NFSv4.1: Fix an NFSv4.1 state renewal regression
[linux-2.6-block.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/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/nfs_idmap.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4session.h"
67 #include "fscache.h"
68
69 #include "nfs4trace.h"
70
71 #define NFSDBG_FACILITY         NFSDBG_PROC
72
73 #define NFS4_POLL_RETRY_MIN     (HZ/10)
74 #define NFS4_POLL_RETRY_MAX     (15*HZ)
75
76 struct nfs4_opendata;
77 static int _nfs4_proc_open(struct nfs4_opendata *data);
78 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
79 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
80 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
85                             struct nfs_fattr *fattr, struct iattr *sattr,
86                             struct nfs4_state *state, struct nfs4_label *ilabel,
87                             struct nfs4_label *olabel);
88 #ifdef CONFIG_NFS_V4_1
89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
90                 struct rpc_cred *);
91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
92                 struct rpc_cred *);
93 #endif
94
95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
96 static inline struct nfs4_label *
97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
98         struct iattr *sattr, struct nfs4_label *label)
99 {
100         int err;
101
102         if (label == NULL)
103                 return NULL;
104
105         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
106                 return NULL;
107
108         err = security_dentry_init_security(dentry, sattr->ia_mode,
109                                 &dentry->d_name, (void **)&label->label, &label->len);
110         if (err == 0)
111                 return label;
112
113         return NULL;
114 }
115 static inline void
116 nfs4_label_release_security(struct nfs4_label *label)
117 {
118         if (label)
119                 security_release_secctx(label->label, label->len);
120 }
121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
122 {
123         if (label)
124                 return server->attr_bitmask;
125
126         return server->attr_bitmask_nl;
127 }
128 #else
129 static inline struct nfs4_label *
130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
131         struct iattr *sattr, struct nfs4_label *l)
132 { return NULL; }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 { return; }
136 static inline u32 *
137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 { return server->attr_bitmask; }
139 #endif
140
141 /* Prevent leaks of NFSv4 errors into userland */
142 static int nfs4_map_errors(int err)
143 {
144         if (err >= -1000)
145                 return err;
146         switch (err) {
147         case -NFS4ERR_RESOURCE:
148         case -NFS4ERR_LAYOUTTRYLATER:
149         case -NFS4ERR_RECALLCONFLICT:
150                 return -EREMOTEIO;
151         case -NFS4ERR_WRONGSEC:
152         case -NFS4ERR_WRONG_CRED:
153                 return -EPERM;
154         case -NFS4ERR_BADOWNER:
155         case -NFS4ERR_BADNAME:
156                 return -EINVAL;
157         case -NFS4ERR_SHARE_DENIED:
158                 return -EACCES;
159         case -NFS4ERR_MINOR_VERS_MISMATCH:
160                 return -EPROTONOSUPPORT;
161         case -NFS4ERR_ACCESS:
162                 return -EACCES;
163         case -NFS4ERR_FILE_OPEN:
164                 return -EBUSY;
165         default:
166                 dprintk("%s could not handle NFSv4 error %d\n",
167                                 __func__, -err);
168                 break;
169         }
170         return -EIO;
171 }
172
173 /*
174  * This is our standard bitmap for GETATTR requests.
175  */
176 const u32 nfs4_fattr_bitmap[3] = {
177         FATTR4_WORD0_TYPE
178         | FATTR4_WORD0_CHANGE
179         | FATTR4_WORD0_SIZE
180         | FATTR4_WORD0_FSID
181         | FATTR4_WORD0_FILEID,
182         FATTR4_WORD1_MODE
183         | FATTR4_WORD1_NUMLINKS
184         | FATTR4_WORD1_OWNER
185         | FATTR4_WORD1_OWNER_GROUP
186         | FATTR4_WORD1_RAWDEV
187         | FATTR4_WORD1_SPACE_USED
188         | FATTR4_WORD1_TIME_ACCESS
189         | FATTR4_WORD1_TIME_METADATA
190         | FATTR4_WORD1_TIME_MODIFY,
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(dentry->d_inode));
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(dentry->d_parent->d_inode));
311
312         readdir->pgbase = (char *)p - (char *)start;
313         readdir->count -= readdir->pgbase;
314         kunmap_atomic(start);
315 }
316
317 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
318 {
319         int res = 0;
320
321         might_sleep();
322
323         if (*timeout <= 0)
324                 *timeout = NFS4_POLL_RETRY_MIN;
325         if (*timeout > NFS4_POLL_RETRY_MAX)
326                 *timeout = NFS4_POLL_RETRY_MAX;
327         freezable_schedule_timeout_killable_unsafe(*timeout);
328         if (fatal_signal_pending(current))
329                 res = -ERESTARTSYS;
330         *timeout <<= 1;
331         return res;
332 }
333
334 /* This is the error handling routine for processes that are allowed
335  * to sleep.
336  */
337 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
338 {
339         struct nfs_client *clp = server->nfs_client;
340         struct nfs4_state *state = exception->state;
341         struct inode *inode = exception->inode;
342         int ret = errorcode;
343
344         exception->retry = 0;
345         switch(errorcode) {
346                 case 0:
347                         return 0;
348                 case -NFS4ERR_OPENMODE:
349                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
350                                 nfs4_inode_return_delegation(inode);
351                                 exception->retry = 1;
352                                 return 0;
353                         }
354                         if (state == NULL)
355                                 break;
356                         ret = nfs4_schedule_stateid_recovery(server, state);
357                         if (ret < 0)
358                                 break;
359                         goto wait_on_recovery;
360                 case -NFS4ERR_DELEG_REVOKED:
361                 case -NFS4ERR_ADMIN_REVOKED:
362                 case -NFS4ERR_BAD_STATEID:
363                         if (inode != NULL && nfs4_have_delegation(inode, FMODE_READ)) {
364                                 nfs_remove_bad_delegation(inode);
365                                 exception->retry = 1;
366                                 break;
367                         }
368                         if (state == NULL)
369                                 break;
370                         ret = nfs4_schedule_stateid_recovery(server, state);
371                         if (ret < 0)
372                                 break;
373                         goto wait_on_recovery;
374                 case -NFS4ERR_EXPIRED:
375                         if (state != NULL) {
376                                 ret = nfs4_schedule_stateid_recovery(server, state);
377                                 if (ret < 0)
378                                         break;
379                         }
380                 case -NFS4ERR_STALE_STATEID:
381                 case -NFS4ERR_STALE_CLIENTID:
382                         nfs4_schedule_lease_recovery(clp);
383                         goto wait_on_recovery;
384                 case -NFS4ERR_MOVED:
385                         ret = nfs4_schedule_migration_recovery(server);
386                         if (ret < 0)
387                                 break;
388                         goto wait_on_recovery;
389                 case -NFS4ERR_LEASE_MOVED:
390                         nfs4_schedule_lease_moved_recovery(clp);
391                         goto wait_on_recovery;
392 #if defined(CONFIG_NFS_V4_1)
393                 case -NFS4ERR_BADSESSION:
394                 case -NFS4ERR_BADSLOT:
395                 case -NFS4ERR_BAD_HIGH_SLOT:
396                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
397                 case -NFS4ERR_DEADSESSION:
398                 case -NFS4ERR_SEQ_FALSE_RETRY:
399                 case -NFS4ERR_SEQ_MISORDERED:
400                         dprintk("%s ERROR: %d Reset session\n", __func__,
401                                 errorcode);
402                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
403                         goto wait_on_recovery;
404 #endif /* defined(CONFIG_NFS_V4_1) */
405                 case -NFS4ERR_FILE_OPEN:
406                         if (exception->timeout > HZ) {
407                                 /* We have retried a decent amount, time to
408                                  * fail
409                                  */
410                                 ret = -EBUSY;
411                                 break;
412                         }
413                 case -NFS4ERR_GRACE:
414                 case -NFS4ERR_DELAY:
415                         ret = nfs4_delay(server->client, &exception->timeout);
416                         if (ret != 0)
417                                 break;
418                 case -NFS4ERR_RETRY_UNCACHED_REP:
419                 case -NFS4ERR_OLD_STATEID:
420                         exception->retry = 1;
421                         break;
422                 case -NFS4ERR_BADOWNER:
423                         /* The following works around a Linux server bug! */
424                 case -NFS4ERR_BADNAME:
425                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
426                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
427                                 exception->retry = 1;
428                                 printk(KERN_WARNING "NFS: v4 server %s "
429                                                 "does not accept raw "
430                                                 "uid/gids. "
431                                                 "Reenabling the idmapper.\n",
432                                                 server->nfs_client->cl_hostname);
433                         }
434         }
435         /* We failed to handle the error */
436         return nfs4_map_errors(ret);
437 wait_on_recovery:
438         ret = nfs4_wait_clnt_recover(clp);
439         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
440                 return -EIO;
441         if (ret == 0)
442                 exception->retry = 1;
443         return ret;
444 }
445
446 /*
447  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
448  * or 'false' otherwise.
449  */
450 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
451 {
452         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
453
454         if (flavor == RPC_AUTH_GSS_KRB5I ||
455             flavor == RPC_AUTH_GSS_KRB5P)
456                 return true;
457
458         return false;
459 }
460
461 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
462 {
463         spin_lock(&clp->cl_lock);
464         if (time_before(clp->cl_last_renewal,timestamp))
465                 clp->cl_last_renewal = timestamp;
466         spin_unlock(&clp->cl_lock);
467 }
468
469 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
470 {
471         do_renew_lease(server->nfs_client, timestamp);
472 }
473
474 struct nfs4_call_sync_data {
475         const struct nfs_server *seq_server;
476         struct nfs4_sequence_args *seq_args;
477         struct nfs4_sequence_res *seq_res;
478 };
479
480 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
481                                struct nfs4_sequence_res *res, int cache_reply)
482 {
483         args->sa_slot = NULL;
484         args->sa_cache_this = cache_reply;
485         args->sa_privileged = 0;
486
487         res->sr_slot = NULL;
488 }
489
490 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
491 {
492         args->sa_privileged = 1;
493 }
494
495 static int nfs40_setup_sequence(const struct nfs_server *server,
496                                 struct nfs4_sequence_args *args,
497                                 struct nfs4_sequence_res *res,
498                                 struct rpc_task *task)
499 {
500         struct nfs4_slot_table *tbl = server->nfs_client->cl_slot_tbl;
501         struct nfs4_slot *slot;
502
503         /* slot already allocated? */
504         if (res->sr_slot != NULL)
505                 goto out_start;
506
507         spin_lock(&tbl->slot_tbl_lock);
508         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
509                 goto out_sleep;
510
511         slot = nfs4_alloc_slot(tbl);
512         if (IS_ERR(slot)) {
513                 if (slot == ERR_PTR(-ENOMEM))
514                         task->tk_timeout = HZ >> 2;
515                 goto out_sleep;
516         }
517         spin_unlock(&tbl->slot_tbl_lock);
518
519         args->sa_slot = slot;
520         res->sr_slot = slot;
521
522 out_start:
523         rpc_call_start(task);
524         return 0;
525
526 out_sleep:
527         if (args->sa_privileged)
528                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
529                                 NULL, RPC_PRIORITY_PRIVILEGED);
530         else
531                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
532         spin_unlock(&tbl->slot_tbl_lock);
533         return -EAGAIN;
534 }
535
536 static int nfs40_sequence_done(struct rpc_task *task,
537                                struct nfs4_sequence_res *res)
538 {
539         struct nfs4_slot *slot = res->sr_slot;
540         struct nfs4_slot_table *tbl;
541
542         if (slot == NULL)
543                 goto out;
544
545         tbl = slot->table;
546         spin_lock(&tbl->slot_tbl_lock);
547         if (!nfs41_wake_and_assign_slot(tbl, slot))
548                 nfs4_free_slot(tbl, slot);
549         spin_unlock(&tbl->slot_tbl_lock);
550
551         res->sr_slot = NULL;
552 out:
553         return 1;
554 }
555
556 #if defined(CONFIG_NFS_V4_1)
557
558 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
559 {
560         struct nfs4_session *session;
561         struct nfs4_slot_table *tbl;
562         struct nfs4_slot *slot = res->sr_slot;
563         bool send_new_highest_used_slotid = false;
564
565         tbl = slot->table;
566         session = tbl->session;
567
568         spin_lock(&tbl->slot_tbl_lock);
569         /* Be nice to the server: try to ensure that the last transmitted
570          * value for highest_user_slotid <= target_highest_slotid
571          */
572         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
573                 send_new_highest_used_slotid = true;
574
575         if (nfs41_wake_and_assign_slot(tbl, slot)) {
576                 send_new_highest_used_slotid = false;
577                 goto out_unlock;
578         }
579         nfs4_free_slot(tbl, slot);
580
581         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
582                 send_new_highest_used_slotid = false;
583 out_unlock:
584         spin_unlock(&tbl->slot_tbl_lock);
585         res->sr_slot = NULL;
586         if (send_new_highest_used_slotid)
587                 nfs41_server_notify_highest_slotid_update(session->clp);
588 }
589
590 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
591 {
592         struct nfs4_session *session;
593         struct nfs4_slot *slot = res->sr_slot;
594         struct nfs_client *clp;
595         bool interrupted = false;
596         int ret = 1;
597
598         if (slot == NULL)
599                 goto out_noaction;
600         /* don't increment the sequence number if the task wasn't sent */
601         if (!RPC_WAS_SENT(task))
602                 goto out;
603
604         session = slot->table->session;
605
606         if (slot->interrupted) {
607                 slot->interrupted = 0;
608                 interrupted = true;
609         }
610
611         trace_nfs4_sequence_done(session, res);
612         /* Check the SEQUENCE operation status */
613         switch (res->sr_status) {
614         case 0:
615                 /* Update the slot's sequence and clientid lease timer */
616                 ++slot->seq_nr;
617                 clp = session->clp;
618                 do_renew_lease(clp, res->sr_timestamp);
619                 /* Check sequence flags */
620                 if (res->sr_status_flags != 0)
621                         nfs4_schedule_lease_recovery(clp);
622                 nfs41_update_target_slotid(slot->table, slot, res);
623                 break;
624         case 1:
625                 /*
626                  * sr_status remains 1 if an RPC level error occurred.
627                  * The server may or may not have processed the sequence
628                  * operation..
629                  * Mark the slot as having hosted an interrupted RPC call.
630                  */
631                 slot->interrupted = 1;
632                 goto out;
633         case -NFS4ERR_DELAY:
634                 /* The server detected a resend of the RPC call and
635                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
636                  * of RFC5661.
637                  */
638                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
639                         __func__,
640                         slot->slot_nr,
641                         slot->seq_nr);
642                 goto out_retry;
643         case -NFS4ERR_BADSLOT:
644                 /*
645                  * The slot id we used was probably retired. Try again
646                  * using a different slot id.
647                  */
648                 goto retry_nowait;
649         case -NFS4ERR_SEQ_MISORDERED:
650                 /*
651                  * Was the last operation on this sequence interrupted?
652                  * If so, retry after bumping the sequence number.
653                  */
654                 if (interrupted) {
655                         ++slot->seq_nr;
656                         goto retry_nowait;
657                 }
658                 /*
659                  * Could this slot have been previously retired?
660                  * If so, then the server may be expecting seq_nr = 1!
661                  */
662                 if (slot->seq_nr != 1) {
663                         slot->seq_nr = 1;
664                         goto retry_nowait;
665                 }
666                 break;
667         case -NFS4ERR_SEQ_FALSE_RETRY:
668                 ++slot->seq_nr;
669                 goto retry_nowait;
670         default:
671                 /* Just update the slot sequence no. */
672                 ++slot->seq_nr;
673         }
674 out:
675         /* The session may be reset by one of the error handlers. */
676         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
677         nfs41_sequence_free_slot(res);
678 out_noaction:
679         return ret;
680 retry_nowait:
681         if (rpc_restart_call_prepare(task)) {
682                 task->tk_status = 0;
683                 ret = 0;
684         }
685         goto out;
686 out_retry:
687         if (!rpc_restart_call(task))
688                 goto out;
689         rpc_delay(task, NFS4_POLL_RETRY_MAX);
690         return 0;
691 }
692 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
693
694 static int nfs4_sequence_done(struct rpc_task *task,
695                                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
704 int nfs41_setup_sequence(struct nfs4_session *session,
705                                 struct nfs4_sequence_args *args,
706                                 struct nfs4_sequence_res *res,
707                                 struct rpc_task *task)
708 {
709         struct nfs4_slot *slot;
710         struct nfs4_slot_table *tbl;
711
712         dprintk("--> %s\n", __func__);
713         /* slot already allocated? */
714         if (res->sr_slot != NULL)
715                 goto out_success;
716
717         tbl = &session->fc_slot_table;
718
719         task->tk_timeout = 0;
720
721         spin_lock(&tbl->slot_tbl_lock);
722         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
723             !args->sa_privileged) {
724                 /* The state manager will wait until the slot table is empty */
725                 dprintk("%s session is draining\n", __func__);
726                 goto out_sleep;
727         }
728
729         slot = nfs4_alloc_slot(tbl);
730         if (IS_ERR(slot)) {
731                 /* If out of memory, try again in 1/4 second */
732                 if (slot == ERR_PTR(-ENOMEM))
733                         task->tk_timeout = HZ >> 2;
734                 dprintk("<-- %s: no free slots\n", __func__);
735                 goto out_sleep;
736         }
737         spin_unlock(&tbl->slot_tbl_lock);
738
739         args->sa_slot = slot;
740
741         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
742                         slot->slot_nr, slot->seq_nr);
743
744         res->sr_slot = slot;
745         res->sr_timestamp = jiffies;
746         res->sr_status_flags = 0;
747         /*
748          * sr_status is only set in decode_sequence, and so will remain
749          * set to 1 if an rpc level failure occurs.
750          */
751         res->sr_status = 1;
752         trace_nfs4_setup_sequence(session, args);
753 out_success:
754         rpc_call_start(task);
755         return 0;
756 out_sleep:
757         /* Privileged tasks are queued with top priority */
758         if (args->sa_privileged)
759                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
760                                 NULL, RPC_PRIORITY_PRIVILEGED);
761         else
762                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
763         spin_unlock(&tbl->slot_tbl_lock);
764         return -EAGAIN;
765 }
766 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
767
768 static int nfs4_setup_sequence(const struct nfs_server *server,
769                                struct nfs4_sequence_args *args,
770                                struct nfs4_sequence_res *res,
771                                struct rpc_task *task)
772 {
773         struct nfs4_session *session = nfs4_get_session(server);
774         int ret = 0;
775
776         if (!session)
777                 return nfs40_setup_sequence(server, args, res, task);
778
779         dprintk("--> %s clp %p session %p sr_slot %u\n",
780                 __func__, session->clp, session, res->sr_slot ?
781                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
782
783         ret = nfs41_setup_sequence(session, args, res, task);
784
785         dprintk("<-- %s status=%d\n", __func__, ret);
786         return ret;
787 }
788
789 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
790 {
791         struct nfs4_call_sync_data *data = calldata;
792         struct nfs4_session *session = nfs4_get_session(data->seq_server);
793
794         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
795
796         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
797 }
798
799 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
800 {
801         struct nfs4_call_sync_data *data = calldata;
802
803         nfs41_sequence_done(task, data->seq_res);
804 }
805
806 static const struct rpc_call_ops nfs41_call_sync_ops = {
807         .rpc_call_prepare = nfs41_call_sync_prepare,
808         .rpc_call_done = nfs41_call_sync_done,
809 };
810
811 #else   /* !CONFIG_NFS_V4_1 */
812
813 static int nfs4_setup_sequence(const struct nfs_server *server,
814                                struct nfs4_sequence_args *args,
815                                struct nfs4_sequence_res *res,
816                                struct rpc_task *task)
817 {
818         return nfs40_setup_sequence(server, args, res, task);
819 }
820
821 static int nfs4_sequence_done(struct rpc_task *task,
822                                struct nfs4_sequence_res *res)
823 {
824         return nfs40_sequence_done(task, res);
825 }
826
827 #endif  /* !CONFIG_NFS_V4_1 */
828
829 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
830 {
831         struct nfs4_call_sync_data *data = calldata;
832         nfs4_setup_sequence(data->seq_server,
833                                 data->seq_args, data->seq_res, task);
834 }
835
836 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
837 {
838         struct nfs4_call_sync_data *data = calldata;
839         nfs4_sequence_done(task, data->seq_res);
840 }
841
842 static const struct rpc_call_ops nfs40_call_sync_ops = {
843         .rpc_call_prepare = nfs40_call_sync_prepare,
844         .rpc_call_done = nfs40_call_sync_done,
845 };
846
847 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
848                                    struct nfs_server *server,
849                                    struct rpc_message *msg,
850                                    struct nfs4_sequence_args *args,
851                                    struct nfs4_sequence_res *res)
852 {
853         int ret;
854         struct rpc_task *task;
855         struct nfs_client *clp = server->nfs_client;
856         struct nfs4_call_sync_data data = {
857                 .seq_server = server,
858                 .seq_args = args,
859                 .seq_res = res,
860         };
861         struct rpc_task_setup task_setup = {
862                 .rpc_client = clnt,
863                 .rpc_message = msg,
864                 .callback_ops = clp->cl_mvops->call_sync_ops,
865                 .callback_data = &data
866         };
867
868         task = rpc_run_task(&task_setup);
869         if (IS_ERR(task))
870                 ret = PTR_ERR(task);
871         else {
872                 ret = task->tk_status;
873                 rpc_put_task(task);
874         }
875         return ret;
876 }
877
878 static
879 int nfs4_call_sync(struct rpc_clnt *clnt,
880                    struct nfs_server *server,
881                    struct rpc_message *msg,
882                    struct nfs4_sequence_args *args,
883                    struct nfs4_sequence_res *res,
884                    int cache_reply)
885 {
886         nfs4_init_sequence(args, res, cache_reply);
887         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
888 }
889
890 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
891 {
892         struct nfs_inode *nfsi = NFS_I(dir);
893
894         spin_lock(&dir->i_lock);
895         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
896         if (!cinfo->atomic || cinfo->before != dir->i_version)
897                 nfs_force_lookup_revalidate(dir);
898         dir->i_version = cinfo->after;
899         nfs_fscache_invalidate(dir);
900         spin_unlock(&dir->i_lock);
901 }
902
903 struct nfs4_opendata {
904         struct kref kref;
905         struct nfs_openargs o_arg;
906         struct nfs_openres o_res;
907         struct nfs_open_confirmargs c_arg;
908         struct nfs_open_confirmres c_res;
909         struct nfs4_string owner_name;
910         struct nfs4_string group_name;
911         struct nfs_fattr f_attr;
912         struct nfs4_label *f_label;
913         struct dentry *dir;
914         struct dentry *dentry;
915         struct nfs4_state_owner *owner;
916         struct nfs4_state *state;
917         struct iattr attrs;
918         unsigned long timestamp;
919         unsigned int rpc_done : 1;
920         unsigned int file_created : 1;
921         unsigned int is_recover : 1;
922         int rpc_status;
923         int cancelled;
924 };
925
926 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
927                 int err, struct nfs4_exception *exception)
928 {
929         if (err != -EINVAL)
930                 return false;
931         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
932                 return false;
933         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
934         exception->retry = 1;
935         return true;
936 }
937
938 static enum open_claim_type4
939 nfs4_map_atomic_open_claim(struct nfs_server *server,
940                 enum open_claim_type4 claim)
941 {
942         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
943                 return claim;
944         switch (claim) {
945         default:
946                 return claim;
947         case NFS4_OPEN_CLAIM_FH:
948                 return NFS4_OPEN_CLAIM_NULL;
949         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
950                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
951         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
952                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
953         }
954 }
955
956 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
957 {
958         p->o_res.f_attr = &p->f_attr;
959         p->o_res.f_label = p->f_label;
960         p->o_res.seqid = p->o_arg.seqid;
961         p->c_res.seqid = p->c_arg.seqid;
962         p->o_res.server = p->o_arg.server;
963         p->o_res.access_request = p->o_arg.access;
964         nfs_fattr_init(&p->f_attr);
965         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
966 }
967
968 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
969                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
970                 const struct iattr *attrs,
971                 struct nfs4_label *label,
972                 enum open_claim_type4 claim,
973                 gfp_t gfp_mask)
974 {
975         struct dentry *parent = dget_parent(dentry);
976         struct inode *dir = parent->d_inode;
977         struct nfs_server *server = NFS_SERVER(dir);
978         struct nfs4_opendata *p;
979
980         p = kzalloc(sizeof(*p), gfp_mask);
981         if (p == NULL)
982                 goto err;
983
984         p->f_label = nfs4_label_alloc(server, gfp_mask);
985         if (IS_ERR(p->f_label))
986                 goto err_free_p;
987
988         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
989         if (p->o_arg.seqid == NULL)
990                 goto err_free_label;
991         nfs_sb_active(dentry->d_sb);
992         p->dentry = dget(dentry);
993         p->dir = parent;
994         p->owner = sp;
995         atomic_inc(&sp->so_count);
996         p->o_arg.open_flags = flags;
997         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
998         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
999          * will return permission denied for all bits until close */
1000         if (!(flags & O_EXCL)) {
1001                 /* ask server to check for all possible rights as results
1002                  * are cached */
1003                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1004                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1005         }
1006         p->o_arg.clientid = server->nfs_client->cl_clientid;
1007         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1008         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1009         p->o_arg.name = &dentry->d_name;
1010         p->o_arg.server = server;
1011         p->o_arg.bitmask = nfs4_bitmask(server, label);
1012         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1013         p->o_arg.label = label;
1014         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1015         switch (p->o_arg.claim) {
1016         case NFS4_OPEN_CLAIM_NULL:
1017         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1018         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1019                 p->o_arg.fh = NFS_FH(dir);
1020                 break;
1021         case NFS4_OPEN_CLAIM_PREVIOUS:
1022         case NFS4_OPEN_CLAIM_FH:
1023         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1024         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1025                 p->o_arg.fh = NFS_FH(dentry->d_inode);
1026         }
1027         if (attrs != NULL && attrs->ia_valid != 0) {
1028                 __u32 verf[2];
1029
1030                 p->o_arg.u.attrs = &p->attrs;
1031                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1032
1033                 verf[0] = jiffies;
1034                 verf[1] = current->pid;
1035                 memcpy(p->o_arg.u.verifier.data, verf,
1036                                 sizeof(p->o_arg.u.verifier.data));
1037         }
1038         p->c_arg.fh = &p->o_res.fh;
1039         p->c_arg.stateid = &p->o_res.stateid;
1040         p->c_arg.seqid = p->o_arg.seqid;
1041         nfs4_init_opendata_res(p);
1042         kref_init(&p->kref);
1043         return p;
1044
1045 err_free_label:
1046         nfs4_label_free(p->f_label);
1047 err_free_p:
1048         kfree(p);
1049 err:
1050         dput(parent);
1051         return NULL;
1052 }
1053
1054 static void nfs4_opendata_free(struct kref *kref)
1055 {
1056         struct nfs4_opendata *p = container_of(kref,
1057                         struct nfs4_opendata, kref);
1058         struct super_block *sb = p->dentry->d_sb;
1059
1060         nfs_free_seqid(p->o_arg.seqid);
1061         if (p->state != NULL)
1062                 nfs4_put_open_state(p->state);
1063         nfs4_put_state_owner(p->owner);
1064
1065         nfs4_label_free(p->f_label);
1066
1067         dput(p->dir);
1068         dput(p->dentry);
1069         nfs_sb_deactive(sb);
1070         nfs_fattr_free_names(&p->f_attr);
1071         kfree(p->f_attr.mdsthreshold);
1072         kfree(p);
1073 }
1074
1075 static void nfs4_opendata_put(struct nfs4_opendata *p)
1076 {
1077         if (p != NULL)
1078                 kref_put(&p->kref, nfs4_opendata_free);
1079 }
1080
1081 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1082 {
1083         int ret;
1084
1085         ret = rpc_wait_for_completion_task(task);
1086         return ret;
1087 }
1088
1089 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1090 {
1091         int ret = 0;
1092
1093         if (open_mode & (O_EXCL|O_TRUNC))
1094                 goto out;
1095         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1096                 case FMODE_READ:
1097                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1098                                 && state->n_rdonly != 0;
1099                         break;
1100                 case FMODE_WRITE:
1101                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1102                                 && state->n_wronly != 0;
1103                         break;
1104                 case FMODE_READ|FMODE_WRITE:
1105                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1106                                 && state->n_rdwr != 0;
1107         }
1108 out:
1109         return ret;
1110 }
1111
1112 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1113 {
1114         if (delegation == NULL)
1115                 return 0;
1116         if ((delegation->type & fmode) != fmode)
1117                 return 0;
1118         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1119                 return 0;
1120         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1121                 return 0;
1122         nfs_mark_delegation_referenced(delegation);
1123         return 1;
1124 }
1125
1126 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1127 {
1128         switch (fmode) {
1129                 case FMODE_WRITE:
1130                         state->n_wronly++;
1131                         break;
1132                 case FMODE_READ:
1133                         state->n_rdonly++;
1134                         break;
1135                 case FMODE_READ|FMODE_WRITE:
1136                         state->n_rdwr++;
1137         }
1138         nfs4_state_set_mode_locked(state, state->state | fmode);
1139 }
1140
1141 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1142 {
1143         struct nfs_client *clp = state->owner->so_server->nfs_client;
1144         bool need_recover = false;
1145
1146         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1147                 need_recover = true;
1148         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1149                 need_recover = true;
1150         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1151                 need_recover = true;
1152         if (need_recover)
1153                 nfs4_state_mark_reclaim_nograce(clp, state);
1154 }
1155
1156 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1157                 nfs4_stateid *stateid)
1158 {
1159         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1160                 return true;
1161         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1162                 nfs_test_and_clear_all_open_stateid(state);
1163                 return true;
1164         }
1165         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1166                 return true;
1167         return false;
1168 }
1169
1170 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1171                 nfs4_stateid *stateid, fmode_t fmode)
1172 {
1173         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1174         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1175         case FMODE_WRITE:
1176                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1177                 break;
1178         case FMODE_READ:
1179                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1180                 break;
1181         case 0:
1182                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1183                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1184                 clear_bit(NFS_OPEN_STATE, &state->flags);
1185         }
1186         if (stateid == NULL)
1187                 return;
1188         if (!nfs_need_update_open_stateid(state, stateid))
1189                 return;
1190         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1191                 nfs4_stateid_copy(&state->stateid, stateid);
1192         nfs4_stateid_copy(&state->open_stateid, stateid);
1193 }
1194
1195 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1196 {
1197         write_seqlock(&state->seqlock);
1198         nfs_clear_open_stateid_locked(state, stateid, fmode);
1199         write_sequnlock(&state->seqlock);
1200         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1201                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1202 }
1203
1204 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1205 {
1206         switch (fmode) {
1207                 case FMODE_READ:
1208                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1209                         break;
1210                 case FMODE_WRITE:
1211                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1212                         break;
1213                 case FMODE_READ|FMODE_WRITE:
1214                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1215         }
1216         if (!nfs_need_update_open_stateid(state, stateid))
1217                 return;
1218         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1219                 nfs4_stateid_copy(&state->stateid, stateid);
1220         nfs4_stateid_copy(&state->open_stateid, stateid);
1221 }
1222
1223 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1224 {
1225         /*
1226          * Protect the call to nfs4_state_set_mode_locked and
1227          * serialise the stateid update
1228          */
1229         write_seqlock(&state->seqlock);
1230         if (deleg_stateid != NULL) {
1231                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1232                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1233         }
1234         if (open_stateid != NULL)
1235                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1236         write_sequnlock(&state->seqlock);
1237         spin_lock(&state->owner->so_lock);
1238         update_open_stateflags(state, fmode);
1239         spin_unlock(&state->owner->so_lock);
1240 }
1241
1242 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1243 {
1244         struct nfs_inode *nfsi = NFS_I(state->inode);
1245         struct nfs_delegation *deleg_cur;
1246         int ret = 0;
1247
1248         fmode &= (FMODE_READ|FMODE_WRITE);
1249
1250         rcu_read_lock();
1251         deleg_cur = rcu_dereference(nfsi->delegation);
1252         if (deleg_cur == NULL)
1253                 goto no_delegation;
1254
1255         spin_lock(&deleg_cur->lock);
1256         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1257            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1258             (deleg_cur->type & fmode) != fmode)
1259                 goto no_delegation_unlock;
1260
1261         if (delegation == NULL)
1262                 delegation = &deleg_cur->stateid;
1263         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1264                 goto no_delegation_unlock;
1265
1266         nfs_mark_delegation_referenced(deleg_cur);
1267         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1268         ret = 1;
1269 no_delegation_unlock:
1270         spin_unlock(&deleg_cur->lock);
1271 no_delegation:
1272         rcu_read_unlock();
1273
1274         if (!ret && open_stateid != NULL) {
1275                 __update_open_stateid(state, open_stateid, NULL, fmode);
1276                 ret = 1;
1277         }
1278         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1279                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1280
1281         return ret;
1282 }
1283
1284
1285 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1286 {
1287         struct nfs_delegation *delegation;
1288
1289         rcu_read_lock();
1290         delegation = rcu_dereference(NFS_I(inode)->delegation);
1291         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1292                 rcu_read_unlock();
1293                 return;
1294         }
1295         rcu_read_unlock();
1296         nfs4_inode_return_delegation(inode);
1297 }
1298
1299 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1300 {
1301         struct nfs4_state *state = opendata->state;
1302         struct nfs_inode *nfsi = NFS_I(state->inode);
1303         struct nfs_delegation *delegation;
1304         int open_mode = opendata->o_arg.open_flags;
1305         fmode_t fmode = opendata->o_arg.fmode;
1306         nfs4_stateid stateid;
1307         int ret = -EAGAIN;
1308
1309         for (;;) {
1310                 if (can_open_cached(state, fmode, open_mode)) {
1311                         spin_lock(&state->owner->so_lock);
1312                         if (can_open_cached(state, fmode, open_mode)) {
1313                                 update_open_stateflags(state, fmode);
1314                                 spin_unlock(&state->owner->so_lock);
1315                                 goto out_return_state;
1316                         }
1317                         spin_unlock(&state->owner->so_lock);
1318                 }
1319                 rcu_read_lock();
1320                 delegation = rcu_dereference(nfsi->delegation);
1321                 if (!can_open_delegated(delegation, fmode)) {
1322                         rcu_read_unlock();
1323                         break;
1324                 }
1325                 /* Save the delegation */
1326                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1327                 rcu_read_unlock();
1328                 nfs_release_seqid(opendata->o_arg.seqid);
1329                 if (!opendata->is_recover) {
1330                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1331                         if (ret != 0)
1332                                 goto out;
1333                 }
1334                 ret = -EAGAIN;
1335
1336                 /* Try to update the stateid using the delegation */
1337                 if (update_open_stateid(state, NULL, &stateid, fmode))
1338                         goto out_return_state;
1339         }
1340 out:
1341         return ERR_PTR(ret);
1342 out_return_state:
1343         atomic_inc(&state->count);
1344         return state;
1345 }
1346
1347 static void
1348 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1349 {
1350         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1351         struct nfs_delegation *delegation;
1352         int delegation_flags = 0;
1353
1354         rcu_read_lock();
1355         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1356         if (delegation)
1357                 delegation_flags = delegation->flags;
1358         rcu_read_unlock();
1359         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1360                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1361                                    "returning a delegation for "
1362                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1363                                    clp->cl_hostname);
1364         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1365                 nfs_inode_set_delegation(state->inode,
1366                                          data->owner->so_cred,
1367                                          &data->o_res);
1368         else
1369                 nfs_inode_reclaim_delegation(state->inode,
1370                                              data->owner->so_cred,
1371                                              &data->o_res);
1372 }
1373
1374 /*
1375  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1376  * and update the nfs4_state.
