]> git.kernelconcepts.de Git - karo-tx-linux.git/blob - fs/nfs/nfs4proc.c
NFSv4: handle nfs4_get_referral failure
[karo-tx-linux.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/file.h>
42 #include <linux/string.h>
43 #include <linux/ratelimit.h>
44 #include <linux/printk.h>
45 #include <linux/slab.h>
46 #include <linux/sunrpc/clnt.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69
70 #include "nfs4trace.h"
71
72 #define NFSDBG_FACILITY         NFSDBG_PROC
73
74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
76
77 struct nfs4_opendata;
78 static int _nfs4_proc_open(struct nfs4_opendata *data);
79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
83 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
84 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
85 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
86                             struct nfs_fattr *fattr, struct iattr *sattr,
87                             struct nfs4_state *state, struct nfs4_label *ilabel,
88                             struct nfs4_label *olabel);
89 #ifdef CONFIG_NFS_V4_1
90 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
91                 struct rpc_cred *);
92 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
93                 struct rpc_cred *);
94 #endif
95
96 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
97 static inline struct nfs4_label *
98 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
99         struct iattr *sattr, struct nfs4_label *label)
100 {
101         int err;
102
103         if (label == NULL)
104                 return NULL;
105
106         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
107                 return NULL;
108
109         err = security_dentry_init_security(dentry, sattr->ia_mode,
110                                 &dentry->d_name, (void **)&label->label, &label->len);
111         if (err == 0)
112                 return label;
113
114         return NULL;
115 }
116 static inline void
117 nfs4_label_release_security(struct nfs4_label *label)
118 {
119         if (label)
120                 security_release_secctx(label->label, label->len);
121 }
122 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
123 {
124         if (label)
125                 return server->attr_bitmask;
126
127         return server->attr_bitmask_nl;
128 }
129 #else
130 static inline struct nfs4_label *
131 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
132         struct iattr *sattr, struct nfs4_label *l)
133 { return NULL; }
134 static inline void
135 nfs4_label_release_security(struct nfs4_label *label)
136 { return; }
137 static inline u32 *
138 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
139 { return server->attr_bitmask; }
140 #endif
141
142 /* Prevent leaks of NFSv4 errors into userland */
143 static int nfs4_map_errors(int err)
144 {
145         if (err >= -1000)
146                 return err;
147         switch (err) {
148         case -NFS4ERR_RESOURCE:
149         case -NFS4ERR_LAYOUTTRYLATER:
150         case -NFS4ERR_RECALLCONFLICT:
151                 return -EREMOTEIO;
152         case -NFS4ERR_WRONGSEC:
153         case -NFS4ERR_WRONG_CRED:
154                 return -EPERM;
155         case -NFS4ERR_BADOWNER:
156         case -NFS4ERR_BADNAME:
157                 return -EINVAL;
158         case -NFS4ERR_SHARE_DENIED:
159                 return -EACCES;
160         case -NFS4ERR_MINOR_VERS_MISMATCH:
161                 return -EPROTONOSUPPORT;
162         case -NFS4ERR_FILE_OPEN:
163                 return -EBUSY;
164         default:
165                 dprintk("%s could not handle NFSv4 error %d\n",
166                                 __func__, -err);
167                 break;
168         }
169         return -EIO;
170 }
171
172 /*
173  * This is our standard bitmap for GETATTR requests.
174  */
175 const u32 nfs4_fattr_bitmap[3] = {
176         FATTR4_WORD0_TYPE
177         | FATTR4_WORD0_CHANGE
178         | FATTR4_WORD0_SIZE
179         | FATTR4_WORD0_FSID
180         | FATTR4_WORD0_FILEID,
181         FATTR4_WORD1_MODE
182         | FATTR4_WORD1_NUMLINKS
183         | FATTR4_WORD1_OWNER
184         | FATTR4_WORD1_OWNER_GROUP
185         | FATTR4_WORD1_RAWDEV
186         | FATTR4_WORD1_SPACE_USED
187         | FATTR4_WORD1_TIME_ACCESS
188         | FATTR4_WORD1_TIME_METADATA
189         | FATTR4_WORD1_TIME_MODIFY
190         | FATTR4_WORD1_MOUNTED_ON_FILEID,
191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
192         FATTR4_WORD2_SECURITY_LABEL
193 #endif
194 };
195
196 static const u32 nfs4_pnfs_open_bitmap[3] = {
197         FATTR4_WORD0_TYPE
198         | FATTR4_WORD0_CHANGE
199         | FATTR4_WORD0_SIZE
200         | FATTR4_WORD0_FSID
201         | FATTR4_WORD0_FILEID,
202         FATTR4_WORD1_MODE
203         | FATTR4_WORD1_NUMLINKS
204         | FATTR4_WORD1_OWNER
205         | FATTR4_WORD1_OWNER_GROUP
206         | FATTR4_WORD1_RAWDEV
207         | FATTR4_WORD1_SPACE_USED
208         | FATTR4_WORD1_TIME_ACCESS
209         | FATTR4_WORD1_TIME_METADATA
210         | FATTR4_WORD1_TIME_MODIFY,
211         FATTR4_WORD2_MDSTHRESHOLD
212 };
213
214 static const u32 nfs4_open_noattr_bitmap[3] = {
215         FATTR4_WORD0_TYPE
216         | FATTR4_WORD0_CHANGE
217         | FATTR4_WORD0_FILEID,
218 };
219
220 const u32 nfs4_statfs_bitmap[3] = {
221         FATTR4_WORD0_FILES_AVAIL
222         | FATTR4_WORD0_FILES_FREE
223         | FATTR4_WORD0_FILES_TOTAL,
224         FATTR4_WORD1_SPACE_AVAIL
225         | FATTR4_WORD1_SPACE_FREE
226         | FATTR4_WORD1_SPACE_TOTAL
227 };
228
229 const u32 nfs4_pathconf_bitmap[3] = {
230         FATTR4_WORD0_MAXLINK
231         | FATTR4_WORD0_MAXNAME,
232         0
233 };
234
235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
236                         | FATTR4_WORD0_MAXREAD
237                         | FATTR4_WORD0_MAXWRITE
238                         | FATTR4_WORD0_LEASE_TIME,
239                         FATTR4_WORD1_TIME_DELTA
240                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
241                         FATTR4_WORD2_LAYOUT_BLKSIZE
242 };
243
244 const u32 nfs4_fs_locations_bitmap[3] = {
245         FATTR4_WORD0_TYPE
246         | FATTR4_WORD0_CHANGE
247         | FATTR4_WORD0_SIZE
248         | FATTR4_WORD0_FSID
249         | FATTR4_WORD0_FILEID
250         | FATTR4_WORD0_FS_LOCATIONS,
251         FATTR4_WORD1_MODE
252         | FATTR4_WORD1_NUMLINKS
253         | FATTR4_WORD1_OWNER
254         | FATTR4_WORD1_OWNER_GROUP
255         | FATTR4_WORD1_RAWDEV
256         | FATTR4_WORD1_SPACE_USED
257         | FATTR4_WORD1_TIME_ACCESS
258         | FATTR4_WORD1_TIME_METADATA
259         | FATTR4_WORD1_TIME_MODIFY
260         | FATTR4_WORD1_MOUNTED_ON_FILEID,
261 };
262
263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
264                 struct nfs4_readdir_arg *readdir)
265 {
266         __be32 *start, *p;
267
268         if (cookie > 2) {
269                 readdir->cookie = cookie;
270                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
271                 return;
272         }
273
274         readdir->cookie = 0;
275         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
276         if (cookie == 2)
277                 return;
278         
279         /*
280          * NFSv4 servers do not return entries for '.' and '..'
281          * Therefore, we fake these entries here.  We let '.'
282          * have cookie 0 and '..' have cookie 1.  Note that
283          * when talking to the server, we always send cookie 0
284          * instead of 1 or 2.
285          */
286         start = p = kmap_atomic(*readdir->pages);
287         
288         if (cookie == 0) {
289                 *p++ = xdr_one;                                  /* next */
290                 *p++ = xdr_zero;                   /* cookie, first word */
291                 *p++ = xdr_one;                   /* cookie, second word */
292                 *p++ = xdr_one;                             /* entry len */
293                 memcpy(p, ".\0\0\0", 4);                        /* entry */
294                 p++;
295                 *p++ = xdr_one;                         /* bitmap length */
296                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
297                 *p++ = htonl(8);              /* attribute buffer length */
298                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
299         }
300         
301         *p++ = xdr_one;                                  /* next */
302         *p++ = xdr_zero;                   /* cookie, first word */
303         *p++ = xdr_two;                   /* cookie, second word */
304         *p++ = xdr_two;                             /* entry len */
305         memcpy(p, "..\0\0", 4);                         /* entry */
306         p++;
307         *p++ = xdr_one;                         /* bitmap length */
308         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
309         *p++ = htonl(8);              /* attribute buffer length */
310         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
311
312         readdir->pgbase = (char *)p - (char *)start;
313         readdir->count -= readdir->pgbase;
314         kunmap_atomic(start);
315 }
316
317 static long nfs4_update_delay(long *timeout)
318 {
319         long ret;
320         if (!timeout)
321                 return NFS4_POLL_RETRY_MAX;
322         if (*timeout <= 0)
323                 *timeout = NFS4_POLL_RETRY_MIN;
324         if (*timeout > NFS4_POLL_RETRY_MAX)
325                 *timeout = NFS4_POLL_RETRY_MAX;
326         ret = *timeout;
327         *timeout <<= 1;
328         return ret;
329 }
330
331 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
332 {
333         int res = 0;
334
335         might_sleep();
336
337         freezable_schedule_timeout_killable_unsafe(
338                 nfs4_update_delay(timeout));
339         if (fatal_signal_pending(current))
340                 res = -ERESTARTSYS;
341         return res;
342 }
343
344 /* This is the error handling routine for processes that are allowed
345  * to sleep.
346  */
347 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
348 {
349         struct nfs_client *clp = server->nfs_client;
350         struct nfs4_state *state = exception->state;
351         struct inode *inode = exception->inode;
352         int ret = errorcode;
353
354         exception->retry = 0;
355         switch(errorcode) {
356                 case 0:
357                         return 0;
358                 case -NFS4ERR_OPENMODE:
359                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
360                                 nfs4_inode_return_delegation(inode);
361                                 exception->retry = 1;
362                                 return 0;
363                         }
364                         if (state == NULL)
365                                 break;
366                         ret = nfs4_schedule_stateid_recovery(server, state);
367                         if (ret < 0)
368                                 break;
369                         goto wait_on_recovery;
370                 case -NFS4ERR_DELEG_REVOKED:
371                 case -NFS4ERR_ADMIN_REVOKED:
372                 case -NFS4ERR_BAD_STATEID:
373                         if (state == NULL)
374                                 break;
375                         ret = nfs4_schedule_stateid_recovery(server, state);
376                         if (ret < 0)
377                                 break;
378                         goto wait_on_recovery;
379                 case -NFS4ERR_EXPIRED:
380                         if (state != NULL) {
381                                 ret = nfs4_schedule_stateid_recovery(server, state);
382                                 if (ret < 0)
383                                         break;
384                         }
385                 case -NFS4ERR_STALE_STATEID:
386                 case -NFS4ERR_STALE_CLIENTID:
387                         nfs4_schedule_lease_recovery(clp);
388                         goto wait_on_recovery;
389                 case -NFS4ERR_MOVED:
390                         ret = nfs4_schedule_migration_recovery(server);
391                         if (ret < 0)
392                                 break;
393                         goto wait_on_recovery;
394                 case -NFS4ERR_LEASE_MOVED:
395                         nfs4_schedule_lease_moved_recovery(clp);
396                         goto wait_on_recovery;
397 #if defined(CONFIG_NFS_V4_1)
398                 case -NFS4ERR_BADSESSION:
399                 case -NFS4ERR_BADSLOT:
400                 case -NFS4ERR_BAD_HIGH_SLOT:
401                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
402                 case -NFS4ERR_DEADSESSION:
403                 case -NFS4ERR_SEQ_FALSE_RETRY:
404                 case -NFS4ERR_SEQ_MISORDERED:
405                         dprintk("%s ERROR: %d Reset session\n", __func__,
406                                 errorcode);
407                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
408                         goto wait_on_recovery;
409 #endif /* defined(CONFIG_NFS_V4_1) */
410                 case -NFS4ERR_FILE_OPEN:
411                         if (exception->timeout > HZ) {
412                                 /* We have retried a decent amount, time to
413                                  * fail
414                                  */
415                                 ret = -EBUSY;
416                                 break;
417                         }
418                 case -NFS4ERR_GRACE:
419                 case -NFS4ERR_DELAY:
420                         ret = nfs4_delay(server->client, &exception->timeout);
421                         if (ret != 0)
422                                 break;
423                 case -NFS4ERR_RETRY_UNCACHED_REP:
424                 case -NFS4ERR_OLD_STATEID:
425                         exception->retry = 1;
426                         break;
427                 case -NFS4ERR_BADOWNER:
428                         /* The following works around a Linux server bug! */
429                 case -NFS4ERR_BADNAME:
430                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
431                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
432                                 exception->retry = 1;
433                                 printk(KERN_WARNING "NFS: v4 server %s "
434                                                 "does not accept raw "
435                                                 "uid/gids. "
436                                                 "Reenabling the idmapper.\n",
437                                                 server->nfs_client->cl_hostname);
438                         }
439         }
440         /* We failed to handle the error */
441         return nfs4_map_errors(ret);
442 wait_on_recovery:
443         ret = nfs4_wait_clnt_recover(clp);
444         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
445                 return -EIO;
446         if (ret == 0)
447                 exception->retry = 1;
448         return ret;
449 }
450
451 /*
452  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
453  * or 'false' otherwise.
454  */
455 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
456 {
457         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
458
459         if (flavor == RPC_AUTH_GSS_KRB5I ||
460             flavor == RPC_AUTH_GSS_KRB5P)
461                 return true;
462
463         return false;
464 }
465
466 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
467 {
468         spin_lock(&clp->cl_lock);
469         if (time_before(clp->cl_last_renewal,timestamp))
470                 clp->cl_last_renewal = timestamp;
471         spin_unlock(&clp->cl_lock);
472 }
473
474 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
475 {
476         do_renew_lease(server->nfs_client, timestamp);
477 }
478
479 struct nfs4_call_sync_data {
480         const struct nfs_server *seq_server;
481         struct nfs4_sequence_args *seq_args;
482         struct nfs4_sequence_res *seq_res;
483 };
484
485 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
486                                struct nfs4_sequence_res *res, int cache_reply)
487 {
488         args->sa_slot = NULL;
489         args->sa_cache_this = cache_reply;
490         args->sa_privileged = 0;
491
492         res->sr_slot = NULL;
493 }
494
495 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
496 {
497         args->sa_privileged = 1;
498 }
499
500 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
501                          struct nfs4_sequence_args *args,
502                          struct nfs4_sequence_res *res,
503                          struct rpc_task *task)
504 {
505         struct nfs4_slot *slot;
506
507         /* slot already allocated? */
508         if (res->sr_slot != NULL)
509                 goto out_start;
510
511         spin_lock(&tbl->slot_tbl_lock);
512         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
513                 goto out_sleep;
514
515         slot = nfs4_alloc_slot(tbl);
516         if (IS_ERR(slot)) {
517                 if (slot == ERR_PTR(-ENOMEM))
518                         task->tk_timeout = HZ >> 2;
519                 goto out_sleep;
520         }
521         spin_unlock(&tbl->slot_tbl_lock);
522
523         args->sa_slot = slot;
524         res->sr_slot = slot;
525
526 out_start:
527         rpc_call_start(task);
528         return 0;
529
530 out_sleep:
531         if (args->sa_privileged)
532                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
533                                 NULL, RPC_PRIORITY_PRIVILEGED);
534         else
535                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
536         spin_unlock(&tbl->slot_tbl_lock);
537         return -EAGAIN;
538 }
539 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
540
541 static int nfs40_sequence_done(struct rpc_task *task,
542                                struct nfs4_sequence_res *res)
543 {
544         struct nfs4_slot *slot = res->sr_slot;
545         struct nfs4_slot_table *tbl;
546
547         if (slot == NULL)
548                 goto out;
549
550         tbl = slot->table;
551         spin_lock(&tbl->slot_tbl_lock);
552         if (!nfs41_wake_and_assign_slot(tbl, slot))
553                 nfs4_free_slot(tbl, slot);
554         spin_unlock(&tbl->slot_tbl_lock);
555
556         res->sr_slot = NULL;
557 out:
558         return 1;
559 }
560
561 #if defined(CONFIG_NFS_V4_1)
562
563 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
564 {
565         struct nfs4_session *session;
566         struct nfs4_slot_table *tbl;
567         struct nfs4_slot *slot = res->sr_slot;
568         bool send_new_highest_used_slotid = false;
569
570         tbl = slot->table;
571         session = tbl->session;
572
573         spin_lock(&tbl->slot_tbl_lock);
574         /* Be nice to the server: try to ensure that the last transmitted
575          * value for highest_user_slotid <= target_highest_slotid
576          */
577         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
578                 send_new_highest_used_slotid = true;
579
580         if (nfs41_wake_and_assign_slot(tbl, slot)) {
581                 send_new_highest_used_slotid = false;
582                 goto out_unlock;
583         }
584         nfs4_free_slot(tbl, slot);
585
586         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
587                 send_new_highest_used_slotid = false;
588 out_unlock:
589         spin_unlock(&tbl->slot_tbl_lock);
590         res->sr_slot = NULL;
591         if (send_new_highest_used_slotid)
592                 nfs41_server_notify_highest_slotid_update(session->clp);
593 }
594
595 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
596 {
597         struct nfs4_session *session;
598         struct nfs4_slot *slot = res->sr_slot;
599         struct nfs_client *clp;
600         bool interrupted = false;
601         int ret = 1;
602
603         if (slot == NULL)
604                 goto out_noaction;
605         /* don't increment the sequence number if the task wasn't sent */
606         if (!RPC_WAS_SENT(task))
607                 goto out;
608
609         session = slot->table->session;
610
611         if (slot->interrupted) {
612                 slot->interrupted = 0;
613                 interrupted = true;
614         }
615
616         trace_nfs4_sequence_done(session, res);
617         /* Check the SEQUENCE operation status */
618         switch (res->sr_status) {
619         case 0:
620                 /* Update the slot's sequence and clientid lease timer */
621                 ++slot->seq_nr;
622                 clp = session->clp;
623                 do_renew_lease(clp, res->sr_timestamp);
624                 /* Check sequence flags */
625                 if (res->sr_status_flags != 0)
626                         nfs4_schedule_lease_recovery(clp);
627                 nfs41_update_target_slotid(slot->table, slot, res);
628                 break;
629         case 1:
630                 /*
631                  * sr_status remains 1 if an RPC level error occurred.
632                  * The server may or may not have processed the sequence
633                  * operation..
634                  * Mark the slot as having hosted an interrupted RPC call.
635                  */
636                 slot->interrupted = 1;
637                 goto out;
638         case -NFS4ERR_DELAY:
639                 /* The server detected a resend of the RPC call and
640                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
641                  * of RFC5661.
642                  */
643                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
644                         __func__,
645                         slot->slot_nr,
646                         slot->seq_nr);
647                 goto out_retry;
648         case -NFS4ERR_BADSLOT:
649                 /*
650                  * The slot id we used was probably retired. Try again
651                  * using a different slot id.
652                  */
653                 goto retry_nowait;
654         case -NFS4ERR_SEQ_MISORDERED:
655                 /*
656                  * Was the last operation on this sequence interrupted?
657                  * If so, retry after bumping the sequence number.
658                  */
659                 if (interrupted) {
660                         ++slot->seq_nr;
661                         goto retry_nowait;
662                 }
663                 /*
664                  * Could this slot have been previously retired?
665                  * If so, then the server may be expecting seq_nr = 1!
666                  */
667                 if (slot->seq_nr != 1) {
668                         slot->seq_nr = 1;
669                         goto retry_nowait;
670                 }
671                 break;
672         case -NFS4ERR_SEQ_FALSE_RETRY:
673                 ++slot->seq_nr;
674                 goto retry_nowait;
675         default:
676                 /* Just update the slot sequence no. */
677                 ++slot->seq_nr;
678         }
679 out:
680         /* The session may be reset by one of the error handlers. */
681         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
682         nfs41_sequence_free_slot(res);
683 out_noaction:
684         return ret;
685 retry_nowait:
686         if (rpc_restart_call_prepare(task)) {
687                 task->tk_status = 0;
688                 ret = 0;
689         }
690         goto out;
691 out_retry:
692         if (!rpc_restart_call(task))
693                 goto out;
694         rpc_delay(task, NFS4_POLL_RETRY_MAX);
695         return 0;
696 }
697 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
698
699 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
700 {
701         if (res->sr_slot == NULL)
702                 return 1;
703         if (!res->sr_slot->table->session)
704                 return nfs40_sequence_done(task, res);
705         return nfs41_sequence_done(task, res);
706 }
707 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
708
709 int nfs41_setup_sequence(struct nfs4_session *session,
710                                 struct nfs4_sequence_args *args,
711                                 struct nfs4_sequence_res *res,
712                                 struct rpc_task *task)
713 {
714         struct nfs4_slot *slot;
715         struct nfs4_slot_table *tbl;
716
717         dprintk("--> %s\n", __func__);
718         /* slot already allocated? */
719         if (res->sr_slot != NULL)
720                 goto out_success;
721
722         tbl = &session->fc_slot_table;
723
724         task->tk_timeout = 0;
725
726         spin_lock(&tbl->slot_tbl_lock);
727         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
728             !args->sa_privileged) {
729                 /* The state manager will wait until the slot table is empty */
730                 dprintk("%s session is draining\n", __func__);
731                 goto out_sleep;
732         }
733
734         slot = nfs4_alloc_slot(tbl);
735         if (IS_ERR(slot)) {
736                 /* If out of memory, try again in 1/4 second */
737                 if (slot == ERR_PTR(-ENOMEM))
738                         task->tk_timeout = HZ >> 2;
739                 dprintk("<-- %s: no free slots\n", __func__);
740                 goto out_sleep;
741         }
742         spin_unlock(&tbl->slot_tbl_lock);
743
744         args->sa_slot = slot;
745
746         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
747                         slot->slot_nr, slot->seq_nr);
748
749         res->sr_slot = slot;
750         res->sr_timestamp = jiffies;
751         res->sr_status_flags = 0;
752         /*
753          * sr_status is only set in decode_sequence, and so will remain
754          * set to 1 if an rpc level failure occurs.
755          */
756         res->sr_status = 1;
757         trace_nfs4_setup_sequence(session, args);
758 out_success:
759         rpc_call_start(task);
760         return 0;
761 out_sleep:
762         /* Privileged tasks are queued with top priority */
763         if (args->sa_privileged)
764                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
765                                 NULL, RPC_PRIORITY_PRIVILEGED);
766         else
767                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
768         spin_unlock(&tbl->slot_tbl_lock);
769         return -EAGAIN;
770 }
771 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
772
773 static int nfs4_setup_sequence(const struct nfs_server *server,
774                                struct nfs4_sequence_args *args,
775                                struct nfs4_sequence_res *res,
776                                struct rpc_task *task)
777 {
778         struct nfs4_session *session = nfs4_get_session(server);
779         int ret = 0;
780
781         if (!session)
782                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
783                                             args, res, task);
784
785         dprintk("--> %s clp %p session %p sr_slot %u\n",
786                 __func__, session->clp, session, res->sr_slot ?
