]> git.kernelconcepts.de Git - karo-tx-linux.git/blob - fs/nfs/nfs4proc.c
ce64a45ab648ae6f0f6d4312e52551397f27a47b
[karo-tx-linux.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/file.h>
42 #include <linux/string.h>
43 #include <linux/ratelimit.h>
44 #include <linux/printk.h>
45 #include <linux/slab.h>
46 #include <linux/sunrpc/clnt.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69
70 #include "nfs4trace.h"
71
72 #define NFSDBG_FACILITY         NFSDBG_PROC
73
74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
76
77 struct nfs4_opendata;
78 static int _nfs4_proc_open(struct nfs4_opendata *data);
79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
83 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
84 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
85 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
86                             struct nfs_fattr *fattr, struct iattr *sattr,
87                             struct nfs4_state *state, struct nfs4_label *ilabel,
88                             struct nfs4_label *olabel);
89 #ifdef CONFIG_NFS_V4_1
90 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
91                 struct rpc_cred *);
92 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
93                 struct rpc_cred *);
94 #endif
95
96 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
97 static inline struct nfs4_label *
98 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
99         struct iattr *sattr, struct nfs4_label *label)
100 {
101         int err;
102
103         if (label == NULL)
104                 return NULL;
105
106         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
107                 return NULL;
108
109         err = security_dentry_init_security(dentry, sattr->ia_mode,
110                                 &dentry->d_name, (void **)&label->label, &label->len);
111         if (err == 0)
112                 return label;
113
114         return NULL;
115 }
116 static inline void
117 nfs4_label_release_security(struct nfs4_label *label)
118 {
119         if (label)
120                 security_release_secctx(label->label, label->len);
121 }
122 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
123 {
124         if (label)
125                 return server->attr_bitmask;
126
127         return server->attr_bitmask_nl;
128 }
129 #else
130 static inline struct nfs4_label *
131 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
132         struct iattr *sattr, struct nfs4_label *l)
133 { return NULL; }
134 static inline void
135 nfs4_label_release_security(struct nfs4_label *label)
136 { return; }
137 static inline u32 *
138 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
139 { return server->attr_bitmask; }
140 #endif
141
142 /* Prevent leaks of NFSv4 errors into userland */
143 static int nfs4_map_errors(int err)
144 {
145         if (err >= -1000)
146                 return err;
147         switch (err) {
148         case -NFS4ERR_RESOURCE:
149         case -NFS4ERR_LAYOUTTRYLATER:
150         case -NFS4ERR_RECALLCONFLICT:
151                 return -EREMOTEIO;
152         case -NFS4ERR_WRONGSEC:
153         case -NFS4ERR_WRONG_CRED:
154                 return -EPERM;
155         case -NFS4ERR_BADOWNER:
156         case -NFS4ERR_BADNAME:
157                 return -EINVAL;
158         case -NFS4ERR_SHARE_DENIED:
159                 return -EACCES;
160         case -NFS4ERR_MINOR_VERS_MISMATCH:
161                 return -EPROTONOSUPPORT;
162         case -NFS4ERR_FILE_OPEN:
163                 return -EBUSY;
164         default:
165                 dprintk("%s could not handle NFSv4 error %d\n",
166                                 __func__, -err);
167                 break;
168         }
169         return -EIO;
170 }
171
172 /*
173  * This is our standard bitmap for GETATTR requests.
174  */
175 const u32 nfs4_fattr_bitmap[3] = {
176         FATTR4_WORD0_TYPE
177         | FATTR4_WORD0_CHANGE
178         | FATTR4_WORD0_SIZE
179         | FATTR4_WORD0_FSID
180         | FATTR4_WORD0_FILEID,
181         FATTR4_WORD1_MODE
182         | FATTR4_WORD1_NUMLINKS
183         | FATTR4_WORD1_OWNER
184         | FATTR4_WORD1_OWNER_GROUP
185         | FATTR4_WORD1_RAWDEV
186         | FATTR4_WORD1_SPACE_USED
187         | FATTR4_WORD1_TIME_ACCESS
188         | FATTR4_WORD1_TIME_METADATA
189         | FATTR4_WORD1_TIME_MODIFY
190         | FATTR4_WORD1_MOUNTED_ON_FILEID,
191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
192         FATTR4_WORD2_SECURITY_LABEL
193 #endif
194 };
195
196 static const u32 nfs4_pnfs_open_bitmap[3] = {
197         FATTR4_WORD0_TYPE
198         | FATTR4_WORD0_CHANGE
199         | FATTR4_WORD0_SIZE
200         | FATTR4_WORD0_FSID
201         | FATTR4_WORD0_FILEID,
202         FATTR4_WORD1_MODE
203         | FATTR4_WORD1_NUMLINKS
204         | FATTR4_WORD1_OWNER
205         | FATTR4_WORD1_OWNER_GROUP
206         | FATTR4_WORD1_RAWDEV
207         | FATTR4_WORD1_SPACE_USED
208         | FATTR4_WORD1_TIME_ACCESS
209         | FATTR4_WORD1_TIME_METADATA
210         | FATTR4_WORD1_TIME_MODIFY,
211         FATTR4_WORD2_MDSTHRESHOLD
212 };
213
214 static const u32 nfs4_open_noattr_bitmap[3] = {
215         FATTR4_WORD0_TYPE
216         | FATTR4_WORD0_CHANGE
217         | FATTR4_WORD0_FILEID,
218 };
219
220 const u32 nfs4_statfs_bitmap[3] = {
221         FATTR4_WORD0_FILES_AVAIL
222         | FATTR4_WORD0_FILES_FREE
223         | FATTR4_WORD0_FILES_TOTAL,
224         FATTR4_WORD1_SPACE_AVAIL
225         | FATTR4_WORD1_SPACE_FREE
226         | FATTR4_WORD1_SPACE_TOTAL
227 };
228
229 const u32 nfs4_pathconf_bitmap[3] = {
230         FATTR4_WORD0_MAXLINK
231         | FATTR4_WORD0_MAXNAME,
232         0
233 };
234
235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
236                         | FATTR4_WORD0_MAXREAD
237                         | FATTR4_WORD0_MAXWRITE
238                         | FATTR4_WORD0_LEASE_TIME,
239                         FATTR4_WORD1_TIME_DELTA
240                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
241                         FATTR4_WORD2_LAYOUT_BLKSIZE
242 };
243
244 const u32 nfs4_fs_locations_bitmap[3] = {
245         FATTR4_WORD0_TYPE
246         | FATTR4_WORD0_CHANGE
247         | FATTR4_WORD0_SIZE
248         | FATTR4_WORD0_FSID
249         | FATTR4_WORD0_FILEID
250         | FATTR4_WORD0_FS_LOCATIONS,
251         FATTR4_WORD1_MODE
252         | FATTR4_WORD1_NUMLINKS
253         | FATTR4_WORD1_OWNER
254         | FATTR4_WORD1_OWNER_GROUP
255         | FATTR4_WORD1_RAWDEV
256         | FATTR4_WORD1_SPACE_USED
257         | FATTR4_WORD1_TIME_ACCESS
258         | FATTR4_WORD1_TIME_METADATA
259         | FATTR4_WORD1_TIME_MODIFY
260         | FATTR4_WORD1_MOUNTED_ON_FILEID,
261 };
262
263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
264                 struct nfs4_readdir_arg *readdir)
265 {
266         __be32 *start, *p;
267
268         if (cookie > 2) {
269                 readdir->cookie = cookie;
270                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
271                 return;
272         }
273
274         readdir->cookie = 0;
275         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
276         if (cookie == 2)
277                 return;
278         
279         /*
280          * NFSv4 servers do not return entries for '.' and '..'
281          * Therefore, we fake these entries here.  We let '.'
282          * have cookie 0 and '..' have cookie 1.  Note that
283          * when talking to the server, we always send cookie 0
284          * instead of 1 or 2.
285          */
286         start = p = kmap_atomic(*readdir->pages);
287         
288         if (cookie == 0) {
289                 *p++ = xdr_one;                                  /* next */
290                 *p++ = xdr_zero;                   /* cookie, first word */
291                 *p++ = xdr_one;                   /* cookie, second word */
292                 *p++ = xdr_one;                             /* entry len */
293                 memcpy(p, ".\0\0\0", 4);                        /* entry */
294                 p++;
295                 *p++ = xdr_one;                         /* bitmap length */
296                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
297                 *p++ = htonl(8);              /* attribute buffer length */
298                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
299         }
300         
301         *p++ = xdr_one;                                  /* next */
302         *p++ = xdr_zero;                   /* cookie, first word */
303         *p++ = xdr_two;                   /* cookie, second word */
304         *p++ = xdr_two;                             /* entry len */
305         memcpy(p, "..\0\0", 4);                         /* entry */
306         p++;
307         *p++ = xdr_one;                         /* bitmap length */
308         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
309         *p++ = htonl(8);              /* attribute buffer length */
310         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
311
312         readdir->pgbase = (char *)p - (char *)start;
313         readdir->count -= readdir->pgbase;
314         kunmap_atomic(start);
315 }
316
317 static long nfs4_update_delay(long *timeout)
318 {
319         long ret;
320         if (!timeout)
321                 return NFS4_POLL_RETRY_MAX;
322         if (*timeout <= 0)
323                 *timeout = NFS4_POLL_RETRY_MIN;
324         if (*timeout > NFS4_POLL_RETRY_MAX)
325                 *timeout = NFS4_POLL_RETRY_MAX;
326         ret = *timeout;
327         *timeout <<= 1;
328         return ret;
329 }
330
331 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
332 {
333         int res = 0;
334
335         might_sleep();
336
337         freezable_schedule_timeout_killable_unsafe(
338                 nfs4_update_delay(timeout));
339         if (fatal_signal_pending(current))
340                 res = -ERESTARTSYS;
341         return res;
342 }
343
344 /* This is the error handling routine for processes that are allowed
345  * to sleep.
346  */
347 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
348 {
349         struct nfs_client *clp = server->nfs_client;
350         struct nfs4_state *state = exception->state;
351         struct inode *inode = exception->inode;
352         int ret = errorcode;
353
354         exception->retry = 0;
355         switch(errorcode) {
356                 case 0:
357                         return 0;
358                 case -NFS4ERR_OPENMODE:
359                 case -NFS4ERR_DELEG_REVOKED:
360                 case -NFS4ERR_ADMIN_REVOKED:
361                 case -NFS4ERR_BAD_STATEID:
362                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
363                                 nfs4_inode_return_delegation(inode);
364                                 exception->retry = 1;
365                                 return 0;
366                         }
367                         if (state == NULL)
368                                 break;
369                         ret = nfs4_schedule_stateid_recovery(server, state);
370                         if (ret < 0)
371                                 break;
372                         goto wait_on_recovery;
373                 case -NFS4ERR_EXPIRED:
374                         if (state != NULL) {
375                                 ret = nfs4_schedule_stateid_recovery(server, state);
376                                 if (ret < 0)
377                                         break;
378                         }
379                 case -NFS4ERR_STALE_STATEID:
380                 case -NFS4ERR_STALE_CLIENTID:
381                         nfs4_schedule_lease_recovery(clp);
382                         goto wait_on_recovery;
383                 case -NFS4ERR_MOVED:
384                         ret = nfs4_schedule_migration_recovery(server);
385                         if (ret < 0)
386                                 break;
387                         goto wait_on_recovery;
388                 case -NFS4ERR_LEASE_MOVED:
389                         nfs4_schedule_lease_moved_recovery(clp);
390                         goto wait_on_recovery;
391 #if defined(CONFIG_NFS_V4_1)
392                 case -NFS4ERR_BADSESSION:
393                 case -NFS4ERR_BADSLOT:
394                 case -NFS4ERR_BAD_HIGH_SLOT:
395                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
396                 case -NFS4ERR_DEADSESSION:
397                 case -NFS4ERR_SEQ_FALSE_RETRY:
398                 case -NFS4ERR_SEQ_MISORDERED:
399                         dprintk("%s ERROR: %d Reset session\n", __func__,
400                                 errorcode);
401                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
402                         goto wait_on_recovery;
403 #endif /* defined(CONFIG_NFS_V4_1) */
404                 case -NFS4ERR_FILE_OPEN:
405                         if (exception->timeout > HZ) {
406                                 /* We have retried a decent amount, time to
407                                  * fail
408                                  */
409                                 ret = -EBUSY;
410                                 break;
411                         }
412                 case -NFS4ERR_GRACE:
413                 case -NFS4ERR_DELAY:
414                         ret = nfs4_delay(server->client, &exception->timeout);
415                         if (ret != 0)
416                                 break;
417                 case -NFS4ERR_RETRY_UNCACHED_REP:
418                 case -NFS4ERR_OLD_STATEID:
419                         exception->retry = 1;
420                         break;
421                 case -NFS4ERR_BADOWNER:
422                         /* The following works around a Linux server bug! */
423                 case -NFS4ERR_BADNAME:
424                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
425                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
426                                 exception->retry = 1;
427                                 printk(KERN_WARNING "NFS: v4 server %s "
428                                                 "does not accept raw "
429                                                 "uid/gids. "
430                                                 "Reenabling the idmapper.\n",
431                                                 server->nfs_client->cl_hostname);
432                         }
433         }
434         /* We failed to handle the error */
435         return nfs4_map_errors(ret);
436 wait_on_recovery:
437         ret = nfs4_wait_clnt_recover(clp);
438         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
439                 return -EIO;
440         if (ret == 0)
441                 exception->retry = 1;
442         return ret;
443 }
444
445 /*
446  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
447  * or 'false' otherwise.
448  */
449 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
450 {
451         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
452
453         if (flavor == RPC_AUTH_GSS_KRB5I ||
454             flavor == RPC_AUTH_GSS_KRB5P)
455                 return true;
456
457         return false;
458 }
459
460 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
461 {
462         spin_lock(&clp->cl_lock);
463         if (time_before(clp->cl_last_renewal,timestamp))
464                 clp->cl_last_renewal = timestamp;
465         spin_unlock(&clp->cl_lock);
466 }
467
468 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
469 {
470         struct nfs_client *clp = server->nfs_client;
471
472         if (!nfs4_has_session(clp))
473                 do_renew_lease(clp, timestamp);
474 }
475
476 struct nfs4_call_sync_data {
477         const struct nfs_server *seq_server;
478         struct nfs4_sequence_args *seq_args;
479         struct nfs4_sequence_res *seq_res;
480 };
481
482 void nfs4_init_sequence(struct nfs4_sequence_args *args,
483                         struct nfs4_sequence_res *res, int cache_reply)
484 {
485         args->sa_slot = NULL;
486         args->sa_cache_this = cache_reply;
487         args->sa_privileged = 0;
488
489         res->sr_slot = NULL;
490 }
491
492 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
493 {
494         args->sa_privileged = 1;
495 }
496
497 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
498                          struct nfs4_sequence_args *args,
499                          struct nfs4_sequence_res *res,
500                          struct rpc_task *task)
501 {
502         struct nfs4_slot *slot;
503
504         /* slot already allocated? */
505         if (res->sr_slot != NULL)
506                 goto out_start;
507
508         spin_lock(&tbl->slot_tbl_lock);
509         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
510                 goto out_sleep;
511
512         slot = nfs4_alloc_slot(tbl);
513         if (IS_ERR(slot)) {
514                 if (slot == ERR_PTR(-ENOMEM))
515                         task->tk_timeout = HZ >> 2;
516                 goto out_sleep;
517         }
518         spin_unlock(&tbl->slot_tbl_lock);
519
520         args->sa_slot = slot;
521         res->sr_slot = slot;
522
523 out_start:
524         rpc_call_start(task);
525         return 0;
526
527 out_sleep:
528         if (args->sa_privileged)
529                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
530                                 NULL, RPC_PRIORITY_PRIVILEGED);
531         else
532                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
533         spin_unlock(&tbl->slot_tbl_lock);
534         return -EAGAIN;
535 }
536 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
537
538 static int nfs40_sequence_done(struct rpc_task *task,
539                                struct nfs4_sequence_res *res)
540 {
541         struct nfs4_slot *slot = res->sr_slot;
542         struct nfs4_slot_table *tbl;
543
544         if (slot == NULL)
545                 goto out;
546
547         tbl = slot->table;
548         spin_lock(&tbl->slot_tbl_lock);
549         if (!nfs41_wake_and_assign_slot(tbl, slot))
550                 nfs4_free_slot(tbl, slot);
551         spin_unlock(&tbl->slot_tbl_lock);
552
553         res->sr_slot = NULL;
554 out:
555         return 1;
556 }
557
558 #if defined(CONFIG_NFS_V4_1)
559
560 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
561 {
562         struct nfs4_session *session;
563         struct nfs4_slot_table *tbl;
564         struct nfs4_slot *slot = res->sr_slot;
565         bool send_new_highest_used_slotid = false;
566
567         tbl = slot->table;
568         session = tbl->session;
569
570         spin_lock(&tbl->slot_tbl_lock);
571         /* Be nice to the server: try to ensure that the last transmitted
572          * value for highest_user_slotid <= target_highest_slotid
573          */
574         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
575                 send_new_highest_used_slotid = true;
576
577         if (nfs41_wake_and_assign_slot(tbl, slot)) {
578                 send_new_highest_used_slotid = false;
579                 goto out_unlock;
580         }
581         nfs4_free_slot(tbl, slot);
582
583         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
584                 send_new_highest_used_slotid = false;
585 out_unlock:
586         spin_unlock(&tbl->slot_tbl_lock);
587         res->sr_slot = NULL;
588         if (send_new_highest_used_slotid)
589                 nfs41_notify_server(session->clp);
590 }
591
592 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
593 {
594         struct nfs4_session *session;
595         struct nfs4_slot *slot = res->sr_slot;
596         struct nfs_client *clp;
597         bool interrupted = false;
598         int ret = 1;
599
600         if (slot == NULL)
601                 goto out_noaction;
602         /* don't increment the sequence number if the task wasn't sent */
603         if (!RPC_WAS_SENT(task))
604                 goto out;
605
606         session = slot->table->session;
607
608         if (slot->interrupted) {
609                 slot->interrupted = 0;
610                 interrupted = true;
611         }
612
613         trace_nfs4_sequence_done(session, res);
614         /* Check the SEQUENCE operation status */
615         switch (res->sr_status) {
616         case 0:
617                 /* Update the slot's sequence and clientid lease timer */
618                 ++slot->seq_nr;
619                 clp = session->clp;
620                 do_renew_lease(clp, res->sr_timestamp);
621                 /* Check sequence flags */
622                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
623                 nfs41_update_target_slotid(slot->table, slot, res);
624                 break;
625         case 1:
626                 /*
627                  * sr_status remains 1 if an RPC level error occurred.
628                  * The server may or may not have processed the sequence
629                  * operation..
630                  * Mark the slot as having hosted an interrupted RPC call.
631                  */
632                 slot->interrupted = 1;
633                 goto out;
634         case -NFS4ERR_DELAY:
635                 /* The server detected a resend of the RPC call and
636                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
637                  * of RFC5661.
638                  */
639                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
640                         __func__,
641                         slot->slot_nr,
642                         slot->seq_nr);
643                 goto out_retry;
644         case -NFS4ERR_BADSLOT:
645                 /*
646                  * The slot id we used was probably retired. Try again
647                  * using a different slot id.
648                  */
649                 goto retry_nowait;
650         case -NFS4ERR_SEQ_MISORDERED:
651                 /*
652                  * Was the last operation on this sequence interrupted?
653                  * If so, retry after bumping the sequence number.
654                  */
655                 if (interrupted) {
656                         ++slot->seq_nr;
657                         goto retry_nowait;
658                 }
659                 /*
660                  * Could this slot have been previously retired?
661                  * If so, then the server may be expecting seq_nr = 1!
662                  */
663                 if (slot->seq_nr != 1) {
664                         slot->seq_nr = 1;
665                         goto retry_nowait;
666                 }
667                 break;
668         case -NFS4ERR_SEQ_FALSE_RETRY:
669                 ++slot->seq_nr;
670                 goto retry_nowait;
671         default:
672                 /* Just update the slot sequence no. */
673                 ++slot->seq_nr;
674         }
675 out:
676         /* The session may be reset by one of the error handlers. */
677         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
678         nfs41_sequence_free_slot(res);
679 out_noaction:
680         return ret;
681 retry_nowait:
682         if (rpc_restart_call_prepare(task)) {
683                 task->tk_status = 0;
684                 ret = 0;
685         }
686         goto out;
687 out_retry:
688         if (!rpc_restart_call(task))
689                 goto out;
690         rpc_delay(task, NFS4_POLL_RETRY_MAX);
691         return 0;
692 }
693 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
694
695 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
696 {
697         if (res->sr_slot == NULL)
698                 return 1;
699         if (!res->sr_slot->table->session)
700                 return nfs40_sequence_done(task, res);
701         return nfs41_sequence_done(task, res);
702 }
703 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
704
705 int nfs41_setup_sequence(struct nfs4_session *session,
706                                 struct nfs4_sequence_args *args,
707                                 struct nfs4_sequence_res *res,
708                                 struct rpc_task *task)
709 {
710         struct nfs4_slot *slot;
711         struct nfs4_slot_table *tbl;
712
713         dprintk("--> %s\n", __func__);
714         /* slot already allocated? */
715         if (res->sr_slot != NULL)
716                 goto out_success;
717
718         tbl = &session->fc_slot_table;
719
720         task->tk_timeout = 0;
721
722         spin_lock(&tbl->slot_tbl_lock);
723         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
724             !args->sa_privileged) {
725                 /* The state manager will wait until the slot table is empty */
726                 dprintk("%s session is draining\n", __func__);
727                 goto out_sleep;
728         }
729
730         slot = nfs4_alloc_slot(tbl);
731         if (IS_ERR(slot)) {
732                 /* If out of memory, try again in 1/4 second */
733                 if (slot == ERR_PTR(-ENOMEM))
734                         task->tk_timeout = HZ >> 2;
735                 dprintk("<-- %s: no free slots\n", __func__);
736                 goto out_sleep;
737         }
738         spin_unlock(&tbl->slot_tbl_lock);
739
740         args->sa_slot = slot;
741
742         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
743                         slot->slot_nr, slot->seq_nr);
744
745         res->sr_slot = slot;
746         res->sr_timestamp = jiffies;
747         res->sr_status_flags = 0;
748         /*
749          * sr_status is only set in decode_sequence, and so will remain
750          * set to 1 if an rpc level failure occurs.
751          */
752         res->sr_status = 1;
753         trace_nfs4_setup_sequence(session, args);
754 out_success:
755         rpc_call_start(task);
756         return 0;
757 out_sleep:
758         /* Privileged tasks are queued with top priority */
759         if (args->sa_privileged)
760                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
761                                 NULL, RPC_PRIORITY_PRIVILEGED);
762         else
763                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
764         spin_unlock(&tbl->slot_tbl_lock);
765         return -EAGAIN;
766 }
767 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
768
769 static int nfs4_setup_sequence(const struct nfs_server *server,
770                                struct nfs4_sequence_args *args,
771                                struct nfs4_sequence_res *res,
772                                struct rpc_task *task)
773 {
774         struct nfs4_session *session = nfs4_get_session(server);
775         int ret = 0;
776
777         if (!session)
778                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
779                                             args, res, task);
780
781         dprintk("--> %s clp %p session %p sr_slot %u\n",
782                 __func__, session->clp, session, res->sr_slot ?