1377  */
1378 static struct nfs4_state *
1379 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1380 {
1381         struct inode *inode = data->state->inode;
1382         struct nfs4_state *state = data->state;
1383         int ret;
1384
1385         if (!data->rpc_done) {
1386                 if (data->rpc_status) {
1387                         ret = data->rpc_status;
1388                         goto err;
1389                 }
1390                 /* cached opens have already been processed */
1391                 goto update;
1392         }
1393
1394         ret = nfs_refresh_inode(inode, &data->f_attr);
1395         if (ret)
1396                 goto err;
1397
1398         if (data->o_res.delegation_type != 0)
1399                 nfs4_opendata_check_deleg(data, state);
1400 update:
1401         update_open_stateid(state, &data->o_res.stateid, NULL,
1402                             data->o_arg.fmode);
1403         atomic_inc(&state->count);
1404
1405         return state;
1406 err:
1407         return ERR_PTR(ret);
1408
1409 }
1410
1411 static struct nfs4_state *
1412 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1413 {
1414         struct inode *inode;
1415         struct nfs4_state *state = NULL;
1416         int ret;
1417
1418         if (!data->rpc_done) {
1419                 state = nfs4_try_open_cached(data);
1420                 goto out;
1421         }
1422
1423         ret = -EAGAIN;
1424         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1425                 goto err;
1426         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1427         ret = PTR_ERR(inode);
1428         if (IS_ERR(inode))
1429                 goto err;
1430         ret = -ENOMEM;
1431         state = nfs4_get_open_state(inode, data->owner);
1432         if (state == NULL)
1433                 goto err_put_inode;
1434         if (data->o_res.delegation_type != 0)
1435                 nfs4_opendata_check_deleg(data, state);
1436         update_open_stateid(state, &data->o_res.stateid, NULL,
1437                         data->o_arg.fmode);
1438         iput(inode);
1439 out:
1440         nfs_release_seqid(data->o_arg.seqid);
1441         return state;
1442 err_put_inode:
1443         iput(inode);
1444 err:
1445         return ERR_PTR(ret);
1446 }
1447
1448 static struct nfs4_state *
1449 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1450 {
1451         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1452                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1453         return _nfs4_opendata_to_nfs4_state(data);
1454 }
1455
1456 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1457 {
1458         struct nfs_inode *nfsi = NFS_I(state->inode);
1459         struct nfs_open_context *ctx;
1460
1461         spin_lock(&state->inode->i_lock);
1462         list_for_each_entry(ctx, &nfsi->open_files, list) {
1463                 if (ctx->state != state)
1464                         continue;
1465                 get_nfs_open_context(ctx);
1466                 spin_unlock(&state->inode->i_lock);
1467                 return ctx;
1468         }
1469         spin_unlock(&state->inode->i_lock);
1470         return ERR_PTR(-ENOENT);
1471 }
1472
1473 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1474                 struct nfs4_state *state, enum open_claim_type4 claim)
1475 {
1476         struct nfs4_opendata *opendata;
1477
1478         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1479                         NULL, NULL, claim, GFP_NOFS);
1480         if (opendata == NULL)
1481                 return ERR_PTR(-ENOMEM);
1482         opendata->state = state;
1483         atomic_inc(&state->count);
1484         return opendata;
1485 }
1486
1487 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1488 {
1489         struct nfs4_state *newstate;
1490         int ret;
1491
1492         opendata->o_arg.open_flags = 0;
1493         opendata->o_arg.fmode = fmode;
1494         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1495         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1496         nfs4_init_opendata_res(opendata);
1497         ret = _nfs4_recover_proc_open(opendata);
1498         if (ret != 0)
1499                 return ret; 
1500         newstate = nfs4_opendata_to_nfs4_state(opendata);
1501         if (IS_ERR(newstate))
1502                 return PTR_ERR(newstate);
1503         nfs4_close_state(newstate, fmode);
1504         *res = newstate;
1505         return 0;
1506 }
1507
1508 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1509 {
1510         struct nfs4_state *newstate;
1511         int ret;
1512
1513         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1514         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1515         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1516         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1517         /* memory barrier prior to reading state->n_* */
1518         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1519         clear_bit(NFS_OPEN_STATE, &state->flags);
1520         smp_rmb();
1521         if (state->n_rdwr != 0) {
1522                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1523                 if (ret != 0)
1524                         return ret;
1525                 if (newstate != state)
1526                         return -ESTALE;
1527         }
1528         if (state->n_wronly != 0) {
1529                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1530                 if (ret != 0)
1531                         return ret;
1532                 if (newstate != state)
1533                         return -ESTALE;
1534         }
1535         if (state->n_rdonly != 0) {
1536                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1537                 if (ret != 0)
1538                         return ret;
1539                 if (newstate != state)
1540                         return -ESTALE;
1541         }
1542         /*
1543          * We may have performed cached opens for all three recoveries.
1544          * Check if we need to update the current stateid.
1545          */
1546         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1547             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1548                 write_seqlock(&state->seqlock);
1549                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1550                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1551                 write_sequnlock(&state->seqlock);
1552         }
1553         return 0;
1554 }
1555
1556 /*
1557  * OPEN_RECLAIM:
1558  *      reclaim state on the server after a reboot.
1559  */
1560 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1561 {
1562         struct nfs_delegation *delegation;
1563         struct nfs4_opendata *opendata;
1564         fmode_t delegation_type = 0;
1565         int status;
1566
1567         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1568                         NFS4_OPEN_CLAIM_PREVIOUS);
1569         if (IS_ERR(opendata))
1570                 return PTR_ERR(opendata);
1571         rcu_read_lock();
1572         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1573         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1574                 delegation_type = delegation->type;
1575         rcu_read_unlock();
1576         opendata->o_arg.u.delegation_type = delegation_type;
1577         status = nfs4_open_recover(opendata, state);
1578         nfs4_opendata_put(opendata);
1579         return status;
1580 }
1581
1582 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1583 {
1584         struct nfs_server *server = NFS_SERVER(state->inode);
1585         struct nfs4_exception exception = { };
1586         int err;
1587         do {
1588                 err = _nfs4_do_open_reclaim(ctx, state);
1589                 trace_nfs4_open_reclaim(ctx, 0, err);
1590                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1591                         continue;
1592                 if (err != -NFS4ERR_DELAY)
1593                         break;
1594                 nfs4_handle_exception(server, err, &exception);
1595         } while (exception.retry);
1596         return err;
1597 }
1598
1599 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1600 {
1601         struct nfs_open_context *ctx;
1602         int ret;
1603
1604         ctx = nfs4_state_find_open_context(state);
1605         if (IS_ERR(ctx))
1606                 return -EAGAIN;
1607         ret = nfs4_do_open_reclaim(ctx, state);
1608         put_nfs_open_context(ctx);
1609         return ret;
1610 }
1611
1612 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1613 {
1614         switch (err) {
1615                 default:
1616                         printk(KERN_ERR "NFS: %s: unhandled error "
1617                                         "%d.\n", __func__, err);
1618                 case 0:
1619                 case -ENOENT:
1620                 case -ESTALE:
1621                         break;
1622                 case -NFS4ERR_BADSESSION:
1623                 case -NFS4ERR_BADSLOT:
1624                 case -NFS4ERR_BAD_HIGH_SLOT:
1625                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1626                 case -NFS4ERR_DEADSESSION:
1627                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1628                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1629                         return -EAGAIN;
1630                 case -NFS4ERR_STALE_CLIENTID:
1631                 case -NFS4ERR_STALE_STATEID:
1632                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1633                 case -NFS4ERR_EXPIRED:
1634                         /* Don't recall a delegation if it was lost */
1635                         nfs4_schedule_lease_recovery(server->nfs_client);
1636                         return -EAGAIN;
1637                 case -NFS4ERR_MOVED:
1638                         nfs4_schedule_migration_recovery(server);
1639                         return -EAGAIN;
1640                 case -NFS4ERR_LEASE_MOVED:
1641                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1642                         return -EAGAIN;
1643                 case -NFS4ERR_DELEG_REVOKED:
1644                 case -NFS4ERR_ADMIN_REVOKED:
1645                 case -NFS4ERR_BAD_STATEID:
1646                 case -NFS4ERR_OPENMODE:
1647                         nfs_inode_find_state_and_recover(state->inode,
1648                                         stateid);
1649                         nfs4_schedule_stateid_recovery(server, state);
1650                         return 0;
1651                 case -NFS4ERR_DELAY:
1652                 case -NFS4ERR_GRACE:
1653                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1654                         ssleep(1);
1655                         return -EAGAIN;
1656                 case -ENOMEM:
1657                 case -NFS4ERR_DENIED:
1658                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1659                         return 0;
1660         }
1661         return err;
1662 }
1663
1664 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1665 {
1666         struct nfs_server *server = NFS_SERVER(state->inode);
1667         struct nfs4_opendata *opendata;
1668         int err;
1669
1670         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1671                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1672         if (IS_ERR(opendata))
1673                 return PTR_ERR(opendata);
1674         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1675         err = nfs4_open_recover(opendata, state);
1676         nfs4_opendata_put(opendata);
1677         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1678 }
1679
1680 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1681 {
1682         struct nfs4_opendata *data = calldata;
1683
1684         nfs40_setup_sequence(data->o_arg.server, &data->c_arg.seq_args,
1685                                 &data->c_res.seq_res, task);
1686 }
1687
1688 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1689 {
1690         struct nfs4_opendata *data = calldata;
1691
1692         nfs40_sequence_done(task, &data->c_res.seq_res);
1693
1694         data->rpc_status = task->tk_status;
1695         if (data->rpc_status == 0) {
1696                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1697                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1698                 renew_lease(data->o_res.server, data->timestamp);
1699                 data->rpc_done = 1;
1700         }
1701 }
1702
1703 static void nfs4_open_confirm_release(void *calldata)
1704 {
1705         struct nfs4_opendata *data = calldata;
1706         struct nfs4_state *state = NULL;
1707
1708         /* If this request hasn't been cancelled, do nothing */
1709         if (data->cancelled == 0)
1710                 goto out_free;
1711         /* In case of error, no cleanup! */
1712         if (!data->rpc_done)
1713                 goto out_free;
1714         state = nfs4_opendata_to_nfs4_state(data);
1715         if (!IS_ERR(state))
1716                 nfs4_close_state(state, data->o_arg.fmode);
1717 out_free:
1718         nfs4_opendata_put(data);
1719 }
1720
1721 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1722         .rpc_call_prepare = nfs4_open_confirm_prepare,
1723         .rpc_call_done = nfs4_open_confirm_done,
1724         .rpc_release = nfs4_open_confirm_release,
1725 };
1726
1727 /*
1728  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1729  */
1730 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1731 {
1732         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1733         struct rpc_task *task;
1734         struct  rpc_message msg = {
1735                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1736                 .rpc_argp = &data->c_arg,
1737                 .rpc_resp = &data->c_res,
1738                 .rpc_cred = data->owner->so_cred,
1739         };
1740         struct rpc_task_setup task_setup_data = {
1741                 .rpc_client = server->client,
1742                 .rpc_message = &msg,
1743                 .callback_ops = &nfs4_open_confirm_ops,
1744                 .callback_data = data,
1745                 .workqueue = nfsiod_workqueue,
1746                 .flags = RPC_TASK_ASYNC,
1747         };
1748         int status;
1749
1750         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1751         kref_get(&data->kref);
1752         data->rpc_done = 0;
1753         data->rpc_status = 0;
1754         data->timestamp = jiffies;
1755         task = rpc_run_task(&task_setup_data);
1756         if (IS_ERR(task))
1757                 return PTR_ERR(task);
1758         status = nfs4_wait_for_completion_rpc_task(task);
1759         if (status != 0) {
1760                 data->cancelled = 1;
1761                 smp_wmb();
1762         } else
1763                 status = data->rpc_status;
1764         rpc_put_task(task);
1765         return status;
1766 }
1767
1768 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1769 {
1770         struct nfs4_opendata *data = calldata;
1771         struct nfs4_state_owner *sp = data->owner;
1772         struct nfs_client *clp = sp->so_server->nfs_client;
1773
1774         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1775                 goto out_wait;
1776         /*
1777          * Check if we still need to send an OPEN call, or if we can use
1778          * a delegation instead.
1779          */
1780         if (data->state != NULL) {
1781                 struct nfs_delegation *delegation;
1782
1783                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1784                         goto out_no_action;
1785                 rcu_read_lock();
1786                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1787                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1788                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1789                     can_open_delegated(delegation, data->o_arg.fmode))
1790                         goto unlock_no_action;
1791                 rcu_read_unlock();
1792         }
1793         /* Update client id. */
1794         data->o_arg.clientid = clp->cl_clientid;
1795         switch (data->o_arg.claim) {
1796         case NFS4_OPEN_CLAIM_PREVIOUS:
1797         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1798         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1799                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1800         case NFS4_OPEN_CLAIM_FH:
1801                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1802                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1803         }
1804         data->timestamp = jiffies;
1805         if (nfs4_setup_sequence(data->o_arg.server,
1806                                 &data->o_arg.seq_args,
1807                                 &data->o_res.seq_res,
1808                                 task) != 0)
1809                 nfs_release_seqid(data->o_arg.seqid);
1810
1811         /* Set the create mode (note dependency on the session type) */
1812         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1813         if (data->o_arg.open_flags & O_EXCL) {
1814                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1815                 if (nfs4_has_persistent_session(clp))
1816                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1817                 else if (clp->cl_mvops->minor_version > 0)
1818                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1819         }
1820         return;
1821 unlock_no_action:
1822         rcu_read_unlock();
1823 out_no_action:
1824         task->tk_action = NULL;
1825 out_wait:
1826         nfs4_sequence_done(task, &data->o_res.seq_res);
1827 }
1828
1829 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1830 {
1831         struct nfs4_opendata *data = calldata;
1832
1833         data->rpc_status = task->tk_status;
1834
1835         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1836                 return;
1837
1838         if (task->tk_status == 0) {
1839                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1840                         switch (data->o_res.f_attr->mode & S_IFMT) {
1841                         case S_IFREG:
1842                                 break;
1843                         case S_IFLNK:
1844                                 data->rpc_status = -ELOOP;
1845                                 break;
1846                         case S_IFDIR:
1847                                 data->rpc_status = -EISDIR;
1848                                 break;
1849                         default:
1850                                 data->rpc_status = -ENOTDIR;
1851                         }
1852                 }
1853                 renew_lease(data->o_res.server, data->timestamp);
1854                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1855                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1856         }
1857         data->rpc_done = 1;
1858 }
1859
1860 static void nfs4_open_release(void *calldata)
1861 {
1862         struct nfs4_opendata *data = calldata;
1863         struct nfs4_state *state = NULL;
1864
1865         /* If this request hasn't been cancelled, do nothing */
1866         if (data->cancelled == 0)
1867                 goto out_free;
1868         /* In case of error, no cleanup! */
1869         if (data->rpc_status != 0 || !data->rpc_done)
1870                 goto out_free;
1871         /* In case we need an open_confirm, no cleanup! */
1872         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1873                 goto out_free;
1874         state = nfs4_opendata_to_nfs4_state(data);
1875         if (!IS_ERR(state))
1876                 nfs4_close_state(state, data->o_arg.fmode);
1877 out_free:
1878         nfs4_opendata_put(data);
1879 }
1880
1881 static const struct rpc_call_ops nfs4_open_ops = {
1882         .rpc_call_prepare = nfs4_open_prepare,
1883         .rpc_call_done = nfs4_open_done,
1884         .rpc_release = nfs4_open_release,
1885 };
1886
1887 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1888 {
1889         struct inode *dir = data->dir->d_inode;
1890         struct nfs_server *server = NFS_SERVER(dir);
1891         struct nfs_openargs *o_arg = &data->o_arg;
1892         struct nfs_openres *o_res = &data->o_res;
1893         struct rpc_task *task;
1894         struct rpc_message msg = {
1895                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1896                 .rpc_argp = o_arg,
1897                 .rpc_resp = o_res,
1898                 .rpc_cred = data->owner->so_cred,
1899         };
1900         struct rpc_task_setup task_setup_data = {
1901                 .rpc_client = server->client,
1902                 .rpc_message = &msg,
1903                 .callback_ops = &nfs4_open_ops,
1904                 .callback_data = data,
1905                 .workqueue = nfsiod_workqueue,
1906                 .flags = RPC_TASK_ASYNC,
1907         };
1908         int status;
1909
1910         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1911         kref_get(&data->kref);
1912         data->rpc_done = 0;
1913         data->rpc_status = 0;
1914         data->cancelled = 0;
1915         data->is_recover = 0;
1916         if (isrecover) {
1917                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1918                 data->is_recover = 1;
1919         }
1920         task = rpc_run_task(&task_setup_data);
1921         if (IS_ERR(task))
1922                 return PTR_ERR(task);
1923         status = nfs4_wait_for_completion_rpc_task(task);
1924         if (status != 0) {
1925                 data->cancelled = 1;
1926                 smp_wmb();
1927         } else
1928                 status = data->rpc_status;
1929         rpc_put_task(task);
1930
1931         return status;
1932 }
1933
1934 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1935 {
1936         struct inode *dir = data->dir->d_inode;
1937         struct nfs_openres *o_res = &data->o_res;
1938         int status;
1939
1940         status = nfs4_run_open_task(data, 1);
1941         if (status != 0 || !data->rpc_done)
1942                 return status;
1943
1944         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1945
1946         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1947                 status = _nfs4_proc_open_confirm(data);
1948                 if (status != 0)
1949                         return status;
1950         }
1951
1952         return status;
1953 }
1954
1955 /*
1956  * Additional permission checks in order to distinguish between an
1957  * open for read, and an open for execute. This works around the
1958  * fact that NFSv4 OPEN treats read and execute permissions as being
1959  * the same.
1960  * Note that in the non-execute case, we want to turn off permission
1961  * checking if we just created a new file (POSIX open() semantics).
1962  */
1963 static int nfs4_opendata_access(struct rpc_cred *cred,
1964                                 struct nfs4_opendata *opendata,
1965                                 struct nfs4_state *state, fmode_t fmode,
1966                                 int openflags)
1967 {
1968         struct nfs_access_entry cache;
1969         u32 mask;
1970
1971         /* access call failed or for some reason the server doesn't
1972          * support any access modes -- defer access call until later */
1973         if (opendata->o_res.access_supported == 0)
1974                 return 0;
1975
1976         mask = 0;
1977         /*
1978          * Use openflags to check for exec, because fmode won't
1979          * always have FMODE_EXEC set when file open for exec.
1980          */
1981         if (openflags & __FMODE_EXEC) {
1982                 /* ONLY check for exec rights */
1983                 mask = MAY_EXEC;
1984         } else if ((fmode & FMODE_READ) && !opendata->file_created)
1985                 mask = MAY_READ;
1986
1987         cache.cred = cred;
1988         cache.jiffies = jiffies;
1989         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1990         nfs_access_add_cache(state->inode, &cache);
1991
1992         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1993                 return 0;
1994
1995         /* even though OPEN succeeded, access is denied. Close the file */
1996         nfs4_close_state(state, fmode);
1997         return -EACCES;
1998 }
1999
2000 /*
2001  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2002  */
2003 static int _nfs4_proc_open(struct nfs4_opendata *data)
2004 {
2005         struct inode *dir = data->dir->d_inode;
2006         struct nfs_server *server = NFS_SERVER(dir);
2007         struct nfs_openargs *o_arg = &data->o_arg;
2008         struct nfs_openres *o_res = &data->o_res;
2009         int status;
2010
2011         status = nfs4_run_open_task(data, 0);
2012         if (!data->rpc_done)
2013                 return status;
2014         if (status != 0) {
2015                 if (status == -NFS4ERR_BADNAME &&
2016                                 !(o_arg->open_flags & O_CREAT))
2017                         return -ENOENT;
2018                 return status;
2019         }
2020
2021         nfs_fattr_map_and_free_names(server, &data->f_attr);
2022
2023         if (o_arg->open_flags & O_CREAT) {
2024                 update_changeattr(dir, &o_res->cinfo);
2025                 if (o_arg->open_flags & O_EXCL)
2026                         data->file_created = 1;
2027                 else if (o_res->cinfo.before != o_res->cinfo.after)
2028                         data->file_created = 1;
2029         }
2030         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2031                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2032         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2033                 status = _nfs4_proc_open_confirm(data);
2034                 if (status != 0)
2035                         return status;
2036         }
2037         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2038                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2039         return 0;
2040 }
2041
2042 static int nfs4_recover_expired_lease(struct nfs_server *server)
2043 {
2044         return nfs4_client_recover_expired_lease(server->nfs_client);
2045 }
2046
2047 /*
2048  * OPEN_EXPIRED:
2049  *      reclaim state on the server after a network partition.
2050  *      Assumes caller holds the appropriate lock
2051  */
2052 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2053 {
2054         struct nfs4_opendata *opendata;
2055         int ret;
2056
2057         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2058                         NFS4_OPEN_CLAIM_FH);
2059         if (IS_ERR(opendata))
2060                 return PTR_ERR(opendata);
2061         ret = nfs4_open_recover(opendata, state);
2062         if (ret == -ESTALE)
2063                 d_drop(ctx->dentry);
2064         nfs4_opendata_put(opendata);
2065         return ret;
2066 }
2067
2068 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2069 {
2070         struct nfs_server *server = NFS_SERVER(state->inode);
2071         struct nfs4_exception exception = { };
2072         int err;
2073
2074         do {
2075                 err = _nfs4_open_expired(ctx, state);
2076                 trace_nfs4_open_expired(ctx, 0, err);
2077                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2078                         continue;
2079                 switch (err) {
2080                 default:
2081                         goto out;
2082                 case -NFS4ERR_GRACE:
2083                 case -NFS4ERR_DELAY:
2084                         nfs4_handle_exception(server, err, &exception);
2085                         err = 0;
2086                 }
2087         } while (exception.retry);
2088 out:
2089         return err;
2090 }
2091
2092 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2093 {
2094         struct nfs_open_context *ctx;
2095         int ret;
2096
2097         ctx = nfs4_state_find_open_context(state);
2098         if (IS_ERR(ctx))
2099                 return -EAGAIN;
2100         ret = nfs4_do_open_expired(ctx, state);
2101         put_nfs_open_context(ctx);
2102         return ret;
2103 }
2104
2105 #if defined(CONFIG_NFS_V4_1)
2106 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
2107 {
2108         struct nfs_server *server = NFS_SERVER(state->inode);
2109         nfs4_stateid *stateid = &state->stateid;
2110         struct nfs_delegation *delegation;
2111         struct rpc_cred *cred = NULL;
2112         int status = -NFS4ERR_BAD_STATEID;
2113
2114         /* If a state reset has been done, test_stateid is unneeded */
2115         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2116                 return;
2117
2118         /* Get the delegation credential for use by test/free_stateid */
2119         rcu_read_lock();
2120         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2121         if (delegation != NULL &&
2122             nfs4_stateid_match(&delegation->stateid, stateid)) {
2123                 cred = get_rpccred(delegation->cred);
2124                 rcu_read_unlock();
2125                 status = nfs41_test_stateid(server, stateid, cred);
2126                 trace_nfs4_test_delegation_stateid(state, NULL, status);
2127         } else
2128                 rcu_read_unlock();
2129
2130         if (status != NFS_OK) {
2131                 /* Free the stateid unless the server explicitly
2132                  * informs us the stateid is unrecognized. */
2133                 if (status != -NFS4ERR_BAD_STATEID)
2134                         nfs41_free_stateid(server, stateid, cred);
2135                 nfs_remove_bad_delegation(state->inode);
2136
2137                 write_seqlock(&state->seqlock);
2138                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2139                 write_sequnlock(&state->seqlock);
2140                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
2141         }
2142
2143         if (cred != NULL)
2144                 put_rpccred(cred);
2145 }
2146
2147 /**
2148  * nfs41_check_open_stateid - possibly free an open stateid
2149  *
2150  * @state: NFSv4 state for an inode
2151  *
2152  * Returns NFS_OK if recovery for this stateid is now finished.
2153  * Otherwise a negative NFS4ERR value is returned.
2154  */
2155 static int nfs41_check_open_stateid(struct nfs4_state *state)
2156 {
2157         struct nfs_server *server = NFS_SERVER(state->inode);
2158         nfs4_stateid *stateid = &state->open_stateid;
2159         struct rpc_cred *cred = state->owner->so_cred;
2160         int status;
2161
2162         /* If a state reset has been done, test_stateid is unneeded */
2163         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2164             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2165             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2166                 return -NFS4ERR_BAD_STATEID;
2167
2168         status = nfs41_test_stateid(server, stateid, cred);
2169         trace_nfs4_test_open_stateid(state, NULL, status);
2170         if (status != NFS_OK) {
2171                 /* Free the stateid unless the server explicitly
2172                  * informs us the stateid is unrecognized. */
2173                 if (status != -NFS4ERR_BAD_STATEID)
2174                         nfs41_free_stateid(server, stateid, cred);
2175
2176                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2177                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2178                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2179                 clear_bit(NFS_OPEN_STATE, &state->flags);
2180         }
2181         return status;
2182 }
2183
2184 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2185 {
2186         int status;
2187
2188         nfs41_clear_delegation_stateid(state);
2189         status = nfs41_check_open_stateid(state);
2190         if (status != NFS_OK)
2191                 status = nfs4_open_expired(sp, state);
2192         return status;
2193 }
2194 #endif
2195
2196 /*
2197  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2198  * fields corresponding to attributes that were used to store the verifier.
