]> git.kernelconcepts.de Git - karo-tx-linux.git/blob - fs/nfs/nfs4proc.c
Merge branch 'next' into for-linus
[karo-tx-linux.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/file.h>
42 #include <linux/string.h>
43 #include <linux/ratelimit.h>
44 #include <linux/printk.h>
45 #include <linux/slab.h>
46 #include <linux/sunrpc/clnt.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69
70 #include "nfs4trace.h"
71
72 #define NFSDBG_FACILITY         NFSDBG_PROC
73
74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
76
77 struct nfs4_opendata;
78 static int _nfs4_proc_open(struct nfs4_opendata *data);
79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
83 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
84 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
85 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
86                             struct nfs_fattr *fattr, struct iattr *sattr,
87                             struct nfs4_state *state, struct nfs4_label *ilabel,
88                             struct nfs4_label *olabel);
89 #ifdef CONFIG_NFS_V4_1
90 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
91                 struct rpc_cred *);
92 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
93                 struct rpc_cred *);
94 #endif
95
96 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
97 static inline struct nfs4_label *
98 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
99         struct iattr *sattr, struct nfs4_label *label)
100 {
101         int err;
102
103         if (label == NULL)
104                 return NULL;
105
106         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
107                 return NULL;
108
109         err = security_dentry_init_security(dentry, sattr->ia_mode,
110                                 &dentry->d_name, (void **)&label->label, &label->len);
111         if (err == 0)
112                 return label;
113
114         return NULL;
115 }
116 static inline void
117 nfs4_label_release_security(struct nfs4_label *label)
118 {
119         if (label)
120                 security_release_secctx(label->label, label->len);
121 }
122 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
123 {
124         if (label)
125                 return server->attr_bitmask;
126
127         return server->attr_bitmask_nl;
128 }
129 #else
130 static inline struct nfs4_label *
131 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
132         struct iattr *sattr, struct nfs4_label *l)
133 { return NULL; }
134 static inline void
135 nfs4_label_release_security(struct nfs4_label *label)
136 { return; }
137 static inline u32 *
138 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
139 { return server->attr_bitmask; }
140 #endif
141
142 /* Prevent leaks of NFSv4 errors into userland */
143 static int nfs4_map_errors(int err)
144 {
145         if (err >= -1000)
146                 return err;
147         switch (err) {
148         case -NFS4ERR_RESOURCE:
149         case -NFS4ERR_LAYOUTTRYLATER:
150         case -NFS4ERR_RECALLCONFLICT:
151                 return -EREMOTEIO;
152         case -NFS4ERR_WRONGSEC:
153         case -NFS4ERR_WRONG_CRED:
154                 return -EPERM;
155         case -NFS4ERR_BADOWNER:
156         case -NFS4ERR_BADNAME:
157                 return -EINVAL;
158         case -NFS4ERR_SHARE_DENIED:
159                 return -EACCES;
160         case -NFS4ERR_MINOR_VERS_MISMATCH:
161                 return -EPROTONOSUPPORT;
162         case -NFS4ERR_FILE_OPEN:
163                 return -EBUSY;
164         default:
165                 dprintk("%s could not handle NFSv4 error %d\n",
166                                 __func__, -err);
167                 break;
168         }
169         return -EIO;
170 }
171
172 /*
173  * This is our standard bitmap for GETATTR requests.
