == NULL)
3799 		goto out;
3800 
3801 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3802 	data->arg.u.symlink.pages = &page;
3803 	data->arg.u.symlink.len = len;
3804 	data->arg.label = label;
3805 
3806 	status = nfs4_do_create(dir, dentry, data);
3807 
3808 	nfs4_free_createdata(data);
3809 out:
3810 	return status;
3811 }
3812 
3813 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3814 		struct page *page, unsigned int len, struct iattr *sattr)
3815 {
3816 	struct nfs4_exception exception = { };
3817 	struct nfs4_label l, *label = NULL;
3818 	int err;
3819 
3820 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3821 
3822 	do {
3823 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3824 		trace_nfs4_symlink(dir, &dentry->d_name, err);
3825 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3826 				&exception);
3827 	} while (exception.retry);
3828 
3829 	nfs4_label_release_security(label);
3830 	return err;
3831 }
3832 
3833 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3834 		struct iattr *sattr, struct nfs4_label *label)
3835 {
3836 	struct nfs4_createdata *data;
3837 	int status = -ENOMEM;
3838 
3839 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3840 	if (data == NULL)
3841 		goto out;
3842 
3843 	data->arg.label = label;
3844 	status = nfs4_do_create(dir, dentry, data);
3845 
3846 	nfs4_free_createdata(data);
3847 out:
3848 	return status;
3849 }
3850 
3851 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3852 		struct iattr *sattr)
3853 {
3854 	struct nfs4_exception exception = { };
3855 	struct nfs4_label l, *label = NULL;
3856 	int err;
3857 
3858 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3859 
3860 	sattr->ia_mode &= ~current_umask();
3861 	do {
3862 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3863 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
3864 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3865 				&exception);
3866 	} while (exception.retry);
3867 	nfs4_label_release_security(label);
3868 
3869 	return err;
3870 }
3871 
3872 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3873 		u64 cookie, struct page **pages, unsigned int count, int plus)
3874 {
3875 	struct inode		*dir = dentry->d_inode;
3876 	struct nfs4_readdir_arg args = {
3877 		.fh = NFS_FH(dir),
3878 		.pages = pages,
3879 		.pgbase = 0,
3880 		.count = count,
3881 		.bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3882 		.plus = plus,
3883 	};
3884 	struct nfs4_readdir_res res;
3885 	struct rpc_message msg = {
3886 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3887 		.rpc_argp = &args,
3888 		.rpc_resp = &res,
3889 		.rpc_cred = cred,
3890 	};
3891 	int			status;
3892 
3893 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3894 			dentry,
3895 			(unsigned long long)cookie);
3896 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3897 	res.pgbase = args.pgbase;
3898 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3899 	if (status >= 0) {
3900 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3901 		status += args.pgbase;
3902 	}
3903 
3904 	nfs_invalidate_atime(dir);
3905 
3906 	dprintk("%s: returns %d\n", __func__, status);
3907 	return status;
3908 }
3909 
3910 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3911 		u64 cookie, struct page **pages, unsigned int count, int plus)
3912 {
3913 	struct nfs4_exception exception = { };
3914 	int err;
3915 	do {
3916 		err = _nfs4_proc_readdir(dentry, cred, cookie,
3917 				pages, count, plus);
3918 		trace_nfs4_readdir(dentry->d_inode, err);
3919 		err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3920 				&exception);
3921 	} while (exception.retry);
3922 	return err;
3923 }
3924 
3925 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3926 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3927 {
3928 	struct nfs4_createdata *data;
3929 	int mode = sattr->ia_mode;
3930 	int status = -ENOMEM;
3931 
3932 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3933 	if (data == NULL)
3934 		goto out;
3935 
3936 	if (S_ISFIFO(mode))
3937 		data->arg.ftype = NF4FIFO;
3938 	else if (S_ISBLK(mode)) {
3939 		data->arg.ftype = NF4BLK;
3940 		data->arg.u.device.specdata1 = MAJOR(rdev);
3941 		data->arg.u.device.specdata2 = MINOR(rdev);
3942 	}
3943 	else if (S_ISCHR(mode)) {
3944 		data->arg.ftype = NF4CHR;
3945 		data->arg.u.device.specdata1 = MAJOR(rdev);
3946 		data->arg.u.device.specdata2 = MINOR(rdev);
3947 	} else if (!S_ISSOCK(mode)) {
3948 		status = -EINVAL;
3949 		goto out_free;
3950 	}
3951 
3952 	data->arg.label = label;
3953 	status = nfs4_do_create(dir, dentry, data);
3954 out_free:
3955 	nfs4_free_createdata(data);
3956 out:
3957 	return status;
3958 }
3959 
3960 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3961 		struct iattr *sattr, dev_t rdev)
3962 {
3963 	struct nfs4_exception exception = { };
3964 	struct nfs4_label l, *label = NULL;
3965 	int err;
3966 
3967 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3968 
3969 	sattr->ia_mode &= ~current_umask();
3970 	do {
3971 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3972 		trace_nfs4_mknod(dir, &dentry->d_name, err);
3973 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3974 				&exception);
3975 	} while (exception.retry);
3976 
3977 	nfs4_label_release_security(label);
3978 
3979 	return err;
3980 }
3981 
3982 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3983 		 struct nfs_fsstat *fsstat)
3984 {
3985 	struct nfs4_statfs_arg args = {
3986 		.fh = fhandle,
3987 		.bitmask = server->attr_bitmask,
3988 	};
3989 	struct nfs4_statfs_res res = {
3990 		.fsstat = fsstat,
3991 	};
3992 	struct rpc_message msg = {
3993 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3994 		.rpc_argp = &args,
3995 		.rpc_resp = &res,
3996 	};
3997 
3998 	nfs_fattr_init(fsstat->fattr);
3999 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4000 }
4001 
4002 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4003 {
4004 	struct nfs4_exception exception = { };
4005 	int err;
4006 	do {
4007 		err = nfs4_handle_exception(server,
4008 				_nfs4_proc_statfs(server, fhandle, fsstat),
4009 				&exception);
4010 	} while (exception.retry);
4011 	return err;
4012 }
4013 
4014 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4015 		struct nfs_fsinfo *fsinfo)
4016 {
4017 	struct nfs4_fsinfo_arg args = {
4018 		.fh = fhandle,
4019 		.bitmask = server->attr_bitmask,
4020 	};
4021 	struct nfs4_fsinfo_res res = {
4022 		.fsinfo = fsinfo,
4023 	};
4024 	struct rpc_message msg = {
4025 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4026 		.rpc_argp = &args,
4027 		.rpc_resp = &res,
4028 	};
4029 
4030 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4031 }
4032 
4033 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4034 {
4035 	struct nfs4_exception exception = { };
4036 	unsigned long now = jiffies;
4037 	int err;
4038 
4039 	do {
4040 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4041 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4042 		if (err == 0) {
4043 			struct nfs_client *clp = server->nfs_client;
4044 
4045 			spin_lock(&clp->cl_lock);
4046 			clp->cl_lease_time = fsinfo->lease_time * HZ;
4047 			clp->cl_last_renewal = now;
4048 			spin_unlock(&clp->cl_lock);
4049 			break;
4050 		}
4051 		err = nfs4_handle_exception(server, err, &exception);
4052 	} while (exception.retry);
4053 	return err;
4054 }
4055 
4056 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4057 {
4058 	int error;
4059 
4060 	nfs_fattr_init(fsinfo->fattr);
4061 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4062 	if (error == 0) {
4063 		/* block layout checks this! */
4064 		server->pnfs_blksize = fsinfo->blksize;
4065 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4066 	}
4067 
4068 	return error;
4069 }
4070 
4071 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4072 		struct nfs_pathconf *pathconf)
4073 {
4074 	struct nfs4_pathconf_arg args = {
4075 		.fh = fhandle,
4076 		.bitmask = server->attr_bitmask,
4077 	};
4078 	struct nfs4_pathconf_res res = {