2199  * Make sure we clobber those fields in the later setattr call
2200  */
2201 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2202 {
2203         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2204             !(sattr->ia_valid & ATTR_ATIME_SET))
2205                 sattr->ia_valid |= ATTR_ATIME;
2206
2207         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2208             !(sattr->ia_valid & ATTR_MTIME_SET))
2209                 sattr->ia_valid |= ATTR_MTIME;
2210 }
2211
2212 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2213                 fmode_t fmode,
2214                 int flags,
2215                 struct nfs_open_context *ctx)
2216 {
2217         struct nfs4_state_owner *sp = opendata->owner;
2218         struct nfs_server *server = sp->so_server;
2219         struct dentry *dentry;
2220         struct nfs4_state *state;
2221         unsigned int seq;
2222         int ret;
2223
2224         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2225
2226         ret = _nfs4_proc_open(opendata);
2227         if (ret != 0) {
2228                 if (ret == -ENOENT) {
2229                         dentry = opendata->dentry;
2230                         if (dentry->d_inode)
2231                                 d_delete(dentry);
2232                         else if (d_unhashed(dentry))
2233                                 d_add(dentry, NULL);
2234
2235                         nfs_set_verifier(dentry,
2236                                          nfs_save_change_attribute(opendata->dir->d_inode));
2237                 }
2238                 goto out;
2239         }
2240
2241         state = nfs4_opendata_to_nfs4_state(opendata);
2242         ret = PTR_ERR(state);
2243         if (IS_ERR(state))
2244                 goto out;
2245         if (server->caps & NFS_CAP_POSIX_LOCK)
2246                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2247
2248         dentry = opendata->dentry;
2249         if (dentry->d_inode == NULL) {
2250                 /* FIXME: Is this d_drop() ever needed? */
2251                 d_drop(dentry);
2252                 dentry = d_add_unique(dentry, igrab(state->inode));
2253                 if (dentry == NULL) {
2254                         dentry = opendata->dentry;
2255                 } else if (dentry != ctx->dentry) {
2256                         dput(ctx->dentry);
2257                         ctx->dentry = dget(dentry);
2258                 }
2259                 nfs_set_verifier(dentry,
2260                                 nfs_save_change_attribute(opendata->dir->d_inode));
2261         }
2262
2263         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2264         if (ret != 0)
2265                 goto out;
2266
2267         ctx->state = state;
2268         if (dentry->d_inode == state->inode) {
2269                 nfs_inode_attach_open_context(ctx);
2270                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2271                         nfs4_schedule_stateid_recovery(server, state);
2272         }
2273 out:
2274         return ret;
2275 }
2276
2277 /*
2278  * Returns a referenced nfs4_state
2279  */
2280 static int _nfs4_do_open(struct inode *dir,
2281                         struct nfs_open_context *ctx,
2282                         int flags,
2283                         struct iattr *sattr,
2284                         struct nfs4_label *label,
2285                         int *opened)
2286 {
2287         struct nfs4_state_owner  *sp;
2288         struct nfs4_state     *state = NULL;
2289         struct nfs_server       *server = NFS_SERVER(dir);
2290         struct nfs4_opendata *opendata;
2291         struct dentry *dentry = ctx->dentry;
2292         struct rpc_cred *cred = ctx->cred;
2293         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2294         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2295         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2296         struct nfs4_label *olabel = NULL;
2297         int status;
2298
2299         /* Protect against reboot recovery conflicts */
2300         status = -ENOMEM;
2301         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2302         if (sp == NULL) {
2303                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2304                 goto out_err;
2305         }
2306         status = nfs4_recover_expired_lease(server);
2307         if (status != 0)
2308                 goto err_put_state_owner;
2309         if (dentry->d_inode != NULL)
2310                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2311         status = -ENOMEM;
2312         if (dentry->d_inode)
2313                 claim = NFS4_OPEN_CLAIM_FH;
2314         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2315                         label, claim, GFP_KERNEL);
2316         if (opendata == NULL)
2317                 goto err_put_state_owner;
2318
2319         if (label) {
2320                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2321                 if (IS_ERR(olabel)) {
2322                         status = PTR_ERR(olabel);
2323                         goto err_opendata_put;
2324                 }
2325         }
2326
2327         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2328                 if (!opendata->f_attr.mdsthreshold) {
2329                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2330                         if (!opendata->f_attr.mdsthreshold)
2331                                 goto err_free_label;
2332                 }
2333                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2334         }
2335         if (dentry->d_inode != NULL)
2336                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2337
2338         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2339         if (status != 0)
2340                 goto err_free_label;
2341         state = ctx->state;
2342
2343         if ((opendata->o_arg.open_flags & O_EXCL) &&
2344             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2345                 nfs4_exclusive_attrset(opendata, sattr);
2346
2347                 nfs_fattr_init(opendata->o_res.f_attr);
2348                 status = nfs4_do_setattr(state->inode, cred,
2349                                 opendata->o_res.f_attr, sattr,
2350                                 state, label, olabel);
2351                 if (status == 0) {
2352                         nfs_setattr_update_inode(state->inode, sattr);
2353                         nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2354                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2355                 }
2356         }
2357         if (opendata->file_created)
2358                 *opened |= FILE_CREATED;
2359
2360         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2361                 *ctx_th = opendata->f_attr.mdsthreshold;
2362                 opendata->f_attr.mdsthreshold = NULL;
2363         }
2364
2365         nfs4_label_free(olabel);
2366
2367         nfs4_opendata_put(opendata);
2368         nfs4_put_state_owner(sp);
2369         return 0;
2370 err_free_label:
2371         nfs4_label_free(olabel);
2372 err_opendata_put:
2373         nfs4_opendata_put(opendata);
2374 err_put_state_owner:
2375         nfs4_put_state_owner(sp);
2376 out_err:
2377         return status;
2378 }
2379
2380
2381 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2382                                         struct nfs_open_context *ctx,
2383                                         int flags,
2384                                         struct iattr *sattr,
2385                                         struct nfs4_label *label,
2386                                         int *opened)
2387 {
2388         struct nfs_server *server = NFS_SERVER(dir);
2389         struct nfs4_exception exception = { };
2390         struct nfs4_state *res;
2391         int status;
2392
2393         do {
2394                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2395                 res = ctx->state;
2396                 trace_nfs4_open_file(ctx, flags, status);
2397                 if (status == 0)
2398                         break;
2399                 /* NOTE: BAD_SEQID means the server and client disagree about the
2400                  * book-keeping w.r.t. state-changing operations
2401                  * (OPEN/CLOSE/LOCK/LOCKU...)
2402                  * It is actually a sign of a bug on the client or on the server.
2403                  *
2404                  * If we receive a BAD_SEQID error in the particular case of
2405                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2406                  * have unhashed the old state_owner for us, and that we can
2407                  * therefore safely retry using a new one. We should still warn
2408                  * the user though...
2409                  */
2410                 if (status == -NFS4ERR_BAD_SEQID) {
2411                         pr_warn_ratelimited("NFS: v4 server %s "
2412                                         " returned a bad sequence-id error!\n",
2413                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2414                         exception.retry = 1;
2415                         continue;
2416                 }
2417                 /*
2418                  * BAD_STATEID on OPEN means that the server cancelled our
2419                  * state before it received the OPEN_CONFIRM.
2420                  * Recover by retrying the request as per the discussion
2421                  * on Page 181 of RFC3530.
2422                  */
2423                 if (status == -NFS4ERR_BAD_STATEID) {
2424                         exception.retry = 1;
2425                         continue;
2426                 }
2427                 if (status == -EAGAIN) {
2428                         /* We must have found a delegation */
2429                         exception.retry = 1;
2430                         continue;
2431                 }
2432                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2433                         continue;
2434                 res = ERR_PTR(nfs4_handle_exception(server,
2435                                         status, &exception));
2436         } while (exception.retry);
2437         return res;
2438 }
2439
2440 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2441                             struct nfs_fattr *fattr, struct iattr *sattr,
2442                             struct nfs4_state *state, struct nfs4_label *ilabel,
2443                             struct nfs4_label *olabel)
2444 {
2445         struct nfs_server *server = NFS_SERVER(inode);
2446         struct nfs_setattrargs  arg = {
2447                 .fh             = NFS_FH(inode),
2448                 .iap            = sattr,
2449                 .server         = server,
2450                 .bitmask = server->attr_bitmask,
2451                 .label          = ilabel,
2452         };
2453         struct nfs_setattrres  res = {
2454                 .fattr          = fattr,
2455                 .label          = olabel,
2456                 .server         = server,
2457         };
2458         struct rpc_message msg = {
2459                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2460                 .rpc_argp       = &arg,
2461                 .rpc_resp       = &res,
2462                 .rpc_cred       = cred,
2463         };
2464         unsigned long timestamp = jiffies;
2465         fmode_t fmode;
2466         bool truncate;
2467         int status;
2468
2469         arg.bitmask = nfs4_bitmask(server, ilabel);
2470         if (ilabel)
2471                 arg.bitmask = nfs4_bitmask(server, olabel);
2472
2473         nfs_fattr_init(fattr);
2474
2475         /* Servers should only apply open mode checks for file size changes */
2476         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2477         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2478
2479         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2480                 /* Use that stateid */
2481         } else if (truncate && state != NULL) {
2482                 struct nfs_lockowner lockowner = {
2483                         .l_owner = current->files,
2484                         .l_pid = current->tgid,
2485                 };
2486                 if (!nfs4_valid_open_stateid(state))
2487                         return -EBADF;
2488                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2489                                 &lockowner) == -EIO)
2490                         return -EBADF;
2491         } else
2492                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2493
2494         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2495         if (status == 0 && state != NULL)
2496                 renew_lease(server, timestamp);
2497         return status;
2498 }
2499
2500 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2501                            struct nfs_fattr *fattr, struct iattr *sattr,
2502                            struct nfs4_state *state, struct nfs4_label *ilabel,
2503                            struct nfs4_label *olabel)
2504 {
2505         struct nfs_server *server = NFS_SERVER(inode);
2506         struct nfs4_exception exception = {
2507                 .state = state,
2508                 .inode = inode,
2509         };
2510         int err;
2511         do {
2512                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2513                 trace_nfs4_setattr(inode, err);
2514                 switch (err) {
2515                 case -NFS4ERR_OPENMODE:
2516                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2517                                 pr_warn_once("NFSv4: server %s is incorrectly "
2518                                                 "applying open mode checks to "
2519                                                 "a SETATTR that is not "
2520                                                 "changing file size.\n",
2521                                                 server->nfs_client->cl_hostname);
2522                         }
2523                         if (state && !(state->state & FMODE_WRITE)) {
2524                                 err = -EBADF;
2525                                 if (sattr->ia_valid & ATTR_OPEN)
2526                                         err = -EACCES;
2527                                 goto out;
2528                         }
2529                 }
2530                 err = nfs4_handle_exception(server, err, &exception);
2531         } while (exception.retry);
2532 out:
2533         return err;
2534 }
2535
2536 struct nfs4_closedata {
2537         struct inode *inode;
2538         struct nfs4_state *state;
2539         struct nfs_closeargs arg;
2540         struct nfs_closeres res;
2541         struct nfs_fattr fattr;
2542         unsigned long timestamp;
2543         bool roc;
2544         u32 roc_barrier;
2545 };
2546
2547 static void nfs4_free_closedata(void *data)
2548 {
2549         struct nfs4_closedata *calldata = data;
2550         struct nfs4_state_owner *sp = calldata->state->owner;
2551         struct super_block *sb = calldata->state->inode->i_sb;
2552
2553         if (calldata->roc)
2554                 pnfs_roc_release(calldata->state->inode);
2555         nfs4_put_open_state(calldata->state);
2556         nfs_free_seqid(calldata->arg.seqid);
2557         nfs4_put_state_owner(sp);
2558         nfs_sb_deactive(sb);
2559         kfree(calldata);
2560 }
2561
2562 static void nfs4_close_done(struct rpc_task *task, void *data)
2563 {
2564         struct nfs4_closedata *calldata = data;
2565         struct nfs4_state *state = calldata->state;
2566         struct nfs_server *server = NFS_SERVER(calldata->inode);
2567         nfs4_stateid *res_stateid = NULL;
2568
2569         dprintk("%s: begin!\n", __func__);
2570         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2571                 return;
2572         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2573         /* hmm. we are done with the inode, and in the process of freeing
2574          * the state_owner. we keep this around to process errors
2575          */
2576         switch (task->tk_status) {
2577                 case 0:
2578                         res_stateid = &calldata->res.stateid;
2579                         if (calldata->arg.fmode == 0 && calldata->roc)
2580                                 pnfs_roc_set_barrier(state->inode,
2581                                                      calldata->roc_barrier);
2582                         renew_lease(server, calldata->timestamp);
2583                         break;
2584                 case -NFS4ERR_ADMIN_REVOKED:
2585                 case -NFS4ERR_STALE_STATEID:
2586                 case -NFS4ERR_OLD_STATEID:
2587                 case -NFS4ERR_BAD_STATEID:
2588                 case -NFS4ERR_EXPIRED:
2589                         if (calldata->arg.fmode == 0)
2590                                 break;
2591                 default:
2592                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN) {
2593                                 rpc_restart_call_prepare(task);
2594                                 goto out_release;
2595                         }
2596         }
2597         nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2598 out_release:
2599         nfs_release_seqid(calldata->arg.seqid);
2600         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2601         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2602 }
2603
2604 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2605 {
2606         struct nfs4_closedata *calldata = data;
2607         struct nfs4_state *state = calldata->state;
2608         struct inode *inode = calldata->inode;
2609         bool is_rdonly, is_wronly, is_rdwr;
2610         int call_close = 0;
2611
2612         dprintk("%s: begin!\n", __func__);
2613         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2614                 goto out_wait;
2615
2616         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2617         spin_lock(&state->owner->so_lock);
2618         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2619         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2620         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2621         /* Calculate the change in open mode */
2622         calldata->arg.fmode = 0;
2623         if (state->n_rdwr == 0) {
2624                 if (state->n_rdonly == 0)
2625                         call_close |= is_rdonly;
2626                 else if (is_rdonly)
2627                         calldata->arg.fmode |= FMODE_READ;
2628                 if (state->n_wronly == 0)
2629                         call_close |= is_wronly;
2630                 else if (is_wronly)
2631                         calldata->arg.fmode |= FMODE_WRITE;
2632         } else if (is_rdwr)
2633                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2634
2635         if (calldata->arg.fmode == 0)
2636                 call_close |= is_rdwr;
2637
2638         if (!nfs4_valid_open_stateid(state))
2639                 call_close = 0;
2640         spin_unlock(&state->owner->so_lock);
2641
2642         if (!call_close) {
2643                 /* Note: exit _without_ calling nfs4_close_done */
2644                 goto out_no_action;
2645         }
2646
2647         if (calldata->arg.fmode == 0) {
2648                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2649                 if (calldata->roc &&
2650                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2651                         nfs_release_seqid(calldata->arg.seqid);
2652                         goto out_wait;
2653                     }
2654         }
2655
2656         nfs_fattr_init(calldata->res.fattr);
2657         calldata->timestamp = jiffies;
2658         if (nfs4_setup_sequence(NFS_SERVER(inode),
2659                                 &calldata->arg.seq_args,
2660                                 &calldata->res.seq_res,
2661                                 task) != 0)
2662                 nfs_release_seqid(calldata->arg.seqid);
2663         dprintk("%s: done!\n", __func__);
2664         return;
2665 out_no_action:
2666         task->tk_action = NULL;
2667 out_wait:
2668         nfs4_sequence_done(task, &calldata->res.seq_res);
2669 }
2670
2671 static const struct rpc_call_ops nfs4_close_ops = {
2672         .rpc_call_prepare = nfs4_close_prepare,
2673         .rpc_call_done = nfs4_close_done,
2674         .rpc_release = nfs4_free_closedata,
2675 };
2676
2677 static bool nfs4_state_has_opener(struct nfs4_state *state)
2678 {
2679         /* first check existing openers */
2680         if (test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0 &&
2681             state->n_rdonly != 0)
2682                 return true;
2683
2684         if (test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0 &&
2685             state->n_wronly != 0)
2686                 return true;
2687
2688         if (test_bit(NFS_O_RDWR_STATE, &state->flags) != 0 &&
2689             state->n_rdwr != 0)
2690                 return true;
2691
2692         return false;
2693 }
2694
2695 static bool nfs4_roc(struct inode *inode)
2696 {
2697         struct nfs_inode *nfsi = NFS_I(inode);
2698         struct nfs_open_context *ctx;
2699         struct nfs4_state *state;
2700
2701         spin_lock(&inode->i_lock);
2702         list_for_each_entry(ctx, &nfsi->open_files, list) {
2703                 state = ctx->state;
2704                 if (state == NULL)
2705                         continue;
2706                 if (nfs4_state_has_opener(state)) {
2707                         spin_unlock(&inode->i_lock);
2708                         return false;
2709                 }
2710         }
2711         spin_unlock(&inode->i_lock);
2712
2713         if (nfs4_check_delegation(inode, FMODE_READ))
2714                 return false;
2715
2716         return pnfs_roc(inode);
2717 }
2718
2719 /* 
2720  * It is possible for data to be read/written from a mem-mapped file 
2721  * after the sys_close call (which hits the vfs layer as a flush).
2722  * This means that we can't safely call nfsv4 close on a file until 
2723  * the inode is cleared. This in turn means that we are not good
2724  * NFSv4 citizens - we do not indicate to the server to update the file's 
2725  * share state even when we are done with one of the three share 
2726  * stateid's in the inode.
2727  *
2728  * NOTE: Caller must be holding the sp->so_owner semaphore!
2729  */
2730 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2731 {
2732         struct nfs_server *server = NFS_SERVER(state->inode);
2733         struct nfs4_closedata *calldata;
2734         struct nfs4_state_owner *sp = state->owner;
2735         struct rpc_task *task;
2736         struct rpc_message msg = {
2737                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2738                 .rpc_cred = state->owner->so_cred,
2739         };
2740         struct rpc_task_setup task_setup_data = {
2741                 .rpc_client = server->client,
2742                 .rpc_message = &msg,
2743                 .callback_ops = &nfs4_close_ops,
2744                 .workqueue = nfsiod_workqueue,
2745                 .flags = RPC_TASK_ASYNC,
2746         };
2747         int status = -ENOMEM;
2748
2749         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2750                 &task_setup_data.rpc_client, &msg);
2751
2752         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2753         if (calldata == NULL)
2754                 goto out;
2755         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2756         calldata->inode = state->inode;
2757         calldata->state = state;
2758         calldata->arg.fh = NFS_FH(state->inode);
2759         calldata->arg.stateid = &state->open_stateid;
2760         /* Serialization for the sequence id */
2761         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2762         if (calldata->arg.seqid == NULL)
2763                 goto out_free_calldata;
2764         calldata->arg.fmode = 0;
2765         calldata->arg.bitmask = server->cache_consistency_bitmask;
2766         calldata->res.fattr = &calldata->fattr;
2767         calldata->res.seqid = calldata->arg.seqid;
2768         calldata->res.server = server;
2769         calldata->roc = nfs4_roc(state->inode);
2770         nfs_sb_active(calldata->inode->i_sb);
2771
2772         msg.rpc_argp = &calldata->arg;
2773         msg.rpc_resp = &calldata->res;
2774         task_setup_data.callback_data = calldata;
2775         task = rpc_run_task(&task_setup_data);
2776         if (IS_ERR(task))
2777                 return PTR_ERR(task);
2778         status = 0;
2779         if (wait)
2780                 status = rpc_wait_for_completion_task(task);
2781         rpc_put_task(task);
2782         return status;
2783 out_free_calldata:
2784         kfree(calldata);
2785 out:
2786         nfs4_put_open_state(state);
2787         nfs4_put_state_owner(sp);
2788         return status;
2789 }
2790
2791 static struct inode *
2792 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2793                 int open_flags, struct iattr *attr, int *opened)
2794 {
2795         struct nfs4_state *state;
2796         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2797
2798         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2799
2800         /* Protect against concurrent sillydeletes */
2801         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2802
2803         nfs4_label_release_security(label);
2804
2805         if (IS_ERR(state))
2806                 return ERR_CAST(state);
2807         return state->inode;
2808 }
2809
2810 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2811 {
2812         if (ctx->state == NULL)
2813                 return;
2814         if (is_sync)
2815                 nfs4_close_sync(ctx->state, ctx->mode);
2816         else
2817                 nfs4_close_state(ctx->state, ctx->mode);
2818 }
2819
2820 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2821 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2822 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2823
2824 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2825 {
2826         struct nfs4_server_caps_arg args = {
2827                 .fhandle = fhandle,
2828         };
2829         struct nfs4_server_caps_res res = {};
2830         struct rpc_message msg = {
2831                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2832                 .rpc_argp = &args,
2833                 .rpc_resp = &res,
2834         };
2835         int status;
2836
2837         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2838         if (status == 0) {
2839                 /* Sanity check the server answers */
2840                 switch (server->nfs_client->cl_minorversion) {
2841                 case 0:
2842                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2843                         res.attr_bitmask[2] = 0;
2844                         break;
2845                 case 1:
2846                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2847                         break;
2848                 case 2:
2849                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2850                 }
2851                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2852                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2853                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2854                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2855                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2856                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2857                                 NFS_CAP_SECURITY_LABEL);
2858                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2859                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2860                         server->caps |= NFS_CAP_ACLS;
2861                 if (res.has_links != 0)
2862                         server->caps |= NFS_CAP_HARDLINKS;
2863                 if (res.has_symlinks != 0)
2864                         server->caps |= NFS_CAP_SYMLINKS;
2865                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2866                         server->caps |= NFS_CAP_FILEID;
2867                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2868                         server->caps |= NFS_CAP_MODE;
2869                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2870                         server->caps |= NFS_CAP_NLINK;
2871                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2872                         server->caps |= NFS_CAP_OWNER;
2873                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2874                         server->caps |= NFS_CAP_OWNER_GROUP;
2875                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2876                         server->caps |= NFS_CAP_ATIME;
2877                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2878                         server->caps |= NFS_CAP_CTIME;
2879                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2880                         server->caps |= NFS_CAP_MTIME;
2881 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2882                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2883                         server->caps |= NFS_CAP_SECURITY_LABEL;
2884 #endif
2885                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2886                                 sizeof(server->attr_bitmask));
2887                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2888
2889                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2890                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2891                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2892                 server->cache_consistency_bitmask[2] = 0;
2893                 server->acl_bitmask = res.acl_bitmask;
2894                 server->fh_expire_type = res.fh_expire_type;
2895         }
2896
2897         return status;
2898 }
2899
2900 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2901 {
2902         struct nfs4_exception exception = { };
2903         int err;
2904         do {
2905                 err = nfs4_handle_exception(server,
2906                                 _nfs4_server_capabilities(server, fhandle),
2907                                 &exception);
2908         } while (exception.retry);
2909         return err;
2910 }
2911
2912 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2913                 struct nfs_fsinfo *info)
2914 {
2915         u32 bitmask[3];
2916         struct nfs4_lookup_root_arg args = {
2917                 .bitmask = bitmask,
2918         };
2919         struct nfs4_lookup_res res = {
2920                 .server = server,
2921                 .fattr = info->fattr,
2922                 .fh = fhandle,
2923         };
2924         struct rpc_message msg = {
2925                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2926                 .rpc_argp = &args,
2927                 .rpc_resp = &res,
2928         };
2929
2930         bitmask[0] = nfs4_fattr_bitmap[0];
2931         bitmask[1] = nfs4_fattr_bitmap[1];
2932         /*
2933          * Process the label in the upcoming getfattr
2934          */
2935         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2936
2937         nfs_fattr_init(info->fattr);
2938         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2939 }
2940
2941 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2942                 struct nfs_fsinfo *info)
2943 {
2944         struct nfs4_exception exception = { };
2945         int err;
2946         do {
2947                 err = _nfs4_lookup_root(server, fhandle, info);
2948                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2949                 switch (err) {
2950                 case 0:
2951                 case -NFS4ERR_WRONGSEC:
2952                         goto out;
2953                 default:
2954                         err = nfs4_handle_exception(server, err, &exception);
2955                 }
2956         } while (exception.retry);
2957 out:
2958         return err;
2959 }
2960
2961 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2962                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2963 {
2964         struct rpc_auth_create_args auth_args = {
2965                 .pseudoflavor = flavor,
2966         };
2967         struct rpc_auth *auth;
2968         int ret;
2969
2970         auth = rpcauth_create(&auth_args, server->client);
2971         if (IS_ERR(auth)) {
2972                 ret = -EACCES;
2973                 goto out;
2974         }
2975         ret = nfs4_lookup_root(server, fhandle, info);
2976 out:
2977         return ret;
2978 }
2979
2980 /*
2981  * Retry pseudoroot lookup with various security flavors.  We do this when:
2982  *
2983  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2984  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2985  *
2986  * Returns zero on success, or a negative NFS4ERR value, or a
2987  * negative errno value.
2988  */
2989 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2990                               struct nfs_fsinfo *info)
2991 {
2992         /* Per 3530bis 15.33.5 */
2993         static const rpc_authflavor_t flav_array[] = {
2994                 RPC_AUTH_GSS_KRB5P,
2995                 RPC_AUTH_GSS_KRB5I,
2996                 RPC_AUTH_GSS_KRB5,
2997                 RPC_AUTH_UNIX,                  /* courtesy */
2998                 RPC_AUTH_NULL,
2999         };
3000         int status = -EPERM;
3001         size_t i;
3002
3003         if (server->auth_info.flavor_len > 0) {
3004                 /* try each flavor specified by user */
3005                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3006                         status = nfs4_lookup_root_sec(server, fhandle, info,
3007                                                 server->auth_info.flavors[i]);
3008                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3009                                 continue;
3010                         break;
3011                 }
3012         } else {
3013                 /* no flavors specified by user, try default list */
3014                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3015                         status = nfs4_lookup_root_sec(server, fhandle, info,
3016                                                       flav_array[i]);
3017                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3018                                 continue;
3019                         break;
3020                 }
3021         }
3022
3023         /*
3024          * -EACCESS could mean that the user doesn't have correct permissions
3025          * to access the mount.  It could also mean that we tried to mount
3026          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3027          * existing mount programs don't handle -EACCES very well so it should
3028          * be mapped to -EPERM instead.
3029          */
3030         if (status == -EACCES)
3031                 status = -EPERM;
3032         return status;
3033 }
3034
3035 static int nfs4_do_find_root_sec(struct nfs_server *server,
3036                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3037 {
3038         int mv = server->nfs_client->cl_minorversion;
3039         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3040 }
3041
3042 /**
3043  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3044  * @server: initialized nfs_server handle
3045  * @fhandle: we fill in the pseudo-fs root file handle
3046  * @info: we fill in an FSINFO struct
3047  * @auth_probe: probe the auth flavours
3048  *
3049  * Returns zero on success, or a negative errno.
3050  */
3051 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3052                          struct nfs_fsinfo *info,
3053                          bool auth_probe)
3054 {
3055         int status;
3056
3057         switch (auth_probe) {
3058         case false:
3059                 status = nfs4_lookup_root(server, fhandle, info);
3060                 if (status != -NFS4ERR_WRONGSEC)
3061                         break;
3062         default:
3063                 status = nfs4_do_find_root_sec(server, fhandle, info);
3064         }
3065
3066         if (status == 0)
3067                 status = nfs4_server_capabilities(server, fhandle);
3068         if (status == 0)
3069                 status = nfs4_do_fsinfo(server, fhandle, info);
3070
3071         return nfs4_map_errors(status);
3072 }
3073
3074 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3075                               struct nfs_fsinfo *info)
3076 {
3077         int error;
3078         struct nfs_fattr *fattr = info->fattr;
3079         struct nfs4_label *label = NULL;
3080
3081         error = nfs4_server_capabilities(server, mntfh);
3082         if (error < 0) {
3083                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3084                 return error;
3085         }
3086
3087         label = nfs4_label_alloc(server, GFP_KERNEL);
3088         if (IS_ERR(label))
3089                 return PTR_ERR(label);
3090
3091         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3092         if (error < 0) {
3093                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3094                 goto err_free_label;
3095         }
3096
3097         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3098             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3099                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3100
3101 err_free_label:
3102         nfs4_label_free(label);
3103
3104         return error;
3105 }
3106
3107 /*
3108  * Get locations and (maybe) other attributes of a referral.
3109  * Note that we'll actually follow the referral later when
3110  * we detect fsid mismatch in inode revalidation
3111  */
3112 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3113                              const struct qstr *name, struct nfs_fattr *fattr,
3114                              struct nfs_fh *fhandle)
3115 {
3116         int status = -ENOMEM;
3117         struct page *page = NULL;
3118         struct nfs4_fs_locations *locations = NULL;
3119
3120         page = alloc_page(GFP_KERNEL);
3121         if (page == NULL)
3122                 goto out;
3123         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3124         if (locations == NULL)
3125                 goto out;
3126
3127         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3128         if (status != 0)
3129                 goto out;
3130
3131         /*
3132          * If the fsid didn't change, this is a migration event, not a
3133          * referral.  Cause us to drop into the exception handler, which
3134          * will kick off migration recovery.
3135          */
3136         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3137                 dprintk("%s: server did not return a different fsid for"
3138                         " a referral at %s\n", __func__, name->name);
3139                 status = -NFS4ERR_MOVED;
3140                 goto out;
3141         }
3142         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3143         nfs_fixup_referral_attributes(&locations->fattr);
3144
3145         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3146         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3147         memset(fhandle, 0, sizeof(struct nfs_fh));
3148 out:
3149         if (page)
3150                 __free_page(page);
3151         kfree(locations);
3152         return status;
3153 }
3154
3155 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3156                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3157 {
3158         struct nfs4_getattr_arg args = {
3159                 .fh = fhandle,
3160                 .bitmask = server->attr_bitmask,
3161         };
3162         struct nfs4_getattr_res res = {
3163                 .fattr = fattr,
3164                 .label = label,
3165                 .server = server,
3166         };
3167         struct rpc_message msg = {
3168                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3169                 .rpc_argp = &args,
3170                 .rpc_resp = &res,
3171         };
3172
3173         args.bitmask = nfs4_bitmask(server, label);
3174
3175         nfs_fattr_init(fattr);
3176         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3177 }
3178
3179 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3180                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3181 {
3182         struct nfs4_exception exception = { };
3183         int err;
3184         do {
3185                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3186                 trace_nfs4_getattr(server, fhandle, fattr, err);
3187                 err = nfs4_handle_exception(server, err,
3188                                 &exception);
3189         } while (exception.retry);
3190         return err;
3191 }
3192
3193 /* 
3194  * The file is not closed if it is opened due to the a request to change
3195  * the size of the file. The open call will not be needed once the
3196  * VFS layer lookup-intents are implemented.
3197  *
3198  * Close is called when the inode is destroyed.
3199  * If we haven't opened the file for O_WRONLY, we
3200  * need to in the size_change case to obtain a stateid.
3201  *
3202  * Got race?
3203  * Because OPEN is always done by name in nfsv4, it is
3204  * possible that we opened a different file by the same
3205  * name.  We can recognize this race condition, but we
3206  * can't do anything about it besides returning an error.
3207  *
3208  * This will be fixed with VFS changes (lookup-intent).