787                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
788
789         ret = nfs41_setup_sequence(session, args, res, task);
790
791         dprintk("<-- %s status=%d\n", __func__, ret);
792         return ret;
793 }
794
795 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
796 {
797         struct nfs4_call_sync_data *data = calldata;
798         struct nfs4_session *session = nfs4_get_session(data->seq_server);
799
800         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
801
802         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
803 }
804
805 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
806 {
807         struct nfs4_call_sync_data *data = calldata;
808
809         nfs41_sequence_done(task, data->seq_res);
810 }
811
812 static const struct rpc_call_ops nfs41_call_sync_ops = {
813         .rpc_call_prepare = nfs41_call_sync_prepare,
814         .rpc_call_done = nfs41_call_sync_done,
815 };
816
817 #else   /* !CONFIG_NFS_V4_1 */
818
819 static int nfs4_setup_sequence(const struct nfs_server *server,
820                                struct nfs4_sequence_args *args,
821                                struct nfs4_sequence_res *res,
822                                struct rpc_task *task)
823 {
824         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
825                                     args, res, task);
826 }
827
828 int nfs4_sequence_done(struct rpc_task *task,
829                        struct nfs4_sequence_res *res)
830 {
831         return nfs40_sequence_done(task, res);
832 }
833 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
834
835 #endif  /* !CONFIG_NFS_V4_1 */
836
837 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
838 {
839         struct nfs4_call_sync_data *data = calldata;
840         nfs4_setup_sequence(data->seq_server,
841                                 data->seq_args, data->seq_res, task);
842 }
843
844 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
845 {
846         struct nfs4_call_sync_data *data = calldata;
847         nfs4_sequence_done(task, data->seq_res);
848 }
849
850 static const struct rpc_call_ops nfs40_call_sync_ops = {
851         .rpc_call_prepare = nfs40_call_sync_prepare,
852         .rpc_call_done = nfs40_call_sync_done,
853 };
854
855 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
856                                    struct nfs_server *server,
857                                    struct rpc_message *msg,
858                                    struct nfs4_sequence_args *args,
859                                    struct nfs4_sequence_res *res)
860 {
861         int ret;
862         struct rpc_task *task;
863         struct nfs_client *clp = server->nfs_client;
864         struct nfs4_call_sync_data data = {
865                 .seq_server = server,
866                 .seq_args = args,
867                 .seq_res = res,
868         };
869         struct rpc_task_setup task_setup = {
870                 .rpc_client = clnt,
871                 .rpc_message = msg,
872                 .callback_ops = clp->cl_mvops->call_sync_ops,
873                 .callback_data = &data
874         };
875
876         task = rpc_run_task(&task_setup);
877         if (IS_ERR(task))
878                 ret = PTR_ERR(task);
879         else {
880                 ret = task->tk_status;
881                 rpc_put_task(task);
882         }
883         return ret;
884 }
885
886 int nfs4_call_sync(struct rpc_clnt *clnt,
887                    struct nfs_server *server,
888                    struct rpc_message *msg,
889                    struct nfs4_sequence_args *args,
890                    struct nfs4_sequence_res *res,
891                    int cache_reply)
892 {
893         nfs4_init_sequence(args, res, cache_reply);
894         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
895 }
896
897 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
898 {
899         struct nfs_inode *nfsi = NFS_I(dir);
900
901         spin_lock(&dir->i_lock);
902         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
903         if (!cinfo->atomic || cinfo->before != dir->i_version)
904                 nfs_force_lookup_revalidate(dir);
905         dir->i_version = cinfo->after;
906         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
907         nfs_fscache_invalidate(dir);
908         spin_unlock(&dir->i_lock);
909 }
910
911 struct nfs4_opendata {
912         struct kref kref;
913         struct nfs_openargs o_arg;
914         struct nfs_openres o_res;
915         struct nfs_open_confirmargs c_arg;
916         struct nfs_open_confirmres c_res;
917         struct nfs4_string owner_name;
918         struct nfs4_string group_name;
919         struct nfs_fattr f_attr;
920         struct nfs4_label *f_label;
921         struct dentry *dir;
922         struct dentry *dentry;
923         struct nfs4_state_owner *owner;
924         struct nfs4_state *state;
925         struct iattr attrs;
926         unsigned long timestamp;
927         unsigned int rpc_done : 1;
928         unsigned int file_created : 1;
929         unsigned int is_recover : 1;
930         int rpc_status;
931         int cancelled;
932 };
933
934 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
935                 int err, struct nfs4_exception *exception)
936 {
937         if (err != -EINVAL)
938                 return false;
939         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
940                 return false;
941         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
942         exception->retry = 1;
943         return true;
944 }
945
946 static u32
947 nfs4_map_atomic_open_share(struct nfs_server *server,
948                 fmode_t fmode, int openflags)
949 {
950         u32 res = 0;
951
952         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
953         case FMODE_READ:
954                 res = NFS4_SHARE_ACCESS_READ;
955                 break;
956         case FMODE_WRITE:
957                 res = NFS4_SHARE_ACCESS_WRITE;
958                 break;
959         case FMODE_READ|FMODE_WRITE:
960                 res = NFS4_SHARE_ACCESS_BOTH;
961         }
962         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
963                 goto out;
964         /* Want no delegation if we're using O_DIRECT */
965         if (openflags & O_DIRECT)
966                 res |= NFS4_SHARE_WANT_NO_DELEG;
967 out:
968         return res;
969 }
970
971 static enum open_claim_type4
972 nfs4_map_atomic_open_claim(struct nfs_server *server,
973                 enum open_claim_type4 claim)
974 {
975         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
976                 return claim;
977         switch (claim) {
978         default:
979                 return claim;
980         case NFS4_OPEN_CLAIM_FH:
981                 return NFS4_OPEN_CLAIM_NULL;
982         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
983                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
984         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
985                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
986         }
987 }
988
989 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
990 {
991         p->o_res.f_attr = &p->f_attr;
992         p->o_res.f_label = p->f_label;
993         p->o_res.seqid = p->o_arg.seqid;
994         p->c_res.seqid = p->c_arg.seqid;
995         p->o_res.server = p->o_arg.server;
996         p->o_res.access_request = p->o_arg.access;
997         nfs_fattr_init(&p->f_attr);
998         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
999 }
1000
1001 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1002                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1003                 const struct iattr *attrs,
1004                 struct nfs4_label *label,
1005                 enum open_claim_type4 claim,
1006                 gfp_t gfp_mask)
1007 {
1008         struct dentry *parent = dget_parent(dentry);
1009         struct inode *dir = d_inode(parent);
1010         struct nfs_server *server = NFS_SERVER(dir);
1011         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1012         struct nfs4_opendata *p;
1013
1014         p = kzalloc(sizeof(*p), gfp_mask);
1015         if (p == NULL)
1016                 goto err;
1017
1018         p->f_label = nfs4_label_alloc(server, gfp_mask);
1019         if (IS_ERR(p->f_label))
1020                 goto err_free_p;
1021
1022         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1023         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1024         if (IS_ERR(p->o_arg.seqid))
1025                 goto err_free_label;
1026         nfs_sb_active(dentry->d_sb);
1027         p->dentry = dget(dentry);
1028         p->dir = parent;
1029         p->owner = sp;
1030         atomic_inc(&sp->so_count);
1031         p->o_arg.open_flags = flags;
1032         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1033         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1034                         fmode, flags);
1035         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1036          * will return permission denied for all bits until close */
1037         if (!(flags & O_EXCL)) {
1038                 /* ask server to check for all possible rights as results
1039                  * are cached */
1040                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1041                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1042         }
1043         p->o_arg.clientid = server->nfs_client->cl_clientid;
1044         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1045         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1046         p->o_arg.name = &dentry->d_name;
1047         p->o_arg.server = server;
1048         p->o_arg.bitmask = nfs4_bitmask(server, label);
1049         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1050         p->o_arg.label = label;
1051         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1052         switch (p->o_arg.claim) {
1053         case NFS4_OPEN_CLAIM_NULL:
1054         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1055         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1056                 p->o_arg.fh = NFS_FH(dir);
1057                 break;
1058         case NFS4_OPEN_CLAIM_PREVIOUS:
1059         case NFS4_OPEN_CLAIM_FH:
1060         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1061         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1062                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1063         }
1064         if (attrs != NULL && attrs->ia_valid != 0) {
1065                 __u32 verf[2];
1066
1067                 p->o_arg.u.attrs = &p->attrs;
1068                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1069
1070                 verf[0] = jiffies;
1071                 verf[1] = current->pid;
1072                 memcpy(p->o_arg.u.verifier.data, verf,
1073                                 sizeof(p->o_arg.u.verifier.data));
1074         }
1075         p->c_arg.fh = &p->o_res.fh;
1076         p->c_arg.stateid = &p->o_res.stateid;
1077         p->c_arg.seqid = p->o_arg.seqid;
1078         nfs4_init_opendata_res(p);
1079         kref_init(&p->kref);
1080         return p;
1081
1082 err_free_label:
1083         nfs4_label_free(p->f_label);
1084 err_free_p:
1085         kfree(p);
1086 err:
1087         dput(parent);
1088         return NULL;
1089 }
1090
1091 static void nfs4_opendata_free(struct kref *kref)
1092 {
1093         struct nfs4_opendata *p = container_of(kref,
1094                         struct nfs4_opendata, kref);
1095         struct super_block *sb = p->dentry->d_sb;
1096
1097         nfs_free_seqid(p->o_arg.seqid);
1098         if (p->state != NULL)
1099                 nfs4_put_open_state(p->state);
1100         nfs4_put_state_owner(p->owner);
1101
1102         nfs4_label_free(p->f_label);
1103
1104         dput(p->dir);
1105         dput(p->dentry);
1106         nfs_sb_deactive(sb);
1107         nfs_fattr_free_names(&p->f_attr);
1108         kfree(p->f_attr.mdsthreshold);
1109         kfree(p);
1110 }
1111
1112 static void nfs4_opendata_put(struct nfs4_opendata *p)
1113 {
1114         if (p != NULL)
1115                 kref_put(&p->kref, nfs4_opendata_free);
1116 }
1117
1118 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1119 {
1120         int ret;
1121
1122         ret = rpc_wait_for_completion_task(task);
1123         return ret;
1124 }
1125
1126 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1127 {
1128         int ret = 0;
1129
1130         if (open_mode & (O_EXCL|O_TRUNC))
1131                 goto out;
1132         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1133                 case FMODE_READ:
1134                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1135                                 && state->n_rdonly != 0;
1136                         break;
1137                 case FMODE_WRITE:
1138                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1139                                 && state->n_wronly != 0;
1140                         break;
1141                 case FMODE_READ|FMODE_WRITE:
1142                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1143                                 && state->n_rdwr != 0;
1144         }
1145 out:
1146         return ret;
1147 }
1148
1149 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1150 {
1151         if (delegation == NULL)
1152                 return 0;
1153         if ((delegation->type & fmode) != fmode)
1154                 return 0;
1155         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1156                 return 0;
1157         nfs_mark_delegation_referenced(delegation);
1158         return 1;
1159 }
1160
1161 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1162 {
1163         switch (fmode) {
1164                 case FMODE_WRITE:
1165                         state->n_wronly++;
1166                         break;
1167                 case FMODE_READ:
1168                         state->n_rdonly++;
1169                         break;
1170                 case FMODE_READ|FMODE_WRITE:
1171                         state->n_rdwr++;
1172         }
1173         nfs4_state_set_mode_locked(state, state->state | fmode);
1174 }
1175
1176 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1177 {
1178         struct nfs_client *clp = state->owner->so_server->nfs_client;
1179         bool need_recover = false;
1180
1181         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1182                 need_recover = true;
1183         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1184                 need_recover = true;
1185         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1186                 need_recover = true;
1187         if (need_recover)
1188                 nfs4_state_mark_reclaim_nograce(clp, state);
1189 }
1190
1191 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1192                 nfs4_stateid *stateid)
1193 {
1194         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1195                 return true;
1196         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1197                 nfs_test_and_clear_all_open_stateid(state);
1198                 return true;
1199         }
1200         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1201                 return true;
1202         return false;
1203 }
1204
1205 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1206 {
1207         if (state->n_wronly)
1208                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1209         if (state->n_rdonly)
1210                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1211         if (state->n_rdwr)
1212                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1213 }
1214
1215 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1216                 nfs4_stateid *stateid, fmode_t fmode)
1217 {
1218         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1219         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1220         case FMODE_WRITE:
1221                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1222                 break;
1223         case FMODE_READ:
1224                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1225                 break;
1226         case 0:
1227                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1228                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1229                 clear_bit(NFS_OPEN_STATE, &state->flags);
1230         }
1231         if (stateid == NULL)
1232                 return;
1233         /* Handle races with OPEN */
1234         if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1235             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1236                 nfs_resync_open_stateid_locked(state);
1237                 return;
1238         }
1239         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1240                 nfs4_stateid_copy(&state->stateid, stateid);
1241         nfs4_stateid_copy(&state->open_stateid, stateid);
1242 }
1243
1244 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1245 {
1246         write_seqlock(&state->seqlock);
1247         nfs_clear_open_stateid_locked(state, stateid, fmode);
1248         write_sequnlock(&state->seqlock);
1249         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1250                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1251 }
1252
1253 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1254 {
1255         switch (fmode) {
1256                 case FMODE_READ:
1257                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1258                         break;
1259                 case FMODE_WRITE:
1260                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1261                         break;
1262                 case FMODE_READ|FMODE_WRITE:
1263                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1264         }
1265         if (!nfs_need_update_open_stateid(state, stateid))
1266                 return;
1267         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1268                 nfs4_stateid_copy(&state->stateid, stateid);
1269         nfs4_stateid_copy(&state->open_stateid, stateid);
1270 }
1271
1272 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1273 {
1274         /*
1275          * Protect the call to nfs4_state_set_mode_locked and
1276          * serialise the stateid update
1277          */
1278         write_seqlock(&state->seqlock);
1279         if (deleg_stateid != NULL) {
1280                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1281                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1282         }
1283         if (open_stateid != NULL)
1284                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1285         write_sequnlock(&state->seqlock);
1286         spin_lock(&state->owner->so_lock);
1287         update_open_stateflags(state, fmode);
1288         spin_unlock(&state->owner->so_lock);
1289 }
1290
1291 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1292 {
1293         struct nfs_inode *nfsi = NFS_I(state->inode);
1294         struct nfs_delegation *deleg_cur;
1295         int ret = 0;
1296
1297         fmode &= (FMODE_READ|FMODE_WRITE);
1298
1299         rcu_read_lock();
1300         deleg_cur = rcu_dereference(nfsi->delegation);
1301         if (deleg_cur == NULL)
1302                 goto no_delegation;
1303
1304         spin_lock(&deleg_cur->lock);
1305         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1306            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1307             (deleg_cur->type & fmode) != fmode)
1308                 goto no_delegation_unlock;
1309
1310         if (delegation == NULL)
1311                 delegation = &deleg_cur->stateid;
1312         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1313                 goto no_delegation_unlock;
1314
1315         nfs_mark_delegation_referenced(deleg_cur);
1316         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1317         ret = 1;
1318 no_delegation_unlock:
1319         spin_unlock(&deleg_cur->lock);
1320 no_delegation:
1321         rcu_read_unlock();
1322
1323         if (!ret && open_stateid != NULL) {
1324                 __update_open_stateid(state, open_stateid, NULL, fmode);
1325                 ret = 1;
1326         }
1327         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1328                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1329
1330         return ret;
1331 }
1332
1333 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1334                 const nfs4_stateid *stateid)
1335 {
1336         struct nfs4_state *state = lsp->ls_state;
1337         bool ret = false;
1338
1339         spin_lock(&state->state_lock);
1340         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1341                 goto out_noupdate;
1342         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1343                 goto out_noupdate;
1344         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1345         ret = true;
1346 out_noupdate:
1347         spin_unlock(&state->state_lock);
1348         return ret;
1349 }
1350
1351 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1352 {
1353         struct nfs_delegation *delegation;
1354
1355         rcu_read_lock();
1356         delegation = rcu_dereference(NFS_I(inode)->delegation);
1357         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1358                 rcu_read_unlock();
1359                 return;
1360         }
1361         rcu_read_unlock();
1362         nfs4_inode_return_delegation(inode);
1363 }
1364
1365 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1366 {
1367         struct nfs4_state *state = opendata->state;
1368         struct nfs_inode *nfsi = NFS_I(state->inode);
1369         struct nfs_delegation *delegation;
1370         int open_mode = opendata->o_arg.open_flags;
1371         fmode_t fmode = opendata->o_arg.fmode;
1372         nfs4_stateid stateid;
1373         int ret = -EAGAIN;
1374
1375         for (;;) {
1376                 spin_lock(&state->owner->so_lock);
1377                 if (can_open_cached(state, fmode, open_mode)) {
1378                         update_open_stateflags(state, fmode);
1379                         spin_unlock(&state->owner->so_lock);
1380                         goto out_return_state;
1381                 }
1382                 spin_unlock(&state->owner->so_lock);
1383                 rcu_read_lock();
1384                 delegation = rcu_dereference(nfsi->delegation);
1385                 if (!can_open_delegated(delegation, fmode)) {
1386                         rcu_read_unlock();
1387                         break;
1388                 }
1389                 /* Save the delegation */
1390                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1391                 rcu_read_unlock();
1392                 nfs_release_seqid(opendata->o_arg.seqid);
1393                 if (!opendata->is_recover) {
1394                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1395                         if (ret != 0)
1396                                 goto out;
1397                 }
1398                 ret = -EAGAIN;
1399
1400                 /* Try to update the stateid using the delegation */
1401                 if (update_open_stateid(state, NULL, &stateid, fmode))
1402                         goto out_return_state;
1403         }
1404 out:
1405         return ERR_PTR(ret);
1406 out_return_state:
1407         atomic_inc(&state->count);
1408         return state;
1409 }
1410
1411 static void
1412 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1413 {
1414         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1415         struct nfs_delegation *delegation;
1416         int delegation_flags = 0;
1417
1418         rcu_read_lock();
1419         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1420         if (delegation)
1421                 delegation_flags = delegation->flags;
1422         rcu_read_unlock();
1423         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1424                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1425                                    "returning a delegation for "
1426                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1427                                    clp->cl_hostname);
1428         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1429                 nfs_inode_set_delegation(state->inode,
1430                                          data->owner->so_cred,
1431                                          &data->o_res);
1432         else
1433                 nfs_inode_reclaim_delegation(state->inode,
1434                                              data->owner->so_cred,
1435                                              &data->o_res);
1436 }
1437
1438 /*
1439  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1440  * and update the nfs4_state.
1441  */
1442 static struct nfs4_state *
1443 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1444 {
1445         struct inode *inode = data->state->inode;
1446         struct nfs4_state *state = data->state;
1447         int ret;
1448
1449         if (!data->rpc_done) {
1450                 if (data->rpc_status) {
1451                         ret = data->rpc_status;
1452                         goto err;
1453                 }
1454                 /* cached opens have already been processed */
1455                 goto update;
1456         }
1457
1458         ret = nfs_refresh_inode(inode, &data->f_attr);
1459         if (ret)
1460                 goto err;
1461
1462         if (data->o_res.delegation_type != 0)
1463                 nfs4_opendata_check_deleg(data, state);
1464 update:
1465         update_open_stateid(state, &data->o_res.stateid, NULL,
1466                             data->o_arg.fmode);
1467         atomic_inc(&state->count);
1468
1469         return state;
1470 err:
1471         return ERR_PTR(ret);
1472
1473 }
1474
1475 static struct nfs4_state *
1476 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1477 {
1478         struct inode *inode;
1479         struct nfs4_state *state = NULL;
1480         int ret;
1481
1482         if (!data->rpc_done) {
1483                 state = nfs4_try_open_cached(data);
1484                 goto out;
1485         }
1486
1487         ret = -EAGAIN;
1488         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1489                 goto err;
1490         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1491         ret = PTR_ERR(inode);
1492         if (IS_ERR(inode))
1493                 goto err;
1494         ret = -ENOMEM;
1495         state = nfs4_get_open_state(inode, data->owner);
1496         if (state == NULL)
1497                 goto err_put_inode;
1498         if (data->o_res.delegation_type != 0)
1499                 nfs4_opendata_check_deleg(data, state);
1500         update_open_stateid(state, &data->o_res.stateid, NULL,
1501                         data->o_arg.fmode);
1502         iput(inode);
1503 out:
1504         nfs_release_seqid(data->o_arg.seqid);
1505         return state;
1506 err_put_inode:
1507         iput(inode);
1508 err:
1509         return ERR_PTR(ret);
1510 }
1511
1512 static struct nfs4_state *
1513 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1514 {
1515         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1516                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1517         return _nfs4_opendata_to_nfs4_state(data);
1518 }
1519
1520 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1521 {
1522         struct nfs_inode *nfsi = NFS_I(state->inode);
1523         struct nfs_open_context *ctx;
1524
1525         spin_lock(&state->inode->i_lock);
1526         list_for_each_entry(ctx, &nfsi->open_files, list) {
1527                 if (ctx->state != state)
1528                         continue;
1529                 get_nfs_open_context(ctx);
1530                 spin_unlock(&state->inode->i_lock);
1531                 return ctx;
1532         }
1533         spin_unlock(&state->inode->i_lock);
1534         return ERR_PTR(-ENOENT);
1535 }
1536
1537 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1538                 struct nfs4_state *state, enum open_claim_type4 claim)
1539 {
1540         struct nfs4_opendata *opendata;
1541
1542         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1543                         NULL, NULL, claim, GFP_NOFS);
1544         if (opendata == NULL)
1545                 return ERR_PTR(-ENOMEM);
1546         opendata->state = state;
1547         atomic_inc(&state->count);
1548         return opendata;
1549 }
1550
1551 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1552 {
1553         struct nfs4_state *newstate;
1554         int ret;
1555
1556         opendata->o_arg.open_flags = 0;
1557         opendata->o_arg.fmode = fmode;
1558         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1559                         NFS_SB(opendata->dentry->d_sb),
1560                         fmode, 0);
1561         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1562         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1563         nfs4_init_opendata_res(opendata);
1564         ret = _nfs4_recover_proc_open(opendata);
1565         if (ret != 0)
1566                 return ret; 
1567         newstate = nfs4_opendata_to_nfs4_state(opendata);
1568         if (IS_ERR(newstate))
1569                 return PTR_ERR(newstate);
1570         nfs4_close_state(newstate, fmode);
1571         *res = newstate;
1572         return 0;
1573 }
1574
1575 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1576 {
1577         struct nfs4_state *newstate;
1578         int ret;
1579
1580         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1581         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1582         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1583         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1584         /* memory barrier prior to reading state->n_* */
1585         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1586         clear_bit(NFS_OPEN_STATE, &state->flags);
1587         smp_rmb();
1588         if (state->n_rdwr != 0) {
1589                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1590                 if (ret != 0)
1591                         return ret;
1592                 if (newstate != state)
1593                         return -ESTALE;
1594         }
1595         if (state->n_wronly != 0) {
1596                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1597                 if (ret != 0)
1598                         return ret;
1599                 if (newstate != state)
1600                         return -ESTALE;
1601         }
1602         if (state->n_rdonly != 0) {
1603                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1604                 if (ret != 0)
1605                         return ret;
1606                 if (newstate != state)
1607                         return -ESTALE;
1608         }
1609         /*
1610          * We may have performed cached opens for all three recoveries.
1611          * Check if we need to update the current stateid.
1612          */
1613         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1614             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1615                 write_seqlock(&state->seqlock);
1616                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1617                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1618                 write_sequnlock(&state->seqlock);
1619         }
1620         return 0;
1621 }
1622
1623 /*
1624  * OPEN_RECLAIM:
1625  *      reclaim state on the server after a reboot.
1626  */
1627 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1628 {
1629         struct nfs_delegation *delegation;
1630         struct nfs4_opendata *opendata;
1631         fmode_t delegation_type = 0;
1632         int status;
1633
1634         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1635                         NFS4_OPEN_CLAIM_PREVIOUS);
1636         if (IS_ERR(opendata))
1637                 return PTR_ERR(opendata);
1638         rcu_read_lock();
1639         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1640         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1641                 delegation_type = delegation->type;
1642         rcu_read_unlock();
1643         opendata->o_arg.u.delegation_type = delegation_type;
1644         status = nfs4_open_recover(opendata, state);
1645         nfs4_opendata_put(opendata);
1646         return status;
1647 }
1648
1649 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1650 {
1651         struct nfs_server *server = NFS_SERVER(state->inode);
1652         struct nfs4_exception exception = { };
1653         int err;
1654         do {
1655                 err = _nfs4_do_open_reclaim(ctx, state);
1656                 trace_nfs4_open_reclaim(ctx, 0, err);
1657                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1658                         continue;
1659                 if (err != -NFS4ERR_DELAY)
1660                         break;
1661                 nfs4_handle_exception(server, err, &exception);
1662         } while (exception.retry);
1663         return err;
1664 }
1665
1666 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1667 {
1668         struct nfs_open_context *ctx;
1669         int ret;
1670
1671         ctx = nfs4_state_find_open_context(state);
1672         if (IS_ERR(ctx))
1673                 return -EAGAIN;
1674         ret = nfs4_do_open_reclaim(ctx, state);
1675         put_nfs_open_context(ctx);
1676         return ret;
1677 }
1678
1679 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1680 {
1681         switch (err) {
1682                 default:
1683                         printk(KERN_ERR "NFS: %s: unhandled error "
1684                                         "%d.\n", __func__, err);
1685                 case 0:
1686                 case -ENOENT:
1687                 case -ESTALE:
1688                         break;
1689                 case -NFS4ERR_BADSESSION:
1690                 case -NFS4ERR_BADSLOT:
1691                 case -NFS4ERR_BAD_HIGH_SLOT:
1692                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1693                 case -NFS4ERR_DEADSESSION:
1694                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1695                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1696                         return -EAGAIN;
1697                 case -NFS4ERR_STALE_CLIENTID:
1698                 case -NFS4ERR_STALE_STATEID:
1699                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1700                 case -NFS4ERR_EXPIRED:
1701                         /* Don't recall a delegation if it was lost */
1702                         nfs4_schedule_lease_recovery(server->nfs_client);
1703                         return -EAGAIN;
1704                 case -NFS4ERR_MOVED:
1705                         nfs4_schedule_migration_recovery(server);
1706                         return -EAGAIN;
1707                 case -NFS4ERR_LEASE_MOVED:
1708                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1709                         return -EAGAIN;
1710                 case -NFS4ERR_DELEG_REVOKED:
1711                 case -NFS4ERR_ADMIN_REVOKED:
1712                 case -NFS4ERR_BAD_STATEID:
1713                 case -NFS4ERR_OPENMODE:
1714                         nfs_inode_find_state_and_recover(state->inode,
1715                                         stateid);
1716                         nfs4_schedule_stateid_recovery(server, state);
1717                         return -EAGAIN;
1718                 case -NFS4ERR_DELAY:
1719                 case -NFS4ERR_GRACE:
1720                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1721                         ssleep(1);
1722                         return -EAGAIN;
1723                 case -ENOMEM:
1724                 case -NFS4ERR_DENIED:
1725                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1726                         return 0;
1727         }
1728         return err;
1729 }
1730
1731 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1732 {
1733         struct nfs_server *server = NFS_SERVER(state->inode);
1734         struct nfs4_opendata *opendata;
1735         int err;
1736
1737         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1738                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1739         if (IS_ERR(opendata))
1740                 return PTR_ERR(opendata);
1741         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1742         err = nfs4_open_recover(opendata, state);
1743         nfs4_opendata_put(opendata);
1744         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1745 }
1746
1747 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1748 {
1749         struct nfs4_opendata *data = calldata;
1750
1751         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1752                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1753 }
1754
1755 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1756 {
1757         struct nfs4_opendata *data = calldata;
1758
1759         nfs40_sequence_done(task, &data->c_res.seq_res);
1760
1761         data->rpc_status = task->tk_status;
1762         if (data->rpc_status == 0) {
1763                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1764                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1765                 renew_lease(data->o_res.server, data->timestamp);
1766                 data->rpc_done = 1;
1767         }
1768 }
1769
1770 static void nfs4_open_confirm_release(void *calldata)
1771 {
1772         struct nfs4_opendata *data = calldata;
1773         struct nfs4_state *state = NULL;
1774
1775         /* If this request hasn't been cancelled, do nothing */
1776         if (data->cancelled == 0)
1777                 goto out_free;
1778         /* In case of error, no cleanup! */
1779         if (!data->rpc_done)
1780                 goto out_free;
1781         state = nfs4_opendata_to_nfs4_state(data);
1782         if (!IS_ERR(state))
1783                 nfs4_close_state(state, data->o_arg.fmode);
1784 out_free:
1785         nfs4_opendata_put(data);
1786 }
1787
1788 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1789         .rpc_call_prepare = nfs4_open_confirm_prepare,
1790         .rpc_call_done = nfs4_open_confirm_done,
1791         .rpc_release = nfs4_open_confirm_release,
1792 };
1793
1794 /*
1795  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1796  */
1797 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1798 {
1799         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1800         struct rpc_task *task;
1801         struct  rpc_message msg = {
1802                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1803                 .rpc_argp = &data->c_arg,
1804                 .rpc_resp = &data->c_res,
1805                 .rpc_cred = data->owner->so_cred,
1806         };
1807         struct rpc_task_setup task_setup_data = {
1808                 .rpc_client = server->client,
1809                 .rpc_message = &msg,
1810                 .callback_ops = &nfs4_open_confirm_ops,
1811                 .callback_data = data,
1812                 .workqueue = nfsiod_workqueue,
1813                 .flags = RPC_TASK_ASYNC,
1814         };
1815         int status;
1816
1817         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1818         kref_get(&data->kref);
1819         data->rpc_done = 0;
1820         data->rpc_status = 0;
1821         data->timestamp = jiffies;
1822         task = rpc_run_task(&task_setup_data);
1823         if (IS_ERR(task))
1824                 return PTR_ERR(task);
1825         status = nfs4_wait_for_completion_rpc_task(task);
1826         if (status != 0) {
1827                 data->cancelled = 1;
1828                 smp_wmb();
1829         } else
1830                 status = data->rpc_status;
1831         rpc_put_task(task);
1832         return status;
1833 }
1834
1835 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1836 {
1837         struct nfs4_opendata *data = calldata;
1838         struct nfs4_state_owner *sp = data->owner;
1839         struct nfs_client *clp = sp->so_server->nfs_client;
1840
1841         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1842                 goto out_wait;
1843         /*
1844          * Check if we still need to send an OPEN call, or if we can use
1845          * a delegation instead.
1846          */
1847         if (data->state != NULL) {
1848                 struct nfs_delegation *delegation;
1849
1850                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1851                         goto out_no_action;
1852                 rcu_read_lock();
1853                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1854                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1855                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1856                     can_open_delegated(delegation, data->o_arg.fmode))
1857                         goto unlock_no_action;
1858                 rcu_read_unlock();
1859         }
1860         /* Update client id. */
1861         data->o_arg.clientid = clp->cl_clientid;
1862         switch (data->o_arg.claim) {
1863         case NFS4_OPEN_CLAIM_PREVIOUS:
1864         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1865         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1866                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1867         case NFS4_OPEN_CLAIM_FH:
1868                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1869                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1870         }
1871         data->timestamp = jiffies;
1872         if (nfs4_setup_sequence(data->o_arg.server,
1873                                 &data->o_arg.seq_args,
1874                                 &data->o_res.seq_res,
1875                                 task) != 0)
1876                 nfs_release_seqid(data->o_arg.seqid);
1877
1878         /* Set the create mode (note dependency on the session type) */
1879         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1880         if (data->o_arg.open_flags & O_EXCL) {
1881                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1882                 if (nfs4_has_persistent_session(clp))
1883                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1884                 else if (clp->cl_mvops->minor_version > 0)
1885                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1886         }
1887         return;
1888 unlock_no_action:
1889         rcu_read_unlock();
1890 out_no_action:
1891         task->tk_action = NULL;
1892 out_wait:
1893         nfs4_sequence_done(task, &data->o_res.seq_res);
1894 }
1895
1896 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1897 {
1898         struct nfs4_opendata *data = calldata;
1899
1900         data->rpc_status = task->tk_status;
1901
1902         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1903                 return;
1904
1905         if (task->tk_status == 0) {
1906                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1907                         switch (data->o_res.f_attr->mode & S_IFMT) {
1908                         case S_IFREG:
1909                                 break;
1910                         case S_IFLNK:
1911                                 data->rpc_status = -ELOOP;
1912                                 break;
1913                         case S_IFDIR:
1914                                 data->rpc_status = -EISDIR;
1915                                 break;
1916                         default:
1917                                 data->rpc_status = -ENOTDIR;
1918                         }
1919                 }
1920                 renew_lease(data->o_res.server, data->timestamp);
1921                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1922                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1923         }
1924         data->rpc_done = 1;
1925 }
1926
1927 static void nfs4_open_release(void *calldata)
1928 {
1929         struct nfs4_opendata *data = calldata;
1930         struct nfs4_state *state = NULL;
1931
1932         /* If this request hasn't been cancelled, do nothing */
1933         if (data->cancelled == 0)
1934                 goto out_free;
1935         /* In case of error, no cleanup! */
1936         if (data->rpc_status != 0 || !data->rpc_done)
1937                 goto out_free;
1938         /* In case we need an open_confirm, no cleanup! */
1939         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1940                 goto out_free;
1941         state = nfs4_opendata_to_nfs4_state(data);
1942         if (!IS_ERR(state))
1943                 nfs4_close_state(state, data->o_arg.fmode);
1944 out_free:
1945         nfs4_opendata_put(data);
1946 }
1947
1948 static const struct rpc_call_ops nfs4_open_ops = {
1949         .rpc_call_prepare = nfs4_open_prepare,
1950         .rpc_call_done = nfs4_open_done,
1951         .rpc_release = nfs4_open_release,
1952 };
1953
1954 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1955 {
1956         struct inode *dir = d_inode(data->dir);
1957         struct nfs_server *server = NFS_SERVER(dir);
1958         struct nfs_openargs *o_arg = &data->o_arg;
1959         struct nfs_openres *o_res = &data->o_res;
1960         struct rpc_task *task;
1961         struct rpc_message msg = {
1962                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1963                 .rpc_argp = o_arg,
1964                 .rpc_resp = o_res,
1965                 .rpc_cred = data->owner->so_cred,
1966         };
1967         struct rpc_task_setup task_setup_data = {
1968                 .rpc_client = server->client,
1969                 .rpc_message = &msg,
1970                 .callback_ops = &nfs4_open_ops,
1971                 .callback_data = data,
1972                 .workqueue = nfsiod_workqueue,
1973                 .flags = RPC_TASK_ASYNC,
1974         };
1975         int status;
1976
1977         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1978         kref_get(&data->kref);
1979         data->rpc_done = 0;
1980         data->rpc_status = 0;
1981         data->cancelled = 0;
1982         data->is_recover = 0;
1983         if (isrecover) {
1984                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1985                 data->is_recover = 1;
1986         }
1987         task = rpc_run_task(&task_setup_data);
1988         if (IS_ERR(task))
1989                 return PTR_ERR(task);
1990         status = nfs4_wait_for_completion_rpc_task(task);
1991         if (status != 0) {
1992                 data->cancelled = 1;
1993                 smp_wmb();
1994         } else
1995                 status = data->rpc_status;
1996         rpc_put_task(task);
1997
1998         return status;
1999 }
2000
2001 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2002 {
2003         struct inode *dir = d_inode(data->dir);
2004         struct nfs_openres *o_res = &data->o_res;
2005         int status;
2006
2007         status = nfs4_run_open_task(data, 1);
2008         if (status != 0 || !data->rpc_done)
2009                 return status;
2010
2011         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2012
2013         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2014                 status = _nfs4_proc_open_confirm(data);
2015                 if (status != 0)
2016                         return status;
2017         }
2018
2019         return status;
2020 }
2021
2022 /*
2023  * Additional permission checks in order to distinguish between an
2024  * open for read, and an open for execute. This works around the
2025  * fact that NFSv4 OPEN treats read and execute permissions as being
2026  * the same.
2027  * Note that in the non-execute case, we want to turn off permission
2028  * checking if we just created a new file (POSIX open() semantics).
2029  */
2030 static int nfs4_opendata_access(struct rpc_cred *cred,
2031                                 struct nfs4_opendata *opendata,
2032                                 struct nfs4_state *state, fmode_t fmode,
2033                                 int openflags)
2034 {
2035         struct nfs_access_entry cache;
2036         u32 mask;
2037
2038         /* access call failed or for some reason the server doesn't
2039          * support any access modes -- defer access call until later */
2040         if (opendata->o_res.access_supported == 0)
2041                 return 0;
2042
2043         mask = 0;
2044         /*
2045          * Use openflags to check for exec, because fmode won't
2046          * always have FMODE_EXEC set when file open for exec.
2047          */
2048         if (openflags & __FMODE_EXEC) {
2049                 /* ONLY check for exec rights */
2050                 mask = MAY_EXEC;
2051         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2052                 mask = MAY_READ;
2053
2054         cache.cred = cred;
2055         cache.jiffies = jiffies;
2056         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2057         nfs_access_add_cache(state->inode, &cache);
2058
2059         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2060                 return 0;
2061
2062         /* even though OPEN succeeded, access is denied. Close the file */
2063         nfs4_close_state(state, fmode);
2064         return -EACCES;
2065 }
2066
2067 /*
2068  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2069  */
2070 static int _nfs4_proc_open(struct nfs4_opendata *data)
2071 {
2072         struct inode *dir = d_inode(data->dir);
2073         struct nfs_server *server = NFS_SERVER(dir);
2074         struct nfs_openargs *o_arg = &data->o_arg;
2075         struct nfs_openres *o_res = &data->o_res;
2076         int status;
2077
2078         status = nfs4_run_open_task(data, 0);
2079         if (!data->rpc_done)
2080                 return status;
2081         if (status != 0) {
2082                 if (status == -NFS4ERR_BADNAME &&
2083                                 !(o_arg->open_flags & O_CREAT))
2084                         return -ENOENT;
2085                 return status;
2086         }
2087
2088         nfs_fattr_map_and_free_names(server, &data->f_attr);
2089
2090         if (o_arg->open_flags & O_CREAT) {
2091                 update_changeattr(dir, &o_res->cinfo);
2092                 if (o_arg->open_flags & O_EXCL)
2093                         data->file_created = 1;
2094                 else if (o_res->cinfo.before != o_res->cinfo.after)
2095                         data->file_created = 1;
2096         }
2097         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2098                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2099         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2100                 status = _nfs4_proc_open_confirm(data);
2101                 if (status != 0)
2102                         return status;
2103         }
2104         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2105                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2106         return 0;
2107 }
2108
2109 static int nfs4_recover_expired_lease(struct nfs_server *server)
2110 {
2111         return nfs4_client_recover_expired_lease(server->nfs_client);
2112 }
2113
2114 /*
2115  * OPEN_EXPIRED:
2116  *      reclaim state on the server after a network partition.
2117  *      Assumes caller holds the appropriate lock
2118  */
2119 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2120 {
2121         struct nfs4_opendata *opendata;
2122         int ret;
2123
2124         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2125                         NFS4_OPEN_CLAIM_FH);
2126         if (IS_ERR(opendata))
2127                 return PTR_ERR(opendata);
2128         ret = nfs4_open_recover(opendata, state);
2129         if (ret == -ESTALE)
2130                 d_drop(ctx->dentry);
2131         nfs4_opendata_put(opendata);
2132         return ret;
2133 }
2134
2135 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2136 {
2137         struct nfs_server *server = NFS_SERVER(state->inode);
2138         struct nfs4_exception exception = { };
2139         int err;
2140
2141         do {
2142                 err = _nfs4_open_expired(ctx, state);
2143                 trace_nfs4_open_expired(ctx, 0, err);
2144                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2145                         continue;
2146                 switch (err) {
2147                 default:
2148                         goto out;
2149                 case -NFS4ERR_GRACE:
2150                 case -NFS4ERR_DELAY:
2151                         nfs4_handle_exception(server, err, &exception);
2152                         err = 0;
2153                 }
2154         } while (exception.retry);
2155 out:
2156         return err;
2157 }
2158
2159 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2160 {
2161         struct nfs_open_context *ctx;
2162         int ret;
2163
2164         ctx = nfs4_state_find_open_context(state);
2165         if (IS_ERR(ctx))
2166                 return -EAGAIN;
2167         ret = nfs4_do_open_expired(ctx, state);
2168         put_nfs_open_context(ctx);
2169         return ret;
2170 }
2171
2172 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2173 {
2174         nfs_remove_bad_delegation(state->inode);
2175         write_seqlock(&state->seqlock);
2176         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2177         write_sequnlock(&state->seqlock);
2178         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2179 }
2180
2181 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2182 {
2183         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2184                 nfs_finish_clear_delegation_stateid(state);
2185 }
2186
2187 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2188 {
2189         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2190         nfs40_clear_delegation_stateid(state);
2191         return nfs4_open_expired(sp, state);
2192 }
2193
2194 #if defined(CONFIG_NFS_V4_1)
2195 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2196 {
2197         struct nfs_server *server = NFS_SERVER(state->inode);
2198         nfs4_stateid stateid;
2199         struct nfs_delegation *delegation;
2200         struct rpc_cred *cred;
2201         int status;
2202
2203         /* Get the delegation credential for use by test/free_stateid */
2204         rcu_read_lock();
2205         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2206         if (delegation == NULL) {
2207                 rcu_read_unlock();
2208                 return;
2209         }
2210
2211         nfs4_stateid_copy(&stateid, &delegation->stateid);
2212         cred = get_rpccred(delegation->cred);
2213         rcu_read_unlock();
2214         status = nfs41_test_stateid(server, &stateid, cred);
2215         trace_nfs4_test_delegation_stateid(state, NULL, status);
2216
2217         if (status != NFS_OK) {
2218                 /* Free the stateid unless the server explicitly
2219                  * informs us the stateid is unrecognized. */
2220                 if (status != -NFS4ERR_BAD_STATEID)
2221                         nfs41_free_stateid(server, &stateid, cred);
2222                 nfs_finish_clear_delegation_stateid(state);
2223         }
2224
2225         put_rpccred(cred);
2226 }
2227
2228 /**
2229  * nfs41_check_open_stateid - possibly free an open stateid
2230  *
2231  * @state: NFSv4 state for an inode
2232  *
2233  * Returns NFS_OK if recovery for this stateid is now finished.