4079 		.pathconf = pathconf,
4080 	};
4081 	struct rpc_message msg = {
4082 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4083 		.rpc_argp = &args,
4084 		.rpc_resp = &res,
4085 	};
4086 
4087 	/* None of the pathconf attributes are mandatory to implement */
4088 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4089 		memset(pathconf, 0, sizeof(*pathconf));
4090 		return 0;
4091 	}
4092 
4093 	nfs_fattr_init(pathconf->fattr);
4094 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4095 }
4096 
4097 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4098 		struct nfs_pathconf *pathconf)
4099 {
4100 	struct nfs4_exception exception = { };
4101 	int err;
4102 
4103 	do {
4104 		err = nfs4_handle_exception(server,
4105 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4106 				&exception);
4107 	} while (exception.retry);
4108 	return err;
4109 }
4110 
4111 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4112 		const struct nfs_open_context *ctx,
4113 		const struct nfs_lock_context *l_ctx,
4114 		fmode_t fmode)
4115 {
4116 	const struct nfs_lockowner *lockowner = NULL;
4117 
4118 	if (l_ctx != NULL)
4119 		lockowner = &l_ctx->lockowner;
4120 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4121 }
4122 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4123 
4124 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4125 		const struct nfs_open_context *ctx,
4126 		const struct nfs_lock_context *l_ctx,
4127 		fmode_t fmode)
4128 {
4129 	nfs4_stateid current_stateid;
4130 
4131 	/* If the current stateid represents a lost lock, then exit */
4132 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4133 		return true;
4134 	return nfs4_stateid_match(stateid, &current_stateid);
4135 }
4136 
4137 static bool nfs4_error_stateid_expired(int err)
4138 {
4139 	switch (err) {
4140 	case -NFS4ERR_DELEG_REVOKED:
4141 	case -NFS4ERR_ADMIN_REVOKED:
4142 	case -NFS4ERR_BAD_STATEID:
4143 	case -NFS4ERR_STALE_STATEID:
4144 	case -NFS4ERR_OLD_STATEID:
4145 	case -NFS4ERR_OPENMODE:
4146 	case -NFS4ERR_EXPIRED:
4147 		return true;
4148 	}
4149 	return false;
4150 }
4151 
4152 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4153 {
4154 	nfs_invalidate_atime(hdr->inode);
4155 }
4156 
4157 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4158 {
4159 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4160 
4161 	trace_nfs4_read(hdr, task->tk_status);
4162 	if (nfs4_async_handle_error(task, server,
4163 				    hdr->args.context->state,
4164 				    NULL) == -EAGAIN) {
4165 		rpc_restart_call_prepare(task);
4166 		return -EAGAIN;
4167 	}
4168 
4169 	__nfs4_read_done_cb(hdr);
4170 	if (task->tk_status > 0)
4171 		renew_lease(server, hdr->timestamp);
4172 	return 0;
4173 }
4174 
4175 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4176 		struct nfs_pgio_args *args)
4177 {
4178 
4179 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4180 		nfs4_stateid_is_current(&args->stateid,
4181 				args->context,
4182 				args->lock_context,
4183 				FMODE_READ))
4184 		return false;
4185 	rpc_restart_call_prepare(task);
4186 	return true;
4187 }
4188 
4189 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4190 {
4191 
4192 	dprintk("--> %s\n", __func__);
4193 
4194 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4195 		return -EAGAIN;
4196 	if (nfs4_read_stateid_changed(task, &hdr->args))
4197 		return -EAGAIN;
4198 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4199 				    nfs4_read_done_cb(task, hdr);
4200 }
4201 
4202 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4203 				 struct rpc_message *msg)
4204 {
4205 	hdr->timestamp   = jiffies;
4206 	hdr->pgio_done_cb = nfs4_read_done_cb;
4207 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4208 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4209 }
4210 
4211 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4212 				      struct nfs_pgio_header *hdr)
4213 {
4214 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4215 			&hdr->args.seq_args,
4216 			&hdr->res.seq_res,
4217 			task))
4218 		return 0;
4219 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4220 				hdr->args.lock_context,
4221 				hdr->rw_ops->rw_mode) == -EIO)
4222 		return -EIO;
4223 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4224 		return -EIO;
4225 	return 0;
4226 }
4227 
4228 static int nfs4_write_done_cb(struct rpc_task *task,
4229 			      struct nfs_pgio_header *hdr)
4230 {
4231 	struct inode *inode = hdr->inode;
4232 
4233 	trace_nfs4_write(hdr, task->tk_status);
4234 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4235 				    hdr->args.context->state,
4236 				    NULL) == -EAGAIN) {
4237 		rpc_restart_call_prepare(task);
4238 		return -EAGAIN;
4239 	}
4240 	if (task->tk_status >= 0) {
4241 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4242 		nfs_writeback_update_inode(hdr);
4243 	}
4244 	return 0;
4245 }
4246 
4247 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4248 		struct nfs_pgio_args *args)
4249 {
4250 
4251 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4252 		nfs4_stateid_is_current(&args->stateid,
4253 				args->context,
4254 				args->lock_context,
4255 				FMODE_WRITE))
4256 		return false;
4257 	rpc_restart_call_prepare(task);
4258 	return true;
4259 }
4260 
4261 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4262 {
4263 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4264 		return -EAGAIN;
4265 	if (nfs4_write_stateid_changed(task, &hdr->args))
4266 		return -EAGAIN;
4267 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4268 		nfs4_write_done_cb(task, hdr);
4269 }
4270 
4271 static
4272 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4273 {
4274 	/* Don't request attributes for pNFS or O_DIRECT writes */
4275 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4276 		return false;
4277 	/* Otherwise, request attributes if and only if we don't hold
4278 	 * a delegation
4279 	 */
4280 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4281 }
4282 
4283 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4284 				  struct rpc_message *msg)
4285 {
4286 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4287 
4288 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4289 		hdr->args.bitmask = NULL;
4290 		hdr->res.fattr = NULL;
4291 	} else
4292 		hdr->args.bitmask = server->cache_consistency_bitmask;
4293 
4294 	if (!hdr->pgio_done_cb)
4295 		hdr->pgio_done_cb = nfs4_write_done_cb;
4296 	hdr->res.server = server;
4297 	hdr->timestamp   = jiffies;
4298 
4299 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4300 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4301 }
4302 
4303 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4304 {
4305 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4306 			&data->args.seq_args,
4307 			&data->res.seq_res,
4308 			task);
4309 }
4310 
4311 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4312 {
4313 	struct inode *inode = data->inode;
4314 
4315 	trace_nfs4_commit(data, task->tk_status);
4316 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4317 				    NULL, NULL) == -EAGAIN) {
4318 		rpc_restart_call_prepare(task);
4319 		return -EAGAIN;
4320 	}
4321 	return 0;
4322 }
4323 
4324 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4325 {
4326 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4327 		return -EAGAIN;
4328 	return data->commit_done_cb(task, data);
4329 }
4330 
4331 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4332 {
4333 	struct nfs_server *server = NFS_SERVER(data->inode);
4334 
4335 	if (data->commit_done_cb == NULL)
4336 		data->commit_done_cb = nfs4_commit_done_cb;
4337 	data->res.server = server;
4338 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4339 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4340 }
4341 
4342 struct nfs4_renewdata {
4343 	struct nfs_client	*client;
4344 	unsigned long		timestamp;
4345 };
4346 
4347 /*
4348  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4349  * standalone procedure for queueing an asynchronous RENEW.