3209  */
3210 static int
3211 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3212                   struct iattr *sattr)
3213 {
3214         struct inode *inode = dentry->d_inode;
3215         struct rpc_cred *cred = NULL;
3216         struct nfs4_state *state = NULL;
3217         struct nfs4_label *label = NULL;
3218         int status;
3219
3220         if (pnfs_ld_layoutret_on_setattr(inode))
3221                 pnfs_commit_and_return_layout(inode);
3222
3223         nfs_fattr_init(fattr);
3224         
3225         /* Deal with open(O_TRUNC) */
3226         if (sattr->ia_valid & ATTR_OPEN)
3227                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3228
3229         /* Optimization: if the end result is no change, don't RPC */
3230         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3231                 return 0;
3232
3233         /* Search for an existing open(O_WRITE) file */
3234         if (sattr->ia_valid & ATTR_FILE) {
3235                 struct nfs_open_context *ctx;
3236
3237                 ctx = nfs_file_open_context(sattr->ia_file);
3238                 if (ctx) {
3239                         cred = ctx->cred;
3240                         state = ctx->state;
3241                 }
3242         }
3243
3244         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3245         if (IS_ERR(label))
3246                 return PTR_ERR(label);
3247
3248         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3249         if (status == 0) {
3250                 nfs_setattr_update_inode(inode, sattr);
3251                 nfs_setsecurity(inode, fattr, label);
3252         }
3253         nfs4_label_free(label);
3254         return status;
3255 }
3256
3257 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3258                 const struct qstr *name, struct nfs_fh *fhandle,
3259                 struct nfs_fattr *fattr, struct nfs4_label *label)
3260 {
3261         struct nfs_server *server = NFS_SERVER(dir);
3262         int                    status;
3263         struct nfs4_lookup_arg args = {
3264                 .bitmask = server->attr_bitmask,
3265                 .dir_fh = NFS_FH(dir),
3266                 .name = name,
3267         };
3268         struct nfs4_lookup_res res = {
3269                 .server = server,
3270                 .fattr = fattr,
3271                 .label = label,
3272                 .fh = fhandle,
3273         };
3274         struct rpc_message msg = {
3275                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3276                 .rpc_argp = &args,
3277                 .rpc_resp = &res,
3278         };
3279
3280         args.bitmask = nfs4_bitmask(server, label);
3281
3282         nfs_fattr_init(fattr);
3283
3284         dprintk("NFS call  lookup %s\n", name->name);
3285         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3286         dprintk("NFS reply lookup: %d\n", status);
3287         return status;
3288 }
3289
3290 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3291 {
3292         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3293                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3294         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3295         fattr->nlink = 2;
3296 }
3297
3298 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3299                                    struct qstr *name, struct nfs_fh *fhandle,
3300                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3301 {
3302         struct nfs4_exception exception = { };
3303         struct rpc_clnt *client = *clnt;
3304         int err;
3305         do {
3306                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3307                 trace_nfs4_lookup(dir, name, err);
3308                 switch (err) {
3309                 case -NFS4ERR_BADNAME:
3310                         err = -ENOENT;
3311                         goto out;
3312                 case -NFS4ERR_MOVED:
3313                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3314                         goto out;
3315                 case -NFS4ERR_WRONGSEC:
3316                         err = -EPERM;
3317                         if (client != *clnt)
3318                                 goto out;
3319                         client = nfs4_negotiate_security(client, dir, name);
3320                         if (IS_ERR(client))
3321                                 return PTR_ERR(client);
3322
3323                         exception.retry = 1;
3324                         break;
3325                 default:
3326                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3327                 }
3328         } while (exception.retry);
3329
3330 out:
3331         if (err == 0)
3332                 *clnt = client;
3333         else if (client != *clnt)
3334                 rpc_shutdown_client(client);
3335
3336         return err;
3337 }
3338
3339 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3340                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3341                             struct nfs4_label *label)
3342 {
3343         int status;
3344         struct rpc_clnt *client = NFS_CLIENT(dir);
3345
3346         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3347         if (client != NFS_CLIENT(dir)) {
3348                 rpc_shutdown_client(client);
3349                 nfs_fixup_secinfo_attributes(fattr);
3350         }
3351         return status;
3352 }
3353
3354 struct rpc_clnt *
3355 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3356                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3357 {
3358         struct rpc_clnt *client = NFS_CLIENT(dir);
3359         int status;
3360
3361         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3362         if (status < 0)
3363                 return ERR_PTR(status);
3364         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3365 }
3366
3367 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3368 {
3369         struct nfs_server *server = NFS_SERVER(inode);
3370         struct nfs4_accessargs args = {
3371                 .fh = NFS_FH(inode),
3372                 .bitmask = server->cache_consistency_bitmask,
3373         };
3374         struct nfs4_accessres res = {
3375                 .server = server,
3376         };
3377         struct rpc_message msg = {
3378                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3379                 .rpc_argp = &args,
3380                 .rpc_resp = &res,
3381                 .rpc_cred = entry->cred,
3382         };
3383         int mode = entry->mask;
3384         int status = 0;
3385
3386         /*
3387          * Determine which access bits we want to ask for...
3388          */
3389         if (mode & MAY_READ)
3390                 args.access |= NFS4_ACCESS_READ;
3391         if (S_ISDIR(inode->i_mode)) {
3392                 if (mode & MAY_WRITE)
3393                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3394                 if (mode & MAY_EXEC)
3395                         args.access |= NFS4_ACCESS_LOOKUP;
3396         } else {
3397                 if (mode & MAY_WRITE)
3398                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3399                 if (mode & MAY_EXEC)
3400                         args.access |= NFS4_ACCESS_EXECUTE;
3401         }
3402
3403         res.fattr = nfs_alloc_fattr();
3404         if (res.fattr == NULL)
3405                 return -ENOMEM;
3406
3407         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3408         if (!status) {
3409                 nfs_access_set_mask(entry, res.access);
3410                 nfs_refresh_inode(inode, res.fattr);
3411         }
3412         nfs_free_fattr(res.fattr);
3413         return status;
3414 }
3415
3416 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3417 {
3418         struct nfs4_exception exception = { };
3419         int err;
3420         do {
3421                 err = _nfs4_proc_access(inode, entry);
3422                 trace_nfs4_access(inode, err);
3423                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3424                                 &exception);
3425         } while (exception.retry);
3426         return err;
3427 }
3428
3429 /*
3430  * TODO: For the time being, we don't try to get any attributes
3431  * along with any of the zero-copy operations READ, READDIR,
3432  * READLINK, WRITE.
3433  *
3434  * In the case of the first three, we want to put the GETATTR
3435  * after the read-type operation -- this is because it is hard
3436  * to predict the length of a GETATTR response in v4, and thus
3437  * align the READ data correctly.  This means that the GETATTR
3438  * may end up partially falling into the page cache, and we should
3439  * shift it into the 'tail' of the xdr_buf before processing.
3440  * To do this efficiently, we need to know the total length
3441  * of data received, which doesn't seem to be available outside
3442  * of the RPC layer.
3443  *
3444  * In the case of WRITE, we also want to put the GETATTR after
3445  * the operation -- in this case because we want to make sure
3446  * we get the post-operation mtime and size.
3447  *
3448  * Both of these changes to the XDR layer would in fact be quite
3449  * minor, but I decided to leave them for a subsequent patch.
3450  */
3451 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3452                 unsigned int pgbase, unsigned int pglen)
3453 {
3454         struct nfs4_readlink args = {
3455                 .fh       = NFS_FH(inode),
3456                 .pgbase   = pgbase,
3457                 .pglen    = pglen,
3458                 .pages    = &page,
3459         };
3460         struct nfs4_readlink_res res;
3461         struct rpc_message msg = {
3462                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3463                 .rpc_argp = &args,
3464                 .rpc_resp = &res,
3465         };
3466
3467         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3468 }
3469
3470 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3471                 unsigned int pgbase, unsigned int pglen)
3472 {
3473         struct nfs4_exception exception = { };
3474         int err;
3475         do {
3476                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3477                 trace_nfs4_readlink(inode, err);
3478                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3479                                 &exception);
3480         } while (exception.retry);
3481         return err;
3482 }
3483
3484 /*
3485  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3486  */
3487 static int
3488 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3489                  int flags)
3490 {
3491         struct nfs4_label l, *ilabel = NULL;
3492         struct nfs_open_context *ctx;
3493         struct nfs4_state *state;
3494         int opened = 0;
3495         int status = 0;
3496
3497         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3498         if (IS_ERR(ctx))
3499                 return PTR_ERR(ctx);
3500
3501         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3502
3503         sattr->ia_mode &= ~current_umask();
3504         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3505         if (IS_ERR(state)) {
3506                 status = PTR_ERR(state);
3507                 goto out;
3508         }
3509 out:
3510         nfs4_label_release_security(ilabel);
3511         put_nfs_open_context(ctx);
3512         return status;
3513 }
3514
3515 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3516 {
3517         struct nfs_server *server = NFS_SERVER(dir);
3518         struct nfs_removeargs args = {
3519                 .fh = NFS_FH(dir),
3520                 .name = *name,
3521         };
3522         struct nfs_removeres res = {
3523                 .server = server,
3524         };
3525         struct rpc_message msg = {
3526                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3527                 .rpc_argp = &args,
3528                 .rpc_resp = &res,
3529         };
3530         int status;
3531
3532         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3533         if (status == 0)
3534                 update_changeattr(dir, &res.cinfo);
3535         return status;
3536 }
3537
3538 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3539 {
3540         struct nfs4_exception exception = { };
3541         int err;
3542         do {
3543                 err = _nfs4_proc_remove(dir, name);
3544                 trace_nfs4_remove(dir, name, err);
3545                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3546                                 &exception);
3547         } while (exception.retry);
3548         return err;
3549 }
3550
3551 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3552 {
3553         struct nfs_server *server = NFS_SERVER(dir);
3554         struct nfs_removeargs *args = msg->rpc_argp;
3555         struct nfs_removeres *res = msg->rpc_resp;
3556
3557         res->server = server;
3558         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3559         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3560
3561         nfs_fattr_init(res->dir_attr);
3562 }
3563
3564 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3565 {
3566         nfs4_setup_sequence(NFS_SERVER(data->dir),
3567                         &data->args.seq_args,
3568                         &data->res.seq_res,
3569                         task);
3570 }
3571
3572 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3573 {
3574         struct nfs_unlinkdata *data = task->tk_calldata;
3575         struct nfs_removeres *res = &data->res;
3576
3577         if (!nfs4_sequence_done(task, &res->seq_res))
3578                 return 0;
3579         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3580                 return 0;
3581         update_changeattr(dir, &res->cinfo);
3582         return 1;
3583 }
3584
3585 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3586 {
3587         struct nfs_server *server = NFS_SERVER(dir);
3588         struct nfs_renameargs *arg = msg->rpc_argp;
3589         struct nfs_renameres *res = msg->rpc_resp;
3590
3591         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3592         res->server = server;
3593         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3594 }
3595
3596 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3597 {
3598         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3599                         &data->args.seq_args,
3600                         &data->res.seq_res,
3601                         task);
3602 }
3603
3604 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3605                                  struct inode *new_dir)
3606 {
3607         struct nfs_renamedata *data = task->tk_calldata;
3608         struct nfs_renameres *res = &data->res;
3609
3610         if (!nfs4_sequence_done(task, &res->seq_res))
3611                 return 0;
3612         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3613                 return 0;
3614
3615         update_changeattr(old_dir, &res->old_cinfo);
3616         update_changeattr(new_dir, &res->new_cinfo);
3617         return 1;
3618 }
3619
3620 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3621 {
3622         struct nfs_server *server = NFS_SERVER(inode);
3623         struct nfs4_link_arg arg = {
3624                 .fh     = NFS_FH(inode),
3625                 .dir_fh = NFS_FH(dir),
3626                 .name   = name,
3627                 .bitmask = server->attr_bitmask,
3628         };
3629         struct nfs4_link_res res = {
3630                 .server = server,
3631                 .label = NULL,
3632         };
3633         struct rpc_message msg = {
3634                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3635                 .rpc_argp = &arg,
3636                 .rpc_resp = &res,
3637         };
3638         int status = -ENOMEM;
3639
3640         res.fattr = nfs_alloc_fattr();
3641         if (res.fattr == NULL)
3642                 goto out;
3643
3644         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3645         if (IS_ERR(res.label)) {
3646                 status = PTR_ERR(res.label);
3647                 goto out;
3648         }
3649         arg.bitmask = nfs4_bitmask(server, res.label);
3650
3651         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3652         if (!status) {
3653                 update_changeattr(dir, &res.cinfo);
3654                 status = nfs_post_op_update_inode(inode, res.fattr);
3655                 if (!status)
3656                         nfs_setsecurity(inode, res.fattr, res.label);
3657         }
3658
3659
3660         nfs4_label_free(res.label);
3661
3662 out:
3663         nfs_free_fattr(res.fattr);
3664         return status;
3665 }
3666
3667 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3668 {
3669         struct nfs4_exception exception = { };
3670         int err;
3671         do {
3672                 err = nfs4_handle_exception(NFS_SERVER(inode),
3673                                 _nfs4_proc_link(inode, dir, name),
3674                                 &exception);
3675         } while (exception.retry);
3676         return err;
3677 }
3678
3679 struct nfs4_createdata {
3680         struct rpc_message msg;
3681         struct nfs4_create_arg arg;
3682         struct nfs4_create_res res;
3683         struct nfs_fh fh;
3684         struct nfs_fattr fattr;
3685         struct nfs4_label *label;
3686 };
3687
3688 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3689                 struct qstr *name, struct iattr *sattr, u32 ftype)
3690 {
3691         struct nfs4_createdata *data;
3692
3693         data = kzalloc(sizeof(*data), GFP_KERNEL);
3694         if (data != NULL) {
3695                 struct nfs_server *server = NFS_SERVER(dir);
3696
3697                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3698                 if (IS_ERR(data->label))
3699                         goto out_free;
3700
3701                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3702                 data->msg.rpc_argp = &data->arg;
3703                 data->msg.rpc_resp = &data->res;
3704                 data->arg.dir_fh = NFS_FH(dir);
3705                 data->arg.server = server;
3706                 data->arg.name = name;
3707                 data->arg.attrs = sattr;
3708                 data->arg.ftype = ftype;
3709                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3710                 data->res.server = server;
3711                 data->res.fh = &data->fh;
3712                 data->res.fattr = &data->fattr;
3713                 data->res.label = data->label;
3714                 nfs_fattr_init(data->res.fattr);
3715         }
3716         return data;
3717 out_free:
3718         kfree(data);
3719         return NULL;
3720 }
3721
3722 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3723 {
3724         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3725                                     &data->arg.seq_args, &data->res.seq_res, 1);
3726         if (status == 0) {
3727                 update_changeattr(dir, &data->res.dir_cinfo);
3728                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3729         }
3730         return status;
3731 }
3732
3733 static void nfs4_free_createdata(struct nfs4_createdata *data)
3734 {
3735         nfs4_label_free(data->label);
3736         kfree(data);
3737 }
3738
3739 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3740                 struct page *page, unsigned int len, struct iattr *sattr,
3741                 struct nfs4_label *label)
3742 {
3743         struct nfs4_createdata *data;
3744         int status = -ENAMETOOLONG;
3745
3746         if (len > NFS4_MAXPATHLEN)
3747                 goto out;
3748
3749         status = -ENOMEM;
3750         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3751         if (data == NULL)
3752                 goto out;
3753
3754         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3755         data->arg.u.symlink.pages = &page;
3756         data->arg.u.symlink.len = len;
3757         data->arg.label = label;
3758         
3759         status = nfs4_do_create(dir, dentry, data);
3760
3761         nfs4_free_createdata(data);
3762 out:
3763         return status;
3764 }
3765
3766 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3767                 struct page *page, unsigned int len, struct iattr *sattr)
3768 {
3769         struct nfs4_exception exception = { };
3770         struct nfs4_label l, *label = NULL;
3771         int err;
3772
3773         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3774
3775         do {
3776                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3777                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3778                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3779                                 &exception);
3780         } while (exception.retry);
3781
3782         nfs4_label_release_security(label);
3783         return err;
3784 }
3785
3786 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3787                 struct iattr *sattr, struct nfs4_label *label)
3788 {
3789         struct nfs4_createdata *data;
3790         int status = -ENOMEM;
3791
3792         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3793         if (data == NULL)
3794                 goto out;
3795
3796         data->arg.label = label;
3797         status = nfs4_do_create(dir, dentry, data);
3798
3799         nfs4_free_createdata(data);
3800 out:
3801         return status;
3802 }
3803
3804 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3805                 struct iattr *sattr)
3806 {
3807         struct nfs4_exception exception = { };
3808         struct nfs4_label l, *label = NULL;
3809         int err;
3810
3811         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3812
3813         sattr->ia_mode &= ~current_umask();
3814         do {
3815                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3816                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3817                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3818                                 &exception);
3819         } while (exception.retry);
3820         nfs4_label_release_security(label);
3821
3822         return err;
3823 }
3824
3825 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3826                 u64 cookie, struct page **pages, unsigned int count, int plus)
3827 {
3828         struct inode            *dir = dentry->d_inode;
3829         struct nfs4_readdir_arg args = {
3830                 .fh = NFS_FH(dir),
3831                 .pages = pages,
3832                 .pgbase = 0,
3833                 .count = count,
3834                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3835                 .plus = plus,
3836         };
3837         struct nfs4_readdir_res res;
3838         struct rpc_message msg = {
3839                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3840                 .rpc_argp = &args,
3841                 .rpc_resp = &res,
3842                 .rpc_cred = cred,
3843         };
3844         int                     status;
3845
3846         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3847                         dentry,
3848                         (unsigned long long)cookie);
3849         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3850         res.pgbase = args.pgbase;
3851         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3852         if (status >= 0) {
3853                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3854                 status += args.pgbase;
3855         }
3856
3857         nfs_invalidate_atime(dir);
3858
3859         dprintk("%s: returns %d\n", __func__, status);
3860         return status;
3861 }
3862
3863 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3864                 u64 cookie, struct page **pages, unsigned int count, int plus)
3865 {
3866         struct nfs4_exception exception = { };
3867         int err;
3868         do {
3869                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3870                                 pages, count, plus);
3871                 trace_nfs4_readdir(dentry->d_inode, err);
3872                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3873                                 &exception);
3874         } while (exception.retry);
3875         return err;
3876 }
3877
3878 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3879                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3880 {
3881         struct nfs4_createdata *data;
3882         int mode = sattr->ia_mode;
3883         int status = -ENOMEM;
3884
3885         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3886         if (data == NULL)
3887                 goto out;
3888
3889         if (S_ISFIFO(mode))
3890                 data->arg.ftype = NF4FIFO;
3891         else if (S_ISBLK(mode)) {
3892                 data->arg.ftype = NF4BLK;
3893                 data->arg.u.device.specdata1 = MAJOR(rdev);
3894                 data->arg.u.device.specdata2 = MINOR(rdev);
3895         }
3896         else if (S_ISCHR(mode)) {
3897                 data->arg.ftype = NF4CHR;
3898                 data->arg.u.device.specdata1 = MAJOR(rdev);
3899                 data->arg.u.device.specdata2 = MINOR(rdev);
3900         } else if (!S_ISSOCK(mode)) {
3901                 status = -EINVAL;
3902                 goto out_free;
3903         }
3904
3905         data->arg.label = label;
3906         status = nfs4_do_create(dir, dentry, data);
3907 out_free:
3908         nfs4_free_createdata(data);
3909 out:
3910         return status;
3911 }
3912
3913 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3914                 struct iattr *sattr, dev_t rdev)
3915 {
3916         struct nfs4_exception exception = { };
3917         struct nfs4_label l, *label = NULL;
3918         int err;
3919
3920         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3921
3922         sattr->ia_mode &= ~current_umask();
3923         do {
3924                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3925                 trace_nfs4_mknod(dir, &dentry->d_name, err);
3926                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3927                                 &exception);
3928         } while (exception.retry);
3929
3930         nfs4_label_release_security(label);
3931
3932         return err;
3933 }
3934
3935 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3936                  struct nfs_fsstat *fsstat)
3937 {
3938         struct nfs4_statfs_arg args = {
3939                 .fh = fhandle,
3940                 .bitmask = server->attr_bitmask,
3941         };
3942         struct nfs4_statfs_res res = {
3943                 .fsstat = fsstat,
3944         };
3945         struct rpc_message msg = {
3946                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3947                 .rpc_argp = &args,
3948                 .rpc_resp = &res,
3949         };
3950
3951         nfs_fattr_init(fsstat->fattr);
3952         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3953 }
3954
3955 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3956 {
3957         struct nfs4_exception exception = { };
3958         int err;
3959         do {
3960                 err = nfs4_handle_exception(server,
3961                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3962                                 &exception);
3963         } while (exception.retry);
3964         return err;
3965 }
3966
3967 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3968                 struct nfs_fsinfo *fsinfo)
3969 {
3970         struct nfs4_fsinfo_arg args = {
3971                 .fh = fhandle,
3972                 .bitmask = server->attr_bitmask,
3973         };
3974         struct nfs4_fsinfo_res res = {
3975                 .fsinfo = fsinfo,
3976         };
3977         struct rpc_message msg = {
3978                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3979                 .rpc_argp = &args,
3980                 .rpc_resp = &res,
3981         };
3982
3983         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3984 }
3985
3986 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3987 {
3988         struct nfs4_exception exception = { };
3989         unsigned long now = jiffies;
3990         int err;
3991
3992         do {
3993                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
3994                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
3995                 if (err == 0) {
3996                         struct nfs_client *clp = server->nfs_client;
3997
3998                         spin_lock(&clp->cl_lock);
3999                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4000                         clp->cl_last_renewal = now;
4001                         spin_unlock(&clp->cl_lock);
4002                         break;
4003                 }
4004                 err = nfs4_handle_exception(server, err, &exception);
4005         } while (exception.retry);
4006         return err;
4007 }
4008
4009 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4010 {
4011         int error;
4012
4013         nfs_fattr_init(fsinfo->fattr);
4014         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4015         if (error == 0) {
4016                 /* block layout checks this! */
4017                 server->pnfs_blksize = fsinfo->blksize;
4018                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4019         }
4020
4021         return error;
4022 }
4023
4024 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4025                 struct nfs_pathconf *pathconf)
4026 {
4027         struct nfs4_pathconf_arg args = {
4028                 .fh = fhandle,
4029                 .bitmask = server->attr_bitmask,
4030         };
4031         struct nfs4_pathconf_res res = {
4032                 .pathconf = pathconf,
4033         };
4034         struct rpc_message msg = {
4035                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4036                 .rpc_argp = &args,
4037                 .rpc_resp = &res,
4038         };
4039
4040         /* None of the pathconf attributes are mandatory to implement */
4041         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4042                 memset(pathconf, 0, sizeof(*pathconf));
4043                 return 0;
4044         }
4045
4046         nfs_fattr_init(pathconf->fattr);
4047         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4048 }
4049
4050 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4051                 struct nfs_pathconf *pathconf)
4052 {
4053         struct nfs4_exception exception = { };
4054         int err;
4055
4056         do {
4057                 err = nfs4_handle_exception(server,
4058                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4059                                 &exception);
4060         } while (exception.retry);
4061         return err;
4062 }
4063
4064 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4065                 const struct nfs_open_context *ctx,
4066                 const struct nfs_lock_context *l_ctx,
4067                 fmode_t fmode)
4068 {
4069         const struct nfs_lockowner *lockowner = NULL;
4070
4071         if (l_ctx != NULL)
4072                 lockowner = &l_ctx->lockowner;
4073         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4074 }
4075 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4076
4077 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4078                 const struct nfs_open_context *ctx,
4079                 const struct nfs_lock_context *l_ctx,
4080                 fmode_t fmode)
4081 {
4082         nfs4_stateid current_stateid;
4083
4084         /* If the current stateid represents a lost lock, then exit */
4085         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4086                 return true;
4087         return nfs4_stateid_match(stateid, &current_stateid);
4088 }
4089
4090 static bool nfs4_error_stateid_expired(int err)
4091 {
4092         switch (err) {
4093         case -NFS4ERR_DELEG_REVOKED:
4094         case -NFS4ERR_ADMIN_REVOKED:
4095         case -NFS4ERR_BAD_STATEID:
4096         case -NFS4ERR_STALE_STATEID:
4097         case -NFS4ERR_OLD_STATEID:
4098         case -NFS4ERR_OPENMODE:
4099         case -NFS4ERR_EXPIRED:
4100                 return true;
4101         }
4102         return false;
4103 }
4104
4105 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4106 {
4107         nfs_invalidate_atime(hdr->inode);
4108 }
4109
4110 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4111 {
4112         struct nfs_server *server = NFS_SERVER(hdr->inode);
4113
4114         trace_nfs4_read(hdr, task->tk_status);
4115         if (nfs4_async_handle_error(task, server,
4116                                     hdr->args.context->state) == -EAGAIN) {
4117                 rpc_restart_call_prepare(task);
4118                 return -EAGAIN;
4119         }
4120
4121         __nfs4_read_done_cb(hdr);
4122         if (task->tk_status > 0)
4123                 renew_lease(server, hdr->timestamp);
4124         return 0;
4125 }
4126
4127 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4128                 struct nfs_pgio_args *args)
4129 {
4130
4131         if (!nfs4_error_stateid_expired(task->tk_status) ||
4132                 nfs4_stateid_is_current(&args->stateid,
4133                                 args->context,
4134                                 args->lock_context,
4135                                 FMODE_READ))
4136                 return false;
4137         rpc_restart_call_prepare(task);
4138         return true;
4139 }
4140
4141 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4142 {
4143
4144         dprintk("--> %s\n", __func__);
4145
4146         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4147                 return -EAGAIN;
4148         if (nfs4_read_stateid_changed(task, &hdr->args))
4149                 return -EAGAIN;
4150         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4151                                     nfs4_read_done_cb(task, hdr);
4152 }
4153
4154 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4155                                  struct rpc_message *msg)
4156 {
4157         hdr->timestamp   = jiffies;
4158         hdr->pgio_done_cb = nfs4_read_done_cb;
4159         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4160         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4161 }
4162
4163 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4164                                       struct nfs_pgio_header *hdr)
4165 {
4166         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4167                         &hdr->args.seq_args,
4168                         &hdr->res.seq_res,
4169                         task))
4170                 return 0;
4171         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4172                                 hdr->args.lock_context,
4173                                 hdr->rw_ops->rw_mode) == -EIO)
4174                 return -EIO;
4175         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4176                 return -EIO;
4177         return 0;
4178 }
4179
4180 static int nfs4_write_done_cb(struct rpc_task *task,
4181                               struct nfs_pgio_header *hdr)
4182 {
4183         struct inode *inode = hdr->inode;
4184         
4185         trace_nfs4_write(hdr, task->tk_status);
4186         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4187                                     hdr->args.context->state) == -EAGAIN) {
4188                 rpc_restart_call_prepare(task);
4189                 return -EAGAIN;
4190         }
4191         if (task->tk_status >= 0) {
4192                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4193                 nfs_post_op_update_inode_force_wcc(inode, &hdr->fattr);
4194         }
4195         return 0;
4196 }
4197
4198 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4199                 struct nfs_pgio_args *args)
4200 {
4201
4202         if (!nfs4_error_stateid_expired(task->tk_status) ||
4203                 nfs4_stateid_is_current(&args->stateid,
4204                                 args->context,
4205                                 args->lock_context,
4206                                 FMODE_WRITE))
4207                 return false;
4208         rpc_restart_call_prepare(task);
4209         return true;
4210 }
4211
4212 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4213 {
4214         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4215                 return -EAGAIN;
4216         if (nfs4_write_stateid_changed(task, &hdr->args))
4217                 return -EAGAIN;
4218         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4219                 nfs4_write_done_cb(task, hdr);
4220 }
4221
4222 static
4223 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4224 {
4225         /* Don't request attributes for pNFS or O_DIRECT writes */
4226         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4227                 return false;
4228         /* Otherwise, request attributes if and only if we don't hold
4229          * a delegation
4230          */
4231         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4232 }
4233
4234 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4235                                   struct rpc_message *msg)
4236 {
4237         struct nfs_server *server = NFS_SERVER(hdr->inode);
4238
4239         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4240                 hdr->args.bitmask = NULL;
4241                 hdr->res.fattr = NULL;
4242         } else
4243                 hdr->args.bitmask = server->cache_consistency_bitmask;
4244
4245         if (!hdr->pgio_done_cb)
4246                 hdr->pgio_done_cb = nfs4_write_done_cb;
4247         hdr->res.server = server;
4248         hdr->timestamp   = jiffies;
4249
4250         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4251         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4252 }
4253
4254 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4255 {
4256         nfs4_setup_sequence(NFS_SERVER(data->inode),
4257                         &data->args.seq_args,
4258                         &data->res.seq_res,
4259                         task);
4260 }
4261
4262 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4263 {
4264         struct inode *inode = data->inode;
4265
4266         trace_nfs4_commit(data, task->tk_status);
4267         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
4268                 rpc_restart_call_prepare(task);
4269                 return -EAGAIN;
4270         }
4271         return 0;
4272 }
4273
4274 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4275 {
4276         if (!nfs4_sequence_done(task, &data->res.seq_res))
4277                 return -EAGAIN;
4278         return data->commit_done_cb(task, data);
4279 }
4280
4281 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4282 {
4283         struct nfs_server *server = NFS_SERVER(data->inode);
4284
4285         if (data->commit_done_cb == NULL)
4286                 data->commit_done_cb = nfs4_commit_done_cb;
4287         data->res.server = server;
4288         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4289         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4290 }
4291
4292 struct nfs4_renewdata {
4293         struct nfs_client       *client;
4294         unsigned long           timestamp;
4295 };
4296
4297 /*
4298  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4299  * standalone procedure for queueing an asynchronous RENEW.
4300  */
4301 static void nfs4_renew_release(void *calldata)
4302 {
4303         struct nfs4_renewdata *data = calldata;
4304         struct nfs_client *clp = data->client;
4305
4306         if (atomic_read(&clp->cl_count) > 1)
4307                 nfs4_schedule_state_renewal(clp);
4308         nfs_put_client(clp);
4309         kfree(data);
4310 }
4311
4312 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4313 {
4314         struct nfs4_renewdata *data = calldata;
4315         struct nfs_client *clp = data->client;
4316         unsigned long timestamp = data->timestamp;
4317
4318         trace_nfs4_renew_async(clp, task->tk_status);
4319         switch (task->tk_status) {
4320         case 0:
4321                 break;
4322         case -NFS4ERR_LEASE_MOVED:
4323                 nfs4_schedule_lease_moved_recovery(clp);
4324                 break;
4325         default:
4326                 /* Unless we're shutting down, schedule state recovery! */
4327                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4328                         return;
4329                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4330                         nfs4_schedule_lease_recovery(clp);
4331                         return;
4332                 }
4333                 nfs4_schedule_path_down_recovery(clp);
4334         }
4335         do_renew_lease(clp, timestamp);
4336 }
4337
4338 static const struct rpc_call_ops nfs4_renew_ops = {
4339         .rpc_call_done = nfs4_renew_done,
4340         .rpc_release = nfs4_renew_release,
4341 };
4342
4343 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4344 {
4345         struct rpc_message msg = {
4346                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4347                 .rpc_argp       = clp,
4348                 .rpc_cred       = cred,
4349         };
4350         struct nfs4_renewdata *data;
4351
4352         if (renew_flags == 0)
4353                 return 0;
4354         if (!atomic_inc_not_zero(&clp->cl_count))
4355                 return -EIO;
4356         data = kmalloc(sizeof(*data), GFP_NOFS);
4357         if (data == NULL)
4358                 return -ENOMEM;
4359         data->client = clp;
4360         data->timestamp = jiffies;
4361         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4362                         &nfs4_renew_ops, data);
4363 }
4364
4365 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4366 {
4367         struct rpc_message msg = {
4368                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4369                 .rpc_argp       = clp,
4370                 .rpc_cred       = cred,
4371         };
4372         unsigned long now = jiffies;
4373         int status;
4374
4375         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4376         if (status < 0)
4377                 return status;
4378         do_renew_lease(clp, now);
4379         return 0;
4380 }
4381
4382 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4383 {
4384         return server->caps & NFS_CAP_ACLS;
4385 }
4386
4387 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4388  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4389  * the stack.
4390  */
4391 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4392
4393 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4394                 struct page **pages, unsigned int *pgbase)
4395 {
4396         struct page *newpage, **spages;
4397         int rc = 0;
4398         size_t len;
4399         spages = pages;
4400
4401         do {
4402                 len = min_t(size_t, PAGE_SIZE, buflen);
4403                 newpage = alloc_page(GFP_KERNEL);
4404
4405                 if (newpage == NULL)
4406                         goto unwind;
4407                 memcpy(page_address(newpage), buf, len);
4408                 buf += len;
4409                 buflen -= len;
4410                 *pages++ = newpage;
4411                 rc++;
4412         } while (buflen != 0);
4413
4414         return rc;
4415
4416 unwind:
4417         for(; rc > 0; rc--)
4418                 __free_page(spages[rc-1]);
4419         return -ENOMEM;
4420 }
4421
4422 struct nfs4_cached_acl {
4423         int cached;
4424         size_t len;
4425         char data[0];
4426 };
4427
4428 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4429 {
4430         struct nfs_inode *nfsi = NFS_I(inode);
4431
4432         spin_lock(&inode->i_lock);
4433         kfree(nfsi->nfs4_acl);
4434         nfsi->nfs4_acl = acl;
4435         spin_unlock(&inode->i_lock);
4436 }
4437
4438 static void nfs4_zap_acl_attr(struct inode *inode)
4439 {
4440         nfs4_set_cached_acl(inode, NULL);
4441 }
4442
4443 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4444 {
4445         struct nfs_inode *nfsi = NFS_I(inode);
4446         struct nfs4_cached_acl *acl;
4447         int ret = -ENOENT;
4448
4449         spin_lock(&inode->i_lock);
4450         acl = nfsi->nfs4_acl;
4451         if (acl == NULL)
4452                 goto out;
4453         if (buf == NULL) /* user is just asking for length */
4454                 goto out_len;
4455         if (acl->cached == 0)
4456                 goto out;
4457         ret = -ERANGE; /* see getxattr(2) man page */
4458         if (acl->len > buflen)
4459                 goto out;
4460         memcpy(buf, acl->data, acl->len);
4461 out_len:
4462         ret = acl->len;
4463 out:
4464         spin_unlock(&inode->i_lock);
4465         return ret;
4466 }
4467
4468 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4469 {
4470         struct nfs4_cached_acl *acl;
4471         size_t buflen = sizeof(*acl) + acl_len;
4472
4473         if (buflen <= PAGE_SIZE) {
4474                 acl = kmalloc(buflen, GFP_KERNEL);
4475                 if (acl == NULL)
4476                         goto out;
4477                 acl->cached = 1;
4478                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4479         } else {
4480                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4481                 if (acl == NULL)
4482                         goto out;
4483                 acl->cached = 0;
4484         }
4485         acl->len = acl_len;
4486 out:
4487         nfs4_set_cached_acl(inode, acl);
4488 }
4489
4490 /*
4491  * The getxattr API returns the required buffer length when called with a
4492  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4493  * the required buf.  On a NULL buf, we send a page of data to the server
4494  * guessing that the ACL request can be serviced by a page. If so, we cache
4495  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4496  * the cache. If not so, we throw away the page, and cache the required
4497  * length. The next getxattr call will then produce another round trip to
4498  * the server, this time with the input buf of the required size.