783                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
784
785         ret = nfs41_setup_sequence(session, args, res, task);
786
787         dprintk("<-- %s status=%d\n", __func__, ret);
788         return ret;
789 }
790
791 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
792 {
793         struct nfs4_call_sync_data *data = calldata;
794         struct nfs4_session *session = nfs4_get_session(data->seq_server);
795
796         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
797
798         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
799 }
800
801 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
802 {
803         struct nfs4_call_sync_data *data = calldata;
804
805         nfs41_sequence_done(task, data->seq_res);
806 }
807
808 static const struct rpc_call_ops nfs41_call_sync_ops = {
809         .rpc_call_prepare = nfs41_call_sync_prepare,
810         .rpc_call_done = nfs41_call_sync_done,
811 };
812
813 #else   /* !CONFIG_NFS_V4_1 */
814
815 static int nfs4_setup_sequence(const struct nfs_server *server,
816                                struct nfs4_sequence_args *args,
817                                struct nfs4_sequence_res *res,
818                                struct rpc_task *task)
819 {
820         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
821                                     args, res, task);
822 }
823
824 int nfs4_sequence_done(struct rpc_task *task,
825                        struct nfs4_sequence_res *res)
826 {
827         return nfs40_sequence_done(task, res);
828 }
829 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
830
831 #endif  /* !CONFIG_NFS_V4_1 */
832
833 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
834 {
835         struct nfs4_call_sync_data *data = calldata;
836         nfs4_setup_sequence(data->seq_server,
837                                 data->seq_args, data->seq_res, task);
838 }
839
840 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
841 {
842         struct nfs4_call_sync_data *data = calldata;
843         nfs4_sequence_done(task, data->seq_res);
844 }
845
846 static const struct rpc_call_ops nfs40_call_sync_ops = {
847         .rpc_call_prepare = nfs40_call_sync_prepare,
848         .rpc_call_done = nfs40_call_sync_done,
849 };
850
851 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
852                                    struct nfs_server *server,
853                                    struct rpc_message *msg,
854                                    struct nfs4_sequence_args *args,
855                                    struct nfs4_sequence_res *res)
856 {
857         int ret;
858         struct rpc_task *task;
859         struct nfs_client *clp = server->nfs_client;
860         struct nfs4_call_sync_data data = {
861                 .seq_server = server,
862                 .seq_args = args,
863                 .seq_res = res,
864         };
865         struct rpc_task_setup task_setup = {
866                 .rpc_client = clnt,
867                 .rpc_message = msg,
868                 .callback_ops = clp->cl_mvops->call_sync_ops,
869                 .callback_data = &data
870         };
871
872         task = rpc_run_task(&task_setup);
873         if (IS_ERR(task))
874                 ret = PTR_ERR(task);
875         else {
876                 ret = task->tk_status;
877                 rpc_put_task(task);
878         }
879         return ret;
880 }
881
882 int nfs4_call_sync(struct rpc_clnt *clnt,
883                    struct nfs_server *server,
884                    struct rpc_message *msg,
885                    struct nfs4_sequence_args *args,
886                    struct nfs4_sequence_res *res,
887                    int cache_reply)
888 {
889         nfs4_init_sequence(args, res, cache_reply);
890         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
891 }
892
893 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
894 {
895         struct nfs_inode *nfsi = NFS_I(dir);
896
897         spin_lock(&dir->i_lock);
898         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
899         if (!cinfo->atomic || cinfo->before != dir->i_version)
900                 nfs_force_lookup_revalidate(dir);
901         dir->i_version = cinfo->after;
902         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
903         nfs_fscache_invalidate(dir);
904         spin_unlock(&dir->i_lock);
905 }
906
907 struct nfs4_opendata {
908         struct kref kref;
909         struct nfs_openargs o_arg;
910         struct nfs_openres o_res;
911         struct nfs_open_confirmargs c_arg;
912         struct nfs_open_confirmres c_res;
913         struct nfs4_string owner_name;
914         struct nfs4_string group_name;
915         struct nfs4_label *a_label;
916         struct nfs_fattr f_attr;
917         struct nfs4_label *f_label;
918         struct dentry *dir;
919         struct dentry *dentry;
920         struct nfs4_state_owner *owner;
921         struct nfs4_state *state;
922         struct iattr attrs;
923         unsigned long timestamp;
924         unsigned int rpc_done : 1;
925         unsigned int file_created : 1;
926         unsigned int is_recover : 1;
927         int rpc_status;
928         int cancelled;
929 };
930
931 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
932                 int err, struct nfs4_exception *exception)
933 {
934         if (err != -EINVAL)
935                 return false;
936         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
937                 return false;
938         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
939         exception->retry = 1;
940         return true;
941 }
942
943 static u32
944 nfs4_map_atomic_open_share(struct nfs_server *server,
945                 fmode_t fmode, int openflags)
946 {
947         u32 res = 0;
948
949         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
950         case FMODE_READ:
951                 res = NFS4_SHARE_ACCESS_READ;
952                 break;
953         case FMODE_WRITE:
954                 res = NFS4_SHARE_ACCESS_WRITE;
955                 break;
956         case FMODE_READ|FMODE_WRITE:
957                 res = NFS4_SHARE_ACCESS_BOTH;
958         }
959         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
960                 goto out;
961         /* Want no delegation if we're using O_DIRECT */
962         if (openflags & O_DIRECT)
963                 res |= NFS4_SHARE_WANT_NO_DELEG;
964 out:
965         return res;
966 }
967
968 static enum open_claim_type4
969 nfs4_map_atomic_open_claim(struct nfs_server *server,
970                 enum open_claim_type4 claim)
971 {
972         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
973                 return claim;
974         switch (claim) {
975         default:
976                 return claim;
977         case NFS4_OPEN_CLAIM_FH:
978                 return NFS4_OPEN_CLAIM_NULL;
979         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
980                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
981         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
982                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
983         }
984 }
985
986 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
987 {
988         p->o_res.f_attr = &p->f_attr;
989         p->o_res.f_label = p->f_label;
990         p->o_res.seqid = p->o_arg.seqid;
991         p->c_res.seqid = p->c_arg.seqid;
992         p->o_res.server = p->o_arg.server;
993         p->o_res.access_request = p->o_arg.access;
994         nfs_fattr_init(&p->f_attr);
995         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
996 }
997
998 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
999                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1000                 const struct iattr *attrs,
1001                 struct nfs4_label *label,
1002                 enum open_claim_type4 claim,
1003                 gfp_t gfp_mask)
1004 {
1005         struct dentry *parent = dget_parent(dentry);
1006         struct inode *dir = d_inode(parent);
1007         struct nfs_server *server = NFS_SERVER(dir);
1008         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1009         struct nfs4_opendata *p;
1010
1011         p = kzalloc(sizeof(*p), gfp_mask);
1012         if (p == NULL)
1013                 goto err;
1014
1015         p->f_label = nfs4_label_alloc(server, gfp_mask);
1016         if (IS_ERR(p->f_label))
1017                 goto err_free_p;
1018
1019         p->a_label = nfs4_label_alloc(server, gfp_mask);
1020         if (IS_ERR(p->a_label))
1021                 goto err_free_f;
1022
1023         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1024         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1025         if (IS_ERR(p->o_arg.seqid))
1026                 goto err_free_label;
1027         nfs_sb_active(dentry->d_sb);
1028         p->dentry = dget(dentry);
1029         p->dir = parent;
1030         p->owner = sp;
1031         atomic_inc(&sp->so_count);
1032         p->o_arg.open_flags = flags;
1033         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1034         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1035                         fmode, flags);
1036         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1037          * will return permission denied for all bits until close */
1038         if (!(flags & O_EXCL)) {
1039                 /* ask server to check for all possible rights as results
1040                  * are cached */
1041                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1042                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1043         }
1044         p->o_arg.clientid = server->nfs_client->cl_clientid;
1045         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1046         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1047         p->o_arg.name = &dentry->d_name;
1048         p->o_arg.server = server;
1049         p->o_arg.bitmask = nfs4_bitmask(server, label);
1050         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1051         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1052         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1053         switch (p->o_arg.claim) {
1054         case NFS4_OPEN_CLAIM_NULL:
1055         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1056         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1057                 p->o_arg.fh = NFS_FH(dir);
1058                 break;
1059         case NFS4_OPEN_CLAIM_PREVIOUS:
1060         case NFS4_OPEN_CLAIM_FH:
1061         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1062         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1063                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1064         }
1065         if (attrs != NULL && attrs->ia_valid != 0) {
1066                 __u32 verf[2];
1067
1068                 p->o_arg.u.attrs = &p->attrs;
1069                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1070
1071                 verf[0] = jiffies;
1072                 verf[1] = current->pid;
1073                 memcpy(p->o_arg.u.verifier.data, verf,
1074                                 sizeof(p->o_arg.u.verifier.data));
1075         }
1076         p->c_arg.fh = &p->o_res.fh;
1077         p->c_arg.stateid = &p->o_res.stateid;
1078         p->c_arg.seqid = p->o_arg.seqid;
1079         nfs4_init_opendata_res(p);
1080         kref_init(&p->kref);
1081         return p;
1082
1083 err_free_label:
1084         nfs4_label_free(p->a_label);
1085 err_free_f:
1086         nfs4_label_free(p->f_label);
1087 err_free_p:
1088         kfree(p);
1089 err:
1090         dput(parent);
1091         return NULL;
1092 }
1093
1094 static void nfs4_opendata_free(struct kref *kref)
1095 {
1096         struct nfs4_opendata *p = container_of(kref,
1097                         struct nfs4_opendata, kref);
1098         struct super_block *sb = p->dentry->d_sb;
1099
1100         nfs_free_seqid(p->o_arg.seqid);
1101         if (p->state != NULL)
1102                 nfs4_put_open_state(p->state);
1103         nfs4_put_state_owner(p->owner);
1104
1105         nfs4_label_free(p->a_label);
1106         nfs4_label_free(p->f_label);
1107
1108         dput(p->dir);
1109         dput(p->dentry);
1110         nfs_sb_deactive(sb);
1111         nfs_fattr_free_names(&p->f_attr);
1112         kfree(p->f_attr.mdsthreshold);
1113         kfree(p);
1114 }
1115
1116 static void nfs4_opendata_put(struct nfs4_opendata *p)
1117 {
1118         if (p != NULL)
1119                 kref_put(&p->kref, nfs4_opendata_free);
1120 }
1121
1122 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1123 {
1124         int ret;
1125
1126         ret = rpc_wait_for_completion_task(task);
1127         return ret;
1128 }
1129
1130 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1131 {
1132         int ret = 0;
1133
1134         if (open_mode & (O_EXCL|O_TRUNC))
1135                 goto out;
1136         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1137                 case FMODE_READ:
1138                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1139                                 && state->n_rdonly != 0;
1140                         break;
1141                 case FMODE_WRITE:
1142                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1143                                 && state->n_wronly != 0;
1144                         break;
1145                 case FMODE_READ|FMODE_WRITE:
1146                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1147                                 && state->n_rdwr != 0;
1148         }
1149 out:
1150         return ret;
1151 }
1152
1153 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1154                 enum open_claim_type4 claim)
1155 {
1156         if (delegation == NULL)
1157                 return 0;
1158         if ((delegation->type & fmode) != fmode)
1159                 return 0;
1160         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1161                 return 0;
1162         switch (claim) {
1163         case NFS4_OPEN_CLAIM_NULL:
1164         case NFS4_OPEN_CLAIM_FH:
1165                 break;
1166         case NFS4_OPEN_CLAIM_PREVIOUS:
1167                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1168                         break;
1169         default:
1170                 return 0;
1171         }
1172         nfs_mark_delegation_referenced(delegation);
1173         return 1;
1174 }
1175
1176 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1177 {
1178         switch (fmode) {
1179                 case FMODE_WRITE:
1180                         state->n_wronly++;
1181                         break;
1182                 case FMODE_READ:
1183                         state->n_rdonly++;
1184                         break;
1185                 case FMODE_READ|FMODE_WRITE:
1186                         state->n_rdwr++;
1187         }
1188         nfs4_state_set_mode_locked(state, state->state | fmode);
1189 }
1190
1191 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1192 {
1193         struct nfs_client *clp = state->owner->so_server->nfs_client;
1194         bool need_recover = false;
1195
1196         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1197                 need_recover = true;
1198         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1199                 need_recover = true;
1200         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1201                 need_recover = true;
1202         if (need_recover)
1203                 nfs4_state_mark_reclaim_nograce(clp, state);
1204 }
1205
1206 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1207                 nfs4_stateid *stateid)
1208 {
1209         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1210                 return true;
1211         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1212                 nfs_test_and_clear_all_open_stateid(state);
1213                 return true;
1214         }
1215         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1216                 return true;
1217         return false;
1218 }
1219
1220 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1221 {
1222         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1223                 return;
1224         if (state->n_wronly)
1225                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1226         if (state->n_rdonly)
1227                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1228         if (state->n_rdwr)
1229                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1230         set_bit(NFS_OPEN_STATE, &state->flags);
1231 }
1232
1233 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1234                 nfs4_stateid *arg_stateid,
1235                 nfs4_stateid *stateid, fmode_t fmode)
1236 {
1237         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1238         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1239         case FMODE_WRITE:
1240                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1241                 break;
1242         case FMODE_READ:
1243                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1244                 break;
1245         case 0:
1246                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1247                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1248                 clear_bit(NFS_OPEN_STATE, &state->flags);
1249         }
1250         if (stateid == NULL)
1251                 return;
1252         /* Handle races with OPEN */
1253         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1254             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1255             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1256                 nfs_resync_open_stateid_locked(state);
1257                 return;
1258         }
1259         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1260                 nfs4_stateid_copy(&state->stateid, stateid);
1261         nfs4_stateid_copy(&state->open_stateid, stateid);
1262 }
1263
1264 static void nfs_clear_open_stateid(struct nfs4_state *state,
1265         nfs4_stateid *arg_stateid,
1266         nfs4_stateid *stateid, fmode_t fmode)
1267 {
1268         write_seqlock(&state->seqlock);
1269         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1270         write_sequnlock(&state->seqlock);
1271         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1272                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1273 }
1274
1275 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1276 {
1277         switch (fmode) {
1278                 case FMODE_READ:
1279                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1280                         break;
1281                 case FMODE_WRITE:
1282                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1283                         break;
1284                 case FMODE_READ|FMODE_WRITE:
1285                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1286         }
1287         if (!nfs_need_update_open_stateid(state, stateid))
1288                 return;
1289         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1290                 nfs4_stateid_copy(&state->stateid, stateid);
1291         nfs4_stateid_copy(&state->open_stateid, stateid);
1292 }
1293
1294 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1295 {
1296         /*
1297          * Protect the call to nfs4_state_set_mode_locked and
1298          * serialise the stateid update
1299          */
1300         write_seqlock(&state->seqlock);
1301         if (deleg_stateid != NULL) {
1302                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1303                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1304         }
1305         if (open_stateid != NULL)
1306                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1307         write_sequnlock(&state->seqlock);
1308         spin_lock(&state->owner->so_lock);
1309         update_open_stateflags(state, fmode);
1310         spin_unlock(&state->owner->so_lock);
1311 }
1312
1313 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1314 {
1315         struct nfs_inode *nfsi = NFS_I(state->inode);
1316         struct nfs_delegation *deleg_cur;
1317         int ret = 0;
1318
1319         fmode &= (FMODE_READ|FMODE_WRITE);
1320
1321         rcu_read_lock();
1322         deleg_cur = rcu_dereference(nfsi->delegation);
1323         if (deleg_cur == NULL)
1324                 goto no_delegation;
1325
1326         spin_lock(&deleg_cur->lock);
1327         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1328            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1329             (deleg_cur->type & fmode) != fmode)
1330                 goto no_delegation_unlock;
1331
1332         if (delegation == NULL)
1333                 delegation = &deleg_cur->stateid;
1334         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1335                 goto no_delegation_unlock;
1336
1337         nfs_mark_delegation_referenced(deleg_cur);
1338         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1339         ret = 1;
1340 no_delegation_unlock:
1341         spin_unlock(&deleg_cur->lock);
1342 no_delegation:
1343         rcu_read_unlock();
1344
1345         if (!ret && open_stateid != NULL) {
1346                 __update_open_stateid(state, open_stateid, NULL, fmode);
1347                 ret = 1;
1348         }
1349         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1350                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1351
1352         return ret;
1353 }
1354
1355 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1356                 const nfs4_stateid *stateid)
1357 {
1358         struct nfs4_state *state = lsp->ls_state;
1359         bool ret = false;
1360
1361         spin_lock(&state->state_lock);
1362         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1363                 goto out_noupdate;
1364         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1365                 goto out_noupdate;
1366         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1367         ret = true;
1368 out_noupdate:
1369         spin_unlock(&state->state_lock);
1370         return ret;
1371 }
1372
1373 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1374 {
1375         struct nfs_delegation *delegation;
1376
1377         rcu_read_lock();
1378         delegation = rcu_dereference(NFS_I(inode)->delegation);
1379         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1380                 rcu_read_unlock();
1381                 return;
1382         }
1383         rcu_read_unlock();
1384         nfs4_inode_return_delegation(inode);
1385 }
1386
1387 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1388 {
1389         struct nfs4_state *state = opendata->state;
1390         struct nfs_inode *nfsi = NFS_I(state->inode);
1391         struct nfs_delegation *delegation;
1392         int open_mode = opendata->o_arg.open_flags;
1393         fmode_t fmode = opendata->o_arg.fmode;
1394         enum open_claim_type4 claim = opendata->o_arg.claim;
1395         nfs4_stateid stateid;
1396         int ret = -EAGAIN;
1397
1398         for (;;) {
1399                 spin_lock(&state->owner->so_lock);
1400                 if (can_open_cached(state, fmode, open_mode)) {
1401                         update_open_stateflags(state, fmode);
1402                         spin_unlock(&state->owner->so_lock);
1403                         goto out_return_state;
1404                 }
1405                 spin_unlock(&state->owner->so_lock);
1406                 rcu_read_lock();
1407                 delegation = rcu_dereference(nfsi->delegation);
1408                 if (!can_open_delegated(delegation, fmode, claim)) {
1409                         rcu_read_unlock();
1410                         break;
1411                 }
1412                 /* Save the delegation */
1413                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1414                 rcu_read_unlock();
1415                 nfs_release_seqid(opendata->o_arg.seqid);
1416                 if (!opendata->is_recover) {
1417                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1418                         if (ret != 0)
1419                                 goto out;
1420                 }
1421                 ret = -EAGAIN;
1422
1423                 /* Try to update the stateid using the delegation */
1424                 if (update_open_stateid(state, NULL, &stateid, fmode))
1425                         goto out_return_state;
1426         }
1427 out:
1428         return ERR_PTR(ret);
1429 out_return_state:
1430         atomic_inc(&state->count);
1431         return state;
1432 }
1433
1434 static void
1435 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1436 {
1437         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1438         struct nfs_delegation *delegation;
1439         int delegation_flags = 0;
1440
1441         rcu_read_lock();
1442         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1443         if (delegation)
1444                 delegation_flags = delegation->flags;
1445         rcu_read_unlock();
1446         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1447                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1448                                    "returning a delegation for "
1449                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1450                                    clp->cl_hostname);
1451         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1452                 nfs_inode_set_delegation(state->inode,
1453                                          data->owner->so_cred,
1454                                          &data->o_res);
1455         else
1456                 nfs_inode_reclaim_delegation(state->inode,
1457                                              data->owner->so_cred,
1458                                              &data->o_res);
1459 }
1460
1461 /*
1462  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1463  * and update the nfs4_state.
1464  */
1465 static struct nfs4_state *
1466 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1467 {
1468         struct inode *inode = data->state->inode;
1469         struct nfs4_state *state = data->state;
1470         int ret;
1471
1472         if (!data->rpc_done) {
1473                 if (data->rpc_status) {
1474                         ret = data->rpc_status;
1475                         goto err;
1476                 }
1477                 /* cached opens have already been processed */
1478                 goto update;
1479         }
1480
1481         ret = nfs_refresh_inode(inode, &data->f_attr);
1482         if (ret)
1483                 goto err;
1484
1485         if (data->o_res.delegation_type != 0)
1486                 nfs4_opendata_check_deleg(data, state);
1487 update:
1488         update_open_stateid(state, &data->o_res.stateid, NULL,
1489                             data->o_arg.fmode);
1490         atomic_inc(&state->count);
1491
1492         return state;
1493 err:
1494         return ERR_PTR(ret);
1495
1496 }
1497
1498 static struct nfs4_state *
1499 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1500 {
1501         struct inode *inode;
1502         struct nfs4_state *state = NULL;
1503         int ret;
1504
1505         if (!data->rpc_done) {
1506                 state = nfs4_try_open_cached(data);
1507                 goto out;
1508         }
1509
1510         ret = -EAGAIN;
1511         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1512                 goto err;
1513         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1514         ret = PTR_ERR(inode);
1515         if (IS_ERR(inode))
1516                 goto err;
1517         ret = -ENOMEM;
1518         state = nfs4_get_open_state(inode, data->owner);
1519         if (state == NULL)
1520                 goto err_put_inode;
1521         if (data->o_res.delegation_type != 0)
1522                 nfs4_opendata_check_deleg(data, state);
1523         update_open_stateid(state, &data->o_res.stateid, NULL,
1524                         data->o_arg.fmode);
1525         iput(inode);
1526 out:
1527         nfs_release_seqid(data->o_arg.seqid);
1528         return state;
1529 err_put_inode:
1530         iput(inode);
1531 err:
1532         return ERR_PTR(ret);
1533 }
1534
1535 static struct nfs4_state *
1536 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1537 {
1538         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1539                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1540         return _nfs4_opendata_to_nfs4_state(data);
1541 }
1542
1543 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1544 {
1545         struct nfs_inode *nfsi = NFS_I(state->inode);
1546         struct nfs_open_context *ctx;
1547
1548         spin_lock(&state->inode->i_lock);
1549         list_for_each_entry(ctx, &nfsi->open_files, list) {
1550                 if (ctx->state != state)
1551                         continue;
1552                 get_nfs_open_context(ctx);
1553                 spin_unlock(&state->inode->i_lock);
1554                 return ctx;
1555         }
1556         spin_unlock(&state->inode->i_lock);
1557         return ERR_PTR(-ENOENT);
1558 }
1559
1560 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1561                 struct nfs4_state *state, enum open_claim_type4 claim)
1562 {
1563         struct nfs4_opendata *opendata;
1564
1565         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1566                         NULL, NULL, claim, GFP_NOFS);
1567         if (opendata == NULL)
1568                 return ERR_PTR(-ENOMEM);
1569         opendata->state = state;
1570         atomic_inc(&state->count);
1571         return opendata;
1572 }
1573
1574 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1575 {
1576         struct nfs4_state *newstate;
1577         int ret;
1578
1579         if ((opendata->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
1580              opendata->o_arg.claim == NFS4_OPEN_CLAIM_DELEG_CUR_FH) &&
1581             (opendata->o_arg.u.delegation_type & fmode) != fmode)
1582                 /* This mode can't have been delegated, so we must have
1583                  * a valid open_stateid to cover it - not need to reclaim.
1584                  */
1585                 return 0;
1586         opendata->o_arg.open_flags = 0;
1587         opendata->o_arg.fmode = fmode;
1588         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1589                         NFS_SB(opendata->dentry->d_sb),
1590                         fmode, 0);
1591         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1592         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1593         nfs4_init_opendata_res(opendata);
1594         ret = _nfs4_recover_proc_open(opendata);
1595         if (ret != 0)
1596                 return ret; 
1597         newstate = nfs4_opendata_to_nfs4_state(opendata);
1598         if (IS_ERR(newstate))
1599                 return PTR_ERR(newstate);
1600         nfs4_close_state(newstate, fmode);
1601         *res = newstate;
1602         return 0;
1603 }
1604
1605 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1606 {
1607         struct nfs4_state *newstate;
1608         int ret;
1609
1610         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1611         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1612         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1613         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1614         /* memory barrier prior to reading state->n_* */
1615         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1616         clear_bit(NFS_OPEN_STATE, &state->flags);
1617         smp_rmb();
1618         if (state->n_rdwr != 0) {
1619                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1620                 if (ret != 0)
1621                         return ret;
1622                 if (newstate != state)
1623                         return -ESTALE;
1624         }
1625         if (state->n_wronly != 0) {
1626                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1627                 if (ret != 0)
1628                         return ret;
1629                 if (newstate != state)
1630                         return -ESTALE;
1631         }
1632         if (state->n_rdonly != 0) {
1633                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1634                 if (ret != 0)
1635                         return ret;
1636                 if (newstate != state)
1637                         return -ESTALE;
1638         }
1639         /*
1640          * We may have performed cached opens for all three recoveries.
1641          * Check if we need to update the current stateid.
1642          */
1643         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1644             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1645                 write_seqlock(&state->seqlock);
1646                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1647                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1648                 write_sequnlock(&state->seqlock);
1649         }
1650         return 0;
1651 }
1652
1653 /*
1654  * OPEN_RECLAIM:
1655  *      reclaim state on the server after a reboot.
1656  */
1657 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1658 {
1659         struct nfs_delegation *delegation;
1660         struct nfs4_opendata *opendata;
1661         fmode_t delegation_type = 0;
1662         int status;
1663
1664         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1665                         NFS4_OPEN_CLAIM_PREVIOUS);
1666         if (IS_ERR(opendata))
1667                 return PTR_ERR(opendata);
1668         rcu_read_lock();
1669         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1670         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1671                 delegation_type = delegation->type;
1672         rcu_read_unlock();
1673         opendata->o_arg.u.delegation_type = delegation_type;
1674         status = nfs4_open_recover(opendata, state);
1675         nfs4_opendata_put(opendata);
1676         return status;
1677 }
1678
1679 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1680 {
1681         struct nfs_server *server = NFS_SERVER(state->inode);
1682         struct nfs4_exception exception = { };
1683         int err;
1684         do {
1685                 err = _nfs4_do_open_reclaim(ctx, state);
1686                 trace_nfs4_open_reclaim(ctx, 0, err);
1687                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1688                         continue;
1689                 if (err != -NFS4ERR_DELAY)
1690                         break;
1691                 nfs4_handle_exception(server, err, &exception);
1692         } while (exception.retry);
1693         return err;
1694 }
1695
1696 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1697 {
1698         struct nfs_open_context *ctx;
1699         int ret;
1700
1701         ctx = nfs4_state_find_open_context(state);
1702         if (IS_ERR(ctx))
1703                 return -EAGAIN;
1704         ret = nfs4_do_open_reclaim(ctx, state);
1705         put_nfs_open_context(ctx);
1706         return ret;
1707 }
1708
1709 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1710 {
1711         switch (err) {
1712                 default:
1713                         printk(KERN_ERR "NFS: %s: unhandled error "
1714                                         "%d.\n", __func__, err);
1715                 case 0:
1716                 case -ENOENT:
1717                 case -EAGAIN:
1718                 case -ESTALE:
1719                         break;
1720                 case -NFS4ERR_BADSESSION:
1721                 case -NFS4ERR_BADSLOT:
1722                 case -NFS4ERR_BAD_HIGH_SLOT:
1723                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1724                 case -NFS4ERR_DEADSESSION:
1725                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1726                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1727                         return -EAGAIN;
1728                 case -NFS4ERR_STALE_CLIENTID:
1729                 case -NFS4ERR_STALE_STATEID:
1730                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1731                 case -NFS4ERR_EXPIRED:
1732                         /* Don't recall a delegation if it was lost */
1733                         nfs4_schedule_lease_recovery(server->nfs_client);
1734                         return -EAGAIN;
1735                 case -NFS4ERR_MOVED:
1736                         nfs4_schedule_migration_recovery(server);
1737                         return -EAGAIN;
1738                 case -NFS4ERR_LEASE_MOVED:
1739                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1740                         return -EAGAIN;
1741                 case -NFS4ERR_DELEG_REVOKED:
1742                 case -NFS4ERR_ADMIN_REVOKED:
1743                 case -NFS4ERR_BAD_STATEID:
1744                 case -NFS4ERR_OPENMODE:
1745                         nfs_inode_find_state_and_recover(state->inode,
1746                                         stateid);
1747                         nfs4_schedule_stateid_recovery(server, state);
1748                         return -EAGAIN;
1749                 case -NFS4ERR_DELAY:
1750                 case -NFS4ERR_GRACE:
1751                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1752                         ssleep(1);
1753                         return -EAGAIN;
1754                 case -ENOMEM:
1755                 case -NFS4ERR_DENIED:
1756                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1757                         return 0;
1758         }
1759         return err;
1760 }
1761
1762 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1763 {
1764         struct nfs_server *server = NFS_SERVER(state->inode);
1765         struct nfs4_opendata *opendata;
1766         int err;
1767
1768         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1769                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1770         if (IS_ERR(opendata))
1771                 return PTR_ERR(opendata);
1772         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1773         err = nfs4_open_recover(opendata, state);
1774         nfs4_opendata_put(opendata);
1775         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1776 }
1777
1778 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1779 {
1780         struct nfs4_opendata *data = calldata;
1781
1782         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1783                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1784 }
1785
1786 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1787 {
1788         struct nfs4_opendata *data = calldata;
1789
1790         nfs40_sequence_done(task, &data->c_res.seq_res);
1791
1792         data->rpc_status = task->tk_status;
1793         if (data->rpc_status == 0) {
1794                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1795                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1796                 renew_lease(data->o_res.server, data->timestamp);
1797                 data->rpc_done = 1;
1798         }
1799 }
1800
1801 static void nfs4_open_confirm_release(void *calldata)
1802 {
1803         struct nfs4_opendata *data = calldata;
1804         struct nfs4_state *state = NULL;
1805
1806         /* If this request hasn't been cancelled, do nothing */
1807         if (data->cancelled == 0)
1808                 goto out_free;
1809         /* In case of error, no cleanup! */
1810         if (!data->rpc_done)
1811                 goto out_free;
1812         state = nfs4_opendata_to_nfs4_state(data);
1813         if (!IS_ERR(state))
1814                 nfs4_close_state(state, data->o_arg.fmode);
1815 out_free:
1816         nfs4_opendata_put(data);
1817 }
1818
1819 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1820         .rpc_call_prepare = nfs4_open_confirm_prepare,
1821         .rpc_call_done = nfs4_open_confirm_done,
1822         .rpc_release = nfs4_open_confirm_release,
1823 };
1824
1825 /*
1826  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1827  */
1828 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1829 {
1830         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1831         struct rpc_task *task;
1832         struct  rpc_message msg = {
1833                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1834                 .rpc_argp = &data->c_arg,
1835                 .rpc_resp = &data->c_res,
1836                 .rpc_cred = data->owner->so_cred,
1837         };
1838         struct rpc_task_setup task_setup_data = {
1839                 .rpc_client = server->client,
1840                 .rpc_message = &msg,
1841                 .callback_ops = &nfs4_open_confirm_ops,
1842                 .callback_data = data,
1843                 .workqueue = nfsiod_workqueue,
1844                 .flags = RPC_TASK_ASYNC,
1845         };
1846         int status;
1847
1848         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1849         kref_get(&data->kref);
1850         data->rpc_done = 0;
1851         data->rpc_status = 0;
1852         data->timestamp = jiffies;
1853         task = rpc_run_task(&task_setup_data);
1854         if (IS_ERR(task))
1855                 return PTR_ERR(task);
1856         status = nfs4_wait_for_completion_rpc_task(task);
1857         if (status != 0) {
1858                 data->cancelled = 1;
1859                 smp_wmb();
1860         } else
1861                 status = data->rpc_status;
1862         rpc_put_task(task);
1863         return status;
1864 }
1865
1866 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1867 {
1868         struct nfs4_opendata *data = calldata;
1869         struct nfs4_state_owner *sp = data->owner;
1870         struct nfs_client *clp = sp->so_server->nfs_client;
1871         enum open_claim_type4 claim = data->o_arg.claim;
1872
1873         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1874                 goto out_wait;
1875         /*
1876          * Check if we still need to send an OPEN call, or if we can use
1877          * a delegation instead.
1878          */
1879         if (data->state != NULL) {
1880                 struct nfs_delegation *delegation;
1881
1882                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1883                         goto out_no_action;
1884                 rcu_read_lock();
1885                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1886                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1887                         goto unlock_no_action;
1888                 rcu_read_unlock();
1889         }
1890         /* Update client id. */
1891         data->o_arg.clientid = clp->cl_clientid;
1892         switch (claim) {
1893         default:
1894                 break;
1895         case NFS4_OPEN_CLAIM_PREVIOUS:
1896         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1897         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1898                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1899         case NFS4_OPEN_CLAIM_FH:
1900                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1901                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1902         }
1903         data->timestamp = jiffies;
1904         if (nfs4_setup_sequence(data->o_arg.server,
1905                                 &data->o_arg.seq_args,
1906                                 &data->o_res.seq_res,
1907                                 task) != 0)
1908                 nfs_release_seqid(data->o_arg.seqid);
1909
1910         /* Set the create mode (note dependency on the session type) */
1911         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1912         if (data->o_arg.open_flags & O_EXCL) {
1913                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1914                 if (nfs4_has_persistent_session(clp))
1915                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1916                 else if (clp->cl_mvops->minor_version > 0)
1917                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1918         }
1919         return;
1920 unlock_no_action:
1921         rcu_read_unlock();
1922 out_no_action:
1923         task->tk_action = NULL;
1924 out_wait:
1925         nfs4_sequence_done(task, &data->o_res.seq_res);
1926 }
1927
1928 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1929 {
1930         struct nfs4_opendata *data = calldata;
1931
1932         data->rpc_status = task->tk_status;
1933
1934         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1935                 return;
1936
1937         if (task->tk_status == 0) {
1938                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1939                         switch (data->o_res.f_attr->mode & S_IFMT) {
1940                         case S_IFREG:
1941                                 break;
1942                         case S_IFLNK:
1943                                 data->rpc_status = -ELOOP;
1944                                 break;
1945                         case S_IFDIR:
1946                                 data->rpc_status = -EISDIR;
1947                                 break;
1948                         default:
1949                                 data->rpc_status = -ENOTDIR;
1950                         }
1951                 }
1952                 renew_lease(data->o_res.server, data->timestamp);
1953                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1954                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1955         }
1956         data->rpc_done = 1;
1957 }
1958
1959 static void nfs4_open_release(void *calldata)
1960 {
1961         struct nfs4_opendata *data = calldata;
1962         struct nfs4_state *state = NULL;
1963
1964         /* If this request hasn't been cancelled, do nothing */
1965         if (data->cancelled == 0)
1966                 goto out_free;
1967         /* In case of error, no cleanup! */
1968         if (data->rpc_status != 0 || !data->rpc_done)
1969                 goto out_free;
1970         /* In case we need an open_confirm, no cleanup! */
1971         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1972                 goto out_free;
1973         state = nfs4_opendata_to_nfs4_state(data);
1974         if (!IS_ERR(state))
1975                 nfs4_close_state(state, data->o_arg.fmode);
1976 out_free:
1977         nfs4_opendata_put(data);
1978 }
1979
1980 static const struct rpc_call_ops nfs4_open_ops = {
1981         .rpc_call_prepare = nfs4_open_prepare,
1982         .rpc_call_done = nfs4_open_done,
1983         .rpc_release = nfs4_open_release,
1984 };
1985
1986 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1987 {
1988         struct inode *dir = d_inode(data->dir);
1989         struct nfs_server *server = NFS_SERVER(dir);
1990         struct nfs_openargs *o_arg = &data->o_arg;
1991         struct nfs_openres *o_res = &data->o_res;
1992         struct rpc_task *task;
1993         struct rpc_message msg = {
1994                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1995                 .rpc_argp = o_arg,
1996                 .rpc_resp = o_res,
1997                 .rpc_cred = data->owner->so_cred,
1998         };
1999         struct rpc_task_setup task_setup_data = {
2000                 .rpc_client = server->client,
2001                 .rpc_message = &msg,
2002                 .callback_ops = &nfs4_open_ops,
2003                 .callback_data = data,
2004                 .workqueue = nfsiod_workqueue,
2005                 .flags = RPC_TASK_ASYNC,
2006         };
2007         int status;
2008
2009         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2010         kref_get(&data->kref);
2011         data->rpc_done = 0;
2012         data->rpc_status = 0;
2013         data->cancelled = 0;
2014         data->is_recover = 0;
2015         if (isrecover) {
2016                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2017                 data->is_recover = 1;
2018         }
2019         task = rpc_run_task(&task_setup_data);
2020         if (IS_ERR(task))
2021                 return PTR_ERR(task);
2022         status = nfs4_wait_for_completion_rpc_task(task);
2023         if (status != 0) {
2024                 data->cancelled = 1;
2025                 smp_wmb();
2026         } else
2027                 status = data->rpc_status;
2028         rpc_put_task(task);
2029
2030         return status;
2031 }
2032
2033 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2034 {
2035         struct inode *dir = d_inode(data->dir);
2036         struct nfs_openres *o_res = &data->o_res;
2037         int status;
2038
2039         status = nfs4_run_open_task(data, 1);
2040         if (status != 0 || !data->rpc_done)
2041                 return status;
2042
2043         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2044
2045         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2046                 status = _nfs4_proc_open_confirm(data);
2047                 if (status != 0)
2048                         return status;
2049         }
2050
2051         return status;
2052 }
2053
2054 /*
2055  * Additional permission checks in order to distinguish between an
2056  * open for read, and an open for execute. This works around the
2057  * fact that NFSv4 OPEN treats read and execute permissions as being
2058  * the same.
2059  * Note that in the non-execute case, we want to turn off permission
2060  * checking if we just created a new file (POSIX open() semantics).
2061  */
2062 static int nfs4_opendata_access(struct rpc_cred *cred,
2063                                 struct nfs4_opendata *opendata,
2064                                 struct nfs4_state *state, fmode_t fmode,
2065                                 int openflags)
2066 {
2067         struct nfs_access_entry cache;
2068         u32 mask;
2069
2070         /* access call failed or for some reason the server doesn't
2071          * support any access modes -- defer access call until later */
2072         if (opendata->o_res.access_supported == 0)
2073                 return 0;
2074
2075         mask = 0;
2076         /*
2077          * Use openflags to check for exec, because fmode won't
2078          * always have FMODE_EXEC set when file open for exec.
2079          */
2080         if (openflags & __FMODE_EXEC) {
2081                 /* ONLY check for exec rights */
2082                 mask = MAY_EXEC;
2083         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2084                 mask = MAY_READ;
2085
2086         cache.cred = cred;
2087         cache.jiffies = jiffies;
2088         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2089         nfs_access_add_cache(state->inode, &cache);
2090
2091         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2092                 return 0;
2093
2094         /* even though OPEN succeeded, access is denied. Close the file */
2095         nfs4_close_state(state, fmode);
2096         return -EACCES;
2097 }
2098
2099 /*
2100  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2101  */
2102 static int _nfs4_proc_open(struct nfs4_opendata *data)
2103 {
2104         struct inode *dir = d_inode(data->dir);
2105         struct nfs_server *server = NFS_SERVER(dir);
2106         struct nfs_openargs *o_arg = &data->o_arg;
2107         struct nfs_openres *o_res = &data->o_res;
2108         int status;
2109
2110         status = nfs4_run_open_task(data, 0);
2111         if (!data->rpc_done)
2112                 return status;
2113         if (status != 0) {
2114                 if (status == -NFS4ERR_BADNAME &&
2115                                 !(o_arg->open_flags & O_CREAT))
2116                         return -ENOENT;
2117                 return status;
2118         }
2119
2120         nfs_fattr_map_and_free_names(server, &data->f_attr);
2121
2122         if (o_arg->open_flags & O_CREAT) {
2123                 update_changeattr(dir, &o_res->cinfo);
2124                 if (o_arg->open_flags & O_EXCL)
2125                         data->file_created = 1;
2126                 else if (o_res->cinfo.before != o_res->cinfo.after)
2127                         data->file_created = 1;
2128         }
2129         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2130                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2131         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2132                 status = _nfs4_proc_open_confirm(data);
2133                 if (status != 0)
2134                         return status;
2135         }
2136         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2137                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2138         return 0;
2139 }
2140
2141 static int nfs4_recover_expired_lease(struct nfs_server *server)
2142 {
2143         return nfs4_client_recover_expired_lease(server->nfs_client);
2144 }
2145
2146 /*
2147  * OPEN_EXPIRED:
2148  *      reclaim state on the server after a network partition.
2149  *      Assumes caller holds the appropriate lock
2150  */
2151 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2152 {
2153         struct nfs4_opendata *opendata;
2154         int ret;
2155
2156         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2157                         NFS4_OPEN_CLAIM_FH);
2158         if (IS_ERR(opendata))
2159                 return PTR_ERR(opendata);
2160         ret = nfs4_open_recover(opendata, state);
2161         if (ret == -ESTALE)
2162                 d_drop(ctx->dentry);
2163         nfs4_opendata_put(opendata);
2164         return ret;
2165 }
2166
2167 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2168 {
2169         struct nfs_server *server = NFS_SERVER(state->inode);
2170         struct nfs4_exception exception = { };
2171         int err;
2172
2173         do {
2174                 err = _nfs4_open_expired(ctx, state);
2175                 trace_nfs4_open_expired(ctx, 0, err);
2176                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2177                         continue;
2178                 switch (err) {
2179                 default:
2180                         goto out;
2181                 case -NFS4ERR_GRACE:
2182                 case -NFS4ERR_DELAY:
2183                         nfs4_handle_exception(server, err, &exception);
2184                         err = 0;
2185                 }
2186         } while (exception.retry);
2187 out:
2188         return err;
2189 }
2190
2191 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2192 {
2193         struct nfs_open_context *ctx;
2194         int ret;
2195
2196         ctx = nfs4_state_find_open_context(state);
2197         if (IS_ERR(ctx))
2198                 return -EAGAIN;
2199         ret = nfs4_do_open_expired(ctx, state);
2200         put_nfs_open_context(ctx);
2201         return ret;
2202 }
2203
2204 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2205 {
2206         nfs_remove_bad_delegation(state->inode);
2207         write_seqlock(&state->seqlock);
2208         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2209         write_sequnlock(&state->seqlock);
2210         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2211 }
2212
2213 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2214 {
2215         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2216                 nfs_finish_clear_delegation_stateid(state);
2217 }
2218
2219 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2220 {
2221         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2222         nfs40_clear_delegation_stateid(state);
2223         return nfs4_open_expired(sp, state);
2224 }
2225
2226 #if defined(CONFIG_NFS_V4_1)
2227 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2228 {
2229         struct nfs_server *server = NFS_SERVER(state->inode);
2230         nfs4_stateid stateid;
2231         struct nfs_delegation *delegation;
2232         struct rpc_cred *cred;
2233         int status;
2234
2235         /* Get the delegation credential for use by test/free_stateid */
2236         rcu_read_lock();
2237         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2238         if (delegation == NULL) {
2239                 rcu_read_unlock();
2240                 return;
2241         }
2242
2243         nfs4_stateid_copy(&stateid, &delegation->stateid);
2244         cred = get_rpccred(delegation->cred);
2245         rcu_read_unlock();
2246         status = nfs41_test_stateid(server, &stateid, cred);
2247         trace_nfs4_test_delegation_stateid(state, NULL, status);
2248
2249         if (status != NFS_OK) {
2250                 /* Free the stateid unless the server explicitly
2251                  * informs us the stateid is unrecognized. */
2252                 if (status != -NFS4ERR_BAD_STATEID)
2253                         nfs41_free_stateid(server, &stateid, cred);
2254                 nfs_finish_clear_delegation_stateid(state);
2255         }
2256
2257         put_rpccred(cred);
2258 }
2259
2260 /**
2261  * nfs41_check_open_stateid - possibly free an open stateid
2262  *
2263  * @state: NFSv4 state for an inode
2264  *
2265  * Returns NFS_OK if recovery for this stateid is now finished.