4350  */
4351 static void nfs4_renew_release(void *calldata)
4352 {
4353 	struct nfs4_renewdata *data = calldata;
4354 	struct nfs_client *clp = data->client;
4355 
4356 	if (atomic_read(&clp->cl_count) > 1)
4357 		nfs4_schedule_state_renewal(clp);
4358 	nfs_put_client(clp);
4359 	kfree(data);
4360 }
4361 
4362 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4363 {
4364 	struct nfs4_renewdata *data = calldata;
4365 	struct nfs_client *clp = data->client;
4366 	unsigned long timestamp = data->timestamp;
4367 
4368 	trace_nfs4_renew_async(clp, task->tk_status);
4369 	switch (task->tk_status) {
4370 	case 0:
4371 		break;
4372 	case -NFS4ERR_LEASE_MOVED:
4373 		nfs4_schedule_lease_moved_recovery(clp);
4374 		break;
4375 	default:
4376 		/* Unless we're shutting down, schedule state recovery! */
4377 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4378 			return;
4379 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4380 			nfs4_schedule_lease_recovery(clp);
4381 			return;
4382 		}
4383 		nfs4_schedule_path_down_recovery(clp);
4384 	}
4385 	do_renew_lease(clp, timestamp);
4386 }
4387 
4388 static const struct rpc_call_ops nfs4_renew_ops = {
4389 	.rpc_call_done = nfs4_renew_done,
4390 	.rpc_release = nfs4_renew_release,
4391 };
4392 
4393 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4394 {
4395 	struct rpc_message msg = {
4396 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4397 		.rpc_argp	= clp,
4398 		.rpc_cred	= cred,
4399 	};
4400 	struct nfs4_renewdata *data;
4401 
4402 	if (renew_flags == 0)
4403 		return 0;
4404 	if (!atomic_inc_not_zero(&clp->cl_count))
4405 		return -EIO;
4406 	data = kmalloc(sizeof(*data), GFP_NOFS);
4407 	if (data == NULL)
4408 		return -ENOMEM;
4409 	data->client = clp;
4410 	data->timestamp = jiffies;
4411 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4412 			&nfs4_renew_ops, data);
4413 }
4414 
4415 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4416 {
4417 	struct rpc_message msg = {
4418 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4419 		.rpc_argp	= clp,
4420 		.rpc_cred	= cred,
4421 	};
4422 	unsigned long now = jiffies;
4423 	int status;
4424 
4425 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4426 	if (status < 0)
4427 		return status;
4428 	do_renew_lease(clp, now);
4429 	return 0;
4430 }
4431 
4432 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4433 {
4434 	return server->caps & NFS_CAP_ACLS;
4435 }
4436 
4437 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4438  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4439  * the stack.
4440  */
4441 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4442 
4443 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4444 		struct page **pages, unsigned int *pgbase)
4445 {
4446 	struct page *newpage, **spages;
4447 	int rc = 0;
4448 	size_t len;
4449 	spages = pages;
4450 
4451 	do {
4452 		len = min_t(size_t, PAGE_SIZE, buflen);
4453 		newpage = alloc_page(GFP_KERNEL);
4454 
4455 		if (newpage == NULL)
4456 			goto unwind;
4457 		memcpy(page_address(newpage), buf, len);
4458                 buf += len;
4459                 buflen -= len;
4460 		*pages++ = newpage;
4461 		rc++;
4462 	} while (buflen != 0);
4463 
4464 	return rc;
4465 
4466 unwind:
4467 	for(; rc > 0; rc--)
4468 		__free_page(spages[rc-1]);
4469 	return -ENOMEM;
4470 }
4471 
4472 struct nfs4_cached_acl {
4473 	int cached;
4474 	size_t len;
4475 	char data[0];
4476 };
4477 
4478 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4479 {
4480 	struct nfs_inode *nfsi = NFS_I(inode);
4481 
4482 	spin_lock(&inode->i_lock);
4483 	kfree(nfsi->nfs4_acl);
4484 	nfsi->nfs4_acl = acl;
4485 	spin_unlock(&inode->i_lock);
4486 }
4487 
4488 static void nfs4_zap_acl_attr(struct inode *inode)
4489 {
4490 	nfs4_set_cached_acl(inode, NULL);
4491 }
4492 
4493 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4494 {
4495 	struct nfs_inode *nfsi = NFS_I(inode);
4496 	struct nfs4_cached_acl *acl;
4497 	int ret = -ENOENT;
4498 
4499 	spin_lock(&inode->i_lock);
4500 	acl = nfsi->nfs4_acl;
4501 	if (acl == NULL)
4502 		goto out;
4503 	if (buf == NULL) /* user is just asking for length */
4504 		goto out_len;
4505 	if (acl->cached == 0)
4506 		goto out;
4507 	ret = -ERANGE; /* see getxattr(2) man page */
4508 	if (acl->len > buflen)
4509 		goto out;
4510 	memcpy(buf, acl->data, acl->len);
4511 out_len:
4512 	ret = acl->len;
4513 out:
4514 	spin_unlock(&inode->i_lock);
4515 	return ret;
4516 }
4517 
4518 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4519 {
4520 	struct nfs4_cached_acl *acl;
4521 	size_t buflen = sizeof(*acl) + acl_len;
4522 
4523 	if (buflen <= PAGE_SIZE) {
4524 		acl = kmalloc(buflen, GFP_KERNEL);
4525 		if (acl == NULL)
4526 			goto out;
4527 		acl->cached = 1;
4528 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4529 	} else {
4530 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4531 		if (acl == NULL)
4532 			goto out;
4533 		acl->cached = 0;
4534 	}
4535 	acl->len = acl_len;
4536 out:
4537 	nfs4_set_cached_acl(inode, acl);
4538 }
4539 
4540 /*
4541  * The getxattr API returns the required buffer length when called with a
4542  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4543  * the required buf.  On a NULL buf, we send a page of data to the server
4544  * guessing that the ACL request can be serviced by a page. If so, we cache
4545  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4546  * the cache. If not so, we throw away the page, and cache the required
4547  * length. The next getxattr call will then produce another round trip to
4548  * the server, this time with the input buf of the required size.