4499  */
4500 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4501 {
4502         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4503         struct nfs_getaclargs args = {
4504                 .fh = NFS_FH(inode),
4505                 .acl_pages = pages,
4506                 .acl_len = buflen,
4507         };
4508         struct nfs_getaclres res = {
4509                 .acl_len = buflen,
4510         };
4511         struct rpc_message msg = {
4512                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4513                 .rpc_argp = &args,
4514                 .rpc_resp = &res,
4515         };
4516         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4517         int ret = -ENOMEM, i;
4518
4519         /* As long as we're doing a round trip to the server anyway,
4520          * let's be prepared for a page of acl data. */
4521         if (npages == 0)
4522                 npages = 1;
4523         if (npages > ARRAY_SIZE(pages))
4524                 return -ERANGE;
4525
4526         for (i = 0; i < npages; i++) {
4527                 pages[i] = alloc_page(GFP_KERNEL);
4528                 if (!pages[i])
4529                         goto out_free;
4530         }
4531
4532         /* for decoding across pages */
4533         res.acl_scratch = alloc_page(GFP_KERNEL);
4534         if (!res.acl_scratch)
4535                 goto out_free;
4536
4537         args.acl_len = npages * PAGE_SIZE;
4538         args.acl_pgbase = 0;
4539
4540         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4541                 __func__, buf, buflen, npages, args.acl_len);
4542         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4543                              &msg, &args.seq_args, &res.seq_res, 0);
4544         if (ret)
4545                 goto out_free;
4546
4547         /* Handle the case where the passed-in buffer is too short */
4548         if (res.acl_flags & NFS4_ACL_TRUNC) {
4549                 /* Did the user only issue a request for the acl length? */
4550                 if (buf == NULL)
4551                         goto out_ok;
4552                 ret = -ERANGE;
4553                 goto out_free;
4554         }
4555         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4556         if (buf) {
4557                 if (res.acl_len > buflen) {
4558                         ret = -ERANGE;
4559                         goto out_free;
4560                 }
4561                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4562         }
4563 out_ok:
4564         ret = res.acl_len;
4565 out_free:
4566         for (i = 0; i < npages; i++)
4567                 if (pages[i])
4568                         __free_page(pages[i]);
4569         if (res.acl_scratch)
4570                 __free_page(res.acl_scratch);
4571         return ret;
4572 }
4573
4574 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4575 {
4576         struct nfs4_exception exception = { };
4577         ssize_t ret;
4578         do {
4579                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4580                 trace_nfs4_get_acl(inode, ret);
4581                 if (ret >= 0)
4582                         break;
4583                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4584         } while (exception.retry);
4585         return ret;
4586 }
4587
4588 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4589 {
4590         struct nfs_server *server = NFS_SERVER(inode);
4591         int ret;
4592
4593         if (!nfs4_server_supports_acls(server))
4594                 return -EOPNOTSUPP;
4595         ret = nfs_revalidate_inode(server, inode);
4596         if (ret < 0)
4597                 return ret;
4598         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4599                 nfs_zap_acl_cache(inode);
4600         ret = nfs4_read_cached_acl(inode, buf, buflen);
4601         if (ret != -ENOENT)
4602                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4603                  * but no cached acl data, just the acl length */
4604                 return ret;
4605         return nfs4_get_acl_uncached(inode, buf, buflen);
4606 }
4607
4608 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4609 {
4610         struct nfs_server *server = NFS_SERVER(inode);
4611         struct page *pages[NFS4ACL_MAXPAGES];
4612         struct nfs_setaclargs arg = {
4613                 .fh             = NFS_FH(inode),
4614                 .acl_pages      = pages,
4615                 .acl_len        = buflen,
4616         };
4617         struct nfs_setaclres res;
4618         struct rpc_message msg = {
4619                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4620                 .rpc_argp       = &arg,
4621                 .rpc_resp       = &res,
4622         };
4623         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4624         int ret, i;
4625
4626         if (!nfs4_server_supports_acls(server))
4627                 return -EOPNOTSUPP;
4628         if (npages > ARRAY_SIZE(pages))
4629                 return -ERANGE;
4630         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4631         if (i < 0)
4632                 return i;
4633         nfs4_inode_return_delegation(inode);
4634         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4635
4636         /*
4637          * Free each page after tx, so the only ref left is
4638          * held by the network stack
4639          */
4640         for (; i > 0; i--)
4641                 put_page(pages[i-1]);
4642
4643         /*
4644          * Acl update can result in inode attribute update.
4645          * so mark the attribute cache invalid.
4646          */
4647         spin_lock(&inode->i_lock);
4648         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4649         spin_unlock(&inode->i_lock);
4650         nfs_access_zap_cache(inode);
4651         nfs_zap_acl_cache(inode);
4652         return ret;
4653 }
4654
4655 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4656 {
4657         struct nfs4_exception exception = { };
4658         int err;
4659         do {
4660                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4661                 trace_nfs4_set_acl(inode, err);
4662                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4663                                 &exception);
4664         } while (exception.retry);
4665         return err;
4666 }
4667
4668 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4669 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4670                                         size_t buflen)
4671 {
4672         struct nfs_server *server = NFS_SERVER(inode);
4673         struct nfs_fattr fattr;
4674         struct nfs4_label label = {0, 0, buflen, buf};
4675
4676         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4677         struct nfs4_getattr_arg arg = {
4678                 .fh             = NFS_FH(inode),
4679                 .bitmask        = bitmask,
4680         };
4681         struct nfs4_getattr_res res = {
4682                 .fattr          = &fattr,
4683                 .label          = &label,
4684                 .server         = server,
4685         };
4686         struct rpc_message msg = {
4687                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4688                 .rpc_argp       = &arg,
4689                 .rpc_resp       = &res,
4690         };
4691         int ret;
4692
4693         nfs_fattr_init(&fattr);
4694
4695         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4696         if (ret)
4697                 return ret;
4698         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4699                 return -ENOENT;
4700         if (buflen < label.len)
4701                 return -ERANGE;
4702         return 0;
4703 }
4704
4705 static int nfs4_get_security_label(struct inode *inode, void *buf,
4706                                         size_t buflen)
4707 {
4708         struct nfs4_exception exception = { };
4709         int err;
4710
4711         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4712                 return -EOPNOTSUPP;
4713
4714         do {
4715                 err = _nfs4_get_security_label(inode, buf, buflen);
4716                 trace_nfs4_get_security_label(inode, err);
4717                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4718                                 &exception);
4719         } while (exception.retry);
4720         return err;
4721 }
4722
4723 static int _nfs4_do_set_security_label(struct inode *inode,
4724                 struct nfs4_label *ilabel,
4725                 struct nfs_fattr *fattr,
4726                 struct nfs4_label *olabel)
4727 {
4728
4729         struct iattr sattr = {0};
4730         struct nfs_server *server = NFS_SERVER(inode);
4731         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4732         struct nfs_setattrargs arg = {
4733                 .fh             = NFS_FH(inode),
4734                 .iap            = &sattr,
4735                 .server         = server,
4736                 .bitmask        = bitmask,
4737                 .label          = ilabel,
4738         };
4739         struct nfs_setattrres res = {
4740                 .fattr          = fattr,
4741                 .label          = olabel,
4742                 .server         = server,
4743         };
4744         struct rpc_message msg = {
4745                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4746                 .rpc_argp       = &arg,
4747                 .rpc_resp       = &res,
4748         };
4749         int status;
4750
4751         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4752
4753         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4754         if (status)
4755                 dprintk("%s failed: %d\n", __func__, status);
4756
4757         return status;
4758 }
4759
4760 static int nfs4_do_set_security_label(struct inode *inode,
4761                 struct nfs4_label *ilabel,
4762                 struct nfs_fattr *fattr,
4763                 struct nfs4_label *olabel)
4764 {
4765         struct nfs4_exception exception = { };
4766         int err;
4767
4768         do {
4769                 err = _nfs4_do_set_security_label(inode, ilabel,
4770                                 fattr, olabel);
4771                 trace_nfs4_set_security_label(inode, err);
4772                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4773                                 &exception);
4774         } while (exception.retry);
4775         return err;
4776 }
4777
4778 static int
4779 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4780 {
4781         struct nfs4_label ilabel, *olabel = NULL;
4782         struct nfs_fattr fattr;
4783         struct rpc_cred *cred;
4784         struct inode *inode = dentry->d_inode;
4785         int status;
4786
4787         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4788                 return -EOPNOTSUPP;
4789
4790         nfs_fattr_init(&fattr);
4791
4792         ilabel.pi = 0;
4793         ilabel.lfs = 0;
4794         ilabel.label = (char *)buf;
4795         ilabel.len = buflen;
4796
4797         cred = rpc_lookup_cred();
4798         if (IS_ERR(cred))
4799                 return PTR_ERR(cred);
4800
4801         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4802         if (IS_ERR(olabel)) {
4803                 status = -PTR_ERR(olabel);
4804                 goto out;
4805         }
4806
4807         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4808         if (status == 0)
4809                 nfs_setsecurity(inode, &fattr, olabel);
4810
4811         nfs4_label_free(olabel);
4812 out:
4813         put_rpccred(cred);
4814         return status;
4815 }
4816 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4817
4818
4819 static int
4820 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4821 {
4822         struct nfs_client *clp = server->nfs_client;
4823
4824         if (task->tk_status >= 0)
4825                 return 0;
4826         switch(task->tk_status) {
4827                 case -NFS4ERR_DELEG_REVOKED:
4828                 case -NFS4ERR_ADMIN_REVOKED:
4829                 case -NFS4ERR_BAD_STATEID:
4830                         if (state == NULL)
4831                                 break;
4832                         nfs_remove_bad_delegation(state->inode);
4833                 case -NFS4ERR_OPENMODE:
4834                         if (state == NULL)
4835                                 break;
4836                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4837                                 goto recovery_failed;
4838                         goto wait_on_recovery;
4839                 case -NFS4ERR_EXPIRED:
4840                         if (state != NULL) {
4841                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4842                                         goto recovery_failed;
4843                         }
4844                 case -NFS4ERR_STALE_STATEID:
4845                 case -NFS4ERR_STALE_CLIENTID:
4846                         nfs4_schedule_lease_recovery(clp);
4847                         goto wait_on_recovery;
4848                 case -NFS4ERR_MOVED:
4849                         if (nfs4_schedule_migration_recovery(server) < 0)
4850                                 goto recovery_failed;
4851                         goto wait_on_recovery;
4852                 case -NFS4ERR_LEASE_MOVED:
4853                         nfs4_schedule_lease_moved_recovery(clp);
4854                         goto wait_on_recovery;
4855 #if defined(CONFIG_NFS_V4_1)
4856                 case -NFS4ERR_BADSESSION:
4857                 case -NFS4ERR_BADSLOT:
4858                 case -NFS4ERR_BAD_HIGH_SLOT:
4859                 case -NFS4ERR_DEADSESSION:
4860                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4861                 case -NFS4ERR_SEQ_FALSE_RETRY:
4862                 case -NFS4ERR_SEQ_MISORDERED:
4863                         dprintk("%s ERROR %d, Reset session\n", __func__,
4864                                 task->tk_status);
4865                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4866                         goto wait_on_recovery;
4867 #endif /* CONFIG_NFS_V4_1 */
4868                 case -NFS4ERR_DELAY:
4869                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4870                 case -NFS4ERR_GRACE:
4871                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4872                 case -NFS4ERR_RETRY_UNCACHED_REP:
4873                 case -NFS4ERR_OLD_STATEID:
4874                         goto restart_call;
4875         }
4876         task->tk_status = nfs4_map_errors(task->tk_status);
4877         return 0;
4878 recovery_failed:
4879         task->tk_status = -EIO;
4880         return 0;
4881 wait_on_recovery:
4882         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4883         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4884                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4885         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4886                 goto recovery_failed;
4887 restart_call:
4888         task->tk_status = 0;
4889         return -EAGAIN;
4890 }
4891
4892 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4893                                     nfs4_verifier *bootverf)
4894 {
4895         __be32 verf[2];
4896
4897         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4898                 /* An impossible timestamp guarantees this value
4899                  * will never match a generated boot time. */
4900                 verf[0] = 0;
4901                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4902         } else {
4903                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4904                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4905                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4906         }
4907         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4908 }
4909
4910 static unsigned int
4911 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4912                                    char *buf, size_t len)
4913 {
4914         unsigned int result;
4915
4916         rcu_read_lock();
4917         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4918                                 clp->cl_ipaddr,
4919                                 rpc_peeraddr2str(clp->cl_rpcclient,
4920                                                         RPC_DISPLAY_ADDR),
4921                                 rpc_peeraddr2str(clp->cl_rpcclient,
4922                                                         RPC_DISPLAY_PROTO));
4923         rcu_read_unlock();
4924         return result;
4925 }
4926
4927 static unsigned int
4928 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4929                                 char *buf, size_t len)
4930 {
4931         const char *nodename = clp->cl_rpcclient->cl_nodename;
4932
4933         if (nfs4_client_id_uniquifier[0] != '\0')
4934                 return scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4935                                 clp->rpc_ops->version,
4936                                 clp->cl_minorversion,
4937                                 nfs4_client_id_uniquifier,
4938                                 nodename);
4939         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4940                                 clp->rpc_ops->version, clp->cl_minorversion,
4941                                 nodename);
4942 }
4943
4944 /*
4945  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
4946  * services.  Advertise one based on the address family of the
4947  * clientaddr.
4948  */
4949 static unsigned int
4950 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
4951 {
4952         if (strchr(clp->cl_ipaddr, ':') != NULL)
4953                 return scnprintf(buf, len, "tcp6");
4954         else
4955                 return scnprintf(buf, len, "tcp");
4956 }
4957
4958 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
4959 {
4960         struct nfs4_setclientid *sc = calldata;
4961
4962         if (task->tk_status == 0)
4963                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
4964 }
4965
4966 static const struct rpc_call_ops nfs4_setclientid_ops = {
4967         .rpc_call_done = nfs4_setclientid_done,
4968 };
4969
4970 /**
4971  * nfs4_proc_setclientid - Negotiate client ID
4972  * @clp: state data structure
4973  * @program: RPC program for NFSv4 callback service
4974  * @port: IP port number for NFS4 callback service
4975  * @cred: RPC credential to use for this call
4976  * @res: where to place the result
4977  *
4978  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4979  */
4980 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4981                 unsigned short port, struct rpc_cred *cred,
4982                 struct nfs4_setclientid_res *res)
4983 {
4984         nfs4_verifier sc_verifier;
4985         struct nfs4_setclientid setclientid = {
4986                 .sc_verifier = &sc_verifier,
4987                 .sc_prog = program,
4988                 .sc_cb_ident = clp->cl_cb_ident,
4989         };
4990         struct rpc_message msg = {
4991                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4992                 .rpc_argp = &setclientid,
4993                 .rpc_resp = res,
4994                 .rpc_cred = cred,
4995         };
4996         struct rpc_task *task;
4997         struct rpc_task_setup task_setup_data = {
4998                 .rpc_client = clp->cl_rpcclient,
4999                 .rpc_message = &msg,
5000                 .callback_ops = &nfs4_setclientid_ops,
5001                 .callback_data = &setclientid,
5002                 .flags = RPC_TASK_TIMEOUT,
5003         };
5004         int status;
5005
5006         /* nfs_client_id4 */
5007         nfs4_init_boot_verifier(clp, &sc_verifier);
5008         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5009                 setclientid.sc_name_len =
5010                                 nfs4_init_uniform_client_string(clp,
5011                                                 setclientid.sc_name,
5012                                                 sizeof(setclientid.sc_name));
5013         else
5014                 setclientid.sc_name_len =
5015                                 nfs4_init_nonuniform_client_string(clp,
5016                                                 setclientid.sc_name,
5017                                                 sizeof(setclientid.sc_name));
5018         /* cb_client4 */
5019         setclientid.sc_netid_len =
5020                                 nfs4_init_callback_netid(clp,
5021                                                 setclientid.sc_netid,
5022                                                 sizeof(setclientid.sc_netid));
5023         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5024                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5025                                 clp->cl_ipaddr, port >> 8, port & 255);
5026
5027         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5028                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5029                 setclientid.sc_name_len, setclientid.sc_name);
5030         task = rpc_run_task(&task_setup_data);
5031         if (IS_ERR(task)) {
5032                 status = PTR_ERR(task);
5033                 goto out;
5034         }
5035         status = task->tk_status;
5036         if (setclientid.sc_cred) {
5037                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5038                 put_rpccred(setclientid.sc_cred);
5039         }
5040         rpc_put_task(task);
5041 out:
5042         trace_nfs4_setclientid(clp, status);
5043         dprintk("NFS reply setclientid: %d\n", status);
5044         return status;
5045 }
5046
5047 /**
5048  * nfs4_proc_setclientid_confirm - Confirm client ID
5049  * @clp: state data structure
5050  * @res: result of a previous SETCLIENTID
5051  * @cred: RPC credential to use for this call
5052  *
5053  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5054  */
5055 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5056                 struct nfs4_setclientid_res *arg,
5057                 struct rpc_cred *cred)
5058 {
5059         struct rpc_message msg = {
5060                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5061                 .rpc_argp = arg,
5062                 .rpc_cred = cred,
5063         };
5064         int status;
5065
5066         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5067                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5068                 clp->cl_clientid);
5069         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5070         trace_nfs4_setclientid_confirm(clp, status);
5071         dprintk("NFS reply setclientid_confirm: %d\n", status);
5072         return status;
5073 }
5074
5075 struct nfs4_delegreturndata {
5076         struct nfs4_delegreturnargs args;
5077         struct nfs4_delegreturnres res;
5078         struct nfs_fh fh;
5079         nfs4_stateid stateid;
5080         unsigned long timestamp;
5081         struct nfs_fattr fattr;
5082         int rpc_status;
5083         struct inode *inode;
5084         bool roc;
5085         u32 roc_barrier;
5086 };
5087
5088 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5089 {
5090         struct nfs4_delegreturndata *data = calldata;
5091
5092         if (!nfs4_sequence_done(task, &data->res.seq_res))
5093                 return;
5094
5095         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5096         switch (task->tk_status) {
5097         case 0:
5098                 renew_lease(data->res.server, data->timestamp);
5099         case -NFS4ERR_ADMIN_REVOKED:
5100         case -NFS4ERR_DELEG_REVOKED:
5101         case -NFS4ERR_BAD_STATEID:
5102         case -NFS4ERR_OLD_STATEID:
5103         case -NFS4ERR_STALE_STATEID:
5104         case -NFS4ERR_EXPIRED:
5105                 task->tk_status = 0;
5106                 if (data->roc)
5107                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5108                 break;
5109         default:
5110                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
5111                                 -EAGAIN) {
5112                         rpc_restart_call_prepare(task);
5113                         return;
5114                 }
5115         }
5116         data->rpc_status = task->tk_status;
5117 }
5118
5119 static void nfs4_delegreturn_release(void *calldata)
5120 {
5121         struct nfs4_delegreturndata *data = calldata;
5122
5123         if (data->roc)
5124                 pnfs_roc_release(data->inode);
5125         kfree(calldata);
5126 }
5127
5128 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5129 {
5130         struct nfs4_delegreturndata *d_data;
5131
5132         d_data = (struct nfs4_delegreturndata *)data;
5133
5134         if (d_data->roc &&
5135             pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5136                 return;
5137
5138         nfs4_setup_sequence(d_data->res.server,
5139                         &d_data->args.seq_args,
5140                         &d_data->res.seq_res,
5141                         task);
5142 }
5143
5144 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5145         .rpc_call_prepare = nfs4_delegreturn_prepare,
5146         .rpc_call_done = nfs4_delegreturn_done,
5147         .rpc_release = nfs4_delegreturn_release,
5148 };
5149
5150 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5151 {
5152         struct nfs4_delegreturndata *data;
5153         struct nfs_server *server = NFS_SERVER(inode);
5154         struct rpc_task *task;
5155         struct rpc_message msg = {
5156                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5157                 .rpc_cred = cred,
5158         };
5159         struct rpc_task_setup task_setup_data = {
5160                 .rpc_client = server->client,
5161                 .rpc_message = &msg,
5162                 .callback_ops = &nfs4_delegreturn_ops,
5163                 .flags = RPC_TASK_ASYNC,
5164         };
5165         int status = 0;
5166
5167         data = kzalloc(sizeof(*data), GFP_NOFS);
5168         if (data == NULL)
5169                 return -ENOMEM;
5170         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5171         data->args.fhandle = &data->fh;
5172         data->args.stateid = &data->stateid;
5173         data->args.bitmask = server->cache_consistency_bitmask;
5174         nfs_copy_fh(&data->fh, NFS_FH(inode));
5175         nfs4_stateid_copy(&data->stateid, stateid);
5176         data->res.fattr = &data->fattr;
5177         data->res.server = server;
5178         nfs_fattr_init(data->res.fattr);
5179         data->timestamp = jiffies;
5180         data->rpc_status = 0;
5181         data->inode = inode;
5182         data->roc = list_empty(&NFS_I(inode)->open_files) ?
5183                     pnfs_roc(inode) : false;
5184
5185         task_setup_data.callback_data = data;
5186         msg.rpc_argp = &data->args;
5187         msg.rpc_resp = &data->res;
5188         task = rpc_run_task(&task_setup_data);
5189         if (IS_ERR(task))
5190                 return PTR_ERR(task);
5191         if (!issync)
5192                 goto out;
5193         status = nfs4_wait_for_completion_rpc_task(task);
5194         if (status != 0)
5195                 goto out;
5196         status = data->rpc_status;
5197         if (status == 0)
5198                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5199         else
5200                 nfs_refresh_inode(inode, &data->fattr);
5201 out:
5202         rpc_put_task(task);
5203         return status;
5204 }
5205
5206 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5207 {
5208         struct nfs_server *server = NFS_SERVER(inode);
5209         struct nfs4_exception exception = { };
5210         int err;
5211         do {
5212                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5213                 trace_nfs4_delegreturn(inode, err);
5214                 switch (err) {
5215                         case -NFS4ERR_STALE_STATEID:
5216                         case -NFS4ERR_EXPIRED:
5217                         case 0:
5218                                 return 0;
5219                 }
5220                 err = nfs4_handle_exception(server, err, &exception);
5221         } while (exception.retry);
5222         return err;
5223 }
5224
5225 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5226 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5227
5228 /* 
5229  * sleep, with exponential backoff, and retry the LOCK operation. 
5230  */
5231 static unsigned long
5232 nfs4_set_lock_task_retry(unsigned long timeout)
5233 {
5234         freezable_schedule_timeout_killable_unsafe(timeout);
5235         timeout <<= 1;
5236         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5237                 return NFS4_LOCK_MAXTIMEOUT;
5238         return timeout;
5239 }
5240
5241 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5242 {
5243         struct inode *inode = state->inode;
5244         struct nfs_server *server = NFS_SERVER(inode);
5245         struct nfs_client *clp = server->nfs_client;
5246         struct nfs_lockt_args arg = {
5247                 .fh = NFS_FH(inode),
5248                 .fl = request,
5249         };
5250         struct nfs_lockt_res res = {
5251                 .denied = request,
5252         };
5253         struct rpc_message msg = {
5254                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5255                 .rpc_argp       = &arg,
5256                 .rpc_resp       = &res,
5257                 .rpc_cred       = state->owner->so_cred,
5258         };
5259         struct nfs4_lock_state *lsp;
5260         int status;
5261
5262         arg.lock_owner.clientid = clp->cl_clientid;
5263         status = nfs4_set_lock_state(state, request);
5264         if (status != 0)
5265                 goto out;
5266         lsp = request->fl_u.nfs4_fl.owner;
5267         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5268         arg.lock_owner.s_dev = server->s_dev;
5269         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5270         switch (status) {
5271                 case 0:
5272                         request->fl_type = F_UNLCK;
5273                         break;
5274                 case -NFS4ERR_DENIED:
5275                         status = 0;
5276         }
5277         request->fl_ops->fl_release_private(request);
5278         request->fl_ops = NULL;
5279 out:
5280         return status;
5281 }
5282
5283 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5284 {
5285         struct nfs4_exception exception = { };
5286         int err;
5287
5288         do {
5289                 err = _nfs4_proc_getlk(state, cmd, request);
5290                 trace_nfs4_get_lock(request, state, cmd, err);
5291                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5292                                 &exception);
5293         } while (exception.retry);
5294         return err;
5295 }
5296
5297 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5298 {
5299         int res = 0;
5300         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5301                 case FL_POSIX:
5302                         res = posix_lock_file_wait(file, fl);
5303                         break;
5304                 case FL_FLOCK:
5305                         res = flock_lock_file_wait(file, fl);
5306                         break;
5307                 default:
5308                         BUG();
5309         }
5310         return res;
5311 }
5312
5313 struct nfs4_unlockdata {
5314         struct nfs_locku_args arg;
5315         struct nfs_locku_res res;
5316         struct nfs4_lock_state *lsp;
5317         struct nfs_open_context *ctx;
5318         struct file_lock fl;
5319         const struct nfs_server *server;
5320         unsigned long timestamp;
5321 };
5322
5323 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5324                 struct nfs_open_context *ctx,
5325                 struct nfs4_lock_state *lsp,
5326                 struct nfs_seqid *seqid)
5327 {
5328         struct nfs4_unlockdata *p;
5329         struct inode *inode = lsp->ls_state->inode;
5330
5331         p = kzalloc(sizeof(*p), GFP_NOFS);
5332         if (p == NULL)
5333                 return NULL;
5334         p->arg.fh = NFS_FH(inode);
5335         p->arg.fl = &p->fl;
5336         p->arg.seqid = seqid;
5337         p->res.seqid = seqid;
5338         p->arg.stateid = &lsp->ls_stateid;
5339         p->lsp = lsp;
5340         atomic_inc(&lsp->ls_count);
5341         /* Ensure we don't close file until we're done freeing locks! */
5342         p->ctx = get_nfs_open_context(ctx);
5343         memcpy(&p->fl, fl, sizeof(p->fl));
5344         p->server = NFS_SERVER(inode);
5345         return p;
5346 }
5347
5348 static void nfs4_locku_release_calldata(void *data)
5349 {
5350         struct nfs4_unlockdata *calldata = data;
5351         nfs_free_seqid(calldata->arg.seqid);
5352         nfs4_put_lock_state(calldata->lsp);
5353         put_nfs_open_context(calldata->ctx);
5354         kfree(calldata);
5355 }
5356
5357 static void nfs4_locku_done(struct rpc_task *task, void *data)
5358 {
5359         struct nfs4_unlockdata *calldata = data;
5360
5361         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5362                 return;
5363         switch (task->tk_status) {
5364                 case 0:
5365                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
5366                                         &calldata->res.stateid);
5367                         renew_lease(calldata->server, calldata->timestamp);
5368                         break;
5369                 case -NFS4ERR_BAD_STATEID:
5370                 case -NFS4ERR_OLD_STATEID:
5371                 case -NFS4ERR_STALE_STATEID:
5372                 case -NFS4ERR_EXPIRED:
5373                         break;
5374                 default:
5375                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
5376                                 rpc_restart_call_prepare(task);
5377         }
5378         nfs_release_seqid(calldata->arg.seqid);
5379 }
5380
5381 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5382 {
5383         struct nfs4_unlockdata *calldata = data;
5384
5385         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5386                 goto out_wait;
5387         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5388                 /* Note: exit _without_ running nfs4_locku_done */
5389                 goto out_no_action;
5390         }
5391         calldata->timestamp = jiffies;
5392         if (nfs4_setup_sequence(calldata->server,
5393                                 &calldata->arg.seq_args,
5394                                 &calldata->res.seq_res,
5395                                 task) != 0)
5396                 nfs_release_seqid(calldata->arg.seqid);
5397         return;
5398 out_no_action:
5399         task->tk_action = NULL;
5400 out_wait:
5401         nfs4_sequence_done(task, &calldata->res.seq_res);
5402 }
5403
5404 static const struct rpc_call_ops nfs4_locku_ops = {
5405         .rpc_call_prepare = nfs4_locku_prepare,
5406         .rpc_call_done = nfs4_locku_done,
5407         .rpc_release = nfs4_locku_release_calldata,
5408 };
5409
5410 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5411                 struct nfs_open_context *ctx,
5412                 struct nfs4_lock_state *lsp,
5413                 struct nfs_seqid *seqid)
5414 {
5415         struct nfs4_unlockdata *data;
5416         struct rpc_message msg = {
5417                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5418                 .rpc_cred = ctx->cred,
5419         };
5420         struct rpc_task_setup task_setup_data = {
5421                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5422                 .rpc_message = &msg,
5423                 .callback_ops = &nfs4_locku_ops,
5424                 .workqueue = nfsiod_workqueue,
5425                 .flags = RPC_TASK_ASYNC,
5426         };
5427
5428         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5429                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5430
5431         /* Ensure this is an unlock - when canceling a lock, the
5432          * canceled lock is passed in, and it won't be an unlock.