2266  * Otherwise a negative NFS4ERR value is returned.
2267  */
2268 static int nfs41_check_open_stateid(struct nfs4_state *state)
2269 {
2270         struct nfs_server *server = NFS_SERVER(state->inode);
2271         nfs4_stateid *stateid = &state->open_stateid;
2272         struct rpc_cred *cred = state->owner->so_cred;
2273         int status;
2274
2275         /* If a state reset has been done, test_stateid is unneeded */
2276         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2277             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2278             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2279                 return -NFS4ERR_BAD_STATEID;
2280
2281         status = nfs41_test_stateid(server, stateid, cred);
2282         trace_nfs4_test_open_stateid(state, NULL, status);
2283         if (status != NFS_OK) {
2284                 /* Free the stateid unless the server explicitly
2285                  * informs us the stateid is unrecognized. */
2286                 if (status != -NFS4ERR_BAD_STATEID)
2287                         nfs41_free_stateid(server, stateid, cred);
2288
2289                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2290                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2291                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2292                 clear_bit(NFS_OPEN_STATE, &state->flags);
2293         }
2294         return status;
2295 }
2296
2297 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2298 {
2299         int status;
2300
2301         nfs41_check_delegation_stateid(state);
2302         status = nfs41_check_open_stateid(state);
2303         if (status != NFS_OK)
2304                 status = nfs4_open_expired(sp, state);
2305         return status;
2306 }
2307 #endif
2308
2309 /*
2310  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2311  * fields corresponding to attributes that were used to store the verifier.
2312  * Make sure we clobber those fields in the later setattr call
2313  */
2314 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2315                                 struct iattr *sattr, struct nfs4_label **label)
2316 {
2317         const u32 *attrset = opendata->o_res.attrset;
2318
2319         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2320             !(sattr->ia_valid & ATTR_ATIME_SET))
2321                 sattr->ia_valid |= ATTR_ATIME;
2322
2323         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2324             !(sattr->ia_valid & ATTR_MTIME_SET))
2325                 sattr->ia_valid |= ATTR_MTIME;
2326
2327         /* Except MODE, it seems harmless of setting twice. */
2328         if ((attrset[1] & FATTR4_WORD1_MODE))
2329                 sattr->ia_valid &= ~ATTR_MODE;
2330
2331         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2332                 *label = NULL;
2333 }
2334
2335 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2336                 fmode_t fmode,
2337                 int flags,
2338                 struct nfs_open_context *ctx)
2339 {
2340         struct nfs4_state_owner *sp = opendata->owner;
2341         struct nfs_server *server = sp->so_server;
2342         struct dentry *dentry;
2343         struct nfs4_state *state;
2344         unsigned int seq;
2345         int ret;
2346
2347         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2348
2349         ret = _nfs4_proc_open(opendata);
2350         if (ret != 0)
2351                 goto out;
2352
2353         state = nfs4_opendata_to_nfs4_state(opendata);
2354         ret = PTR_ERR(state);
2355         if (IS_ERR(state))
2356                 goto out;
2357         if (server->caps & NFS_CAP_POSIX_LOCK)
2358                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2359
2360         dentry = opendata->dentry;
2361         if (d_really_is_negative(dentry)) {
2362                 /* FIXME: Is this d_drop() ever needed? */
2363                 d_drop(dentry);
2364                 dentry = d_add_unique(dentry, igrab(state->inode));
2365                 if (dentry == NULL) {
2366                         dentry = opendata->dentry;
2367                 } else if (dentry != ctx->dentry) {
2368                         dput(ctx->dentry);
2369                         ctx->dentry = dget(dentry);
2370                 }
2371                 nfs_set_verifier(dentry,
2372                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2373         }
2374
2375         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2376         if (ret != 0)
2377                 goto out;
2378
2379         ctx->state = state;
2380         if (d_inode(dentry) == state->inode) {
2381                 nfs_inode_attach_open_context(ctx);
2382                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2383                         nfs4_schedule_stateid_recovery(server, state);
2384         }
2385 out:
2386         return ret;
2387 }
2388
2389 /*
2390  * Returns a referenced nfs4_state
2391  */
2392 static int _nfs4_do_open(struct inode *dir,
2393                         struct nfs_open_context *ctx,
2394                         int flags,
2395                         struct iattr *sattr,
2396                         struct nfs4_label *label,
2397                         int *opened)
2398 {
2399         struct nfs4_state_owner  *sp;
2400         struct nfs4_state     *state = NULL;
2401         struct nfs_server       *server = NFS_SERVER(dir);
2402         struct nfs4_opendata *opendata;
2403         struct dentry *dentry = ctx->dentry;
2404         struct rpc_cred *cred = ctx->cred;
2405         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2406         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2407         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2408         struct nfs4_label *olabel = NULL;
2409         int status;
2410
2411         /* Protect against reboot recovery conflicts */
2412         status = -ENOMEM;
2413         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2414         if (sp == NULL) {
2415                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2416                 goto out_err;
2417         }
2418         status = nfs4_recover_expired_lease(server);
2419         if (status != 0)
2420                 goto err_put_state_owner;
2421         if (d_really_is_positive(dentry))
2422                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2423         status = -ENOMEM;
2424         if (d_really_is_positive(dentry))
2425                 claim = NFS4_OPEN_CLAIM_FH;
2426         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2427                         label, claim, GFP_KERNEL);
2428         if (opendata == NULL)
2429                 goto err_put_state_owner;
2430
2431         if (label) {
2432                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2433                 if (IS_ERR(olabel)) {
2434                         status = PTR_ERR(olabel);
2435                         goto err_opendata_put;
2436                 }
2437         }
2438
2439         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2440                 if (!opendata->f_attr.mdsthreshold) {
2441                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2442                         if (!opendata->f_attr.mdsthreshold)
2443                                 goto err_free_label;
2444                 }
2445                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2446         }
2447         if (d_really_is_positive(dentry))
2448                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2449
2450         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2451         if (status != 0)
2452                 goto err_free_label;
2453         state = ctx->state;
2454
2455         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2456             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2457                 nfs4_exclusive_attrset(opendata, sattr, &label);
2458
2459                 nfs_fattr_init(opendata->o_res.f_attr);
2460                 status = nfs4_do_setattr(state->inode, cred,
2461                                 opendata->o_res.f_attr, sattr,
2462                                 state, label, olabel);
2463                 if (status == 0) {
2464                         nfs_setattr_update_inode(state->inode, sattr,
2465                                         opendata->o_res.f_attr);
2466                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2467                 }
2468         }
2469         if (opened && opendata->file_created)
2470                 *opened |= FILE_CREATED;
2471
2472         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2473                 *ctx_th = opendata->f_attr.mdsthreshold;
2474                 opendata->f_attr.mdsthreshold = NULL;
2475         }
2476
2477         nfs4_label_free(olabel);
2478
2479         nfs4_opendata_put(opendata);
2480         nfs4_put_state_owner(sp);
2481         return 0;
2482 err_free_label:
2483         nfs4_label_free(olabel);
2484 err_opendata_put:
2485         nfs4_opendata_put(opendata);
2486 err_put_state_owner:
2487         nfs4_put_state_owner(sp);
2488 out_err:
2489         return status;
2490 }
2491
2492
2493 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2494                                         struct nfs_open_context *ctx,
2495                                         int flags,
2496                                         struct iattr *sattr,
2497                                         struct nfs4_label *label,
2498                                         int *opened)
2499 {
2500         struct nfs_server *server = NFS_SERVER(dir);
2501         struct nfs4_exception exception = { };
2502         struct nfs4_state *res;
2503         int status;
2504
2505         do {
2506                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2507                 res = ctx->state;
2508                 trace_nfs4_open_file(ctx, flags, status);
2509                 if (status == 0)
2510                         break;
2511                 /* NOTE: BAD_SEQID means the server and client disagree about the
2512                  * book-keeping w.r.t. state-changing operations
2513                  * (OPEN/CLOSE/LOCK/LOCKU...)
2514                  * It is actually a sign of a bug on the client or on the server.
2515                  *
2516                  * If we receive a BAD_SEQID error in the particular case of
2517                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2518                  * have unhashed the old state_owner for us, and that we can
2519                  * therefore safely retry using a new one. We should still warn
2520                  * the user though...
2521                  */
2522                 if (status == -NFS4ERR_BAD_SEQID) {
2523                         pr_warn_ratelimited("NFS: v4 server %s "
2524                                         " returned a bad sequence-id error!\n",
2525                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2526                         exception.retry = 1;
2527                         continue;
2528                 }
2529                 /*
2530                  * BAD_STATEID on OPEN means that the server cancelled our
2531                  * state before it received the OPEN_CONFIRM.
2532                  * Recover by retrying the request as per the discussion
2533                  * on Page 181 of RFC3530.
2534                  */
2535                 if (status == -NFS4ERR_BAD_STATEID) {
2536                         exception.retry = 1;
2537                         continue;
2538                 }
2539                 if (status == -EAGAIN) {
2540                         /* We must have found a delegation */
2541                         exception.retry = 1;
2542                         continue;
2543                 }
2544                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2545                         continue;
2546                 res = ERR_PTR(nfs4_handle_exception(server,
2547                                         status, &exception));
2548         } while (exception.retry);
2549         return res;
2550 }
2551
2552 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2553                             struct nfs_fattr *fattr, struct iattr *sattr,
2554                             struct nfs4_state *state, struct nfs4_label *ilabel,
2555                             struct nfs4_label *olabel)
2556 {
2557         struct nfs_server *server = NFS_SERVER(inode);
2558         struct nfs_setattrargs  arg = {
2559                 .fh             = NFS_FH(inode),
2560                 .iap            = sattr,
2561                 .server         = server,
2562                 .bitmask = server->attr_bitmask,
2563                 .label          = ilabel,
2564         };
2565         struct nfs_setattrres  res = {
2566                 .fattr          = fattr,
2567                 .label          = olabel,
2568                 .server         = server,
2569         };
2570         struct rpc_message msg = {
2571                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2572                 .rpc_argp       = &arg,
2573                 .rpc_resp       = &res,
2574                 .rpc_cred       = cred,
2575         };
2576         unsigned long timestamp = jiffies;
2577         fmode_t fmode;
2578         bool truncate;
2579         int status;
2580
2581         arg.bitmask = nfs4_bitmask(server, ilabel);
2582         if (ilabel)
2583                 arg.bitmask = nfs4_bitmask(server, olabel);
2584
2585         nfs_fattr_init(fattr);
2586
2587         /* Servers should only apply open mode checks for file size changes */
2588         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2589         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2590
2591         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2592                 /* Use that stateid */
2593         } else if (truncate && state != NULL) {
2594                 struct nfs_lockowner lockowner = {
2595                         .l_owner = current->files,
2596                         .l_pid = current->tgid,
2597                 };
2598                 if (!nfs4_valid_open_stateid(state))
2599                         return -EBADF;
2600                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2601                                 &lockowner) == -EIO)
2602                         return -EBADF;
2603         } else
2604                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2605
2606         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2607         if (status == 0 && state != NULL)
2608                 renew_lease(server, timestamp);
2609         return status;
2610 }
2611
2612 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2613                            struct nfs_fattr *fattr, struct iattr *sattr,
2614                            struct nfs4_state *state, struct nfs4_label *ilabel,
2615                            struct nfs4_label *olabel)
2616 {
2617         struct nfs_server *server = NFS_SERVER(inode);
2618         struct nfs4_exception exception = {
2619                 .state = state,
2620                 .inode = inode,
2621         };
2622         int err;
2623         do {
2624                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2625                 trace_nfs4_setattr(inode, err);
2626                 switch (err) {
2627                 case -NFS4ERR_OPENMODE:
2628                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2629                                 pr_warn_once("NFSv4: server %s is incorrectly "
2630                                                 "applying open mode checks to "
2631                                                 "a SETATTR that is not "
2632                                                 "changing file size.\n",
2633                                                 server->nfs_client->cl_hostname);
2634                         }
2635                         if (state && !(state->state & FMODE_WRITE)) {
2636                                 err = -EBADF;
2637                                 if (sattr->ia_valid & ATTR_OPEN)
2638                                         err = -EACCES;
2639                                 goto out;
2640                         }
2641                 }
2642                 err = nfs4_handle_exception(server, err, &exception);
2643         } while (exception.retry);
2644 out:
2645         return err;
2646 }
2647
2648 struct nfs4_closedata {
2649         struct inode *inode;
2650         struct nfs4_state *state;
2651         struct nfs_closeargs arg;
2652         struct nfs_closeres res;
2653         struct nfs_fattr fattr;
2654         unsigned long timestamp;
2655         bool roc;
2656         u32 roc_barrier;
2657 };
2658
2659 static void nfs4_free_closedata(void *data)
2660 {
2661         struct nfs4_closedata *calldata = data;
2662         struct nfs4_state_owner *sp = calldata->state->owner;
2663         struct super_block *sb = calldata->state->inode->i_sb;
2664
2665         if (calldata->roc)
2666                 pnfs_roc_release(calldata->state->inode);
2667         nfs4_put_open_state(calldata->state);
2668         nfs_free_seqid(calldata->arg.seqid);
2669         nfs4_put_state_owner(sp);
2670         nfs_sb_deactive(sb);
2671         kfree(calldata);
2672 }
2673
2674 static void nfs4_close_done(struct rpc_task *task, void *data)
2675 {
2676         struct nfs4_closedata *calldata = data;
2677         struct nfs4_state *state = calldata->state;
2678         struct nfs_server *server = NFS_SERVER(calldata->inode);
2679         nfs4_stateid *res_stateid = NULL;
2680
2681         dprintk("%s: begin!\n", __func__);
2682         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2683                 return;
2684         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2685         /* hmm. we are done with the inode, and in the process of freeing
2686          * the state_owner. we keep this around to process errors
2687          */
2688         switch (task->tk_status) {
2689                 case 0:
2690                         res_stateid = &calldata->res.stateid;
2691                         if (calldata->roc)
2692                                 pnfs_roc_set_barrier(state->inode,
2693                                                      calldata->roc_barrier);
2694                         renew_lease(server, calldata->timestamp);
2695                         break;
2696                 case -NFS4ERR_ADMIN_REVOKED:
2697                 case -NFS4ERR_STALE_STATEID:
2698                 case -NFS4ERR_OLD_STATEID:
2699                 case -NFS4ERR_BAD_STATEID:
2700                 case -NFS4ERR_EXPIRED:
2701                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2702                                                 &state->open_stateid)) {
2703                                 rpc_restart_call_prepare(task);
2704                                 goto out_release;
2705                         }
2706                         if (calldata->arg.fmode == 0)
2707                                 break;
2708                 default:
2709                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2710                                 rpc_restart_call_prepare(task);
2711                                 goto out_release;
2712                         }
2713         }
2714         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2715                         res_stateid, calldata->arg.fmode);
2716 out_release:
2717         nfs_release_seqid(calldata->arg.seqid);
2718         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2719         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2720 }
2721
2722 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2723 {
2724         struct nfs4_closedata *calldata = data;
2725         struct nfs4_state *state = calldata->state;
2726         struct inode *inode = calldata->inode;
2727         bool is_rdonly, is_wronly, is_rdwr;
2728         int call_close = 0;
2729
2730         dprintk("%s: begin!\n", __func__);
2731         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2732                 goto out_wait;
2733
2734         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2735         spin_lock(&state->owner->so_lock);
2736         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2737         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2738         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2739         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2740         /* Calculate the change in open mode */
2741         calldata->arg.fmode = 0;
2742         if (state->n_rdwr == 0) {
2743                 if (state->n_rdonly == 0)
2744                         call_close |= is_rdonly;
2745                 else if (is_rdonly)
2746                         calldata->arg.fmode |= FMODE_READ;
2747                 if (state->n_wronly == 0)
2748                         call_close |= is_wronly;
2749                 else if (is_wronly)
2750                         calldata->arg.fmode |= FMODE_WRITE;
2751         } else if (is_rdwr)
2752                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2753
2754         if (calldata->arg.fmode == 0)
2755                 call_close |= is_rdwr;
2756
2757         if (!nfs4_valid_open_stateid(state))
2758                 call_close = 0;
2759         spin_unlock(&state->owner->so_lock);
2760
2761         if (!call_close) {
2762                 /* Note: exit _without_ calling nfs4_close_done */
2763                 goto out_no_action;
2764         }
2765
2766         if (calldata->arg.fmode == 0)
2767                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2768         if (calldata->roc)
2769                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2770
2771         calldata->arg.share_access =
2772                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2773                                 calldata->arg.fmode, 0);
2774
2775         nfs_fattr_init(calldata->res.fattr);
2776         calldata->timestamp = jiffies;
2777         if (nfs4_setup_sequence(NFS_SERVER(inode),
2778                                 &calldata->arg.seq_args,
2779                                 &calldata->res.seq_res,
2780                                 task) != 0)
2781                 nfs_release_seqid(calldata->arg.seqid);
2782         dprintk("%s: done!\n", __func__);
2783         return;
2784 out_no_action:
2785         task->tk_action = NULL;
2786 out_wait:
2787         nfs4_sequence_done(task, &calldata->res.seq_res);
2788 }
2789
2790 static const struct rpc_call_ops nfs4_close_ops = {
2791         .rpc_call_prepare = nfs4_close_prepare,
2792         .rpc_call_done = nfs4_close_done,
2793         .rpc_release = nfs4_free_closedata,
2794 };
2795
2796 static bool nfs4_roc(struct inode *inode)
2797 {
2798         if (!nfs_have_layout(inode))
2799                 return false;
2800         return pnfs_roc(inode);
2801 }
2802
2803 /* 
2804  * It is possible for data to be read/written from a mem-mapped file 
2805  * after the sys_close call (which hits the vfs layer as a flush).
2806  * This means that we can't safely call nfsv4 close on a file until 
2807  * the inode is cleared. This in turn means that we are not good
2808  * NFSv4 citizens - we do not indicate to the server to update the file's 
2809  * share state even when we are done with one of the three share 
2810  * stateid's in the inode.
2811  *
2812  * NOTE: Caller must be holding the sp->so_owner semaphore!
2813  */
2814 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2815 {
2816         struct nfs_server *server = NFS_SERVER(state->inode);
2817         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2818         struct nfs4_closedata *calldata;
2819         struct nfs4_state_owner *sp = state->owner;
2820         struct rpc_task *task;
2821         struct rpc_message msg = {
2822                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2823                 .rpc_cred = state->owner->so_cred,
2824         };
2825         struct rpc_task_setup task_setup_data = {
2826                 .rpc_client = server->client,
2827                 .rpc_message = &msg,
2828                 .callback_ops = &nfs4_close_ops,
2829                 .workqueue = nfsiod_workqueue,
2830                 .flags = RPC_TASK_ASYNC,
2831         };
2832         int status = -ENOMEM;
2833
2834         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2835                 &task_setup_data.rpc_client, &msg);
2836
2837         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2838         if (calldata == NULL)
2839                 goto out;
2840         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2841         calldata->inode = state->inode;
2842         calldata->state = state;
2843         calldata->arg.fh = NFS_FH(state->inode);
2844         /* Serialization for the sequence id */
2845         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2846         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2847         if (IS_ERR(calldata->arg.seqid))
2848                 goto out_free_calldata;
2849         calldata->arg.fmode = 0;
2850         calldata->arg.bitmask = server->cache_consistency_bitmask;
2851         calldata->res.fattr = &calldata->fattr;
2852         calldata->res.seqid = calldata->arg.seqid;
2853         calldata->res.server = server;
2854         calldata->roc = nfs4_roc(state->inode);
2855         nfs_sb_active(calldata->inode->i_sb);
2856
2857         msg.rpc_argp = &calldata->arg;
2858         msg.rpc_resp = &calldata->res;
2859         task_setup_data.callback_data = calldata;
2860         task = rpc_run_task(&task_setup_data);
2861         if (IS_ERR(task))
2862                 return PTR_ERR(task);
2863         status = 0;
2864         if (wait)
2865                 status = rpc_wait_for_completion_task(task);
2866         rpc_put_task(task);
2867         return status;
2868 out_free_calldata:
2869         kfree(calldata);
2870 out:
2871         nfs4_put_open_state(state);
2872         nfs4_put_state_owner(sp);
2873         return status;
2874 }
2875
2876 static struct inode *
2877 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2878                 int open_flags, struct iattr *attr, int *opened)
2879 {
2880         struct nfs4_state *state;
2881         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2882
2883         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2884
2885         /* Protect against concurrent sillydeletes */
2886         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2887
2888         nfs4_label_release_security(label);
2889
2890         if (IS_ERR(state))
2891                 return ERR_CAST(state);
2892         return state->inode;
2893 }
2894
2895 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2896 {
2897         if (ctx->state == NULL)
2898                 return;
2899         if (is_sync)
2900                 nfs4_close_sync(ctx->state, ctx->mode);
2901         else
2902                 nfs4_close_state(ctx->state, ctx->mode);
2903 }
2904
2905 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2906 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2907 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2908
2909 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2910 {
2911         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
2912         struct nfs4_server_caps_arg args = {
2913                 .fhandle = fhandle,
2914                 .bitmask = bitmask,
2915         };
2916         struct nfs4_server_caps_res res = {};
2917         struct rpc_message msg = {
2918                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2919                 .rpc_argp = &args,
2920                 .rpc_resp = &res,
2921         };
2922         int status;
2923
2924         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
2925                      FATTR4_WORD0_FH_EXPIRE_TYPE |
2926                      FATTR4_WORD0_LINK_SUPPORT |
2927                      FATTR4_WORD0_SYMLINK_SUPPORT |
2928                      FATTR4_WORD0_ACLSUPPORT;
2929         if (minorversion)
2930                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
2931
2932         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2933         if (status == 0) {
2934                 /* Sanity check the server answers */
2935                 switch (minorversion) {
2936                 case 0:
2937                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2938                         res.attr_bitmask[2] = 0;
2939                         break;
2940                 case 1:
2941                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2942                         break;
2943                 case 2:
2944                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2945                 }
2946                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2947                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2948                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2949                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2950                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2951                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2952                                 NFS_CAP_SECURITY_LABEL);
2953                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2954                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2955                         server->caps |= NFS_CAP_ACLS;
2956                 if (res.has_links != 0)
2957                         server->caps |= NFS_CAP_HARDLINKS;
2958                 if (res.has_symlinks != 0)
2959                         server->caps |= NFS_CAP_SYMLINKS;
2960                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2961                         server->caps |= NFS_CAP_FILEID;
2962                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2963                         server->caps |= NFS_CAP_MODE;
2964                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2965                         server->caps |= NFS_CAP_NLINK;
2966                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2967                         server->caps |= NFS_CAP_OWNER;
2968                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2969                         server->caps |= NFS_CAP_OWNER_GROUP;
2970                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2971                         server->caps |= NFS_CAP_ATIME;
2972                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2973                         server->caps |= NFS_CAP_CTIME;
2974                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2975                         server->caps |= NFS_CAP_MTIME;
2976 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2977                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2978                         server->caps |= NFS_CAP_SECURITY_LABEL;
2979 #endif
2980                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2981                                 sizeof(server->attr_bitmask));
2982                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2983
2984                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2985                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2986                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2987                 server->cache_consistency_bitmask[2] = 0;
2988                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
2989                         sizeof(server->exclcreat_bitmask));
2990                 server->acl_bitmask = res.acl_bitmask;
2991                 server->fh_expire_type = res.fh_expire_type;
2992         }
2993
2994         return status;
2995 }
2996
2997 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2998 {
2999         struct nfs4_exception exception = { };
3000         int err;
3001         do {
3002                 err = nfs4_handle_exception(server,
3003                                 _nfs4_server_capabilities(server, fhandle),
3004                                 &exception);
3005         } while (exception.retry);
3006         return err;
3007 }
3008
3009 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3010                 struct nfs_fsinfo *info)
3011 {
3012         u32 bitmask[3];
3013         struct nfs4_lookup_root_arg args = {
3014                 .bitmask = bitmask,
3015         };
3016         struct nfs4_lookup_res res = {
3017                 .server = server,
3018                 .fattr = info->fattr,
3019                 .fh = fhandle,
3020         };
3021         struct rpc_message msg = {
3022                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3023                 .rpc_argp = &args,
3024                 .rpc_resp = &res,
3025         };
3026
3027         bitmask[0] = nfs4_fattr_bitmap[0];
3028         bitmask[1] = nfs4_fattr_bitmap[1];
3029         /*
3030          * Process the label in the upcoming getfattr
3031          */
3032         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3033
3034         nfs_fattr_init(info->fattr);
3035         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3036 }
3037
3038 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3039                 struct nfs_fsinfo *info)
3040 {
3041         struct nfs4_exception exception = { };
3042         int err;
3043         do {
3044                 err = _nfs4_lookup_root(server, fhandle, info);
3045                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3046                 switch (err) {
3047                 case 0:
3048                 case -NFS4ERR_WRONGSEC:
3049                         goto out;
3050                 default:
3051                         err = nfs4_handle_exception(server, err, &exception);
3052                 }
3053         } while (exception.retry);
3054 out:
3055         return err;
3056 }
3057
3058 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3059                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3060 {
3061         struct rpc_auth_create_args auth_args = {
3062                 .pseudoflavor = flavor,
3063         };
3064         struct rpc_auth *auth;
3065         int ret;
3066
3067         auth = rpcauth_create(&auth_args, server->client);
3068         if (IS_ERR(auth)) {
3069                 ret = -EACCES;
3070                 goto out;
3071         }
3072         ret = nfs4_lookup_root(server, fhandle, info);
3073 out:
3074         return ret;
3075 }
3076
3077 /*
3078  * Retry pseudoroot lookup with various security flavors.  We do this when:
3079  *
3080  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3081  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3082  *
3083  * Returns zero on success, or a negative NFS4ERR value, or a
3084  * negative errno value.
3085  */
3086 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3087                               struct nfs_fsinfo *info)
3088 {
3089         /* Per 3530bis 15.33.5 */
3090         static const rpc_authflavor_t flav_array[] = {
3091                 RPC_AUTH_GSS_KRB5P,
3092                 RPC_AUTH_GSS_KRB5I,
3093                 RPC_AUTH_GSS_KRB5,
3094                 RPC_AUTH_UNIX,                  /* courtesy */
3095                 RPC_AUTH_NULL,
3096         };
3097         int status = -EPERM;
3098         size_t i;
3099
3100         if (server->auth_info.flavor_len > 0) {
3101                 /* try each flavor specified by user */
3102                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3103                         status = nfs4_lookup_root_sec(server, fhandle, info,
3104                                                 server->auth_info.flavors[i]);
3105                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3106                                 continue;
3107                         break;
3108                 }
3109         } else {
3110                 /* no flavors specified by user, try default list */
3111                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3112                         status = nfs4_lookup_root_sec(server, fhandle, info,
3113                                                       flav_array[i]);
3114                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3115                                 continue;
3116                         break;
3117                 }
3118         }
3119
3120         /*
3121          * -EACCESS could mean that the user doesn't have correct permissions
3122          * to access the mount.  It could also mean that we tried to mount
3123          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3124          * existing mount programs don't handle -EACCES very well so it should
3125          * be mapped to -EPERM instead.
3126          */
3127         if (status == -EACCES)
3128                 status = -EPERM;
3129         return status;
3130 }
3131
3132 static int nfs4_do_find_root_sec(struct nfs_server *server,
3133                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3134 {
3135         int mv = server->nfs_client->cl_minorversion;
3136         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3137 }
3138
3139 /**
3140  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3141  * @server: initialized nfs_server handle
3142  * @fhandle: we fill in the pseudo-fs root file handle
3143  * @info: we fill in an FSINFO struct
3144  * @auth_probe: probe the auth flavours
3145  *
3146  * Returns zero on success, or a negative errno.
3147  */
3148 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3149                          struct nfs_fsinfo *info,
3150                          bool auth_probe)
3151 {
3152         int status = 0;
3153
3154         if (!auth_probe)
3155                 status = nfs4_lookup_root(server, fhandle, info);
3156
3157         if (auth_probe || status == NFS4ERR_WRONGSEC)
3158                 status = nfs4_do_find_root_sec(server, fhandle, info);
3159
3160         if (status == 0)
3161                 status = nfs4_server_capabilities(server, fhandle);
3162         if (status == 0)
3163                 status = nfs4_do_fsinfo(server, fhandle, info);
3164
3165         return nfs4_map_errors(status);
3166 }
3167
3168 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3169                               struct nfs_fsinfo *info)
3170 {
3171         int error;
3172         struct nfs_fattr *fattr = info->fattr;
3173         struct nfs4_label *label = NULL;
3174
3175         error = nfs4_server_capabilities(server, mntfh);
3176         if (error < 0) {
3177                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3178                 return error;
3179         }
3180
3181         label = nfs4_label_alloc(server, GFP_KERNEL);
3182         if (IS_ERR(label))
3183                 return PTR_ERR(label);
3184
3185         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3186         if (error < 0) {
3187                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3188                 goto err_free_label;
3189         }
3190
3191         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3192             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3193                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3194
3195 err_free_label:
3196         nfs4_label_free(label);
3197
3198         return error;
3199 }
3200
3201 /*
3202  * Get locations and (maybe) other attributes of a referral.
3203  * Note that we'll actually follow the referral later when
3204  * we detect fsid mismatch in inode revalidation
3205  */
3206 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3207                              const struct qstr *name, struct nfs_fattr *fattr,
3208                              struct nfs_fh *fhandle)
3209 {
3210         int status = -ENOMEM;
3211         struct page *page = NULL;
3212         struct nfs4_fs_locations *locations = NULL;
3213
3214         page = alloc_page(GFP_KERNEL);
3215         if (page == NULL)
3216                 goto out;
3217         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3218         if (locations == NULL)
3219                 goto out;
3220
3221         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3222         if (status != 0)
3223                 goto out;
3224
3225         /*
3226          * If the fsid didn't change, this is a migration event, not a
3227          * referral.  Cause us to drop into the exception handler, which
3228          * will kick off migration recovery.
3229          */
3230         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3231                 dprintk("%s: server did not return a different fsid for"
3232                         " a referral at %s\n", __func__, name->name);
3233                 status = -NFS4ERR_MOVED;
3234                 goto out;
3235         }
3236         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3237         nfs_fixup_referral_attributes(&locations->fattr);
3238
3239         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3240         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3241         memset(fhandle, 0, sizeof(struct nfs_fh));
3242 out:
3243         if (page)
3244                 __free_page(page);
3245         kfree(locations);
3246         return status;
3247 }
3248
3249 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3250                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3251 {
3252         struct nfs4_getattr_arg args = {
3253                 .fh = fhandle,
3254                 .bitmask = server->attr_bitmask,
3255         };
3256         struct nfs4_getattr_res res = {
3257                 .fattr = fattr,
3258                 .label = label,
3259                 .server = server,
3260         };
3261         struct rpc_message msg = {
3262                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3263                 .rpc_argp = &args,
3264                 .rpc_resp = &res,
3265         };
3266
3267         args.bitmask = nfs4_bitmask(server, label);
3268
3269         nfs_fattr_init(fattr);
3270         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3271 }
3272
3273 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3274                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3275 {
3276         struct nfs4_exception exception = { };
3277         int err;
3278         do {
3279                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3280                 trace_nfs4_getattr(server, fhandle, fattr, err);
3281                 err = nfs4_handle_exception(server, err,
3282                                 &exception);
3283         } while (exception.retry);
3284         return err;
3285 }
3286
3287 /* 
3288  * The file is not closed if it is opened due to the a request to change
3289  * the size of the file. The open call will not be needed once the
3290  * VFS layer lookup-intents are implemented.