4549  */
4550 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4551 {
4552 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4553 	struct nfs_getaclargs args = {
4554 		.fh = NFS_FH(inode),
4555 		.acl_pages = pages,
4556 		.acl_len = buflen,
4557 	};
4558 	struct nfs_getaclres res = {
4559 		.acl_len = buflen,
4560 	};
4561 	struct rpc_message msg = {
4562 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4563 		.rpc_argp = &args,
4564 		.rpc_resp = &res,
4565 	};
4566 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4567 	int ret = -ENOMEM, i;
4568 
4569 	/* As long as we're doing a round trip to the server anyway,
4570 	 * let's be prepared for a page of acl data. */
4571 	if (npages == 0)
4572 		npages = 1;
4573 	if (npages > ARRAY_SIZE(pages))
4574 		return -ERANGE;
4575 
4576 	for (i = 0; i < npages; i++) {
4577 		pages[i] = alloc_page(GFP_KERNEL);
4578 		if (!pages[i])
4579 			goto out_free;
4580 	}
4581 
4582 	/* for decoding across pages */
4583 	res.acl_scratch = alloc_page(GFP_KERNEL);
4584 	if (!res.acl_scratch)
4585 		goto out_free;
4586 
4587 	args.acl_len = npages * PAGE_SIZE;
4588 	args.acl_pgbase = 0;
4589 
4590 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4591 		__func__, buf, buflen, npages, args.acl_len);
4592 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4593 			     &msg, &args.seq_args, &res.seq_res, 0);
4594 	if (ret)
4595 		goto out_free;
4596 
4597 	/* Handle the case where the passed-in buffer is too short */
4598 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4599 		/* Did the user only issue a request for the acl length? */
4600 		if (buf == NULL)
4601 			goto out_ok;
4602 		ret = -ERANGE;
4603 		goto out_free;
4604 	}
4605 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4606 	if (buf) {
4607 		if (res.acl_len > buflen) {
4608 			ret = -ERANGE;
4609 			goto out_free;
4610 		}
4611 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4612 	}
4613 out_ok:
4614 	ret = res.acl_len;
4615 out_free:
4616 	for (i = 0; i < npages; i++)
4617 		if (pages[i])
4618 			__free_page(pages[i]);
4619 	if (res.acl_scratch)
4620 		__free_page(res.acl_scratch);
4621 	return ret;
4622 }
4623 
4624 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4625 {
4626 	struct nfs4_exception exception = { };
4627 	ssize_t ret;
4628 	do {
4629 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4630 		trace_nfs4_get_acl(inode, ret);
4631 		if (ret >= 0)
4632 			break;
4633 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4634 	} while (exception.retry);
4635 	return ret;
4636 }
4637 
4638 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4639 {
4640 	struct nfs_server *server = NFS_SERVER(inode);
4641 	int ret;
4642 
4643 	if (!nfs4_server_supports_acls(server))
4644 		return -EOPNOTSUPP;
4645 	ret = nfs_revalidate_inode(server, inode);
4646 	if (ret < 0)
4647 		return ret;
4648 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4649 		nfs_zap_acl_cache(inode);
4650 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4651 	if (ret != -ENOENT)
4652 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4653 		 * but no cached acl data, just the acl length */
4654 		return ret;
4655 	return nfs4_get_acl_uncached(inode, buf, buflen);
4656 }
4657 
4658 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4659 {
4660 	struct nfs_server *server = NFS_SERVER(inode);
4661 	struct page *pages[NFS4ACL_MAXPAGES];
4662 	struct nfs_setaclargs arg = {
4663 		.fh		= NFS_FH(inode),
4664 		.acl_pages	= pages,
4665 		.acl_len	= buflen,
4666 	};
4667 	struct nfs_setaclres res;
4668 	struct rpc_message msg = {
4669 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4670 		.rpc_argp	= &arg,
4671 		.rpc_resp	= &res,
4672 	};
4673 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4674 	int ret, i;
4675 
4676 	if (!nfs4_server_supports_acls(server))
4677 		return -EOPNOTSUPP;
4678 	if (npages > ARRAY_SIZE(pages))
4679 		return -ERANGE;
4680 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4681 	if (i < 0)
4682 		return i;
4683 	nfs4_inode_return_delegation(inode);
4684 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4685 
4686 	/*
4687 	 * Free each page after tx, so the only ref left is
4688 	 * held by the network stack
4689 	 */
4690 	for (; i > 0; i--)
4691 		put_page(pages[i-1]);
4692 
4693 	/*
4694 	 * Acl update can result in inode attribute update.
4695 	 * so mark the attribute cache invalid.
4696 	 */
4697 	spin_lock(&inode->i_lock);
4698 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4699 	spin_unlock(&inode->i_lock);
4700 	nfs_access_zap_cache(inode);
4701 	nfs_zap_acl_cache(inode);
4702 	return ret;
4703 }
4704 
4705 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4706 {
4707 	struct nfs4_exception exception = { };
4708 	int err;
4709 	do {
4710 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4711 		trace_nfs4_set_acl(inode, err);
4712 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4713 				&exception);
4714 	} while (exception.retry);
4715 	return err;
4716 }
4717 
4718 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4719 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4720 					size_t buflen)
4721 {
4722 	struct nfs_server *server = NFS_SERVER(inode);
4723 	struct nfs_fattr fattr;
4724 	struct nfs4_label label = {0, 0, buflen, buf};
4725 
4726 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4727 	struct nfs4_getattr_arg arg = {
4728 		.fh		= NFS_FH(inode),
4729 		.bitmask	= bitmask,
4730 	};
4731 	struct nfs4_getattr_res res = {
4732 		.fattr		= &fattr,
4733 		.label		= &label,
4734 		.server		= server,
4735 	};
4736 	struct rpc_message msg = {
4737 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4738 		.rpc_argp	= &arg,
4739 		.rpc_resp	= &res,
4740 	};
4741 	int ret;
4742 
4743 	nfs_fattr_init(&fattr);
4744 
4745 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4746 	if (ret)
4747 		return ret;
4748 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4749 		return -ENOENT;
4750 	if (buflen < label.len)
4751 		return -ERANGE;
4752 	return 0;
4753 }
4754 
4755 static int nfs4_get_security_label(struct inode *inode, void *buf,
4756 					size_t buflen)
4757 {
4758 	struct nfs4_exception exception = { };
4759 	int err;
4760 
4761 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4762 		return -EOPNOTSUPP;
4763 
4764 	do {
4765 		err = _nfs4_get_security_label(inode, buf, buflen);
4766 		trace_nfs4_get_security_label(inode, err);
4767 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4768 				&exception);
4769 	} while (exception.retry);
4770 	return err;
4771 }
4772 
4773 static int _nfs4_do_set_security_label(struct inode *inode,
4774 		struct nfs4_label *ilabel,
4775 		struct nfs_fattr *fattr,
4776 		struct nfs4_label *olabel)
4777 {
4778 
4779 	struct iattr sattr = {0};
4780 	struct nfs_server *server = NFS_SERVER(inode);
4781 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4782 	struct nfs_setattrargs arg = {