5433          */
5434         fl->fl_type = F_UNLCK;
5435
5436         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5437         if (data == NULL) {
5438                 nfs_free_seqid(seqid);
5439                 return ERR_PTR(-ENOMEM);
5440         }
5441
5442         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5443         msg.rpc_argp = &data->arg;
5444         msg.rpc_resp = &data->res;
5445         task_setup_data.callback_data = data;
5446         return rpc_run_task(&task_setup_data);
5447 }
5448
5449 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5450 {
5451         struct inode *inode = state->inode;
5452         struct nfs4_state_owner *sp = state->owner;
5453         struct nfs_inode *nfsi = NFS_I(inode);
5454         struct nfs_seqid *seqid;
5455         struct nfs4_lock_state *lsp;
5456         struct rpc_task *task;
5457         int status = 0;
5458         unsigned char fl_flags = request->fl_flags;
5459
5460         status = nfs4_set_lock_state(state, request);
5461         /* Unlock _before_ we do the RPC call */
5462         request->fl_flags |= FL_EXISTS;
5463         /* Exclude nfs_delegation_claim_locks() */
5464         mutex_lock(&sp->so_delegreturn_mutex);
5465         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5466         down_read(&nfsi->rwsem);
5467         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5468                 up_read(&nfsi->rwsem);
5469                 mutex_unlock(&sp->so_delegreturn_mutex);
5470                 goto out;
5471         }
5472         up_read(&nfsi->rwsem);
5473         mutex_unlock(&sp->so_delegreturn_mutex);
5474         if (status != 0)
5475                 goto out;
5476         /* Is this a delegated lock? */
5477         lsp = request->fl_u.nfs4_fl.owner;
5478         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5479                 goto out;
5480         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5481         status = -ENOMEM;
5482         if (seqid == NULL)
5483                 goto out;
5484         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5485         status = PTR_ERR(task);
5486         if (IS_ERR(task))
5487                 goto out;
5488         status = nfs4_wait_for_completion_rpc_task(task);
5489         rpc_put_task(task);
5490 out:
5491         request->fl_flags = fl_flags;
5492         trace_nfs4_unlock(request, state, F_SETLK, status);
5493         return status;
5494 }
5495
5496 struct nfs4_lockdata {
5497         struct nfs_lock_args arg;
5498         struct nfs_lock_res res;
5499         struct nfs4_lock_state *lsp;
5500         struct nfs_open_context *ctx;
5501         struct file_lock fl;
5502         unsigned long timestamp;
5503         int rpc_status;
5504         int cancelled;
5505         struct nfs_server *server;
5506 };
5507
5508 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5509                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5510                 gfp_t gfp_mask)
5511 {
5512         struct nfs4_lockdata *p;
5513         struct inode *inode = lsp->ls_state->inode;
5514         struct nfs_server *server = NFS_SERVER(inode);
5515
5516         p = kzalloc(sizeof(*p), gfp_mask);
5517         if (p == NULL)
5518                 return NULL;
5519
5520         p->arg.fh = NFS_FH(inode);
5521         p->arg.fl = &p->fl;
5522         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5523         if (p->arg.open_seqid == NULL)
5524                 goto out_free;
5525         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
5526         if (p->arg.lock_seqid == NULL)
5527                 goto out_free_seqid;
5528         p->arg.lock_stateid = &lsp->ls_stateid;
5529         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5530         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5531         p->arg.lock_owner.s_dev = server->s_dev;
5532         p->res.lock_seqid = p->arg.lock_seqid;
5533         p->lsp = lsp;
5534         p->server = server;
5535         atomic_inc(&lsp->ls_count);
5536         p->ctx = get_nfs_open_context(ctx);
5537         memcpy(&p->fl, fl, sizeof(p->fl));
5538         return p;
5539 out_free_seqid:
5540         nfs_free_seqid(p->arg.open_seqid);
5541 out_free:
5542         kfree(p);
5543         return NULL;
5544 }
5545
5546 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5547 {
5548         struct nfs4_lockdata *data = calldata;
5549         struct nfs4_state *state = data->lsp->ls_state;
5550
5551         dprintk("%s: begin!\n", __func__);
5552         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5553                 goto out_wait;
5554         /* Do we need to do an open_to_lock_owner? */
5555         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
5556                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5557                         goto out_release_lock_seqid;
5558                 }
5559                 data->arg.open_stateid = &state->open_stateid;
5560                 data->arg.new_lock_owner = 1;
5561                 data->res.open_seqid = data->arg.open_seqid;
5562         } else
5563                 data->arg.new_lock_owner = 0;
5564         if (!nfs4_valid_open_stateid(state)) {
5565                 data->rpc_status = -EBADF;
5566                 task->tk_action = NULL;
5567                 goto out_release_open_seqid;
5568         }
5569         data->timestamp = jiffies;
5570         if (nfs4_setup_sequence(data->server,
5571                                 &data->arg.seq_args,
5572                                 &data->res.seq_res,
5573                                 task) == 0)
5574                 return;
5575 out_release_open_seqid:
5576         nfs_release_seqid(data->arg.open_seqid);
5577 out_release_lock_seqid:
5578         nfs_release_seqid(data->arg.lock_seqid);
5579 out_wait:
5580         nfs4_sequence_done(task, &data->res.seq_res);
5581         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5582 }
5583
5584 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5585 {
5586         struct nfs4_lockdata *data = calldata;
5587
5588         dprintk("%s: begin!\n", __func__);
5589
5590         if (!nfs4_sequence_done(task, &data->res.seq_res))
5591                 return;
5592
5593         data->rpc_status = task->tk_status;
5594         if (data->arg.new_lock_owner != 0) {
5595                 if (data->rpc_status == 0)
5596                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
5597                 else
5598                         goto out;
5599         }
5600         if (data->rpc_status == 0) {
5601                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
5602                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
5603                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
5604         }
5605 out:
5606         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5607 }
5608
5609 static void nfs4_lock_release(void *calldata)
5610 {
5611         struct nfs4_lockdata *data = calldata;
5612
5613         dprintk("%s: begin!\n", __func__);
5614         nfs_free_seqid(data->arg.open_seqid);
5615         if (data->cancelled != 0) {
5616                 struct rpc_task *task;
5617                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5618                                 data->arg.lock_seqid);
5619                 if (!IS_ERR(task))
5620                         rpc_put_task_async(task);
5621                 dprintk("%s: cancelling lock!\n", __func__);
5622         } else
5623                 nfs_free_seqid(data->arg.lock_seqid);
5624         nfs4_put_lock_state(data->lsp);
5625         put_nfs_open_context(data->ctx);
5626         kfree(data);
5627         dprintk("%s: done!\n", __func__);
5628 }
5629
5630 static const struct rpc_call_ops nfs4_lock_ops = {
5631         .rpc_call_prepare = nfs4_lock_prepare,
5632         .rpc_call_done = nfs4_lock_done,
5633         .rpc_release = nfs4_lock_release,
5634 };
5635
5636 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5637 {
5638         switch (error) {
5639         case -NFS4ERR_ADMIN_REVOKED:
5640         case -NFS4ERR_BAD_STATEID:
5641                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5642                 if (new_lock_owner != 0 ||
5643                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5644                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5645                 break;
5646         case -NFS4ERR_STALE_STATEID:
5647                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5648         case -NFS4ERR_EXPIRED:
5649                 nfs4_schedule_lease_recovery(server->nfs_client);
5650         };
5651 }
5652
5653 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5654 {
5655         struct nfs4_lockdata *data;
5656         struct rpc_task *task;
5657         struct rpc_message msg = {
5658                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5659                 .rpc_cred = state->owner->so_cred,
5660         };
5661         struct rpc_task_setup task_setup_data = {
5662                 .rpc_client = NFS_CLIENT(state->inode),
5663                 .rpc_message = &msg,
5664                 .callback_ops = &nfs4_lock_ops,
5665                 .workqueue = nfsiod_workqueue,
5666                 .flags = RPC_TASK_ASYNC,
5667         };
5668         int ret;
5669
5670         dprintk("%s: begin!\n", __func__);
5671         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5672                         fl->fl_u.nfs4_fl.owner,
5673                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5674         if (data == NULL)
5675                 return -ENOMEM;
5676         if (IS_SETLKW(cmd))
5677                 data->arg.block = 1;
5678         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5679         msg.rpc_argp = &data->arg;
5680         msg.rpc_resp = &data->res;
5681         task_setup_data.callback_data = data;
5682         if (recovery_type > NFS_LOCK_NEW) {
5683                 if (recovery_type == NFS_LOCK_RECLAIM)
5684                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5685                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5686         }
5687         task = rpc_run_task(&task_setup_data);
5688         if (IS_ERR(task))
5689                 return PTR_ERR(task);
5690         ret = nfs4_wait_for_completion_rpc_task(task);
5691         if (ret == 0) {
5692                 ret = data->rpc_status;
5693                 if (ret)
5694                         nfs4_handle_setlk_error(data->server, data->lsp,
5695                                         data->arg.new_lock_owner, ret);
5696         } else
5697                 data->cancelled = 1;
5698         rpc_put_task(task);
5699         dprintk("%s: done, ret = %d!\n", __func__, ret);
5700         return ret;
5701 }
5702
5703 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5704 {
5705         struct nfs_server *server = NFS_SERVER(state->inode);
5706         struct nfs4_exception exception = {
5707                 .inode = state->inode,
5708         };
5709         int err;
5710
5711         do {
5712                 /* Cache the lock if possible... */
5713                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5714                         return 0;
5715                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5716                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5717                 if (err != -NFS4ERR_DELAY)
5718                         break;
5719                 nfs4_handle_exception(server, err, &exception);
5720         } while (exception.retry);
5721         return err;
5722 }
5723
5724 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5725 {
5726         struct nfs_server *server = NFS_SERVER(state->inode);
5727         struct nfs4_exception exception = {
5728                 .inode = state->inode,
5729         };
5730         int err;
5731
5732         err = nfs4_set_lock_state(state, request);
5733         if (err != 0)
5734                 return err;
5735         if (!recover_lost_locks) {
5736                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5737                 return 0;
5738         }
5739         do {
5740                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5741                         return 0;
5742                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5743                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5744                 switch (err) {
5745                 default:
5746                         goto out;
5747                 case -NFS4ERR_GRACE:
5748                 case -NFS4ERR_DELAY:
5749                         nfs4_handle_exception(server, err, &exception);
5750                         err = 0;
5751                 }
5752         } while (exception.retry);
5753 out:
5754         return err;
5755 }
5756
5757 #if defined(CONFIG_NFS_V4_1)
5758 /**
5759  * nfs41_check_expired_locks - possibly free a lock stateid
5760  *
5761  * @state: NFSv4 state for an inode
5762  *
5763  * Returns NFS_OK if recovery for this stateid is now finished.
5764  * Otherwise a negative NFS4ERR value is returned.
5765  */
5766 static int nfs41_check_expired_locks(struct nfs4_state *state)
5767 {
5768         int status, ret = -NFS4ERR_BAD_STATEID;
5769         struct nfs4_lock_state *lsp;
5770         struct nfs_server *server = NFS_SERVER(state->inode);
5771
5772         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5773                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5774                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5775
5776                         status = nfs41_test_stateid(server,
5777                                         &lsp->ls_stateid,
5778                                         cred);
5779                         trace_nfs4_test_lock_stateid(state, lsp, status);
5780                         if (status != NFS_OK) {
5781                                 /* Free the stateid unless the server
5782                                  * informs us the stateid is unrecognized. */
5783                                 if (status != -NFS4ERR_BAD_STATEID)
5784                                         nfs41_free_stateid(server,
5785                                                         &lsp->ls_stateid,
5786                                                         cred);
5787                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5788                                 ret = status;
5789                         }
5790                 }
5791         };
5792
5793         return ret;
5794 }
5795
5796 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5797 {
5798         int status = NFS_OK;
5799
5800         if (test_bit(LK_STATE_IN_USE, &state->flags))
5801                 status = nfs41_check_expired_locks(state);
5802         if (status != NFS_OK)
5803                 status = nfs4_lock_expired(state, request);
5804         return status;
5805 }
5806 #endif
5807
5808 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5809 {
5810         struct nfs4_state_owner *sp = state->owner;
5811         struct nfs_inode *nfsi = NFS_I(state->inode);
5812         unsigned char fl_flags = request->fl_flags;
5813         unsigned int seq;
5814         int status = -ENOLCK;
5815
5816         if ((fl_flags & FL_POSIX) &&
5817                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5818                 goto out;
5819         /* Is this a delegated open? */
5820         status = nfs4_set_lock_state(state, request);
5821         if (status != 0)
5822                 goto out;
5823         request->fl_flags |= FL_ACCESS;
5824         status = do_vfs_lock(request->fl_file, request);
5825         if (status < 0)
5826                 goto out;
5827         down_read(&nfsi->rwsem);
5828         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5829                 /* Yes: cache locks! */
5830                 /* ...but avoid races with delegation recall... */
5831                 request->fl_flags = fl_flags & ~FL_SLEEP;
5832                 status = do_vfs_lock(request->fl_file, request);
5833                 goto out_unlock;
5834         }
5835         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5836         up_read(&nfsi->rwsem);
5837         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5838         if (status != 0)
5839                 goto out;
5840         down_read(&nfsi->rwsem);
5841         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5842                 status = -NFS4ERR_DELAY;
5843                 goto out_unlock;
5844         }
5845         /* Note: we always want to sleep here! */
5846         request->fl_flags = fl_flags | FL_SLEEP;
5847         if (do_vfs_lock(request->fl_file, request) < 0)
5848                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5849                         "manager!\n", __func__);
5850 out_unlock:
5851         up_read(&nfsi->rwsem);
5852 out:
5853         request->fl_flags = fl_flags;
5854         return status;
5855 }
5856
5857 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5858 {
5859         struct nfs4_exception exception = {
5860                 .state = state,
5861                 .inode = state->inode,
5862         };
5863         int err;
5864
5865         do {
5866                 err = _nfs4_proc_setlk(state, cmd, request);
5867                 trace_nfs4_set_lock(request, state, cmd, err);
5868                 if (err == -NFS4ERR_DENIED)
5869                         err = -EAGAIN;
5870                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5871                                 err, &exception);
5872         } while (exception.retry);
5873         return err;
5874 }
5875
5876 static int
5877 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5878 {
5879         struct nfs_open_context *ctx;
5880         struct nfs4_state *state;
5881         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5882         int status;
5883
5884         /* verify open state */
5885         ctx = nfs_file_open_context(filp);
5886         state = ctx->state;
5887
5888         if (request->fl_start < 0 || request->fl_end < 0)
5889                 return -EINVAL;
5890
5891         if (IS_GETLK(cmd)) {
5892                 if (state != NULL)
5893                         return nfs4_proc_getlk(state, F_GETLK, request);
5894                 return 0;
5895         }
5896
5897         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5898                 return -EINVAL;
5899
5900         if (request->fl_type == F_UNLCK) {
5901                 if (state != NULL)
5902                         return nfs4_proc_unlck(state, cmd, request);
5903                 return 0;
5904         }
5905
5906         if (state == NULL)
5907                 return -ENOLCK;
5908         /*
5909          * Don't rely on the VFS having checked the file open mode,
5910          * since it won't do this for flock() locks.
5911          */
5912         switch (request->fl_type) {
5913         case F_RDLCK:
5914                 if (!(filp->f_mode & FMODE_READ))
5915                         return -EBADF;
5916                 break;
5917         case F_WRLCK:
5918                 if (!(filp->f_mode & FMODE_WRITE))
5919                         return -EBADF;
5920         }
5921
5922         do {
5923                 status = nfs4_proc_setlk(state, cmd, request);
5924                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5925                         break;
5926                 timeout = nfs4_set_lock_task_retry(timeout);
5927                 status = -ERESTARTSYS;
5928                 if (signalled())
5929                         break;
5930         } while(status < 0);
5931         return status;
5932 }
5933
5934 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5935 {
5936         struct nfs_server *server = NFS_SERVER(state->inode);
5937         int err;
5938
5939         err = nfs4_set_lock_state(state, fl);
5940         if (err != 0)
5941                 return err;
5942         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5943         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5944 }
5945
5946 struct nfs_release_lockowner_data {
5947         struct nfs4_lock_state *lsp;
5948         struct nfs_server *server;
5949         struct nfs_release_lockowner_args args;
5950         struct nfs_release_lockowner_res res;
5951         unsigned long timestamp;
5952 };
5953
5954 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
5955 {
5956         struct nfs_release_lockowner_data *data = calldata;
5957         struct nfs_server *server = data->server;
5958         nfs40_setup_sequence(server, &data->args.seq_args,
5959                                 &data->res.seq_res, task);
5960         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5961         data->timestamp = jiffies;
5962 }
5963
5964 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
5965 {
5966         struct nfs_release_lockowner_data *data = calldata;
5967         struct nfs_server *server = data->server;
5968
5969         nfs40_sequence_done(task, &data->res.seq_res);
5970
5971         switch (task->tk_status) {
5972         case 0:
5973                 renew_lease(server, data->timestamp);
5974                 break;
5975         case -NFS4ERR_STALE_CLIENTID:
5976         case -NFS4ERR_EXPIRED:
5977                 nfs4_schedule_lease_recovery(server->nfs_client);
5978                 break;
5979         case -NFS4ERR_LEASE_MOVED:
5980         case -NFS4ERR_DELAY:
5981                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN)
5982                         rpc_restart_call_prepare(task);
5983         }
5984 }
5985
5986 static void nfs4_release_lockowner_release(void *calldata)
5987 {
5988         struct nfs_release_lockowner_data *data = calldata;
5989         nfs4_free_lock_state(data->server, data->lsp);
5990         kfree(calldata);
5991 }
5992
5993 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5994         .rpc_call_prepare = nfs4_release_lockowner_prepare,
5995         .rpc_call_done = nfs4_release_lockowner_done,
5996         .rpc_release = nfs4_release_lockowner_release,
5997 };
5998
5999 static void
6000 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6001 {
6002         struct nfs_release_lockowner_data *data;
6003         struct rpc_message msg = {
6004                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6005         };
6006
6007         if (server->nfs_client->cl_mvops->minor_version != 0)
6008                 return;
6009
6010         data = kmalloc(sizeof(*data), GFP_NOFS);
6011         if (!data)
6012                 return;
6013         data->lsp = lsp;
6014         data->server = server;
6015         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6016         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6017         data->args.lock_owner.s_dev = server->s_dev;
6018
6019         msg.rpc_argp = &data->args;
6020         msg.rpc_resp = &data->res;
6021         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6022         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6023 }
6024
6025 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6026
6027 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6028                                    const void *buf, size_t buflen,
6029                                    int flags, int type)
6030 {
6031         if (strcmp(key, "") != 0)
6032                 return -EINVAL;
6033
6034         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
6035 }
6036
6037 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6038                                    void *buf, size_t buflen, int type)
6039 {
6040         if (strcmp(key, "") != 0)
6041                 return -EINVAL;
6042
6043         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
6044 }
6045
6046 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6047                                        size_t list_len, const char *name,
6048                                        size_t name_len, int type)
6049 {
6050         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6051
6052         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
6053                 return 0;
6054
6055         if (list && len <= list_len)
6056                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6057         return len;
6058 }
6059
6060 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6061 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6062 {
6063         return server->caps & NFS_CAP_SECURITY_LABEL;
6064 }
6065
6066 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6067                                    const void *buf, size_t buflen,
6068                                    int flags, int type)
6069 {
6070         if (security_ismaclabel(key))
6071                 return nfs4_set_security_label(dentry, buf, buflen);
6072
6073         return -EOPNOTSUPP;
6074 }
6075
6076 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6077                                    void *buf, size_t buflen, int type)
6078 {
6079         if (security_ismaclabel(key))
6080                 return nfs4_get_security_label(dentry->d_inode, buf, buflen);
6081         return -EOPNOTSUPP;
6082 }
6083
6084 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6085                                        size_t list_len, const char *name,
6086                                        size_t name_len, int type)
6087 {
6088         size_t len = 0;
6089
6090         if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
6091                 len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
6092                 if (list && len <= list_len)
6093                         security_inode_listsecurity(dentry->d_inode, list, len);
6094         }
6095         return len;
6096 }
6097
6098 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6099         .prefix = XATTR_SECURITY_PREFIX,
6100         .list   = nfs4_xattr_list_nfs4_label,
6101         .get    = nfs4_xattr_get_nfs4_label,
6102         .set    = nfs4_xattr_set_nfs4_label,
6103 };
6104 #endif
6105
6106
6107 /*
6108  * nfs_fhget will use either the mounted_on_fileid or the fileid
6109  */
6110 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6111 {
6112         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6113                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6114               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6115               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6116                 return;
6117
6118         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6119                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6120         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6121         fattr->nlink = 2;
6122 }
6123
6124 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6125                                    const struct qstr *name,
6126                                    struct nfs4_fs_locations *fs_locations,
6127                                    struct page *page)
6128 {
6129         struct nfs_server *server = NFS_SERVER(dir);
6130         u32 bitmask[3] = {
6131                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6132         };
6133         struct nfs4_fs_locations_arg args = {
6134                 .dir_fh = NFS_FH(dir),
6135                 .name = name,
6136                 .page = page,
6137                 .bitmask = bitmask,
6138         };
6139         struct nfs4_fs_locations_res res = {
6140                 .fs_locations = fs_locations,
6141         };
6142         struct rpc_message msg = {
6143                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6144                 .rpc_argp = &args,
6145                 .rpc_resp = &res,
6146         };
6147         int status;
6148
6149         dprintk("%s: start\n", __func__);
6150
6151         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6152          * is not supported */
6153         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6154                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6155         else
6156                 bitmask[0] |= FATTR4_WORD0_FILEID;
6157
6158         nfs_fattr_init(&fs_locations->fattr);
6159         fs_locations->server = server;
6160         fs_locations->nlocations = 0;
6161         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6162         dprintk("%s: returned status = %d\n", __func__, status);
6163         return status;
6164 }
6165
6166 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6167                            const struct qstr *name,
6168                            struct nfs4_fs_locations *fs_locations,
6169                            struct page *page)
6170 {
6171         struct nfs4_exception exception = { };
6172         int err;
6173         do {
6174                 err = _nfs4_proc_fs_locations(client, dir, name,
6175                                 fs_locations, page);
6176                 trace_nfs4_get_fs_locations(dir, name, err);
6177                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6178                                 &exception);
6179         } while (exception.retry);
6180         return err;
6181 }
6182
6183 /*
6184  * This operation also signals the server that this client is
6185  * performing migration recovery.  The server can stop returning
6186  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6187  * appended to this compound to identify the client ID which is
6188  * performing recovery.
6189  */
6190 static int _nfs40_proc_get_locations(struct inode *inode,
6191                                      struct nfs4_fs_locations *locations,
6192                                      struct page *page, struct rpc_cred *cred)
6193 {
6194         struct nfs_server *server = NFS_SERVER(inode);
6195         struct rpc_clnt *clnt = server->client;
6196         u32 bitmask[2] = {
6197                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6198         };
6199         struct nfs4_fs_locations_arg args = {
6200                 .clientid       = server->nfs_client->cl_clientid,
6201                 .fh             = NFS_FH(inode),
6202                 .page           = page,
6203                 .bitmask        = bitmask,
6204                 .migration      = 1,            /* skip LOOKUP */
6205                 .renew          = 1,            /* append RENEW */
6206         };
6207         struct nfs4_fs_locations_res res = {
6208                 .fs_locations   = locations,
6209                 .migration      = 1,
6210                 .renew          = 1,
6211         };
6212         struct rpc_message msg = {
6213                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6214                 .rpc_argp       = &args,
6215                 .rpc_resp       = &res,
6216                 .rpc_cred       = cred,
6217         };
6218         unsigned long now = jiffies;
6219         int status;
6220
6221         nfs_fattr_init(&locations->fattr);
6222         locations->server = server;
6223         locations->nlocations = 0;
6224
6225         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6226         nfs4_set_sequence_privileged(&args.seq_args);
6227         status = nfs4_call_sync_sequence(clnt, server, &msg,
6228                                         &args.seq_args, &res.seq_res);
6229         if (status)
6230                 return status;
6231
6232         renew_lease(server, now);
6233         return 0;
6234 }
6235
6236 #ifdef CONFIG_NFS_V4_1
6237
6238 /*
6239  * This operation also signals the server that this client is
6240  * performing migration recovery.  The server can stop asserting
6241  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6242  * performing this operation is identified in the SEQUENCE
6243  * operation in this compound.
6244  *
6245  * When the client supports GETATTR(fs_locations_info), it can
6246  * be plumbed in here.
6247  */
6248 static int _nfs41_proc_get_locations(struct inode *inode,
6249                                      struct nfs4_fs_locations *locations,
6250                                      struct page *page, struct rpc_cred *cred)
6251 {
6252         struct nfs_server *server = NFS_SERVER(inode);
6253         struct rpc_clnt *clnt = server->client;
6254         u32 bitmask[2] = {
6255                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6256         };
6257         struct nfs4_fs_locations_arg args = {
6258                 .fh             = NFS_FH(inode),
6259                 .page           = page,
6260                 .bitmask        = bitmask,
6261                 .migration      = 1,            /* skip LOOKUP */
6262         };
6263         struct nfs4_fs_locations_res res = {
6264                 .fs_locations   = locations,
6265                 .migration      = 1,
6266         };
6267         struct rpc_message msg = {
6268                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6269                 .rpc_argp       = &args,
6270                 .rpc_resp       = &res,
6271                 .rpc_cred       = cred,
6272         };
6273         int status;
6274
6275         nfs_fattr_init(&locations->fattr);
6276         locations->server = server;
6277         locations->nlocations = 0;
6278
6279         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6280         nfs4_set_sequence_privileged(&args.seq_args);
6281         status = nfs4_call_sync_sequence(clnt, server, &msg,
6282                                         &args.seq_args, &res.seq_res);
6283         if (status == NFS4_OK &&
6284             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6285                 status = -NFS4ERR_LEASE_MOVED;
6286         return status;
6287 }
6288
6289 #endif  /* CONFIG_NFS_V4_1 */
6290
6291 /**
6292  * nfs4_proc_get_locations - discover locations for a migrated FSID
6293  * @inode: inode on FSID that is migrating
6294  * @locations: result of query
6295  * @page: buffer
6296  * @cred: credential to use for this operation
6297  *
6298  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6299  * operation failed, or a negative errno if a local error occurred.
6300  *
6301  * On success, "locations" is filled in, but if the server has
6302  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6303  * asserted.
6304  *
6305  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6306  * from this client that require migration recovery.
6307  */
6308 int nfs4_proc_get_locations(struct inode *inode,
6309                             struct nfs4_fs_locations *locations,
6310                             struct page *page, struct rpc_cred *cred)
6311 {
6312         struct nfs_server *server = NFS_SERVER(inode);
6313         struct nfs_client *clp = server->nfs_client;
6314         const struct nfs4_mig_recovery_ops *ops =
6315                                         clp->cl_mvops->mig_recovery_ops;
6316         struct nfs4_exception exception = { };
6317         int status;
6318
6319         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6320                 (unsigned long long)server->fsid.major,
6321                 (unsigned long long)server->fsid.minor,
6322                 clp->cl_hostname);
6323         nfs_display_fhandle(NFS_FH(inode), __func__);
6324
6325         do {
6326                 status = ops->get_locations(inode, locations, page, cred);
6327                 if (status != -NFS4ERR_DELAY)
6328                         break;
6329                 nfs4_handle_exception(server, status, &exception);
6330         } while (exception.retry);
6331         return status;
6332 }
6333
6334 /*
6335  * This operation also signals the server that this client is
6336  * performing "lease moved" recovery.  The server can stop
6337  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6338  * is appended to this compound to identify the client ID which is
6339  * performing recovery.
6340  */
6341 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6342 {
6343         struct nfs_server *server = NFS_SERVER(inode);
6344         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6345         struct rpc_clnt *clnt = server->client;
6346         struct nfs4_fsid_present_arg args = {
6347                 .fh             = NFS_FH(inode),
6348                 .clientid       = clp->cl_clientid,
6349                 .renew          = 1,            /* append RENEW */
6350         };
6351         struct nfs4_fsid_present_res res = {
6352                 .renew          = 1,
6353         };
6354         struct rpc_message msg = {
6355                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6356                 .rpc_argp       = &args,
6357                 .rpc_resp       = &res,
6358                 .rpc_cred       = cred,
6359         };
6360         unsigned long now = jiffies;
6361         int status;
6362
6363         res.fh = nfs_alloc_fhandle();
6364         if (res.fh == NULL)
6365                 return -ENOMEM;
6366
6367         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6368         nfs4_set_sequence_privileged(&args.seq_args);
6369         status = nfs4_call_sync_sequence(clnt, server, &msg,
6370                                                 &args.seq_args, &res.seq_res);
6371         nfs_free_fhandle(res.fh);
6372         if (status)
6373                 return status;
6374
6375         do_renew_lease(clp, now);
6376         return 0;
6377 }
6378
6379 #ifdef CONFIG_NFS_V4_1
6380
6381 /*
6382  * This operation also signals the server that this client is
6383  * performing "lease moved" recovery.  The server can stop asserting
6384  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6385  * this operation is identified in the SEQUENCE operation in this
6386  * compound.
6387  */
6388 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6389 {
6390         struct nfs_server *server = NFS_SERVER(inode);
6391         struct rpc_clnt *clnt = server->client;
6392         struct nfs4_fsid_present_arg args = {
6393                 .fh             = NFS_FH(inode),
6394         };
6395         struct nfs4_fsid_present_res res = {
6396         };
6397         struct rpc_message msg = {
6398                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6399                 .rpc_argp       = &args,
6400                 .rpc_resp       = &res,
6401                 .rpc_cred       = cred,
6402         };
6403         int status;
6404
6405         res.fh = nfs_alloc_fhandle();
6406         if (res.fh == NULL)
6407                 return -ENOMEM;
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         nfs_free_fhandle(res.fh);
6414         if (status == NFS4_OK &&
6415             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6416                 status = -NFS4ERR_LEASE_MOVED;
6417         return status;
6418 }
6419
6420 #endif  /* CONFIG_NFS_V4_1 */
6421
6422 /**
6423  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6424  * @inode: inode on FSID to check
6425  * @cred: credential to use for this operation
6426  *
6427  * Server indicates whether the FSID is present, moved, or not
6428  * recognized.  This operation is necessary to clear a LEASE_MOVED
6429  * condition for this client ID.
6430  *
6431  * Returns NFS4_OK if the FSID is present on this server,
6432  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6433  *  NFS4ERR code if some error occurred on the server, or a
6434  *  negative errno if a local failure occurred.
6435  */
6436 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6437 {
6438         struct nfs_server *server = NFS_SERVER(inode);
6439         struct nfs_client *clp = server->nfs_client;
6440         const struct nfs4_mig_recovery_ops *ops =
6441                                         clp->cl_mvops->mig_recovery_ops;
6442         struct nfs4_exception exception = { };
6443         int status;
6444
6445         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6446                 (unsigned long long)server->fsid.major,
6447                 (unsigned long long)server->fsid.minor,
6448                 clp->cl_hostname);
6449         nfs_display_fhandle(NFS_FH(inode), __func__);
6450
6451         do {
6452                 status = ops->fsid_present(inode, cred);
6453                 if (status != -NFS4ERR_DELAY)
6454                         break;
6455                 nfs4_handle_exception(server, status, &exception);
6456         } while (exception.retry);
6457         return status;
6458 }
6459
6460 /**
6461  * If 'use_integrity' is true and the state managment nfs_client
6462  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6463  * and the machine credential as per RFC3530bis and RFC5661 Security
6464  * Considerations sections. Otherwise, just use the user cred with the
6465  * filesystem's rpc_client.
6466  */
6467 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6468 {
6469         int status;
6470         struct nfs4_secinfo_arg args = {
6471                 .dir_fh = NFS_FH(dir),
6472                 .name   = name,
6473         };
6474         struct nfs4_secinfo_res res = {
6475                 .flavors     = flavors,
6476         };
6477         struct rpc_message msg = {
6478                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6479                 .rpc_argp = &args,
6480                 .rpc_resp = &res,
6481         };
6482         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6483         struct rpc_cred *cred = NULL;
6484
6485         if (use_integrity) {
6486                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6487                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6488                 msg.rpc_cred = cred;
6489         }
6490
6491         dprintk("NFS call  secinfo %s\n", name->name);
6492
6493         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6494                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6495
6496         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6497                                 &res.seq_res, 0);
6498         dprintk("NFS reply  secinfo: %d\n", status);
6499
6500         if (cred)
6501                 put_rpccred(cred);
6502
6503         return status;
6504 }
6505
6506 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6507                       struct nfs4_secinfo_flavors *flavors)
6508 {
6509         struct nfs4_exception exception = { };
6510         int err;
6511         do {
6512                 err = -NFS4ERR_WRONGSEC;
6513
6514                 /* try to use integrity protection with machine cred */
6515                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6516                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6517
6518                 /*
6519                  * if unable to use integrity protection, or SECINFO with
6520                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6521                  * disallowed by spec, but exists in deployed servers) use
6522                  * the current filesystem's rpc_client and the user cred.