3291  *
3292  * Close is called when the inode is destroyed.
3293  * If we haven't opened the file for O_WRONLY, we
3294  * need to in the size_change case to obtain a stateid.
3295  *
3296  * Got race?
3297  * Because OPEN is always done by name in nfsv4, it is
3298  * possible that we opened a different file by the same
3299  * name.  We can recognize this race condition, but we
3300  * can't do anything about it besides returning an error.
3301  *
3302  * This will be fixed with VFS changes (lookup-intent).
3303  */
3304 static int
3305 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3306                   struct iattr *sattr)
3307 {
3308         struct inode *inode = d_inode(dentry);
3309         struct rpc_cred *cred = NULL;
3310         struct nfs4_state *state = NULL;
3311         struct nfs4_label *label = NULL;
3312         int status;
3313
3314         if (pnfs_ld_layoutret_on_setattr(inode) &&
3315             sattr->ia_valid & ATTR_SIZE &&
3316             sattr->ia_size < i_size_read(inode))
3317                 pnfs_commit_and_return_layout(inode);
3318
3319         nfs_fattr_init(fattr);
3320         
3321         /* Deal with open(O_TRUNC) */
3322         if (sattr->ia_valid & ATTR_OPEN)
3323                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3324
3325         /* Optimization: if the end result is no change, don't RPC */
3326         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3327                 return 0;
3328
3329         /* Search for an existing open(O_WRITE) file */
3330         if (sattr->ia_valid & ATTR_FILE) {
3331                 struct nfs_open_context *ctx;
3332
3333                 ctx = nfs_file_open_context(sattr->ia_file);
3334                 if (ctx) {
3335                         cred = ctx->cred;
3336                         state = ctx->state;
3337                 }
3338         }
3339
3340         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3341         if (IS_ERR(label))
3342                 return PTR_ERR(label);
3343
3344         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3345         if (status == 0) {
3346                 nfs_setattr_update_inode(inode, sattr, fattr);
3347                 nfs_setsecurity(inode, fattr, label);
3348         }
3349         nfs4_label_free(label);
3350         return status;
3351 }
3352
3353 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3354                 const struct qstr *name, struct nfs_fh *fhandle,
3355                 struct nfs_fattr *fattr, struct nfs4_label *label)
3356 {
3357         struct nfs_server *server = NFS_SERVER(dir);
3358         int                    status;
3359         struct nfs4_lookup_arg args = {
3360                 .bitmask = server->attr_bitmask,
3361                 .dir_fh = NFS_FH(dir),
3362                 .name = name,
3363         };
3364         struct nfs4_lookup_res res = {
3365                 .server = server,
3366                 .fattr = fattr,
3367                 .label = label,
3368                 .fh = fhandle,
3369         };
3370         struct rpc_message msg = {
3371                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3372                 .rpc_argp = &args,
3373                 .rpc_resp = &res,
3374         };
3375
3376         args.bitmask = nfs4_bitmask(server, label);
3377
3378         nfs_fattr_init(fattr);
3379
3380         dprintk("NFS call  lookup %s\n", name->name);
3381         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3382         dprintk("NFS reply lookup: %d\n", status);
3383         return status;
3384 }
3385
3386 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3387 {
3388         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3389                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3390         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3391         fattr->nlink = 2;
3392 }
3393
3394 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3395                                    struct qstr *name, struct nfs_fh *fhandle,
3396                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3397 {
3398         struct nfs4_exception exception = { };
3399         struct rpc_clnt *client = *clnt;
3400         int err;
3401         do {
3402                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3403                 trace_nfs4_lookup(dir, name, err);
3404                 switch (err) {
3405                 case -NFS4ERR_BADNAME:
3406                         err = -ENOENT;
3407                         goto out;
3408                 case -NFS4ERR_MOVED:
3409                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3410                         if (err == -NFS4ERR_MOVED)
3411                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3412                         goto out;
3413                 case -NFS4ERR_WRONGSEC:
3414                         err = -EPERM;
3415                         if (client != *clnt)
3416                                 goto out;
3417                         client = nfs4_negotiate_security(client, dir, name);
3418                         if (IS_ERR(client))
3419                                 return PTR_ERR(client);
3420
3421                         exception.retry = 1;
3422                         break;
3423                 default:
3424                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3425                 }
3426         } while (exception.retry);
3427
3428 out:
3429         if (err == 0)
3430                 *clnt = client;
3431         else if (client != *clnt)
3432                 rpc_shutdown_client(client);
3433
3434         return err;
3435 }
3436
3437 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3438                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3439                             struct nfs4_label *label)
3440 {
3441         int status;
3442         struct rpc_clnt *client = NFS_CLIENT(dir);
3443
3444         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3445         if (client != NFS_CLIENT(dir)) {
3446                 rpc_shutdown_client(client);
3447                 nfs_fixup_secinfo_attributes(fattr);
3448         }
3449         return status;
3450 }
3451
3452 struct rpc_clnt *
3453 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3454                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3455 {
3456         struct rpc_clnt *client = NFS_CLIENT(dir);
3457         int status;
3458
3459         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3460         if (status < 0)
3461                 return ERR_PTR(status);
3462         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3463 }
3464
3465 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3466 {
3467         struct nfs_server *server = NFS_SERVER(inode);
3468         struct nfs4_accessargs args = {
3469                 .fh = NFS_FH(inode),
3470                 .bitmask = server->cache_consistency_bitmask,
3471         };
3472         struct nfs4_accessres res = {
3473                 .server = server,
3474         };
3475         struct rpc_message msg = {
3476                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3477                 .rpc_argp = &args,
3478                 .rpc_resp = &res,
3479                 .rpc_cred = entry->cred,
3480         };
3481         int mode = entry->mask;
3482         int status = 0;
3483
3484         /*
3485          * Determine which access bits we want to ask for...
3486          */
3487         if (mode & MAY_READ)
3488                 args.access |= NFS4_ACCESS_READ;
3489         if (S_ISDIR(inode->i_mode)) {
3490                 if (mode & MAY_WRITE)
3491                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3492                 if (mode & MAY_EXEC)
3493                         args.access |= NFS4_ACCESS_LOOKUP;
3494         } else {
3495                 if (mode & MAY_WRITE)
3496                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3497                 if (mode & MAY_EXEC)
3498                         args.access |= NFS4_ACCESS_EXECUTE;
3499         }
3500
3501         res.fattr = nfs_alloc_fattr();
3502         if (res.fattr == NULL)
3503                 return -ENOMEM;
3504
3505         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3506         if (!status) {
3507                 nfs_access_set_mask(entry, res.access);
3508                 nfs_refresh_inode(inode, res.fattr);
3509         }
3510         nfs_free_fattr(res.fattr);
3511         return status;
3512 }
3513
3514 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3515 {
3516         struct nfs4_exception exception = { };
3517         int err;
3518         do {
3519                 err = _nfs4_proc_access(inode, entry);
3520                 trace_nfs4_access(inode, err);
3521                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3522                                 &exception);
3523         } while (exception.retry);
3524         return err;
3525 }
3526
3527 /*
3528  * TODO: For the time being, we don't try to get any attributes
3529  * along with any of the zero-copy operations READ, READDIR,
3530  * READLINK, WRITE.
3531  *
3532  * In the case of the first three, we want to put the GETATTR
3533  * after the read-type operation -- this is because it is hard
3534  * to predict the length of a GETATTR response in v4, and thus
3535  * align the READ data correctly.  This means that the GETATTR
3536  * may end up partially falling into the page cache, and we should
3537  * shift it into the 'tail' of the xdr_buf before processing.
3538  * To do this efficiently, we need to know the total length
3539  * of data received, which doesn't seem to be available outside
3540  * of the RPC layer.
3541  *
3542  * In the case of WRITE, we also want to put the GETATTR after
3543  * the operation -- in this case because we want to make sure
3544  * we get the post-operation mtime and size.
3545  *
3546  * Both of these changes to the XDR layer would in fact be quite
3547  * minor, but I decided to leave them for a subsequent patch.
3548  */
3549 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3550                 unsigned int pgbase, unsigned int pglen)
3551 {
3552         struct nfs4_readlink args = {
3553                 .fh       = NFS_FH(inode),
3554                 .pgbase   = pgbase,
3555                 .pglen    = pglen,
3556                 .pages    = &page,
3557         };
3558         struct nfs4_readlink_res res;
3559         struct rpc_message msg = {
3560                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3561                 .rpc_argp = &args,
3562                 .rpc_resp = &res,
3563         };
3564
3565         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3566 }
3567
3568 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3569                 unsigned int pgbase, unsigned int pglen)
3570 {
3571         struct nfs4_exception exception = { };
3572         int err;
3573         do {
3574                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3575                 trace_nfs4_readlink(inode, err);
3576                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3577                                 &exception);
3578         } while (exception.retry);
3579         return err;
3580 }
3581
3582 /*
3583  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3584  */
3585 static int
3586 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3587                  int flags)
3588 {
3589         struct nfs4_label l, *ilabel = NULL;
3590         struct nfs_open_context *ctx;
3591         struct nfs4_state *state;
3592         int status = 0;
3593
3594         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3595         if (IS_ERR(ctx))
3596                 return PTR_ERR(ctx);
3597
3598         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3599
3600         sattr->ia_mode &= ~current_umask();
3601         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3602         if (IS_ERR(state)) {
3603                 status = PTR_ERR(state);
3604                 goto out;
3605         }
3606 out:
3607         nfs4_label_release_security(ilabel);
3608         put_nfs_open_context(ctx);
3609         return status;
3610 }
3611
3612 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3613 {
3614         struct nfs_server *server = NFS_SERVER(dir);
3615         struct nfs_removeargs args = {
3616                 .fh = NFS_FH(dir),
3617                 .name = *name,
3618         };
3619         struct nfs_removeres res = {
3620                 .server = server,
3621         };
3622         struct rpc_message msg = {
3623                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3624                 .rpc_argp = &args,
3625                 .rpc_resp = &res,
3626         };
3627         int status;
3628
3629         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3630         if (status == 0)
3631                 update_changeattr(dir, &res.cinfo);
3632         return status;
3633 }
3634
3635 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3636 {
3637         struct nfs4_exception exception = { };
3638         int err;
3639         do {
3640                 err = _nfs4_proc_remove(dir, name);
3641                 trace_nfs4_remove(dir, name, err);
3642                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3643                                 &exception);
3644         } while (exception.retry);
3645         return err;
3646 }
3647
3648 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3649 {
3650         struct nfs_server *server = NFS_SERVER(dir);
3651         struct nfs_removeargs *args = msg->rpc_argp;
3652         struct nfs_removeres *res = msg->rpc_resp;
3653
3654         res->server = server;
3655         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3656         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3657
3658         nfs_fattr_init(res->dir_attr);
3659 }
3660
3661 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3662 {
3663         nfs4_setup_sequence(NFS_SERVER(data->dir),
3664                         &data->args.seq_args,
3665                         &data->res.seq_res,
3666                         task);
3667 }
3668
3669 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3670 {
3671         struct nfs_unlinkdata *data = task->tk_calldata;
3672         struct nfs_removeres *res = &data->res;
3673
3674         if (!nfs4_sequence_done(task, &res->seq_res))
3675                 return 0;
3676         if (nfs4_async_handle_error(task, res->server, NULL,
3677                                     &data->timeout) == -EAGAIN)
3678                 return 0;
3679         update_changeattr(dir, &res->cinfo);
3680         return 1;
3681 }
3682
3683 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3684 {
3685         struct nfs_server *server = NFS_SERVER(dir);
3686         struct nfs_renameargs *arg = msg->rpc_argp;
3687         struct nfs_renameres *res = msg->rpc_resp;
3688
3689         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3690         res->server = server;
3691         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3692 }
3693
3694 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3695 {
3696         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3697                         &data->args.seq_args,
3698                         &data->res.seq_res,
3699                         task);
3700 }
3701
3702 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3703                                  struct inode *new_dir)
3704 {
3705         struct nfs_renamedata *data = task->tk_calldata;
3706         struct nfs_renameres *res = &data->res;
3707
3708         if (!nfs4_sequence_done(task, &res->seq_res))
3709                 return 0;
3710         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3711                 return 0;
3712
3713         update_changeattr(old_dir, &res->old_cinfo);
3714         update_changeattr(new_dir, &res->new_cinfo);
3715         return 1;
3716 }
3717
3718 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3719 {
3720         struct nfs_server *server = NFS_SERVER(inode);
3721         struct nfs4_link_arg arg = {
3722                 .fh     = NFS_FH(inode),
3723                 .dir_fh = NFS_FH(dir),
3724                 .name   = name,
3725                 .bitmask = server->attr_bitmask,
3726         };
3727         struct nfs4_link_res res = {
3728                 .server = server,
3729                 .label = NULL,
3730         };
3731         struct rpc_message msg = {
3732                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3733                 .rpc_argp = &arg,
3734                 .rpc_resp = &res,
3735         };
3736         int status = -ENOMEM;
3737
3738         res.fattr = nfs_alloc_fattr();
3739         if (res.fattr == NULL)
3740                 goto out;
3741
3742         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3743         if (IS_ERR(res.label)) {
3744                 status = PTR_ERR(res.label);
3745                 goto out;
3746         }
3747         arg.bitmask = nfs4_bitmask(server, res.label);
3748
3749         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3750         if (!status) {
3751                 update_changeattr(dir, &res.cinfo);
3752                 status = nfs_post_op_update_inode(inode, res.fattr);
3753                 if (!status)
3754                         nfs_setsecurity(inode, res.fattr, res.label);
3755         }
3756
3757
3758         nfs4_label_free(res.label);
3759
3760 out:
3761         nfs_free_fattr(res.fattr);
3762         return status;
3763 }
3764
3765 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3766 {
3767         struct nfs4_exception exception = { };
3768         int err;
3769         do {
3770                 err = nfs4_handle_exception(NFS_SERVER(inode),
3771                                 _nfs4_proc_link(inode, dir, name),
3772                                 &exception);
3773         } while (exception.retry);
3774         return err;
3775 }
3776
3777 struct nfs4_createdata {
3778         struct rpc_message msg;
3779         struct nfs4_create_arg arg;
3780         struct nfs4_create_res res;
3781         struct nfs_fh fh;
3782         struct nfs_fattr fattr;
3783         struct nfs4_label *label;
3784 };
3785
3786 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3787                 struct qstr *name, struct iattr *sattr, u32 ftype)
3788 {
3789         struct nfs4_createdata *data;
3790
3791         data = kzalloc(sizeof(*data), GFP_KERNEL);
3792         if (data != NULL) {
3793                 struct nfs_server *server = NFS_SERVER(dir);
3794
3795                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3796                 if (IS_ERR(data->label))
3797                         goto out_free;
3798
3799                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3800                 data->msg.rpc_argp = &data->arg;
3801                 data->msg.rpc_resp = &data->res;
3802                 data->arg.dir_fh = NFS_FH(dir);
3803                 data->arg.server = server;
3804                 data->arg.name = name;
3805                 data->arg.attrs = sattr;
3806                 data->arg.ftype = ftype;
3807                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3808                 data->res.server = server;
3809                 data->res.fh = &data->fh;
3810                 data->res.fattr = &data->fattr;
3811                 data->res.label = data->label;
3812                 nfs_fattr_init(data->res.fattr);
3813         }
3814         return data;
3815 out_free:
3816         kfree(data);
3817         return NULL;
3818 }
3819
3820 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3821 {
3822         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3823                                     &data->arg.seq_args, &data->res.seq_res, 1);
3824         if (status == 0) {
3825                 update_changeattr(dir, &data->res.dir_cinfo);
3826                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3827         }
3828         return status;
3829 }
3830
3831 static void nfs4_free_createdata(struct nfs4_createdata *data)
3832 {
3833         nfs4_label_free(data->label);
3834         kfree(data);
3835 }
3836
3837 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3838                 struct page *page, unsigned int len, struct iattr *sattr,
3839                 struct nfs4_label *label)
3840 {
3841         struct nfs4_createdata *data;
3842         int status = -ENAMETOOLONG;
3843
3844         if (len > NFS4_MAXPATHLEN)
3845                 goto out;
3846
3847         status = -ENOMEM;
3848         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3849         if (data == NULL)
3850                 goto out;
3851
3852         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3853         data->arg.u.symlink.pages = &page;
3854         data->arg.u.symlink.len = len;
3855         data->arg.label = label;
3856         
3857         status = nfs4_do_create(dir, dentry, data);
3858
3859         nfs4_free_createdata(data);
3860 out:
3861         return status;
3862 }
3863
3864 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3865                 struct page *page, unsigned int len, struct iattr *sattr)
3866 {
3867         struct nfs4_exception exception = { };
3868         struct nfs4_label l, *label = NULL;
3869         int err;
3870
3871         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3872
3873         do {
3874                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3875                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3876                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3877                                 &exception);
3878         } while (exception.retry);
3879
3880         nfs4_label_release_security(label);
3881         return err;
3882 }
3883
3884 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3885                 struct iattr *sattr, struct nfs4_label *label)
3886 {
3887         struct nfs4_createdata *data;
3888         int status = -ENOMEM;
3889
3890         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3891         if (data == NULL)
3892                 goto out;
3893
3894         data->arg.label = label;
3895         status = nfs4_do_create(dir, dentry, data);
3896
3897         nfs4_free_createdata(data);
3898 out:
3899         return status;
3900 }
3901
3902 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3903                 struct iattr *sattr)
3904 {
3905         struct nfs4_exception exception = { };
3906         struct nfs4_label l, *label = NULL;
3907         int err;
3908
3909         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3910
3911         sattr->ia_mode &= ~current_umask();
3912         do {
3913                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3914                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3915                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3916                                 &exception);
3917         } while (exception.retry);
3918         nfs4_label_release_security(label);
3919
3920         return err;
3921 }
3922
3923 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3924                 u64 cookie, struct page **pages, unsigned int count, int plus)
3925 {
3926         struct inode            *dir = d_inode(dentry);
3927         struct nfs4_readdir_arg args = {
3928                 .fh = NFS_FH(dir),
3929                 .pages = pages,
3930                 .pgbase = 0,
3931                 .count = count,
3932                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3933                 .plus = plus,
3934         };
3935         struct nfs4_readdir_res res;
3936         struct rpc_message msg = {
3937                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3938                 .rpc_argp = &args,
3939                 .rpc_resp = &res,
3940                 .rpc_cred = cred,
3941         };
3942         int                     status;
3943
3944         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3945                         dentry,
3946                         (unsigned long long)cookie);
3947         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3948         res.pgbase = args.pgbase;
3949         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3950         if (status >= 0) {
3951                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3952                 status += args.pgbase;
3953         }
3954
3955         nfs_invalidate_atime(dir);
3956
3957         dprintk("%s: returns %d\n", __func__, status);
3958         return status;
3959 }
3960
3961 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3962                 u64 cookie, struct page **pages, unsigned int count, int plus)
3963 {
3964         struct nfs4_exception exception = { };
3965         int err;
3966         do {
3967                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3968                                 pages, count, plus);
3969                 trace_nfs4_readdir(d_inode(dentry), err);
3970                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3971                                 &exception);
3972         } while (exception.retry);
3973         return err;
3974 }
3975
3976 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3977                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3978 {
3979         struct nfs4_createdata *data;
3980         int mode = sattr->ia_mode;
3981         int status = -ENOMEM;
3982
3983         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3984         if (data == NULL)
3985                 goto out;
3986
3987         if (S_ISFIFO(mode))
3988                 data->arg.ftype = NF4FIFO;
3989         else if (S_ISBLK(mode)) {
3990                 data->arg.ftype = NF4BLK;
3991                 data->arg.u.device.specdata1 = MAJOR(rdev);
3992                 data->arg.u.device.specdata2 = MINOR(rdev);
3993         }
3994         else if (S_ISCHR(mode)) {
3995                 data->arg.ftype = NF4CHR;
3996                 data->arg.u.device.specdata1 = MAJOR(rdev);
3997                 data->arg.u.device.specdata2 = MINOR(rdev);
3998         } else if (!S_ISSOCK(mode)) {
3999                 status = -EINVAL;
4000                 goto out_free;
4001         }
4002
4003         data->arg.label = label;
4004         status = nfs4_do_create(dir, dentry, data);
4005 out_free:
4006         nfs4_free_createdata(data);
4007 out:
4008         return status;
4009 }
4010
4011 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4012                 struct iattr *sattr, dev_t rdev)
4013 {
4014         struct nfs4_exception exception = { };
4015         struct nfs4_label l, *label = NULL;
4016         int err;
4017
4018         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4019
4020         sattr->ia_mode &= ~current_umask();
4021         do {
4022                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4023                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4024                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4025                                 &exception);
4026         } while (exception.retry);
4027
4028         nfs4_label_release_security(label);
4029
4030         return err;
4031 }
4032
4033 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4034                  struct nfs_fsstat *fsstat)
4035 {
4036         struct nfs4_statfs_arg args = {
4037                 .fh = fhandle,
4038                 .bitmask = server->attr_bitmask,
4039         };
4040         struct nfs4_statfs_res res = {
4041                 .fsstat = fsstat,
4042         };
4043         struct rpc_message msg = {
4044                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4045                 .rpc_argp = &args,
4046                 .rpc_resp = &res,
4047         };
4048
4049         nfs_fattr_init(fsstat->fattr);
4050         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4051 }
4052
4053 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4054 {
4055         struct nfs4_exception exception = { };
4056         int err;
4057         do {
4058                 err = nfs4_handle_exception(server,
4059                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4060                                 &exception);
4061         } while (exception.retry);
4062         return err;
4063 }
4064
4065 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4066                 struct nfs_fsinfo *fsinfo)
4067 {
4068         struct nfs4_fsinfo_arg args = {
4069                 .fh = fhandle,
4070                 .bitmask = server->attr_bitmask,
4071         };
4072         struct nfs4_fsinfo_res res = {
4073                 .fsinfo = fsinfo,
4074         };
4075         struct rpc_message msg = {
4076                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4077                 .rpc_argp = &args,
4078                 .rpc_resp = &res,
4079         };
4080
4081         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4082 }
4083
4084 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4085 {
4086         struct nfs4_exception exception = { };
4087         unsigned long now = jiffies;
4088         int err;
4089
4090         do {
4091                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4092                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4093                 if (err == 0) {
4094                         struct nfs_client *clp = server->nfs_client;
4095
4096                         spin_lock(&clp->cl_lock);
4097                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4098                         clp->cl_last_renewal = now;
4099                         spin_unlock(&clp->cl_lock);
4100                         break;
4101                 }
4102                 err = nfs4_handle_exception(server, err, &exception);
4103         } while (exception.retry);
4104         return err;
4105 }
4106
4107 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4108 {
4109         int error;
4110
4111         nfs_fattr_init(fsinfo->fattr);
4112         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4113         if (error == 0) {
4114                 /* block layout checks this! */
4115                 server->pnfs_blksize = fsinfo->blksize;
4116                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4117         }
4118
4119         return error;
4120 }
4121
4122 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4123                 struct nfs_pathconf *pathconf)
4124 {
4125         struct nfs4_pathconf_arg args = {
4126                 .fh = fhandle,
4127                 .bitmask = server->attr_bitmask,
4128         };
4129         struct nfs4_pathconf_res res = {
4130                 .pathconf = pathconf,
4131         };
4132         struct rpc_message msg = {
4133                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4134                 .rpc_argp = &args,
4135                 .rpc_resp = &res,
4136         };
4137
4138         /* None of the pathconf attributes are mandatory to implement */
4139         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4140                 memset(pathconf, 0, sizeof(*pathconf));
4141                 return 0;
4142         }
4143
4144         nfs_fattr_init(pathconf->fattr);
4145         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4146 }
4147
4148 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4149                 struct nfs_pathconf *pathconf)
4150 {
4151         struct nfs4_exception exception = { };
4152         int err;
4153
4154         do {
4155                 err = nfs4_handle_exception(server,
4156                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4157                                 &exception);
4158         } while (exception.retry);
4159         return err;
4160 }
4161
4162 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4163                 const struct nfs_open_context *ctx,
4164                 const struct nfs_lock_context *l_ctx,
4165                 fmode_t fmode)
4166 {
4167         const struct nfs_lockowner *lockowner = NULL;
4168
4169         if (l_ctx != NULL)
4170                 lockowner = &l_ctx->lockowner;
4171         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4172 }
4173 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4174
4175 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4176                 const struct nfs_open_context *ctx,
4177                 const struct nfs_lock_context *l_ctx,
4178                 fmode_t fmode)
4179 {
4180         nfs4_stateid current_stateid;
4181
4182         /* If the current stateid represents a lost lock, then exit */
4183         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4184                 return true;
4185         return nfs4_stateid_match(stateid, &current_stateid);
4186 }
4187
4188 static bool nfs4_error_stateid_expired(int err)
4189 {
4190         switch (err) {
4191         case -NFS4ERR_DELEG_REVOKED:
4192         case -NFS4ERR_ADMIN_REVOKED:
4193         case -NFS4ERR_BAD_STATEID:
4194         case -NFS4ERR_STALE_STATEID:
4195         case -NFS4ERR_OLD_STATEID:
4196         case -NFS4ERR_OPENMODE:
4197         case -NFS4ERR_EXPIRED:
4198                 return true;
4199         }
4200         return false;
4201 }
4202
4203 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4204 {
4205         nfs_invalidate_atime(hdr->inode);
4206 }
4207
4208 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4209 {
4210         struct nfs_server *server = NFS_SERVER(hdr->inode);
4211
4212         trace_nfs4_read(hdr, task->tk_status);
4213         if (nfs4_async_handle_error(task, server,
4214                                     hdr->args.context->state,
4215                                     NULL) == -EAGAIN) {
4216                 rpc_restart_call_prepare(task);
4217                 return -EAGAIN;
4218         }
4219
4220         __nfs4_read_done_cb(hdr);
4221         if (task->tk_status > 0)
4222                 renew_lease(server, hdr->timestamp);
4223         return 0;
4224 }
4225
4226 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4227                 struct nfs_pgio_args *args)
4228 {
4229
4230         if (!nfs4_error_stateid_expired(task->tk_status) ||
4231                 nfs4_stateid_is_current(&args->stateid,
4232                                 args->context,
4233                                 args->lock_context,
4234                                 FMODE_READ))
4235                 return false;
4236         rpc_restart_call_prepare(task);
4237         return true;
4238 }
4239
4240 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4241 {
4242
4243         dprintk("--> %s\n", __func__);
4244
4245         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4246                 return -EAGAIN;
4247         if (nfs4_read_stateid_changed(task, &hdr->args))
4248                 return -EAGAIN;
4249         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4250                                     nfs4_read_done_cb(task, hdr);
4251 }
4252
4253 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4254                                  struct rpc_message *msg)
4255 {
4256         hdr->timestamp   = jiffies;
4257         hdr->pgio_done_cb = nfs4_read_done_cb;
4258         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4259         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4260 }
4261
4262 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4263                                       struct nfs_pgio_header *hdr)
4264 {
4265         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4266                         &hdr->args.seq_args,
4267                         &hdr->res.seq_res,
4268                         task))
4269                 return 0;
4270         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4271                                 hdr->args.lock_context,
4272                                 hdr->rw_ops->rw_mode) == -EIO)
4273                 return -EIO;
4274         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4275                 return -EIO;
4276         return 0;
4277 }
4278
4279 static int nfs4_write_done_cb(struct rpc_task *task,
4280                               struct nfs_pgio_header *hdr)
4281 {
4282         struct inode *inode = hdr->inode;
4283
4284         trace_nfs4_write(hdr, task->tk_status);
4285         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4286                                     hdr->args.context->state,
4287                                     NULL) == -EAGAIN) {
4288                 rpc_restart_call_prepare(task);
4289                 return -EAGAIN;
4290         }
4291         if (task->tk_status >= 0) {
4292                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4293                 nfs_writeback_update_inode(hdr);
4294         }
4295         return 0;
4296 }
4297
4298 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4299                 struct nfs_pgio_args *args)
4300 {
4301
4302         if (!nfs4_error_stateid_expired(task->tk_status) ||
4303                 nfs4_stateid_is_current(&args->stateid,
4304                                 args->context,
4305                                 args->lock_context,
4306                                 FMODE_WRITE))
4307                 return false;
4308         rpc_restart_call_prepare(task);
4309         return true;
4310 }
4311
4312 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4313 {
4314         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4315                 return -EAGAIN;
4316         if (nfs4_write_stateid_changed(task, &hdr->args))
4317                 return -EAGAIN;
4318         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4319                 nfs4_write_done_cb(task, hdr);
4320 }
4321
4322 static
4323 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4324 {
4325         /* Don't request attributes for pNFS or O_DIRECT writes */
4326         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4327                 return false;
4328         /* Otherwise, request attributes if and only if we don't hold
4329          * a delegation
4330          */
4331         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4332 }
4333
4334 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4335                                   struct rpc_message *msg)
4336 {
4337         struct nfs_server *server = NFS_SERVER(hdr->inode);
4338
4339         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4340                 hdr->args.bitmask = NULL;
4341                 hdr->res.fattr = NULL;
4342         } else
4343                 hdr->args.bitmask = server->cache_consistency_bitmask;
4344
4345         if (!hdr->pgio_done_cb)
4346                 hdr->pgio_done_cb = nfs4_write_done_cb;
4347         hdr->res.server = server;
4348         hdr->timestamp   = jiffies;
4349
4350         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4351         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4352 }
4353
4354 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4355 {
4356         nfs4_setup_sequence(NFS_SERVER(data->inode),
4357                         &data->args.seq_args,
4358                         &data->res.seq_res,
4359                         task);
4360 }
4361
4362 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4363 {
4364         struct inode *inode = data->inode;
4365
4366         trace_nfs4_commit(data, task->tk_status);
4367         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4368                                     NULL, NULL) == -EAGAIN) {
4369                 rpc_restart_call_prepare(task);
4370                 return -EAGAIN;
4371         }
4372         return 0;
4373 }
4374
4375 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4376 {
4377         if (!nfs4_sequence_done(task, &data->res.seq_res))
4378                 return -EAGAIN;
4379         return data->commit_done_cb(task, data);
4380 }
4381
4382 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4383 {
4384         struct nfs_server *server = NFS_SERVER(data->inode);
4385
4386         if (data->commit_done_cb == NULL)
4387                 data->commit_done_cb = nfs4_commit_done_cb;
4388         data->res.server = server;
4389         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4390         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4391 }
4392
4393 struct nfs4_renewdata {
4394         struct nfs_client       *client;
4395         unsigned long           timestamp;
4396 };
4397
4398 /*
4399  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4400  * standalone procedure for queueing an asynchronous RENEW.
4401  */
4402 static void nfs4_renew_release(void *calldata)
4403 {
4404         struct nfs4_renewdata *data = calldata;
4405         struct nfs_client *clp = data->client;
4406
4407         if (atomic_read(&clp->cl_count) > 1)
4408                 nfs4_schedule_state_renewal(clp);
4409         nfs_put_client(clp);
4410         kfree(data);
4411 }
4412
4413 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4414 {
4415         struct nfs4_renewdata *data = calldata;
4416         struct nfs_client *clp = data->client;
4417         unsigned long timestamp = data->timestamp;
4418
4419         trace_nfs4_renew_async(clp, task->tk_status);
4420         switch (task->tk_status) {
4421         case 0:
4422                 break;
4423         case -NFS4ERR_LEASE_MOVED:
4424                 nfs4_schedule_lease_moved_recovery(clp);
4425                 break;
4426         default:
4427                 /* Unless we're shutting down, schedule state recovery! */
4428                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4429                         return;
4430                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4431                         nfs4_schedule_lease_recovery(clp);
4432                         return;
4433                 }
4434                 nfs4_schedule_path_down_recovery(clp);
4435         }
4436         do_renew_lease(clp, timestamp);
4437 }
4438
4439 static const struct rpc_call_ops nfs4_renew_ops = {
4440         .rpc_call_done = nfs4_renew_done,
4441         .rpc_release = nfs4_renew_release,
4442 };
4443
4444 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4445 {
4446         struct rpc_message msg = {
4447                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4448                 .rpc_argp       = clp,
4449                 .rpc_cred       = cred,
4450         };
4451         struct nfs4_renewdata *data;
4452
4453         if (renew_flags == 0)
4454                 return 0;
4455         if (!atomic_inc_not_zero(&clp->cl_count))
4456                 return -EIO;
4457         data = kmalloc(sizeof(*data), GFP_NOFS);
4458         if (data == NULL)
4459                 return -ENOMEM;
4460         data->client = clp;
4461         data->timestamp = jiffies;
4462         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4463                         &nfs4_renew_ops, data);
4464 }
4465
4466 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4467 {
4468         struct rpc_message msg = {
4469                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4470                 .rpc_argp       = clp,
4471                 .rpc_cred       = cred,
4472         };
4473         unsigned long now = jiffies;
4474         int status;
4475
4476         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4477         if (status < 0)
4478                 return status;
4479         do_renew_lease(clp, now);
4480         return 0;
4481 }
4482
4483 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4484 {
4485         return server->caps & NFS_CAP_ACLS;
4486 }
4487
4488 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4489  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4490  * the stack.