174  */
175 const u32 nfs4_fattr_bitmap[3] = {
176         FATTR4_WORD0_TYPE
177         | FATTR4_WORD0_CHANGE
178         | FATTR4_WORD0_SIZE
179         | FATTR4_WORD0_FSID
180         | FATTR4_WORD0_FILEID,
181         FATTR4_WORD1_MODE
182         | FATTR4_WORD1_NUMLINKS
183         | FATTR4_WORD1_OWNER
184         | FATTR4_WORD1_OWNER_GROUP
185         | FATTR4_WORD1_RAWDEV
186         | FATTR4_WORD1_SPACE_USED
187         | FATTR4_WORD1_TIME_ACCESS
188         | FATTR4_WORD1_TIME_METADATA
189         | FATTR4_WORD1_TIME_MODIFY
190         | FATTR4_WORD1_MOUNTED_ON_FILEID,
191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
192         FATTR4_WORD2_SECURITY_LABEL
193 #endif
194 };
195
196 static const u32 nfs4_pnfs_open_bitmap[3] = {
197         FATTR4_WORD0_TYPE
198         | FATTR4_WORD0_CHANGE
199         | FATTR4_WORD0_SIZE
200         | FATTR4_WORD0_FSID
201         | FATTR4_WORD0_FILEID,
202         FATTR4_WORD1_MODE
203         | FATTR4_WORD1_NUMLINKS
204         | FATTR4_WORD1_OWNER
205         | FATTR4_WORD1_OWNER_GROUP
206         | FATTR4_WORD1_RAWDEV
207         | FATTR4_WORD1_SPACE_USED
208         | FATTR4_WORD1_TIME_ACCESS
209         | FATTR4_WORD1_TIME_METADATA
210         | FATTR4_WORD1_TIME_MODIFY,
211         FATTR4_WORD2_MDSTHRESHOLD
212 };
213
214 static const u32 nfs4_open_noattr_bitmap[3] = {
215         FATTR4_WORD0_TYPE
216         | FATTR4_WORD0_CHANGE
217         | FATTR4_WORD0_FILEID,
218 };
219
220 const u32 nfs4_statfs_bitmap[3] = {
221         FATTR4_WORD0_FILES_AVAIL
222         | FATTR4_WORD0_FILES_FREE
223         | FATTR4_WORD0_FILES_TOTAL,
224         FATTR4_WORD1_SPACE_AVAIL
225         | FATTR4_WORD1_SPACE_FREE
226         | FATTR4_WORD1_SPACE_TOTAL
227 };
228
229 const u32 nfs4_pathconf_bitmap[3] = {
230         FATTR4_WORD0_MAXLINK
231         | FATTR4_WORD0_MAXNAME,
232         0
233 };
234
235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
236                         | FATTR4_WORD0_MAXREAD
237                         | FATTR4_WORD0_MAXWRITE
238                         | FATTR4_WORD0_LEASE_TIME,
239                         FATTR4_WORD1_TIME_DELTA
240                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
241                         FATTR4_WORD2_LAYOUT_BLKSIZE
242 };
243
244 const u32 nfs4_fs_locations_bitmap[3] = {
245         FATTR4_WORD0_TYPE
246         | FATTR4_WORD0_CHANGE
247         | FATTR4_WORD0_SIZE
248         | FATTR4_WORD0_FSID
249         | FATTR4_WORD0_FILEID
250         | FATTR4_WORD0_FS_LOCATIONS,
251         FATTR4_WORD1_MODE
252         | FATTR4_WORD1_NUMLINKS
253         | FATTR4_WORD1_OWNER
254         | FATTR4_WORD1_OWNER_GROUP
255         | FATTR4_WORD1_RAWDEV
256         | FATTR4_WORD1_SPACE_USED
257         | FATTR4_WORD1_TIME_ACCESS
258         | FATTR4_WORD1_TIME_METADATA
259         | FATTR4_WORD1_TIME_MODIFY
260         | FATTR4_WORD1_MOUNTED_ON_FILEID,
261 };
262
263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
264                 struct nfs4_readdir_arg *readdir)
265 {
266         __be32 *start, *p;
267
268         if (cookie > 2) {
269                 readdir->cookie = cookie;
270                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
271                 return;
272         }
273
274         readdir->cookie = 0;
275         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
276         if (cookie == 2)
277                 return;
278         
279         /*
280          * NFSv4 servers do not return entries for '.' and '..'
281          * Therefore, we fake these entries here.  We let '.'
282          * have cookie 0 and '..' have cookie 1.  Note that
283          * when talking to the server, we always send cookie 0
284          * instead of 1 or 2.