6523                  */
6524                 if (err == -NFS4ERR_WRONGSEC)
6525                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6526
6527                 trace_nfs4_secinfo(dir, name, err);
6528                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6529                                 &exception);
6530         } while (exception.retry);
6531         return err;
6532 }
6533
6534 #ifdef CONFIG_NFS_V4_1
6535 /*
6536  * Check the exchange flags returned by the server for invalid flags, having
6537  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6538  * DS flags set.
6539  */
6540 static int nfs4_check_cl_exchange_flags(u32 flags)
6541 {
6542         if (flags & ~EXCHGID4_FLAG_MASK_R)
6543                 goto out_inval;
6544         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6545             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6546                 goto out_inval;
6547         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6548                 goto out_inval;
6549         return NFS_OK;
6550 out_inval:
6551         return -NFS4ERR_INVAL;
6552 }
6553
6554 static bool
6555 nfs41_same_server_scope(struct nfs41_server_scope *a,
6556                         struct nfs41_server_scope *b)
6557 {
6558         if (a->server_scope_sz == b->server_scope_sz &&
6559             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6560                 return true;
6561
6562         return false;
6563 }
6564
6565 /*
6566  * nfs4_proc_bind_conn_to_session()
6567  *
6568  * The 4.1 client currently uses the same TCP connection for the
6569  * fore and backchannel.
6570  */
6571 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6572 {
6573         int status;
6574         struct nfs41_bind_conn_to_session_res res;
6575         struct rpc_message msg = {
6576                 .rpc_proc =
6577                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6578                 .rpc_argp = clp,
6579                 .rpc_resp = &res,
6580                 .rpc_cred = cred,
6581         };
6582
6583         dprintk("--> %s\n", __func__);
6584
6585         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
6586         if (unlikely(res.session == NULL)) {
6587                 status = -ENOMEM;
6588                 goto out;
6589         }
6590
6591         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6592         trace_nfs4_bind_conn_to_session(clp, status);
6593         if (status == 0) {
6594                 if (memcmp(res.session->sess_id.data,
6595                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6596                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6597                         status = -EIO;
6598                         goto out_session;
6599                 }
6600                 if (res.dir != NFS4_CDFS4_BOTH) {
6601                         dprintk("NFS: %s: Unexpected direction from server\n",
6602                                 __func__);
6603                         status = -EIO;
6604                         goto out_session;
6605                 }
6606                 if (res.use_conn_in_rdma_mode) {
6607                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6608                                 __func__);
6609                         status = -EIO;
6610                         goto out_session;
6611                 }
6612         }
6613 out_session:
6614         kfree(res.session);
6615 out:
6616         dprintk("<-- %s status= %d\n", __func__, status);
6617         return status;
6618 }
6619
6620 /*
6621  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6622  * and operations we'd like to see to enable certain features in the allow map
6623  */
6624 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6625         .how = SP4_MACH_CRED,
6626         .enforce.u.words = {
6627                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6628                       1 << (OP_EXCHANGE_ID - 32) |
6629                       1 << (OP_CREATE_SESSION - 32) |
6630                       1 << (OP_DESTROY_SESSION - 32) |
6631                       1 << (OP_DESTROY_CLIENTID - 32)
6632         },
6633         .allow.u.words = {
6634                 [0] = 1 << (OP_CLOSE) |
6635                       1 << (OP_LOCKU) |
6636                       1 << (OP_COMMIT),
6637                 [1] = 1 << (OP_SECINFO - 32) |
6638                       1 << (OP_SECINFO_NO_NAME - 32) |
6639                       1 << (OP_TEST_STATEID - 32) |
6640                       1 << (OP_FREE_STATEID - 32) |
6641                       1 << (OP_WRITE - 32)
6642         }
6643 };
6644
6645 /*
6646  * Select the state protection mode for client `clp' given the server results
6647  * from exchange_id in `sp'.
6648  *
6649  * Returns 0 on success, negative errno otherwise.
6650  */
6651 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6652                                  struct nfs41_state_protection *sp)
6653 {
6654         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6655                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6656                       1 << (OP_EXCHANGE_ID - 32) |
6657                       1 << (OP_CREATE_SESSION - 32) |
6658                       1 << (OP_DESTROY_SESSION - 32) |
6659                       1 << (OP_DESTROY_CLIENTID - 32)
6660         };
6661         unsigned int i;
6662
6663         if (sp->how == SP4_MACH_CRED) {
6664                 /* Print state protect result */
6665                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6666                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6667                         if (test_bit(i, sp->enforce.u.longs))
6668                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6669                         if (test_bit(i, sp->allow.u.longs))
6670                                 dfprintk(MOUNT, "  allow op %d\n", i);
6671                 }
6672
6673                 /* make sure nothing is on enforce list that isn't supported */
6674                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6675                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6676                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6677                                 return -EINVAL;
6678                         }
6679                 }
6680
6681                 /*
6682                  * Minimal mode - state operations are allowed to use machine
6683                  * credential.  Note this already happens by default, so the
6684                  * client doesn't have to do anything more than the negotiation.
6685                  *
6686                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6687                  *       we're already using the machine cred for exchange_id
6688                  *       and will never use a different cred.
6689                  */
6690                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6691                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6692                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6693                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6694                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6695                         dfprintk(MOUNT, "  minimal mode enabled\n");
6696                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6697                 } else {
6698                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6699                         return -EINVAL;
6700                 }
6701
6702                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6703                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6704                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6705                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6706                 }
6707
6708                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6709                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6710                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6711                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6712                 }
6713
6714                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6715                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6716                         dfprintk(MOUNT, "  stateid mode enabled\n");
6717                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6718                 }
6719
6720                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6721                         dfprintk(MOUNT, "  write mode enabled\n");
6722                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6723                 }
6724
6725                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6726                         dfprintk(MOUNT, "  commit mode enabled\n");
6727                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6728                 }
6729         }
6730
6731         return 0;
6732 }
6733
6734 /*
6735  * _nfs4_proc_exchange_id()
6736  *
6737  * Wrapper for EXCHANGE_ID operation.
6738  */
6739 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6740         u32 sp4_how)
6741 {
6742         nfs4_verifier verifier;
6743         struct nfs41_exchange_id_args args = {
6744                 .verifier = &verifier,
6745                 .client = clp,
6746 #ifdef CONFIG_NFS_V4_1_MIGRATION
6747                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6748                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6749                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6750 #else
6751                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6752                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6753 #endif
6754         };
6755         struct nfs41_exchange_id_res res = {
6756                 0
6757         };
6758         int status;
6759         struct rpc_message msg = {
6760                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6761                 .rpc_argp = &args,
6762                 .rpc_resp = &res,
6763                 .rpc_cred = cred,
6764         };
6765
6766         nfs4_init_boot_verifier(clp, &verifier);
6767         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6768                                                         sizeof(args.id));
6769         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6770                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6771                 args.id_len, args.id);
6772
6773         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6774                                         GFP_NOFS);
6775         if (unlikely(res.server_owner == NULL)) {
6776                 status = -ENOMEM;
6777                 goto out;
6778         }
6779
6780         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6781                                         GFP_NOFS);
6782         if (unlikely(res.server_scope == NULL)) {
6783                 status = -ENOMEM;
6784                 goto out_server_owner;
6785         }
6786
6787         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6788         if (unlikely(res.impl_id == NULL)) {
6789                 status = -ENOMEM;
6790                 goto out_server_scope;
6791         }
6792
6793         switch (sp4_how) {
6794         case SP4_NONE:
6795                 args.state_protect.how = SP4_NONE;
6796                 break;
6797
6798         case SP4_MACH_CRED:
6799                 args.state_protect = nfs4_sp4_mach_cred_request;
6800                 break;
6801
6802         default:
6803                 /* unsupported! */
6804                 WARN_ON_ONCE(1);
6805                 status = -EINVAL;
6806                 goto out_server_scope;
6807         }
6808
6809         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6810         trace_nfs4_exchange_id(clp, status);
6811         if (status == 0)
6812                 status = nfs4_check_cl_exchange_flags(res.flags);
6813
6814         if (status == 0)
6815                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
6816
6817         if (status == 0) {
6818                 clp->cl_clientid = res.clientid;
6819                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
6820                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
6821                         clp->cl_seqid = res.seqid;
6822
6823                 kfree(clp->cl_serverowner);
6824                 clp->cl_serverowner = res.server_owner;
6825                 res.server_owner = NULL;
6826
6827                 /* use the most recent implementation id */
6828                 kfree(clp->cl_implid);
6829                 clp->cl_implid = res.impl_id;
6830
6831                 if (clp->cl_serverscope != NULL &&
6832                     !nfs41_same_server_scope(clp->cl_serverscope,
6833                                              res.server_scope)) {
6834                         dprintk("%s: server_scope mismatch detected\n",
6835                                 __func__);
6836                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6837                         kfree(clp->cl_serverscope);
6838                         clp->cl_serverscope = NULL;
6839                 }
6840
6841                 if (clp->cl_serverscope == NULL) {
6842                         clp->cl_serverscope = res.server_scope;
6843                         goto out;
6844                 }
6845         } else
6846                 kfree(res.impl_id);
6847
6848 out_server_owner:
6849         kfree(res.server_owner);
6850 out_server_scope:
6851         kfree(res.server_scope);
6852 out:
6853         if (clp->cl_implid != NULL)
6854                 dprintk("NFS reply exchange_id: Server Implementation ID: "
6855                         "domain: %s, name: %s, date: %llu,%u\n",
6856                         clp->cl_implid->domain, clp->cl_implid->name,
6857                         clp->cl_implid->date.seconds,
6858                         clp->cl_implid->date.nseconds);
6859         dprintk("NFS reply exchange_id: %d\n", status);
6860         return status;
6861 }
6862
6863 /*
6864  * nfs4_proc_exchange_id()
6865  *
6866  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6867  *
6868  * Since the clientid has expired, all compounds using sessions
6869  * associated with the stale clientid will be returning
6870  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6871  * be in some phase of session reset.
6872  *
6873  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6874  */
6875 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6876 {
6877         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6878         int status;
6879
6880         /* try SP4_MACH_CRED if krb5i/p */
6881         if (authflavor == RPC_AUTH_GSS_KRB5I ||
6882             authflavor == RPC_AUTH_GSS_KRB5P) {
6883                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6884                 if (!status)
6885                         return 0;
6886         }
6887
6888         /* try SP4_NONE */
6889         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6890 }
6891
6892 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6893                 struct rpc_cred *cred)
6894 {
6895         struct rpc_message msg = {
6896                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6897                 .rpc_argp = clp,
6898                 .rpc_cred = cred,
6899         };
6900         int status;
6901
6902         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6903         trace_nfs4_destroy_clientid(clp, status);
6904         if (status)
6905                 dprintk("NFS: Got error %d from the server %s on "
6906                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
6907         return status;
6908 }
6909
6910 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6911                 struct rpc_cred *cred)
6912 {
6913         unsigned int loop;
6914         int ret;
6915
6916         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6917                 ret = _nfs4_proc_destroy_clientid(clp, cred);
6918                 switch (ret) {
6919                 case -NFS4ERR_DELAY:
6920                 case -NFS4ERR_CLIENTID_BUSY:
6921                         ssleep(1);
6922                         break;
6923                 default:
6924                         return ret;
6925                 }
6926         }
6927         return 0;
6928 }
6929
6930 int nfs4_destroy_clientid(struct nfs_client *clp)
6931 {
6932         struct rpc_cred *cred;
6933         int ret = 0;
6934
6935         if (clp->cl_mvops->minor_version < 1)
6936                 goto out;
6937         if (clp->cl_exchange_flags == 0)
6938                 goto out;
6939         if (clp->cl_preserve_clid)
6940                 goto out;
6941         cred = nfs4_get_clid_cred(clp);
6942         ret = nfs4_proc_destroy_clientid(clp, cred);
6943         if (cred)
6944                 put_rpccred(cred);
6945         switch (ret) {
6946         case 0:
6947         case -NFS4ERR_STALE_CLIENTID:
6948                 clp->cl_exchange_flags = 0;
6949         }
6950 out:
6951         return ret;
6952 }
6953
6954 struct nfs4_get_lease_time_data {
6955         struct nfs4_get_lease_time_args *args;
6956         struct nfs4_get_lease_time_res *res;
6957         struct nfs_client *clp;
6958 };
6959
6960 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
6961                                         void *calldata)
6962 {
6963         struct nfs4_get_lease_time_data *data =
6964                         (struct nfs4_get_lease_time_data *)calldata;
6965
6966         dprintk("--> %s\n", __func__);
6967         /* just setup sequence, do not trigger session recovery
6968            since we're invoked within one */
6969         nfs41_setup_sequence(data->clp->cl_session,
6970                         &data->args->la_seq_args,
6971                         &data->res->lr_seq_res,
6972                         task);
6973         dprintk("<-- %s\n", __func__);
6974 }
6975
6976 /*
6977  * Called from nfs4_state_manager thread for session setup, so don't recover
6978  * from sequence operation or clientid errors.
6979  */
6980 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
6981 {
6982         struct nfs4_get_lease_time_data *data =
6983                         (struct nfs4_get_lease_time_data *)calldata;
6984
6985         dprintk("--> %s\n", __func__);
6986         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
6987                 return;
6988         switch (task->tk_status) {
6989         case -NFS4ERR_DELAY:
6990         case -NFS4ERR_GRACE:
6991                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
6992                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
6993                 task->tk_status = 0;
6994                 /* fall through */
6995         case -NFS4ERR_RETRY_UNCACHED_REP:
6996                 rpc_restart_call_prepare(task);
6997                 return;
6998         }
6999         dprintk("<-- %s\n", __func__);
7000 }
7001
7002 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7003         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7004         .rpc_call_done = nfs4_get_lease_time_done,
7005 };
7006
7007 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7008 {
7009         struct rpc_task *task;
7010         struct nfs4_get_lease_time_args args;
7011         struct nfs4_get_lease_time_res res = {
7012                 .lr_fsinfo = fsinfo,
7013         };
7014         struct nfs4_get_lease_time_data data = {
7015                 .args = &args,
7016                 .res = &res,
7017                 .clp = clp,
7018         };
7019         struct rpc_message msg = {
7020                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7021                 .rpc_argp = &args,
7022                 .rpc_resp = &res,
7023         };
7024         struct rpc_task_setup task_setup = {
7025                 .rpc_client = clp->cl_rpcclient,
7026                 .rpc_message = &msg,
7027                 .callback_ops = &nfs4_get_lease_time_ops,
7028                 .callback_data = &data,
7029                 .flags = RPC_TASK_TIMEOUT,
7030         };
7031         int status;
7032
7033         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7034         nfs4_set_sequence_privileged(&args.la_seq_args);
7035         dprintk("--> %s\n", __func__);
7036         task = rpc_run_task(&task_setup);
7037
7038         if (IS_ERR(task))
7039                 status = PTR_ERR(task);
7040         else {
7041                 status = task->tk_status;
7042                 rpc_put_task(task);
7043         }
7044         dprintk("<-- %s return %d\n", __func__, status);
7045
7046         return status;
7047 }
7048
7049 /*
7050  * Initialize the values to be used by the client in CREATE_SESSION
7051  * If nfs4_init_session set the fore channel request and response sizes,
7052  * use them.
7053  *
7054  * Set the back channel max_resp_sz_cached to zero to force the client to
7055  * always set csa_cachethis to FALSE because the current implementation
7056  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7057  */
7058 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7059 {
7060         unsigned int max_rqst_sz, max_resp_sz;
7061
7062         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7063         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7064
7065         /* Fore channel attributes */
7066         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7067         args->fc_attrs.max_resp_sz = max_resp_sz;
7068         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7069         args->fc_attrs.max_reqs = max_session_slots;
7070
7071         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7072                 "max_ops=%u max_reqs=%u\n",
7073                 __func__,
7074                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7075                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7076
7077         /* Back channel attributes */
7078         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7079         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7080         args->bc_attrs.max_resp_sz_cached = 0;
7081         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7082         args->bc_attrs.max_reqs = 1;
7083
7084         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7085                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7086                 __func__,
7087                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7088                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7089                 args->bc_attrs.max_reqs);
7090 }
7091
7092 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7093 {
7094         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7095         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
7096
7097         if (rcvd->max_resp_sz > sent->max_resp_sz)
7098                 return -EINVAL;
7099         /*
7100          * Our requested max_ops is the minimum we need; we're not
7101          * prepared to break up compounds into smaller pieces than that.
7102          * So, no point even trying to continue if the server won't
7103          * cooperate:
7104          */
7105         if (rcvd->max_ops < sent->max_ops)
7106                 return -EINVAL;
7107         if (rcvd->max_reqs == 0)
7108                 return -EINVAL;
7109         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7110                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7111         return 0;
7112 }
7113
7114 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7115 {
7116         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7117         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
7118
7119         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7120                 return -EINVAL;
7121         if (rcvd->max_resp_sz < sent->max_resp_sz)
7122                 return -EINVAL;
7123         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7124                 return -EINVAL;
7125         /* These would render the backchannel useless: */
7126         if (rcvd->max_ops != sent->max_ops)
7127                 return -EINVAL;
7128         if (rcvd->max_reqs != sent->max_reqs)
7129                 return -EINVAL;
7130         return 0;
7131 }
7132
7133 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7134                                      struct nfs4_session *session)
7135 {
7136         int ret;
7137
7138         ret = nfs4_verify_fore_channel_attrs(args, session);
7139         if (ret)
7140                 return ret;
7141         return nfs4_verify_back_channel_attrs(args, session);
7142 }
7143
7144 static int _nfs4_proc_create_session(struct nfs_client *clp,
7145                 struct rpc_cred *cred)
7146 {
7147         struct nfs4_session *session = clp->cl_session;
7148         struct nfs41_create_session_args args = {
7149                 .client = clp,
7150                 .cb_program = NFS4_CALLBACK,
7151         };
7152         struct nfs41_create_session_res res = {
7153                 .client = clp,
7154         };
7155         struct rpc_message msg = {
7156                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7157                 .rpc_argp = &args,
7158                 .rpc_resp = &res,
7159                 .rpc_cred = cred,
7160         };
7161         int status;
7162
7163         nfs4_init_channel_attrs(&args);
7164         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7165
7166         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7167         trace_nfs4_create_session(clp, status);
7168
7169         if (!status) {
7170                 /* Verify the session's negotiated channel_attrs values */
7171                 status = nfs4_verify_channel_attrs(&args, session);
7172                 /* Increment the clientid slot sequence id */
7173                 clp->cl_seqid++;
7174         }
7175
7176         return status;
7177 }
7178
7179 /*
7180  * Issues a CREATE_SESSION operation to the server.
7181  * It is the responsibility of the caller to verify the session is
7182  * expired before calling this routine.
7183  */
7184 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7185 {
7186         int status;
7187         unsigned *ptr;
7188         struct nfs4_session *session = clp->cl_session;
7189
7190         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7191
7192         status = _nfs4_proc_create_session(clp, cred);
7193         if (status)
7194                 goto out;
7195
7196         /* Init or reset the session slot tables */
7197         status = nfs4_setup_session_slot_tables(session);
7198         dprintk("slot table setup returned %d\n", status);
7199         if (status)
7200                 goto out;
7201
7202         ptr = (unsigned *)&session->sess_id.data[0];
7203         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7204                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7205 out:
7206         dprintk("<-- %s\n", __func__);
7207         return status;
7208 }
7209
7210 /*
7211  * Issue the over-the-wire RPC DESTROY_SESSION.
7212  * The caller must serialize access to this routine.
7213  */
7214 int nfs4_proc_destroy_session(struct nfs4_session *session,
7215                 struct rpc_cred *cred)
7216 {
7217         struct rpc_message msg = {
7218                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7219                 .rpc_argp = session,
7220                 .rpc_cred = cred,
7221         };
7222         int status = 0;
7223
7224         dprintk("--> nfs4_proc_destroy_session\n");
7225
7226         /* session is still being setup */
7227         if (session->clp->cl_cons_state != NFS_CS_READY)
7228                 return status;
7229
7230         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7231         trace_nfs4_destroy_session(session->clp, status);
7232
7233         if (status)
7234                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7235                         "Session has been destroyed regardless...\n", status);
7236
7237         dprintk("<-- nfs4_proc_destroy_session\n");
7238         return status;
7239 }
7240
7241 /*
7242  * Renew the cl_session lease.
7243  */
7244 struct nfs4_sequence_data {
7245         struct nfs_client *clp;
7246         struct nfs4_sequence_args args;
7247         struct nfs4_sequence_res res;
7248 };
7249
7250 static void nfs41_sequence_release(void *data)
7251 {
7252         struct nfs4_sequence_data *calldata = data;
7253         struct nfs_client *clp = calldata->clp;
7254
7255         if (atomic_read(&clp->cl_count) > 1)
7256                 nfs4_schedule_state_renewal(clp);
7257         nfs_put_client(clp);
7258         kfree(calldata);
7259 }
7260
7261 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7262 {
7263         switch(task->tk_status) {
7264         case -NFS4ERR_DELAY:
7265                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7266                 return -EAGAIN;
7267         default:
7268                 nfs4_schedule_lease_recovery(clp);
7269         }
7270         return 0;
7271 }
7272
7273 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7274 {
7275         struct nfs4_sequence_data *calldata = data;
7276         struct nfs_client *clp = calldata->clp;
7277
7278         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7279                 return;
7280
7281         trace_nfs4_sequence(clp, task->tk_status);
7282         if (task->tk_status < 0) {
7283                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7284                 if (atomic_read(&clp->cl_count) == 1)
7285                         goto out;
7286
7287                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7288                         rpc_restart_call_prepare(task);
7289                         return;
7290                 }
7291         }
7292         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7293 out:
7294         dprintk("<-- %s\n", __func__);
7295 }
7296
7297 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7298 {
7299         struct nfs4_sequence_data *calldata = data;
7300         struct nfs_client *clp = calldata->clp;
7301         struct nfs4_sequence_args *args;
7302         struct nfs4_sequence_res *res;
7303
7304         args = task->tk_msg.rpc_argp;
7305         res = task->tk_msg.rpc_resp;
7306
7307         nfs41_setup_sequence(clp->cl_session, args, res, task);
7308 }
7309
7310 static const struct rpc_call_ops nfs41_sequence_ops = {
7311         .rpc_call_done = nfs41_sequence_call_done,
7312         .rpc_call_prepare = nfs41_sequence_prepare,
7313         .rpc_release = nfs41_sequence_release,
7314 };
7315
7316 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7317                 struct rpc_cred *cred,
7318                 bool is_privileged)
7319 {
7320         struct nfs4_sequence_data *calldata;
7321         struct rpc_message msg = {
7322                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7323                 .rpc_cred = cred,
7324         };
7325         struct rpc_task_setup task_setup_data = {
7326                 .rpc_client = clp->cl_rpcclient,
7327                 .rpc_message = &msg,
7328                 .callback_ops = &nfs41_sequence_ops,
7329                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7330         };
7331
7332         if (!atomic_inc_not_zero(&clp->cl_count))
7333                 return ERR_PTR(-EIO);
7334         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7335         if (calldata == NULL) {
7336                 nfs_put_client(clp);
7337                 return ERR_PTR(-ENOMEM);
7338         }
7339         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7340         if (is_privileged)
7341                 nfs4_set_sequence_privileged(&calldata->args);
7342         msg.rpc_argp = &calldata->args;
7343         msg.rpc_resp = &calldata->res;
7344         calldata->clp = clp;
7345         task_setup_data.callback_data = calldata;
7346
7347         return rpc_run_task(&task_setup_data);
7348 }
7349
7350 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7351 {
7352         struct rpc_task *task;
7353         int ret = 0;
7354
7355         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7356                 return -EAGAIN;
7357         task = _nfs41_proc_sequence(clp, cred, false);
7358         if (IS_ERR(task))
7359                 ret = PTR_ERR(task);
7360         else
7361                 rpc_put_task_async(task);
7362         dprintk("<-- %s status=%d\n", __func__, ret);
7363         return ret;
7364 }
7365
7366 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7367 {
7368         struct rpc_task *task;
7369         int ret;
7370
7371         task = _nfs41_proc_sequence(clp, cred, true);
7372         if (IS_ERR(task)) {
7373                 ret = PTR_ERR(task);
7374                 goto out;
7375         }
7376         ret = rpc_wait_for_completion_task(task);
7377         if (!ret) {
7378                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7379
7380                 if (task->tk_status == 0)
7381                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7382                 ret = task->tk_status;
7383         }
7384         rpc_put_task(task);
7385 out:
7386         dprintk("<-- %s status=%d\n", __func__, ret);
7387         return ret;
7388 }
7389
7390 struct nfs4_reclaim_complete_data {
7391         struct nfs_client *clp;
7392         struct nfs41_reclaim_complete_args arg;
7393         struct nfs41_reclaim_complete_res res;
7394 };
7395
7396 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7397 {
7398         struct nfs4_reclaim_complete_data *calldata = data;
7399
7400         nfs41_setup_sequence(calldata->clp->cl_session,
7401                         &calldata->arg.seq_args,
7402                         &calldata->res.seq_res,
7403                         task);
7404 }
7405
7406 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7407 {
7408         switch(task->tk_status) {
7409         case 0:
7410         case -NFS4ERR_COMPLETE_ALREADY:
7411         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7412                 break;
7413         case -NFS4ERR_DELAY:
7414                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7415                 /* fall through */
7416         case -NFS4ERR_RETRY_UNCACHED_REP:
7417                 return -EAGAIN;
7418         default:
7419                 nfs4_schedule_lease_recovery(clp);
7420         }
7421         return 0;
7422 }
7423
7424 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7425 {
7426         struct nfs4_reclaim_complete_data *calldata = data;
7427         struct nfs_client *clp = calldata->clp;
7428         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7429
7430         dprintk("--> %s\n", __func__);
7431         if (!nfs41_sequence_done(task, res))
7432                 return;
7433
7434         trace_nfs4_reclaim_complete(clp, task->tk_status);
7435         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7436                 rpc_restart_call_prepare(task);
7437                 return;
7438         }
7439         dprintk("<-- %s\n", __func__);
7440 }
7441
7442 static void nfs4_free_reclaim_complete_data(void *data)
7443 {
7444         struct nfs4_reclaim_complete_data *calldata = data;
7445
7446         kfree(calldata);
7447 }
7448
7449 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7450         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7451         .rpc_call_done = nfs4_reclaim_complete_done,
7452         .rpc_release = nfs4_free_reclaim_complete_data,
7453 };
7454
7455 /*
7456  * Issue a global reclaim complete.
7457  */
7458 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7459                 struct rpc_cred *cred)
7460 {
7461         struct nfs4_reclaim_complete_data *calldata;
7462         struct rpc_task *task;
7463         struct rpc_message msg = {
7464                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7465                 .rpc_cred = cred,
7466         };
7467         struct rpc_task_setup task_setup_data = {
7468                 .rpc_client = clp->cl_rpcclient,
7469                 .rpc_message = &msg,
7470                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7471                 .flags = RPC_TASK_ASYNC,
7472         };
7473         int status = -ENOMEM;
7474
7475         dprintk("--> %s\n", __func__);
7476         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7477         if (calldata == NULL)
7478                 goto out;
7479         calldata->clp = clp;
7480         calldata->arg.one_fs = 0;
7481
7482         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7483         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7484         msg.rpc_argp = &calldata->arg;
7485         msg.rpc_resp = &calldata->res;
7486         task_setup_data.callback_data = calldata;
7487         task = rpc_run_task(&task_setup_data);
7488         if (IS_ERR(task)) {
7489                 status = PTR_ERR(task);
7490                 goto out;
7491         }
7492         status = nfs4_wait_for_completion_rpc_task(task);
7493         if (status == 0)
7494                 status = task->tk_status;
7495         rpc_put_task(task);
7496         return 0;
7497 out:
7498         dprintk("<-- %s status=%d\n", __func__, status);
7499         return status;
7500 }
7501
7502 static void
7503 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7504 {
7505         struct nfs4_layoutget *lgp = calldata;
7506         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7507         struct nfs4_session *session = nfs4_get_session(server);
7508
7509         dprintk("--> %s\n", __func__);
7510         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7511          * right now covering the LAYOUTGET we are about to send.
7512          * However, that is not so catastrophic, and there seems
7513          * to be no way to prevent it completely.
7514          */
7515         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7516                                 &lgp->res.seq_res, task))
7517                 return;
7518         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7519                                           NFS_I(lgp->args.inode)->layout,
7520                                           lgp->args.ctx->state)) {
7521                 rpc_exit(task, NFS4_OK);
7522         }
7523 }
7524
7525 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7526 {
7527         struct nfs4_layoutget *lgp = calldata;
7528         struct inode *inode = lgp->args.inode;
7529         struct nfs_server *server = NFS_SERVER(inode);
7530         struct pnfs_layout_hdr *lo;
7531         struct nfs4_state *state = NULL;
7532         unsigned long timeo, now, giveup;
7533
7534         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7535
7536         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7537                 goto out;
7538
7539         switch (task->tk_status) {
7540         case 0:
7541                 goto out;
7542         /*
7543          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7544          * (or clients) writing to the same RAID stripe
7545          */
7546         case -NFS4ERR_LAYOUTTRYLATER:
7547         /*
7548          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7549          * existing layout before getting a new one).
7550          */
7551         case -NFS4ERR_RECALLCONFLICT:
7552                 timeo = rpc_get_timeout(task->tk_client);
7553                 giveup = lgp->args.timestamp + timeo;
7554                 now = jiffies;
7555                 if (time_after(giveup, now)) {
7556                         unsigned long delay;
7557
7558                         /* Delay for:
7559                          * - Not less then NFS4_POLL_RETRY_MIN.