4491  */
4492 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4493
4494 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4495                 struct page **pages, unsigned int *pgbase)
4496 {
4497         struct page *newpage, **spages;
4498         int rc = 0;
4499         size_t len;
4500         spages = pages;
4501
4502         do {
4503                 len = min_t(size_t, PAGE_SIZE, buflen);
4504                 newpage = alloc_page(GFP_KERNEL);
4505
4506                 if (newpage == NULL)
4507                         goto unwind;
4508                 memcpy(page_address(newpage), buf, len);
4509                 buf += len;
4510                 buflen -= len;
4511                 *pages++ = newpage;
4512                 rc++;
4513         } while (buflen != 0);
4514
4515         return rc;
4516
4517 unwind:
4518         for(; rc > 0; rc--)
4519                 __free_page(spages[rc-1]);
4520         return -ENOMEM;
4521 }
4522
4523 struct nfs4_cached_acl {
4524         int cached;
4525         size_t len;
4526         char data[0];
4527 };
4528
4529 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4530 {
4531         struct nfs_inode *nfsi = NFS_I(inode);
4532
4533         spin_lock(&inode->i_lock);
4534         kfree(nfsi->nfs4_acl);
4535         nfsi->nfs4_acl = acl;
4536         spin_unlock(&inode->i_lock);
4537 }
4538
4539 static void nfs4_zap_acl_attr(struct inode *inode)
4540 {
4541         nfs4_set_cached_acl(inode, NULL);
4542 }
4543
4544 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4545 {
4546         struct nfs_inode *nfsi = NFS_I(inode);
4547         struct nfs4_cached_acl *acl;
4548         int ret = -ENOENT;
4549
4550         spin_lock(&inode->i_lock);
4551         acl = nfsi->nfs4_acl;
4552         if (acl == NULL)
4553                 goto out;
4554         if (buf == NULL) /* user is just asking for length */
4555                 goto out_len;
4556         if (acl->cached == 0)
4557                 goto out;
4558         ret = -ERANGE; /* see getxattr(2) man page */
4559         if (acl->len > buflen)
4560                 goto out;
4561         memcpy(buf, acl->data, acl->len);
4562 out_len:
4563         ret = acl->len;
4564 out:
4565         spin_unlock(&inode->i_lock);
4566         return ret;
4567 }
4568
4569 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4570 {
4571         struct nfs4_cached_acl *acl;
4572         size_t buflen = sizeof(*acl) + acl_len;
4573
4574         if (buflen <= PAGE_SIZE) {
4575                 acl = kmalloc(buflen, GFP_KERNEL);
4576                 if (acl == NULL)
4577                         goto out;
4578                 acl->cached = 1;
4579                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4580         } else {
4581                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4582                 if (acl == NULL)
4583                         goto out;
4584                 acl->cached = 0;
4585         }
4586         acl->len = acl_len;
4587 out:
4588         nfs4_set_cached_acl(inode, acl);
4589 }
4590
4591 /*
4592  * The getxattr API returns the required buffer length when called with a
4593  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4594  * the required buf.  On a NULL buf, we send a page of data to the server
4595  * guessing that the ACL request can be serviced by a page. If so, we cache
4596  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4597  * the cache. If not so, we throw away the page, and cache the required
4598  * length. The next getxattr call will then produce another round trip to
4599  * the server, this time with the input buf of the required size.
4600  */
4601 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4602 {
4603         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4604         struct nfs_getaclargs args = {
4605                 .fh = NFS_FH(inode),
4606                 .acl_pages = pages,
4607                 .acl_len = buflen,
4608         };
4609         struct nfs_getaclres res = {
4610                 .acl_len = buflen,
4611         };
4612         struct rpc_message msg = {
4613                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4614                 .rpc_argp = &args,
4615                 .rpc_resp = &res,
4616         };
4617         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4618         int ret = -ENOMEM, i;
4619
4620         /* As long as we're doing a round trip to the server anyway,
4621          * let's be prepared for a page of acl data. */
4622         if (npages == 0)
4623                 npages = 1;
4624         if (npages > ARRAY_SIZE(pages))
4625                 return -ERANGE;
4626
4627         for (i = 0; i < npages; i++) {
4628                 pages[i] = alloc_page(GFP_KERNEL);
4629                 if (!pages[i])
4630                         goto out_free;
4631         }
4632
4633         /* for decoding across pages */
4634         res.acl_scratch = alloc_page(GFP_KERNEL);
4635         if (!res.acl_scratch)
4636                 goto out_free;
4637
4638         args.acl_len = npages * PAGE_SIZE;
4639         args.acl_pgbase = 0;
4640
4641         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4642                 __func__, buf, buflen, npages, args.acl_len);
4643         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4644                              &msg, &args.seq_args, &res.seq_res, 0);
4645         if (ret)
4646                 goto out_free;
4647
4648         /* Handle the case where the passed-in buffer is too short */
4649         if (res.acl_flags & NFS4_ACL_TRUNC) {
4650                 /* Did the user only issue a request for the acl length? */
4651                 if (buf == NULL)
4652                         goto out_ok;
4653                 ret = -ERANGE;
4654                 goto out_free;
4655         }
4656         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4657         if (buf) {
4658                 if (res.acl_len > buflen) {
4659                         ret = -ERANGE;
4660                         goto out_free;
4661                 }
4662                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4663         }
4664 out_ok:
4665         ret = res.acl_len;
4666 out_free:
4667         for (i = 0; i < npages; i++)
4668                 if (pages[i])
4669                         __free_page(pages[i]);
4670         if (res.acl_scratch)
4671                 __free_page(res.acl_scratch);
4672         return ret;
4673 }
4674
4675 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4676 {
4677         struct nfs4_exception exception = { };
4678         ssize_t ret;
4679         do {
4680                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4681                 trace_nfs4_get_acl(inode, ret);
4682                 if (ret >= 0)
4683                         break;
4684                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4685         } while (exception.retry);
4686         return ret;
4687 }
4688
4689 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4690 {
4691         struct nfs_server *server = NFS_SERVER(inode);
4692         int ret;
4693
4694         if (!nfs4_server_supports_acls(server))
4695                 return -EOPNOTSUPP;
4696         ret = nfs_revalidate_inode(server, inode);
4697         if (ret < 0)
4698                 return ret;
4699         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4700                 nfs_zap_acl_cache(inode);
4701         ret = nfs4_read_cached_acl(inode, buf, buflen);
4702         if (ret != -ENOENT)
4703                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4704                  * but no cached acl data, just the acl length */
4705                 return ret;
4706         return nfs4_get_acl_uncached(inode, buf, buflen);
4707 }
4708
4709 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4710 {
4711         struct nfs_server *server = NFS_SERVER(inode);
4712         struct page *pages[NFS4ACL_MAXPAGES];
4713         struct nfs_setaclargs arg = {
4714                 .fh             = NFS_FH(inode),
4715                 .acl_pages      = pages,
4716                 .acl_len        = buflen,
4717         };
4718         struct nfs_setaclres res;
4719         struct rpc_message msg = {
4720                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4721                 .rpc_argp       = &arg,
4722                 .rpc_resp       = &res,
4723         };
4724         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4725         int ret, i;
4726
4727         if (!nfs4_server_supports_acls(server))
4728                 return -EOPNOTSUPP;
4729         if (npages > ARRAY_SIZE(pages))
4730                 return -ERANGE;
4731         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4732         if (i < 0)
4733                 return i;
4734         nfs4_inode_return_delegation(inode);
4735         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4736
4737         /*
4738          * Free each page after tx, so the only ref left is
4739          * held by the network stack
4740          */
4741         for (; i > 0; i--)
4742                 put_page(pages[i-1]);
4743
4744         /*
4745          * Acl update can result in inode attribute update.
4746          * so mark the attribute cache invalid.
4747          */
4748         spin_lock(&inode->i_lock);
4749         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4750         spin_unlock(&inode->i_lock);
4751         nfs_access_zap_cache(inode);
4752         nfs_zap_acl_cache(inode);
4753         return ret;
4754 }
4755
4756 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4757 {
4758         struct nfs4_exception exception = { };
4759         int err;
4760         do {
4761                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4762                 trace_nfs4_set_acl(inode, err);
4763                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4764                                 &exception);
4765         } while (exception.retry);
4766         return err;
4767 }
4768
4769 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4770 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4771                                         size_t buflen)
4772 {
4773         struct nfs_server *server = NFS_SERVER(inode);
4774         struct nfs_fattr fattr;
4775         struct nfs4_label label = {0, 0, buflen, buf};
4776
4777         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4778         struct nfs4_getattr_arg arg = {
4779                 .fh             = NFS_FH(inode),
4780                 .bitmask        = bitmask,
4781         };
4782         struct nfs4_getattr_res res = {
4783                 .fattr          = &fattr,
4784                 .label          = &label,
4785                 .server         = server,
4786         };
4787         struct rpc_message msg = {
4788                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4789                 .rpc_argp       = &arg,
4790                 .rpc_resp       = &res,
4791         };
4792         int ret;
4793
4794         nfs_fattr_init(&fattr);
4795
4796         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4797         if (ret)
4798                 return ret;
4799         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4800                 return -ENOENT;
4801         if (buflen < label.len)
4802                 return -ERANGE;
4803         return 0;
4804 }
4805
4806 static int nfs4_get_security_label(struct inode *inode, void *buf,
4807                                         size_t buflen)
4808 {
4809         struct nfs4_exception exception = { };
4810         int err;
4811
4812         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4813                 return -EOPNOTSUPP;
4814
4815         do {
4816                 err = _nfs4_get_security_label(inode, buf, buflen);
4817                 trace_nfs4_get_security_label(inode, err);
4818                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4819                                 &exception);
4820         } while (exception.retry);
4821         return err;
4822 }
4823
4824 static int _nfs4_do_set_security_label(struct inode *inode,
4825                 struct nfs4_label *ilabel,
4826                 struct nfs_fattr *fattr,
4827                 struct nfs4_label *olabel)
4828 {
4829
4830         struct iattr sattr = {0};
4831         struct nfs_server *server = NFS_SERVER(inode);
4832         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4833         struct nfs_setattrargs arg = {
4834                 .fh             = NFS_FH(inode),
4835                 .iap            = &sattr,
4836                 .server         = server,
4837                 .bitmask        = bitmask,
4838                 .label          = ilabel,
4839         };
4840         struct nfs_setattrres res = {
4841                 .fattr          = fattr,
4842                 .label          = olabel,
4843                 .server         = server,
4844         };
4845         struct rpc_message msg = {
4846                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4847                 .rpc_argp       = &arg,
4848                 .rpc_resp       = &res,
4849         };
4850         int status;
4851
4852         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4853
4854         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4855         if (status)
4856                 dprintk("%s failed: %d\n", __func__, status);
4857
4858         return status;
4859 }
4860
4861 static int nfs4_do_set_security_label(struct inode *inode,
4862                 struct nfs4_label *ilabel,
4863                 struct nfs_fattr *fattr,
4864                 struct nfs4_label *olabel)
4865 {
4866         struct nfs4_exception exception = { };
4867         int err;
4868
4869         do {
4870                 err = _nfs4_do_set_security_label(inode, ilabel,
4871                                 fattr, olabel);
4872                 trace_nfs4_set_security_label(inode, err);
4873                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4874                                 &exception);
4875         } while (exception.retry);
4876         return err;
4877 }
4878
4879 static int
4880 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4881 {
4882         struct nfs4_label ilabel, *olabel = NULL;
4883         struct nfs_fattr fattr;
4884         struct rpc_cred *cred;
4885         struct inode *inode = d_inode(dentry);
4886         int status;
4887
4888         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4889                 return -EOPNOTSUPP;
4890
4891         nfs_fattr_init(&fattr);
4892
4893         ilabel.pi = 0;
4894         ilabel.lfs = 0;
4895         ilabel.label = (char *)buf;
4896         ilabel.len = buflen;
4897
4898         cred = rpc_lookup_cred();
4899         if (IS_ERR(cred))
4900                 return PTR_ERR(cred);
4901
4902         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4903         if (IS_ERR(olabel)) {
4904                 status = -PTR_ERR(olabel);
4905                 goto out;
4906         }
4907
4908         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4909         if (status == 0)
4910                 nfs_setsecurity(inode, &fattr, olabel);
4911
4912         nfs4_label_free(olabel);
4913 out:
4914         put_rpccred(cred);
4915         return status;
4916 }
4917 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4918
4919
4920 static int
4921 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4922                         struct nfs4_state *state, long *timeout)
4923 {
4924         struct nfs_client *clp = server->nfs_client;
4925
4926         if (task->tk_status >= 0)
4927                 return 0;
4928         switch(task->tk_status) {
4929                 case -NFS4ERR_DELEG_REVOKED:
4930                 case -NFS4ERR_ADMIN_REVOKED:
4931                 case -NFS4ERR_BAD_STATEID:
4932                 case -NFS4ERR_OPENMODE:
4933                         if (state == NULL)
4934                                 break;
4935                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4936                                 goto recovery_failed;
4937                         goto wait_on_recovery;
4938                 case -NFS4ERR_EXPIRED:
4939                         if (state != NULL) {
4940                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4941                                         goto recovery_failed;
4942                         }
4943                 case -NFS4ERR_STALE_STATEID:
4944                 case -NFS4ERR_STALE_CLIENTID:
4945                         nfs4_schedule_lease_recovery(clp);
4946                         goto wait_on_recovery;
4947                 case -NFS4ERR_MOVED:
4948                         if (nfs4_schedule_migration_recovery(server) < 0)
4949                                 goto recovery_failed;
4950                         goto wait_on_recovery;
4951                 case -NFS4ERR_LEASE_MOVED:
4952                         nfs4_schedule_lease_moved_recovery(clp);
4953                         goto wait_on_recovery;
4954 #if defined(CONFIG_NFS_V4_1)
4955                 case -NFS4ERR_BADSESSION:
4956                 case -NFS4ERR_BADSLOT:
4957                 case -NFS4ERR_BAD_HIGH_SLOT:
4958                 case -NFS4ERR_DEADSESSION:
4959                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4960                 case -NFS4ERR_SEQ_FALSE_RETRY:
4961                 case -NFS4ERR_SEQ_MISORDERED:
4962                         dprintk("%s ERROR %d, Reset session\n", __func__,
4963                                 task->tk_status);
4964                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4965                         goto wait_on_recovery;
4966 #endif /* CONFIG_NFS_V4_1 */
4967                 case -NFS4ERR_DELAY:
4968                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4969                         rpc_delay(task, nfs4_update_delay(timeout));
4970                         goto restart_call;
4971                 case -NFS4ERR_GRACE:
4972                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4973                 case -NFS4ERR_RETRY_UNCACHED_REP:
4974                 case -NFS4ERR_OLD_STATEID:
4975                         goto restart_call;
4976         }
4977         task->tk_status = nfs4_map_errors(task->tk_status);
4978         return 0;
4979 recovery_failed:
4980         task->tk_status = -EIO;
4981         return 0;
4982 wait_on_recovery:
4983         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4984         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4985                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4986         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4987                 goto recovery_failed;
4988 restart_call:
4989         task->tk_status = 0;
4990         return -EAGAIN;
4991 }
4992
4993 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4994                                     nfs4_verifier *bootverf)
4995 {
4996         __be32 verf[2];
4997
4998         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4999                 /* An impossible timestamp guarantees this value
5000                  * will never match a generated boot time. */
5001                 verf[0] = 0;
5002                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5003         } else {
5004                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5005                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5006                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5007         }
5008         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5009 }
5010
5011 static int
5012 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5013 {
5014         int result;
5015         size_t len;
5016         char *str;
5017
5018         if (clp->cl_owner_id != NULL)
5019                 return 0;
5020
5021         rcu_read_lock();
5022         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5023                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5024                 1 +
5025                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5026                 1;
5027         rcu_read_unlock();
5028
5029         if (len > NFS4_OPAQUE_LIMIT + 1)
5030                 return -EINVAL;
5031
5032         /*
5033          * Since this string is allocated at mount time, and held until the
5034          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5035          * about a memory-reclaim deadlock.
5036          */
5037         str = kmalloc(len, GFP_KERNEL);
5038         if (!str)
5039                 return -ENOMEM;
5040
5041         rcu_read_lock();
5042         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5043                         clp->cl_ipaddr,
5044                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5045                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5046         rcu_read_unlock();
5047
5048         clp->cl_owner_id = str;
5049         return 0;
5050 }
5051
5052 static int
5053 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5054 {
5055         int result;
5056         size_t len;
5057         char *str;
5058
5059         len = 10 + 10 + 1 + 10 + 1 +
5060                 strlen(nfs4_client_id_uniquifier) + 1 +
5061                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5062
5063         if (len > NFS4_OPAQUE_LIMIT + 1)
5064                 return -EINVAL;
5065
5066         /*
5067          * Since this string is allocated at mount time, and held until the
5068          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5069          * about a memory-reclaim deadlock.
5070          */
5071         str = kmalloc(len, GFP_KERNEL);
5072         if (!str)
5073                 return -ENOMEM;
5074
5075         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5076                         clp->rpc_ops->version, clp->cl_minorversion,
5077                         nfs4_client_id_uniquifier,
5078                         clp->cl_rpcclient->cl_nodename);
5079         clp->cl_owner_id = str;
5080         return 0;
5081 }
5082
5083 static int
5084 nfs4_init_uniform_client_string(struct nfs_client *clp)
5085 {
5086         int result;
5087         size_t len;
5088         char *str;
5089
5090         if (clp->cl_owner_id != NULL)
5091                 return 0;
5092
5093         if (nfs4_client_id_uniquifier[0] != '\0')
5094                 return nfs4_init_uniquifier_client_string(clp);
5095
5096         len = 10 + 10 + 1 + 10 + 1 +
5097                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5098
5099         if (len > NFS4_OPAQUE_LIMIT + 1)
5100                 return -EINVAL;
5101
5102         /*
5103          * Since this string is allocated at mount time, and held until the
5104          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5105          * about a memory-reclaim deadlock.
5106          */
5107         str = kmalloc(len, GFP_KERNEL);
5108         if (!str)
5109                 return -ENOMEM;
5110
5111         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5112                         clp->rpc_ops->version, clp->cl_minorversion,
5113                         clp->cl_rpcclient->cl_nodename);
5114         clp->cl_owner_id = str;
5115         return 0;
5116 }
5117
5118 /*
5119  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5120  * services.  Advertise one based on the address family of the
5121  * clientaddr.
5122  */
5123 static unsigned int
5124 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5125 {
5126         if (strchr(clp->cl_ipaddr, ':') != NULL)
5127                 return scnprintf(buf, len, "tcp6");
5128         else
5129                 return scnprintf(buf, len, "tcp");
5130 }
5131
5132 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5133 {
5134         struct nfs4_setclientid *sc = calldata;
5135
5136         if (task->tk_status == 0)
5137                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5138 }
5139
5140 static const struct rpc_call_ops nfs4_setclientid_ops = {
5141         .rpc_call_done = nfs4_setclientid_done,
5142 };
5143
5144 /**
5145  * nfs4_proc_setclientid - Negotiate client ID
5146  * @clp: state data structure
5147  * @program: RPC program for NFSv4 callback service
5148  * @port: IP port number for NFS4 callback service
5149  * @cred: RPC credential to use for this call
5150  * @res: where to place the result
5151  *
5152  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5153  */
5154 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5155                 unsigned short port, struct rpc_cred *cred,
5156                 struct nfs4_setclientid_res *res)
5157 {
5158         nfs4_verifier sc_verifier;
5159         struct nfs4_setclientid setclientid = {
5160                 .sc_verifier = &sc_verifier,
5161                 .sc_prog = program,
5162                 .sc_clnt = clp,
5163         };
5164         struct rpc_message msg = {
5165                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5166                 .rpc_argp = &setclientid,
5167                 .rpc_resp = res,
5168                 .rpc_cred = cred,
5169         };
5170         struct rpc_task *task;
5171         struct rpc_task_setup task_setup_data = {
5172                 .rpc_client = clp->cl_rpcclient,
5173                 .rpc_message = &msg,
5174                 .callback_ops = &nfs4_setclientid_ops,
5175                 .callback_data = &setclientid,
5176                 .flags = RPC_TASK_TIMEOUT,
5177         };
5178         int status;
5179
5180         /* nfs_client_id4 */
5181         nfs4_init_boot_verifier(clp, &sc_verifier);
5182
5183         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5184                 status = nfs4_init_uniform_client_string(clp);
5185         else
5186                 status = nfs4_init_nonuniform_client_string(clp);
5187
5188         if (status)
5189                 goto out;
5190
5191         /* cb_client4 */
5192         setclientid.sc_netid_len =
5193                                 nfs4_init_callback_netid(clp,
5194                                                 setclientid.sc_netid,
5195                                                 sizeof(setclientid.sc_netid));
5196         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5197                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5198                                 clp->cl_ipaddr, port >> 8, port & 255);
5199
5200         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5201                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5202                 clp->cl_owner_id);
5203         task = rpc_run_task(&task_setup_data);
5204         if (IS_ERR(task)) {
5205                 status = PTR_ERR(task);
5206                 goto out;
5207         }
5208         status = task->tk_status;
5209         if (setclientid.sc_cred) {
5210                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5211                 put_rpccred(setclientid.sc_cred);
5212         }
5213         rpc_put_task(task);
5214 out:
5215         trace_nfs4_setclientid(clp, status);
5216         dprintk("NFS reply setclientid: %d\n", status);
5217         return status;
5218 }
5219
5220 /**
5221  * nfs4_proc_setclientid_confirm - Confirm client ID
5222  * @clp: state data structure
5223  * @res: result of a previous SETCLIENTID
5224  * @cred: RPC credential to use for this call
5225  *
5226  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5227  */
5228 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5229                 struct nfs4_setclientid_res *arg,
5230                 struct rpc_cred *cred)
5231 {
5232         struct rpc_message msg = {
5233                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5234                 .rpc_argp = arg,
5235                 .rpc_cred = cred,
5236         };
5237         int status;
5238
5239         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5240                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5241                 clp->cl_clientid);
5242         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5243         trace_nfs4_setclientid_confirm(clp, status);
5244         dprintk("NFS reply setclientid_confirm: %d\n", status);
5245         return status;
5246 }
5247
5248 struct nfs4_delegreturndata {
5249         struct nfs4_delegreturnargs args;
5250         struct nfs4_delegreturnres res;
5251         struct nfs_fh fh;
5252         nfs4_stateid stateid;
5253         unsigned long timestamp;
5254         struct nfs_fattr fattr;
5255         int rpc_status;
5256         struct inode *inode;
5257         bool roc;
5258         u32 roc_barrier;
5259 };
5260
5261 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5262 {
5263         struct nfs4_delegreturndata *data = calldata;
5264
5265         if (!nfs4_sequence_done(task, &data->res.seq_res))
5266                 return;
5267
5268         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5269         switch (task->tk_status) {
5270         case 0:
5271                 renew_lease(data->res.server, data->timestamp);
5272         case -NFS4ERR_ADMIN_REVOKED:
5273         case -NFS4ERR_DELEG_REVOKED:
5274         case -NFS4ERR_BAD_STATEID:
5275         case -NFS4ERR_OLD_STATEID:
5276         case -NFS4ERR_STALE_STATEID:
5277         case -NFS4ERR_EXPIRED:
5278                 task->tk_status = 0;
5279                 if (data->roc)
5280                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5281                 break;
5282         default:
5283                 if (nfs4_async_handle_error(task, data->res.server,
5284                                             NULL, NULL) == -EAGAIN) {
5285                         rpc_restart_call_prepare(task);
5286                         return;
5287                 }
5288         }
5289         data->rpc_status = task->tk_status;
5290 }
5291
5292 static void nfs4_delegreturn_release(void *calldata)
5293 {
5294         struct nfs4_delegreturndata *data = calldata;
5295         struct inode *inode = data->inode;
5296
5297         if (inode) {
5298                 if (data->roc)
5299                         pnfs_roc_release(inode);
5300                 nfs_iput_and_deactive(inode);
5301         }
5302         kfree(calldata);
5303 }
5304
5305 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5306 {
5307         struct nfs4_delegreturndata *d_data;
5308
5309         d_data = (struct nfs4_delegreturndata *)data;
5310
5311         if (d_data->roc)
5312                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5313
5314         nfs4_setup_sequence(d_data->res.server,
5315                         &d_data->args.seq_args,
5316                         &d_data->res.seq_res,
5317                         task);
5318 }
5319
5320 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5321         .rpc_call_prepare = nfs4_delegreturn_prepare,
5322         .rpc_call_done = nfs4_delegreturn_done,
5323         .rpc_release = nfs4_delegreturn_release,
5324 };
5325
5326 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5327 {
5328         struct nfs4_delegreturndata *data;
5329         struct nfs_server *server = NFS_SERVER(inode);
5330         struct rpc_task *task;
5331         struct rpc_message msg = {
5332                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5333                 .rpc_cred = cred,
5334         };
5335         struct rpc_task_setup task_setup_data = {
5336                 .rpc_client = server->client,
5337                 .rpc_message = &msg,
5338                 .callback_ops = &nfs4_delegreturn_ops,
5339                 .flags = RPC_TASK_ASYNC,
5340         };
5341         int status = 0;
5342
5343         data = kzalloc(sizeof(*data), GFP_NOFS);
5344         if (data == NULL)
5345                 return -ENOMEM;
5346         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5347         data->args.fhandle = &data->fh;
5348         data->args.stateid = &data->stateid;
5349         data->args.bitmask = server->cache_consistency_bitmask;
5350         nfs_copy_fh(&data->fh, NFS_FH(inode));
5351         nfs4_stateid_copy(&data->stateid, stateid);
5352         data->res.fattr = &data->fattr;
5353         data->res.server = server;
5354         nfs_fattr_init(data->res.fattr);
5355         data->timestamp = jiffies;
5356         data->rpc_status = 0;
5357         data->inode = nfs_igrab_and_active(inode);
5358         if (data->inode)
5359                 data->roc = nfs4_roc(inode);
5360
5361         task_setup_data.callback_data = data;
5362         msg.rpc_argp = &data->args;
5363         msg.rpc_resp = &data->res;
5364         task = rpc_run_task(&task_setup_data);
5365         if (IS_ERR(task))
5366                 return PTR_ERR(task);
5367         if (!issync)
5368                 goto out;
5369         status = nfs4_wait_for_completion_rpc_task(task);
5370         if (status != 0)
5371                 goto out;
5372         status = data->rpc_status;
5373         if (status == 0)
5374                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5375         else
5376                 nfs_refresh_inode(inode, &data->fattr);
5377 out:
5378         rpc_put_task(task);
5379         return status;
5380 }
5381
5382 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5383 {
5384         struct nfs_server *server = NFS_SERVER(inode);
5385         struct nfs4_exception exception = { };
5386         int err;
5387         do {
5388                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5389                 trace_nfs4_delegreturn(inode, err);
5390                 switch (err) {
5391                         case -NFS4ERR_STALE_STATEID:
5392                         case -NFS4ERR_EXPIRED:
5393                         case 0:
5394                                 return 0;
5395                 }
5396                 err = nfs4_handle_exception(server, err, &exception);
5397         } while (exception.retry);
5398         return err;
5399 }
5400
5401 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5402 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5403
5404 /* 
5405  * sleep, with exponential backoff, and retry the LOCK operation. 
5406  */
5407 static unsigned long
5408 nfs4_set_lock_task_retry(unsigned long timeout)
5409 {
5410         freezable_schedule_timeout_killable_unsafe(timeout);
5411         timeout <<= 1;
5412         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5413                 return NFS4_LOCK_MAXTIMEOUT;
5414         return timeout;
5415 }
5416
5417 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5418 {
5419         struct inode *inode = state->inode;
5420         struct nfs_server *server = NFS_SERVER(inode);
5421         struct nfs_client *clp = server->nfs_client;
5422         struct nfs_lockt_args arg = {
5423                 .fh = NFS_FH(inode),
5424                 .fl = request,
5425         };
5426         struct nfs_lockt_res res = {
5427                 .denied = request,
5428         };
5429         struct rpc_message msg = {
5430                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5431                 .rpc_argp       = &arg,
5432                 .rpc_resp       = &res,
5433                 .rpc_cred       = state->owner->so_cred,
5434         };
5435         struct nfs4_lock_state *lsp;
5436         int status;
5437
5438         arg.lock_owner.clientid = clp->cl_clientid;
5439         status = nfs4_set_lock_state(state, request);
5440         if (status != 0)
5441                 goto out;
5442         lsp = request->fl_u.nfs4_fl.owner;
5443         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5444         arg.lock_owner.s_dev = server->s_dev;
5445         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5446         switch (status) {
5447                 case 0:
5448                         request->fl_type = F_UNLCK;
5449                         break;
5450                 case -NFS4ERR_DENIED:
5451                         status = 0;
5452         }
5453         request->fl_ops->fl_release_private(request);
5454         request->fl_ops = NULL;
5455 out:
5456         return status;
5457 }
5458
5459 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5460 {
5461         struct nfs4_exception exception = { };
5462         int err;
5463
5464         do {
5465                 err = _nfs4_proc_getlk(state, cmd, request);
5466                 trace_nfs4_get_lock(request, state, cmd, err);
5467                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5468                                 &exception);
5469         } while (exception.retry);
5470         return err;
5471 }
5472
5473 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5474 {
5475         return locks_lock_inode_wait(inode, fl);
5476 }
5477
5478 struct nfs4_unlockdata {
5479         struct nfs_locku_args arg;
5480         struct nfs_locku_res res;
5481         struct nfs4_lock_state *lsp;
5482         struct nfs_open_context *ctx;
5483         struct file_lock fl;
5484         const struct nfs_server *server;
5485         unsigned long timestamp;
5486 };
5487
5488 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5489                 struct nfs_open_context *ctx,
5490                 struct nfs4_lock_state *lsp,
5491                 struct nfs_seqid *seqid)
5492 {
5493         struct nfs4_unlockdata *p;
5494         struct inode *inode = lsp->ls_state->inode;
5495
5496         p = kzalloc(sizeof(*p), GFP_NOFS);
5497         if (p == NULL)
5498                 return NULL;
5499         p->arg.fh = NFS_FH(inode);
5500         p->arg.fl = &p->fl;
5501         p->arg.seqid = seqid;
5502         p->res.seqid = seqid;
5503         p->lsp = lsp;
5504         atomic_inc(&lsp->ls_count);
5505         /* Ensure we don't close file until we're done freeing locks! */
5506         p->ctx = get_nfs_open_context(ctx);
5507         memcpy(&p->fl, fl, sizeof(p->fl));
5508         p->server = NFS_SERVER(inode);
5509         return p;
5510 }
5511
5512 static void nfs4_locku_release_calldata(void *data)
5513 {
5514         struct nfs4_unlockdata *calldata = data;
5515         nfs_free_seqid(calldata->arg.seqid);
5516         nfs4_put_lock_state(calldata->lsp);
5517         put_nfs_open_context(calldata->ctx);
5518         kfree(calldata);
5519 }
5520
5521 static void nfs4_locku_done(struct rpc_task *task, void *data)
5522 {
5523         struct nfs4_unlockdata *calldata = data;
5524
5525         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5526                 return;
5527         switch (task->tk_status) {
5528                 case 0:
5529                         renew_lease(calldata->server, calldata->timestamp);
5530                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5531                         if (nfs4_update_lock_stateid(calldata->lsp,
5532                                         &calldata->res.stateid))
5533                                 break;
5534                 case -NFS4ERR_BAD_STATEID:
5535                 case -NFS4ERR_OLD_STATEID:
5536                 case -NFS4ERR_STALE_STATEID:
5537                 case -NFS4ERR_EXPIRED:
5538                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5539                                                 &calldata->lsp->ls_stateid))
5540                                 rpc_restart_call_prepare(task);
5541                         break;
5542                 default:
5543                         if (nfs4_async_handle_error(task, calldata->server,
5544                                                     NULL, NULL) == -EAGAIN)
5545                                 rpc_restart_call_prepare(task);
5546         }
5547         nfs_release_seqid(calldata->arg.seqid);
5548 }
5549
5550 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5551 {
5552         struct nfs4_unlockdata *calldata = data;
5553
5554         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5555                 goto out_wait;
5556         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5557         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5558                 /* Note: exit _without_ running nfs4_locku_done */
5559                 goto out_no_action;
5560         }
5561         calldata->timestamp = jiffies;
5562         if (nfs4_setup_sequence(calldata->server,
5563                                 &calldata->arg.seq_args,
5564                                 &calldata->res.seq_res,
5565                                 task) != 0)
5566                 nfs_release_seqid(calldata->arg.seqid);
5567         return;
5568 out_no_action:
5569         task->tk_action = NULL;
5570 out_wait:
5571         nfs4_sequence_done(task, &calldata->res.seq_res);
5572 }
5573
5574 static const struct rpc_call_ops nfs4_locku_ops = {
5575         .rpc_call_prepare = nfs4_locku_prepare,
5576         .rpc_call_done = nfs4_locku_done,
5577         .rpc_release = nfs4_locku_release_calldata,
5578 };
5579
5580 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5581                 struct nfs_open_context *ctx,
5582                 struct nfs4_lock_state *lsp,
5583                 struct nfs_seqid *seqid)
5584 {
5585         struct nfs4_unlockdata *data;
5586         struct rpc_message msg = {
5587                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5588                 .rpc_cred = ctx->cred,
5589         };
5590         struct rpc_task_setup task_setup_data = {
5591                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5592                 .rpc_message = &msg,
5593                 .callback_ops = &nfs4_locku_ops,
5594                 .workqueue = nfsiod_workqueue,
5595                 .flags = RPC_TASK_ASYNC,
5596         };
5597
5598         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5599                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5600
5601         /* Ensure this is an unlock - when canceling a lock, the
5602          * canceled lock is passed in, and it won't be an unlock.