2234  * Otherwise a negative NFS4ERR value is returned.
2235  */
2236 static int nfs41_check_open_stateid(struct nfs4_state *state)
2237 {
2238         struct nfs_server *server = NFS_SERVER(state->inode);
2239         nfs4_stateid *stateid = &state->open_stateid;
2240         struct rpc_cred *cred = state->owner->so_cred;
2241         int status;
2242
2243         /* If a state reset has been done, test_stateid is unneeded */
2244         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2245             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2246             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2247                 return -NFS4ERR_BAD_STATEID;
2248
2249         status = nfs41_test_stateid(server, stateid, cred);
2250         trace_nfs4_test_open_stateid(state, NULL, status);
2251         if (status != NFS_OK) {
2252                 /* Free the stateid unless the server explicitly
2253                  * informs us the stateid is unrecognized. */
2254                 if (status != -NFS4ERR_BAD_STATEID)
2255                         nfs41_free_stateid(server, stateid, cred);
2256
2257                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2258                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2259                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2260                 clear_bit(NFS_OPEN_STATE, &state->flags);
2261         }
2262         return status;
2263 }
2264
2265 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2266 {
2267         int status;
2268
2269         nfs41_check_delegation_stateid(state);
2270         status = nfs41_check_open_stateid(state);
2271         if (status != NFS_OK)
2272                 status = nfs4_open_expired(sp, state);
2273         return status;
2274 }
2275 #endif
2276
2277 /*
2278  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2279  * fields corresponding to attributes that were used to store the verifier.
2280  * Make sure we clobber those fields in the later setattr call
2281  */
2282 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2283 {
2284         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2285             !(sattr->ia_valid & ATTR_ATIME_SET))
2286                 sattr->ia_valid |= ATTR_ATIME;
2287
2288         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2289             !(sattr->ia_valid & ATTR_MTIME_SET))
2290                 sattr->ia_valid |= ATTR_MTIME;
2291 }
2292
2293 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2294                 fmode_t fmode,
2295                 int flags,
2296                 struct nfs_open_context *ctx)
2297 {
2298         struct nfs4_state_owner *sp = opendata->owner;
2299         struct nfs_server *server = sp->so_server;
2300         struct dentry *dentry;
2301         struct nfs4_state *state;
2302         unsigned int seq;
2303         int ret;
2304
2305         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2306
2307         ret = _nfs4_proc_open(opendata);
2308         if (ret != 0)
2309                 goto out;
2310
2311         state = nfs4_opendata_to_nfs4_state(opendata);
2312         ret = PTR_ERR(state);
2313         if (IS_ERR(state))
2314                 goto out;
2315         if (server->caps & NFS_CAP_POSIX_LOCK)
2316                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2317
2318         dentry = opendata->dentry;
2319         if (d_really_is_negative(dentry)) {
2320                 /* FIXME: Is this d_drop() ever needed? */
2321                 d_drop(dentry);
2322                 dentry = d_add_unique(dentry, igrab(state->inode));
2323                 if (dentry == NULL) {
2324                         dentry = opendata->dentry;
2325                 } else if (dentry != ctx->dentry) {
2326                         dput(ctx->dentry);
2327                         ctx->dentry = dget(dentry);
2328                 }
2329                 nfs_set_verifier(dentry,
2330                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2331         }
2332
2333         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2334         if (ret != 0)
2335                 goto out;
2336
2337         ctx->state = state;
2338         if (d_inode(dentry) == state->inode) {
2339                 nfs_inode_attach_open_context(ctx);
2340                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2341                         nfs4_schedule_stateid_recovery(server, state);
2342         }
2343 out:
2344         return ret;
2345 }
2346
2347 /*
2348  * Returns a referenced nfs4_state
2349  */
2350 static int _nfs4_do_open(struct inode *dir,
2351                         struct nfs_open_context *ctx,
2352                         int flags,
2353                         struct iattr *sattr,
2354                         struct nfs4_label *label,
2355                         int *opened)
2356 {
2357         struct nfs4_state_owner  *sp;
2358         struct nfs4_state     *state = NULL;
2359         struct nfs_server       *server = NFS_SERVER(dir);
2360         struct nfs4_opendata *opendata;
2361         struct dentry *dentry = ctx->dentry;
2362         struct rpc_cred *cred = ctx->cred;
2363         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2364         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2365         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2366         struct nfs4_label *olabel = NULL;
2367         int status;
2368
2369         /* Protect against reboot recovery conflicts */
2370         status = -ENOMEM;
2371         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2372         if (sp == NULL) {
2373                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2374                 goto out_err;
2375         }
2376         status = nfs4_recover_expired_lease(server);
2377         if (status != 0)
2378                 goto err_put_state_owner;
2379         if (d_really_is_positive(dentry))
2380                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2381         status = -ENOMEM;
2382         if (d_really_is_positive(dentry))
2383                 claim = NFS4_OPEN_CLAIM_FH;
2384         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2385                         label, claim, GFP_KERNEL);
2386         if (opendata == NULL)
2387                 goto err_put_state_owner;
2388
2389         if (label) {
2390                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2391                 if (IS_ERR(olabel)) {
2392                         status = PTR_ERR(olabel);
2393                         goto err_opendata_put;
2394                 }
2395         }
2396
2397         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2398                 if (!opendata->f_attr.mdsthreshold) {
2399                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2400                         if (!opendata->f_attr.mdsthreshold)
2401                                 goto err_free_label;
2402                 }
2403                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2404         }
2405         if (d_really_is_positive(dentry))
2406                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2407
2408         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2409         if (status != 0)
2410                 goto err_free_label;
2411         state = ctx->state;
2412
2413         if ((opendata->o_arg.open_flags & O_EXCL) &&
2414             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2415                 nfs4_exclusive_attrset(opendata, sattr);
2416
2417                 nfs_fattr_init(opendata->o_res.f_attr);
2418                 status = nfs4_do_setattr(state->inode, cred,
2419                                 opendata->o_res.f_attr, sattr,
2420                                 state, label, olabel);
2421                 if (status == 0) {
2422                         nfs_setattr_update_inode(state->inode, sattr,
2423                                         opendata->o_res.f_attr);
2424                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2425                 }
2426         }
2427         if (opendata->file_created)
2428                 *opened |= FILE_CREATED;
2429
2430         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2431                 *ctx_th = opendata->f_attr.mdsthreshold;
2432                 opendata->f_attr.mdsthreshold = NULL;
2433         }
2434
2435         nfs4_label_free(olabel);
2436
2437         nfs4_opendata_put(opendata);
2438         nfs4_put_state_owner(sp);
2439         return 0;
2440 err_free_label:
2441         nfs4_label_free(olabel);
2442 err_opendata_put:
2443         nfs4_opendata_put(opendata);
2444 err_put_state_owner:
2445         nfs4_put_state_owner(sp);
2446 out_err:
2447         return status;
2448 }
2449
2450
2451 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2452                                         struct nfs_open_context *ctx,
2453                                         int flags,
2454                                         struct iattr *sattr,
2455                                         struct nfs4_label *label,
2456                                         int *opened)
2457 {
2458         struct nfs_server *server = NFS_SERVER(dir);
2459         struct nfs4_exception exception = { };
2460         struct nfs4_state *res;
2461         int status;
2462
2463         do {
2464                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2465                 res = ctx->state;
2466                 trace_nfs4_open_file(ctx, flags, status);
2467                 if (status == 0)
2468                         break;
2469                 /* NOTE: BAD_SEQID means the server and client disagree about the
2470                  * book-keeping w.r.t. state-changing operations
2471                  * (OPEN/CLOSE/LOCK/LOCKU...)
2472                  * It is actually a sign of a bug on the client or on the server.
2473                  *
2474                  * If we receive a BAD_SEQID error in the particular case of
2475                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2476                  * have unhashed the old state_owner for us, and that we can
2477                  * therefore safely retry using a new one. We should still warn
2478                  * the user though...
2479                  */
2480                 if (status == -NFS4ERR_BAD_SEQID) {
2481                         pr_warn_ratelimited("NFS: v4 server %s "
2482                                         " returned a bad sequence-id error!\n",
2483                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2484                         exception.retry = 1;
2485                         continue;
2486                 }
2487                 /*
2488                  * BAD_STATEID on OPEN means that the server cancelled our
2489                  * state before it received the OPEN_CONFIRM.
2490                  * Recover by retrying the request as per the discussion
2491                  * on Page 181 of RFC3530.
2492                  */
2493                 if (status == -NFS4ERR_BAD_STATEID) {
2494                         exception.retry = 1;
2495                         continue;
2496                 }
2497                 if (status == -EAGAIN) {
2498                         /* We must have found a delegation */
2499                         exception.retry = 1;
2500                         continue;
2501                 }
2502                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2503                         continue;
2504                 res = ERR_PTR(nfs4_handle_exception(server,
2505                                         status, &exception));
2506         } while (exception.retry);
2507         return res;
2508 }
2509
2510 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2511                             struct nfs_fattr *fattr, struct iattr *sattr,
2512                             struct nfs4_state *state, struct nfs4_label *ilabel,
2513                             struct nfs4_label *olabel)
2514 {
2515         struct nfs_server *server = NFS_SERVER(inode);
2516         struct nfs_setattrargs  arg = {
2517                 .fh             = NFS_FH(inode),
2518                 .iap            = sattr,
2519                 .server         = server,
2520                 .bitmask = server->attr_bitmask,
2521                 .label          = ilabel,
2522         };
2523         struct nfs_setattrres  res = {
2524                 .fattr          = fattr,
2525                 .label          = olabel,
2526                 .server         = server,
2527         };
2528         struct rpc_message msg = {
2529                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2530                 .rpc_argp       = &arg,
2531                 .rpc_resp       = &res,
2532                 .rpc_cred       = cred,
2533         };
2534         unsigned long timestamp = jiffies;
2535         fmode_t fmode;
2536         bool truncate;
2537         int status;
2538
2539         arg.bitmask = nfs4_bitmask(server, ilabel);
2540         if (ilabel)
2541                 arg.bitmask = nfs4_bitmask(server, olabel);
2542
2543         nfs_fattr_init(fattr);
2544
2545         /* Servers should only apply open mode checks for file size changes */
2546         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2547         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2548
2549         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2550                 /* Use that stateid */
2551         } else if (truncate && state != NULL) {
2552                 struct nfs_lockowner lockowner = {
2553                         .l_owner = current->files,
2554                         .l_pid = current->tgid,
2555                 };
2556                 if (!nfs4_valid_open_stateid(state))
2557                         return -EBADF;
2558                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2559                                 &lockowner) == -EIO)
2560                         return -EBADF;
2561         } else
2562                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2563
2564         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2565         if (status == 0 && state != NULL)
2566                 renew_lease(server, timestamp);
2567         return status;
2568 }
2569
2570 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2571                            struct nfs_fattr *fattr, struct iattr *sattr,
2572                            struct nfs4_state *state, struct nfs4_label *ilabel,
2573                            struct nfs4_label *olabel)
2574 {
2575         struct nfs_server *server = NFS_SERVER(inode);
2576         struct nfs4_exception exception = {
2577                 .state = state,
2578                 .inode = inode,
2579         };
2580         int err;
2581         do {
2582                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2583                 trace_nfs4_setattr(inode, err);
2584                 switch (err) {
2585                 case -NFS4ERR_OPENMODE:
2586                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2587                                 pr_warn_once("NFSv4: server %s is incorrectly "
2588                                                 "applying open mode checks to "
2589                                                 "a SETATTR that is not "
2590                                                 "changing file size.\n",
2591                                                 server->nfs_client->cl_hostname);
2592                         }
2593                         if (state && !(state->state & FMODE_WRITE)) {
2594                                 err = -EBADF;
2595                                 if (sattr->ia_valid & ATTR_OPEN)
2596                                         err = -EACCES;
2597                                 goto out;
2598                         }
2599                 }
2600                 err = nfs4_handle_exception(server, err, &exception);
2601         } while (exception.retry);
2602 out:
2603         return err;
2604 }
2605
2606 struct nfs4_closedata {
2607         struct inode *inode;
2608         struct nfs4_state *state;
2609         struct nfs_closeargs arg;
2610         struct nfs_closeres res;
2611         struct nfs_fattr fattr;
2612         unsigned long timestamp;
2613         bool roc;
2614         u32 roc_barrier;
2615 };
2616
2617 static void nfs4_free_closedata(void *data)
2618 {
2619         struct nfs4_closedata *calldata = data;
2620         struct nfs4_state_owner *sp = calldata->state->owner;
2621         struct super_block *sb = calldata->state->inode->i_sb;
2622
2623         if (calldata->roc)
2624                 pnfs_roc_release(calldata->state->inode);
2625         nfs4_put_open_state(calldata->state);
2626         nfs_free_seqid(calldata->arg.seqid);
2627         nfs4_put_state_owner(sp);
2628         nfs_sb_deactive(sb);
2629         kfree(calldata);
2630 }
2631
2632 static void nfs4_close_done(struct rpc_task *task, void *data)
2633 {
2634         struct nfs4_closedata *calldata = data;
2635         struct nfs4_state *state = calldata->state;
2636         struct nfs_server *server = NFS_SERVER(calldata->inode);
2637         nfs4_stateid *res_stateid = NULL;
2638
2639         dprintk("%s: begin!\n", __func__);
2640         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2641                 return;
2642         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2643         /* hmm. we are done with the inode, and in the process of freeing
2644          * the state_owner. we keep this around to process errors
2645          */
2646         switch (task->tk_status) {
2647                 case 0:
2648                         res_stateid = &calldata->res.stateid;
2649                         if (calldata->arg.fmode == 0 && calldata->roc)
2650                                 pnfs_roc_set_barrier(state->inode,
2651                                                      calldata->roc_barrier);
2652                         renew_lease(server, calldata->timestamp);
2653                         break;
2654                 case -NFS4ERR_ADMIN_REVOKED:
2655                 case -NFS4ERR_STALE_STATEID:
2656                 case -NFS4ERR_OLD_STATEID:
2657                 case -NFS4ERR_BAD_STATEID:
2658                 case -NFS4ERR_EXPIRED:
2659                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2660                                                 &state->open_stateid)) {
2661                                 rpc_restart_call_prepare(task);
2662                                 goto out_release;
2663                         }
2664                         if (calldata->arg.fmode == 0)
2665                                 break;
2666                 default:
2667                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2668                                 rpc_restart_call_prepare(task);
2669                                 goto out_release;
2670                         }
2671         }
2672         nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2673 out_release:
2674         nfs_release_seqid(calldata->arg.seqid);
2675         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2676         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2677 }
2678
2679 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2680 {
2681         struct nfs4_closedata *calldata = data;
2682         struct nfs4_state *state = calldata->state;
2683         struct inode *inode = calldata->inode;
2684         bool is_rdonly, is_wronly, is_rdwr;
2685         int call_close = 0;
2686
2687         dprintk("%s: begin!\n", __func__);
2688         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2689                 goto out_wait;
2690
2691         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2692         spin_lock(&state->owner->so_lock);
2693         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2694         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2695         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2696         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2697         /* Calculate the change in open mode */
2698         calldata->arg.fmode = 0;
2699         if (state->n_rdwr == 0) {
2700                 if (state->n_rdonly == 0)
2701                         call_close |= is_rdonly;
2702                 else if (is_rdonly)
2703                         calldata->arg.fmode |= FMODE_READ;
2704                 if (state->n_wronly == 0)
2705                         call_close |= is_wronly;
2706                 else if (is_wronly)
2707                         calldata->arg.fmode |= FMODE_WRITE;
2708         } else if (is_rdwr)
2709                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2710
2711         if (calldata->arg.fmode == 0)
2712                 call_close |= is_rdwr;
2713
2714         if (!nfs4_valid_open_stateid(state))
2715                 call_close = 0;
2716         spin_unlock(&state->owner->so_lock);
2717
2718         if (!call_close) {
2719                 /* Note: exit _without_ calling nfs4_close_done */
2720                 goto out_no_action;
2721         }
2722
2723         if (calldata->arg.fmode == 0) {
2724                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2725                 if (calldata->roc &&
2726                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2727                         nfs_release_seqid(calldata->arg.seqid);
2728                         goto out_wait;
2729                     }
2730         }
2731         calldata->arg.share_access =
2732                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2733                                 calldata->arg.fmode, 0);
2734
2735         nfs_fattr_init(calldata->res.fattr);
2736         calldata->timestamp = jiffies;
2737         if (nfs4_setup_sequence(NFS_SERVER(inode),
2738                                 &calldata->arg.seq_args,
2739                                 &calldata->res.seq_res,
2740                                 task) != 0)
2741                 nfs_release_seqid(calldata->arg.seqid);
2742         dprintk("%s: done!\n", __func__);
2743         return;
2744 out_no_action:
2745         task->tk_action = NULL;
2746 out_wait:
2747         nfs4_sequence_done(task, &calldata->res.seq_res);
2748 }
2749
2750 static const struct rpc_call_ops nfs4_close_ops = {
2751         .rpc_call_prepare = nfs4_close_prepare,
2752         .rpc_call_done = nfs4_close_done,
2753         .rpc_release = nfs4_free_closedata,
2754 };
2755
2756 static bool nfs4_roc(struct inode *inode)
2757 {
2758         if (!nfs_have_layout(inode))
2759                 return false;
2760         return pnfs_roc(inode);
2761 }
2762
2763 /* 
2764  * It is possible for data to be read/written from a mem-mapped file 
2765  * after the sys_close call (which hits the vfs layer as a flush).
2766  * This means that we can't safely call nfsv4 close on a file until 
2767  * the inode is cleared. This in turn means that we are not good
2768  * NFSv4 citizens - we do not indicate to the server to update the file's 
2769  * share state even when we are done with one of the three share 
2770  * stateid's in the inode.
2771  *
2772  * NOTE: Caller must be holding the sp->so_owner semaphore!
2773  */
2774 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2775 {
2776         struct nfs_server *server = NFS_SERVER(state->inode);
2777         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2778         struct nfs4_closedata *calldata;
2779         struct nfs4_state_owner *sp = state->owner;
2780         struct rpc_task *task;
2781         struct rpc_message msg = {
2782                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2783                 .rpc_cred = state->owner->so_cred,
2784         };
2785         struct rpc_task_setup task_setup_data = {
2786                 .rpc_client = server->client,
2787                 .rpc_message = &msg,
2788                 .callback_ops = &nfs4_close_ops,
2789                 .workqueue = nfsiod_workqueue,
2790                 .flags = RPC_TASK_ASYNC,
2791         };
2792         int status = -ENOMEM;
2793
2794         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2795                 &task_setup_data.rpc_client, &msg);
2796
2797         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2798         if (calldata == NULL)
2799                 goto out;
2800         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2801         calldata->inode = state->inode;
2802         calldata->state = state;
2803         calldata->arg.fh = NFS_FH(state->inode);
2804         /* Serialization for the sequence id */
2805         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2806         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2807         if (IS_ERR(calldata->arg.seqid))
2808                 goto out_free_calldata;
2809         calldata->arg.fmode = 0;
2810         calldata->arg.bitmask = server->cache_consistency_bitmask;
2811         calldata->res.fattr = &calldata->fattr;
2812         calldata->res.seqid = calldata->arg.seqid;
2813         calldata->res.server = server;
2814         calldata->roc = nfs4_roc(state->inode);
2815         nfs_sb_active(calldata->inode->i_sb);
2816
2817         msg.rpc_argp = &calldata->arg;
2818         msg.rpc_resp = &calldata->res;
2819         task_setup_data.callback_data = calldata;
2820         task = rpc_run_task(&task_setup_data);
2821         if (IS_ERR(task))
2822                 return PTR_ERR(task);
2823         status = 0;
2824         if (wait)
2825                 status = rpc_wait_for_completion_task(task);
2826         rpc_put_task(task);
2827         return status;
2828 out_free_calldata:
2829         kfree(calldata);
2830 out:
2831         nfs4_put_open_state(state);
2832         nfs4_put_state_owner(sp);
2833         return status;
2834 }
2835
2836 static struct inode *
2837 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2838                 int open_flags, struct iattr *attr, int *opened)
2839 {
2840         struct nfs4_state *state;
2841         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2842
2843         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2844
2845         /* Protect against concurrent sillydeletes */
2846         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2847
2848         nfs4_label_release_security(label);
2849
2850         if (IS_ERR(state))
2851                 return ERR_CAST(state);
2852         return state->inode;
2853 }
2854
2855 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2856 {
2857         if (ctx->state == NULL)
2858                 return;
2859         if (is_sync)
2860                 nfs4_close_sync(ctx->state, ctx->mode);
2861         else
2862                 nfs4_close_state(ctx->state, ctx->mode);
2863 }
2864
2865 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2866 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2867 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2868
2869 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2870 {
2871         struct nfs4_server_caps_arg args = {
2872                 .fhandle = fhandle,
2873         };
2874         struct nfs4_server_caps_res res = {};
2875         struct rpc_message msg = {
2876                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2877                 .rpc_argp = &args,
2878                 .rpc_resp = &res,
2879         };
2880         int status;
2881
2882         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2883         if (status == 0) {
2884                 /* Sanity check the server answers */
2885                 switch (server->nfs_client->cl_minorversion) {
2886                 case 0:
2887                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2888                         res.attr_bitmask[2] = 0;
2889                         break;
2890                 case 1:
2891                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2892                         break;
2893                 case 2:
2894                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2895                 }
2896                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2897                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2898                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2899                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2900                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2901                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2902                                 NFS_CAP_SECURITY_LABEL);
2903                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2904                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2905                         server->caps |= NFS_CAP_ACLS;
2906                 if (res.has_links != 0)
2907                         server->caps |= NFS_CAP_HARDLINKS;
2908                 if (res.has_symlinks != 0)
2909                         server->caps |= NFS_CAP_SYMLINKS;
2910                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2911                         server->caps |= NFS_CAP_FILEID;
2912                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2913                         server->caps |= NFS_CAP_MODE;
2914                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2915                         server->caps |= NFS_CAP_NLINK;
2916                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2917                         server->caps |= NFS_CAP_OWNER;
2918                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2919                         server->caps |= NFS_CAP_OWNER_GROUP;
2920                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2921                         server->caps |= NFS_CAP_ATIME;
2922                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2923                         server->caps |= NFS_CAP_CTIME;
2924                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2925                         server->caps |= NFS_CAP_MTIME;
2926 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2927                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2928                         server->caps |= NFS_CAP_SECURITY_LABEL;
2929 #endif
2930                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2931                                 sizeof(server->attr_bitmask));
2932                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2933
2934                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2935                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2936                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2937                 server->cache_consistency_bitmask[2] = 0;
2938                 server->acl_bitmask = res.acl_bitmask;
2939                 server->fh_expire_type = res.fh_expire_type;
2940         }
2941
2942         return status;
2943 }
2944
2945 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2946 {
2947         struct nfs4_exception exception = { };
2948         int err;
2949         do {
2950                 err = nfs4_handle_exception(server,
2951                                 _nfs4_server_capabilities(server, fhandle),
2952                                 &exception);
2953         } while (exception.retry);
2954         return err;
2955 }
2956
2957 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2958                 struct nfs_fsinfo *info)
2959 {
2960         u32 bitmask[3];
2961         struct nfs4_lookup_root_arg args = {
2962                 .bitmask = bitmask,
2963         };
2964         struct nfs4_lookup_res res = {
2965                 .server = server,
2966                 .fattr = info->fattr,
2967                 .fh = fhandle,
2968         };
2969         struct rpc_message msg = {
2970                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2971                 .rpc_argp = &args,
2972                 .rpc_resp = &res,
2973         };
2974
2975         bitmask[0] = nfs4_fattr_bitmap[0];
2976         bitmask[1] = nfs4_fattr_bitmap[1];
2977         /*
2978          * Process the label in the upcoming getfattr
2979          */
2980         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2981
2982         nfs_fattr_init(info->fattr);
2983         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2984 }
2985
2986 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2987                 struct nfs_fsinfo *info)
2988 {
2989         struct nfs4_exception exception = { };
2990         int err;
2991         do {
2992                 err = _nfs4_lookup_root(server, fhandle, info);
2993                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2994                 switch (err) {
2995                 case 0:
2996                 case -NFS4ERR_WRONGSEC:
2997                         goto out;
2998                 default:
2999                         err = nfs4_handle_exception(server, err, &exception);
3000                 }
3001         } while (exception.retry);
3002 out:
3003         return err;
3004 }
3005
3006 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3007                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3008 {
3009         struct rpc_auth_create_args auth_args = {
3010                 .pseudoflavor = flavor,
3011         };
3012         struct rpc_auth *auth;
3013         int ret;
3014
3015         auth = rpcauth_create(&auth_args, server->client);
3016         if (IS_ERR(auth)) {
3017                 ret = -EACCES;
3018                 goto out;
3019         }
3020         ret = nfs4_lookup_root(server, fhandle, info);
3021 out:
3022         return ret;
3023 }
3024
3025 /*
3026  * Retry pseudoroot lookup with various security flavors.  We do this when:
3027  *
3028  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3029  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3030  *
3031  * Returns zero on success, or a negative NFS4ERR value, or a
3032  * negative errno value.
3033  */
3034 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3035                               struct nfs_fsinfo *info)
3036 {
3037         /* Per 3530bis 15.33.5 */
3038         static const rpc_authflavor_t flav_array[] = {
3039                 RPC_AUTH_GSS_KRB5P,
3040                 RPC_AUTH_GSS_KRB5I,
3041                 RPC_AUTH_GSS_KRB5,
3042                 RPC_AUTH_UNIX,                  /* courtesy */
3043                 RPC_AUTH_NULL,
3044         };
3045         int status = -EPERM;
3046         size_t i;
3047
3048         if (server->auth_info.flavor_len > 0) {
3049                 /* try each flavor specified by user */
3050                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3051                         status = nfs4_lookup_root_sec(server, fhandle, info,
3052                                                 server->auth_info.flavors[i]);
3053                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3054                                 continue;
3055                         break;
3056                 }
3057         } else {
3058                 /* no flavors specified by user, try default list */
3059                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3060                         status = nfs4_lookup_root_sec(server, fhandle, info,
3061                                                       flav_array[i]);
3062                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3063                                 continue;
3064                         break;
3065                 }
3066         }
3067
3068         /*
3069          * -EACCESS could mean that the user doesn't have correct permissions
3070          * to access the mount.  It could also mean that we tried to mount
3071          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3072          * existing mount programs don't handle -EACCES very well so it should
3073          * be mapped to -EPERM instead.
3074          */
3075         if (status == -EACCES)
3076                 status = -EPERM;
3077         return status;
3078 }
3079
3080 static int nfs4_do_find_root_sec(struct nfs_server *server,
3081                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3082 {
3083         int mv = server->nfs_client->cl_minorversion;
3084         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3085 }
3086
3087 /**
3088  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3089  * @server: initialized nfs_server handle
3090  * @fhandle: we fill in the pseudo-fs root file handle
3091  * @info: we fill in an FSINFO struct
3092  * @auth_probe: probe the auth flavours
3093  *
3094  * Returns zero on success, or a negative errno.
3095  */
3096 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3097                          struct nfs_fsinfo *info,
3098                          bool auth_probe)
3099 {
3100         int status = 0;
3101
3102         if (!auth_probe)
3103                 status = nfs4_lookup_root(server, fhandle, info);
3104
3105         if (auth_probe || status == NFS4ERR_WRONGSEC)
3106                 status = nfs4_do_find_root_sec(server, fhandle, info);
3107
3108         if (status == 0)
3109                 status = nfs4_server_capabilities(server, fhandle);
3110         if (status == 0)
3111                 status = nfs4_do_fsinfo(server, fhandle, info);
3112
3113         return nfs4_map_errors(status);
3114 }
3115
3116 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3117                               struct nfs_fsinfo *info)
3118 {
3119         int error;
3120         struct nfs_fattr *fattr = info->fattr;
3121         struct nfs4_label *label = NULL;
3122
3123         error = nfs4_server_capabilities(server, mntfh);
3124         if (error < 0) {
3125                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3126                 return error;
3127         }
3128
3129         label = nfs4_label_alloc(server, GFP_KERNEL);
3130         if (IS_ERR(label))
3131                 return PTR_ERR(label);
3132
3133         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3134         if (error < 0) {
3135                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3136                 goto err_free_label;
3137         }
3138
3139         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3140             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3141                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3142
3143 err_free_label:
3144         nfs4_label_free(label);
3145
3146         return error;
3147 }
3148
3149 /*
3150  * Get locations and (maybe) other attributes of a referral.
3151  * Note that we'll actually follow the referral later when
3152  * we detect fsid mismatch in inode revalidation
3153  */
3154 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3155                              const struct qstr *name, struct nfs_fattr *fattr,
3156                              struct nfs_fh *fhandle)
3157 {
3158         int status = -ENOMEM;
3159         struct page *page = NULL;
3160         struct nfs4_fs_locations *locations = NULL;
3161
3162         page = alloc_page(GFP_KERNEL);
3163         if (page == NULL)
3164                 goto out;
3165         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3166         if (locations == NULL)
3167                 goto out;
3168
3169         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3170         if (status != 0)
3171                 goto out;
3172
3173         /*
3174          * If the fsid didn't change, this is a migration event, not a
3175          * referral.  Cause us to drop into the exception handler, which
3176          * will kick off migration recovery.
3177          */
3178         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3179                 dprintk("%s: server did not return a different fsid for"
3180                         " a referral at %s\n", __func__, name->name);
3181                 status = -NFS4ERR_MOVED;
3182                 goto out;
3183         }
3184         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3185         nfs_fixup_referral_attributes(&locations->fattr);
3186
3187         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3188         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3189         memset(fhandle, 0, sizeof(struct nfs_fh));
3190 out:
3191         if (page)
3192                 __free_page(page);
3193         kfree(locations);
3194         return status;
3195 }
3196
3197 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3198                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3199 {
3200         struct nfs4_getattr_arg args = {
3201                 .fh = fhandle,
3202                 .bitmask = server->attr_bitmask,
3203         };
3204         struct nfs4_getattr_res res = {
3205                 .fattr = fattr,
3206                 .label = label,
3207                 .server = server,
3208         };
3209         struct rpc_message msg = {
3210                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3211                 .rpc_argp = &args,
3212                 .rpc_resp = &res,
3213         };
3214
3215         args.bitmask = nfs4_bitmask(server, label);
3216
3217         nfs_fattr_init(fattr);
3218         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3219 }
3220
3221 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3222                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3223 {
3224         struct nfs4_exception exception = { };
3225         int err;
3226         do {
3227                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3228                 trace_nfs4_getattr(server, fhandle, fattr, err);
3229                 err = nfs4_handle_exception(server, err,
3230                                 &exception);
3231         } while (exception.retry);
3232         return err;
3233 }
3234
3235 /* 
3236  * The file is not closed if it is opened due to the a request to change
3237  * the size of the file. The open call will not be needed once the
3238  * VFS layer lookup-intents are implemented.
3239  *
3240  * Close is called when the inode is destroyed.
3241  * If we haven't opened the file for O_WRONLY, we
3242  * need to in the size_change case to obtain a stateid.
3243  *
3244  * Got race?
3245  * Because OPEN is always done by name in nfsv4, it is
3246  * possible that we opened a different file by the same
3247  * name.  We can recognize this race condition, but we
3248  * can't do anything about it besides returning an error.
3249  *
3250  * This will be fixed with VFS changes (lookup-intent).