4783 		.fh             = NFS_FH(inode),
4784 		.iap            = &sattr,
4785 		.server		= server,
4786 		.bitmask	= bitmask,
4787 		.label		= ilabel,
4788 	};
4789 	struct nfs_setattrres res = {
4790 		.fattr		= fattr,
4791 		.label		= olabel,
4792 		.server		= server,
4793 	};
4794 	struct rpc_message msg = {
4795 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4796 		.rpc_argp       = &arg,
4797 		.rpc_resp       = &res,
4798 	};
4799 	int status;
4800 
4801 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4802 
4803 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4804 	if (status)
4805 		dprintk("%s failed: %d\n", __func__, status);
4806 
4807 	return status;
4808 }
4809 
4810 static int nfs4_do_set_security_label(struct inode *inode,
4811 		struct nfs4_label *ilabel,
4812 		struct nfs_fattr *fattr,
4813 		struct nfs4_label *olabel)
4814 {
4815 	struct nfs4_exception exception = { };
4816 	int err;
4817 
4818 	do {
4819 		err = _nfs4_do_set_security_label(inode, ilabel,
4820 				fattr, olabel);
4821 		trace_nfs4_set_security_label(inode, err);
4822 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4823 				&exception);
4824 	} while (exception.retry);
4825 	return err;
4826 }
4827 
4828 static int
4829 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4830 {
4831 	struct nfs4_label ilabel, *olabel = NULL;
4832 	struct nfs_fattr fattr;
4833 	struct rpc_cred *cred;
4834 	struct inode *inode = dentry->d_inode;
4835 	int status;
4836 
4837 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4838 		return -EOPNOTSUPP;
4839 
4840 	nfs_fattr_init(&fattr);
4841 
4842 	ilabel.pi = 0;
4843 	ilabel.lfs = 0;
4844 	ilabel.label = (char *)buf;
4845 	ilabel.len = buflen;
4846 
4847 	cred = rpc_lookup_cred();
4848 	if (IS_ERR(cred))
4849 		return PTR_ERR(cred);
4850 
4851 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4852 	if (IS_ERR(olabel)) {
4853 		status = -PTR_ERR(olabel);
4854 		goto out;
4855 	}
4856 
4857 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4858 	if (status == 0)
4859 		nfs_setsecurity(inode, &fattr, olabel);
4860 
4861 	nfs4_label_free(olabel);
4862 out:
4863 	put_rpccred(cred);
4864 	return status;
4865 }
4866 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4867 
4868 
4869 static int
4870 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4871 			struct nfs4_state *state, long *timeout)
4872 {
4873 	struct nfs_client *clp = server->nfs_client;
4874 
4875 	if (task->tk_status >= 0)
4876 		return 0;
4877 	switch(task->tk_status) {
4878 		case -NFS4ERR_DELEG_REVOKED:
4879 		case -NFS4ERR_ADMIN_REVOKED:
4880 		case -NFS4ERR_BAD_STATEID:
4881 		case -NFS4ERR_OPENMODE:
4882 			if (state == NULL)
4883 				break;
4884 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4885 				goto recovery_failed;
4886 			goto wait_on_recovery;
4887 		case -NFS4ERR_EXPIRED:
4888 			if (state != NULL) {
4889 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4890 					goto recovery_failed;
4891 			}
4892 		case -NFS4ERR_STALE_STATEID:
4893 		case -NFS4ERR_STALE_CLIENTID:
4894 			nfs4_schedule_lease_recovery(clp);
4895 			goto wait_on_recovery;
4896 		case -NFS4ERR_MOVED:
4897 			if (nfs4_schedule_migration_recovery(server) < 0)
4898 				goto recovery_failed;
4899 			goto wait_on_recovery;
4900 		case -NFS4ERR_LEASE_MOVED:
4901 			nfs4_schedule_lease_moved_recovery(clp);
4902 			goto wait_on_recovery;
4903 #if defined(CONFIG_NFS_V4_1)
4904 		case -NFS4ERR_BADSESSION:
4905 		case -NFS4ERR_BADSLOT:
4906 		case -NFS4ERR_BAD_HIGH_SLOT:
4907 		case -NFS4ERR_DEADSESSION:
4908 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4909 		case -NFS4ERR_SEQ_FALSE_RETRY:
4910 		case -NFS4ERR_SEQ_MISORDERED:
4911 			dprintk("%s ERROR %d, Reset session\n", __func__,
4912 				task->tk_status);
4913 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4914 			goto wait_on_recovery;
4915 #endif /* CONFIG_NFS_V4_1 */
4916 		case -NFS4ERR_DELAY:
4917 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4918 			rpc_delay(task, nfs4_update_delay(timeout));
4919 			goto restart_call;
4920 		case -NFS4ERR_GRACE:
4921 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4922 		case -NFS4ERR_RETRY_UNCACHED_REP:
4923 		case -NFS4ERR_OLD_STATEID:
4924 			goto restart_call;
4925 	}
4926 	task->tk_status = nfs4_map_errors(task->tk_status);
4927 	return 0;
4928 recovery_failed:
4929 	task->tk_status = -EIO;
4930 	return 0;
4931 wait_on_recovery:
4932 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4933 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4934 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4935 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4936 		goto recovery_failed;
4937 restart_call:
4938 	task->tk_status = 0;
4939 	return -EAGAIN;
4940 }
4941 
4942 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4943 				    nfs4_verifier *bootverf)
4944 {
4945 	__be32 verf[2];
4946 
4947 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4948 		/* An impossible timestamp guarantees this value
4949 		 * will never match a generated boot time. */
4950 		verf[0] = 0;
4951 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4952 	} else {
4953 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4954 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4955 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4956 	}
4957 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4958 }
4959 
4960 static unsigned int
4961 nfs4_init_nonuniform_client_string(struct nfs_client *clp,
4962 				   char *buf, size_t len)
4963 {
4964 	unsigned int result;
4965 
4966 	if (clp->cl_owner_id != NULL)
4967 		return strlcpy(buf, clp->cl_owner_id, len);
4968 
4969 	rcu_read_lock();
4970 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4971 				clp->cl_ipaddr,
4972 				rpc_peeraddr2str(clp->cl_rpcclient,
4973 							RPC_DISPLAY_ADDR),
4974 				rpc_peeraddr2str(clp->cl_rpcclient,
4975 							RPC_DISPLAY_PROTO));
4976 	rcu_read_unlock();
4977 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4978 	return result;
4979 }
4980 
4981 static unsigned int
4982 nfs4_init_uniform_client_string(struct nfs_client *clp,
4983 				char *buf, size_t len)
4984 {
4985 	const char *nodename = clp->cl_rpcclient->cl_nodename;
4986 	unsigned int result;
4987 
4988 	if (clp->cl_owner_id != NULL)
4989 		return strlcpy(buf, clp->cl_owner_id, len);
4990 
4991 	if (nfs4_client_id_uniquifier[0] != '\0')
4992 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4993 				clp->rpc_ops->version,
4994 				clp->cl_minorversion,
4995 				nfs4_client_id_uniquifier,
4996 				nodename);
4997 	else
4998 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s",
4999 				clp->rpc_ops->version, clp->cl_minorversion,
5000 				nodename);
5001 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
5002 	return result;
5003 }
5004 
5005 /*
5006  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5007  * services.  Advertise one based on the address family of the
5008  * clientaddr.