5603          */
5604         fl->fl_type = F_UNLCK;
5605
5606         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5607         if (data == NULL) {
5608                 nfs_free_seqid(seqid);
5609                 return ERR_PTR(-ENOMEM);
5610         }
5611
5612         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5613         msg.rpc_argp = &data->arg;
5614         msg.rpc_resp = &data->res;
5615         task_setup_data.callback_data = data;
5616         return rpc_run_task(&task_setup_data);
5617 }
5618
5619 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5620 {
5621         struct inode *inode = state->inode;
5622         struct nfs4_state_owner *sp = state->owner;
5623         struct nfs_inode *nfsi = NFS_I(inode);
5624         struct nfs_seqid *seqid;
5625         struct nfs4_lock_state *lsp;
5626         struct rpc_task *task;
5627         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5628         int status = 0;
5629         unsigned char fl_flags = request->fl_flags;
5630
5631         status = nfs4_set_lock_state(state, request);
5632         /* Unlock _before_ we do the RPC call */
5633         request->fl_flags |= FL_EXISTS;
5634         /* Exclude nfs_delegation_claim_locks() */
5635         mutex_lock(&sp->so_delegreturn_mutex);
5636         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5637         down_read(&nfsi->rwsem);
5638         if (do_vfs_lock(inode, request) == -ENOENT) {
5639                 up_read(&nfsi->rwsem);
5640                 mutex_unlock(&sp->so_delegreturn_mutex);
5641                 goto out;
5642         }
5643         up_read(&nfsi->rwsem);
5644         mutex_unlock(&sp->so_delegreturn_mutex);
5645         if (status != 0)
5646                 goto out;
5647         /* Is this a delegated lock? */
5648         lsp = request->fl_u.nfs4_fl.owner;
5649         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5650                 goto out;
5651         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5652         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5653         status = -ENOMEM;
5654         if (IS_ERR(seqid))
5655                 goto out;
5656         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5657         status = PTR_ERR(task);
5658         if (IS_ERR(task))
5659                 goto out;
5660         status = nfs4_wait_for_completion_rpc_task(task);
5661         rpc_put_task(task);
5662 out:
5663         request->fl_flags = fl_flags;
5664         trace_nfs4_unlock(request, state, F_SETLK, status);
5665         return status;
5666 }
5667
5668 struct nfs4_lockdata {
5669         struct nfs_lock_args arg;
5670         struct nfs_lock_res res;
5671         struct nfs4_lock_state *lsp;
5672         struct nfs_open_context *ctx;
5673         struct file_lock fl;
5674         unsigned long timestamp;
5675         int rpc_status;
5676         int cancelled;
5677         struct nfs_server *server;
5678 };
5679
5680 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5681                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5682                 gfp_t gfp_mask)
5683 {
5684         struct nfs4_lockdata *p;
5685         struct inode *inode = lsp->ls_state->inode;
5686         struct nfs_server *server = NFS_SERVER(inode);
5687         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5688
5689         p = kzalloc(sizeof(*p), gfp_mask);
5690         if (p == NULL)
5691                 return NULL;
5692
5693         p->arg.fh = NFS_FH(inode);
5694         p->arg.fl = &p->fl;
5695         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5696         if (IS_ERR(p->arg.open_seqid))
5697                 goto out_free;
5698         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5699         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5700         if (IS_ERR(p->arg.lock_seqid))
5701                 goto out_free_seqid;
5702         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5703         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5704         p->arg.lock_owner.s_dev = server->s_dev;
5705         p->res.lock_seqid = p->arg.lock_seqid;
5706         p->lsp = lsp;
5707         p->server = server;
5708         atomic_inc(&lsp->ls_count);
5709         p->ctx = get_nfs_open_context(ctx);
5710         get_file(fl->fl_file);
5711         memcpy(&p->fl, fl, sizeof(p->fl));
5712         return p;
5713 out_free_seqid:
5714         nfs_free_seqid(p->arg.open_seqid);
5715 out_free:
5716         kfree(p);
5717         return NULL;
5718 }
5719
5720 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5721 {
5722         struct nfs4_lockdata *data = calldata;
5723         struct nfs4_state *state = data->lsp->ls_state;
5724
5725         dprintk("%s: begin!\n", __func__);
5726         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5727                 goto out_wait;
5728         /* Do we need to do an open_to_lock_owner? */
5729         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5730                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5731                         goto out_release_lock_seqid;
5732                 }
5733                 nfs4_stateid_copy(&data->arg.open_stateid,
5734                                 &state->open_stateid);
5735                 data->arg.new_lock_owner = 1;
5736                 data->res.open_seqid = data->arg.open_seqid;
5737         } else {
5738                 data->arg.new_lock_owner = 0;
5739                 nfs4_stateid_copy(&data->arg.lock_stateid,
5740                                 &data->lsp->ls_stateid);
5741         }
5742         if (!nfs4_valid_open_stateid(state)) {
5743                 data->rpc_status = -EBADF;
5744                 task->tk_action = NULL;
5745                 goto out_release_open_seqid;
5746         }
5747         data->timestamp = jiffies;
5748         if (nfs4_setup_sequence(data->server,
5749                                 &data->arg.seq_args,
5750                                 &data->res.seq_res,
5751                                 task) == 0)
5752                 return;
5753 out_release_open_seqid:
5754         nfs_release_seqid(data->arg.open_seqid);
5755 out_release_lock_seqid:
5756         nfs_release_seqid(data->arg.lock_seqid);
5757 out_wait:
5758         nfs4_sequence_done(task, &data->res.seq_res);
5759         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5760 }
5761
5762 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5763 {
5764         struct nfs4_lockdata *data = calldata;
5765         struct nfs4_lock_state *lsp = data->lsp;
5766
5767         dprintk("%s: begin!\n", __func__);
5768
5769         if (!nfs4_sequence_done(task, &data->res.seq_res))
5770                 return;
5771
5772         data->rpc_status = task->tk_status;
5773         switch (task->tk_status) {
5774         case 0:
5775                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5776                                 data->timestamp);
5777                 if (data->arg.new_lock) {
5778                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5779                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5780                                 rpc_restart_call_prepare(task);
5781                                 break;
5782                         }
5783                 }
5784                 if (data->arg.new_lock_owner != 0) {
5785                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5786                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5787                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5788                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5789                         rpc_restart_call_prepare(task);
5790                 break;
5791         case -NFS4ERR_BAD_STATEID:
5792         case -NFS4ERR_OLD_STATEID:
5793         case -NFS4ERR_STALE_STATEID:
5794         case -NFS4ERR_EXPIRED:
5795                 if (data->arg.new_lock_owner != 0) {
5796                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5797                                                 &lsp->ls_state->open_stateid))
5798                                 rpc_restart_call_prepare(task);
5799                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5800                                                 &lsp->ls_stateid))
5801                                 rpc_restart_call_prepare(task);
5802         }
5803         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5804 }
5805
5806 static void nfs4_lock_release(void *calldata)
5807 {
5808         struct nfs4_lockdata *data = calldata;
5809
5810         dprintk("%s: begin!\n", __func__);
5811         nfs_free_seqid(data->arg.open_seqid);
5812         if (data->cancelled != 0) {
5813                 struct rpc_task *task;
5814                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5815                                 data->arg.lock_seqid);
5816                 if (!IS_ERR(task))
5817                         rpc_put_task_async(task);
5818                 dprintk("%s: cancelling lock!\n", __func__);
5819         } else
5820                 nfs_free_seqid(data->arg.lock_seqid);
5821         nfs4_put_lock_state(data->lsp);
5822         put_nfs_open_context(data->ctx);
5823         fput(data->fl.fl_file);
5824         kfree(data);
5825         dprintk("%s: done!\n", __func__);
5826 }
5827
5828 static const struct rpc_call_ops nfs4_lock_ops = {
5829         .rpc_call_prepare = nfs4_lock_prepare,
5830         .rpc_call_done = nfs4_lock_done,
5831         .rpc_release = nfs4_lock_release,
5832 };
5833
5834 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5835 {
5836         switch (error) {
5837         case -NFS4ERR_ADMIN_REVOKED:
5838         case -NFS4ERR_BAD_STATEID:
5839                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5840                 if (new_lock_owner != 0 ||
5841                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5842                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5843                 break;
5844         case -NFS4ERR_STALE_STATEID:
5845                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5846         case -NFS4ERR_EXPIRED:
5847                 nfs4_schedule_lease_recovery(server->nfs_client);
5848         };
5849 }
5850
5851 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5852 {
5853         struct nfs4_lockdata *data;
5854         struct rpc_task *task;
5855         struct rpc_message msg = {
5856                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5857                 .rpc_cred = state->owner->so_cred,
5858         };
5859         struct rpc_task_setup task_setup_data = {
5860                 .rpc_client = NFS_CLIENT(state->inode),
5861                 .rpc_message = &msg,
5862                 .callback_ops = &nfs4_lock_ops,
5863                 .workqueue = nfsiod_workqueue,
5864                 .flags = RPC_TASK_ASYNC,
5865         };
5866         int ret;
5867
5868         dprintk("%s: begin!\n", __func__);
5869         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5870                         fl->fl_u.nfs4_fl.owner,
5871                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5872         if (data == NULL)
5873                 return -ENOMEM;
5874         if (IS_SETLKW(cmd))
5875                 data->arg.block = 1;
5876         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5877         msg.rpc_argp = &data->arg;
5878         msg.rpc_resp = &data->res;
5879         task_setup_data.callback_data = data;
5880         if (recovery_type > NFS_LOCK_NEW) {
5881                 if (recovery_type == NFS_LOCK_RECLAIM)
5882                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5883                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5884         } else
5885                 data->arg.new_lock = 1;
5886         task = rpc_run_task(&task_setup_data);
5887         if (IS_ERR(task))
5888                 return PTR_ERR(task);
5889         ret = nfs4_wait_for_completion_rpc_task(task);
5890         if (ret == 0) {
5891                 ret = data->rpc_status;
5892                 if (ret)
5893                         nfs4_handle_setlk_error(data->server, data->lsp,
5894                                         data->arg.new_lock_owner, ret);
5895         } else
5896                 data->cancelled = 1;
5897         rpc_put_task(task);
5898         dprintk("%s: done, ret = %d!\n", __func__, ret);
5899         return ret;
5900 }
5901
5902 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5903 {
5904         struct nfs_server *server = NFS_SERVER(state->inode);
5905         struct nfs4_exception exception = {
5906                 .inode = state->inode,
5907         };
5908         int err;
5909
5910         do {
5911                 /* Cache the lock if possible... */
5912                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5913                         return 0;
5914                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5915                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5916                 if (err != -NFS4ERR_DELAY)
5917                         break;
5918                 nfs4_handle_exception(server, err, &exception);
5919         } while (exception.retry);
5920         return err;
5921 }
5922
5923 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5924 {
5925         struct nfs_server *server = NFS_SERVER(state->inode);
5926         struct nfs4_exception exception = {
5927                 .inode = state->inode,
5928         };
5929         int err;
5930
5931         err = nfs4_set_lock_state(state, request);
5932         if (err != 0)
5933                 return err;
5934         if (!recover_lost_locks) {
5935                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5936                 return 0;
5937         }
5938         do {
5939                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5940                         return 0;
5941                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5942                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5943                 switch (err) {
5944                 default:
5945                         goto out;
5946                 case -NFS4ERR_GRACE:
5947                 case -NFS4ERR_DELAY:
5948                         nfs4_handle_exception(server, err, &exception);
5949                         err = 0;
5950                 }
5951         } while (exception.retry);
5952 out:
5953         return err;
5954 }
5955
5956 #if defined(CONFIG_NFS_V4_1)
5957 /**
5958  * nfs41_check_expired_locks - possibly free a lock stateid
5959  *
5960  * @state: NFSv4 state for an inode
5961  *
5962  * Returns NFS_OK if recovery for this stateid is now finished.
5963  * Otherwise a negative NFS4ERR value is returned.
5964  */
5965 static int nfs41_check_expired_locks(struct nfs4_state *state)
5966 {
5967         int status, ret = -NFS4ERR_BAD_STATEID;
5968         struct nfs4_lock_state *lsp;
5969         struct nfs_server *server = NFS_SERVER(state->inode);
5970
5971         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5972                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5973                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5974
5975                         status = nfs41_test_stateid(server,
5976                                         &lsp->ls_stateid,
5977                                         cred);
5978                         trace_nfs4_test_lock_stateid(state, lsp, status);
5979                         if (status != NFS_OK) {
5980                                 /* Free the stateid unless the server
5981                                  * informs us the stateid is unrecognized. */
5982                                 if (status != -NFS4ERR_BAD_STATEID)
5983                                         nfs41_free_stateid(server,
5984                                                         &lsp->ls_stateid,
5985                                                         cred);
5986                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5987                                 ret = status;
5988                         }
5989                 }
5990         };
5991
5992         return ret;
5993 }
5994
5995 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5996 {
5997         int status = NFS_OK;
5998
5999         if (test_bit(LK_STATE_IN_USE, &state->flags))
6000                 status = nfs41_check_expired_locks(state);
6001         if (status != NFS_OK)
6002                 status = nfs4_lock_expired(state, request);
6003         return status;
6004 }
6005 #endif
6006
6007 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6008 {
6009         struct nfs_inode *nfsi = NFS_I(state->inode);
6010         unsigned char fl_flags = request->fl_flags;
6011         int status = -ENOLCK;
6012
6013         if ((fl_flags & FL_POSIX) &&
6014                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6015                 goto out;
6016         /* Is this a delegated open? */
6017         status = nfs4_set_lock_state(state, request);
6018         if (status != 0)
6019                 goto out;
6020         request->fl_flags |= FL_ACCESS;
6021         status = do_vfs_lock(state->inode, request);
6022         if (status < 0)
6023                 goto out;
6024         down_read(&nfsi->rwsem);
6025         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6026                 /* Yes: cache locks! */
6027                 /* ...but avoid races with delegation recall... */
6028                 request->fl_flags = fl_flags & ~FL_SLEEP;
6029                 status = do_vfs_lock(state->inode, request);
6030                 up_read(&nfsi->rwsem);
6031                 goto out;
6032         }
6033         up_read(&nfsi->rwsem);
6034         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6035 out:
6036         request->fl_flags = fl_flags;
6037         return status;
6038 }
6039
6040 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6041 {
6042         struct nfs4_exception exception = {
6043                 .state = state,
6044                 .inode = state->inode,
6045         };
6046         int err;
6047
6048         do {
6049                 err = _nfs4_proc_setlk(state, cmd, request);
6050                 trace_nfs4_set_lock(request, state, cmd, err);
6051                 if (err == -NFS4ERR_DENIED)
6052                         err = -EAGAIN;
6053                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6054                                 err, &exception);
6055         } while (exception.retry);
6056         return err;
6057 }
6058
6059 static int
6060 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6061 {
6062         struct nfs_open_context *ctx;
6063         struct nfs4_state *state;
6064         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6065         int status;
6066
6067         /* verify open state */
6068         ctx = nfs_file_open_context(filp);
6069         state = ctx->state;
6070
6071         if (request->fl_start < 0 || request->fl_end < 0)
6072                 return -EINVAL;
6073
6074         if (IS_GETLK(cmd)) {
6075                 if (state != NULL)
6076                         return nfs4_proc_getlk(state, F_GETLK, request);
6077                 return 0;
6078         }
6079
6080         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6081                 return -EINVAL;
6082
6083         if (request->fl_type == F_UNLCK) {
6084                 if (state != NULL)
6085                         return nfs4_proc_unlck(state, cmd, request);
6086                 return 0;
6087         }
6088
6089         if (state == NULL)
6090                 return -ENOLCK;
6091         /*
6092          * Don't rely on the VFS having checked the file open mode,
6093          * since it won't do this for flock() locks.
6094          */
6095         switch (request->fl_type) {
6096         case F_RDLCK:
6097                 if (!(filp->f_mode & FMODE_READ))
6098                         return -EBADF;
6099                 break;
6100         case F_WRLCK:
6101                 if (!(filp->f_mode & FMODE_WRITE))
6102                         return -EBADF;
6103         }
6104
6105         do {
6106                 status = nfs4_proc_setlk(state, cmd, request);
6107                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6108                         break;
6109                 timeout = nfs4_set_lock_task_retry(timeout);
6110                 status = -ERESTARTSYS;
6111                 if (signalled())
6112                         break;
6113         } while(status < 0);
6114         return status;
6115 }
6116
6117 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6118 {
6119         struct nfs_server *server = NFS_SERVER(state->inode);
6120         int err;
6121
6122         err = nfs4_set_lock_state(state, fl);
6123         if (err != 0)
6124                 return err;
6125         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6126         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6127 }
6128
6129 struct nfs_release_lockowner_data {
6130         struct nfs4_lock_state *lsp;
6131         struct nfs_server *server;
6132         struct nfs_release_lockowner_args args;
6133         struct nfs_release_lockowner_res res;
6134         unsigned long timestamp;
6135 };
6136
6137 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6138 {
6139         struct nfs_release_lockowner_data *data = calldata;
6140         struct nfs_server *server = data->server;
6141         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6142                              &data->args.seq_args, &data->res.seq_res, task);
6143         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6144         data->timestamp = jiffies;
6145 }
6146
6147 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6148 {
6149         struct nfs_release_lockowner_data *data = calldata;
6150         struct nfs_server *server = data->server;
6151
6152         nfs40_sequence_done(task, &data->res.seq_res);
6153
6154         switch (task->tk_status) {
6155         case 0:
6156                 renew_lease(server, data->timestamp);
6157                 break;
6158         case -NFS4ERR_STALE_CLIENTID:
6159         case -NFS4ERR_EXPIRED:
6160                 nfs4_schedule_lease_recovery(server->nfs_client);
6161                 break;
6162         case -NFS4ERR_LEASE_MOVED:
6163         case -NFS4ERR_DELAY:
6164                 if (nfs4_async_handle_error(task, server,
6165                                             NULL, NULL) == -EAGAIN)
6166                         rpc_restart_call_prepare(task);
6167         }
6168 }
6169
6170 static void nfs4_release_lockowner_release(void *calldata)
6171 {
6172         struct nfs_release_lockowner_data *data = calldata;
6173         nfs4_free_lock_state(data->server, data->lsp);
6174         kfree(calldata);
6175 }
6176
6177 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6178         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6179         .rpc_call_done = nfs4_release_lockowner_done,
6180         .rpc_release = nfs4_release_lockowner_release,
6181 };
6182
6183 static void
6184 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6185 {
6186         struct nfs_release_lockowner_data *data;
6187         struct rpc_message msg = {
6188                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6189         };
6190
6191         if (server->nfs_client->cl_mvops->minor_version != 0)
6192                 return;
6193
6194         data = kmalloc(sizeof(*data), GFP_NOFS);
6195         if (!data)
6196                 return;
6197         data->lsp = lsp;
6198         data->server = server;
6199         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6200         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6201         data->args.lock_owner.s_dev = server->s_dev;
6202
6203         msg.rpc_argp = &data->args;
6204         msg.rpc_resp = &data->res;
6205         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6206         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6207 }
6208
6209 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6210
6211 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6212                                    const void *buf, size_t buflen,
6213                                    int flags, int type)
6214 {
6215         if (strcmp(key, "") != 0)
6216                 return -EINVAL;
6217
6218         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6219 }
6220
6221 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6222                                    void *buf, size_t buflen, int type)
6223 {
6224         if (strcmp(key, "") != 0)
6225                 return -EINVAL;
6226
6227         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6228 }
6229
6230 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6231                                        size_t list_len, const char *name,
6232                                        size_t name_len, int type)
6233 {
6234         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6235
6236         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6237                 return 0;
6238
6239         if (list && len <= list_len)
6240                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6241         return len;
6242 }
6243
6244 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6245 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6246 {
6247         return server->caps & NFS_CAP_SECURITY_LABEL;
6248 }
6249
6250 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6251                                    const void *buf, size_t buflen,
6252                                    int flags, int type)
6253 {
6254         if (security_ismaclabel(key))
6255                 return nfs4_set_security_label(dentry, buf, buflen);
6256
6257         return -EOPNOTSUPP;
6258 }
6259
6260 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6261                                    void *buf, size_t buflen, int type)
6262 {
6263         if (security_ismaclabel(key))
6264                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6265         return -EOPNOTSUPP;
6266 }
6267
6268 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6269                                        size_t list_len, const char *name,
6270                                        size_t name_len, int type)
6271 {
6272         size_t len = 0;
6273
6274         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6275                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6276                 if (list && len <= list_len)
6277                         security_inode_listsecurity(d_inode(dentry), list, len);
6278         }
6279         return len;
6280 }
6281
6282 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6283         .prefix = XATTR_SECURITY_PREFIX,
6284         .list   = nfs4_xattr_list_nfs4_label,
6285         .get    = nfs4_xattr_get_nfs4_label,
6286         .set    = nfs4_xattr_set_nfs4_label,
6287 };
6288 #endif
6289
6290
6291 /*
6292  * nfs_fhget will use either the mounted_on_fileid or the fileid
6293  */
6294 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6295 {
6296         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6297                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6298               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6299               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6300                 return;
6301
6302         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6303                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6304         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6305         fattr->nlink = 2;
6306 }
6307
6308 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6309                                    const struct qstr *name,
6310                                    struct nfs4_fs_locations *fs_locations,
6311                                    struct page *page)
6312 {
6313         struct nfs_server *server = NFS_SERVER(dir);
6314         u32 bitmask[3] = {
6315                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6316         };
6317         struct nfs4_fs_locations_arg args = {
6318                 .dir_fh = NFS_FH(dir),
6319                 .name = name,
6320                 .page = page,
6321                 .bitmask = bitmask,
6322         };
6323         struct nfs4_fs_locations_res res = {
6324                 .fs_locations = fs_locations,
6325         };
6326         struct rpc_message msg = {
6327                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6328                 .rpc_argp = &args,
6329                 .rpc_resp = &res,
6330         };
6331         int status;
6332
6333         dprintk("%s: start\n", __func__);
6334
6335         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6336          * is not supported */
6337         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6338                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6339         else
6340                 bitmask[0] |= FATTR4_WORD0_FILEID;
6341
6342         nfs_fattr_init(&fs_locations->fattr);
6343         fs_locations->server = server;
6344         fs_locations->nlocations = 0;
6345         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6346         dprintk("%s: returned status = %d\n", __func__, status);
6347         return status;
6348 }
6349
6350 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6351                            const struct qstr *name,
6352                            struct nfs4_fs_locations *fs_locations,
6353                            struct page *page)
6354 {
6355         struct nfs4_exception exception = { };
6356         int err;
6357         do {
6358                 err = _nfs4_proc_fs_locations(client, dir, name,
6359                                 fs_locations, page);
6360                 trace_nfs4_get_fs_locations(dir, name, err);
6361                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6362                                 &exception);
6363         } while (exception.retry);
6364         return err;
6365 }
6366
6367 /*
6368  * This operation also signals the server that this client is
6369  * performing migration recovery.  The server can stop returning
6370  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6371  * appended to this compound to identify the client ID which is
6372  * performing recovery.
6373  */
6374 static int _nfs40_proc_get_locations(struct inode *inode,
6375                                      struct nfs4_fs_locations *locations,
6376                                      struct page *page, struct rpc_cred *cred)
6377 {
6378         struct nfs_server *server = NFS_SERVER(inode);
6379         struct rpc_clnt *clnt = server->client;
6380         u32 bitmask[2] = {
6381                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6382         };
6383         struct nfs4_fs_locations_arg args = {
6384                 .clientid       = server->nfs_client->cl_clientid,
6385                 .fh             = NFS_FH(inode),
6386                 .page           = page,
6387                 .bitmask        = bitmask,
6388                 .migration      = 1,            /* skip LOOKUP */
6389                 .renew          = 1,            /* append RENEW */
6390         };
6391         struct nfs4_fs_locations_res res = {
6392                 .fs_locations   = locations,
6393                 .migration      = 1,
6394                 .renew          = 1,
6395         };
6396         struct rpc_message msg = {
6397                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6398                 .rpc_argp       = &args,
6399                 .rpc_resp       = &res,
6400                 .rpc_cred       = cred,
6401         };
6402         unsigned long now = jiffies;
6403         int status;
6404
6405         nfs_fattr_init(&locations->fattr);
6406         locations->server = server;
6407         locations->nlocations = 0;
6408
6409         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6410         nfs4_set_sequence_privileged(&args.seq_args);
6411         status = nfs4_call_sync_sequence(clnt, server, &msg,
6412                                         &args.seq_args, &res.seq_res);
6413         if (status)
6414                 return status;
6415
6416         renew_lease(server, now);
6417         return 0;
6418 }
6419
6420 #ifdef CONFIG_NFS_V4_1
6421
6422 /*
6423  * This operation also signals the server that this client is
6424  * performing migration recovery.  The server can stop asserting
6425  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6426  * performing this operation is identified in the SEQUENCE
6427  * operation in this compound.
6428  *
6429  * When the client supports GETATTR(fs_locations_info), it can
6430  * be plumbed in here.
6431  */
6432 static int _nfs41_proc_get_locations(struct inode *inode,
6433                                      struct nfs4_fs_locations *locations,
6434                                      struct page *page, struct rpc_cred *cred)
6435 {
6436         struct nfs_server *server = NFS_SERVER(inode);
6437         struct rpc_clnt *clnt = server->client;
6438         u32 bitmask[2] = {
6439                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6440         };
6441         struct nfs4_fs_locations_arg args = {
6442                 .fh             = NFS_FH(inode),
6443                 .page           = page,
6444                 .bitmask        = bitmask,
6445                 .migration      = 1,            /* skip LOOKUP */
6446         };
6447         struct nfs4_fs_locations_res res = {
6448                 .fs_locations   = locations,
6449                 .migration      = 1,
6450         };
6451         struct rpc_message msg = {
6452                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6453                 .rpc_argp       = &args,
6454                 .rpc_resp       = &res,
6455                 .rpc_cred       = cred,
6456         };
6457         int status;
6458
6459         nfs_fattr_init(&locations->fattr);
6460         locations->server = server;
6461         locations->nlocations = 0;
6462
6463         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6464         nfs4_set_sequence_privileged(&args.seq_args);
6465         status = nfs4_call_sync_sequence(clnt, server, &msg,
6466                                         &args.seq_args, &res.seq_res);
6467         if (status == NFS4_OK &&
6468             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6469                 status = -NFS4ERR_LEASE_MOVED;
6470         return status;
6471 }
6472
6473 #endif  /* CONFIG_NFS_V4_1 */
6474
6475 /**
6476  * nfs4_proc_get_locations - discover locations for a migrated FSID
6477  * @inode: inode on FSID that is migrating
6478  * @locations: result of query
6479  * @page: buffer
6480  * @cred: credential to use for this operation
6481  *
6482  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6483  * operation failed, or a negative errno if a local error occurred.
6484  *
6485  * On success, "locations" is filled in, but if the server has
6486  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6487  * asserted.
6488  *
6489  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6490  * from this client that require migration recovery.
6491  */
6492 int nfs4_proc_get_locations(struct inode *inode,
6493                             struct nfs4_fs_locations *locations,
6494                             struct page *page, struct rpc_cred *cred)
6495 {
6496         struct nfs_server *server = NFS_SERVER(inode);
6497         struct nfs_client *clp = server->nfs_client;
6498         const struct nfs4_mig_recovery_ops *ops =
6499                                         clp->cl_mvops->mig_recovery_ops;
6500         struct nfs4_exception exception = { };
6501         int status;
6502
6503         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6504                 (unsigned long long)server->fsid.major,
6505                 (unsigned long long)server->fsid.minor,
6506                 clp->cl_hostname);
6507         nfs_display_fhandle(NFS_FH(inode), __func__);
6508
6509         do {
6510                 status = ops->get_locations(inode, locations, page, cred);
6511                 if (status != -NFS4ERR_DELAY)
6512                         break;
6513                 nfs4_handle_exception(server, status, &exception);
6514         } while (exception.retry);
6515         return status;
6516 }
6517
6518 /*
6519  * This operation also signals the server that this client is
6520  * performing "lease moved" recovery.  The server can stop
6521  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6522  * is appended to this compound to identify the client ID which is
6523  * performing recovery.
6524  */
6525 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6526 {
6527         struct nfs_server *server = NFS_SERVER(inode);
6528         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6529         struct rpc_clnt *clnt = server->client;
6530         struct nfs4_fsid_present_arg args = {
6531                 .fh             = NFS_FH(inode),
6532                 .clientid       = clp->cl_clientid,
6533                 .renew          = 1,            /* append RENEW */
6534         };
6535         struct nfs4_fsid_present_res res = {
6536                 .renew          = 1,
6537         };
6538         struct rpc_message msg = {
6539                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6540                 .rpc_argp       = &args,
6541                 .rpc_resp       = &res,
6542                 .rpc_cred       = cred,
6543         };
6544         unsigned long now = jiffies;
6545         int status;
6546
6547         res.fh = nfs_alloc_fhandle();
6548         if (res.fh == NULL)
6549                 return -ENOMEM;
6550
6551         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6552         nfs4_set_sequence_privileged(&args.seq_args);
6553         status = nfs4_call_sync_sequence(clnt, server, &msg,
6554                                                 &args.seq_args, &res.seq_res);
6555         nfs_free_fhandle(res.fh);
6556         if (status)
6557                 return status;
6558
6559         do_renew_lease(clp, now);
6560         return 0;
6561 }
6562
6563 #ifdef CONFIG_NFS_V4_1
6564
6565 /*
6566  * This operation also signals the server that this client is
6567  * performing "lease moved" recovery.  The server can stop asserting
6568  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6569  * this operation is identified in the SEQUENCE operation in this
6570  * compound.
6571  */
6572 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6573 {
6574         struct nfs_server *server = NFS_SERVER(inode);
6575         struct rpc_clnt *clnt = server->client;
6576         struct nfs4_fsid_present_arg args = {
6577                 .fh             = NFS_FH(inode),
6578         };
6579         struct nfs4_fsid_present_res res = {
6580         };
6581         struct rpc_message msg = {
6582                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6583                 .rpc_argp       = &args,
6584                 .rpc_resp       = &res,
6585                 .rpc_cred       = cred,
6586         };
6587         int status;
6588
6589         res.fh = nfs_alloc_fhandle();
6590         if (res.fh == NULL)
6591                 return -ENOMEM;
6592
6593         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6594         nfs4_set_sequence_privileged(&args.seq_args);
6595         status = nfs4_call_sync_sequence(clnt, server, &msg,
6596                                                 &args.seq_args, &res.seq_res);
6597         nfs_free_fhandle(res.fh);
6598         if (status == NFS4_OK &&
6599             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6600                 status = -NFS4ERR_LEASE_MOVED;
6601         return status;
6602 }
6603
6604 #endif  /* CONFIG_NFS_V4_1 */
6605
6606 /**
6607  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6608  * @inode: inode on FSID to check
6609  * @cred: credential to use for this operation
6610  *
6611  * Server indicates whether the FSID is present, moved, or not
6612  * recognized.  This operation is necessary to clear a LEASE_MOVED
6613  * condition for this client ID.
6614  *
6615  * Returns NFS4_OK if the FSID is present on this server,
6616  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6617  *  NFS4ERR code if some error occurred on the server, or a
6618  *  negative errno if a local failure occurred.