3251  */
3252 static int
3253 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3254                   struct iattr *sattr)
3255 {
3256         struct inode *inode = d_inode(dentry);
3257         struct rpc_cred *cred = NULL;
3258         struct nfs4_state *state = NULL;
3259         struct nfs4_label *label = NULL;
3260         int status;
3261
3262         if (pnfs_ld_layoutret_on_setattr(inode) &&
3263             sattr->ia_valid & ATTR_SIZE &&
3264             sattr->ia_size < i_size_read(inode))
3265                 pnfs_commit_and_return_layout(inode);
3266
3267         nfs_fattr_init(fattr);
3268         
3269         /* Deal with open(O_TRUNC) */
3270         if (sattr->ia_valid & ATTR_OPEN)
3271                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3272
3273         /* Optimization: if the end result is no change, don't RPC */
3274         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3275                 return 0;
3276
3277         /* Search for an existing open(O_WRITE) file */
3278         if (sattr->ia_valid & ATTR_FILE) {
3279                 struct nfs_open_context *ctx;
3280
3281                 ctx = nfs_file_open_context(sattr->ia_file);
3282                 if (ctx) {
3283                         cred = ctx->cred;
3284                         state = ctx->state;
3285                 }
3286         }
3287
3288         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3289         if (IS_ERR(label))
3290                 return PTR_ERR(label);
3291
3292         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3293         if (status == 0) {
3294                 nfs_setattr_update_inode(inode, sattr, fattr);
3295                 nfs_setsecurity(inode, fattr, label);
3296         }
3297         nfs4_label_free(label);
3298         return status;
3299 }
3300
3301 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3302                 const struct qstr *name, struct nfs_fh *fhandle,
3303                 struct nfs_fattr *fattr, struct nfs4_label *label)
3304 {
3305         struct nfs_server *server = NFS_SERVER(dir);
3306         int                    status;
3307         struct nfs4_lookup_arg args = {
3308                 .bitmask = server->attr_bitmask,
3309                 .dir_fh = NFS_FH(dir),
3310                 .name = name,
3311         };
3312         struct nfs4_lookup_res res = {
3313                 .server = server,
3314                 .fattr = fattr,
3315                 .label = label,
3316                 .fh = fhandle,
3317         };
3318         struct rpc_message msg = {
3319                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3320                 .rpc_argp = &args,
3321                 .rpc_resp = &res,
3322         };
3323
3324         args.bitmask = nfs4_bitmask(server, label);
3325
3326         nfs_fattr_init(fattr);
3327
3328         dprintk("NFS call  lookup %s\n", name->name);
3329         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3330         dprintk("NFS reply lookup: %d\n", status);
3331         return status;
3332 }
3333
3334 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3335 {
3336         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3337                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3338         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3339         fattr->nlink = 2;
3340 }
3341
3342 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3343                                    struct qstr *name, struct nfs_fh *fhandle,
3344                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3345 {
3346         struct nfs4_exception exception = { };
3347         struct rpc_clnt *client = *clnt;
3348         int err;
3349         do {
3350                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3351                 trace_nfs4_lookup(dir, name, err);
3352                 switch (err) {
3353                 case -NFS4ERR_BADNAME:
3354                         err = -ENOENT;
3355                         goto out;
3356                 case -NFS4ERR_MOVED:
3357                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3358                         if (err == -NFS4ERR_MOVED)
3359                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3360                         goto out;
3361                 case -NFS4ERR_WRONGSEC:
3362                         err = -EPERM;
3363                         if (client != *clnt)
3364                                 goto out;
3365                         client = nfs4_negotiate_security(client, dir, name);
3366                         if (IS_ERR(client))
3367                                 return PTR_ERR(client);
3368
3369                         exception.retry = 1;
3370                         break;
3371                 default:
3372                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3373                 }
3374         } while (exception.retry);
3375
3376 out:
3377         if (err == 0)
3378                 *clnt = client;
3379         else if (client != *clnt)
3380                 rpc_shutdown_client(client);
3381
3382         return err;
3383 }
3384
3385 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3386                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3387                             struct nfs4_label *label)
3388 {
3389         int status;
3390         struct rpc_clnt *client = NFS_CLIENT(dir);
3391
3392         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3393         if (client != NFS_CLIENT(dir)) {
3394                 rpc_shutdown_client(client);
3395                 nfs_fixup_secinfo_attributes(fattr);
3396         }
3397         return status;
3398 }
3399
3400 struct rpc_clnt *
3401 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3402                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3403 {
3404         struct rpc_clnt *client = NFS_CLIENT(dir);
3405         int status;
3406
3407         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3408         if (status < 0)
3409                 return ERR_PTR(status);
3410         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3411 }
3412
3413 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3414 {
3415         struct nfs_server *server = NFS_SERVER(inode);
3416         struct nfs4_accessargs args = {
3417                 .fh = NFS_FH(inode),
3418                 .bitmask = server->cache_consistency_bitmask,
3419         };
3420         struct nfs4_accessres res = {
3421                 .server = server,
3422         };
3423         struct rpc_message msg = {
3424                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3425                 .rpc_argp = &args,
3426                 .rpc_resp = &res,
3427                 .rpc_cred = entry->cred,
3428         };
3429         int mode = entry->mask;
3430         int status = 0;
3431
3432         /*
3433          * Determine which access bits we want to ask for...
3434          */
3435         if (mode & MAY_READ)
3436                 args.access |= NFS4_ACCESS_READ;
3437         if (S_ISDIR(inode->i_mode)) {
3438                 if (mode & MAY_WRITE)
3439                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3440                 if (mode & MAY_EXEC)
3441                         args.access |= NFS4_ACCESS_LOOKUP;
3442         } else {
3443                 if (mode & MAY_WRITE)
3444                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3445                 if (mode & MAY_EXEC)
3446                         args.access |= NFS4_ACCESS_EXECUTE;
3447         }
3448
3449         res.fattr = nfs_alloc_fattr();
3450         if (res.fattr == NULL)
3451                 return -ENOMEM;
3452
3453         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3454         if (!status) {
3455                 nfs_access_set_mask(entry, res.access);
3456                 nfs_refresh_inode(inode, res.fattr);
3457         }
3458         nfs_free_fattr(res.fattr);
3459         return status;
3460 }
3461
3462 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3463 {
3464         struct nfs4_exception exception = { };
3465         int err;
3466         do {
3467                 err = _nfs4_proc_access(inode, entry);
3468                 trace_nfs4_access(inode, err);
3469                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3470                                 &exception);
3471         } while (exception.retry);
3472         return err;
3473 }
3474
3475 /*
3476  * TODO: For the time being, we don't try to get any attributes
3477  * along with any of the zero-copy operations READ, READDIR,
3478  * READLINK, WRITE.
3479  *
3480  * In the case of the first three, we want to put the GETATTR
3481  * after the read-type operation -- this is because it is hard
3482  * to predict the length of a GETATTR response in v4, and thus
3483  * align the READ data correctly.  This means that the GETATTR
3484  * may end up partially falling into the page cache, and we should
3485  * shift it into the 'tail' of the xdr_buf before processing.
3486  * To do this efficiently, we need to know the total length
3487  * of data received, which doesn't seem to be available outside
3488  * of the RPC layer.
3489  *
3490  * In the case of WRITE, we also want to put the GETATTR after
3491  * the operation -- in this case because we want to make sure
3492  * we get the post-operation mtime and size.
3493  *
3494  * Both of these changes to the XDR layer would in fact be quite
3495  * minor, but I decided to leave them for a subsequent patch.
3496  */
3497 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3498                 unsigned int pgbase, unsigned int pglen)
3499 {
3500         struct nfs4_readlink args = {
3501                 .fh       = NFS_FH(inode),
3502                 .pgbase   = pgbase,
3503                 .pglen    = pglen,
3504                 .pages    = &page,
3505         };
3506         struct nfs4_readlink_res res;
3507         struct rpc_message msg = {
3508                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3509                 .rpc_argp = &args,
3510                 .rpc_resp = &res,
3511         };
3512
3513         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3514 }
3515
3516 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3517                 unsigned int pgbase, unsigned int pglen)
3518 {
3519         struct nfs4_exception exception = { };
3520         int err;
3521         do {
3522                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3523                 trace_nfs4_readlink(inode, err);
3524                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3525                                 &exception);
3526         } while (exception.retry);
3527         return err;
3528 }
3529
3530 /*
3531  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3532  */
3533 static int
3534 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3535                  int flags)
3536 {
3537         struct nfs4_label l, *ilabel = NULL;
3538         struct nfs_open_context *ctx;
3539         struct nfs4_state *state;
3540         int opened = 0;
3541         int status = 0;
3542
3543         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3544         if (IS_ERR(ctx))
3545                 return PTR_ERR(ctx);
3546
3547         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3548
3549         sattr->ia_mode &= ~current_umask();
3550         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3551         if (IS_ERR(state)) {
3552                 status = PTR_ERR(state);
3553                 goto out;
3554         }
3555 out:
3556         nfs4_label_release_security(ilabel);
3557         put_nfs_open_context(ctx);
3558         return status;
3559 }
3560
3561 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3562 {
3563         struct nfs_server *server = NFS_SERVER(dir);
3564         struct nfs_removeargs args = {
3565                 .fh = NFS_FH(dir),
3566                 .name = *name,
3567         };
3568         struct nfs_removeres res = {
3569                 .server = server,
3570         };
3571         struct rpc_message msg = {
3572                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3573                 .rpc_argp = &args,
3574                 .rpc_resp = &res,
3575         };
3576         int status;
3577
3578         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3579         if (status == 0)
3580                 update_changeattr(dir, &res.cinfo);
3581         return status;
3582 }
3583
3584 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3585 {
3586         struct nfs4_exception exception = { };
3587         int err;
3588         do {
3589                 err = _nfs4_proc_remove(dir, name);
3590                 trace_nfs4_remove(dir, name, err);
3591                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3592                                 &exception);
3593         } while (exception.retry);
3594         return err;
3595 }
3596
3597 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3598 {
3599         struct nfs_server *server = NFS_SERVER(dir);
3600         struct nfs_removeargs *args = msg->rpc_argp;
3601         struct nfs_removeres *res = msg->rpc_resp;
3602
3603         res->server = server;
3604         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3605         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3606
3607         nfs_fattr_init(res->dir_attr);
3608 }
3609
3610 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3611 {
3612         nfs4_setup_sequence(NFS_SERVER(data->dir),
3613                         &data->args.seq_args,
3614                         &data->res.seq_res,
3615                         task);
3616 }
3617
3618 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3619 {
3620         struct nfs_unlinkdata *data = task->tk_calldata;
3621         struct nfs_removeres *res = &data->res;
3622
3623         if (!nfs4_sequence_done(task, &res->seq_res))
3624                 return 0;
3625         if (nfs4_async_handle_error(task, res->server, NULL,
3626                                     &data->timeout) == -EAGAIN)
3627                 return 0;
3628         update_changeattr(dir, &res->cinfo);
3629         return 1;
3630 }
3631
3632 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3633 {
3634         struct nfs_server *server = NFS_SERVER(dir);
3635         struct nfs_renameargs *arg = msg->rpc_argp;
3636         struct nfs_renameres *res = msg->rpc_resp;
3637
3638         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3639         res->server = server;
3640         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3641 }
3642
3643 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3644 {
3645         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3646                         &data->args.seq_args,
3647                         &data->res.seq_res,
3648                         task);
3649 }
3650
3651 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3652                                  struct inode *new_dir)
3653 {
3654         struct nfs_renamedata *data = task->tk_calldata;
3655         struct nfs_renameres *res = &data->res;
3656
3657         if (!nfs4_sequence_done(task, &res->seq_res))
3658                 return 0;
3659         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3660                 return 0;
3661
3662         update_changeattr(old_dir, &res->old_cinfo);
3663         update_changeattr(new_dir, &res->new_cinfo);
3664         return 1;
3665 }
3666
3667 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3668 {
3669         struct nfs_server *server = NFS_SERVER(inode);
3670         struct nfs4_link_arg arg = {
3671                 .fh     = NFS_FH(inode),
3672                 .dir_fh = NFS_FH(dir),
3673                 .name   = name,
3674                 .bitmask = server->attr_bitmask,
3675         };
3676         struct nfs4_link_res res = {
3677                 .server = server,
3678                 .label = NULL,
3679         };
3680         struct rpc_message msg = {
3681                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3682                 .rpc_argp = &arg,
3683                 .rpc_resp = &res,
3684         };
3685         int status = -ENOMEM;
3686
3687         res.fattr = nfs_alloc_fattr();
3688         if (res.fattr == NULL)
3689                 goto out;
3690
3691         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3692         if (IS_ERR(res.label)) {
3693                 status = PTR_ERR(res.label);
3694                 goto out;
3695         }
3696         arg.bitmask = nfs4_bitmask(server, res.label);
3697
3698         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3699         if (!status) {
3700                 update_changeattr(dir, &res.cinfo);
3701                 status = nfs_post_op_update_inode(inode, res.fattr);
3702                 if (!status)
3703                         nfs_setsecurity(inode, res.fattr, res.label);
3704         }
3705
3706
3707         nfs4_label_free(res.label);
3708
3709 out:
3710         nfs_free_fattr(res.fattr);
3711         return status;
3712 }
3713
3714 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3715 {
3716         struct nfs4_exception exception = { };
3717         int err;
3718         do {
3719                 err = nfs4_handle_exception(NFS_SERVER(inode),
3720                                 _nfs4_proc_link(inode, dir, name),
3721                                 &exception);
3722         } while (exception.retry);
3723         return err;
3724 }
3725
3726 struct nfs4_createdata {
3727         struct rpc_message msg;
3728         struct nfs4_create_arg arg;
3729         struct nfs4_create_res res;
3730         struct nfs_fh fh;
3731         struct nfs_fattr fattr;
3732         struct nfs4_label *label;
3733 };
3734
3735 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3736                 struct qstr *name, struct iattr *sattr, u32 ftype)
3737 {
3738         struct nfs4_createdata *data;
3739
3740         data = kzalloc(sizeof(*data), GFP_KERNEL);
3741         if (data != NULL) {
3742                 struct nfs_server *server = NFS_SERVER(dir);
3743
3744                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3745                 if (IS_ERR(data->label))
3746                         goto out_free;
3747
3748                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3749                 data->msg.rpc_argp = &data->arg;
3750                 data->msg.rpc_resp = &data->res;
3751                 data->arg.dir_fh = NFS_FH(dir);
3752                 data->arg.server = server;
3753                 data->arg.name = name;
3754                 data->arg.attrs = sattr;
3755                 data->arg.ftype = ftype;
3756                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3757                 data->res.server = server;
3758                 data->res.fh = &data->fh;
3759                 data->res.fattr = &data->fattr;
3760                 data->res.label = data->label;
3761                 nfs_fattr_init(data->res.fattr);
3762         }
3763         return data;
3764 out_free:
3765         kfree(data);
3766         return NULL;
3767 }
3768
3769 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3770 {
3771         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3772                                     &data->arg.seq_args, &data->res.seq_res, 1);
3773         if (status == 0) {
3774                 update_changeattr(dir, &data->res.dir_cinfo);
3775                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3776         }
3777         return status;
3778 }
3779
3780 static void nfs4_free_createdata(struct nfs4_createdata *data)
3781 {
3782         nfs4_label_free(data->label);
3783         kfree(data);
3784 }
3785
3786 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3787                 struct page *page, unsigned int len, struct iattr *sattr,
3788                 struct nfs4_label *label)
3789 {
3790         struct nfs4_createdata *data;
3791         int status = -ENAMETOOLONG;
3792
3793         if (len > NFS4_MAXPATHLEN)
3794                 goto out;
3795
3796         status = -ENOMEM;
3797         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3798         if (data == NULL)
3799                 goto out;
3800
3801         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3802         data->arg.u.symlink.pages = &page;
3803         data->arg.u.symlink.len = len;
3804         data->arg.label = label;
3805         
3806         status = nfs4_do_create(dir, dentry, data);
3807
3808         nfs4_free_createdata(data);
3809 out:
3810         return status;
3811 }
3812
3813 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3814                 struct page *page, unsigned int len, struct iattr *sattr)
3815 {
3816         struct nfs4_exception exception = { };
3817         struct nfs4_label l, *label = NULL;
3818         int err;
3819
3820         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3821
3822         do {
3823                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3824                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3825                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3826                                 &exception);
3827         } while (exception.retry);
3828
3829         nfs4_label_release_security(label);
3830         return err;
3831 }
3832
3833 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3834                 struct iattr *sattr, struct nfs4_label *label)
3835 {
3836         struct nfs4_createdata *data;
3837         int status = -ENOMEM;
3838
3839         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3840         if (data == NULL)
3841                 goto out;
3842
3843         data->arg.label = label;
3844         status = nfs4_do_create(dir, dentry, data);
3845
3846         nfs4_free_createdata(data);
3847 out:
3848         return status;
3849 }
3850
3851 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3852                 struct iattr *sattr)
3853 {
3854         struct nfs4_exception exception = { };
3855         struct nfs4_label l, *label = NULL;
3856         int err;
3857
3858         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3859
3860         sattr->ia_mode &= ~current_umask();
3861         do {
3862                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3863                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3864                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3865                                 &exception);
3866         } while (exception.retry);
3867         nfs4_label_release_security(label);
3868
3869         return err;
3870 }
3871
3872 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3873                 u64 cookie, struct page **pages, unsigned int count, int plus)
3874 {
3875         struct inode            *dir = d_inode(dentry);
3876         struct nfs4_readdir_arg args = {
3877                 .fh = NFS_FH(dir),
3878                 .pages = pages,
3879                 .pgbase = 0,
3880                 .count = count,
3881                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3882                 .plus = plus,
3883         };
3884         struct nfs4_readdir_res res;
3885         struct rpc_message msg = {
3886                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3887                 .rpc_argp = &args,
3888                 .rpc_resp = &res,
3889                 .rpc_cred = cred,
3890         };
3891         int                     status;
3892
3893         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3894                         dentry,
3895                         (unsigned long long)cookie);
3896         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3897         res.pgbase = args.pgbase;
3898         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3899         if (status >= 0) {
3900                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3901                 status += args.pgbase;
3902         }
3903
3904         nfs_invalidate_atime(dir);
3905
3906         dprintk("%s: returns %d\n", __func__, status);
3907         return status;
3908 }
3909
3910 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3911                 u64 cookie, struct page **pages, unsigned int count, int plus)
3912 {
3913         struct nfs4_exception exception = { };
3914         int err;
3915         do {
3916                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3917                                 pages, count, plus);
3918                 trace_nfs4_readdir(d_inode(dentry), err);
3919                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3920                                 &exception);
3921         } while (exception.retry);
3922         return err;
3923 }
3924
3925 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3926                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3927 {
3928         struct nfs4_createdata *data;
3929         int mode = sattr->ia_mode;
3930         int status = -ENOMEM;
3931
3932         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3933         if (data == NULL)
3934                 goto out;
3935
3936         if (S_ISFIFO(mode))
3937                 data->arg.ftype = NF4FIFO;
3938         else if (S_ISBLK(mode)) {
3939                 data->arg.ftype = NF4BLK;
3940                 data->arg.u.device.specdata1 = MAJOR(rdev);
3941                 data->arg.u.device.specdata2 = MINOR(rdev);
3942         }
3943         else if (S_ISCHR(mode)) {
3944                 data->arg.ftype = NF4CHR;
3945                 data->arg.u.device.specdata1 = MAJOR(rdev);
3946                 data->arg.u.device.specdata2 = MINOR(rdev);
3947         } else if (!S_ISSOCK(mode)) {
3948                 status = -EINVAL;
3949                 goto out_free;
3950         }
3951
3952         data->arg.label = label;
3953         status = nfs4_do_create(dir, dentry, data);
3954 out_free:
3955         nfs4_free_createdata(data);
3956 out:
3957         return status;
3958 }
3959
3960 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3961                 struct iattr *sattr, dev_t rdev)
3962 {
3963         struct nfs4_exception exception = { };
3964         struct nfs4_label l, *label = NULL;
3965         int err;
3966
3967         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3968
3969         sattr->ia_mode &= ~current_umask();
3970         do {
3971                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3972                 trace_nfs4_mknod(dir, &dentry->d_name, err);
3973                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3974                                 &exception);
3975         } while (exception.retry);
3976
3977         nfs4_label_release_security(label);
3978
3979         return err;
3980 }
3981
3982 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3983                  struct nfs_fsstat *fsstat)
3984 {
3985         struct nfs4_statfs_arg args = {
3986                 .fh = fhandle,
3987                 .bitmask = server->attr_bitmask,
3988         };
3989         struct nfs4_statfs_res res = {
3990                 .fsstat = fsstat,
3991         };
3992         struct rpc_message msg = {
3993                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3994                 .rpc_argp = &args,
3995                 .rpc_resp = &res,
3996         };
3997
3998         nfs_fattr_init(fsstat->fattr);
3999         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4000 }
4001
4002 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4003 {
4004         struct nfs4_exception exception = { };
4005         int err;
4006         do {
4007                 err = nfs4_handle_exception(server,
4008                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4009                                 &exception);
4010         } while (exception.retry);
4011         return err;
4012 }
4013
4014 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4015                 struct nfs_fsinfo *fsinfo)
4016 {
4017         struct nfs4_fsinfo_arg args = {
4018                 .fh = fhandle,
4019                 .bitmask = server->attr_bitmask,
4020         };
4021         struct nfs4_fsinfo_res res = {
4022                 .fsinfo = fsinfo,
4023         };
4024         struct rpc_message msg = {
4025                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4026                 .rpc_argp = &args,
4027                 .rpc_resp = &res,
4028         };
4029
4030         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4031 }
4032
4033 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4034 {
4035         struct nfs4_exception exception = { };
4036         unsigned long now = jiffies;
4037         int err;
4038
4039         do {
4040                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4041                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4042                 if (err == 0) {
4043                         struct nfs_client *clp = server->nfs_client;
4044
4045                         spin_lock(&clp->cl_lock);
4046                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4047                         clp->cl_last_renewal = now;
4048                         spin_unlock(&clp->cl_lock);
4049                         break;
4050                 }
4051                 err = nfs4_handle_exception(server, err, &exception);
4052         } while (exception.retry);
4053         return err;
4054 }
4055
4056 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4057 {
4058         int error;
4059
4060         nfs_fattr_init(fsinfo->fattr);
4061         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4062         if (error == 0) {
4063                 /* block layout checks this! */
4064                 server->pnfs_blksize = fsinfo->blksize;
4065                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4066         }
4067
4068         return error;
4069 }
4070
4071 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4072                 struct nfs_pathconf *pathconf)
4073 {
4074         struct nfs4_pathconf_arg args = {
4075                 .fh = fhandle,
4076                 .bitmask = server->attr_bitmask,
4077         };
4078         struct nfs4_pathconf_res res = {
4079                 .pathconf = pathconf,
4080         };
4081         struct rpc_message msg = {
4082                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4083                 .rpc_argp = &args,
4084                 .rpc_resp = &res,
4085         };
4086
4087         /* None of the pathconf attributes are mandatory to implement */
4088         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4089                 memset(pathconf, 0, sizeof(*pathconf));
4090                 return 0;
4091         }
4092
4093         nfs_fattr_init(pathconf->fattr);
4094         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4095 }
4096
4097 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4098                 struct nfs_pathconf *pathconf)
4099 {
4100         struct nfs4_exception exception = { };
4101         int err;
4102
4103         do {
4104                 err = nfs4_handle_exception(server,
4105                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4106                                 &exception);
4107         } while (exception.retry);
4108         return err;
4109 }
4110
4111 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4112                 const struct nfs_open_context *ctx,
4113                 const struct nfs_lock_context *l_ctx,
4114                 fmode_t fmode)
4115 {
4116         const struct nfs_lockowner *lockowner = NULL;
4117
4118         if (l_ctx != NULL)
4119                 lockowner = &l_ctx->lockowner;
4120         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4121 }
4122 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4123
4124 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4125                 const struct nfs_open_context *ctx,
4126                 const struct nfs_lock_context *l_ctx,
4127                 fmode_t fmode)
4128 {
4129         nfs4_stateid current_stateid;
4130
4131         /* If the current stateid represents a lost lock, then exit */
4132         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4133                 return true;
4134         return nfs4_stateid_match(stateid, &current_stateid);
4135 }
4136
4137 static bool nfs4_error_stateid_expired(int err)
4138 {
4139         switch (err) {
4140         case -NFS4ERR_DELEG_REVOKED:
4141         case -NFS4ERR_ADMIN_REVOKED:
4142         case -NFS4ERR_BAD_STATEID:
4143         case -NFS4ERR_STALE_STATEID:
4144         case -NFS4ERR_OLD_STATEID:
4145         case -NFS4ERR_OPENMODE:
4146         case -NFS4ERR_EXPIRED:
4147                 return true;
4148         }
4149         return false;
4150 }
4151
4152 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4153 {
4154         nfs_invalidate_atime(hdr->inode);
4155 }
4156
4157 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4158 {
4159         struct nfs_server *server = NFS_SERVER(hdr->inode);
4160
4161         trace_nfs4_read(hdr, task->tk_status);
4162         if (nfs4_async_handle_error(task, server,
4163                                     hdr->args.context->state,
4164                                     NULL) == -EAGAIN) {
4165                 rpc_restart_call_prepare(task);
4166                 return -EAGAIN;
4167         }
4168
4169         __nfs4_read_done_cb(hdr);
4170         if (task->tk_status > 0)
4171                 renew_lease(server, hdr->timestamp);
4172         return 0;
4173 }
4174
4175 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4176                 struct nfs_pgio_args *args)
4177 {
4178
4179         if (!nfs4_error_stateid_expired(task->tk_status) ||
4180                 nfs4_stateid_is_current(&args->stateid,
4181                                 args->context,
4182                                 args->lock_context,
4183                                 FMODE_READ))
4184                 return false;
4185         rpc_restart_call_prepare(task);
4186         return true;
4187 }
4188
4189 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4190 {
4191
4192         dprintk("--> %s\n", __func__);
4193
4194         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4195                 return -EAGAIN;
4196         if (nfs4_read_stateid_changed(task, &hdr->args))
4197                 return -EAGAIN;
4198         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4199                                     nfs4_read_done_cb(task, hdr);
4200 }
4201
4202 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4203                                  struct rpc_message *msg)
4204 {
4205         hdr->timestamp   = jiffies;
4206         hdr->pgio_done_cb = nfs4_read_done_cb;
4207         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4208         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4209 }
4210
4211 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4212                                       struct nfs_pgio_header *hdr)
4213 {
4214         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4215                         &hdr->args.seq_args,
4216                         &hdr->res.seq_res,
4217                         task))
4218                 return 0;
4219         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4220                                 hdr->args.lock_context,
4221                                 hdr->rw_ops->rw_mode) == -EIO)
4222                 return -EIO;
4223         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4224                 return -EIO;
4225         return 0;
4226 }
4227
4228 static int nfs4_write_done_cb(struct rpc_task *task,
4229                               struct nfs_pgio_header *hdr)
4230 {
4231         struct inode *inode = hdr->inode;
4232
4233         trace_nfs4_write(hdr, task->tk_status);
4234         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4235                                     hdr->args.context->state,
4236                                     NULL) == -EAGAIN) {
4237                 rpc_restart_call_prepare(task);
4238                 return -EAGAIN;
4239         }
4240         if (task->tk_status >= 0) {
4241                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4242                 nfs_writeback_update_inode(hdr);
4243         }
4244         return 0;
4245 }
4246
4247 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4248                 struct nfs_pgio_args *args)
4249 {
4250
4251         if (!nfs4_error_stateid_expired(task->tk_status) ||
4252                 nfs4_stateid_is_current(&args->stateid,
4253                                 args->context,
4254                                 args->lock_context,
4255                                 FMODE_WRITE))
4256                 return false;
4257         rpc_restart_call_prepare(task);
4258         return true;
4259 }
4260
4261 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4262 {
4263         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4264                 return -EAGAIN;
4265         if (nfs4_write_stateid_changed(task, &hdr->args))
4266                 return -EAGAIN;
4267         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4268                 nfs4_write_done_cb(task, hdr);
4269 }
4270
4271 static
4272 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4273 {
4274         /* Don't request attributes for pNFS or O_DIRECT writes */
4275         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4276                 return false;
4277         /* Otherwise, request attributes if and only if we don't hold
4278          * a delegation
4279          */
4280         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4281 }
4282
4283 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4284                                   struct rpc_message *msg)
4285 {
4286         struct nfs_server *server = NFS_SERVER(hdr->inode);
4287
4288         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4289                 hdr->args.bitmask = NULL;
4290                 hdr->res.fattr = NULL;
4291         } else
4292                 hdr->args.bitmask = server->cache_consistency_bitmask;
4293
4294         if (!hdr->pgio_done_cb)
4295                 hdr->pgio_done_cb = nfs4_write_done_cb;
4296         hdr->res.server = server;
4297         hdr->timestamp   = jiffies;
4298
4299         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4300         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4301 }
4302
4303 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4304 {
4305         nfs4_setup_sequence(NFS_SERVER(data->inode),
4306                         &data->args.seq_args,
4307                         &data->res.seq_res,
4308                         task);
4309 }
4310
4311 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4312 {
4313         struct inode *inode = data->inode;
4314
4315         trace_nfs4_commit(data, task->tk_status);
4316         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4317                                     NULL, NULL) == -EAGAIN) {
4318                 rpc_restart_call_prepare(task);
4319                 return -EAGAIN;
4320         }
4321         return 0;
4322 }
4323
4324 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4325 {
4326         if (!nfs4_sequence_done(task, &data->res.seq_res))
4327                 return -EAGAIN;
4328         return data->commit_done_cb(task, data);
4329 }
4330
4331 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4332 {
4333         struct nfs_server *server = NFS_SERVER(data->inode);
4334
4335         if (data->commit_done_cb == NULL)
4336                 data->commit_done_cb = nfs4_commit_done_cb;
4337         data->res.server = server;
4338         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4339         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4340 }
4341
4342 struct nfs4_renewdata {
4343         struct nfs_client       *client;
4344         unsigned long           timestamp;
4345 };
4346
4347 /*
4348  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4349  * standalone procedure for queueing an asynchronous RENEW.
4350  */
4351 static void nfs4_renew_release(void *calldata)
4352 {
4353         struct nfs4_renewdata *data = calldata;
4354         struct nfs_client *clp = data->client;
4355
4356         if (atomic_read(&clp->cl_count) > 1)
4357                 nfs4_schedule_state_renewal(clp);
4358         nfs_put_client(clp);
4359         kfree(data);
4360 }
4361
4362 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4363 {
4364         struct nfs4_renewdata *data = calldata;
4365         struct nfs_client *clp = data->client;
4366         unsigned long timestamp = data->timestamp;
4367
4368         trace_nfs4_renew_async(clp, task->tk_status);
4369         switch (task->tk_status) {
4370         case 0:
4371                 break;
4372         case -NFS4ERR_LEASE_MOVED:
4373                 nfs4_schedule_lease_moved_recovery(clp);
4374                 break;
4375         default:
4376                 /* Unless we're shutting down, schedule state recovery! */
4377                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4378                         return;
4379                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4380                         nfs4_schedule_lease_recovery(clp);
4381                         return;
4382                 }
4383                 nfs4_schedule_path_down_recovery(clp);
4384         }
4385         do_renew_lease(clp, timestamp);
4386 }
4387
4388 static const struct rpc_call_ops nfs4_renew_ops = {
4389         .rpc_call_done = nfs4_renew_done,
4390         .rpc_release = nfs4_renew_release,
4391 };
4392
4393 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4394 {
4395         struct rpc_message msg = {
4396                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4397                 .rpc_argp       = clp,
4398                 .rpc_cred       = cred,
4399         };
4400         struct nfs4_renewdata *data;
4401
4402         if (renew_flags == 0)
4403                 return 0;
4404         if (!atomic_inc_not_zero(&clp->cl_count))
4405                 return -EIO;
4406         data = kmalloc(sizeof(*data), GFP_NOFS);
4407         if (data == NULL)
4408                 return -ENOMEM;
4409         data->client = clp;
4410         data->timestamp = jiffies;
4411         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4412                         &nfs4_renew_ops, data);
4413 }
4414
4415 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4416 {
4417         struct rpc_message msg = {
4418                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4419                 .rpc_argp       = clp,
4420                 .rpc_cred       = cred,
4421         };
4422         unsigned long now = jiffies;
4423         int status;
4424
4425         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4426         if (status < 0)
4427                 return status;
4428         do_renew_lease(clp, now);
4429         return 0;
4430 }
4431
4432 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4433 {
4434         return server->caps & NFS_CAP_ACLS;
4435 }
4436
4437 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4438  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4439  * the stack.