5009  */
5010 static unsigned int
5011 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5012 {
5013 	if (strchr(clp->cl_ipaddr, ':') != NULL)
5014 		return scnprintf(buf, len, "tcp6");
5015 	else
5016 		return scnprintf(buf, len, "tcp");
5017 }
5018 
5019 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5020 {
5021 	struct nfs4_setclientid *sc = calldata;
5022 
5023 	if (task->tk_status == 0)
5024 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5025 }
5026 
5027 static const struct rpc_call_ops nfs4_setclientid_ops = {
5028 	.rpc_call_done = nfs4_setclientid_done,
5029 };
5030 
5031 /**
5032  * nfs4_proc_setclientid - Negotiate client ID
5033  * @clp: state data structure
5034  * @program: RPC program for NFSv4 callback service
5035  * @port: IP port number for NFS4 callback service
5036  * @cred: RPC credential to use for this call
5037  * @res: where to place the result
5038  *
5039  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5040  */
5041 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5042 		unsigned short port, struct rpc_cred *cred,
5043 		struct nfs4_setclientid_res *res)
5044 {
5045 	nfs4_verifier sc_verifier;
5046 	struct nfs4_setclientid setclientid = {
5047 		.sc_verifier = &sc_verifier,
5048 		.sc_prog = program,
5049 		.sc_cb_ident = clp->cl_cb_ident,
5050 	};
5051 	struct rpc_message msg = {
5052 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5053 		.rpc_argp = &setclientid,
5054 		.rpc_resp = res,
5055 		.rpc_cred = cred,
5056 	};
5057 	struct rpc_task *task;
5058 	struct rpc_task_setup task_setup_data = {
5059 		.rpc_client = clp->cl_rpcclient,
5060 		.rpc_message = &msg,
5061 		.callback_ops = &nfs4_setclientid_ops,
5062 		.callback_data = &setclientid,
5063 		.flags = RPC_TASK_TIMEOUT,
5064 	};
5065 	int status;
5066 
5067 	/* nfs_client_id4 */
5068 	nfs4_init_boot_verifier(clp, &sc_verifier);
5069 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5070 		setclientid.sc_name_len =
5071 				nfs4_init_uniform_client_string(clp,
5072 						setclientid.sc_name,
5073 						sizeof(setclientid.sc_name));
5074 	else
5075 		setclientid.sc_name_len =
5076 				nfs4_init_nonuniform_client_string(clp,
5077 						setclientid.sc_name,
5078 						sizeof(setclientid.sc_name));
5079 	/* cb_client4 */
5080 	setclientid.sc_netid_len =
5081 				nfs4_init_callback_netid(clp,
5082 						setclientid.sc_netid,
5083 						sizeof(setclientid.sc_netid));
5084 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5085 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5086 				clp->cl_ipaddr, port >> 8, port & 255);
5087 
5088 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5089 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5090 		setclientid.sc_name_len, setclientid.sc_name);
5091 	task = rpc_run_task(&task_setup_data);
5092 	if (IS_ERR(task)) {
5093 		status = PTR_ERR(task);
5094 		goto out;
5095 	}
5096 	status = task->tk_status;
5097 	if (setclientid.sc_cred) {
5098 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5099 		put_rpccred(setclientid.sc_cred);
5100 	}
5101 	rpc_put_task(task);
5102 out:
5103 	trace_nfs4_setclientid(clp, status);
5104 	dprintk("NFS reply setclientid: %d\n", status);
5105 	return status;
5106 }
5107 
5108 /**
5109  * nfs4_proc_setclientid_confirm - Confirm client ID
5110  * @clp: state data structure
5111  * @res: result of a previous SETCLIENTID
5112  * @cred: RPC credential to use for this call
5113  *
5114  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5115  */
5116 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5117 		struct nfs4_setclientid_res *arg,
5118 		struct rpc_cred *cred)
5119 {
5120 	struct rpc_message msg = {
5121 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5122 		.rpc_argp = arg,
5123 		.rpc_cred = cred,
5124 	};
5125 	int status;
5126 
5127 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5128 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5129 		clp->cl_clientid);
5130 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5131 	trace_nfs4_setclientid_confirm(clp, status);
5132 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5133 	return status;
5134 }
5135 
5136 struct nfs4_delegreturndata {
5137 	struct nfs4_delegreturnargs args;
5138 	struct nfs4_delegreturnres res;
5139 	struct nfs_fh fh;
5140 	nfs4_stateid stateid;
5141 	unsigned long timestamp;
5142 	struct nfs_fattr fattr;
5143 	int rpc_status;
5144 	struct inode *inode;
5145 	bool roc;
5146 	u32 roc_barrier;
5147 };
5148 
5149 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5150 {
5151 	struct nfs4_delegreturndata *data = calldata;
5152 
5153 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5154 		return;
5155 
5156 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5157 	switch (task->tk_status) {
5158 	case 0:
5159 		renew_lease(data->res.server, data->timestamp);
5160 	case -NFS4ERR_ADMIN_REVOKED:
5161 	case -NFS4ERR_DELEG_REVOKED:
5162 	case -NFS4ERR_BAD_STATEID:
5163 	case -NFS4ERR_OLD_STATEID:
5164 	case -NFS4ERR_STALE_STATEID:
5165 	case -NFS4ERR_EXPIRED:
5166 		task->tk_status = 0;
5167 		if (data->roc)
5168 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5169 		break;
5170 	default:
5171 		if (nfs4_async_handle_error(task, data->res.server,
5172 					    NULL, NULL) == -EAGAIN) {
5173 			rpc_restart_call_prepare(task);
5174 			return;
5175 		}
5176 	}
5177 	data->rpc_status = task->tk_status;
5178 }
5179 
5180 static void nfs4_delegreturn_release(void *calldata)
5181 {
5182 	struct nfs4_delegreturndata *data = calldata;
5183 	struct inode *inode = data->inode;
5184 
5185 	if (inode) {
5186 		if (data->roc)
5187 			pnfs_roc_release(inode);
5188 		nfs_iput_and_deactive(inode);
5189 	}
5190 	kfree(calldata);
5191 }
5192 
5193 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5194 {
5195 	struct nfs4_delegreturndata *d_data;
5196 
5197 	d_data = (struct nfs4_delegreturndata *)data;
5198 
5199 	if (d_data->roc &&
5200 	    pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5201 		return;
5202 
5203 	nfs4_setup_sequence(d_data->res.server,
5204 			&d_data->args.seq_args,
5205 			&d_data->res.seq_res,
5206 			task);
5207 }
5208 
5209 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5210 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5211 	.rpc_call_done = nfs4_delegreturn_done,
5212 	.rpc_release = nfs4_delegreturn_release,
5213 };
5214 
5215 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5216 {
5217 	struct nfs4_delegreturndata *data;
5218 	struct nfs_server *server = NFS_SERVER(inode);
5219 	struct rpc_task *task;
5220 	struct rpc_message msg = {
5221 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5222 		.rpc_cred = cred,
5223 	};
5224 	struct rpc_task_setup task_setup_data = {
5225 		.rpc_client = server->client,
5226 		.rpc_message = &msg,
5227 		.callback_ops = &nfs4_delegreturn_ops,
5228 		.flags = RPC_TASK_ASYNC,
5229 	};
5230 	int status = 0;
5231 
5232 	data = kzalloc(sizeof(*data), GFP_NOFS);
5233 	if (data == NULL)
5234 		return -ENOMEM;
5235 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5236 	data->args.fhandle = &data->fh;
5237 	data->args.stateid = &data->stateid;
5238 	data->args.bitmask = server->cache_consistency_bitmask;
5239 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5240 	nfs4_stateid_copy(&data->stateid, stateid);
5241 	data->res.fattr = &data->fattr;
5242 	data->res.server = server;
5243 	nfs_fattr_init(data->res.fattr);
5244 	data->timestamp = jiffies;
5245 	data->rpc_status = 0;
5246 	data->inode = nfs_igrab_and_active(inode);
5247 	if (data->inode)
5248 		data->roc = nfs4_roc(inode);
5249 
5250 	task_setup_data.callback_data = data;
5251 	msg.rpc_argp = &data->args;
5252 	msg.rpc_resp = &data->res;
5253 	task = rpc_run_task(&task_setup_data);
5254 	if (IS_ERR(task))
5255 		return PTR_ERR(task);
5256 	if (!issync)
5257 		goto out;
5258 	status = nfs4_wait_for_completion_rpc_task(task);
5259 	if (status != 0)
5260 		goto out;
5261 	status = data->rpc_status;
5262 	if (status == 0)
5263 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5264 	else
5265 		nfs_refresh_inode(inode, &data->fattr);
5266 out:
5267 	rpc_put_task(task);
5268 	return status;
5269 }
5270 
5271 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5272 {
5273 	struct nfs_server *server = NFS_SERVER(inode);
5274 	struct nfs4_exception exception = { };
5275 	int err;
5276 	do {
5277 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5278 		trace_nfs4_delegreturn(inode, err);
5279 		switch (err) {
5280 			case -NFS4ERR_STALE_STATEID:
5281 			case -NFS4ERR_EXPIRED:
5282 			case 0:
5283 				return 0;
5284 		}
5285 		err = nfs4_handle_exception(server, err, &exception);
5286 	} while (exception.retry);
5287 	return err;
5288 }
5289 
5290 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5291 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5292 
5293 /*
5294  * sleep, with exponential backoff, and retry the LOCK operation.