6619  */
6620 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6621 {
6622         struct nfs_server *server = NFS_SERVER(inode);
6623         struct nfs_client *clp = server->nfs_client;
6624         const struct nfs4_mig_recovery_ops *ops =
6625                                         clp->cl_mvops->mig_recovery_ops;
6626         struct nfs4_exception exception = { };
6627         int status;
6628
6629         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6630                 (unsigned long long)server->fsid.major,
6631                 (unsigned long long)server->fsid.minor,
6632                 clp->cl_hostname);
6633         nfs_display_fhandle(NFS_FH(inode), __func__);
6634
6635         do {
6636                 status = ops->fsid_present(inode, cred);
6637                 if (status != -NFS4ERR_DELAY)
6638                         break;
6639                 nfs4_handle_exception(server, status, &exception);
6640         } while (exception.retry);
6641         return status;
6642 }
6643
6644 /**
6645  * If 'use_integrity' is true and the state managment nfs_client
6646  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6647  * and the machine credential as per RFC3530bis and RFC5661 Security
6648  * Considerations sections. Otherwise, just use the user cred with the
6649  * filesystem's rpc_client.
6650  */
6651 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6652 {
6653         int status;
6654         struct nfs4_secinfo_arg args = {
6655                 .dir_fh = NFS_FH(dir),
6656                 .name   = name,
6657         };
6658         struct nfs4_secinfo_res res = {
6659                 .flavors     = flavors,
6660         };
6661         struct rpc_message msg = {
6662                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6663                 .rpc_argp = &args,
6664                 .rpc_resp = &res,
6665         };
6666         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6667         struct rpc_cred *cred = NULL;
6668
6669         if (use_integrity) {
6670                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6671                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6672                 msg.rpc_cred = cred;
6673         }
6674
6675         dprintk("NFS call  secinfo %s\n", name->name);
6676
6677         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6678                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6679
6680         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6681                                 &res.seq_res, 0);
6682         dprintk("NFS reply  secinfo: %d\n", status);
6683
6684         if (cred)
6685                 put_rpccred(cred);
6686
6687         return status;
6688 }
6689
6690 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6691                       struct nfs4_secinfo_flavors *flavors)
6692 {
6693         struct nfs4_exception exception = { };
6694         int err;
6695         do {
6696                 err = -NFS4ERR_WRONGSEC;
6697
6698                 /* try to use integrity protection with machine cred */
6699                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6700                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6701
6702                 /*
6703                  * if unable to use integrity protection, or SECINFO with
6704                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6705                  * disallowed by spec, but exists in deployed servers) use
6706                  * the current filesystem's rpc_client and the user cred.
6707                  */
6708                 if (err == -NFS4ERR_WRONGSEC)
6709                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6710
6711                 trace_nfs4_secinfo(dir, name, err);
6712                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6713                                 &exception);
6714         } while (exception.retry);
6715         return err;
6716 }
6717
6718 #ifdef CONFIG_NFS_V4_1
6719 /*
6720  * Check the exchange flags returned by the server for invalid flags, having
6721  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6722  * DS flags set.
6723  */
6724 static int nfs4_check_cl_exchange_flags(u32 flags)
6725 {
6726         if (flags & ~EXCHGID4_FLAG_MASK_R)
6727                 goto out_inval;
6728         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6729             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6730                 goto out_inval;
6731         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6732                 goto out_inval;
6733         return NFS_OK;
6734 out_inval:
6735         return -NFS4ERR_INVAL;
6736 }
6737
6738 static bool
6739 nfs41_same_server_scope(struct nfs41_server_scope *a,
6740                         struct nfs41_server_scope *b)
6741 {
6742         if (a->server_scope_sz == b->server_scope_sz &&
6743             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6744                 return true;
6745
6746         return false;
6747 }
6748
6749 /*
6750  * nfs4_proc_bind_conn_to_session()
6751  *
6752  * The 4.1 client currently uses the same TCP connection for the
6753  * fore and backchannel.
6754  */
6755 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6756 {
6757         int status;
6758         struct nfs41_bind_conn_to_session_args args = {
6759                 .client = clp,
6760                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6761         };
6762         struct nfs41_bind_conn_to_session_res res;
6763         struct rpc_message msg = {
6764                 .rpc_proc =
6765                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6766                 .rpc_argp = &args,
6767                 .rpc_resp = &res,
6768                 .rpc_cred = cred,
6769         };
6770
6771         dprintk("--> %s\n", __func__);
6772
6773         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6774         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6775                 args.dir = NFS4_CDFC4_FORE;
6776
6777         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6778         trace_nfs4_bind_conn_to_session(clp, status);
6779         if (status == 0) {
6780                 if (memcmp(res.sessionid.data,
6781                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6782                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6783                         status = -EIO;
6784                         goto out;
6785                 }
6786                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6787                         dprintk("NFS: %s: Unexpected direction from server\n",
6788                                 __func__);
6789                         status = -EIO;
6790                         goto out;
6791                 }
6792                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6793                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6794                                 __func__);
6795                         status = -EIO;
6796                         goto out;
6797                 }
6798         }
6799 out:
6800         dprintk("<-- %s status= %d\n", __func__, status);
6801         return status;
6802 }
6803
6804 /*
6805  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6806  * and operations we'd like to see to enable certain features in the allow map
6807  */
6808 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6809         .how = SP4_MACH_CRED,
6810         .enforce.u.words = {
6811                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6812                       1 << (OP_EXCHANGE_ID - 32) |
6813                       1 << (OP_CREATE_SESSION - 32) |
6814                       1 << (OP_DESTROY_SESSION - 32) |
6815                       1 << (OP_DESTROY_CLIENTID - 32)
6816         },
6817         .allow.u.words = {
6818                 [0] = 1 << (OP_CLOSE) |
6819                       1 << (OP_LOCKU) |
6820                       1 << (OP_COMMIT),
6821                 [1] = 1 << (OP_SECINFO - 32) |
6822                       1 << (OP_SECINFO_NO_NAME - 32) |
6823                       1 << (OP_TEST_STATEID - 32) |
6824                       1 << (OP_FREE_STATEID - 32) |
6825                       1 << (OP_WRITE - 32)
6826         }
6827 };
6828
6829 /*
6830  * Select the state protection mode for client `clp' given the server results
6831  * from exchange_id in `sp'.
6832  *
6833  * Returns 0 on success, negative errno otherwise.
6834  */
6835 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6836                                  struct nfs41_state_protection *sp)
6837 {
6838         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6839                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6840                       1 << (OP_EXCHANGE_ID - 32) |
6841                       1 << (OP_CREATE_SESSION - 32) |
6842                       1 << (OP_DESTROY_SESSION - 32) |
6843                       1 << (OP_DESTROY_CLIENTID - 32)
6844         };
6845         unsigned int i;
6846
6847         if (sp->how == SP4_MACH_CRED) {
6848                 /* Print state protect result */
6849                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6850                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6851                         if (test_bit(i, sp->enforce.u.longs))
6852                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6853                         if (test_bit(i, sp->allow.u.longs))
6854                                 dfprintk(MOUNT, "  allow op %d\n", i);
6855                 }
6856
6857                 /* make sure nothing is on enforce list that isn't supported */
6858                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6859                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6860                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6861                                 return -EINVAL;
6862                         }
6863                 }
6864
6865                 /*
6866                  * Minimal mode - state operations are allowed to use machine
6867                  * credential.  Note this already happens by default, so the
6868                  * client doesn't have to do anything more than the negotiation.
6869                  *
6870                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6871                  *       we're already using the machine cred for exchange_id
6872                  *       and will never use a different cred.
6873                  */
6874                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6875                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6876                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6877                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6878                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6879                         dfprintk(MOUNT, "  minimal mode enabled\n");
6880                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6881                 } else {
6882                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6883                         return -EINVAL;
6884                 }
6885
6886                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6887                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6888                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6889                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6890                 }
6891
6892                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6893                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6894                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6895                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6896                 }
6897
6898                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6899                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6900                         dfprintk(MOUNT, "  stateid mode enabled\n");
6901                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6902                 }
6903
6904                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6905                         dfprintk(MOUNT, "  write mode enabled\n");
6906                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6907                 }
6908
6909                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6910                         dfprintk(MOUNT, "  commit mode enabled\n");
6911                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6912                 }
6913         }
6914
6915         return 0;
6916 }
6917
6918 /*
6919  * _nfs4_proc_exchange_id()
6920  *
6921  * Wrapper for EXCHANGE_ID operation.
6922  */
6923 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6924         u32 sp4_how)
6925 {
6926         nfs4_verifier verifier;
6927         struct nfs41_exchange_id_args args = {
6928                 .verifier = &verifier,
6929                 .client = clp,
6930 #ifdef CONFIG_NFS_V4_1_MIGRATION
6931                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6932                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6933                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6934 #else
6935                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6936                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6937 #endif
6938         };
6939         struct nfs41_exchange_id_res res = {
6940                 0
6941         };
6942         int status;
6943         struct rpc_message msg = {
6944                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6945                 .rpc_argp = &args,
6946                 .rpc_resp = &res,
6947                 .rpc_cred = cred,
6948         };
6949
6950         nfs4_init_boot_verifier(clp, &verifier);
6951
6952         status = nfs4_init_uniform_client_string(clp);
6953         if (status)
6954                 goto out;
6955
6956         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
6957                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6958                 clp->cl_owner_id);
6959
6960         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6961                                         GFP_NOFS);
6962         if (unlikely(res.server_owner == NULL)) {
6963                 status = -ENOMEM;
6964                 goto out;
6965         }
6966
6967         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6968                                         GFP_NOFS);
6969         if (unlikely(res.server_scope == NULL)) {
6970                 status = -ENOMEM;
6971                 goto out_server_owner;
6972         }
6973
6974         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6975         if (unlikely(res.impl_id == NULL)) {
6976                 status = -ENOMEM;
6977                 goto out_server_scope;
6978         }
6979
6980         switch (sp4_how) {
6981         case SP4_NONE:
6982                 args.state_protect.how = SP4_NONE;
6983                 break;
6984
6985         case SP4_MACH_CRED:
6986                 args.state_protect = nfs4_sp4_mach_cred_request;
6987                 break;
6988
6989         default:
6990                 /* unsupported! */
6991                 WARN_ON_ONCE(1);
6992                 status = -EINVAL;
6993                 goto out_impl_id;
6994         }
6995
6996         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6997         trace_nfs4_exchange_id(clp, status);
6998         if (status == 0)
6999                 status = nfs4_check_cl_exchange_flags(res.flags);
7000
7001         if (status == 0)
7002                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7003
7004         if (status == 0) {
7005                 clp->cl_clientid = res.clientid;
7006                 clp->cl_exchange_flags = res.flags;
7007                 /* Client ID is not confirmed */
7008                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7009                         clear_bit(NFS4_SESSION_ESTABLISHED,
7010                                         &clp->cl_session->session_state);
7011                         clp->cl_seqid = res.seqid;
7012                 }
7013
7014                 kfree(clp->cl_serverowner);
7015                 clp->cl_serverowner = res.server_owner;
7016                 res.server_owner = NULL;
7017
7018                 /* use the most recent implementation id */
7019                 kfree(clp->cl_implid);
7020                 clp->cl_implid = res.impl_id;
7021                 res.impl_id = NULL;
7022
7023                 if (clp->cl_serverscope != NULL &&
7024                     !nfs41_same_server_scope(clp->cl_serverscope,
7025                                              res.server_scope)) {
7026                         dprintk("%s: server_scope mismatch detected\n",
7027                                 __func__);
7028                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7029                         kfree(clp->cl_serverscope);
7030                         clp->cl_serverscope = NULL;
7031                 }
7032
7033                 if (clp->cl_serverscope == NULL) {
7034                         clp->cl_serverscope = res.server_scope;
7035                         res.server_scope = NULL;
7036                 }
7037         }
7038
7039 out_impl_id:
7040         kfree(res.impl_id);
7041 out_server_scope:
7042         kfree(res.server_scope);
7043 out_server_owner:
7044         kfree(res.server_owner);
7045 out:
7046         if (clp->cl_implid != NULL)
7047                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7048                         "domain: %s, name: %s, date: %llu,%u\n",
7049                         clp->cl_implid->domain, clp->cl_implid->name,
7050                         clp->cl_implid->date.seconds,
7051                         clp->cl_implid->date.nseconds);
7052         dprintk("NFS reply exchange_id: %d\n", status);
7053         return status;
7054 }
7055
7056 /*
7057  * nfs4_proc_exchange_id()
7058  *
7059  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7060  *
7061  * Since the clientid has expired, all compounds using sessions
7062  * associated with the stale clientid will be returning
7063  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7064  * be in some phase of session reset.
7065  *
7066  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7067  */
7068 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7069 {
7070         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7071         int status;
7072
7073         /* try SP4_MACH_CRED if krb5i/p */
7074         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7075             authflavor == RPC_AUTH_GSS_KRB5P) {
7076                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7077                 if (!status)
7078                         return 0;
7079         }
7080
7081         /* try SP4_NONE */
7082         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7083 }
7084
7085 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7086                 struct rpc_cred *cred)
7087 {
7088         struct rpc_message msg = {
7089                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7090                 .rpc_argp = clp,
7091                 .rpc_cred = cred,
7092         };
7093         int status;
7094
7095         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7096         trace_nfs4_destroy_clientid(clp, status);
7097         if (status)
7098                 dprintk("NFS: Got error %d from the server %s on "
7099                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7100         return status;
7101 }
7102
7103 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7104                 struct rpc_cred *cred)
7105 {
7106         unsigned int loop;
7107         int ret;
7108
7109         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7110                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7111                 switch (ret) {
7112                 case -NFS4ERR_DELAY:
7113                 case -NFS4ERR_CLIENTID_BUSY:
7114                         ssleep(1);
7115                         break;
7116                 default:
7117                         return ret;
7118                 }
7119         }
7120         return 0;
7121 }
7122
7123 int nfs4_destroy_clientid(struct nfs_client *clp)
7124 {
7125         struct rpc_cred *cred;
7126         int ret = 0;
7127
7128         if (clp->cl_mvops->minor_version < 1)
7129                 goto out;
7130         if (clp->cl_exchange_flags == 0)
7131                 goto out;
7132         if (clp->cl_preserve_clid)
7133                 goto out;
7134         cred = nfs4_get_clid_cred(clp);
7135         ret = nfs4_proc_destroy_clientid(clp, cred);
7136         if (cred)
7137                 put_rpccred(cred);
7138         switch (ret) {
7139         case 0:
7140         case -NFS4ERR_STALE_CLIENTID:
7141                 clp->cl_exchange_flags = 0;
7142         }
7143 out:
7144         return ret;
7145 }
7146
7147 struct nfs4_get_lease_time_data {
7148         struct nfs4_get_lease_time_args *args;
7149         struct nfs4_get_lease_time_res *res;
7150         struct nfs_client *clp;
7151 };
7152
7153 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7154                                         void *calldata)
7155 {
7156         struct nfs4_get_lease_time_data *data =
7157                         (struct nfs4_get_lease_time_data *)calldata;
7158
7159         dprintk("--> %s\n", __func__);
7160         /* just setup sequence, do not trigger session recovery
7161            since we're invoked within one */
7162         nfs41_setup_sequence(data->clp->cl_session,
7163                         &data->args->la_seq_args,
7164                         &data->res->lr_seq_res,
7165                         task);
7166         dprintk("<-- %s\n", __func__);
7167 }
7168
7169 /*
7170  * Called from nfs4_state_manager thread for session setup, so don't recover
7171  * from sequence operation or clientid errors.
7172  */
7173 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7174 {
7175         struct nfs4_get_lease_time_data *data =
7176                         (struct nfs4_get_lease_time_data *)calldata;
7177
7178         dprintk("--> %s\n", __func__);
7179         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7180                 return;
7181         switch (task->tk_status) {
7182         case -NFS4ERR_DELAY:
7183         case -NFS4ERR_GRACE:
7184                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7185                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7186                 task->tk_status = 0;
7187                 /* fall through */
7188         case -NFS4ERR_RETRY_UNCACHED_REP:
7189                 rpc_restart_call_prepare(task);
7190                 return;
7191         }
7192         dprintk("<-- %s\n", __func__);
7193 }
7194
7195 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7196         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7197         .rpc_call_done = nfs4_get_lease_time_done,
7198 };
7199
7200 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7201 {
7202         struct rpc_task *task;
7203         struct nfs4_get_lease_time_args args;
7204         struct nfs4_get_lease_time_res res = {
7205                 .lr_fsinfo = fsinfo,
7206         };
7207         struct nfs4_get_lease_time_data data = {
7208                 .args = &args,
7209                 .res = &res,
7210                 .clp = clp,
7211         };
7212         struct rpc_message msg = {
7213                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7214                 .rpc_argp = &args,
7215                 .rpc_resp = &res,
7216         };
7217         struct rpc_task_setup task_setup = {
7218                 .rpc_client = clp->cl_rpcclient,
7219                 .rpc_message = &msg,
7220                 .callback_ops = &nfs4_get_lease_time_ops,
7221                 .callback_data = &data,
7222                 .flags = RPC_TASK_TIMEOUT,
7223         };
7224         int status;
7225
7226         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7227         nfs4_set_sequence_privileged(&args.la_seq_args);
7228         dprintk("--> %s\n", __func__);
7229         task = rpc_run_task(&task_setup);
7230
7231         if (IS_ERR(task))
7232                 status = PTR_ERR(task);
7233         else {
7234                 status = task->tk_status;
7235                 rpc_put_task(task);
7236         }
7237         dprintk("<-- %s return %d\n", __func__, status);
7238
7239         return status;
7240 }
7241
7242 /*
7243  * Initialize the values to be used by the client in CREATE_SESSION
7244  * If nfs4_init_session set the fore channel request and response sizes,
7245  * use them.
7246  *
7247  * Set the back channel max_resp_sz_cached to zero to force the client to
7248  * always set csa_cachethis to FALSE because the current implementation
7249  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7250  */
7251 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7252 {
7253         unsigned int max_rqst_sz, max_resp_sz;
7254
7255         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7256         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7257
7258         /* Fore channel attributes */
7259         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7260         args->fc_attrs.max_resp_sz = max_resp_sz;
7261         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7262         args->fc_attrs.max_reqs = max_session_slots;
7263
7264         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7265                 "max_ops=%u max_reqs=%u\n",
7266                 __func__,
7267                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7268                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7269
7270         /* Back channel attributes */
7271         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7272         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7273         args->bc_attrs.max_resp_sz_cached = 0;
7274         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7275         args->bc_attrs.max_reqs = 1;
7276
7277         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7278                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7279                 __func__,
7280                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7281                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7282                 args->bc_attrs.max_reqs);
7283 }
7284
7285 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7286                 struct nfs41_create_session_res *res)
7287 {
7288         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7289         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7290
7291         if (rcvd->max_resp_sz > sent->max_resp_sz)
7292                 return -EINVAL;
7293         /*
7294          * Our requested max_ops is the minimum we need; we're not
7295          * prepared to break up compounds into smaller pieces than that.
7296          * So, no point even trying to continue if the server won't
7297          * cooperate:
7298          */
7299         if (rcvd->max_ops < sent->max_ops)
7300                 return -EINVAL;
7301         if (rcvd->max_reqs == 0)
7302                 return -EINVAL;
7303         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7304                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7305         return 0;
7306 }
7307
7308 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7309                 struct nfs41_create_session_res *res)
7310 {
7311         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7312         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7313
7314         if (!(res->flags & SESSION4_BACK_CHAN))
7315                 goto out;
7316         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7317                 return -EINVAL;
7318         if (rcvd->max_resp_sz < sent->max_resp_sz)
7319                 return -EINVAL;
7320         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7321                 return -EINVAL;
7322         /* These would render the backchannel useless: */
7323         if (rcvd->max_ops != sent->max_ops)
7324                 return -EINVAL;
7325         if (rcvd->max_reqs != sent->max_reqs)
7326                 return -EINVAL;
7327 out:
7328         return 0;
7329 }
7330
7331 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7332                                      struct nfs41_create_session_res *res)
7333 {
7334         int ret;
7335
7336         ret = nfs4_verify_fore_channel_attrs(args, res);
7337         if (ret)
7338                 return ret;
7339         return nfs4_verify_back_channel_attrs(args, res);
7340 }
7341
7342 static void nfs4_update_session(struct nfs4_session *session,
7343                 struct nfs41_create_session_res *res)
7344 {
7345         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7346         /* Mark client id and session as being confirmed */
7347         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7348         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7349         session->flags = res->flags;
7350         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7351         if (res->flags & SESSION4_BACK_CHAN)
7352                 memcpy(&session->bc_attrs, &res->bc_attrs,
7353                                 sizeof(session->bc_attrs));
7354 }
7355
7356 static int _nfs4_proc_create_session(struct nfs_client *clp,
7357                 struct rpc_cred *cred)
7358 {
7359         struct nfs4_session *session = clp->cl_session;
7360         struct nfs41_create_session_args args = {
7361                 .client = clp,
7362                 .clientid = clp->cl_clientid,
7363                 .seqid = clp->cl_seqid,
7364                 .cb_program = NFS4_CALLBACK,
7365         };
7366         struct nfs41_create_session_res res;
7367
7368         struct rpc_message msg = {
7369                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7370                 .rpc_argp = &args,
7371                 .rpc_resp = &res,
7372                 .rpc_cred = cred,
7373         };
7374         int status;
7375
7376         nfs4_init_channel_attrs(&args);
7377         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7378
7379         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7380         trace_nfs4_create_session(clp, status);
7381
7382         if (!status) {
7383                 /* Verify the session's negotiated channel_attrs values */
7384                 status = nfs4_verify_channel_attrs(&args, &res);
7385                 /* Increment the clientid slot sequence id */
7386                 if (clp->cl_seqid == res.seqid)
7387                         clp->cl_seqid++;
7388                 if (status)
7389                         goto out;
7390                 nfs4_update_session(session, &res);
7391         }
7392 out:
7393         return status;
7394 }
7395
7396 /*
7397  * Issues a CREATE_SESSION operation to the server.
7398  * It is the responsibility of the caller to verify the session is
7399  * expired before calling this routine.
7400  */
7401 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7402 {
7403         int status;
7404         unsigned *ptr;
7405         struct nfs4_session *session = clp->cl_session;
7406
7407         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7408
7409         status = _nfs4_proc_create_session(clp, cred);
7410         if (status)
7411                 goto out;
7412
7413         /* Init or reset the session slot tables */
7414         status = nfs4_setup_session_slot_tables(session);
7415         dprintk("slot table setup returned %d\n", status);
7416         if (status)
7417                 goto out;
7418
7419         ptr = (unsigned *)&session->sess_id.data[0];
7420         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7421                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7422 out:
7423         dprintk("<-- %s\n", __func__);
7424         return status;
7425 }
7426
7427 /*
7428  * Issue the over-the-wire RPC DESTROY_SESSION.
7429  * The caller must serialize access to this routine.
7430  */
7431 int nfs4_proc_destroy_session(struct nfs4_session *session,
7432                 struct rpc_cred *cred)
7433 {
7434         struct rpc_message msg = {
7435                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7436                 .rpc_argp = session,
7437                 .rpc_cred = cred,
7438         };
7439         int status = 0;
7440
7441         dprintk("--> nfs4_proc_destroy_session\n");
7442
7443         /* session is still being setup */
7444         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7445                 return 0;
7446
7447         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7448         trace_nfs4_destroy_session(session->clp, status);
7449
7450         if (status)
7451                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7452                         "Session has been destroyed regardless...\n", status);
7453
7454         dprintk("<-- nfs4_proc_destroy_session\n");
7455         return status;
7456 }
7457
7458 /*
7459  * Renew the cl_session lease.
7460  */
7461 struct nfs4_sequence_data {
7462         struct nfs_client *clp;
7463         struct nfs4_sequence_args args;
7464         struct nfs4_sequence_res res;
7465 };
7466
7467 static void nfs41_sequence_release(void *data)
7468 {
7469         struct nfs4_sequence_data *calldata = data;
7470         struct nfs_client *clp = calldata->clp;
7471
7472         if (atomic_read(&clp->cl_count) > 1)
7473                 nfs4_schedule_state_renewal(clp);
7474         nfs_put_client(clp);
7475         kfree(calldata);
7476 }
7477
7478 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7479 {
7480         switch(task->tk_status) {
7481         case -NFS4ERR_DELAY:
7482                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7483                 return -EAGAIN;
7484         default:
7485                 nfs4_schedule_lease_recovery(clp);
7486         }
7487         return 0;
7488 }
7489
7490 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7491 {
7492         struct nfs4_sequence_data *calldata = data;
7493         struct nfs_client *clp = calldata->clp;
7494
7495         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7496                 return;
7497
7498         trace_nfs4_sequence(clp, task->tk_status);
7499         if (task->tk_status < 0) {
7500                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7501                 if (atomic_read(&clp->cl_count) == 1)
7502                         goto out;
7503
7504                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7505                         rpc_restart_call_prepare(task);
7506                         return;
7507                 }
7508         }
7509         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7510 out:
7511         dprintk("<-- %s\n", __func__);
7512 }
7513
7514 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7515 {
7516         struct nfs4_sequence_data *calldata = data;
7517         struct nfs_client *clp = calldata->clp;
7518         struct nfs4_sequence_args *args;
7519         struct nfs4_sequence_res *res;
7520
7521         args = task->tk_msg.rpc_argp;
7522         res = task->tk_msg.rpc_resp;
7523
7524         nfs41_setup_sequence(clp->cl_session, args, res, task);
7525 }
7526
7527 static const struct rpc_call_ops nfs41_sequence_ops = {
7528         .rpc_call_done = nfs41_sequence_call_done,
7529         .rpc_call_prepare = nfs41_sequence_prepare,
7530         .rpc_release = nfs41_sequence_release,
7531 };
7532
7533 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7534                 struct rpc_cred *cred,
7535                 bool is_privileged)
7536 {
7537         struct nfs4_sequence_data *calldata;
7538         struct rpc_message msg = {
7539                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7540                 .rpc_cred = cred,
7541         };
7542         struct rpc_task_setup task_setup_data = {
7543                 .rpc_client = clp->cl_rpcclient,
7544                 .rpc_message = &msg,
7545                 .callback_ops = &nfs41_sequence_ops,
7546                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7547         };
7548
7549         if (!atomic_inc_not_zero(&clp->cl_count))
7550                 return ERR_PTR(-EIO);
7551         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7552         if (calldata == NULL) {
7553                 nfs_put_client(clp);
7554                 return ERR_PTR(-ENOMEM);
7555         }
7556         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7557         if (is_privileged)
7558                 nfs4_set_sequence_privileged(&calldata->args);
7559         msg.rpc_argp = &calldata->args;
7560         msg.rpc_resp = &calldata->res;
7561         calldata->clp = clp;
7562         task_setup_data.callback_data = calldata;
7563
7564         return rpc_run_task(&task_setup_data);
7565 }
7566
7567 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7568 {
7569         struct rpc_task *task;
7570         int ret = 0;
7571
7572         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7573                 return -EAGAIN;
7574         task = _nfs41_proc_sequence(clp, cred, false);
7575         if (IS_ERR(task))
7576                 ret = PTR_ERR(task);
7577         else
7578                 rpc_put_task_async(task);
7579         dprintk("<-- %s status=%d\n", __func__, ret);
7580         return ret;
7581 }
7582
7583 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7584 {
7585         struct rpc_task *task;
7586         int ret;
7587
7588         task = _nfs41_proc_sequence(clp, cred, true);
7589         if (IS_ERR(task)) {
7590                 ret = PTR_ERR(task);
7591                 goto out;
7592         }
7593         ret = rpc_wait_for_completion_task(task);
7594         if (!ret)
7595                 ret = task->tk_status;
7596         rpc_put_task(task);
7597 out:
7598         dprintk("<-- %s status=%d\n", __func__, ret);
7599         return ret;
7600 }
7601
7602 struct nfs4_reclaim_complete_data {
7603         struct nfs_client *clp;
7604         struct nfs41_reclaim_complete_args arg;
7605         struct nfs41_reclaim_complete_res res;
7606 };
7607
7608 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7609 {
7610         struct nfs4_reclaim_complete_data *calldata = data;
7611
7612         nfs41_setup_sequence(calldata->clp->cl_session,
7613                         &calldata->arg.seq_args,
7614                         &calldata->res.seq_res,
7615                         task);
7616 }
7617
7618 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7619 {
7620         switch(task->tk_status) {
7621         case 0:
7622         case -NFS4ERR_COMPLETE_ALREADY:
7623         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7624                 break;
7625         case -NFS4ERR_DELAY:
7626                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7627                 /* fall through */
7628         case -NFS4ERR_RETRY_UNCACHED_REP:
7629                 return -EAGAIN;
7630         default:
7631                 nfs4_schedule_lease_recovery(clp);
7632         }
7633         return 0;
7634 }
7635
7636 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7637 {
7638         struct nfs4_reclaim_complete_data *calldata = data;
7639         struct nfs_client *clp = calldata->clp;
7640         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7641
7642         dprintk("--> %s\n", __func__);
7643         if (!nfs41_sequence_done(task, res))
7644                 return;
7645
7646         trace_nfs4_reclaim_complete(clp, task->tk_status);
7647         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7648                 rpc_restart_call_prepare(task);
7649                 return;
7650         }
7651         dprintk("<-- %s\n", __func__);
7652 }
7653
7654 static void nfs4_free_reclaim_complete_data(void *data)
7655 {
7656         struct nfs4_reclaim_complete_data *calldata = data;
7657
7658         kfree(calldata);
7659 }
7660
7661 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7662         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7663         .rpc_call_done = nfs4_reclaim_complete_done,
7664         .rpc_release = nfs4_free_reclaim_complete_data,
7665 };
7666
7667 /*
7668  * Issue a global reclaim complete.
7669  */
7670 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7671                 struct rpc_cred *cred)
7672 {
7673         struct nfs4_reclaim_complete_data *calldata;
7674         struct rpc_task *task;
7675         struct rpc_message msg = {
7676                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7677                 .rpc_cred = cred,
7678         };
7679         struct rpc_task_setup task_setup_data = {
7680                 .rpc_client = clp->cl_rpcclient,
7681                 .rpc_message = &msg,
7682                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7683                 .flags = RPC_TASK_ASYNC,
7684         };
7685         int status = -ENOMEM;
7686
7687         dprintk("--> %s\n", __func__);
7688         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7689         if (calldata == NULL)
7690                 goto out;
7691         calldata->clp = clp;
7692         calldata->arg.one_fs = 0;
7693
7694         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7695         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7696         msg.rpc_argp = &calldata->arg;
7697         msg.rpc_resp = &calldata->res;
7698         task_setup_data.callback_data = calldata;
7699         task = rpc_run_task(&task_setup_data);
7700         if (IS_ERR(task)) {
7701                 status = PTR_ERR(task);
7702                 goto out;
7703         }
7704         status = nfs4_wait_for_completion_rpc_task(task);
7705         if (status == 0)
7706                 status = task->tk_status;
7707         rpc_put_task(task);
7708         return 0;
7709 out:
7710         dprintk("<-- %s status=%d\n", __func__, status);
7711         return status;
7712 }
7713
7714 static void
7715 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7716 {
7717         struct nfs4_layoutget *lgp = calldata;
7718         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7719         struct nfs4_session *session = nfs4_get_session(server);
7720
7721         dprintk("--> %s\n", __func__);
7722         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7723          * right now covering the LAYOUTGET we are about to send.
7724          * However, that is not so catastrophic, and there seems
7725          * to be no way to prevent it completely.
7726          */
7727         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7728                                 &lgp->res.seq_res, task))
7729                 return;
7730         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7731                                           NFS_I(lgp->args.inode)->layout,
7732                                           &lgp->args.range,
7733                                           lgp->args.ctx->state)) {
7734                 rpc_exit(task, NFS4_OK);
7735         }
7736 }
7737
7738 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7739 {
7740         struct nfs4_layoutget *lgp = calldata;
7741         struct inode *inode = lgp->args.inode;
7742         struct nfs_server *server = NFS_SERVER(inode);
7743         struct pnfs_layout_hdr *lo;
7744         struct nfs4_state *state = NULL;
7745         unsigned long timeo, now, giveup;
7746
7747         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7748
7749         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7750                 goto out;
7751
7752         switch (task->tk_status) {
7753         case 0:
7754                 goto out;
7755         /*
7756          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7757          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7758          */
7759         case -NFS4ERR_BADLAYOUT:
7760                 goto out_overflow;
7761         /*
7762          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7763          * (or clients) writing to the same RAID stripe except when
7764          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7765          */
7766         case -NFS4ERR_LAYOUTTRYLATER:
7767                 if (lgp->args.minlength == 0)
7768                         goto out_overflow;
7769         /*
7770          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7771          * existing layout before getting a new one).
7772          */
7773         case -NFS4ERR_RECALLCONFLICT:
7774                 timeo = rpc_get_timeout(task->tk_client);
7775                 giveup = lgp->args.timestamp + timeo;
7776                 now = jiffies;
7777                 if (time_after(giveup, now)) {
7778                         unsigned long delay;
7779
7780                         /* Delay for:
7781                          * - Not less then NFS4_POLL_RETRY_MIN.