4440  */
4441 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4442
4443 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4444                 struct page **pages, unsigned int *pgbase)
4445 {
4446         struct page *newpage, **spages;
4447         int rc = 0;
4448         size_t len;
4449         spages = pages;
4450
4451         do {
4452                 len = min_t(size_t, PAGE_SIZE, buflen);
4453                 newpage = alloc_page(GFP_KERNEL);
4454
4455                 if (newpage == NULL)
4456                         goto unwind;
4457                 memcpy(page_address(newpage), buf, len);
4458                 buf += len;
4459                 buflen -= len;
4460                 *pages++ = newpage;
4461                 rc++;
4462         } while (buflen != 0);
4463
4464         return rc;
4465
4466 unwind:
4467         for(; rc > 0; rc--)
4468                 __free_page(spages[rc-1]);
4469         return -ENOMEM;
4470 }
4471
4472 struct nfs4_cached_acl {
4473         int cached;
4474         size_t len;
4475         char data[0];
4476 };
4477
4478 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4479 {
4480         struct nfs_inode *nfsi = NFS_I(inode);
4481
4482         spin_lock(&inode->i_lock);
4483         kfree(nfsi->nfs4_acl);
4484         nfsi->nfs4_acl = acl;
4485         spin_unlock(&inode->i_lock);
4486 }
4487
4488 static void nfs4_zap_acl_attr(struct inode *inode)
4489 {
4490         nfs4_set_cached_acl(inode, NULL);
4491 }
4492
4493 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4494 {
4495         struct nfs_inode *nfsi = NFS_I(inode);
4496         struct nfs4_cached_acl *acl;
4497         int ret = -ENOENT;
4498
4499         spin_lock(&inode->i_lock);
4500         acl = nfsi->nfs4_acl;
4501         if (acl == NULL)
4502                 goto out;
4503         if (buf == NULL) /* user is just asking for length */
4504                 goto out_len;
4505         if (acl->cached == 0)
4506                 goto out;
4507         ret = -ERANGE; /* see getxattr(2) man page */
4508         if (acl->len > buflen)
4509                 goto out;
4510         memcpy(buf, acl->data, acl->len);
4511 out_len:
4512         ret = acl->len;
4513 out:
4514         spin_unlock(&inode->i_lock);
4515         return ret;
4516 }
4517
4518 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4519 {
4520         struct nfs4_cached_acl *acl;
4521         size_t buflen = sizeof(*acl) + acl_len;
4522
4523         if (buflen <= PAGE_SIZE) {
4524                 acl = kmalloc(buflen, GFP_KERNEL);
4525                 if (acl == NULL)
4526                         goto out;
4527                 acl->cached = 1;
4528                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4529         } else {
4530                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4531                 if (acl == NULL)
4532                         goto out;
4533                 acl->cached = 0;
4534         }
4535         acl->len = acl_len;
4536 out:
4537         nfs4_set_cached_acl(inode, acl);
4538 }
4539
4540 /*
4541  * The getxattr API returns the required buffer length when called with a
4542  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4543  * the required buf.  On a NULL buf, we send a page of data to the server
4544  * guessing that the ACL request can be serviced by a page. If so, we cache
4545  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4546  * the cache. If not so, we throw away the page, and cache the required
4547  * length. The next getxattr call will then produce another round trip to
4548  * the server, this time with the input buf of the required size.
4549  */
4550 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4551 {
4552         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4553         struct nfs_getaclargs args = {
4554                 .fh = NFS_FH(inode),
4555                 .acl_pages = pages,
4556                 .acl_len = buflen,
4557         };
4558         struct nfs_getaclres res = {
4559                 .acl_len = buflen,
4560         };
4561         struct rpc_message msg = {
4562                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4563                 .rpc_argp = &args,
4564                 .rpc_resp = &res,
4565         };
4566         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4567         int ret = -ENOMEM, i;
4568
4569         /* As long as we're doing a round trip to the server anyway,
4570          * let's be prepared for a page of acl data. */
4571         if (npages == 0)
4572                 npages = 1;
4573         if (npages > ARRAY_SIZE(pages))
4574                 return -ERANGE;
4575
4576         for (i = 0; i < npages; i++) {
4577                 pages[i] = alloc_page(GFP_KERNEL);
4578                 if (!pages[i])
4579                         goto out_free;
4580         }
4581
4582         /* for decoding across pages */
4583         res.acl_scratch = alloc_page(GFP_KERNEL);
4584         if (!res.acl_scratch)
4585                 goto out_free;
4586
4587         args.acl_len = npages * PAGE_SIZE;
4588         args.acl_pgbase = 0;
4589
4590         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4591                 __func__, buf, buflen, npages, args.acl_len);
4592         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4593                              &msg, &args.seq_args, &res.seq_res, 0);
4594         if (ret)
4595                 goto out_free;
4596
4597         /* Handle the case where the passed-in buffer is too short */
4598         if (res.acl_flags & NFS4_ACL_TRUNC) {
4599                 /* Did the user only issue a request for the acl length? */
4600                 if (buf == NULL)
4601                         goto out_ok;
4602                 ret = -ERANGE;
4603                 goto out_free;
4604         }
4605         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4606         if (buf) {
4607                 if (res.acl_len > buflen) {
4608                         ret = -ERANGE;
4609                         goto out_free;
4610                 }
4611                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4612         }
4613 out_ok:
4614         ret = res.acl_len;
4615 out_free:
4616         for (i = 0; i < npages; i++)
4617                 if (pages[i])
4618                         __free_page(pages[i]);
4619         if (res.acl_scratch)
4620                 __free_page(res.acl_scratch);
4621         return ret;
4622 }
4623
4624 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4625 {
4626         struct nfs4_exception exception = { };
4627         ssize_t ret;
4628         do {
4629                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4630                 trace_nfs4_get_acl(inode, ret);
4631                 if (ret >= 0)
4632                         break;
4633                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4634         } while (exception.retry);
4635         return ret;
4636 }
4637
4638 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4639 {
4640         struct nfs_server *server = NFS_SERVER(inode);
4641         int ret;
4642
4643         if (!nfs4_server_supports_acls(server))
4644                 return -EOPNOTSUPP;
4645         ret = nfs_revalidate_inode(server, inode);
4646         if (ret < 0)
4647                 return ret;
4648         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4649                 nfs_zap_acl_cache(inode);
4650         ret = nfs4_read_cached_acl(inode, buf, buflen);
4651         if (ret != -ENOENT)
4652                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4653                  * but no cached acl data, just the acl length */
4654                 return ret;
4655         return nfs4_get_acl_uncached(inode, buf, buflen);
4656 }
4657
4658 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4659 {
4660         struct nfs_server *server = NFS_SERVER(inode);
4661         struct page *pages[NFS4ACL_MAXPAGES];
4662         struct nfs_setaclargs arg = {
4663                 .fh             = NFS_FH(inode),
4664                 .acl_pages      = pages,
4665                 .acl_len        = buflen,
4666         };
4667         struct nfs_setaclres res;
4668         struct rpc_message msg = {
4669                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4670                 .rpc_argp       = &arg,
4671                 .rpc_resp       = &res,
4672         };
4673         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4674         int ret, i;
4675
4676         if (!nfs4_server_supports_acls(server))
4677                 return -EOPNOTSUPP;
4678         if (npages > ARRAY_SIZE(pages))
4679                 return -ERANGE;
4680         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4681         if (i < 0)
4682                 return i;
4683         nfs4_inode_return_delegation(inode);
4684         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4685
4686         /*
4687          * Free each page after tx, so the only ref left is
4688          * held by the network stack
4689          */
4690         for (; i > 0; i--)
4691                 put_page(pages[i-1]);
4692
4693         /*
4694          * Acl update can result in inode attribute update.
4695          * so mark the attribute cache invalid.
4696          */
4697         spin_lock(&inode->i_lock);
4698         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4699         spin_unlock(&inode->i_lock);
4700         nfs_access_zap_cache(inode);
4701         nfs_zap_acl_cache(inode);
4702         return ret;
4703 }
4704
4705 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4706 {
4707         struct nfs4_exception exception = { };
4708         int err;
4709         do {
4710                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4711                 trace_nfs4_set_acl(inode, err);
4712                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4713                                 &exception);
4714         } while (exception.retry);
4715         return err;
4716 }
4717
4718 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4719 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4720                                         size_t buflen)
4721 {
4722         struct nfs_server *server = NFS_SERVER(inode);
4723         struct nfs_fattr fattr;
4724         struct nfs4_label label = {0, 0, buflen, buf};
4725
4726         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4727         struct nfs4_getattr_arg arg = {
4728                 .fh             = NFS_FH(inode),
4729                 .bitmask        = bitmask,
4730         };
4731         struct nfs4_getattr_res res = {
4732                 .fattr          = &fattr,
4733                 .label          = &label,
4734                 .server         = server,
4735         };
4736         struct rpc_message msg = {
4737                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4738                 .rpc_argp       = &arg,
4739                 .rpc_resp       = &res,
4740         };
4741         int ret;
4742
4743         nfs_fattr_init(&fattr);
4744
4745         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4746         if (ret)
4747                 return ret;
4748         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4749                 return -ENOENT;
4750         if (buflen < label.len)
4751                 return -ERANGE;
4752         return 0;
4753 }
4754
4755 static int nfs4_get_security_label(struct inode *inode, void *buf,
4756                                         size_t buflen)
4757 {
4758         struct nfs4_exception exception = { };
4759         int err;
4760
4761         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4762                 return -EOPNOTSUPP;
4763
4764         do {
4765                 err = _nfs4_get_security_label(inode, buf, buflen);
4766                 trace_nfs4_get_security_label(inode, err);
4767                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4768                                 &exception);
4769         } while (exception.retry);
4770         return err;
4771 }
4772
4773 static int _nfs4_do_set_security_label(struct inode *inode,
4774                 struct nfs4_label *ilabel,
4775                 struct nfs_fattr *fattr,
4776                 struct nfs4_label *olabel)
4777 {
4778
4779         struct iattr sattr = {0};
4780         struct nfs_server *server = NFS_SERVER(inode);
4781         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4782         struct nfs_setattrargs arg = {
4783                 .fh             = NFS_FH(inode),
4784                 .iap            = &sattr,
4785                 .server         = server,
4786                 .bitmask        = bitmask,
4787                 .label          = ilabel,
4788         };
4789         struct nfs_setattrres res = {
4790                 .fattr          = fattr,
4791                 .label          = olabel,
4792                 .server         = server,
4793         };
4794         struct rpc_message msg = {
4795                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4796                 .rpc_argp       = &arg,
4797                 .rpc_resp       = &res,
4798         };
4799         int status;
4800
4801         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4802
4803         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4804         if (status)
4805                 dprintk("%s failed: %d\n", __func__, status);
4806
4807         return status;
4808 }
4809
4810 static int nfs4_do_set_security_label(struct inode *inode,
4811                 struct nfs4_label *ilabel,
4812                 struct nfs_fattr *fattr,
4813                 struct nfs4_label *olabel)
4814 {
4815         struct nfs4_exception exception = { };
4816         int err;
4817
4818         do {
4819                 err = _nfs4_do_set_security_label(inode, ilabel,
4820                                 fattr, olabel);
4821                 trace_nfs4_set_security_label(inode, err);
4822                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4823                                 &exception);
4824         } while (exception.retry);
4825         return err;
4826 }
4827
4828 static int
4829 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4830 {
4831         struct nfs4_label ilabel, *olabel = NULL;
4832         struct nfs_fattr fattr;
4833         struct rpc_cred *cred;
4834         struct inode *inode = d_inode(dentry);
4835         int status;
4836
4837         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4838                 return -EOPNOTSUPP;
4839
4840         nfs_fattr_init(&fattr);
4841
4842         ilabel.pi = 0;
4843         ilabel.lfs = 0;
4844         ilabel.label = (char *)buf;
4845         ilabel.len = buflen;
4846
4847         cred = rpc_lookup_cred();
4848         if (IS_ERR(cred))
4849                 return PTR_ERR(cred);
4850
4851         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4852         if (IS_ERR(olabel)) {
4853                 status = -PTR_ERR(olabel);
4854                 goto out;
4855         }
4856
4857         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4858         if (status == 0)
4859                 nfs_setsecurity(inode, &fattr, olabel);
4860
4861         nfs4_label_free(olabel);
4862 out:
4863         put_rpccred(cred);
4864         return status;
4865 }
4866 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4867
4868
4869 static int
4870 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4871                         struct nfs4_state *state, long *timeout)
4872 {
4873         struct nfs_client *clp = server->nfs_client;
4874
4875         if (task->tk_status >= 0)
4876                 return 0;
4877         switch(task->tk_status) {
4878                 case -NFS4ERR_DELEG_REVOKED:
4879                 case -NFS4ERR_ADMIN_REVOKED:
4880                 case -NFS4ERR_BAD_STATEID:
4881                 case -NFS4ERR_OPENMODE:
4882                         if (state == NULL)
4883                                 break;
4884                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4885                                 goto recovery_failed;
4886                         goto wait_on_recovery;
4887                 case -NFS4ERR_EXPIRED:
4888                         if (state != NULL) {
4889                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4890                                         goto recovery_failed;
4891                         }
4892                 case -NFS4ERR_STALE_STATEID:
4893                 case -NFS4ERR_STALE_CLIENTID:
4894                         nfs4_schedule_lease_recovery(clp);
4895                         goto wait_on_recovery;
4896                 case -NFS4ERR_MOVED:
4897                         if (nfs4_schedule_migration_recovery(server) < 0)
4898                                 goto recovery_failed;
4899                         goto wait_on_recovery;
4900                 case -NFS4ERR_LEASE_MOVED:
4901                         nfs4_schedule_lease_moved_recovery(clp);
4902                         goto wait_on_recovery;
4903 #if defined(CONFIG_NFS_V4_1)
4904                 case -NFS4ERR_BADSESSION:
4905                 case -NFS4ERR_BADSLOT:
4906                 case -NFS4ERR_BAD_HIGH_SLOT:
4907                 case -NFS4ERR_DEADSESSION:
4908                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4909                 case -NFS4ERR_SEQ_FALSE_RETRY:
4910                 case -NFS4ERR_SEQ_MISORDERED:
4911                         dprintk("%s ERROR %d, Reset session\n", __func__,
4912                                 task->tk_status);
4913                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4914                         goto wait_on_recovery;
4915 #endif /* CONFIG_NFS_V4_1 */
4916                 case -NFS4ERR_DELAY:
4917                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4918                         rpc_delay(task, nfs4_update_delay(timeout));
4919                         goto restart_call;
4920                 case -NFS4ERR_GRACE:
4921                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4922                 case -NFS4ERR_RETRY_UNCACHED_REP:
4923                 case -NFS4ERR_OLD_STATEID:
4924                         goto restart_call;
4925         }
4926         task->tk_status = nfs4_map_errors(task->tk_status);
4927         return 0;
4928 recovery_failed:
4929         task->tk_status = -EIO;
4930         return 0;
4931 wait_on_recovery:
4932         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4933         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4934                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4935         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4936                 goto recovery_failed;
4937 restart_call:
4938         task->tk_status = 0;
4939         return -EAGAIN;
4940 }
4941
4942 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4943                                     nfs4_verifier *bootverf)
4944 {
4945         __be32 verf[2];
4946
4947         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4948                 /* An impossible timestamp guarantees this value
4949                  * will never match a generated boot time. */
4950                 verf[0] = 0;
4951                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4952         } else {
4953                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4954                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4955                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4956         }
4957         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4958 }
4959
4960 static unsigned int
4961 nfs4_init_nonuniform_client_string(struct nfs_client *clp,
4962                                    char *buf, size_t len)
4963 {
4964         unsigned int result;
4965
4966         if (clp->cl_owner_id != NULL)
4967                 return strlcpy(buf, clp->cl_owner_id, len);
4968
4969         rcu_read_lock();
4970         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4971                                 clp->cl_ipaddr,
4972                                 rpc_peeraddr2str(clp->cl_rpcclient,
4973                                                         RPC_DISPLAY_ADDR),
4974                                 rpc_peeraddr2str(clp->cl_rpcclient,
4975                                                         RPC_DISPLAY_PROTO));
4976         rcu_read_unlock();
4977         clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4978         return result;
4979 }
4980
4981 static unsigned int
4982 nfs4_init_uniform_client_string(struct nfs_client *clp,
4983                                 char *buf, size_t len)
4984 {
4985         const char *nodename = clp->cl_rpcclient->cl_nodename;
4986         unsigned int result;
4987
4988         if (clp->cl_owner_id != NULL)
4989                 return strlcpy(buf, clp->cl_owner_id, len);
4990
4991         if (nfs4_client_id_uniquifier[0] != '\0')
4992                 result = scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4993                                 clp->rpc_ops->version,
4994                                 clp->cl_minorversion,
4995                                 nfs4_client_id_uniquifier,
4996                                 nodename);
4997         else
4998                 result = scnprintf(buf, len, "Linux NFSv%u.%u %s",
4999                                 clp->rpc_ops->version, clp->cl_minorversion,
5000                                 nodename);
5001         clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
5002         return result;
5003 }
5004
5005 /*
5006  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5007  * services.  Advertise one based on the address family of the
5008  * clientaddr.
5009  */
5010 static unsigned int
5011 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5012 {
5013         if (strchr(clp->cl_ipaddr, ':') != NULL)
5014                 return scnprintf(buf, len, "tcp6");
5015         else
5016                 return scnprintf(buf, len, "tcp");
5017 }
5018
5019 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5020 {
5021         struct nfs4_setclientid *sc = calldata;
5022
5023         if (task->tk_status == 0)
5024                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5025 }
5026
5027 static const struct rpc_call_ops nfs4_setclientid_ops = {
5028         .rpc_call_done = nfs4_setclientid_done,
5029 };
5030
5031 /**
5032  * nfs4_proc_setclientid - Negotiate client ID
5033  * @clp: state data structure
5034  * @program: RPC program for NFSv4 callback service
5035  * @port: IP port number for NFS4 callback service
5036  * @cred: RPC credential to use for this call
5037  * @res: where to place the result
5038  *
5039  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5040  */
5041 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5042                 unsigned short port, struct rpc_cred *cred,
5043                 struct nfs4_setclientid_res *res)
5044 {
5045         nfs4_verifier sc_verifier;
5046         struct nfs4_setclientid setclientid = {
5047                 .sc_verifier = &sc_verifier,
5048                 .sc_prog = program,
5049                 .sc_cb_ident = clp->cl_cb_ident,
5050         };
5051         struct rpc_message msg = {
5052                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5053                 .rpc_argp = &setclientid,
5054                 .rpc_resp = res,
5055                 .rpc_cred = cred,
5056         };
5057         struct rpc_task *task;
5058         struct rpc_task_setup task_setup_data = {
5059                 .rpc_client = clp->cl_rpcclient,
5060                 .rpc_message = &msg,
5061                 .callback_ops = &nfs4_setclientid_ops,
5062                 .callback_data = &setclientid,
5063                 .flags = RPC_TASK_TIMEOUT,
5064         };
5065         int status;
5066
5067         /* nfs_client_id4 */
5068         nfs4_init_boot_verifier(clp, &sc_verifier);
5069         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5070                 setclientid.sc_name_len =
5071                                 nfs4_init_uniform_client_string(clp,
5072                                                 setclientid.sc_name,
5073                                                 sizeof(setclientid.sc_name));
5074         else
5075                 setclientid.sc_name_len =
5076                                 nfs4_init_nonuniform_client_string(clp,
5077                                                 setclientid.sc_name,
5078                                                 sizeof(setclientid.sc_name));
5079         /* cb_client4 */
5080         setclientid.sc_netid_len =
5081                                 nfs4_init_callback_netid(clp,
5082                                                 setclientid.sc_netid,
5083                                                 sizeof(setclientid.sc_netid));
5084         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5085                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5086                                 clp->cl_ipaddr, port >> 8, port & 255);
5087
5088         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5089                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5090                 setclientid.sc_name_len, setclientid.sc_name);
5091         task = rpc_run_task(&task_setup_data);
5092         if (IS_ERR(task)) {
5093                 status = PTR_ERR(task);
5094                 goto out;
5095         }
5096         status = task->tk_status;
5097         if (setclientid.sc_cred) {
5098                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5099                 put_rpccred(setclientid.sc_cred);
5100         }
5101         rpc_put_task(task);
5102 out:
5103         trace_nfs4_setclientid(clp, status);
5104         dprintk("NFS reply setclientid: %d\n", status);
5105         return status;
5106 }
5107
5108 /**
5109  * nfs4_proc_setclientid_confirm - Confirm client ID
5110  * @clp: state data structure
5111  * @res: result of a previous SETCLIENTID
5112  * @cred: RPC credential to use for this call
5113  *
5114  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5115  */
5116 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5117                 struct nfs4_setclientid_res *arg,
5118                 struct rpc_cred *cred)
5119 {
5120         struct rpc_message msg = {
5121                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5122                 .rpc_argp = arg,
5123                 .rpc_cred = cred,
5124         };
5125         int status;
5126
5127         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5128                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5129                 clp->cl_clientid);
5130         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5131         trace_nfs4_setclientid_confirm(clp, status);
5132         dprintk("NFS reply setclientid_confirm: %d\n", status);
5133         return status;
5134 }
5135
5136 struct nfs4_delegreturndata {
5137         struct nfs4_delegreturnargs args;
5138         struct nfs4_delegreturnres res;
5139         struct nfs_fh fh;
5140         nfs4_stateid stateid;
5141         unsigned long timestamp;
5142         struct nfs_fattr fattr;
5143         int rpc_status;
5144         struct inode *inode;
5145         bool roc;
5146         u32 roc_barrier;
5147 };
5148
5149 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5150 {
5151         struct nfs4_delegreturndata *data = calldata;
5152
5153         if (!nfs4_sequence_done(task, &data->res.seq_res))
5154                 return;
5155
5156         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5157         switch (task->tk_status) {
5158         case 0:
5159                 renew_lease(data->res.server, data->timestamp);
5160         case -NFS4ERR_ADMIN_REVOKED:
5161         case -NFS4ERR_DELEG_REVOKED:
5162         case -NFS4ERR_BAD_STATEID:
5163         case -NFS4ERR_OLD_STATEID:
5164         case -NFS4ERR_STALE_STATEID:
5165         case -NFS4ERR_EXPIRED:
5166                 task->tk_status = 0;
5167                 if (data->roc)
5168                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5169                 break;
5170         default:
5171                 if (nfs4_async_handle_error(task, data->res.server,
5172                                             NULL, NULL) == -EAGAIN) {
5173                         rpc_restart_call_prepare(task);
5174                         return;
5175                 }
5176         }
5177         data->rpc_status = task->tk_status;
5178 }
5179
5180 static void nfs4_delegreturn_release(void *calldata)
5181 {
5182         struct nfs4_delegreturndata *data = calldata;
5183         struct inode *inode = data->inode;
5184
5185         if (inode) {
5186                 if (data->roc)
5187                         pnfs_roc_release(inode);
5188                 nfs_iput_and_deactive(inode);
5189         }
5190         kfree(calldata);
5191 }
5192
5193 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5194 {
5195         struct nfs4_delegreturndata *d_data;
5196
5197         d_data = (struct nfs4_delegreturndata *)data;
5198
5199         if (d_data->roc &&
5200             pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5201                 return;
5202
5203         nfs4_setup_sequence(d_data->res.server,
5204                         &d_data->args.seq_args,
5205                         &d_data->res.seq_res,
5206                         task);
5207 }
5208
5209 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5210         .rpc_call_prepare = nfs4_delegreturn_prepare,
5211         .rpc_call_done = nfs4_delegreturn_done,
5212         .rpc_release = nfs4_delegreturn_release,
5213 };
5214
5215 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5216 {
5217         struct nfs4_delegreturndata *data;
5218         struct nfs_server *server = NFS_SERVER(inode);
5219         struct rpc_task *task;
5220         struct rpc_message msg = {
5221                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5222                 .rpc_cred = cred,
5223         };
5224         struct rpc_task_setup task_setup_data = {
5225                 .rpc_client = server->client,
5226                 .rpc_message = &msg,
5227                 .callback_ops = &nfs4_delegreturn_ops,
5228                 .flags = RPC_TASK_ASYNC,
5229         };
5230         int status = 0;
5231
5232         data = kzalloc(sizeof(*data), GFP_NOFS);
5233         if (data == NULL)
5234                 return -ENOMEM;
5235         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5236         data->args.fhandle = &data->fh;
5237         data->args.stateid = &data->stateid;
5238         data->args.bitmask = server->cache_consistency_bitmask;
5239         nfs_copy_fh(&data->fh, NFS_FH(inode));
5240         nfs4_stateid_copy(&data->stateid, stateid);
5241         data->res.fattr = &data->fattr;
5242         data->res.server = server;
5243         nfs_fattr_init(data->res.fattr);
5244         data->timestamp = jiffies;
5245         data->rpc_status = 0;
5246         data->inode = nfs_igrab_and_active(inode);
5247         if (data->inode)
5248                 data->roc = nfs4_roc(inode);
5249
5250         task_setup_data.callback_data = data;
5251         msg.rpc_argp = &data->args;
5252         msg.rpc_resp = &data->res;
5253         task = rpc_run_task(&task_setup_data);
5254         if (IS_ERR(task))
5255                 return PTR_ERR(task);
5256         if (!issync)
5257                 goto out;
5258         status = nfs4_wait_for_completion_rpc_task(task);
5259         if (status != 0)
5260                 goto out;
5261         status = data->rpc_status;
5262         if (status == 0)
5263                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5264         else
5265                 nfs_refresh_inode(inode, &data->fattr);
5266 out:
5267         rpc_put_task(task);
5268         return status;
5269 }
5270
5271 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5272 {
5273         struct nfs_server *server = NFS_SERVER(inode);
5274         struct nfs4_exception exception = { };
5275         int err;
5276         do {
5277                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5278                 trace_nfs4_delegreturn(inode, err);
5279                 switch (err) {
5280                         case -NFS4ERR_STALE_STATEID:
5281                         case -NFS4ERR_EXPIRED:
5282                         case 0:
5283                                 return 0;
5284                 }
5285                 err = nfs4_handle_exception(server, err, &exception);
5286         } while (exception.retry);
5287         return err;
5288 }
5289
5290 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5291 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5292
5293 /* 
5294  * sleep, with exponential backoff, and retry the LOCK operation. 
5295  */
5296 static unsigned long
5297 nfs4_set_lock_task_retry(unsigned long timeout)
5298 {
5299         freezable_schedule_timeout_killable_unsafe(timeout);
5300         timeout <<= 1;
5301         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5302                 return NFS4_LOCK_MAXTIMEOUT;
5303         return timeout;
5304 }
5305
5306 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5307 {
5308         struct inode *inode = state->inode;
5309         struct nfs_server *server = NFS_SERVER(inode);
5310         struct nfs_client *clp = server->nfs_client;
5311         struct nfs_lockt_args arg = {
5312                 .fh = NFS_FH(inode),
5313                 .fl = request,
5314         };
5315         struct nfs_lockt_res res = {
5316                 .denied = request,
5317         };
5318         struct rpc_message msg = {
5319                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5320                 .rpc_argp       = &arg,
5321                 .rpc_resp       = &res,
5322                 .rpc_cred       = state->owner->so_cred,
5323         };
5324         struct nfs4_lock_state *lsp;
5325         int status;
5326
5327         arg.lock_owner.clientid = clp->cl_clientid;
5328         status = nfs4_set_lock_state(state, request);
5329         if (status != 0)
5330                 goto out;
5331         lsp = request->fl_u.nfs4_fl.owner;
5332         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5333         arg.lock_owner.s_dev = server->s_dev;
5334         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5335         switch (status) {
5336                 case 0:
5337                         request->fl_type = F_UNLCK;
5338                         break;
5339                 case -NFS4ERR_DENIED:
5340                         status = 0;
5341         }
5342         request->fl_ops->fl_release_private(request);
5343         request->fl_ops = NULL;
5344 out:
5345         return status;
5346 }
5347
5348 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5349 {
5350         struct nfs4_exception exception = { };
5351         int err;
5352
5353         do {
5354                 err = _nfs4_proc_getlk(state, cmd, request);
5355                 trace_nfs4_get_lock(request, state, cmd, err);
5356                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5357                                 &exception);
5358         } while (exception.retry);
5359         return err;
5360 }
5361
5362 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5363 {
5364         int res = 0;
5365         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5366                 case FL_POSIX:
5367                         res = posix_lock_file_wait(file, fl);
5368                         break;
5369                 case FL_FLOCK:
5370                         res = flock_lock_file_wait(file, fl);
5371                         break;
5372                 default:
5373                         BUG();
5374         }
5375         return res;
5376 }
5377
5378 struct nfs4_unlockdata {
5379         struct nfs_locku_args arg;
5380         struct nfs_locku_res res;
5381         struct nfs4_lock_state *lsp;
5382         struct nfs_open_context *ctx;
5383         struct file_lock fl;
5384         const struct nfs_server *server;
5385         unsigned long timestamp;
5386 };
5387
5388 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5389                 struct nfs_open_context *ctx,
5390                 struct nfs4_lock_state *lsp,
5391                 struct nfs_seqid *seqid)
5392 {
5393         struct nfs4_unlockdata *p;
5394         struct inode *inode = lsp->ls_state->inode;
5395
5396         p = kzalloc(sizeof(*p), GFP_NOFS);
5397         if (p == NULL)
5398                 return NULL;
5399         p->arg.fh = NFS_FH(inode);
5400         p->arg.fl = &p->fl;
5401         p->arg.seqid = seqid;
5402         p->res.seqid = seqid;
5403         p->lsp = lsp;
5404         atomic_inc(&lsp->ls_count);
5405         /* Ensure we don't close file until we're done freeing locks! */
5406         p->ctx = get_nfs_open_context(ctx);
5407         memcpy(&p->fl, fl, sizeof(p->fl));
5408         p->server = NFS_SERVER(inode);
5409         return p;
5410 }
5411
5412 static void nfs4_locku_release_calldata(void *data)
5413 {
5414         struct nfs4_unlockdata *calldata = data;
5415         nfs_free_seqid(calldata->arg.seqid);
5416         nfs4_put_lock_state(calldata->lsp);
5417         put_nfs_open_context(calldata->ctx);
5418         kfree(calldata);
5419 }
5420
5421 static void nfs4_locku_done(struct rpc_task *task, void *data)
5422 {
5423         struct nfs4_unlockdata *calldata = data;
5424
5425         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5426                 return;
5427         switch (task->tk_status) {
5428                 case 0:
5429                         renew_lease(calldata->server, calldata->timestamp);
5430                         do_vfs_lock(calldata->fl.fl_file, &calldata->fl);
5431                         if (nfs4_update_lock_stateid(calldata->lsp,
5432                                         &calldata->res.stateid))
5433                                 break;
5434                 case -NFS4ERR_BAD_STATEID:
5435                 case -NFS4ERR_OLD_STATEID:
5436                 case -NFS4ERR_STALE_STATEID:
5437                 case -NFS4ERR_EXPIRED:
5438                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5439                                                 &calldata->lsp->ls_stateid))
5440                                 rpc_restart_call_prepare(task);
5441                         break;
5442                 default:
5443                         if (nfs4_async_handle_error(task, calldata->server,
5444                                                     NULL, NULL) == -EAGAIN)
5445                                 rpc_restart_call_prepare(task);
5446         }
5447         nfs_release_seqid(calldata->arg.seqid);
5448 }
5449
5450 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5451 {
5452         struct nfs4_unlockdata *calldata = data;
5453
5454         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5455                 goto out_wait;
5456         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5457         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5458                 /* Note: exit _without_ running nfs4_locku_done */
5459                 goto out_no_action;
5460         }
5461         calldata->timestamp = jiffies;
5462         if (nfs4_setup_sequence(calldata->server,
5463                                 &calldata->arg.seq_args,
5464                                 &calldata->res.seq_res,
5465                                 task) != 0)
5466                 nfs_release_seqid(calldata->arg.seqid);
5467         return;
5468 out_no_action:
5469         task->tk_action = NULL;
5470 out_wait:
5471         nfs4_sequence_done(task, &calldata->res.seq_res);
5472 }
5473
5474 static const struct rpc_call_ops nfs4_locku_ops = {
5475         .rpc_call_prepare = nfs4_locku_prepare,
5476         .rpc_call_done = nfs4_locku_done,
5477         .rpc_release = nfs4_locku_release_calldata,
5478 };
5479
5480 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5481                 struct nfs_open_context *ctx,
5482                 struct nfs4_lock_state *lsp,
5483                 struct nfs_seqid *seqid)
5484 {
5485         struct nfs4_unlockdata *data;
5486         struct rpc_message msg = {
5487                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5488                 .rpc_cred = ctx->cred,
5489         };
5490         struct rpc_task_setup task_setup_data = {
5491                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5492                 .rpc_message = &msg,
5493                 .callback_ops = &nfs4_locku_ops,
5494                 .workqueue = nfsiod_workqueue,
5495                 .flags = RPC_TASK_ASYNC,
5496         };
5497
5498         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5499                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5500
5501         /* Ensure this is an unlock - when canceling a lock, the
5502          * canceled lock is passed in, and it won't be an unlock.