5295  */
5296 static unsigned long
5297 nfs4_set_lock_task_retry(unsigned long timeout)
5298 {
5299 	freezable_schedule_timeout_killable_unsafe(timeout);
5300 	timeout <<= 1;
5301 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5302 		return NFS4_LOCK_MAXTIMEOUT;
5303 	return timeout;
5304 }
5305 
5306 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5307 {
5308 	struct inode *inode = state->inode;
5309 	struct nfs_server *server = NFS_SERVER(inode);
5310 	struct nfs_client *clp = server->nfs_client;
5311 	struct nfs_lockt_args arg = {
5312 		.fh = NFS_FH(inode),
5313 		.fl = request,
5314 	};
5315 	struct nfs_lockt_res res = {
5316 		.denied = request,
5317 	};
5318 	struct rpc_message msg = {
5319 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5320 		.rpc_argp       = &arg,
5321 		.rpc_resp       = &res,
5322 		.rpc_cred	= state->owner->so_cred,
5323 	};
5324 	struct nfs4_lock_state *lsp;
5325 	int status;
5326 
5327 	arg.lock_owner.clientid = clp->cl_clientid;
5328 	status = nfs4_set_lock_state(state, request);
5329 	if (status != 0)
5330 		goto out;
5331 	lsp = request->fl_u.nfs4_fl.owner;
5332 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5333 	arg.lock_owner.s_dev = server->s_dev;
5334 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5335 	switch (status) {
5336 		case 0:
5337 			request->fl_type = F_UNLCK;
5338 			break;
5339 		case -NFS4ERR_DENIED:
5340 			status = 0;
5341 	}
5342 	request->fl_ops->fl_release_private(request);
5343 	request->fl_ops = NULL;
5344 out:
5345 	return status;
5346 }
5347 
5348 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5349 {
5350 	struct nfs4_exception exception = { };
5351 	int err;
5352 
5353 	do {
5354 		err = _nfs4_proc_getlk(state, cmd, request);
5355 		trace_nfs4_get_lock(request, state, cmd, err);
5356 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5357 				&exception);
5358 	} while (exception.retry);
5359 	return err;
5360 }
5361 
5362 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5363 {
5364 	int res = 0;
5365 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5366 		case FL_POSIX:
5367 			res = posix_lock_file_wait(file, fl);
5368 			break;
5369 		case FL_FLOCK:
5370 			res = flock_lock_file_wait(file, fl);
5371 			break;
5372 		default:
5373 			BUG();
5374 	}
5375 	return res;
5376 }
5377 
5378 struct nfs4_unlockdata {
5379 	struct nfs_locku_args arg;
5380 	struct nfs_locku_res res;
5381 	struct nfs4_lock_state *lsp;
5382 	struct nfs_open_context *ctx;
5383 	struct file_lock fl;
5384 	const struct nfs_server *server;
5385 	unsigned long timestamp;
5386 };
5387 
5388 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5389 		struct nfs_open_context *ctx,
5390 		struct nfs4_lock_state *lsp,
5391 		struct nfs_seqid *seqid)
5392 {
5393 	struct nfs4_unlockdata *p;
5394 	struct inode *inode = lsp->ls_state->inode;
5395 
5396 	p = kzalloc(sizeof(*p), GFP_NOFS);
5397 	if (p == NULL)
5398 		return NULL;
5399 	p->arg.fh = NFS_FH(inode);
5400 	p->arg.fl = &p->fl;
5401 	p->arg.seqid = seqid;
5402 	p->res.seqid = seqid;
5403 	p->lsp = lsp;
5404 	atomic_inc(&lsp->ls_count);
5405 	/* Ensure we don't close file until we're done freeing locks! */
5406 	p->ctx = get_nfs_open_context(ctx);
5407 	memcpy(&p->fl, fl, sizeof(p->fl));
5408 	p->server = NFS_SERVER(inode);
5409 	return p;
5410 }
5411 
5412 static void nfs4_locku_release_calldata(void *data)
5413 {
5414 	struct nfs4_unlockdata *calldata = data;
5415 	nfs_free_seqid(calldata->arg.seqid);
5416 	nfs4_put_lock_state(calldata->lsp);
5417 	put_nfs_open_context(calldata->ctx);
5418 	kfree(calldata);
5419 }
5420 
5421 static void nfs4_locku_done(struct rpc_task *task, void *data)
5422 {
5423 	struct nfs4_unlockdata *calldata = data;
5424 
5425 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5426 		return;
5427 	switch (task->tk_status) {
5428 		case 0:
5429 			renew_lease(calldata->server, calldata->timestamp);
5430 			do_vfs_lock(calldata->fl.fl_file, &calldata->fl);
5431 			if (nfs4_update_lock_stateid(calldata->lsp,
5432 					&calldata->res.stateid))
5433 				break;
5434 		case -NFS4ERR_BAD_STATEID:
5435 		case -NFS4ERR_OLD_STATEID:
5436 		case -NFS4ERR_STALE_STATEID:
5437 		case -NFS4ERR_EXPIRED:
5438 			if (!nfs4_stateid_match(&calldata->arg.stateid,
5439 						&calldata->lsp->ls_stateid))
5440 				rpc_restart_call_prepare(task);
5441 			break;
5442 		default:
5443 			if (nfs4_async_handle_error(task, calldata->server,
5444 						    NULL, NULL) == -EAGAIN)
5445 				rpc_restart_call_prepare(task);
5446 	}
5447 	nfs_release_seqid(calldata->arg.seqid);
5448 }
5449 
5450 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5451 {
5452 	struct nfs4_unlockdata *calldata = data;
5453 
5454 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5455 		goto out_wait;
5456 	nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5457 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5458 		/* Note: exit _without_ running nfs4_locku_done */
5459 		goto out_no_action;
5460 	}
5461 	calldata->timestamp = jiffies;
5462 	if (nfs4_setup_sequence(calldata->server,
5463 				&calldata->arg.seq_args,
5464 				&calldata->res.seq_res,
5465 				task) != 0)
5466 		nfs_release_seqid(calldata->arg.seqid);
5467 	return;
5468 out_no_action:
5469 	task->tk_action = NULL;
5470 out_wait:
5471 	nfs4_sequence_done(task, &calldata->res.seq_res);
5472 }
5473 
5474 static const struct rpc_call_ops nfs4_locku_ops = {
5475 	.rpc_call_prepare = nfs4_locku_prepare,
5476 	.rpc_call_done = nfs4_locku_done,
5477 	.rpc_release = nfs4_locku_release_calldata,
5478 };
5479 
5480 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5481 		struct nfs_open_context *ctx,
5482 		struct nfs4_lock_state *lsp,
5483 		struct nfs_seqid *seqid)
5484 {
5485 	struct nfs4_unlockdata *data;
5486 	struct rpc_message msg = {
5487 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5488 		.rpc_cred = ctx->cred,
5489 	};
5490 	struct rpc_task_setup task_setup_data = {
5491 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5492 		.rpc_message = &msg,
5493 		.callback_ops = &nfs4_locku_ops,
5494 		.workqueue = nfsiod_workqueue,
5495 		.flags = RPC_TASK_ASYNC,
5496 	};
5497 
5498 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5499 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5500 
5501 	/* Ensure this is an unlock - when canceling a lock, the
5502 	 * canceled lock is passed in, and it won't be an unlock.