7782                          * - One last time a jiffie before we give up
7783                          * - exponential backoff (time_now minus start_attempt)
7784                          */
7785                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7786                                     min((giveup - now - 1),
7787                                         now - lgp->args.timestamp));
7788
7789                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7790                                 __func__, delay);
7791                         rpc_delay(task, delay);
7792                         task->tk_status = 0;
7793                         rpc_restart_call_prepare(task);
7794                         goto out; /* Do not call nfs4_async_handle_error() */
7795                 }
7796                 break;
7797         case -NFS4ERR_EXPIRED:
7798         case -NFS4ERR_BAD_STATEID:
7799                 spin_lock(&inode->i_lock);
7800                 lo = NFS_I(inode)->layout;
7801                 if (!lo || list_empty(&lo->plh_segs)) {
7802                         spin_unlock(&inode->i_lock);
7803                         /* If the open stateid was bad, then recover it. */
7804                         state = lgp->args.ctx->state;
7805                 } else {
7806                         LIST_HEAD(head);
7807
7808                         /*
7809                          * Mark the bad layout state as invalid, then retry
7810                          * with the current stateid.
7811                          */
7812                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7813                         spin_unlock(&inode->i_lock);
7814                         pnfs_free_lseg_list(&head);
7815         
7816                         task->tk_status = 0;
7817                         rpc_restart_call_prepare(task);
7818                 }
7819         }
7820         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7821                 rpc_restart_call_prepare(task);
7822 out:
7823         dprintk("<-- %s\n", __func__);
7824         return;
7825 out_overflow:
7826         task->tk_status = -EOVERFLOW;
7827         goto out;
7828 }
7829
7830 static size_t max_response_pages(struct nfs_server *server)
7831 {
7832         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7833         return nfs_page_array_len(0, max_resp_sz);
7834 }
7835
7836 static void nfs4_free_pages(struct page **pages, size_t size)
7837 {
7838         int i;
7839
7840         if (!pages)
7841                 return;
7842
7843         for (i = 0; i < size; i++) {
7844                 if (!pages[i])
7845                         break;
7846                 __free_page(pages[i]);
7847         }
7848         kfree(pages);
7849 }
7850
7851 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7852 {
7853         struct page **pages;
7854         int i;
7855
7856         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7857         if (!pages) {
7858                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7859                 return NULL;
7860         }
7861
7862         for (i = 0; i < size; i++) {
7863                 pages[i] = alloc_page(gfp_flags);
7864                 if (!pages[i]) {
7865                         dprintk("%s: failed to allocate page\n", __func__);
7866                         nfs4_free_pages(pages, size);
7867                         return NULL;
7868                 }
7869         }
7870
7871         return pages;
7872 }
7873
7874 static void nfs4_layoutget_release(void *calldata)
7875 {
7876         struct nfs4_layoutget *lgp = calldata;
7877         struct inode *inode = lgp->args.inode;
7878         struct nfs_server *server = NFS_SERVER(inode);
7879         size_t max_pages = max_response_pages(server);
7880
7881         dprintk("--> %s\n", __func__);
7882         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7883         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7884         put_nfs_open_context(lgp->args.ctx);
7885         kfree(calldata);
7886         dprintk("<-- %s\n", __func__);
7887 }
7888
7889 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7890         .rpc_call_prepare = nfs4_layoutget_prepare,
7891         .rpc_call_done = nfs4_layoutget_done,
7892         .rpc_release = nfs4_layoutget_release,
7893 };
7894
7895 struct pnfs_layout_segment *
7896 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7897 {
7898         struct inode *inode = lgp->args.inode;
7899         struct nfs_server *server = NFS_SERVER(inode);
7900         size_t max_pages = max_response_pages(server);
7901         struct rpc_task *task;
7902         struct rpc_message msg = {
7903                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7904                 .rpc_argp = &lgp->args,
7905                 .rpc_resp = &lgp->res,
7906                 .rpc_cred = lgp->cred,
7907         };
7908         struct rpc_task_setup task_setup_data = {
7909                 .rpc_client = server->client,
7910                 .rpc_message = &msg,
7911                 .callback_ops = &nfs4_layoutget_call_ops,
7912                 .callback_data = lgp,
7913                 .flags = RPC_TASK_ASYNC,
7914         };
7915         struct pnfs_layout_segment *lseg = NULL;
7916         int status = 0;
7917
7918         dprintk("--> %s\n", __func__);
7919
7920         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7921         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7922
7923         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7924         if (!lgp->args.layout.pages) {
7925                 nfs4_layoutget_release(lgp);
7926                 return ERR_PTR(-ENOMEM);
7927         }
7928         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7929         lgp->args.timestamp = jiffies;
7930
7931         lgp->res.layoutp = &lgp->args.layout;
7932         lgp->res.seq_res.sr_slot = NULL;
7933         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7934
7935         task = rpc_run_task(&task_setup_data);
7936         if (IS_ERR(task))
7937                 return ERR_CAST(task);
7938         status = nfs4_wait_for_completion_rpc_task(task);
7939         if (status == 0)
7940                 status = task->tk_status;
7941         trace_nfs4_layoutget(lgp->args.ctx,
7942                         &lgp->args.range,
7943                         &lgp->res.range,
7944                         status);
7945         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7946         if (status == 0 && lgp->res.layoutp->len)
7947                 lseg = pnfs_layout_process(lgp);
7948         rpc_put_task(task);
7949         dprintk("<-- %s status=%d\n", __func__, status);
7950         if (status)
7951                 return ERR_PTR(status);
7952         return lseg;
7953 }
7954
7955 static void
7956 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7957 {
7958         struct nfs4_layoutreturn *lrp = calldata;
7959
7960         dprintk("--> %s\n", __func__);
7961         nfs41_setup_sequence(lrp->clp->cl_session,
7962                         &lrp->args.seq_args,
7963                         &lrp->res.seq_res,
7964                         task);
7965 }
7966
7967 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7968 {
7969         struct nfs4_layoutreturn *lrp = calldata;
7970         struct nfs_server *server;
7971
7972         dprintk("--> %s\n", __func__);
7973
7974         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7975                 return;
7976
7977         server = NFS_SERVER(lrp->args.inode);
7978         switch (task->tk_status) {
7979         default:
7980                 task->tk_status = 0;
7981         case 0:
7982                 break;
7983         case -NFS4ERR_DELAY:
7984                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7985                         break;
7986                 rpc_restart_call_prepare(task);
7987                 return;
7988         }
7989         dprintk("<-- %s\n", __func__);
7990 }
7991
7992 static void nfs4_layoutreturn_release(void *calldata)
7993 {
7994         struct nfs4_layoutreturn *lrp = calldata;
7995         struct pnfs_layout_hdr *lo = lrp->args.layout;
7996         LIST_HEAD(freeme);
7997
7998         dprintk("--> %s\n", __func__);
7999         spin_lock(&lo->plh_inode->i_lock);
8000         if (lrp->res.lrs_present)
8001                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8002         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
8003         pnfs_clear_layoutreturn_waitbit(lo);
8004         lo->plh_block_lgets--;
8005         spin_unlock(&lo->plh_inode->i_lock);
8006         pnfs_free_lseg_list(&freeme);
8007         pnfs_put_layout_hdr(lrp->args.layout);
8008         nfs_iput_and_deactive(lrp->inode);
8009         kfree(calldata);
8010         dprintk("<-- %s\n", __func__);
8011 }
8012
8013 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8014         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8015         .rpc_call_done = nfs4_layoutreturn_done,
8016         .rpc_release = nfs4_layoutreturn_release,
8017 };
8018
8019 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8020 {
8021         struct rpc_task *task;
8022         struct rpc_message msg = {
8023                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8024                 .rpc_argp = &lrp->args,
8025                 .rpc_resp = &lrp->res,
8026                 .rpc_cred = lrp->cred,
8027         };
8028         struct rpc_task_setup task_setup_data = {
8029                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8030                 .rpc_message = &msg,
8031                 .callback_ops = &nfs4_layoutreturn_call_ops,
8032                 .callback_data = lrp,
8033         };
8034         int status = 0;
8035
8036         dprintk("--> %s\n", __func__);
8037         if (!sync) {
8038                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8039                 if (!lrp->inode) {
8040                         nfs4_layoutreturn_release(lrp);
8041                         return -EAGAIN;
8042                 }
8043                 task_setup_data.flags |= RPC_TASK_ASYNC;
8044         }
8045         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8046         task = rpc_run_task(&task_setup_data);
8047         if (IS_ERR(task))
8048                 return PTR_ERR(task);
8049         if (sync)
8050                 status = task->tk_status;
8051         trace_nfs4_layoutreturn(lrp->args.inode, status);
8052         dprintk("<-- %s status=%d\n", __func__, status);
8053         rpc_put_task(task);
8054         return status;
8055 }
8056
8057 static int
8058 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8059                 struct pnfs_device *pdev,
8060                 struct rpc_cred *cred)
8061 {
8062         struct nfs4_getdeviceinfo_args args = {
8063                 .pdev = pdev,
8064                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8065                         NOTIFY_DEVICEID4_DELETE,
8066         };
8067         struct nfs4_getdeviceinfo_res res = {
8068                 .pdev = pdev,
8069         };
8070         struct rpc_message msg = {
8071                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8072                 .rpc_argp = &args,
8073                 .rpc_resp = &res,
8074                 .rpc_cred = cred,
8075         };
8076         int status;
8077
8078         dprintk("--> %s\n", __func__);
8079         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8080         if (res.notification & ~args.notify_types)
8081                 dprintk("%s: unsupported notification\n", __func__);
8082         if (res.notification != args.notify_types)
8083                 pdev->nocache = 1;
8084
8085         dprintk("<-- %s status=%d\n", __func__, status);
8086
8087         return status;
8088 }
8089
8090 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8091                 struct pnfs_device *pdev,
8092                 struct rpc_cred *cred)
8093 {
8094         struct nfs4_exception exception = { };
8095         int err;
8096
8097         do {
8098                 err = nfs4_handle_exception(server,
8099                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8100                                         &exception);
8101         } while (exception.retry);
8102         return err;
8103 }
8104 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8105
8106 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8107 {
8108         struct nfs4_layoutcommit_data *data = calldata;
8109         struct nfs_server *server = NFS_SERVER(data->args.inode);
8110         struct nfs4_session *session = nfs4_get_session(server);
8111
8112         nfs41_setup_sequence(session,
8113                         &data->args.seq_args,
8114                         &data->res.seq_res,
8115                         task);
8116 }
8117
8118 static void
8119 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8120 {
8121         struct nfs4_layoutcommit_data *data = calldata;
8122         struct nfs_server *server = NFS_SERVER(data->args.inode);
8123
8124         if (!nfs41_sequence_done(task, &data->res.seq_res))
8125                 return;
8126
8127         switch (task->tk_status) { /* Just ignore these failures */
8128         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8129         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8130         case -NFS4ERR_BADLAYOUT:     /* no layout */
8131         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8132                 task->tk_status = 0;
8133         case 0:
8134                 break;
8135         default:
8136                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8137                         rpc_restart_call_prepare(task);
8138                         return;
8139                 }
8140         }
8141 }
8142
8143 static void nfs4_layoutcommit_release(void *calldata)
8144 {
8145         struct nfs4_layoutcommit_data *data = calldata;
8146
8147         pnfs_cleanup_layoutcommit(data);
8148         nfs_post_op_update_inode_force_wcc(data->args.inode,
8149                                            data->res.fattr);
8150         put_rpccred(data->cred);
8151         nfs_iput_and_deactive(data->inode);
8152         kfree(data);
8153 }
8154
8155 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8156         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8157         .rpc_call_done = nfs4_layoutcommit_done,
8158         .rpc_release = nfs4_layoutcommit_release,
8159 };
8160
8161 int
8162 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8163 {
8164         struct rpc_message msg = {
8165                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8166                 .rpc_argp = &data->args,
8167                 .rpc_resp = &data->res,
8168                 .rpc_cred = data->cred,
8169         };
8170         struct rpc_task_setup task_setup_data = {
8171                 .task = &data->task,
8172                 .rpc_client = NFS_CLIENT(data->args.inode),
8173                 .rpc_message = &msg,
8174                 .callback_ops = &nfs4_layoutcommit_ops,
8175                 .callback_data = data,
8176         };
8177         struct rpc_task *task;
8178         int status = 0;
8179
8180         dprintk("NFS: initiating layoutcommit call. sync %d "
8181                 "lbw: %llu inode %lu\n", sync,
8182                 data->args.lastbytewritten,
8183                 data->args.inode->i_ino);
8184
8185         if (!sync) {
8186                 data->inode = nfs_igrab_and_active(data->args.inode);
8187                 if (data->inode == NULL) {
8188                         nfs4_layoutcommit_release(data);
8189                         return -EAGAIN;
8190                 }
8191                 task_setup_data.flags = RPC_TASK_ASYNC;
8192         }
8193         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8194         task = rpc_run_task(&task_setup_data);
8195         if (IS_ERR(task))
8196                 return PTR_ERR(task);
8197         if (sync)
8198                 status = task->tk_status;
8199         trace_nfs4_layoutcommit(data->args.inode, status);
8200         dprintk("%s: status %d\n", __func__, status);
8201         rpc_put_task(task);
8202         return status;
8203 }
8204
8205 /**
8206  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8207  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8208  */
8209 static int
8210 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8211                     struct nfs_fsinfo *info,
8212                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8213 {
8214         struct nfs41_secinfo_no_name_args args = {
8215                 .style = SECINFO_STYLE_CURRENT_FH,
8216         };
8217         struct nfs4_secinfo_res res = {
8218                 .flavors = flavors,
8219         };
8220         struct rpc_message msg = {
8221                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8222                 .rpc_argp = &args,
8223                 .rpc_resp = &res,
8224         };
8225         struct rpc_clnt *clnt = server->client;
8226         struct rpc_cred *cred = NULL;
8227         int status;
8228
8229         if (use_integrity) {
8230                 clnt = server->nfs_client->cl_rpcclient;
8231                 cred = nfs4_get_clid_cred(server->nfs_client);
8232                 msg.rpc_cred = cred;
8233         }
8234
8235         dprintk("--> %s\n", __func__);
8236         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8237                                 &res.seq_res, 0);
8238         dprintk("<-- %s status=%d\n", __func__, status);
8239
8240         if (cred)
8241                 put_rpccred(cred);
8242
8243         return status;
8244 }
8245
8246 static int
8247 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8248                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8249 {
8250         struct nfs4_exception exception = { };
8251         int err;
8252         do {
8253                 /* first try using integrity protection */
8254                 err = -NFS4ERR_WRONGSEC;
8255
8256                 /* try to use integrity protection with machine cred */
8257                 if (_nfs4_is_integrity_protected(server->nfs_client))
8258                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8259                                                           flavors, true);
8260
8261                 /*
8262                  * if unable to use integrity protection, or SECINFO with
8263                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8264                  * disallowed by spec, but exists in deployed servers) use
8265                  * the current filesystem's rpc_client and the user cred.
8266                  */
8267                 if (err == -NFS4ERR_WRONGSEC)
8268                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8269                                                           flavors, false);
8270
8271                 switch (err) {
8272                 case 0:
8273                 case -NFS4ERR_WRONGSEC:
8274                 case -ENOTSUPP:
8275                         goto out;
8276                 default:
8277                         err = nfs4_handle_exception(server, err, &exception);
8278                 }
8279         } while (exception.retry);
8280 out:
8281         return err;
8282 }
8283
8284 static int
8285 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8286                     struct nfs_fsinfo *info)
8287 {
8288         int err;
8289         struct page *page;
8290         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8291         struct nfs4_secinfo_flavors *flavors;
8292         struct nfs4_secinfo4 *secinfo;
8293         int i;
8294
8295         page = alloc_page(GFP_KERNEL);
8296         if (!page) {
8297                 err = -ENOMEM;
8298                 goto out;
8299         }
8300
8301         flavors = page_address(page);
8302         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8303
8304         /*
8305          * Fall back on "guess and check" method if
8306          * the server doesn't support SECINFO_NO_NAME
8307          */
8308         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8309                 err = nfs4_find_root_sec(server, fhandle, info);
8310                 goto out_freepage;
8311         }
8312         if (err)
8313                 goto out_freepage;
8314
8315         for (i = 0; i < flavors->num_flavors; i++) {
8316                 secinfo = &flavors->flavors[i];
8317
8318                 switch (secinfo->flavor) {
8319                 case RPC_AUTH_NULL:
8320                 case RPC_AUTH_UNIX:
8321                 case RPC_AUTH_GSS:
8322                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8323                                         &secinfo->flavor_info);
8324                         break;
8325                 default:
8326                         flavor = RPC_AUTH_MAXFLAVOR;
8327                         break;
8328                 }
8329
8330                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8331                         flavor = RPC_AUTH_MAXFLAVOR;
8332
8333                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8334                         err = nfs4_lookup_root_sec(server, fhandle,
8335                                                    info, flavor);
8336                         if (!err)
8337                                 break;
8338                 }
8339         }
8340
8341         if (flavor == RPC_AUTH_MAXFLAVOR)
8342                 err = -EPERM;
8343
8344 out_freepage:
8345         put_page(page);
8346         if (err == -EACCES)
8347                 return -EPERM;
8348 out:
8349         return err;
8350 }
8351
8352 static int _nfs41_test_stateid(struct nfs_server *server,
8353                 nfs4_stateid *stateid,
8354                 struct rpc_cred *cred)
8355 {
8356         int status;
8357         struct nfs41_test_stateid_args args = {
8358                 .stateid = stateid,
8359         };
8360         struct nfs41_test_stateid_res res;
8361         struct rpc_message msg = {
8362                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8363                 .rpc_argp = &args,
8364                 .rpc_resp = &res,
8365                 .rpc_cred = cred,
8366         };
8367         struct rpc_clnt *rpc_client = server->client;
8368
8369         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8370                 &rpc_client, &msg);
8371
8372         dprintk("NFS call  test_stateid %p\n", stateid);
8373         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8374         nfs4_set_sequence_privileged(&args.seq_args);
8375         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8376                         &args.seq_args, &res.seq_res);
8377         if (status != NFS_OK) {
8378                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8379                 return status;
8380         }
8381         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8382         return -res.status;
8383 }
8384
8385 /**
8386  * nfs41_test_stateid - perform a TEST_STATEID operation
8387  *
8388  * @server: server / transport on which to perform the operation
8389  * @stateid: state ID to test
8390  * @cred: credential
8391  *
8392  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8393  * Otherwise a negative NFS4ERR value is returned if the operation
8394  * failed or the state ID is not currently valid.
8395  */
8396 static int nfs41_test_stateid(struct nfs_server *server,
8397                 nfs4_stateid *stateid,
8398                 struct rpc_cred *cred)
8399 {
8400         struct nfs4_exception exception = { };
8401         int err;
8402         do {
8403                 err = _nfs41_test_stateid(server, stateid, cred);
8404                 if (err != -NFS4ERR_DELAY)
8405                         break;
8406                 nfs4_handle_exception(server, err, &exception);
8407         } while (exception.retry);
8408         return err;
8409 }
8410
8411 struct nfs_free_stateid_data {
8412         struct nfs_server *server;
8413         struct nfs41_free_stateid_args args;
8414         struct nfs41_free_stateid_res res;
8415 };
8416
8417 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8418 {
8419         struct nfs_free_stateid_data *data = calldata;
8420         nfs41_setup_sequence(nfs4_get_session(data->server),
8421                         &data->args.seq_args,
8422                         &data->res.seq_res,
8423                         task);
8424 }
8425
8426 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8427 {
8428         struct nfs_free_stateid_data *data = calldata;
8429
8430         nfs41_sequence_done(task, &data->res.seq_res);
8431
8432         switch (task->tk_status) {
8433         case -NFS4ERR_DELAY:
8434                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8435                         rpc_restart_call_prepare(task);
8436         }
8437 }
8438
8439 static void nfs41_free_stateid_release(void *calldata)
8440 {
8441         kfree(calldata);
8442 }
8443
8444 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8445         .rpc_call_prepare = nfs41_free_stateid_prepare,
8446         .rpc_call_done = nfs41_free_stateid_done,
8447         .rpc_release = nfs41_free_stateid_release,
8448 };
8449
8450 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8451                 nfs4_stateid *stateid,
8452                 struct rpc_cred *cred,
8453                 bool privileged)
8454 {
8455         struct rpc_message msg = {
8456                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8457                 .rpc_cred = cred,
8458         };
8459         struct rpc_task_setup task_setup = {
8460                 .rpc_client = server->client,
8461                 .rpc_message = &msg,
8462                 .callback_ops = &nfs41_free_stateid_ops,
8463                 .flags = RPC_TASK_ASYNC,
8464         };
8465         struct nfs_free_stateid_data *data;
8466
8467         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8468                 &task_setup.rpc_client, &msg);
8469
8470         dprintk("NFS call  free_stateid %p\n", stateid);
8471         data = kmalloc(sizeof(*data), GFP_NOFS);
8472         if (!data)
8473                 return ERR_PTR(-ENOMEM);
8474         data->server = server;
8475         nfs4_stateid_copy(&data->args.stateid, stateid);
8476
8477         task_setup.callback_data = data;
8478
8479         msg.rpc_argp = &data->args;
8480         msg.rpc_resp = &data->res;
8481         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8482         if (privileged)
8483                 nfs4_set_sequence_privileged(&data->args.seq_args);
8484
8485         return rpc_run_task(&task_setup);
8486 }
8487
8488 /**
8489  * nfs41_free_stateid - perform a FREE_STATEID operation
8490  *
8491  * @server: server / transport on which to perform the operation
8492  * @stateid: state ID to release
8493  * @cred: credential
8494  *
8495  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8496  * negative NFS4ERR value is returned.
8497  */
8498 static int nfs41_free_stateid(struct nfs_server *server,
8499                 nfs4_stateid *stateid,
8500                 struct rpc_cred *cred)
8501 {
8502         struct rpc_task *task;
8503         int ret;
8504
8505         task = _nfs41_free_stateid(server, stateid, cred, true);
8506         if (IS_ERR(task))
8507                 return PTR_ERR(task);
8508         ret = rpc_wait_for_completion_task(task);
8509         if (!ret)
8510                 ret = task->tk_status;
8511         rpc_put_task(task);
8512         return ret;
8513 }
8514
8515 static void
8516 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8517 {
8518         struct rpc_task *task;
8519         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8520
8521         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8522         nfs4_free_lock_state(server, lsp);
8523         if (IS_ERR(task))
8524                 return;
8525         rpc_put_task(task);
8526 }
8527
8528 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8529                 const nfs4_stateid *s2)
8530 {
8531         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8532                 return false;
8533
8534         if (s1->seqid == s2->seqid)
8535                 return true;
8536         if (s1->seqid == 0 || s2->seqid == 0)
8537                 return true;
8538
8539         return false;
8540 }
8541
8542 #endif /* CONFIG_NFS_V4_1 */
8543
8544 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8545                 const nfs4_stateid *s2)
8546 {
8547         return nfs4_stateid_match(s1, s2);
8548 }
8549
8550
8551 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8552         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8553         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8554         .recover_open   = nfs4_open_reclaim,
8555         .recover_lock   = nfs4_lock_reclaim,
8556         .establish_clid = nfs4_init_clientid,
8557         .detect_trunking = nfs40_discover_server_trunking,
8558 };
8559
8560 #if defined(CONFIG_NFS_V4_1)
8561 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8562         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8563         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8564         .recover_open   = nfs4_open_reclaim,
8565         .recover_lock   = nfs4_lock_reclaim,
8566         .establish_clid = nfs41_init_clientid,
8567         .reclaim_complete = nfs41_proc_reclaim_complete,
8568         .detect_trunking = nfs41_discover_server_trunking,
8569 };
8570 #endif /* CONFIG_NFS_V4_1 */
8571
8572 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8573         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8574         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8575         .recover_open   = nfs40_open_expired,
8576         .recover_lock   = nfs4_lock_expired,
8577         .establish_clid = nfs4_init_clientid,
8578 };
8579
8580 #if defined(CONFIG_NFS_V4_1)
8581 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8582         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8583         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8584         .recover_open   = nfs41_open_expired,
8585         .recover_lock   = nfs41_lock_expired,
8586         .establish_clid = nfs41_init_clientid,
8587 };
8588 #endif /* CONFIG_NFS_V4_1 */
8589
8590 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8591         .sched_state_renewal = nfs4_proc_async_renew,
8592         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8593         .renew_lease = nfs4_proc_renew,
8594 };
8595
8596 #if defined(CONFIG_NFS_V4_1)
8597 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8598         .sched_state_renewal = nfs41_proc_async_sequence,
8599         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8600         .renew_lease = nfs4_proc_sequence,
8601 };
8602 #endif
8603
8604 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8605         .get_locations = _nfs40_proc_get_locations,
8606         .fsid_present = _nfs40_proc_fsid_present,
8607 };
8608
8609 #if defined(CONFIG_NFS_V4_1)
8610 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8611         .get_locations = _nfs41_proc_get_locations,
8612         .fsid_present = _nfs41_proc_fsid_present,
8613 };
8614 #endif  /* CONFIG_NFS_V4_1 */
8615
8616 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8617         .minor_version = 0,
8618         .init_caps = NFS_CAP_READDIRPLUS
8619                 | NFS_CAP_ATOMIC_OPEN
8620                 | NFS_CAP_POSIX_LOCK,
8621         .init_client = nfs40_init_client,
8622         .shutdown_client = nfs40_shutdown_client,
8623         .match_stateid = nfs4_match_stateid,
8624         .find_root_sec = nfs4_find_root_sec,
8625         .free_lock_state = nfs4_release_lockowner,
8626         .alloc_seqid = nfs_alloc_seqid,
8627         .call_sync_ops = &nfs40_call_sync_ops,
8628         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8629         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8630         .state_renewal_ops = &nfs40_state_renewal_ops,
8631         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8632 };
8633
8634 #if defined(CONFIG_NFS_V4_1)
8635 static struct nfs_seqid *
8636 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8637 {
8638         return NULL;
8639 }
8640
8641 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8642         .minor_version = 1,
8643         .init_caps = NFS_CAP_READDIRPLUS
8644                 | NFS_CAP_ATOMIC_OPEN
8645                 | NFS_CAP_POSIX_LOCK
8646                 | NFS_CAP_STATEID_NFSV41
8647                 | NFS_CAP_ATOMIC_OPEN_V1,
8648         .init_client = nfs41_init_client,
8649         .shutdown_client = nfs41_shutdown_client,
8650         .match_stateid = nfs41_match_stateid,
8651         .find_root_sec = nfs41_find_root_sec,
8652         .free_lock_state = nfs41_free_lock_state,
8653         .alloc_seqid = nfs_alloc_no_seqid,
8654         .call_sync_ops = &nfs41_call_sync_ops,
8655         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8656         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8657         .state_renewal_ops = &nfs41_state_renewal_ops,
8658         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8659 };
8660 #endif
8661
8662 #if defined(CONFIG_NFS_V4_2)
8663 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8664         .minor_version = 2,
8665         .init_caps = NFS_CAP_READDIRPLUS
8666                 | NFS_CAP_ATOMIC_OPEN
8667                 | NFS_CAP_POSIX_LOCK
8668                 | NFS_CAP_STATEID_NFSV41
8669                 | NFS_CAP_ATOMIC_OPEN_V1
8670                 | NFS_CAP_ALLOCATE
8671                 | NFS_CAP_DEALLOCATE
8672                 | NFS_CAP_SEEK
8673                 | NFS_CAP_LAYOUTSTATS,
8674         .init_client = nfs41_init_client,
8675         .shutdown_client = nfs41_shutdown_client,
8676         .match_stateid = nfs41_match_stateid,
8677         .find_root_sec = nfs41_find_root_sec,
8678         .free_lock_state = nfs41_free_lock_state,
8679         .call_sync_ops = &nfs41_call_sync_ops,
8680         .alloc_seqid = nfs_alloc_no_seqid,
8681         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8682         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8683         .state_renewal_ops = &nfs41_state_renewal_ops,
8684         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8685 };
8686 #endif
8687
8688 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8689         [0] = &nfs_v4_0_minor_ops,
8690 #if defined(CONFIG_NFS_V4_1)
8691         [1] = &nfs_v4_1_minor_ops,
8692 #endif
8693 #if defined(CONFIG_NFS_V4_2)
8694         [2] = &nfs_v4_2_minor_ops,
8695 #endif
8696 };
8697
8698 static const struct inode_operations nfs4_dir_inode_operations = {
8699         .create         = nfs_create,
8700         .lookup         = nfs_lookup,
8701         .atomic_open    = nfs_atomic_open,
8702         .link           = nfs_link,
8703         .unlink         = nfs_unlink,
8704         .symlink        = nfs_symlink,
8705         .mkdir          = nfs_mkdir,
8706         .rmdir          = nfs_rmdir,
8707         .mknod          = nfs_mknod,
8708         .rename         = nfs_rename,
8709         .permission     = nfs_permission,
8710         .getattr        = nfs_getattr,
8711         .setattr        = nfs_setattr,
8712         .getxattr       = generic_getxattr,
8713         .setxattr       = generic_setxattr,
8714         .listxattr      = generic_listxattr,
8715         .removexattr    = generic_removexattr,
8716 };
8717
8718 static const struct inode_operations nfs4_file_inode_operations = {
8719         .permission     = nfs_permission,
8720         .getattr        = nfs_getattr,
8721         .setattr        = nfs_setattr,
8722         .getxattr       = generic_getxattr,
8723         .setxattr       = generic_setxattr,
8724         .listxattr      = generic_listxattr,
8725         .removexattr    = generic_removexattr,
8726 };
8727
8728 const struct nfs_rpc_ops nfs_v4_clientops = {
8729         .version        = 4,                    /* protocol version */
8730         .dentry_ops     = &nfs4_dentry_operations,
8731         .dir_inode_ops  = &nfs4_dir_inode_operations,
8732         .file_inode_ops = &nfs4_file_inode_operations,
8733         .file_ops       = &nfs4_file_operations,
8734         .getroot        = nfs4_proc_get_root,
8735         .submount       = nfs4_submount,
8736         .try_mount      = nfs4_try_mount,
8737         .getattr        = nfs4_proc_getattr,
8738         .setattr        = nfs4_proc_setattr,
8739         .lookup         = nfs4_proc_lookup,
8740         .access         = nfs4_proc_access,
8741         .readlink       = nfs4_proc_readlink,
8742         .create         = nfs4_proc_create,
8743         .remove         = nfs4_proc_remove,
8744         .unlink_setup   = nfs4_proc_unlink_setup,
8745         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8746         .unlink_done    = nfs4_proc_unlink_done,
8747         .rename_setup   = nfs4_proc_rename_setup,
8748         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8749         .rename_done    = nfs4_proc_rename_done,
8750         .link           = nfs4_proc_link,
8751         .symlink        = nfs4_proc_symlink,
8752         .mkdir          = nfs4_proc_mkdir,
8753         .rmdir          = nfs4_proc_remove,
8754         .readdir        = nfs4_proc_readdir,
8755         .mknod          = nfs4_proc_mknod,
8756         .statfs         = nfs4_proc_statfs,
8757         .fsinfo         = nfs4_proc_fsinfo,
8758         .pathconf       = nfs4_proc_pathconf,
8759         .set_capabilities = nfs4_server_capabilities,
8760         .decode_dirent  = nfs4_decode_dirent,
8761         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8762         .read_setup     = nfs4_proc_read_setup,
8763         .read_done      = nfs4_read_done,
8764         .write_setup    = nfs4_proc_write_setup,
8765         .write_done     = nfs4_write_done,
8766         .commit_setup   = nfs4_proc_commit_setup,
8767         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8768         .commit_done    = nfs4_commit_done,
8769         .lock           = nfs4_proc_lock,
8770         .clear_acl_cache = nfs4_zap_acl_attr,
8771         .close_context  = nfs4_close_context,
8772         .open_context   = nfs4_atomic_open,
8773         .have_delegation = nfs4_have_delegation,
8774         .return_delegation = nfs4_inode_return_delegation,
8775         .alloc_client   = nfs4_alloc_client,
8776         .init_client    = nfs4_init_client,
8777         .free_client    = nfs4_free_client,
8778         .create_server  = nfs4_create_server,
8779         .clone_server   = nfs_clone_server,
8780 };
8781
8782 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8783         .prefix = XATTR_NAME_NFSV4_ACL,
8784         .list   = nfs4_xattr_list_nfs4_acl,
8785         .get    = nfs4_xattr_get_nfs4_acl,
8786         .set    = nfs4_xattr_set_nfs4_acl,
8787 };
8788
8789 const struct xattr_handler *nfs4_xattr_handlers[] = {
8790         &nfs4_xattr_nfs4_acl_handler,
8791 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8792         &nfs4_xattr_nfs4_label_handler,
8793 #endif
8794         NULL
8795 };
8796
8797 /*
8798  * Local variables:
8799  *  c-basic-offset: 8
8800  * End:
8801  */