5503          */
5504         fl->fl_type = F_UNLCK;
5505
5506         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5507         if (data == NULL) {
5508                 nfs_free_seqid(seqid);
5509                 return ERR_PTR(-ENOMEM);
5510         }
5511
5512         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5513         msg.rpc_argp = &data->arg;
5514         msg.rpc_resp = &data->res;
5515         task_setup_data.callback_data = data;
5516         return rpc_run_task(&task_setup_data);
5517 }
5518
5519 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5520 {
5521         struct inode *inode = state->inode;
5522         struct nfs4_state_owner *sp = state->owner;
5523         struct nfs_inode *nfsi = NFS_I(inode);
5524         struct nfs_seqid *seqid;
5525         struct nfs4_lock_state *lsp;
5526         struct rpc_task *task;
5527         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5528         int status = 0;
5529         unsigned char fl_flags = request->fl_flags;
5530
5531         status = nfs4_set_lock_state(state, request);
5532         /* Unlock _before_ we do the RPC call */
5533         request->fl_flags |= FL_EXISTS;
5534         /* Exclude nfs_delegation_claim_locks() */
5535         mutex_lock(&sp->so_delegreturn_mutex);
5536         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5537         down_read(&nfsi->rwsem);
5538         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5539                 up_read(&nfsi->rwsem);
5540                 mutex_unlock(&sp->so_delegreturn_mutex);
5541                 goto out;
5542         }
5543         up_read(&nfsi->rwsem);
5544         mutex_unlock(&sp->so_delegreturn_mutex);
5545         if (status != 0)
5546                 goto out;
5547         /* Is this a delegated lock? */
5548         lsp = request->fl_u.nfs4_fl.owner;
5549         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5550                 goto out;
5551         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5552         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5553         status = -ENOMEM;
5554         if (IS_ERR(seqid))
5555                 goto out;
5556         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5557         status = PTR_ERR(task);
5558         if (IS_ERR(task))
5559                 goto out;
5560         status = nfs4_wait_for_completion_rpc_task(task);
5561         rpc_put_task(task);
5562 out:
5563         request->fl_flags = fl_flags;
5564         trace_nfs4_unlock(request, state, F_SETLK, status);
5565         return status;
5566 }
5567
5568 struct nfs4_lockdata {
5569         struct nfs_lock_args arg;
5570         struct nfs_lock_res res;
5571         struct nfs4_lock_state *lsp;
5572         struct nfs_open_context *ctx;
5573         struct file_lock fl;
5574         unsigned long timestamp;
5575         int rpc_status;
5576         int cancelled;
5577         struct nfs_server *server;
5578 };
5579
5580 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5581                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5582                 gfp_t gfp_mask)
5583 {
5584         struct nfs4_lockdata *p;
5585         struct inode *inode = lsp->ls_state->inode;
5586         struct nfs_server *server = NFS_SERVER(inode);
5587         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5588
5589         p = kzalloc(sizeof(*p), gfp_mask);
5590         if (p == NULL)
5591                 return NULL;
5592
5593         p->arg.fh = NFS_FH(inode);
5594         p->arg.fl = &p->fl;
5595         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5596         if (IS_ERR(p->arg.open_seqid))
5597                 goto out_free;
5598         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5599         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5600         if (IS_ERR(p->arg.lock_seqid))
5601                 goto out_free_seqid;
5602         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5603         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5604         p->arg.lock_owner.s_dev = server->s_dev;
5605         p->res.lock_seqid = p->arg.lock_seqid;
5606         p->lsp = lsp;
5607         p->server = server;
5608         atomic_inc(&lsp->ls_count);
5609         p->ctx = get_nfs_open_context(ctx);
5610         get_file(fl->fl_file);
5611         memcpy(&p->fl, fl, sizeof(p->fl));
5612         return p;
5613 out_free_seqid:
5614         nfs_free_seqid(p->arg.open_seqid);
5615 out_free:
5616         kfree(p);
5617         return NULL;
5618 }
5619
5620 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5621 {
5622         struct nfs4_lockdata *data = calldata;
5623         struct nfs4_state *state = data->lsp->ls_state;
5624
5625         dprintk("%s: begin!\n", __func__);
5626         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5627                 goto out_wait;
5628         /* Do we need to do an open_to_lock_owner? */
5629         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5630                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5631                         goto out_release_lock_seqid;
5632                 }
5633                 nfs4_stateid_copy(&data->arg.open_stateid,
5634                                 &state->open_stateid);
5635                 data->arg.new_lock_owner = 1;
5636                 data->res.open_seqid = data->arg.open_seqid;
5637         } else {
5638                 data->arg.new_lock_owner = 0;
5639                 nfs4_stateid_copy(&data->arg.lock_stateid,
5640                                 &data->lsp->ls_stateid);
5641         }
5642         if (!nfs4_valid_open_stateid(state)) {
5643                 data->rpc_status = -EBADF;
5644                 task->tk_action = NULL;
5645                 goto out_release_open_seqid;
5646         }
5647         data->timestamp = jiffies;
5648         if (nfs4_setup_sequence(data->server,
5649                                 &data->arg.seq_args,
5650                                 &data->res.seq_res,
5651                                 task) == 0)
5652                 return;
5653 out_release_open_seqid:
5654         nfs_release_seqid(data->arg.open_seqid);
5655 out_release_lock_seqid:
5656         nfs_release_seqid(data->arg.lock_seqid);
5657 out_wait:
5658         nfs4_sequence_done(task, &data->res.seq_res);
5659         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5660 }
5661
5662 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5663 {
5664         struct nfs4_lockdata *data = calldata;
5665         struct nfs4_lock_state *lsp = data->lsp;
5666
5667         dprintk("%s: begin!\n", __func__);
5668
5669         if (!nfs4_sequence_done(task, &data->res.seq_res))
5670                 return;
5671
5672         data->rpc_status = task->tk_status;
5673         switch (task->tk_status) {
5674         case 0:
5675                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5676                                 data->timestamp);
5677                 if (data->arg.new_lock) {
5678                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5679                         if (do_vfs_lock(data->fl.fl_file, &data->fl) < 0) {
5680                                 rpc_restart_call_prepare(task);
5681                                 break;
5682                         }
5683                 }
5684                 if (data->arg.new_lock_owner != 0) {
5685                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5686                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5687                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5688                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5689                         rpc_restart_call_prepare(task);
5690                 break;
5691         case -NFS4ERR_BAD_STATEID:
5692         case -NFS4ERR_OLD_STATEID:
5693         case -NFS4ERR_STALE_STATEID:
5694         case -NFS4ERR_EXPIRED:
5695                 if (data->arg.new_lock_owner != 0) {
5696                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5697                                                 &lsp->ls_state->open_stateid))
5698                                 rpc_restart_call_prepare(task);
5699                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5700                                                 &lsp->ls_stateid))
5701                                 rpc_restart_call_prepare(task);
5702         }
5703         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5704 }
5705
5706 static void nfs4_lock_release(void *calldata)
5707 {
5708         struct nfs4_lockdata *data = calldata;
5709
5710         dprintk("%s: begin!\n", __func__);
5711         nfs_free_seqid(data->arg.open_seqid);
5712         if (data->cancelled != 0) {
5713                 struct rpc_task *task;
5714                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5715                                 data->arg.lock_seqid);
5716                 if (!IS_ERR(task))
5717                         rpc_put_task_async(task);
5718                 dprintk("%s: cancelling lock!\n", __func__);
5719         } else
5720                 nfs_free_seqid(data->arg.lock_seqid);
5721         nfs4_put_lock_state(data->lsp);
5722         put_nfs_open_context(data->ctx);
5723         fput(data->fl.fl_file);
5724         kfree(data);
5725         dprintk("%s: done!\n", __func__);
5726 }
5727
5728 static const struct rpc_call_ops nfs4_lock_ops = {
5729         .rpc_call_prepare = nfs4_lock_prepare,
5730         .rpc_call_done = nfs4_lock_done,
5731         .rpc_release = nfs4_lock_release,
5732 };
5733
5734 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5735 {
5736         switch (error) {
5737         case -NFS4ERR_ADMIN_REVOKED:
5738         case -NFS4ERR_BAD_STATEID:
5739                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5740                 if (new_lock_owner != 0 ||
5741                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5742                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5743                 break;
5744         case -NFS4ERR_STALE_STATEID:
5745                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5746         case -NFS4ERR_EXPIRED:
5747                 nfs4_schedule_lease_recovery(server->nfs_client);
5748         };
5749 }
5750
5751 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5752 {
5753         struct nfs4_lockdata *data;
5754         struct rpc_task *task;
5755         struct rpc_message msg = {
5756                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5757                 .rpc_cred = state->owner->so_cred,
5758         };
5759         struct rpc_task_setup task_setup_data = {
5760                 .rpc_client = NFS_CLIENT(state->inode),
5761                 .rpc_message = &msg,
5762                 .callback_ops = &nfs4_lock_ops,
5763                 .workqueue = nfsiod_workqueue,
5764                 .flags = RPC_TASK_ASYNC,
5765         };
5766         int ret;
5767
5768         dprintk("%s: begin!\n", __func__);
5769         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5770                         fl->fl_u.nfs4_fl.owner,
5771                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5772         if (data == NULL)
5773                 return -ENOMEM;
5774         if (IS_SETLKW(cmd))
5775                 data->arg.block = 1;
5776         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5777         msg.rpc_argp = &data->arg;
5778         msg.rpc_resp = &data->res;
5779         task_setup_data.callback_data = data;
5780         if (recovery_type > NFS_LOCK_NEW) {
5781                 if (recovery_type == NFS_LOCK_RECLAIM)
5782                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5783                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5784         } else
5785                 data->arg.new_lock = 1;
5786         task = rpc_run_task(&task_setup_data);
5787         if (IS_ERR(task))
5788                 return PTR_ERR(task);
5789         ret = nfs4_wait_for_completion_rpc_task(task);
5790         if (ret == 0) {
5791                 ret = data->rpc_status;
5792                 if (ret)
5793                         nfs4_handle_setlk_error(data->server, data->lsp,
5794                                         data->arg.new_lock_owner, ret);
5795         } else
5796                 data->cancelled = 1;
5797         rpc_put_task(task);
5798         dprintk("%s: done, ret = %d!\n", __func__, ret);
5799         return ret;
5800 }
5801
5802 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5803 {
5804         struct nfs_server *server = NFS_SERVER(state->inode);
5805         struct nfs4_exception exception = {
5806                 .inode = state->inode,
5807         };
5808         int err;
5809
5810         do {
5811                 /* Cache the lock if possible... */
5812                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5813                         return 0;
5814                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5815                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5816                 if (err != -NFS4ERR_DELAY)
5817                         break;
5818                 nfs4_handle_exception(server, err, &exception);
5819         } while (exception.retry);
5820         return err;
5821 }
5822
5823 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5824 {
5825         struct nfs_server *server = NFS_SERVER(state->inode);
5826         struct nfs4_exception exception = {
5827                 .inode = state->inode,
5828         };
5829         int err;
5830
5831         err = nfs4_set_lock_state(state, request);
5832         if (err != 0)
5833                 return err;
5834         if (!recover_lost_locks) {
5835                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5836                 return 0;
5837         }
5838         do {
5839                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5840                         return 0;
5841                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5842                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5843                 switch (err) {
5844                 default:
5845                         goto out;
5846                 case -NFS4ERR_GRACE:
5847                 case -NFS4ERR_DELAY:
5848                         nfs4_handle_exception(server, err, &exception);
5849                         err = 0;
5850                 }
5851         } while (exception.retry);
5852 out:
5853         return err;
5854 }
5855
5856 #if defined(CONFIG_NFS_V4_1)
5857 /**
5858  * nfs41_check_expired_locks - possibly free a lock stateid
5859  *
5860  * @state: NFSv4 state for an inode
5861  *
5862  * Returns NFS_OK if recovery for this stateid is now finished.
5863  * Otherwise a negative NFS4ERR value is returned.
5864  */
5865 static int nfs41_check_expired_locks(struct nfs4_state *state)
5866 {
5867         int status, ret = -NFS4ERR_BAD_STATEID;
5868         struct nfs4_lock_state *lsp;
5869         struct nfs_server *server = NFS_SERVER(state->inode);
5870
5871         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5872                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5873                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5874
5875                         status = nfs41_test_stateid(server,
5876                                         &lsp->ls_stateid,
5877                                         cred);
5878                         trace_nfs4_test_lock_stateid(state, lsp, status);
5879                         if (status != NFS_OK) {
5880                                 /* Free the stateid unless the server
5881                                  * informs us the stateid is unrecognized. */
5882                                 if (status != -NFS4ERR_BAD_STATEID)
5883                                         nfs41_free_stateid(server,
5884                                                         &lsp->ls_stateid,
5885                                                         cred);
5886                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5887                                 ret = status;
5888                         }
5889                 }
5890         };
5891
5892         return ret;
5893 }
5894
5895 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5896 {
5897         int status = NFS_OK;
5898
5899         if (test_bit(LK_STATE_IN_USE, &state->flags))
5900                 status = nfs41_check_expired_locks(state);
5901         if (status != NFS_OK)
5902                 status = nfs4_lock_expired(state, request);
5903         return status;
5904 }
5905 #endif
5906
5907 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5908 {
5909         struct nfs_inode *nfsi = NFS_I(state->inode);
5910         unsigned char fl_flags = request->fl_flags;
5911         int status = -ENOLCK;
5912
5913         if ((fl_flags & FL_POSIX) &&
5914                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5915                 goto out;
5916         /* Is this a delegated open? */
5917         status = nfs4_set_lock_state(state, request);
5918         if (status != 0)
5919                 goto out;
5920         request->fl_flags |= FL_ACCESS;
5921         status = do_vfs_lock(request->fl_file, request);
5922         if (status < 0)
5923                 goto out;
5924         down_read(&nfsi->rwsem);
5925         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5926                 /* Yes: cache locks! */
5927                 /* ...but avoid races with delegation recall... */
5928                 request->fl_flags = fl_flags & ~FL_SLEEP;
5929                 status = do_vfs_lock(request->fl_file, request);
5930                 up_read(&nfsi->rwsem);
5931                 goto out;
5932         }
5933         up_read(&nfsi->rwsem);
5934         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5935 out:
5936         request->fl_flags = fl_flags;
5937         return status;
5938 }
5939
5940 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5941 {
5942         struct nfs4_exception exception = {
5943                 .state = state,
5944                 .inode = state->inode,
5945         };
5946         int err;
5947
5948         do {
5949                 err = _nfs4_proc_setlk(state, cmd, request);
5950                 trace_nfs4_set_lock(request, state, cmd, err);
5951                 if (err == -NFS4ERR_DENIED)
5952                         err = -EAGAIN;
5953                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5954                                 err, &exception);
5955         } while (exception.retry);
5956         return err;
5957 }
5958
5959 static int
5960 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5961 {
5962         struct nfs_open_context *ctx;
5963         struct nfs4_state *state;
5964         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5965         int status;
5966
5967         /* verify open state */
5968         ctx = nfs_file_open_context(filp);
5969         state = ctx->state;
5970
5971         if (request->fl_start < 0 || request->fl_end < 0)
5972                 return -EINVAL;
5973
5974         if (IS_GETLK(cmd)) {
5975                 if (state != NULL)
5976                         return nfs4_proc_getlk(state, F_GETLK, request);
5977                 return 0;
5978         }
5979
5980         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5981                 return -EINVAL;
5982
5983         if (request->fl_type == F_UNLCK) {
5984                 if (state != NULL)
5985                         return nfs4_proc_unlck(state, cmd, request);
5986                 return 0;
5987         }
5988
5989         if (state == NULL)
5990                 return -ENOLCK;
5991         /*
5992          * Don't rely on the VFS having checked the file open mode,
5993          * since it won't do this for flock() locks.
5994          */
5995         switch (request->fl_type) {
5996         case F_RDLCK:
5997                 if (!(filp->f_mode & FMODE_READ))
5998                         return -EBADF;
5999                 break;
6000         case F_WRLCK:
6001                 if (!(filp->f_mode & FMODE_WRITE))
6002                         return -EBADF;
6003         }
6004
6005         do {
6006                 status = nfs4_proc_setlk(state, cmd, request);
6007                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6008                         break;
6009                 timeout = nfs4_set_lock_task_retry(timeout);
6010                 status = -ERESTARTSYS;
6011                 if (signalled())
6012                         break;
6013         } while(status < 0);
6014         return status;
6015 }
6016
6017 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6018 {
6019         struct nfs_server *server = NFS_SERVER(state->inode);
6020         int err;
6021
6022         err = nfs4_set_lock_state(state, fl);
6023         if (err != 0)
6024                 return err;
6025         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6026         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6027 }
6028
6029 struct nfs_release_lockowner_data {
6030         struct nfs4_lock_state *lsp;
6031         struct nfs_server *server;
6032         struct nfs_release_lockowner_args args;
6033         struct nfs_release_lockowner_res res;
6034         unsigned long timestamp;
6035 };
6036
6037 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6038 {
6039         struct nfs_release_lockowner_data *data = calldata;
6040         struct nfs_server *server = data->server;
6041         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6042                              &data->args.seq_args, &data->res.seq_res, task);
6043         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6044         data->timestamp = jiffies;
6045 }
6046
6047 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6048 {
6049         struct nfs_release_lockowner_data *data = calldata;
6050         struct nfs_server *server = data->server;
6051
6052         nfs40_sequence_done(task, &data->res.seq_res);
6053
6054         switch (task->tk_status) {
6055         case 0:
6056                 renew_lease(server, data->timestamp);
6057                 break;
6058         case -NFS4ERR_STALE_CLIENTID:
6059         case -NFS4ERR_EXPIRED:
6060                 nfs4_schedule_lease_recovery(server->nfs_client);
6061                 break;
6062         case -NFS4ERR_LEASE_MOVED:
6063         case -NFS4ERR_DELAY:
6064                 if (nfs4_async_handle_error(task, server,
6065                                             NULL, NULL) == -EAGAIN)
6066                         rpc_restart_call_prepare(task);
6067         }
6068 }
6069
6070 static void nfs4_release_lockowner_release(void *calldata)
6071 {
6072         struct nfs_release_lockowner_data *data = calldata;
6073         nfs4_free_lock_state(data->server, data->lsp);
6074         kfree(calldata);
6075 }
6076
6077 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6078         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6079         .rpc_call_done = nfs4_release_lockowner_done,
6080         .rpc_release = nfs4_release_lockowner_release,
6081 };
6082
6083 static void
6084 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6085 {
6086         struct nfs_release_lockowner_data *data;
6087         struct rpc_message msg = {
6088                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6089         };
6090
6091         if (server->nfs_client->cl_mvops->minor_version != 0)
6092                 return;
6093
6094         data = kmalloc(sizeof(*data), GFP_NOFS);
6095         if (!data)
6096                 return;
6097         data->lsp = lsp;
6098         data->server = server;
6099         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6100         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6101         data->args.lock_owner.s_dev = server->s_dev;
6102
6103         msg.rpc_argp = &data->args;
6104         msg.rpc_resp = &data->res;
6105         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6106         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6107 }
6108
6109 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6110
6111 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6112                                    const void *buf, size_t buflen,
6113                                    int flags, int type)
6114 {
6115         if (strcmp(key, "") != 0)
6116                 return -EINVAL;
6117
6118         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6119 }
6120
6121 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6122                                    void *buf, size_t buflen, int type)
6123 {
6124         if (strcmp(key, "") != 0)
6125                 return -EINVAL;
6126
6127         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6128 }
6129
6130 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6131                                        size_t list_len, const char *name,
6132                                        size_t name_len, int type)
6133 {
6134         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6135
6136         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6137                 return 0;
6138
6139         if (list && len <= list_len)
6140                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6141         return len;
6142 }
6143
6144 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6145 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6146 {
6147         return server->caps & NFS_CAP_SECURITY_LABEL;
6148 }
6149
6150 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6151                                    const void *buf, size_t buflen,
6152                                    int flags, int type)
6153 {
6154         if (security_ismaclabel(key))
6155                 return nfs4_set_security_label(dentry, buf, buflen);
6156
6157         return -EOPNOTSUPP;
6158 }
6159
6160 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6161                                    void *buf, size_t buflen, int type)
6162 {
6163         if (security_ismaclabel(key))
6164                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6165         return -EOPNOTSUPP;
6166 }
6167
6168 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6169                                        size_t list_len, const char *name,
6170                                        size_t name_len, int type)
6171 {
6172         size_t len = 0;
6173
6174         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6175                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6176                 if (list && len <= list_len)
6177                         security_inode_listsecurity(d_inode(dentry), list, len);
6178         }
6179         return len;
6180 }
6181
6182 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6183         .prefix = XATTR_SECURITY_PREFIX,
6184         .list   = nfs4_xattr_list_nfs4_label,
6185         .get    = nfs4_xattr_get_nfs4_label,
6186         .set    = nfs4_xattr_set_nfs4_label,
6187 };
6188 #endif
6189
6190
6191 /*
6192  * nfs_fhget will use either the mounted_on_fileid or the fileid
6193  */
6194 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6195 {
6196         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6197                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6198               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6199               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6200                 return;
6201
6202         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6203                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6204         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6205         fattr->nlink = 2;
6206 }
6207
6208 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6209                                    const struct qstr *name,
6210                                    struct nfs4_fs_locations *fs_locations,
6211                                    struct page *page)
6212 {
6213         struct nfs_server *server = NFS_SERVER(dir);
6214         u32 bitmask[3] = {
6215                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6216         };
6217         struct nfs4_fs_locations_arg args = {
6218                 .dir_fh = NFS_FH(dir),
6219                 .name = name,
6220                 .page = page,
6221                 .bitmask = bitmask,
6222         };
6223         struct nfs4_fs_locations_res res = {
6224                 .fs_locations = fs_locations,
6225         };
6226         struct rpc_message msg = {
6227                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6228                 .rpc_argp = &args,
6229                 .rpc_resp = &res,
6230         };
6231         int status;
6232
6233         dprintk("%s: start\n", __func__);
6234
6235         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6236          * is not supported */
6237         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6238                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6239         else
6240                 bitmask[0] |= FATTR4_WORD0_FILEID;
6241
6242         nfs_fattr_init(&fs_locations->fattr);
6243         fs_locations->server = server;
6244         fs_locations->nlocations = 0;
6245         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6246         dprintk("%s: returned status = %d\n", __func__, status);
6247         return status;
6248 }
6249
6250 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6251                            const struct qstr *name,
6252                            struct nfs4_fs_locations *fs_locations,
6253                            struct page *page)
6254 {
6255         struct nfs4_exception exception = { };
6256         int err;
6257         do {
6258                 err = _nfs4_proc_fs_locations(client, dir, name,
6259                                 fs_locations, page);
6260                 trace_nfs4_get_fs_locations(dir, name, err);
6261                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6262                                 &exception);
6263         } while (exception.retry);
6264         return err;
6265 }
6266
6267 /*
6268  * This operation also signals the server that this client is
6269  * performing migration recovery.  The server can stop returning
6270  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6271  * appended to this compound to identify the client ID which is
6272  * performing recovery.
6273  */
6274 static int _nfs40_proc_get_locations(struct inode *inode,
6275                                      struct nfs4_fs_locations *locations,
6276                                      struct page *page, struct rpc_cred *cred)
6277 {
6278         struct nfs_server *server = NFS_SERVER(inode);
6279         struct rpc_clnt *clnt = server->client;
6280         u32 bitmask[2] = {
6281                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6282         };
6283         struct nfs4_fs_locations_arg args = {
6284                 .clientid       = server->nfs_client->cl_clientid,
6285                 .fh             = NFS_FH(inode),
6286                 .page           = page,
6287                 .bitmask        = bitmask,
6288                 .migration      = 1,            /* skip LOOKUP */
6289                 .renew          = 1,            /* append RENEW */
6290         };
6291         struct nfs4_fs_locations_res res = {
6292                 .fs_locations   = locations,
6293                 .migration      = 1,
6294                 .renew          = 1,
6295         };
6296         struct rpc_message msg = {
6297                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6298                 .rpc_argp       = &args,
6299                 .rpc_resp       = &res,
6300                 .rpc_cred       = cred,
6301         };
6302         unsigned long now = jiffies;
6303         int status;
6304
6305         nfs_fattr_init(&locations->fattr);
6306         locations->server = server;
6307         locations->nlocations = 0;
6308
6309         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6310         nfs4_set_sequence_privileged(&args.seq_args);
6311         status = nfs4_call_sync_sequence(clnt, server, &msg,
6312                                         &args.seq_args, &res.seq_res);
6313         if (status)
6314                 return status;
6315
6316         renew_lease(server, now);
6317         return 0;
6318 }
6319
6320 #ifdef CONFIG_NFS_V4_1
6321
6322 /*
6323  * This operation also signals the server that this client is
6324  * performing migration recovery.  The server can stop asserting
6325  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6326  * performing this operation is identified in the SEQUENCE
6327  * operation in this compound.
6328  *
6329  * When the client supports GETATTR(fs_locations_info), it can
6330  * be plumbed in here.
6331  */
6332 static int _nfs41_proc_get_locations(struct inode *inode,
6333                                      struct nfs4_fs_locations *locations,
6334                                      struct page *page, struct rpc_cred *cred)
6335 {
6336         struct nfs_server *server = NFS_SERVER(inode);
6337         struct rpc_clnt *clnt = server->client;
6338         u32 bitmask[2] = {
6339                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6340         };
6341         struct nfs4_fs_locations_arg args = {
6342                 .fh             = NFS_FH(inode),
6343                 .page           = page,
6344                 .bitmask        = bitmask,
6345                 .migration      = 1,            /* skip LOOKUP */
6346         };
6347         struct nfs4_fs_locations_res res = {
6348                 .fs_locations   = locations,
6349                 .migration      = 1,
6350         };
6351         struct rpc_message msg = {
6352                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6353                 .rpc_argp       = &args,
6354                 .rpc_resp       = &res,
6355                 .rpc_cred       = cred,
6356         };
6357         int status;
6358
6359         nfs_fattr_init(&locations->fattr);
6360         locations->server = server;
6361         locations->nlocations = 0;
6362
6363         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6364         nfs4_set_sequence_privileged(&args.seq_args);
6365         status = nfs4_call_sync_sequence(clnt, server, &msg,
6366                                         &args.seq_args, &res.seq_res);
6367         if (status == NFS4_OK &&
6368             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6369                 status = -NFS4ERR_LEASE_MOVED;
6370         return status;
6371 }
6372
6373 #endif  /* CONFIG_NFS_V4_1 */
6374
6375 /**
6376  * nfs4_proc_get_locations - discover locations for a migrated FSID
6377  * @inode: inode on FSID that is migrating
6378  * @locations: result of query
6379  * @page: buffer
6380  * @cred: credential to use for this operation
6381  *
6382  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6383  * operation failed, or a negative errno if a local error occurred.
6384  *
6385  * On success, "locations" is filled in, but if the server has
6386  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6387  * asserted.
6388  *
6389  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6390  * from this client that require migration recovery.
6391  */
6392 int nfs4_proc_get_locations(struct inode *inode,
6393                             struct nfs4_fs_locations *locations,
6394                             struct page *page, struct rpc_cred *cred)
6395 {
6396         struct nfs_server *server = NFS_SERVER(inode);
6397         struct nfs_client *clp = server->nfs_client;
6398         const struct nfs4_mig_recovery_ops *ops =
6399                                         clp->cl_mvops->mig_recovery_ops;
6400         struct nfs4_exception exception = { };
6401         int status;
6402
6403         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6404                 (unsigned long long)server->fsid.major,
6405                 (unsigned long long)server->fsid.minor,
6406                 clp->cl_hostname);
6407         nfs_display_fhandle(NFS_FH(inode), __func__);
6408
6409         do {
6410                 status = ops->get_locations(inode, locations, page, cred);
6411                 if (status != -NFS4ERR_DELAY)
6412                         break;
6413                 nfs4_handle_exception(server, status, &exception);
6414         } while (exception.retry);
6415         return status;
6416 }
6417
6418 /*
6419  * This operation also signals the server that this client is
6420  * performing "lease moved" recovery.  The server can stop
6421  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6422  * is appended to this compound to identify the client ID which is
6423  * performing recovery.
6424  */
6425 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6426 {
6427         struct nfs_server *server = NFS_SERVER(inode);
6428         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6429         struct rpc_clnt *clnt = server->client;
6430         struct nfs4_fsid_present_arg args = {
6431                 .fh             = NFS_FH(inode),
6432                 .clientid       = clp->cl_clientid,
6433                 .renew          = 1,            /* append RENEW */
6434         };
6435         struct nfs4_fsid_present_res res = {
6436                 .renew          = 1,
6437         };
6438         struct rpc_message msg = {
6439                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6440                 .rpc_argp       = &args,
6441                 .rpc_resp       = &res,
6442                 .rpc_cred       = cred,
6443         };
6444         unsigned long now = jiffies;
6445         int status;
6446
6447         res.fh = nfs_alloc_fhandle();
6448         if (res.fh == NULL)
6449                 return -ENOMEM;
6450
6451         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6452         nfs4_set_sequence_privileged(&args.seq_args);
6453         status = nfs4_call_sync_sequence(clnt, server, &msg,
6454                                                 &args.seq_args, &res.seq_res);
6455         nfs_free_fhandle(res.fh);
6456         if (status)
6457                 return status;
6458
6459         do_renew_lease(clp, now);
6460         return 0;
6461 }
6462
6463 #ifdef CONFIG_NFS_V4_1
6464
6465 /*
6466  * This operation also signals the server that this client is
6467  * performing "lease moved" recovery.  The server can stop asserting
6468  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6469  * this operation is identified in the SEQUENCE operation in this
6470  * compound.
6471  */
6472 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6473 {
6474         struct nfs_server *server = NFS_SERVER(inode);
6475         struct rpc_clnt *clnt = server->client;
6476         struct nfs4_fsid_present_arg args = {
6477                 .fh             = NFS_FH(inode),
6478         };
6479         struct nfs4_fsid_present_res res = {
6480         };
6481         struct rpc_message msg = {
6482                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6483                 .rpc_argp       = &args,
6484                 .rpc_resp       = &res,
6485                 .rpc_cred       = cred,
6486         };
6487         int status;
6488
6489         res.fh = nfs_alloc_fhandle();
6490         if (res.fh == NULL)
6491                 return -ENOMEM;
6492
6493         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6494         nfs4_set_sequence_privileged(&args.seq_args);
6495         status = nfs4_call_sync_sequence(clnt, server, &msg,
6496                                                 &args.seq_args, &res.seq_res);
6497         nfs_free_fhandle(res.fh);
6498         if (status == NFS4_OK &&
6499             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6500                 status = -NFS4ERR_LEASE_MOVED;
6501         return status;
6502 }
6503
6504 #endif  /* CONFIG_NFS_V4_1 */
6505
6506 /**
6507  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6508  * @inode: inode on FSID to check
6509  * @cred: credential to use for this operation
6510  *
6511  * Server indicates whether the FSID is present, moved, or not
6512  * recognized.  This operation is necessary to clear a LEASE_MOVED
6513  * condition for this client ID.
6514  *
6515  * Returns NFS4_OK if the FSID is present on this server,
6516  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6517  *  NFS4ERR code if some error occurred on the server, or a
6518  *  negative errno if a local failure occurred.
6519  */
6520 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6521 {
6522         struct nfs_server *server = NFS_SERVER(inode);
6523         struct nfs_client *clp = server->nfs_client;
6524         const struct nfs4_mig_recovery_ops *ops =
6525                                         clp->cl_mvops->mig_recovery_ops;
6526         struct nfs4_exception exception = { };
6527         int status;
6528
6529         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6530                 (unsigned long long)server->fsid.major,
6531                 (unsigned long long)server->fsid.minor,
6532                 clp->cl_hostname);
6533         nfs_display_fhandle(NFS_FH(inode), __func__);
6534
6535         do {
6536                 status = ops->fsid_present(inode, cred);
6537                 if (status != -NFS4ERR_DELAY)
6538                         break;
6539                 nfs4_handle_exception(server, status, &exception);
6540         } while (exception.retry);
6541         return status;
6542 }
6543
6544 /**
6545  * If 'use_integrity' is true and the state managment nfs_client
6546  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6547  * and the machine credential as per RFC3530bis and RFC5661 Security
6548  * Considerations sections. Otherwise, just use the user cred with the
6549  * filesystem's rpc_client.