5503 	 */
5504 	fl->fl_type = F_UNLCK;
5505 
5506 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5507 	if (data == NULL) {
5508 		nfs_free_seqid(seqid);
5509 		return ERR_PTR(-ENOMEM);
5510 	}
5511 
5512 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5513 	msg.rpc_argp = &data->arg;
5514 	msg.rpc_resp = &data->res;
5515 	task_setup_data.callback_data = data;
5516 	return rpc_run_task(&task_setup_data);
5517 }
5518 
5519 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5520 {
5521 	struct inode *inode = state->inode;
5522 	struct nfs4_state_owner *sp = state->owner;
5523 	struct nfs_inode *nfsi = NFS_I(inode);
5524 	struct nfs_seqid *seqid;
5525 	struct nfs4_lock_state *lsp;
5526 	struct rpc_task *task;
5527 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5528 	int status = 0;
5529 	unsigned char fl_flags = request->fl_flags;
5530 
5531 	status = nfs4_set_lock_state(state, request);
5532 	/* Unlock _before_ we do the RPC call */
5533 	request->fl_flags |= FL_EXISTS;
5534 	/* Exclude nfs_delegation_claim_locks() */
5535 	mutex_lock(&sp->so_delegreturn_mutex);
5536 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5537 	down_read(&nfsi->rwsem);
5538 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5539 		up_read(&nfsi->rwsem);
5540 		mutex_unlock(&sp->so_delegreturn_mutex);
5541 		goto out;
5542 	}
5543 	up_read(&nfsi->rwsem);
5544 	mutex_unlock(&sp->so_delegreturn_mutex);
5545 	if (status != 0)
5546 		goto out;
5547 	/* Is this a delegated lock? */
5548 	lsp = request->fl_u.nfs4_fl.owner;
5549 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5550 		goto out;
5551 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5552 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5553 	status = -ENOMEM;
5554 	if (IS_ERR(seqid))
5555 		goto out;
5556 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5557 	status = PTR_ERR(task);
5558 	if (IS_ERR(task))
5559 		goto out;
5560 	status = nfs4_wait_for_completion_rpc_task(task);
5561 	rpc_put_task(task);
5562 out:
5563 	request->fl_flags = fl_flags;
5564 	trace_nfs4_unlock(request, state, F_SETLK, status);
5565 	return status;
5566 }
5567 
5568 struct nfs4_lockdata {
5569 	struct nfs_lock_args arg;
5570 	struct nfs_lock_res res;
5571 	struct nfs4_lock_state *lsp;
5572 	struct nfs_open_context *ctx;
5573 	struct file_lock fl;
5574 	unsigned long timestamp;
5575 	int rpc_status;
5576 	int cancelled;
5577 	struct nfs_server *server;
5578 };
5579 
5580 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5581 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5582 		gfp_t gfp_mask)
5583 {
5584 	struct nfs4_lockdata *p;
5585 	struct inode *inode = lsp->ls_state->inode;
5586 	struct nfs_server *server = NFS_SERVER(inode);
5587 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5588 
5589 	p = kzalloc(sizeof(*p), gfp_mask);
5590 	if (p == NULL)
5591 		return NULL;
5592 
5593 	p->arg.fh = NFS_FH(inode);
5594 	p->arg.fl = &p->fl;
5595 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5596 	if (IS_ERR(p->arg.open_seqid))
5597 		goto out_free;
5598 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5599 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5600 	if (IS_ERR(p->arg.lock_seqid))
5601 		goto out_free_seqid;
5602 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5603 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5604 	p->arg.lock_owner.s_dev = server->s_dev;
5605 	p->res.lock_seqid = p->arg.lock_seqid;
5606 	p->lsp = lsp;
5607 	p->server = server;
5608 	atomic_inc(&lsp->ls_count);
5609 	p->ctx = get_nfs_open_context(ctx);
5610 	memcpy(&p->fl, fl, sizeof(p->fl));
5611 	return p;
5612 out_free_seqid:
5613 	nfs_free_seqid(p->arg.open_seqid);
5614 out_free:
5615 	kfree(p);
5616 	return NULL;
5617 }
5618 
5619 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5620 {
5621 	struct nfs4_lockdata *data = calldata;
5622 	struct nfs4_state *state = data->lsp->ls_state;
5623 
5624 	dprintk("%s: begin!\n", __func__);
5625 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5626 		goto out_wait;
5627 	/* Do we need to do an open_to_lock_owner? */
5628 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5629 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5630 			goto out_release_lock_seqid;
5631 		}
5632 		nfs4_stateid_copy(&data->arg.open_stateid,
5633 				&state->open_stateid);
5634 		data->arg.new_lock_owner = 1;
5635 		data->res.open_seqid = data->arg.open_seqid;
5636 	} else {
5637 		data->arg.new_lock_owner = 0;
5638 		nfs4_stateid_copy(&data->arg.lock_stateid,
5639 				&data->lsp->ls_stateid);
5640 	}
5641 	if (!nfs4_valid_open_stateid(state)) {
5642 		data->rpc_status = -EBADF;
5643 		task->tk_action = NULL;
5644 		goto out_release_open_seqid;
5645 	}
5646 	data->timestamp = jiffies;
5647 	if (nfs4_setup_sequence(data->server,
5648 				&data->arg.seq_args,
5649 				&data->res.seq_res,
5650 				task) == 0)
5651 		return;
5652 out_release_open_seqid:
5653 	nfs_release_seqid(data->arg.open_seqid);
5654 out_release_lock_seqid:
5655 	nfs_release_seqid(data->arg.lock_seqid);
5656 out_wait:
5657 	nfs4_sequence_done(task, &data->res.seq_res);
5658 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5659 }
5660 
5661 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5662 {
5663 	struct nfs4_lockdata *data = calldata;
5664 	struct nfs4_lock_state *lsp = data->lsp;
5665 
5666 	dprintk("%s: begin!\n", __func__);
5667 
5668 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5669 		return;
5670 
5671 	data->rpc_status = task->tk_status;
5672 	switch (task->tk_status) {
5673 	case 0:
5674 		renew_lease(NFS_SERVER(data->ctx->dentry->d_inode),
5675 				data->timestamp);
5676 		if (data->arg.new_lock) {
5677 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5678 			if (do_vfs_lock(data->fl.fl_file, &data->fl) < 0) {
5679 				rpc_restart_call_prepare(task);
5680 				break;
5681 			}
5682 		}
5683 		if (data->arg.new_lock_owner != 0) {
5684 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
5685 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5686 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5687 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5688 			rpc_restart_call_prepare(task);
5689 		break;
5690 	case -NFS4ERR_BAD_STATEID:
5691 	case -NFS4ERR_OLD_STATEID:
5692 	case -NFS4ERR_STALE_STATEID:
5693 	case -NFS4ERR_EXPIRED:
5694 		if (data->arg.new_lock_owner != 0) {
5695 			if (!nfs4_stateid_match(&data->arg.open_stateid,
5696 						&lsp->ls_state->open_stateid))
5697 				rpc_restart_call_prepare(task);
5698 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5699 						&lsp->ls_stateid))
5700 				rpc_restart_call_prepare(task);
5701 	}
5702 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5703 }
5704 
5705 static void nfs4_lock_release(void *calldata)
5706 {
5707 	struct nfs4_lockdata *data = calldata;
5708 
5709 	dprintk("%s: begin!\n", __func__);
5710 	nfs_free_seqid(data->arg.open_seqid);
5711 	if (data->cancelled != 0) {
5712 		struct rpc_task *task;
5713 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5714 				data->arg.lock_seqid);
5715 		if (!IS_ERR(task))
5716 			rpc_put_task_async(task);
5717 		dprintk("%s: cancelling lock!\n", __func__);
5718 	} else
5719 		nfs_free_seqid(data->arg.lock_seqid);
5720 	nfs4_put_lock_state(data->lsp);
5721 	put_nfs_open_context(data->ctx);
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(dentry->d_inode, 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(dentry->d_inode, 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(dentry->d_inode)))
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(dentry->d_inode, 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(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
6173 		len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
6174 		if (list && len <= list_len)
6175 			security_inode_listsecurity(dentry->d_inode, 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