7560                          * - One last time a jiffie before we give up
7561                          * - exponential backoff (time_now minus start_attempt)
7562                          */
7563                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7564                                     min((giveup - now - 1),
7565                                         now - lgp->args.timestamp));
7566
7567                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7568                                 __func__, delay);
7569                         rpc_delay(task, delay);
7570                         task->tk_status = 0;
7571                         rpc_restart_call_prepare(task);
7572                         goto out; /* Do not call nfs4_async_handle_error() */
7573                 }
7574                 break;
7575         case -NFS4ERR_EXPIRED:
7576         case -NFS4ERR_BAD_STATEID:
7577                 spin_lock(&inode->i_lock);
7578                 lo = NFS_I(inode)->layout;
7579                 if (!lo || list_empty(&lo->plh_segs)) {
7580                         spin_unlock(&inode->i_lock);
7581                         /* If the open stateid was bad, then recover it. */
7582                         state = lgp->args.ctx->state;
7583                 } else {
7584                         LIST_HEAD(head);
7585
7586                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7587                         spin_unlock(&inode->i_lock);
7588                         /* Mark the bad layout state as invalid, then
7589                          * retry using the open stateid. */
7590                         pnfs_free_lseg_list(&head);
7591                 }
7592         }
7593         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
7594                 rpc_restart_call_prepare(task);
7595 out:
7596         dprintk("<-- %s\n", __func__);
7597 }
7598
7599 static size_t max_response_pages(struct nfs_server *server)
7600 {
7601         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7602         return nfs_page_array_len(0, max_resp_sz);
7603 }
7604
7605 static void nfs4_free_pages(struct page **pages, size_t size)
7606 {
7607         int i;
7608
7609         if (!pages)
7610                 return;
7611
7612         for (i = 0; i < size; i++) {
7613                 if (!pages[i])
7614                         break;
7615                 __free_page(pages[i]);
7616         }
7617         kfree(pages);
7618 }
7619
7620 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7621 {
7622         struct page **pages;
7623         int i;
7624
7625         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7626         if (!pages) {
7627                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7628                 return NULL;
7629         }
7630
7631         for (i = 0; i < size; i++) {
7632                 pages[i] = alloc_page(gfp_flags);
7633                 if (!pages[i]) {
7634                         dprintk("%s: failed to allocate page\n", __func__);
7635                         nfs4_free_pages(pages, size);
7636                         return NULL;
7637                 }
7638         }
7639
7640         return pages;
7641 }
7642
7643 static void nfs4_layoutget_release(void *calldata)
7644 {
7645         struct nfs4_layoutget *lgp = calldata;
7646         struct inode *inode = lgp->args.inode;
7647         struct nfs_server *server = NFS_SERVER(inode);
7648         size_t max_pages = max_response_pages(server);
7649
7650         dprintk("--> %s\n", __func__);
7651         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7652         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7653         put_nfs_open_context(lgp->args.ctx);
7654         kfree(calldata);
7655         dprintk("<-- %s\n", __func__);
7656 }
7657
7658 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7659         .rpc_call_prepare = nfs4_layoutget_prepare,
7660         .rpc_call_done = nfs4_layoutget_done,
7661         .rpc_release = nfs4_layoutget_release,
7662 };
7663
7664 struct pnfs_layout_segment *
7665 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7666 {
7667         struct inode *inode = lgp->args.inode;
7668         struct nfs_server *server = NFS_SERVER(inode);
7669         size_t max_pages = max_response_pages(server);
7670         struct rpc_task *task;
7671         struct rpc_message msg = {
7672                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7673                 .rpc_argp = &lgp->args,
7674                 .rpc_resp = &lgp->res,
7675                 .rpc_cred = lgp->cred,
7676         };
7677         struct rpc_task_setup task_setup_data = {
7678                 .rpc_client = server->client,
7679                 .rpc_message = &msg,
7680                 .callback_ops = &nfs4_layoutget_call_ops,
7681                 .callback_data = lgp,
7682                 .flags = RPC_TASK_ASYNC,
7683         };
7684         struct pnfs_layout_segment *lseg = NULL;
7685         int status = 0;
7686
7687         dprintk("--> %s\n", __func__);
7688
7689         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7690         if (!lgp->args.layout.pages) {
7691                 nfs4_layoutget_release(lgp);
7692                 return ERR_PTR(-ENOMEM);
7693         }
7694         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7695         lgp->args.timestamp = jiffies;
7696
7697         lgp->res.layoutp = &lgp->args.layout;
7698         lgp->res.seq_res.sr_slot = NULL;
7699         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7700
7701         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7702         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7703
7704         task = rpc_run_task(&task_setup_data);
7705         if (IS_ERR(task))
7706                 return ERR_CAST(task);
7707         status = nfs4_wait_for_completion_rpc_task(task);
7708         if (status == 0)
7709                 status = task->tk_status;
7710         trace_nfs4_layoutget(lgp->args.ctx,
7711                         &lgp->args.range,
7712                         &lgp->res.range,
7713                         status);
7714         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7715         if (status == 0 && lgp->res.layoutp->len)
7716                 lseg = pnfs_layout_process(lgp);
7717         rpc_put_task(task);
7718         dprintk("<-- %s status=%d\n", __func__, status);
7719         if (status)
7720                 return ERR_PTR(status);
7721         return lseg;
7722 }
7723
7724 static void
7725 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7726 {
7727         struct nfs4_layoutreturn *lrp = calldata;
7728
7729         dprintk("--> %s\n", __func__);
7730         nfs41_setup_sequence(lrp->clp->cl_session,
7731                         &lrp->args.seq_args,
7732                         &lrp->res.seq_res,
7733                         task);
7734 }
7735
7736 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7737 {
7738         struct nfs4_layoutreturn *lrp = calldata;
7739         struct nfs_server *server;
7740
7741         dprintk("--> %s\n", __func__);
7742
7743         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7744                 return;
7745
7746         server = NFS_SERVER(lrp->args.inode);
7747         switch (task->tk_status) {
7748         default:
7749                 task->tk_status = 0;
7750         case 0:
7751                 break;
7752         case -NFS4ERR_DELAY:
7753                 if (nfs4_async_handle_error(task, server, NULL) != -EAGAIN)
7754                         break;
7755                 rpc_restart_call_prepare(task);
7756                 return;
7757         }
7758         dprintk("<-- %s\n", __func__);
7759 }
7760
7761 static void nfs4_layoutreturn_release(void *calldata)
7762 {
7763         struct nfs4_layoutreturn *lrp = calldata;
7764         struct pnfs_layout_hdr *lo = lrp->args.layout;
7765
7766         dprintk("--> %s\n", __func__);
7767         spin_lock(&lo->plh_inode->i_lock);
7768         if (lrp->res.lrs_present)
7769                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7770         lo->plh_block_lgets--;
7771         spin_unlock(&lo->plh_inode->i_lock);
7772         pnfs_put_layout_hdr(lrp->args.layout);
7773         kfree(calldata);
7774         dprintk("<-- %s\n", __func__);
7775 }
7776
7777 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7778         .rpc_call_prepare = nfs4_layoutreturn_prepare,
7779         .rpc_call_done = nfs4_layoutreturn_done,
7780         .rpc_release = nfs4_layoutreturn_release,
7781 };
7782
7783 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
7784 {
7785         struct rpc_task *task;
7786         struct rpc_message msg = {
7787                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7788                 .rpc_argp = &lrp->args,
7789                 .rpc_resp = &lrp->res,
7790                 .rpc_cred = lrp->cred,
7791         };
7792         struct rpc_task_setup task_setup_data = {
7793                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
7794                 .rpc_message = &msg,
7795                 .callback_ops = &nfs4_layoutreturn_call_ops,
7796                 .callback_data = lrp,
7797         };
7798         int status;
7799
7800         dprintk("--> %s\n", __func__);
7801         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7802         task = rpc_run_task(&task_setup_data);
7803         if (IS_ERR(task))
7804                 return PTR_ERR(task);
7805         status = task->tk_status;
7806         trace_nfs4_layoutreturn(lrp->args.inode, status);
7807         dprintk("<-- %s status=%d\n", __func__, status);
7808         rpc_put_task(task);
7809         return status;
7810 }
7811
7812 /*
7813  * Retrieve the list of Data Server devices from the MDS.
7814  */
7815 static int _nfs4_getdevicelist(struct nfs_server *server,
7816                                     const struct nfs_fh *fh,
7817                                     struct pnfs_devicelist *devlist)
7818 {
7819         struct nfs4_getdevicelist_args args = {
7820                 .fh = fh,
7821                 .layoutclass = server->pnfs_curr_ld->id,
7822         };
7823         struct nfs4_getdevicelist_res res = {
7824                 .devlist = devlist,
7825         };
7826         struct rpc_message msg = {
7827                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
7828                 .rpc_argp = &args,
7829                 .rpc_resp = &res,
7830         };
7831         int status;
7832
7833         dprintk("--> %s\n", __func__);
7834         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
7835                                 &res.seq_res, 0);
7836         dprintk("<-- %s status=%d\n", __func__, status);
7837         return status;
7838 }
7839
7840 int nfs4_proc_getdevicelist(struct nfs_server *server,
7841                             const struct nfs_fh *fh,
7842                             struct pnfs_devicelist *devlist)
7843 {
7844         struct nfs4_exception exception = { };
7845         int err;
7846
7847         do {
7848                 err = nfs4_handle_exception(server,
7849                                 _nfs4_getdevicelist(server, fh, devlist),
7850                                 &exception);
7851         } while (exception.retry);
7852
7853         dprintk("%s: err=%d, num_devs=%u\n", __func__,
7854                 err, devlist->num_devs);
7855
7856         return err;
7857 }
7858 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
7859
7860 static int
7861 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7862                 struct pnfs_device *pdev,
7863                 struct rpc_cred *cred)
7864 {
7865         struct nfs4_getdeviceinfo_args args = {
7866                 .pdev = pdev,
7867         };
7868         struct nfs4_getdeviceinfo_res res = {
7869                 .pdev = pdev,
7870         };
7871         struct rpc_message msg = {
7872                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7873                 .rpc_argp = &args,
7874                 .rpc_resp = &res,
7875                 .rpc_cred = cred,
7876         };
7877         int status;
7878
7879         dprintk("--> %s\n", __func__);
7880         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7881         dprintk("<-- %s status=%d\n", __func__, status);
7882
7883         return status;
7884 }
7885
7886 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7887                 struct pnfs_device *pdev,
7888                 struct rpc_cred *cred)
7889 {
7890         struct nfs4_exception exception = { };
7891         int err;
7892
7893         do {
7894                 err = nfs4_handle_exception(server,
7895                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
7896                                         &exception);
7897         } while (exception.retry);
7898         return err;
7899 }
7900 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7901
7902 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7903 {
7904         struct nfs4_layoutcommit_data *data = calldata;
7905         struct nfs_server *server = NFS_SERVER(data->args.inode);
7906         struct nfs4_session *session = nfs4_get_session(server);
7907
7908         nfs41_setup_sequence(session,
7909                         &data->args.seq_args,
7910                         &data->res.seq_res,
7911                         task);
7912 }
7913
7914 static void
7915 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7916 {
7917         struct nfs4_layoutcommit_data *data = calldata;
7918         struct nfs_server *server = NFS_SERVER(data->args.inode);
7919
7920         if (!nfs41_sequence_done(task, &data->res.seq_res))
7921                 return;
7922
7923         switch (task->tk_status) { /* Just ignore these failures */
7924         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7925         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7926         case -NFS4ERR_BADLAYOUT:     /* no layout */
7927         case -NFS4ERR_GRACE:        /* loca_recalim always false */
7928                 task->tk_status = 0;
7929         case 0:
7930                 break;
7931         default:
7932                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
7933                         rpc_restart_call_prepare(task);
7934                         return;
7935                 }
7936         }
7937 }
7938
7939 static void nfs4_layoutcommit_release(void *calldata)
7940 {
7941         struct nfs4_layoutcommit_data *data = calldata;
7942
7943         pnfs_cleanup_layoutcommit(data);
7944         nfs_post_op_update_inode_force_wcc(data->args.inode,
7945                                            data->res.fattr);
7946         put_rpccred(data->cred);
7947         kfree(data);
7948 }
7949
7950 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7951         .rpc_call_prepare = nfs4_layoutcommit_prepare,
7952         .rpc_call_done = nfs4_layoutcommit_done,
7953         .rpc_release = nfs4_layoutcommit_release,
7954 };
7955
7956 int
7957 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7958 {
7959         struct rpc_message msg = {
7960                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7961                 .rpc_argp = &data->args,
7962                 .rpc_resp = &data->res,
7963                 .rpc_cred = data->cred,
7964         };
7965         struct rpc_task_setup task_setup_data = {
7966                 .task = &data->task,
7967                 .rpc_client = NFS_CLIENT(data->args.inode),
7968                 .rpc_message = &msg,
7969                 .callback_ops = &nfs4_layoutcommit_ops,
7970                 .callback_data = data,
7971                 .flags = RPC_TASK_ASYNC,
7972         };
7973         struct rpc_task *task;
7974         int status = 0;
7975
7976         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7977                 "lbw: %llu inode %lu\n",
7978                 data->task.tk_pid, sync,
7979                 data->args.lastbytewritten,
7980                 data->args.inode->i_ino);
7981
7982         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7983         task = rpc_run_task(&task_setup_data);
7984         if (IS_ERR(task))
7985                 return PTR_ERR(task);
7986         if (sync == false)
7987                 goto out;
7988         status = nfs4_wait_for_completion_rpc_task(task);
7989         if (status != 0)
7990                 goto out;
7991         status = task->tk_status;
7992         trace_nfs4_layoutcommit(data->args.inode, status);
7993 out:
7994         dprintk("%s: status %d\n", __func__, status);
7995         rpc_put_task(task);
7996         return status;
7997 }
7998
7999 /**
8000  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8001  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8002  */
8003 static int
8004 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8005                     struct nfs_fsinfo *info,
8006                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8007 {
8008         struct nfs41_secinfo_no_name_args args = {
8009                 .style = SECINFO_STYLE_CURRENT_FH,
8010         };
8011         struct nfs4_secinfo_res res = {
8012                 .flavors = flavors,
8013         };
8014         struct rpc_message msg = {
8015                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8016                 .rpc_argp = &args,
8017                 .rpc_resp = &res,
8018         };
8019         struct rpc_clnt *clnt = server->client;
8020         struct rpc_cred *cred = NULL;
8021         int status;
8022
8023         if (use_integrity) {
8024                 clnt = server->nfs_client->cl_rpcclient;
8025                 cred = nfs4_get_clid_cred(server->nfs_client);
8026                 msg.rpc_cred = cred;
8027         }
8028
8029         dprintk("--> %s\n", __func__);
8030         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8031                                 &res.seq_res, 0);
8032         dprintk("<-- %s status=%d\n", __func__, status);
8033
8034         if (cred)
8035                 put_rpccred(cred);
8036
8037         return status;
8038 }
8039
8040 static int
8041 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8042                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8043 {
8044         struct nfs4_exception exception = { };
8045         int err;
8046         do {
8047                 /* first try using integrity protection */
8048                 err = -NFS4ERR_WRONGSEC;
8049
8050                 /* try to use integrity protection with machine cred */
8051                 if (_nfs4_is_integrity_protected(server->nfs_client))
8052                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8053                                                           flavors, true);
8054
8055                 /*
8056                  * if unable to use integrity protection, or SECINFO with
8057                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8058                  * disallowed by spec, but exists in deployed servers) use
8059                  * the current filesystem's rpc_client and the user cred.
8060                  */
8061                 if (err == -NFS4ERR_WRONGSEC)
8062                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8063                                                           flavors, false);
8064
8065                 switch (err) {
8066                 case 0:
8067                 case -NFS4ERR_WRONGSEC:
8068                 case -ENOTSUPP:
8069                         goto out;
8070                 default:
8071                         err = nfs4_handle_exception(server, err, &exception);
8072                 }
8073         } while (exception.retry);
8074 out:
8075         return err;
8076 }
8077
8078 static int
8079 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8080                     struct nfs_fsinfo *info)
8081 {
8082         int err;
8083         struct page *page;
8084         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8085         struct nfs4_secinfo_flavors *flavors;
8086         struct nfs4_secinfo4 *secinfo;
8087         int i;
8088
8089         page = alloc_page(GFP_KERNEL);
8090         if (!page) {
8091                 err = -ENOMEM;
8092                 goto out;
8093         }
8094
8095         flavors = page_address(page);
8096         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8097
8098         /*
8099          * Fall back on "guess and check" method if
8100          * the server doesn't support SECINFO_NO_NAME
8101          */
8102         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8103                 err = nfs4_find_root_sec(server, fhandle, info);
8104                 goto out_freepage;
8105         }
8106         if (err)
8107                 goto out_freepage;
8108
8109         for (i = 0; i < flavors->num_flavors; i++) {
8110                 secinfo = &flavors->flavors[i];
8111
8112                 switch (secinfo->flavor) {
8113                 case RPC_AUTH_NULL:
8114                 case RPC_AUTH_UNIX:
8115                 case RPC_AUTH_GSS:
8116                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8117                                         &secinfo->flavor_info);
8118                         break;
8119                 default:
8120                         flavor = RPC_AUTH_MAXFLAVOR;
8121                         break;
8122                 }
8123
8124                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8125                         flavor = RPC_AUTH_MAXFLAVOR;
8126
8127                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8128                         err = nfs4_lookup_root_sec(server, fhandle,
8129                                                    info, flavor);
8130                         if (!err)
8131                                 break;
8132                 }
8133         }
8134
8135         if (flavor == RPC_AUTH_MAXFLAVOR)
8136                 err = -EPERM;
8137
8138 out_freepage:
8139         put_page(page);
8140         if (err == -EACCES)
8141                 return -EPERM;
8142 out:
8143         return err;
8144 }
8145
8146 static int _nfs41_test_stateid(struct nfs_server *server,
8147                 nfs4_stateid *stateid,
8148                 struct rpc_cred *cred)
8149 {
8150         int status;
8151         struct nfs41_test_stateid_args args = {
8152                 .stateid = stateid,
8153         };
8154         struct nfs41_test_stateid_res res;
8155         struct rpc_message msg = {
8156                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8157                 .rpc_argp = &args,
8158                 .rpc_resp = &res,
8159                 .rpc_cred = cred,
8160         };
8161         struct rpc_clnt *rpc_client = server->client;
8162
8163         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8164                 &rpc_client, &msg);
8165
8166         dprintk("NFS call  test_stateid %p\n", stateid);
8167         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8168         nfs4_set_sequence_privileged(&args.seq_args);
8169         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8170                         &args.seq_args, &res.seq_res);
8171         if (status != NFS_OK) {
8172                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8173                 return status;
8174         }
8175         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8176         return -res.status;
8177 }
8178
8179 /**
8180  * nfs41_test_stateid - perform a TEST_STATEID operation
8181  *
8182  * @server: server / transport on which to perform the operation
8183  * @stateid: state ID to test
8184  * @cred: credential
8185  *
8186  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8187  * Otherwise a negative NFS4ERR value is returned if the operation
8188  * failed or the state ID is not currently valid.
8189  */
8190 static int nfs41_test_stateid(struct nfs_server *server,
8191                 nfs4_stateid *stateid,
8192                 struct rpc_cred *cred)
8193 {
8194         struct nfs4_exception exception = { };
8195         int err;
8196         do {
8197                 err = _nfs41_test_stateid(server, stateid, cred);
8198                 if (err != -NFS4ERR_DELAY)
8199                         break;
8200                 nfs4_handle_exception(server, err, &exception);
8201         } while (exception.retry);
8202         return err;
8203 }
8204
8205 struct nfs_free_stateid_data {
8206         struct nfs_server *server;
8207         struct nfs41_free_stateid_args args;
8208         struct nfs41_free_stateid_res res;
8209 };
8210
8211 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8212 {
8213         struct nfs_free_stateid_data *data = calldata;
8214         nfs41_setup_sequence(nfs4_get_session(data->server),
8215                         &data->args.seq_args,
8216                         &data->res.seq_res,
8217                         task);
8218 }
8219
8220 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8221 {
8222         struct nfs_free_stateid_data *data = calldata;
8223
8224         nfs41_sequence_done(task, &data->res.seq_res);
8225
8226         switch (task->tk_status) {
8227         case -NFS4ERR_DELAY:
8228                 if (nfs4_async_handle_error(task, data->server, NULL) == -EAGAIN)
8229                         rpc_restart_call_prepare(task);
8230         }
8231 }
8232
8233 static void nfs41_free_stateid_release(void *calldata)
8234 {
8235         kfree(calldata);
8236 }
8237
8238 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8239         .rpc_call_prepare = nfs41_free_stateid_prepare,
8240         .rpc_call_done = nfs41_free_stateid_done,
8241         .rpc_release = nfs41_free_stateid_release,
8242 };
8243
8244 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8245                 nfs4_stateid *stateid,
8246                 struct rpc_cred *cred,
8247                 bool privileged)
8248 {
8249         struct rpc_message msg = {
8250                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8251                 .rpc_cred = cred,
8252         };
8253         struct rpc_task_setup task_setup = {
8254                 .rpc_client = server->client,
8255                 .rpc_message = &msg,
8256                 .callback_ops = &nfs41_free_stateid_ops,
8257                 .flags = RPC_TASK_ASYNC,
8258         };
8259         struct nfs_free_stateid_data *data;
8260
8261         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8262                 &task_setup.rpc_client, &msg);
8263
8264         dprintk("NFS call  free_stateid %p\n", stateid);
8265         data = kmalloc(sizeof(*data), GFP_NOFS);
8266         if (!data)
8267                 return ERR_PTR(-ENOMEM);
8268         data->server = server;
8269         nfs4_stateid_copy(&data->args.stateid, stateid);
8270
8271         task_setup.callback_data = data;
8272
8273         msg.rpc_argp = &data->args;
8274         msg.rpc_resp = &data->res;
8275         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8276         if (privileged)
8277                 nfs4_set_sequence_privileged(&data->args.seq_args);
8278
8279         return rpc_run_task(&task_setup);
8280 }
8281
8282 /**
8283  * nfs41_free_stateid - perform a FREE_STATEID operation
8284  *
8285  * @server: server / transport on which to perform the operation
8286  * @stateid: state ID to release
8287  * @cred: credential
8288  *
8289  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8290  * negative NFS4ERR value is returned.
8291  */
8292 static int nfs41_free_stateid(struct nfs_server *server,
8293                 nfs4_stateid *stateid,
8294                 struct rpc_cred *cred)
8295 {
8296         struct rpc_task *task;
8297         int ret;
8298
8299         task = _nfs41_free_stateid(server, stateid, cred, true);
8300         if (IS_ERR(task))
8301                 return PTR_ERR(task);
8302         ret = rpc_wait_for_completion_task(task);
8303         if (!ret)
8304                 ret = task->tk_status;
8305         rpc_put_task(task);
8306         return ret;
8307 }
8308
8309 static void
8310 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8311 {
8312         struct rpc_task *task;
8313         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8314
8315         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8316         nfs4_free_lock_state(server, lsp);
8317         if (IS_ERR(task))
8318                 return;
8319         rpc_put_task(task);
8320 }
8321
8322 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8323                 const nfs4_stateid *s2)
8324 {
8325         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8326                 return false;
8327
8328         if (s1->seqid == s2->seqid)
8329                 return true;
8330         if (s1->seqid == 0 || s2->seqid == 0)
8331                 return true;
8332
8333         return false;
8334 }
8335
8336 #endif /* CONFIG_NFS_V4_1 */
8337
8338 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8339                 const nfs4_stateid *s2)
8340 {
8341         return nfs4_stateid_match(s1, s2);
8342 }
8343
8344
8345 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8346         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8347         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8348         .recover_open   = nfs4_open_reclaim,
8349         .recover_lock   = nfs4_lock_reclaim,
8350         .establish_clid = nfs4_init_clientid,
8351         .detect_trunking = nfs40_discover_server_trunking,
8352 };
8353
8354 #if defined(CONFIG_NFS_V4_1)
8355 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8356         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8357         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8358         .recover_open   = nfs4_open_reclaim,
8359         .recover_lock   = nfs4_lock_reclaim,
8360         .establish_clid = nfs41_init_clientid,
8361         .reclaim_complete = nfs41_proc_reclaim_complete,
8362         .detect_trunking = nfs41_discover_server_trunking,
8363 };
8364 #endif /* CONFIG_NFS_V4_1 */
8365
8366 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8367         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8368         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8369         .recover_open   = nfs4_open_expired,
8370         .recover_lock   = nfs4_lock_expired,
8371         .establish_clid = nfs4_init_clientid,
8372 };
8373
8374 #if defined(CONFIG_NFS_V4_1)
8375 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8376         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8377         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8378         .recover_open   = nfs41_open_expired,
8379         .recover_lock   = nfs41_lock_expired,
8380         .establish_clid = nfs41_init_clientid,
8381 };
8382 #endif /* CONFIG_NFS_V4_1 */
8383
8384 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8385         .sched_state_renewal = nfs4_proc_async_renew,
8386         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8387         .renew_lease = nfs4_proc_renew,
8388 };
8389
8390 #if defined(CONFIG_NFS_V4_1)
8391 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8392         .sched_state_renewal = nfs41_proc_async_sequence,
8393         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8394         .renew_lease = nfs4_proc_sequence,
8395 };
8396 #endif
8397
8398 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8399         .get_locations = _nfs40_proc_get_locations,
8400         .fsid_present = _nfs40_proc_fsid_present,
8401 };
8402
8403 #if defined(CONFIG_NFS_V4_1)
8404 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8405         .get_locations = _nfs41_proc_get_locations,
8406         .fsid_present = _nfs41_proc_fsid_present,
8407 };
8408 #endif  /* CONFIG_NFS_V4_1 */
8409
8410 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8411         .minor_version = 0,
8412         .init_caps = NFS_CAP_READDIRPLUS
8413                 | NFS_CAP_ATOMIC_OPEN
8414                 | NFS_CAP_CHANGE_ATTR
8415                 | NFS_CAP_POSIX_LOCK,
8416         .init_client = nfs40_init_client,
8417         .shutdown_client = nfs40_shutdown_client,
8418         .match_stateid = nfs4_match_stateid,
8419         .find_root_sec = nfs4_find_root_sec,
8420         .free_lock_state = nfs4_release_lockowner,
8421         .call_sync_ops = &nfs40_call_sync_ops,
8422         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8423         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8424         .state_renewal_ops = &nfs40_state_renewal_ops,
8425         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8426 };
8427
8428 #if defined(CONFIG_NFS_V4_1)
8429 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8430         .minor_version = 1,
8431         .init_caps = NFS_CAP_READDIRPLUS
8432                 | NFS_CAP_ATOMIC_OPEN
8433                 | NFS_CAP_CHANGE_ATTR
8434                 | NFS_CAP_POSIX_LOCK
8435                 | NFS_CAP_STATEID_NFSV41
8436                 | NFS_CAP_ATOMIC_OPEN_V1,
8437         .init_client = nfs41_init_client,
8438         .shutdown_client = nfs41_shutdown_client,
8439         .match_stateid = nfs41_match_stateid,
8440         .find_root_sec = nfs41_find_root_sec,
8441         .free_lock_state = nfs41_free_lock_state,
8442         .call_sync_ops = &nfs41_call_sync_ops,
8443         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8444         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8445         .state_renewal_ops = &nfs41_state_renewal_ops,
8446         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8447 };
8448 #endif
8449
8450 #if defined(CONFIG_NFS_V4_2)
8451 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8452         .minor_version = 2,
8453         .init_caps = NFS_CAP_READDIRPLUS
8454                 | NFS_CAP_ATOMIC_OPEN
8455                 | NFS_CAP_CHANGE_ATTR
8456                 | NFS_CAP_POSIX_LOCK
8457                 | NFS_CAP_STATEID_NFSV41
8458                 | NFS_CAP_ATOMIC_OPEN_V1,
8459         .init_client = nfs41_init_client,
8460         .shutdown_client = nfs41_shutdown_client,
8461         .match_stateid = nfs41_match_stateid,
8462         .find_root_sec = nfs41_find_root_sec,
8463         .free_lock_state = nfs41_free_lock_state,
8464         .call_sync_ops = &nfs41_call_sync_ops,
8465         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8466         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8467         .state_renewal_ops = &nfs41_state_renewal_ops,
8468 };
8469 #endif
8470
8471 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8472         [0] = &nfs_v4_0_minor_ops,
8473 #if defined(CONFIG_NFS_V4_1)
8474         [1] = &nfs_v4_1_minor_ops,
8475 #endif
8476 #if defined(CONFIG_NFS_V4_2)
8477         [2] = &nfs_v4_2_minor_ops,
8478 #endif
8479 };
8480
8481 static const struct inode_operations nfs4_dir_inode_operations = {
8482         .create         = nfs_create,
8483         .lookup         = nfs_lookup,
8484         .atomic_open    = nfs_atomic_open,
8485         .link           = nfs_link,
8486         .unlink         = nfs_unlink,
8487         .symlink        = nfs_symlink,
8488         .mkdir          = nfs_mkdir,
8489         .rmdir          = nfs_rmdir,
8490         .mknod          = nfs_mknod,
8491         .rename         = nfs_rename,
8492         .permission     = nfs_permission,
8493         .getattr        = nfs_getattr,
8494         .setattr        = nfs_setattr,
8495         .getxattr       = generic_getxattr,
8496         .setxattr       = generic_setxattr,
8497         .listxattr      = generic_listxattr,
8498         .removexattr    = generic_removexattr,
8499 };
8500
8501 static const struct inode_operations nfs4_file_inode_operations = {
8502         .permission     = nfs_permission,
8503         .getattr        = nfs_getattr,
8504         .setattr        = nfs_setattr,
8505         .getxattr       = generic_getxattr,
8506         .setxattr       = generic_setxattr,
8507         .listxattr      = generic_listxattr,
8508         .removexattr    = generic_removexattr,
8509 };
8510
8511 const struct nfs_rpc_ops nfs_v4_clientops = {
8512         .version        = 4,                    /* protocol version */
8513         .dentry_ops     = &nfs4_dentry_operations,
8514         .dir_inode_ops  = &nfs4_dir_inode_operations,
8515         .file_inode_ops = &nfs4_file_inode_operations,
8516         .file_ops       = &nfs4_file_operations,
8517         .getroot        = nfs4_proc_get_root,
8518         .submount       = nfs4_submount,
8519         .try_mount      = nfs4_try_mount,
8520         .getattr        = nfs4_proc_getattr,
8521         .setattr        = nfs4_proc_setattr,
8522         .lookup         = nfs4_proc_lookup,
8523         .access         = nfs4_proc_access,
8524         .readlink       = nfs4_proc_readlink,
8525         .create         = nfs4_proc_create,
8526         .remove         = nfs4_proc_remove,
8527         .unlink_setup   = nfs4_proc_unlink_setup,
8528         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8529         .unlink_done    = nfs4_proc_unlink_done,
8530         .rename_setup   = nfs4_proc_rename_setup,
8531         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8532         .rename_done    = nfs4_proc_rename_done,
8533         .link           = nfs4_proc_link,
8534         .symlink        = nfs4_proc_symlink,
8535         .mkdir          = nfs4_proc_mkdir,
8536         .rmdir          = nfs4_proc_remove,
8537         .readdir        = nfs4_proc_readdir,
8538         .mknod          = nfs4_proc_mknod,
8539         .statfs         = nfs4_proc_statfs,
8540         .fsinfo         = nfs4_proc_fsinfo,
8541         .pathconf       = nfs4_proc_pathconf,
8542         .set_capabilities = nfs4_server_capabilities,
8543         .decode_dirent  = nfs4_decode_dirent,
8544         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8545         .read_setup     = nfs4_proc_read_setup,
8546         .read_done      = nfs4_read_done,
8547         .write_setup    = nfs4_proc_write_setup,
8548         .write_done     = nfs4_write_done,
8549         .commit_setup   = nfs4_proc_commit_setup,
8550         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8551         .commit_done    = nfs4_commit_done,
8552         .lock           = nfs4_proc_lock,
8553         .clear_acl_cache = nfs4_zap_acl_attr,
8554         .close_context  = nfs4_close_context,
8555         .open_context   = nfs4_atomic_open,
8556         .have_delegation = nfs4_have_delegation,
8557         .return_delegation = nfs4_inode_return_delegation,
8558         .alloc_client   = nfs4_alloc_client,
8559         .init_client    = nfs4_init_client,
8560         .free_client    = nfs4_free_client,
8561         .create_server  = nfs4_create_server,
8562         .clone_server   = nfs_clone_server,
8563 };
8564
8565 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8566         .prefix = XATTR_NAME_NFSV4_ACL,
8567         .list   = nfs4_xattr_list_nfs4_acl,
8568         .get    = nfs4_xattr_get_nfs4_acl,
8569         .set    = nfs4_xattr_set_nfs4_acl,
8570 };
8571
8572 const struct xattr_handler *nfs4_xattr_handlers[] = {
8573         &nfs4_xattr_nfs4_acl_handler,
8574 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8575         &nfs4_xattr_nfs4_label_handler,
8576 #endif
8577         NULL
8578 };
8579
8580 /*
8581  * Local variables:
8582  *  c-basic-offset: 8
8583  * End:
8584  */