6550  */
6551 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6552 {
6553         int status;
6554         struct nfs4_secinfo_arg args = {
6555                 .dir_fh = NFS_FH(dir),
6556                 .name   = name,
6557         };
6558         struct nfs4_secinfo_res res = {
6559                 .flavors     = flavors,
6560         };
6561         struct rpc_message msg = {
6562                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6563                 .rpc_argp = &args,
6564                 .rpc_resp = &res,
6565         };
6566         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6567         struct rpc_cred *cred = NULL;
6568
6569         if (use_integrity) {
6570                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6571                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6572                 msg.rpc_cred = cred;
6573         }
6574
6575         dprintk("NFS call  secinfo %s\n", name->name);
6576
6577         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6578                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6579
6580         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6581                                 &res.seq_res, 0);
6582         dprintk("NFS reply  secinfo: %d\n", status);
6583
6584         if (cred)
6585                 put_rpccred(cred);
6586
6587         return status;
6588 }
6589
6590 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6591                       struct nfs4_secinfo_flavors *flavors)
6592 {
6593         struct nfs4_exception exception = { };
6594         int err;
6595         do {
6596                 err = -NFS4ERR_WRONGSEC;
6597
6598                 /* try to use integrity protection with machine cred */
6599                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6600                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6601
6602                 /*
6603                  * if unable to use integrity protection, or SECINFO with
6604                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6605                  * disallowed by spec, but exists in deployed servers) use
6606                  * the current filesystem's rpc_client and the user cred.
6607                  */
6608                 if (err == -NFS4ERR_WRONGSEC)
6609                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6610
6611                 trace_nfs4_secinfo(dir, name, err);
6612                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6613                                 &exception);
6614         } while (exception.retry);
6615         return err;
6616 }
6617
6618 #ifdef CONFIG_NFS_V4_1
6619 /*
6620  * Check the exchange flags returned by the server for invalid flags, having
6621  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6622  * DS flags set.
6623  */
6624 static int nfs4_check_cl_exchange_flags(u32 flags)
6625 {
6626         if (flags & ~EXCHGID4_FLAG_MASK_R)
6627                 goto out_inval;
6628         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6629             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6630                 goto out_inval;
6631         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6632                 goto out_inval;
6633         return NFS_OK;
6634 out_inval:
6635         return -NFS4ERR_INVAL;
6636 }
6637
6638 static bool
6639 nfs41_same_server_scope(struct nfs41_server_scope *a,
6640                         struct nfs41_server_scope *b)
6641 {
6642         if (a->server_scope_sz == b->server_scope_sz &&
6643             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6644                 return true;
6645
6646         return false;
6647 }
6648
6649 /*
6650  * nfs4_proc_bind_conn_to_session()
6651  *
6652  * The 4.1 client currently uses the same TCP connection for the
6653  * fore and backchannel.
6654  */
6655 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6656 {
6657         int status;
6658         struct nfs41_bind_conn_to_session_args args = {
6659                 .client = clp,
6660                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6661         };
6662         struct nfs41_bind_conn_to_session_res res;
6663         struct rpc_message msg = {
6664                 .rpc_proc =
6665                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6666                 .rpc_argp = &args,
6667                 .rpc_resp = &res,
6668                 .rpc_cred = cred,
6669         };
6670
6671         dprintk("--> %s\n", __func__);
6672
6673         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6674         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6675                 args.dir = NFS4_CDFC4_FORE;
6676
6677         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6678         trace_nfs4_bind_conn_to_session(clp, status);
6679         if (status == 0) {
6680                 if (memcmp(res.sessionid.data,
6681                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6682                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6683                         status = -EIO;
6684                         goto out;
6685                 }
6686                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6687                         dprintk("NFS: %s: Unexpected direction from server\n",
6688                                 __func__);
6689                         status = -EIO;
6690                         goto out;
6691                 }
6692                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6693                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6694                                 __func__);
6695                         status = -EIO;
6696                         goto out;
6697                 }
6698         }
6699 out:
6700         dprintk("<-- %s status= %d\n", __func__, status);
6701         return status;
6702 }
6703
6704 /*
6705  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6706  * and operations we'd like to see to enable certain features in the allow map
6707  */
6708 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6709         .how = SP4_MACH_CRED,
6710         .enforce.u.words = {
6711                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6712                       1 << (OP_EXCHANGE_ID - 32) |
6713                       1 << (OP_CREATE_SESSION - 32) |
6714                       1 << (OP_DESTROY_SESSION - 32) |
6715                       1 << (OP_DESTROY_CLIENTID - 32)
6716         },
6717         .allow.u.words = {
6718                 [0] = 1 << (OP_CLOSE) |
6719                       1 << (OP_LOCKU) |
6720                       1 << (OP_COMMIT),
6721                 [1] = 1 << (OP_SECINFO - 32) |
6722                       1 << (OP_SECINFO_NO_NAME - 32) |
6723                       1 << (OP_TEST_STATEID - 32) |
6724                       1 << (OP_FREE_STATEID - 32) |
6725                       1 << (OP_WRITE - 32)
6726         }
6727 };
6728
6729 /*
6730  * Select the state protection mode for client `clp' given the server results
6731  * from exchange_id in `sp'.
6732  *
6733  * Returns 0 on success, negative errno otherwise.
6734  */
6735 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6736                                  struct nfs41_state_protection *sp)
6737 {
6738         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6739                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6740                       1 << (OP_EXCHANGE_ID - 32) |
6741                       1 << (OP_CREATE_SESSION - 32) |
6742                       1 << (OP_DESTROY_SESSION - 32) |
6743                       1 << (OP_DESTROY_CLIENTID - 32)
6744         };
6745         unsigned int i;
6746
6747         if (sp->how == SP4_MACH_CRED) {
6748                 /* Print state protect result */
6749                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6750                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6751                         if (test_bit(i, sp->enforce.u.longs))
6752                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6753                         if (test_bit(i, sp->allow.u.longs))
6754                                 dfprintk(MOUNT, "  allow op %d\n", i);
6755                 }
6756
6757                 /* make sure nothing is on enforce list that isn't supported */
6758                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6759                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6760                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6761                                 return -EINVAL;
6762                         }
6763                 }
6764
6765                 /*
6766                  * Minimal mode - state operations are allowed to use machine
6767                  * credential.  Note this already happens by default, so the
6768                  * client doesn't have to do anything more than the negotiation.
6769                  *
6770                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6771                  *       we're already using the machine cred for exchange_id
6772                  *       and will never use a different cred.
6773                  */
6774                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6775                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6776                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6777                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6778                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6779                         dfprintk(MOUNT, "  minimal mode enabled\n");
6780                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6781                 } else {
6782                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6783                         return -EINVAL;
6784                 }
6785
6786                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6787                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6788                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6789                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6790                 }
6791
6792                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6793                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6794                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6795                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6796                 }
6797
6798                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6799                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6800                         dfprintk(MOUNT, "  stateid mode enabled\n");
6801                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6802                 }
6803
6804                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6805                         dfprintk(MOUNT, "  write mode enabled\n");
6806                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6807                 }
6808
6809                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6810                         dfprintk(MOUNT, "  commit mode enabled\n");
6811                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6812                 }
6813         }
6814
6815         return 0;
6816 }
6817
6818 /*
6819  * _nfs4_proc_exchange_id()
6820  *
6821  * Wrapper for EXCHANGE_ID operation.
6822  */
6823 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6824         u32 sp4_how)
6825 {
6826         nfs4_verifier verifier;
6827         struct nfs41_exchange_id_args args = {
6828                 .verifier = &verifier,
6829                 .client = clp,
6830 #ifdef CONFIG_NFS_V4_1_MIGRATION
6831                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6832                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6833                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6834 #else
6835                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6836                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6837 #endif
6838         };
6839         struct nfs41_exchange_id_res res = {
6840                 0
6841         };
6842         int status;
6843         struct rpc_message msg = {
6844                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6845                 .rpc_argp = &args,
6846                 .rpc_resp = &res,
6847                 .rpc_cred = cred,
6848         };
6849
6850         nfs4_init_boot_verifier(clp, &verifier);
6851         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6852                                                         sizeof(args.id));
6853         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6854                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6855                 args.id_len, args.id);
6856
6857         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6858                                         GFP_NOFS);
6859         if (unlikely(res.server_owner == NULL)) {
6860                 status = -ENOMEM;
6861                 goto out;
6862         }
6863
6864         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6865                                         GFP_NOFS);
6866         if (unlikely(res.server_scope == NULL)) {
6867                 status = -ENOMEM;
6868                 goto out_server_owner;
6869         }
6870
6871         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6872         if (unlikely(res.impl_id == NULL)) {
6873                 status = -ENOMEM;
6874                 goto out_server_scope;
6875         }
6876
6877         switch (sp4_how) {
6878         case SP4_NONE:
6879                 args.state_protect.how = SP4_NONE;
6880                 break;
6881
6882         case SP4_MACH_CRED:
6883                 args.state_protect = nfs4_sp4_mach_cred_request;
6884                 break;
6885
6886         default:
6887                 /* unsupported! */
6888                 WARN_ON_ONCE(1);
6889                 status = -EINVAL;
6890                 goto out_server_scope;
6891         }
6892
6893         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6894         trace_nfs4_exchange_id(clp, status);
6895         if (status == 0)
6896                 status = nfs4_check_cl_exchange_flags(res.flags);
6897
6898         if (status == 0)
6899                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
6900
6901         if (status == 0) {
6902                 clp->cl_clientid = res.clientid;
6903                 clp->cl_exchange_flags = res.flags;
6904                 /* Client ID is not confirmed */
6905                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
6906                         clear_bit(NFS4_SESSION_ESTABLISHED,
6907                                         &clp->cl_session->session_state);
6908                         clp->cl_seqid = res.seqid;
6909                 }
6910
6911                 kfree(clp->cl_serverowner);
6912                 clp->cl_serverowner = res.server_owner;
6913                 res.server_owner = NULL;
6914
6915                 /* use the most recent implementation id */
6916                 kfree(clp->cl_implid);
6917                 clp->cl_implid = res.impl_id;
6918
6919                 if (clp->cl_serverscope != NULL &&
6920                     !nfs41_same_server_scope(clp->cl_serverscope,
6921                                              res.server_scope)) {
6922                         dprintk("%s: server_scope mismatch detected\n",
6923                                 __func__);
6924                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6925                         kfree(clp->cl_serverscope);
6926                         clp->cl_serverscope = NULL;
6927                 }
6928
6929                 if (clp->cl_serverscope == NULL) {
6930                         clp->cl_serverscope = res.server_scope;
6931                         goto out;
6932                 }
6933         } else
6934                 kfree(res.impl_id);
6935
6936 out_server_owner:
6937         kfree(res.server_owner);
6938 out_server_scope:
6939         kfree(res.server_scope);
6940 out:
6941         if (clp->cl_implid != NULL)
6942                 dprintk("NFS reply exchange_id: Server Implementation ID: "
6943                         "domain: %s, name: %s, date: %llu,%u\n",
6944                         clp->cl_implid->domain, clp->cl_implid->name,
6945                         clp->cl_implid->date.seconds,
6946                         clp->cl_implid->date.nseconds);
6947         dprintk("NFS reply exchange_id: %d\n", status);
6948         return status;
6949 }
6950
6951 /*
6952  * nfs4_proc_exchange_id()
6953  *
6954  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6955  *
6956  * Since the clientid has expired, all compounds using sessions
6957  * associated with the stale clientid will be returning
6958  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6959  * be in some phase of session reset.
6960  *
6961  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6962  */
6963 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6964 {
6965         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6966         int status;
6967
6968         /* try SP4_MACH_CRED if krb5i/p */
6969         if (authflavor == RPC_AUTH_GSS_KRB5I ||
6970             authflavor == RPC_AUTH_GSS_KRB5P) {
6971                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6972                 if (!status)
6973                         return 0;
6974         }
6975
6976         /* try SP4_NONE */
6977         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6978 }
6979
6980 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6981                 struct rpc_cred *cred)
6982 {
6983         struct rpc_message msg = {
6984                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6985                 .rpc_argp = clp,
6986                 .rpc_cred = cred,
6987         };
6988         int status;
6989
6990         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6991         trace_nfs4_destroy_clientid(clp, status);
6992         if (status)
6993                 dprintk("NFS: Got error %d from the server %s on "
6994                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
6995         return status;
6996 }
6997
6998 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6999                 struct rpc_cred *cred)
7000 {
7001         unsigned int loop;
7002         int ret;
7003
7004         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7005                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7006                 switch (ret) {
7007                 case -NFS4ERR_DELAY:
7008                 case -NFS4ERR_CLIENTID_BUSY:
7009                         ssleep(1);
7010                         break;
7011                 default:
7012                         return ret;
7013                 }
7014         }
7015         return 0;
7016 }
7017
7018 int nfs4_destroy_clientid(struct nfs_client *clp)
7019 {
7020         struct rpc_cred *cred;
7021         int ret = 0;
7022
7023         if (clp->cl_mvops->minor_version < 1)
7024                 goto out;
7025         if (clp->cl_exchange_flags == 0)
7026                 goto out;
7027         if (clp->cl_preserve_clid)
7028                 goto out;
7029         cred = nfs4_get_clid_cred(clp);
7030         ret = nfs4_proc_destroy_clientid(clp, cred);
7031         if (cred)
7032                 put_rpccred(cred);
7033         switch (ret) {
7034         case 0:
7035         case -NFS4ERR_STALE_CLIENTID:
7036                 clp->cl_exchange_flags = 0;
7037         }
7038 out:
7039         return ret;
7040 }
7041
7042 struct nfs4_get_lease_time_data {
7043         struct nfs4_get_lease_time_args *args;
7044         struct nfs4_get_lease_time_res *res;
7045         struct nfs_client *clp;
7046 };
7047
7048 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7049                                         void *calldata)
7050 {
7051         struct nfs4_get_lease_time_data *data =
7052                         (struct nfs4_get_lease_time_data *)calldata;
7053
7054         dprintk("--> %s\n", __func__);
7055         /* just setup sequence, do not trigger session recovery
7056            since we're invoked within one */
7057         nfs41_setup_sequence(data->clp->cl_session,
7058                         &data->args->la_seq_args,
7059                         &data->res->lr_seq_res,
7060                         task);
7061         dprintk("<-- %s\n", __func__);
7062 }
7063
7064 /*
7065  * Called from nfs4_state_manager thread for session setup, so don't recover
7066  * from sequence operation or clientid errors.
7067  */
7068 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7069 {
7070         struct nfs4_get_lease_time_data *data =
7071                         (struct nfs4_get_lease_time_data *)calldata;
7072
7073         dprintk("--> %s\n", __func__);
7074         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7075                 return;
7076         switch (task->tk_status) {
7077         case -NFS4ERR_DELAY:
7078         case -NFS4ERR_GRACE:
7079                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7080                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7081                 task->tk_status = 0;
7082                 /* fall through */
7083         case -NFS4ERR_RETRY_UNCACHED_REP:
7084                 rpc_restart_call_prepare(task);
7085                 return;
7086         }
7087         dprintk("<-- %s\n", __func__);
7088 }
7089
7090 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7091         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7092         .rpc_call_done = nfs4_get_lease_time_done,
7093 };
7094
7095 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7096 {
7097         struct rpc_task *task;
7098         struct nfs4_get_lease_time_args args;
7099         struct nfs4_get_lease_time_res res = {
7100                 .lr_fsinfo = fsinfo,
7101         };
7102         struct nfs4_get_lease_time_data data = {
7103                 .args = &args,
7104                 .res = &res,
7105                 .clp = clp,
7106         };
7107         struct rpc_message msg = {
7108                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7109                 .rpc_argp = &args,
7110                 .rpc_resp = &res,
7111         };
7112         struct rpc_task_setup task_setup = {
7113                 .rpc_client = clp->cl_rpcclient,
7114                 .rpc_message = &msg,
7115                 .callback_ops = &nfs4_get_lease_time_ops,
7116                 .callback_data = &data,
7117                 .flags = RPC_TASK_TIMEOUT,
7118         };
7119         int status;
7120
7121         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7122         nfs4_set_sequence_privileged(&args.la_seq_args);
7123         dprintk("--> %s\n", __func__);
7124         task = rpc_run_task(&task_setup);
7125
7126         if (IS_ERR(task))
7127                 status = PTR_ERR(task);
7128         else {
7129                 status = task->tk_status;
7130                 rpc_put_task(task);
7131         }
7132         dprintk("<-- %s return %d\n", __func__, status);
7133
7134         return status;
7135 }
7136
7137 /*
7138  * Initialize the values to be used by the client in CREATE_SESSION
7139  * If nfs4_init_session set the fore channel request and response sizes,
7140  * use them.
7141  *
7142  * Set the back channel max_resp_sz_cached to zero to force the client to
7143  * always set csa_cachethis to FALSE because the current implementation
7144  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7145  */
7146 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7147 {
7148         unsigned int max_rqst_sz, max_resp_sz;
7149
7150         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7151         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7152
7153         /* Fore channel attributes */
7154         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7155         args->fc_attrs.max_resp_sz = max_resp_sz;
7156         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7157         args->fc_attrs.max_reqs = max_session_slots;
7158
7159         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7160                 "max_ops=%u max_reqs=%u\n",
7161                 __func__,
7162                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7163                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7164
7165         /* Back channel attributes */
7166         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7167         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7168         args->bc_attrs.max_resp_sz_cached = 0;
7169         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7170         args->bc_attrs.max_reqs = 1;
7171
7172         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7173                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7174                 __func__,
7175                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7176                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7177                 args->bc_attrs.max_reqs);
7178 }
7179
7180 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7181                 struct nfs41_create_session_res *res)
7182 {
7183         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7184         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7185
7186         if (rcvd->max_resp_sz > sent->max_resp_sz)
7187                 return -EINVAL;
7188         /*
7189          * Our requested max_ops is the minimum we need; we're not
7190          * prepared to break up compounds into smaller pieces than that.
7191          * So, no point even trying to continue if the server won't
7192          * cooperate:
7193          */
7194         if (rcvd->max_ops < sent->max_ops)
7195                 return -EINVAL;
7196         if (rcvd->max_reqs == 0)
7197                 return -EINVAL;
7198         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7199                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7200         return 0;
7201 }
7202
7203 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7204                 struct nfs41_create_session_res *res)
7205 {
7206         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7207         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7208
7209         if (!(res->flags & SESSION4_BACK_CHAN))
7210                 goto out;
7211         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7212                 return -EINVAL;
7213         if (rcvd->max_resp_sz < sent->max_resp_sz)
7214                 return -EINVAL;
7215         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7216                 return -EINVAL;
7217         /* These would render the backchannel useless: */
7218         if (rcvd->max_ops != sent->max_ops)
7219                 return -EINVAL;
7220         if (rcvd->max_reqs != sent->max_reqs)
7221                 return -EINVAL;
7222 out:
7223         return 0;
7224 }
7225
7226 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7227                                      struct nfs41_create_session_res *res)
7228 {
7229         int ret;
7230
7231         ret = nfs4_verify_fore_channel_attrs(args, res);
7232         if (ret)
7233                 return ret;
7234         return nfs4_verify_back_channel_attrs(args, res);
7235 }
7236
7237 static void nfs4_update_session(struct nfs4_session *session,
7238                 struct nfs41_create_session_res *res)
7239 {
7240         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7241         /* Mark client id and session as being confirmed */
7242         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7243         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7244         session->flags = res->flags;
7245         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7246         if (res->flags & SESSION4_BACK_CHAN)
7247                 memcpy(&session->bc_attrs, &res->bc_attrs,
7248                                 sizeof(session->bc_attrs));
7249 }
7250
7251 static int _nfs4_proc_create_session(struct nfs_client *clp,
7252                 struct rpc_cred *cred)
7253 {
7254         struct nfs4_session *session = clp->cl_session;
7255         struct nfs41_create_session_args args = {
7256                 .client = clp,
7257                 .clientid = clp->cl_clientid,
7258                 .seqid = clp->cl_seqid,
7259                 .cb_program = NFS4_CALLBACK,
7260         };
7261         struct nfs41_create_session_res res;
7262
7263         struct rpc_message msg = {
7264                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7265                 .rpc_argp = &args,
7266                 .rpc_resp = &res,
7267                 .rpc_cred = cred,
7268         };
7269         int status;
7270
7271         nfs4_init_channel_attrs(&args);
7272         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7273
7274         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7275         trace_nfs4_create_session(clp, status);
7276
7277         if (!status) {
7278                 /* Verify the session's negotiated channel_attrs values */
7279                 status = nfs4_verify_channel_attrs(&args, &res);
7280                 /* Increment the clientid slot sequence id */
7281                 if (clp->cl_seqid == res.seqid)
7282                         clp->cl_seqid++;
7283                 if (status)
7284                         goto out;
7285                 nfs4_update_session(session, &res);
7286         }
7287 out:
7288         return status;
7289 }
7290
7291 /*
7292  * Issues a CREATE_SESSION operation to the server.
7293  * It is the responsibility of the caller to verify the session is
7294  * expired before calling this routine.
7295  */
7296 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7297 {
7298         int status;
7299         unsigned *ptr;
7300         struct nfs4_session *session = clp->cl_session;
7301
7302         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7303
7304         status = _nfs4_proc_create_session(clp, cred);
7305         if (status)
7306                 goto out;
7307
7308         /* Init or reset the session slot tables */
7309         status = nfs4_setup_session_slot_tables(session);
7310         dprintk("slot table setup returned %d\n", status);
7311         if (status)
7312                 goto out;
7313
7314         ptr = (unsigned *)&session->sess_id.data[0];
7315         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7316                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7317 out:
7318         dprintk("<-- %s\n", __func__);
7319         return status;
7320 }
7321
7322 /*
7323  * Issue the over-the-wire RPC DESTROY_SESSION.
7324  * The caller must serialize access to this routine.
7325  */
7326 int nfs4_proc_destroy_session(struct nfs4_session *session,
7327                 struct rpc_cred *cred)
7328 {
7329         struct rpc_message msg = {
7330                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7331                 .rpc_argp = session,
7332                 .rpc_cred = cred,
7333         };
7334         int status = 0;
7335
7336         dprintk("--> nfs4_proc_destroy_session\n");
7337
7338         /* session is still being setup */
7339         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7340                 return 0;
7341
7342         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7343         trace_nfs4_destroy_session(session->clp, status);
7344
7345         if (status)
7346                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7347                         "Session has been destroyed regardless...\n", status);
7348
7349         dprintk("<-- nfs4_proc_destroy_session\n");
7350         return status;
7351 }
7352
7353 /*
7354  * Renew the cl_session lease.
7355  */
7356 struct nfs4_sequence_data {
7357         struct nfs_client *clp;
7358         struct nfs4_sequence_args args;
7359         struct nfs4_sequence_res res;
7360 };
7361
7362 static void nfs41_sequence_release(void *data)
7363 {
7364         struct nfs4_sequence_data *calldata = data;
7365         struct nfs_client *clp = calldata->clp;
7366
7367         if (atomic_read(&clp->cl_count) > 1)
7368                 nfs4_schedule_state_renewal(clp);
7369         nfs_put_client(clp);
7370         kfree(calldata);
7371 }
7372
7373 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7374 {
7375         switch(task->tk_status) {
7376         case -NFS4ERR_DELAY:
7377                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7378                 return -EAGAIN;
7379         default:
7380                 nfs4_schedule_lease_recovery(clp);
7381         }
7382         return 0;
7383 }
7384
7385 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7386 {
7387         struct nfs4_sequence_data *calldata = data;
7388         struct nfs_client *clp = calldata->clp;
7389
7390         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7391                 return;
7392
7393         trace_nfs4_sequence(clp, task->tk_status);
7394         if (task->tk_status < 0) {
7395                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7396                 if (atomic_read(&clp->cl_count) == 1)
7397                         goto out;
7398
7399                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7400                         rpc_restart_call_prepare(task);
7401                         return;
7402                 }
7403         }
7404         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7405 out:
7406         dprintk("<-- %s\n", __func__);
7407 }
7408
7409 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7410 {
7411         struct nfs4_sequence_data *calldata = data;
7412         struct nfs_client *clp = calldata->clp;
7413         struct nfs4_sequence_args *args;
7414         struct nfs4_sequence_res *res;
7415
7416         args = task->tk_msg.rpc_argp;
7417         res = task->tk_msg.rpc_resp;
7418
7419         nfs41_setup_sequence(clp->cl_session, args, res, task);
7420 }
7421
7422 static const struct rpc_call_ops nfs41_sequence_ops = {
7423         .rpc_call_done = nfs41_sequence_call_done,
7424         .rpc_call_prepare = nfs41_sequence_prepare,
7425         .rpc_release = nfs41_sequence_release,
7426 };
7427
7428 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7429                 struct rpc_cred *cred,
7430                 bool is_privileged)
7431 {
7432         struct nfs4_sequence_data *calldata;
7433         struct rpc_message msg = {
7434                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7435                 .rpc_cred = cred,
7436         };
7437         struct rpc_task_setup task_setup_data = {
7438                 .rpc_client = clp->cl_rpcclient,
7439                 .rpc_message = &msg,
7440                 .callback_ops = &nfs41_sequence_ops,
7441                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7442         };
7443
7444         if (!atomic_inc_not_zero(&clp->cl_count))
7445                 return ERR_PTR(-EIO);
7446         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7447         if (calldata == NULL) {
7448                 nfs_put_client(clp);
7449                 return ERR_PTR(-ENOMEM);
7450         }
7451         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7452         if (is_privileged)
7453                 nfs4_set_sequence_privileged(&calldata->args);
7454         msg.rpc_argp = &calldata->args;
7455         msg.rpc_resp = &calldata->res;
7456         calldata->clp = clp;
7457         task_setup_data.callback_data = calldata;
7458
7459         return rpc_run_task(&task_setup_data);
7460 }
7461
7462 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7463 {
7464         struct rpc_task *task;
7465         int ret = 0;
7466
7467         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7468                 return -EAGAIN;
7469         task = _nfs41_proc_sequence(clp, cred, false);
7470         if (IS_ERR(task))
7471                 ret = PTR_ERR(task);
7472         else
7473                 rpc_put_task_async(task);
7474         dprintk("<-- %s status=%d\n", __func__, ret);
7475         return ret;
7476 }
7477
7478 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7479 {
7480         struct rpc_task *task;
7481         int ret;
7482
7483         task = _nfs41_proc_sequence(clp, cred, true);
7484         if (IS_ERR(task)) {
7485                 ret = PTR_ERR(task);
7486                 goto out;
7487         }
7488         ret = rpc_wait_for_completion_task(task);
7489         if (!ret) {
7490                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7491
7492                 if (task->tk_status == 0)
7493                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7494                 ret = task->tk_status;
7495         }
7496         rpc_put_task(task);
7497 out:
7498         dprintk("<-- %s status=%d\n", __func__, ret);
7499         return ret;
7500 }
7501
7502 struct nfs4_reclaim_complete_data {
7503         struct nfs_client *clp;
7504         struct nfs41_reclaim_complete_args arg;
7505         struct nfs41_reclaim_complete_res res;
7506 };
7507
7508 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7509 {
7510         struct nfs4_reclaim_complete_data *calldata = data;
7511
7512         nfs41_setup_sequence(calldata->clp->cl_session,
7513                         &calldata->arg.seq_args,
7514                         &calldata->res.seq_res,
7515                         task);
7516 }
7517
7518 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7519 {
7520         switch(task->tk_status) {
7521         case 0:
7522         case -NFS4ERR_COMPLETE_ALREADY:
7523         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7524                 break;
7525         case -NFS4ERR_DELAY:
7526                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7527                 /* fall through */
7528         case -NFS4ERR_RETRY_UNCACHED_REP:
7529                 return -EAGAIN;
7530         default:
7531                 nfs4_schedule_lease_recovery(clp);
7532         }
7533         return 0;
7534 }
7535
7536 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7537 {
7538         struct nfs4_reclaim_complete_data *calldata = data;
7539         struct nfs_client *clp = calldata->clp;
7540         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7541
7542         dprintk("--> %s\n", __func__);
7543         if (!nfs41_sequence_done(task, res))
7544                 return;
7545
7546         trace_nfs4_reclaim_complete(clp, task->tk_status);
7547         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7548                 rpc_restart_call_prepare(task);
7549                 return;
7550         }
7551         dprintk("<-- %s\n", __func__);
7552 }
7553
7554 static void nfs4_free_reclaim_complete_data(void *data)
7555 {
7556         struct nfs4_reclaim_complete_data *calldata = data;
7557
7558         kfree(calldata);
7559 }
7560
7561 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7562         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7563         .rpc_call_done = nfs4_reclaim_complete_done,
7564         .rpc_release = nfs4_free_reclaim_complete_data,
7565 };
7566
7567 /*
7568  * Issue a global reclaim complete.
7569  */
7570 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7571                 struct rpc_cred *cred)
7572 {
7573         struct nfs4_reclaim_complete_data *calldata;
7574         struct rpc_task *task;
7575         struct rpc_message msg = {
7576                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7577                 .rpc_cred = cred,
7578         };
7579         struct rpc_task_setup task_setup_data = {
7580                 .rpc_client = clp->cl_rpcclient,
7581                 .rpc_message = &msg,
7582                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7583                 .flags = RPC_TASK_ASYNC,
7584         };
7585         int status = -ENOMEM;
7586
7587         dprintk("--> %s\n", __func__);
7588         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7589         if (calldata == NULL)
7590                 goto out;
7591         calldata->clp = clp;
7592         calldata->arg.one_fs = 0;
7593
7594         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7595         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7596         msg.rpc_argp = &calldata->arg;
7597         msg.rpc_resp = &calldata->res;
7598         task_setup_data.callback_data = calldata;
7599         task = rpc_run_task(&task_setup_data);
7600         if (IS_ERR(task)) {
7601                 status = PTR_ERR(task);
7602                 goto out;
7603         }
7604         status = nfs4_wait_for_completion_rpc_task(task);
7605         if (status == 0)
7606                 status = task->tk_status;
7607         rpc_put_task(task);
7608         return 0;
7609 out:
7610         dprintk("<-- %s status=%d\n", __func__, status);
7611         return status;
7612 }
7613
7614 static void
7615 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7616 {
7617         struct nfs4_layoutget *lgp = calldata;
7618         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7619         struct nfs4_session *session = nfs4_get_session(server);
7620
7621         dprintk("--> %s\n", __func__);
7622         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7623          * right now covering the LAYOUTGET we are about to send.
7624          * However, that is not so catastrophic, and there seems
7625          * to be no way to prevent it completely.
7626          */
7627         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7628                                 &lgp->res.seq_res, task))
7629                 return;
7630         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7631                                           NFS_I(lgp->args.inode)->layout,
7632                                           &lgp->args.range,
7633                                           lgp->args.ctx->state)) {
7634                 rpc_exit(task, NFS4_OK);
7635         }
7636 }
7637
7638 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7639 {
7640         struct nfs4_layoutget *lgp = calldata;
7641         struct inode *inode = lgp->args.inode;
7642         struct nfs_server *server = NFS_SERVER(inode);
7643         struct pnfs_layout_hdr *lo;
7644         struct nfs4_state *state = NULL;
7645         unsigned long timeo, now, giveup;
7646
7647         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7648
7649         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7650                 goto out;
7651
7652         switch (task->tk_status) {
7653         case 0:
7654                 goto out;
7655         /*
7656          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7657          * (or clients) writing to the same RAID stripe
7658          */
7659         case -NFS4ERR_LAYOUTTRYLATER:
7660         /*
7661          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7662          * existing layout before getting a new one).
7663          */
7664         case -NFS4ERR_RECALLCONFLICT:
7665                 timeo = rpc_get_timeout(task->tk_client);
7666                 giveup = lgp->args.timestamp + timeo;
7667                 now = jiffies;
7668                 if (time_after(giveup, now)) {
7669                         unsigned long delay;
7670
7671                         /* Delay for:
7672                          * - Not less then NFS4_POLL_RETRY_MIN.
7673                          * - One last time a jiffie before we give up
7674                          * - exponential backoff (time_now minus start_attempt)
7675                          */
7676                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7677                                     min((giveup - now - 1),
7678                                         now - lgp->args.timestamp));
7679
7680                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7681                                 __func__, delay);
7682                         rpc_delay(task, delay);
7683                         task->tk_status = 0;
7684                         rpc_restart_call_prepare(task);
7685                         goto out; /* Do not call nfs4_async_handle_error() */
7686                 }
7687                 break;
7688         case -NFS4ERR_EXPIRED:
7689         case -NFS4ERR_BAD_STATEID:
7690                 spin_lock(&inode->i_lock);
7691                 lo = NFS_I(inode)->layout;
7692                 if (!lo || list_empty(&lo->plh_segs)) {
7693                         spin_unlock(&inode->i_lock);
7694                         /* If the open stateid was bad, then recover it. */
7695                         state = lgp->args.ctx->state;
7696                 } else {
7697                         LIST_HEAD(head);
7698
7699                         /*
7700                          * Mark the bad layout state as invalid, then retry
7701                          * with the current stateid.
7702                          */
7703                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7704                         spin_unlock(&inode->i_lock);
7705                         pnfs_free_lseg_list(&head);
7706         
7707                         task->tk_status = 0;
7708                         rpc_restart_call_prepare(task);
7709                 }
7710         }
7711         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7712                 rpc_restart_call_prepare(task);
7713 out:
7714         dprintk("<-- %s\n", __func__);
7715 }
7716
7717 static size_t max_response_pages(struct nfs_server *server)
7718 {
7719         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7720         return nfs_page_array_len(0, max_resp_sz);
7721 }
7722
7723 static void nfs4_free_pages(struct page **pages, size_t size)
7724 {
7725         int i;
7726
7727         if (!pages)
7728                 return;
7729
7730         for (i = 0; i < size; i++) {
7731                 if (!pages[i])
7732                         break;
7733                 __free_page(pages[i]);
7734         }
7735         kfree(pages);
7736 }
7737
7738 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7739 {
7740         struct page **pages;
7741         int i;
7742
7743         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7744         if (!pages) {
7745                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7746                 return NULL;
7747         }
7748
7749         for (i = 0; i < size; i++) {
7750                 pages[i] = alloc_page(gfp_flags);
7751                 if (!pages[i]) {
7752                         dprintk("%s: failed to allocate page\n", __func__);
7753                         nfs4_free_pages(pages, size);
7754                         return NULL;
7755                 }
7756         }
7757
7758         return pages;
7759 }
7760
7761 static void nfs4_layoutget_release(void *calldata)
7762 {
7763         struct nfs4_layoutget *lgp = calldata;
7764         struct inode *inode = lgp->args.inode;
7765         struct nfs_server *server = NFS_SERVER(inode);
7766         size_t max_pages = max_response_pages(server);
7767
7768         dprintk("--> %s\n", __func__);
7769         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7770         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7771         put_nfs_open_context(lgp->args.ctx);
7772         kfree(calldata);
7773         dprintk("<-- %s\n", __func__);
7774 }
7775
7776 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7777         .rpc_call_prepare = nfs4_layoutget_prepare,
7778         .rpc_call_done = nfs4_layoutget_done,
7779         .rpc_release = nfs4_layoutget_release,
7780 };
7781
7782 struct pnfs_layout_segment *
7783 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7784 {
7785         struct inode *inode = lgp->args.inode;
7786         struct nfs_server *server = NFS_SERVER(inode);
7787         size_t max_pages = max_response_pages(server);
7788         struct rpc_task *task;
7789         struct rpc_message msg = {
7790                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7791                 .rpc_argp = &lgp->args,
7792                 .rpc_resp = &lgp->res,
7793                 .rpc_cred = lgp->cred,
7794         };
7795         struct rpc_task_setup task_setup_data = {
7796                 .rpc_client = server->client,
7797                 .rpc_message = &msg,
7798                 .callback_ops = &nfs4_layoutget_call_ops,
7799                 .callback_data = lgp,
7800                 .flags = RPC_TASK_ASYNC,
7801         };
7802         struct pnfs_layout_segment *lseg = NULL;
7803         int status = 0;
7804
7805         dprintk("--> %s\n", __func__);
7806
7807         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7808         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7809
7810         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7811         if (!lgp->args.layout.pages) {
7812                 nfs4_layoutget_release(lgp);
7813                 return ERR_PTR(-ENOMEM);
7814         }
7815         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7816         lgp->args.timestamp = jiffies;
7817
7818         lgp->res.layoutp = &lgp->args.layout;
7819         lgp->res.seq_res.sr_slot = NULL;
7820         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7821
7822         task = rpc_run_task(&task_setup_data);
7823         if (IS_ERR(task))
7824                 return ERR_CAST(task);
7825         status = nfs4_wait_for_completion_rpc_task(task);
7826         if (status == 0)
7827                 status = task->tk_status;
7828         trace_nfs4_layoutget(lgp->args.ctx,
7829                         &lgp->args.range,
7830                         &lgp->res.range,
7831                         status);
7832         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7833         if (status == 0 && lgp->res.layoutp->len)
7834                 lseg = pnfs_layout_process(lgp);
7835         rpc_put_task(task);
7836         dprintk("<-- %s status=%d\n", __func__, status);
7837         if (status)
7838                 return ERR_PTR(status);
7839         return lseg;
7840 }
7841
7842 static void
7843 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7844 {
7845         struct nfs4_layoutreturn *lrp = calldata;
7846
7847         dprintk("--> %s\n", __func__);
7848         nfs41_setup_sequence(lrp->clp->cl_session,
7849                         &lrp->args.seq_args,
7850                         &lrp->res.seq_res,
7851                         task);
7852 }
7853
7854 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7855 {
7856         struct nfs4_layoutreturn *lrp = calldata;
7857         struct nfs_server *server;
7858
7859         dprintk("--> %s\n", __func__);
7860
7861         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7862                 return;
7863
7864         server = NFS_SERVER(lrp->args.inode);
7865         switch (task->tk_status) {
7866         default:
7867                 task->tk_status = 0;
7868         case 0:
7869                 break;
7870         case -NFS4ERR_DELAY:
7871                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7872                         break;
7873                 rpc_restart_call_prepare(task);
7874                 return;
7875         }
7876         dprintk("<-- %s\n", __func__);
7877 }
7878
7879 static void nfs4_layoutreturn_release(void *calldata)
7880 {
7881         struct nfs4_layoutreturn *lrp = calldata;
7882         struct pnfs_layout_hdr *lo = lrp->args.layout;
7883
7884         dprintk("--> %s\n", __func__);
7885         spin_lock(&lo->plh_inode->i_lock);
7886         if (lrp->res.lrs_present)
7887                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7888         pnfs_clear_layoutreturn_waitbit(lo);
7889         clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, &lo->plh_flags);
7890         rpc_wake_up(&NFS_SERVER(lo->plh_inode)->roc_rpcwaitq);
7891         lo->plh_block_lgets--;
7892         spin_unlock(&lo->plh_inode->i_lock);
7893         pnfs_put_layout_hdr(lrp->args.layout);
7894         nfs_iput_and_deactive(lrp->inode);
7895         kfree(calldata);
7896         dprintk("<-- %s\n", __func__);
7897 }
7898
7899 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7900         .rpc_call_prepare = nfs4_layoutreturn_prepare,
7901         .rpc_call_done = nfs4_layoutreturn_done,
7902         .rpc_release = nfs4_layoutreturn_release,
7903 };
7904
7905 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
7906 {
7907         struct rpc_task *task;
7908         struct rpc_message msg = {
7909                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7910                 .rpc_argp = &lrp->args,
7911                 .rpc_resp = &lrp->res,
7912                 .rpc_cred = lrp->cred,
7913         };
7914         struct rpc_task_setup task_setup_data = {
7915                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
7916                 .rpc_message = &msg,
7917                 .callback_ops = &nfs4_layoutreturn_call_ops,
7918                 .callback_data = lrp,
7919         };
7920         int status = 0;
7921
7922         dprintk("--> %s\n", __func__);
7923         if (!sync) {
7924                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
7925                 if (!lrp->inode) {
7926                         nfs4_layoutreturn_release(lrp);
7927                         return -EAGAIN;
7928                 }
7929                 task_setup_data.flags |= RPC_TASK_ASYNC;
7930         }
7931         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7932         task = rpc_run_task(&task_setup_data);
7933         if (IS_ERR(task))
7934                 return PTR_ERR(task);
7935         if (sync)
7936                 status = task->tk_status;
7937         trace_nfs4_layoutreturn(lrp->args.inode, status);
7938         dprintk("<-- %s status=%d\n", __func__, status);
7939         rpc_put_task(task);
7940         return status;
7941 }
7942
7943 static int
7944 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7945                 struct pnfs_device *pdev,
7946                 struct rpc_cred *cred)
7947 {
7948         struct nfs4_getdeviceinfo_args args = {
7949                 .pdev = pdev,
7950                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
7951                         NOTIFY_DEVICEID4_DELETE,
7952         };
7953         struct nfs4_getdeviceinfo_res res = {
7954                 .pdev = pdev,
7955         };
7956         struct rpc_message msg = {
7957                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7958                 .rpc_argp = &args,
7959                 .rpc_resp = &res,
7960                 .rpc_cred = cred,
7961         };
7962         int status;
7963
7964         dprintk("--> %s\n", __func__);
7965         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7966         if (res.notification & ~args.notify_types)
7967                 dprintk("%s: unsupported notification\n", __func__);
7968         if (res.notification != args.notify_types)
7969                 pdev->nocache = 1;
7970
7971         dprintk("<-- %s status=%d\n", __func__, status);
7972
7973         return status;
7974 }
7975
7976 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7977                 struct pnfs_device *pdev,
7978                 struct rpc_cred *cred)
7979 {
7980         struct nfs4_exception exception = { };
7981         int err;
7982
7983         do {
7984                 err = nfs4_handle_exception(server,
7985                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
7986                                         &exception);
7987         } while (exception.retry);
7988         return err;
7989 }
7990 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7991
7992 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7993 {
7994         struct nfs4_layoutcommit_data *data = calldata;
7995         struct nfs_server *server = NFS_SERVER(data->args.inode);
7996         struct nfs4_session *session = nfs4_get_session(server);
7997
7998         nfs41_setup_sequence(session,
7999                         &data->args.seq_args,
8000                         &data->res.seq_res,
8001                         task);
8002 }
8003
8004 static void
8005 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8006 {
8007         struct nfs4_layoutcommit_data *data = calldata;
8008         struct nfs_server *server = NFS_SERVER(data->args.inode);
8009
8010         if (!nfs41_sequence_done(task, &data->res.seq_res))
8011                 return;
8012
8013         switch (task->tk_status) { /* Just ignore these failures */
8014         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8015         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8016         case -NFS4ERR_BADLAYOUT:     /* no layout */
8017         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8018                 task->tk_status = 0;
8019         case 0:
8020                 break;
8021         default:
8022                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8023                         rpc_restart_call_prepare(task);
8024                         return;
8025                 }
8026         }
8027 }
8028
8029 static void nfs4_layoutcommit_release(void *calldata)
8030 {
8031         struct nfs4_layoutcommit_data *data = calldata;
8032
8033         pnfs_cleanup_layoutcommit(data);
8034         nfs_post_op_update_inode_force_wcc(data->args.inode,
8035                                            data->res.fattr);
8036         put_rpccred(data->cred);
8037         nfs_iput_and_deactive(data->inode);
8038         kfree(data);
8039 }
8040
8041 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8042         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8043         .rpc_call_done = nfs4_layoutcommit_done,
8044         .rpc_release = nfs4_layoutcommit_release,
8045 };
8046
8047 int
8048 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8049 {
8050         struct rpc_message msg = {
8051                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8052                 .rpc_argp = &data->args,
8053                 .rpc_resp = &data->res,
8054                 .rpc_cred = data->cred,
8055         };
8056         struct rpc_task_setup task_setup_data = {
8057                 .task = &data->task,
8058                 .rpc_client = NFS_CLIENT(data->args.inode),
8059                 .rpc_message = &msg,
8060                 .callback_ops = &nfs4_layoutcommit_ops,
8061                 .callback_data = data,
8062         };
8063         struct rpc_task *task;
8064         int status = 0;
8065
8066         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
8067                 "lbw: %llu inode %lu\n",
8068                 data->task.tk_pid, sync,
8069                 data->args.lastbytewritten,
8070                 data->args.inode->i_ino);
8071
8072         if (!sync) {
8073                 data->inode = nfs_igrab_and_active(data->args.inode);
8074                 if (data->inode == NULL) {
8075                         nfs4_layoutcommit_release(data);
8076                         return -EAGAIN;
8077                 }
8078                 task_setup_data.flags = RPC_TASK_ASYNC;
8079         }
8080         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8081         task = rpc_run_task(&task_setup_data);
8082         if (IS_ERR(task))
8083                 return PTR_ERR(task);
8084         if (sync)
8085                 status = task->tk_status;
8086         trace_nfs4_layoutcommit(data->args.inode, status);
8087         dprintk("%s: status %d\n", __func__, status);
8088         rpc_put_task(task);
8089         return status;
8090 }
8091
8092 /**
8093  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8094  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8095  */
8096 static int
8097 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8098                     struct nfs_fsinfo *info,
8099                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8100 {
8101         struct nfs41_secinfo_no_name_args args = {
8102                 .style = SECINFO_STYLE_CURRENT_FH,
8103         };
8104         struct nfs4_secinfo_res res = {
8105                 .flavors = flavors,
8106         };
8107         struct rpc_message msg = {
8108                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8109                 .rpc_argp = &args,
8110                 .rpc_resp = &res,
8111         };
8112         struct rpc_clnt *clnt = server->client;
8113         struct rpc_cred *cred = NULL;
8114         int status;
8115
8116         if (use_integrity) {
8117                 clnt = server->nfs_client->cl_rpcclient;
8118                 cred = nfs4_get_clid_cred(server->nfs_client);
8119                 msg.rpc_cred = cred;
8120         }
8121
8122         dprintk("--> %s\n", __func__);
8123         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8124                                 &res.seq_res, 0);
8125         dprintk("<-- %s status=%d\n", __func__, status);
8126
8127         if (cred)
8128                 put_rpccred(cred);
8129
8130         return status;
8131 }
8132
8133 static int
8134 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8135                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8136 {
8137         struct nfs4_exception exception = { };
8138         int err;
8139         do {
8140                 /* first try using integrity protection */
8141                 err = -NFS4ERR_WRONGSEC;
8142
8143                 /* try to use integrity protection with machine cred */
8144                 if (_nfs4_is_integrity_protected(server->nfs_client))
8145                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8146                                                           flavors, true);
8147
8148                 /*
8149                  * if unable to use integrity protection, or SECINFO with
8150                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8151                  * disallowed by spec, but exists in deployed servers) use
8152                  * the current filesystem's rpc_client and the user cred.
8153                  */
8154                 if (err == -NFS4ERR_WRONGSEC)
8155                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8156                                                           flavors, false);
8157
8158                 switch (err) {
8159                 case 0:
8160                 case -NFS4ERR_WRONGSEC:
8161                 case -ENOTSUPP:
8162                         goto out;
8163                 default:
8164                         err = nfs4_handle_exception(server, err, &exception);
8165                 }
8166         } while (exception.retry);
8167 out:
8168         return err;
8169 }
8170
8171 static int
8172 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8173                     struct nfs_fsinfo *info)
8174 {
8175         int err;
8176         struct page *page;
8177         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8178         struct nfs4_secinfo_flavors *flavors;
8179         struct nfs4_secinfo4 *secinfo;
8180         int i;
8181
8182         page = alloc_page(GFP_KERNEL);
8183         if (!page) {
8184                 err = -ENOMEM;
8185                 goto out;
8186         }
8187
8188         flavors = page_address(page);
8189         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8190
8191         /*
8192          * Fall back on "guess and check" method if
8193          * the server doesn't support SECINFO_NO_NAME
8194          */
8195         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8196                 err = nfs4_find_root_sec(server, fhandle, info);
8197                 goto out_freepage;
8198         }
8199         if (err)
8200                 goto out_freepage;
8201
8202         for (i = 0; i < flavors->num_flavors; i++) {
8203                 secinfo = &flavors->flavors[i];
8204
8205                 switch (secinfo->flavor) {
8206                 case RPC_AUTH_NULL:
8207                 case RPC_AUTH_UNIX:
8208                 case RPC_AUTH_GSS:
8209                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8210                                         &secinfo->flavor_info);
8211                         break;
8212                 default:
8213                         flavor = RPC_AUTH_MAXFLAVOR;
8214                         break;
8215                 }
8216
8217                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8218                         flavor = RPC_AUTH_MAXFLAVOR;
8219
8220                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8221                         err = nfs4_lookup_root_sec(server, fhandle,
8222                                                    info, flavor);
8223                         if (!err)
8224                                 break;
8225                 }
8226         }
8227
8228         if (flavor == RPC_AUTH_MAXFLAVOR)
8229                 err = -EPERM;
8230
8231 out_freepage:
8232         put_page(page);
8233         if (err == -EACCES)
8234                 return -EPERM;
8235 out:
8236         return err;
8237 }
8238
8239 static int _nfs41_test_stateid(struct nfs_server *server,
8240                 nfs4_stateid *stateid,
8241                 struct rpc_cred *cred)
8242 {
8243         int status;
8244         struct nfs41_test_stateid_args args = {
8245                 .stateid = stateid,
8246         };
8247         struct nfs41_test_stateid_res res;
8248         struct rpc_message msg = {
8249                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8250                 .rpc_argp = &args,
8251                 .rpc_resp = &res,
8252                 .rpc_cred = cred,
8253         };
8254         struct rpc_clnt *rpc_client = server->client;
8255
8256         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8257                 &rpc_client, &msg);
8258
8259         dprintk("NFS call  test_stateid %p\n", stateid);
8260         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8261         nfs4_set_sequence_privileged(&args.seq_args);
8262         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8263                         &args.seq_args, &res.seq_res);
8264         if (status != NFS_OK) {
8265                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8266                 return status;
8267         }
8268         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8269         return -res.status;
8270 }
8271
8272 /**
8273  * nfs41_test_stateid - perform a TEST_STATEID operation
8274  *
8275  * @server: server / transport on which to perform the operation
8276  * @stateid: state ID to test
8277  * @cred: credential
8278  *
8279  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8280  * Otherwise a negative NFS4ERR value is returned if the operation
8281  * failed or the state ID is not currently valid.
8282  */
8283 static int nfs41_test_stateid(struct nfs_server *server,
8284                 nfs4_stateid *stateid,
8285                 struct rpc_cred *cred)
8286 {
8287         struct nfs4_exception exception = { };
8288         int err;
8289         do {
8290                 err = _nfs41_test_stateid(server, stateid, cred);
8291                 if (err != -NFS4ERR_DELAY)
8292                         break;
8293                 nfs4_handle_exception(server, err, &exception);
8294         } while (exception.retry);
8295         return err;
8296 }
8297
8298 struct nfs_free_stateid_data {
8299         struct nfs_server *server;
8300         struct nfs41_free_stateid_args args;
8301         struct nfs41_free_stateid_res res;
8302 };
8303
8304 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8305 {
8306         struct nfs_free_stateid_data *data = calldata;
8307         nfs41_setup_sequence(nfs4_get_session(data->server),
8308                         &data->args.seq_args,
8309                         &data->res.seq_res,
8310                         task);
8311 }
8312
8313 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8314 {
8315         struct nfs_free_stateid_data *data = calldata;
8316
8317         nfs41_sequence_done(task, &data->res.seq_res);
8318
8319         switch (task->tk_status) {
8320         case -NFS4ERR_DELAY:
8321                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8322                         rpc_restart_call_prepare(task);
8323         }
8324 }
8325
8326 static void nfs41_free_stateid_release(void *calldata)
8327 {
8328         kfree(calldata);
8329 }
8330
8331 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8332         .rpc_call_prepare = nfs41_free_stateid_prepare,
8333         .rpc_call_done = nfs41_free_stateid_done,
8334         .rpc_release = nfs41_free_stateid_release,
8335 };
8336
8337 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8338                 nfs4_stateid *stateid,
8339                 struct rpc_cred *cred,
8340                 bool privileged)
8341 {
8342         struct rpc_message msg = {
8343                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8344                 .rpc_cred = cred,
8345         };
8346         struct rpc_task_setup task_setup = {
8347                 .rpc_client = server->client,
8348                 .rpc_message = &msg,
8349                 .callback_ops = &nfs41_free_stateid_ops,
8350                 .flags = RPC_TASK_ASYNC,
8351         };
8352         struct nfs_free_stateid_data *data;
8353
8354         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8355                 &task_setup.rpc_client, &msg);
8356
8357         dprintk("NFS call  free_stateid %p\n", stateid);
8358         data = kmalloc(sizeof(*data), GFP_NOFS);
8359         if (!data)
8360                 return ERR_PTR(-ENOMEM);
8361         data->server = server;
8362         nfs4_stateid_copy(&data->args.stateid, stateid);
8363
8364         task_setup.callback_data = data;
8365
8366         msg.rpc_argp = &data->args;
8367         msg.rpc_resp = &data->res;
8368         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8369         if (privileged)
8370                 nfs4_set_sequence_privileged(&data->args.seq_args);
8371
8372         return rpc_run_task(&task_setup);
8373 }
8374
8375 /**
8376  * nfs41_free_stateid - perform a FREE_STATEID operation
8377  *
8378  * @server: server / transport on which to perform the operation
8379  * @stateid: state ID to release
8380  * @cred: credential
8381  *
8382  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8383  * negative NFS4ERR value is returned.
8384  */
8385 static int nfs41_free_stateid(struct nfs_server *server,
8386                 nfs4_stateid *stateid,
8387                 struct rpc_cred *cred)
8388 {
8389         struct rpc_task *task;
8390         int ret;
8391
8392         task = _nfs41_free_stateid(server, stateid, cred, true);
8393         if (IS_ERR(task))
8394                 return PTR_ERR(task);
8395         ret = rpc_wait_for_completion_task(task);
8396         if (!ret)
8397                 ret = task->tk_status;
8398         rpc_put_task(task);
8399         return ret;
8400 }
8401
8402 static void
8403 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8404 {
8405         struct rpc_task *task;
8406         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8407
8408         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8409         nfs4_free_lock_state(server, lsp);
8410         if (IS_ERR(task))
8411                 return;
8412         rpc_put_task(task);
8413 }
8414
8415 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8416                 const nfs4_stateid *s2)
8417 {
8418         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8419                 return false;
8420
8421         if (s1->seqid == s2->seqid)
8422                 return true;
8423         if (s1->seqid == 0 || s2->seqid == 0)
8424                 return true;
8425
8426         return false;
8427 }
8428
8429 #endif /* CONFIG_NFS_V4_1 */
8430
8431 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8432                 const nfs4_stateid *s2)
8433 {
8434         return nfs4_stateid_match(s1, s2);
8435 }
8436
8437
8438 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8439         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8440         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8441         .recover_open   = nfs4_open_reclaim,
8442         .recover_lock   = nfs4_lock_reclaim,
8443         .establish_clid = nfs4_init_clientid,
8444         .detect_trunking = nfs40_discover_server_trunking,
8445 };
8446
8447 #if defined(CONFIG_NFS_V4_1)
8448 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8449         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8450         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8451         .recover_open   = nfs4_open_reclaim,
8452         .recover_lock   = nfs4_lock_reclaim,
8453         .establish_clid = nfs41_init_clientid,
8454         .reclaim_complete = nfs41_proc_reclaim_complete,
8455         .detect_trunking = nfs41_discover_server_trunking,
8456 };
8457 #endif /* CONFIG_NFS_V4_1 */
8458
8459 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8460         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8461         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8462         .recover_open   = nfs40_open_expired,
8463         .recover_lock   = nfs4_lock_expired,
8464         .establish_clid = nfs4_init_clientid,
8465 };
8466
8467 #if defined(CONFIG_NFS_V4_1)
8468 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8469         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8470         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8471         .recover_open   = nfs41_open_expired,
8472         .recover_lock   = nfs41_lock_expired,
8473         .establish_clid = nfs41_init_clientid,
8474 };
8475 #endif /* CONFIG_NFS_V4_1 */
8476
8477 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8478         .sched_state_renewal = nfs4_proc_async_renew,
8479         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8480         .renew_lease = nfs4_proc_renew,
8481 };
8482
8483 #if defined(CONFIG_NFS_V4_1)
8484 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8485         .sched_state_renewal = nfs41_proc_async_sequence,
8486         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8487         .renew_lease = nfs4_proc_sequence,
8488 };
8489 #endif
8490
8491 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8492         .get_locations = _nfs40_proc_get_locations,
8493         .fsid_present = _nfs40_proc_fsid_present,
8494 };
8495
8496 #if defined(CONFIG_NFS_V4_1)
8497 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8498         .get_locations = _nfs41_proc_get_locations,
8499         .fsid_present = _nfs41_proc_fsid_present,
8500 };
8501 #endif  /* CONFIG_NFS_V4_1 */
8502
8503 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8504         .minor_version = 0,
8505         .init_caps = NFS_CAP_READDIRPLUS
8506                 | NFS_CAP_ATOMIC_OPEN
8507                 | NFS_CAP_CHANGE_ATTR
8508                 | NFS_CAP_POSIX_LOCK,
8509         .init_client = nfs40_init_client,
8510         .shutdown_client = nfs40_shutdown_client,
8511         .match_stateid = nfs4_match_stateid,
8512         .find_root_sec = nfs4_find_root_sec,
8513         .free_lock_state = nfs4_release_lockowner,
8514         .alloc_seqid = nfs_alloc_seqid,
8515         .call_sync_ops = &nfs40_call_sync_ops,
8516         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8517         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8518         .state_renewal_ops = &nfs40_state_renewal_ops,
8519         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8520 };
8521
8522 #if defined(CONFIG_NFS_V4_1)
8523 static struct nfs_seqid *
8524 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8525 {
8526         return NULL;
8527 }
8528
8529 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8530         .minor_version = 1,
8531         .init_caps = NFS_CAP_READDIRPLUS
8532                 | NFS_CAP_ATOMIC_OPEN
8533                 | NFS_CAP_CHANGE_ATTR
8534                 | NFS_CAP_POSIX_LOCK
8535                 | NFS_CAP_STATEID_NFSV41
8536                 | NFS_CAP_ATOMIC_OPEN_V1,
8537         .init_client = nfs41_init_client,
8538         .shutdown_client = nfs41_shutdown_client,
8539         .match_stateid = nfs41_match_stateid,
8540         .find_root_sec = nfs41_find_root_sec,
8541         .free_lock_state = nfs41_free_lock_state,
8542         .alloc_seqid = nfs_alloc_no_seqid,
8543         .call_sync_ops = &nfs41_call_sync_ops,
8544         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8545         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8546         .state_renewal_ops = &nfs41_state_renewal_ops,
8547         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8548 };
8549 #endif
8550
8551 #if defined(CONFIG_NFS_V4_2)
8552 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8553         .minor_version = 2,
8554         .init_caps = NFS_CAP_READDIRPLUS
8555                 | NFS_CAP_ATOMIC_OPEN
8556                 | NFS_CAP_CHANGE_ATTR
8557                 | NFS_CAP_POSIX_LOCK
8558                 | NFS_CAP_STATEID_NFSV41
8559                 | NFS_CAP_ATOMIC_OPEN_V1
8560                 | NFS_CAP_ALLOCATE
8561                 | NFS_CAP_DEALLOCATE
8562                 | NFS_CAP_SEEK,
8563         .init_client = nfs41_init_client,
8564         .shutdown_client = nfs41_shutdown_client,
8565         .match_stateid = nfs41_match_stateid,
8566         .find_root_sec = nfs41_find_root_sec,
8567         .free_lock_state = nfs41_free_lock_state,
8568         .call_sync_ops = &nfs41_call_sync_ops,
8569         .alloc_seqid = nfs_alloc_no_seqid,
8570         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8571         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8572         .state_renewal_ops = &nfs41_state_renewal_ops,
8573 };
8574 #endif
8575
8576 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8577         [0] = &nfs_v4_0_minor_ops,
8578 #if defined(CONFIG_NFS_V4_1)
8579         [1] = &nfs_v4_1_minor_ops,
8580 #endif
8581 #if defined(CONFIG_NFS_V4_2)
8582         [2] = &nfs_v4_2_minor_ops,
8583 #endif
8584 };
8585
8586 static const struct inode_operations nfs4_dir_inode_operations = {
8587         .create         = nfs_create,
8588         .lookup         = nfs_lookup,
8589         .atomic_open    = nfs_atomic_open,
8590         .link           = nfs_link,
8591         .unlink         = nfs_unlink,
8592         .symlink        = nfs_symlink,
8593         .mkdir          = nfs_mkdir,
8594         .rmdir          = nfs_rmdir,
8595         .mknod          = nfs_mknod,
8596         .rename         = nfs_rename,
8597         .permission     = nfs_permission,
8598         .getattr        = nfs_getattr,
8599         .setattr        = nfs_setattr,
8600         .getxattr       = generic_getxattr,
8601         .setxattr       = generic_setxattr,
8602         .listxattr      = generic_listxattr,
8603         .removexattr    = generic_removexattr,
8604 };
8605
8606 static const struct inode_operations nfs4_file_inode_operations = {
8607         .permission     = nfs_permission,
8608         .getattr        = nfs_getattr,
8609         .setattr        = nfs_setattr,
8610         .getxattr       = generic_getxattr,
8611         .setxattr       = generic_setxattr,
8612         .listxattr      = generic_listxattr,
8613         .removexattr    = generic_removexattr,
8614 };
8615
8616 const struct nfs_rpc_ops nfs_v4_clientops = {
8617         .version        = 4,                    /* protocol version */
8618         .dentry_ops     = &nfs4_dentry_operations,
8619         .dir_inode_ops  = &nfs4_dir_inode_operations,
8620         .file_inode_ops = &nfs4_file_inode_operations,
8621         .file_ops       = &nfs4_file_operations,
8622         .getroot        = nfs4_proc_get_root,
8623         .submount       = nfs4_submount,
8624         .try_mount      = nfs4_try_mount,
8625         .getattr        = nfs4_proc_getattr,
8626         .setattr        = nfs4_proc_setattr,
8627         .lookup         = nfs4_proc_lookup,
8628         .access         = nfs4_proc_access,
8629         .readlink       = nfs4_proc_readlink,
8630         .create         = nfs4_proc_create,
8631         .remove         = nfs4_proc_remove,
8632         .unlink_setup   = nfs4_proc_unlink_setup,
8633         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8634         .unlink_done    = nfs4_proc_unlink_done,
8635         .rename_setup   = nfs4_proc_rename_setup,
8636         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8637         .rename_done    = nfs4_proc_rename_done,
8638         .link           = nfs4_proc_link,
8639         .symlink        = nfs4_proc_symlink,
8640         .mkdir          = nfs4_proc_mkdir,
8641         .rmdir          = nfs4_proc_remove,
8642         .readdir        = nfs4_proc_readdir,
8643         .mknod          = nfs4_proc_mknod,
8644         .statfs         = nfs4_proc_statfs,
8645         .fsinfo         = nfs4_proc_fsinfo,
8646         .pathconf       = nfs4_proc_pathconf,
8647         .set_capabilities = nfs4_server_capabilities,
8648         .decode_dirent  = nfs4_decode_dirent,
8649         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8650         .read_setup     = nfs4_proc_read_setup,
8651         .read_done      = nfs4_read_done,
8652         .write_setup    = nfs4_proc_write_setup,
8653         .write_done     = nfs4_write_done,
8654         .commit_setup   = nfs4_proc_commit_setup,
8655         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8656         .commit_done    = nfs4_commit_done,
8657         .lock           = nfs4_proc_lock,
8658         .clear_acl_cache = nfs4_zap_acl_attr,
8659         .close_context  = nfs4_close_context,
8660         .open_context   = nfs4_atomic_open,
8661         .have_delegation = nfs4_have_delegation,
8662         .return_delegation = nfs4_inode_return_delegation,
8663         .alloc_client   = nfs4_alloc_client,
8664         .init_client    = nfs4_init_client,
8665         .free_client    = nfs4_free_client,
8666         .create_server  = nfs4_create_server,
8667         .clone_server   = nfs_clone_server,
8668 };
8669
8670 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8671         .prefix = XATTR_NAME_NFSV4_ACL,
8672         .list   = nfs4_xattr_list_nfs4_acl,
8673         .get    = nfs4_xattr_get_nfs4_acl,
8674         .set    = nfs4_xattr_set_nfs4_acl,
8675 };
8676
8677 const struct xattr_handler *nfs4_xattr_handlers[] = {
8678         &nfs4_xattr_nfs4_acl_handler,
8679 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8680         &nfs4_xattr_nfs4_label_handler,
8681 #endif
8682         NULL
8683 };
8684
8685 /*
8686  * Local variables:
8687  *  c-basic-offset: 8
8688  * End:
8689  */