285          */
286         start = p = kmap_atomic(*readdir->pages);
287         
288         if (cookie == 0) {
289                 *p++ = xdr_one;                                  /* next */
290                 *p++ = xdr_zero;                   /* cookie, first word */
291                 *p++ = xdr_one;                   /* cookie, second word */
292                 *p++ = xdr_one;                             /* entry len */
293                 memcpy(p, ".\0\0\0", 4);                        /* entry */
294                 p++;
295                 *p++ = xdr_one;                         /* bitmap length */
296                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
297                 *p++ = htonl(8);              /* attribute buffer length */
298                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
299         }
300         
301         *p++ = xdr_one;                                  /* next */
302         *p++ = xdr_zero;                   /* cookie, first word */
303         *p++ = xdr_two;                   /* cookie, second word */
304         *p++ = xdr_two;                             /* entry len */
305         memcpy(p, "..\0\0", 4);                         /* entry */
306         p++;
307         *p++ = xdr_one;                         /* bitmap length */
308         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
309         *p++ = htonl(8);              /* attribute buffer length */
310         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
311
312         readdir->pgbase = (char *)p - (char *)start;
313         readdir->count -= readdir->pgbase;
314         kunmap_atomic(start);
315 }
316
317 static long nfs4_update_delay(long *timeout)
318 {
319         long ret;
320         if (!timeout)
321                 return NFS4_POLL_RETRY_MAX;
322         if (*timeout <= 0)
323                 *timeout = NFS4_POLL_RETRY_MIN;
324         if (*timeout > NFS4_POLL_RETRY_MAX)
325                 *timeout = NFS4_POLL_RETRY_MAX;
326         ret = *timeout;
327         *timeout <<= 1;
328         return ret;
329 }
330
331 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
332 {
333         int res = 0;
334
335         might_sleep();
336
337         freezable_schedule_timeout_killable_unsafe(
338                 nfs4_update_delay(timeout));
339         if (fatal_signal_pending(current))
340                 res = -ERESTARTSYS;
341         return res;
342 }
343
344 /* This is the error handling routine for processes that are allowed
345  * to sleep.
346  */
347 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
348 {
349         struct nfs_client *clp = server->nfs_client;
350         struct nfs4_state *state = exception->state;
351         struct inode *inode = exception->inode;
352         int ret = errorcode;
353
354         exception->retry = 0;
355         switch(errorcode) {
356                 case 0:
357                         return 0;
358                 case -NFS4ERR_OPENMODE:
359                 case -NFS4ERR_DELEG_REVOKED:
360                 case -NFS4ERR_ADMIN_REVOKED:
361                 case -NFS4ERR_BAD_STATEID:
362                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
363                                 nfs4_inode_return_delegation(inode);
364                                 exception->retry = 1;
365                                 return 0;
366                         }
367                         if (state == NULL)
368                                 break;
369                         ret = nfs4_schedule_stateid_recovery(server, state);
370                         if (ret < 0)
371                                 break;
372                         goto wait_on_recovery;
373                 case -NFS4ERR_EXPIRED:
374                         if (state != NULL) {
375                                 ret = nfs4_schedule_stateid_recovery(server, state);
376                                 if (ret < 0)
377                                         break;
378                         }
379                 case -NFS4ERR_STALE_STATEID:
380                 case -NFS4ERR_STALE_CLIENTID:
381                         nfs4_schedule_lease_recovery(clp);
382                         goto wait_on_recovery;
383                 case -NFS4ERR_MOVED:
384                         ret = nfs4_schedule_migration_recovery(server);
385                         if (ret < 0)
386                                 break;
387                         goto wait_on_recovery;
388                 case -NFS4ERR_LEASE_MOVED:
389                         nfs4_schedule_lease_moved_recovery(clp);
390                         goto wait_on_recovery;
391 #if defined(CONFIG_NFS_V4_1)
392                 case -NFS4ERR_BADSESSION:
393                 case -NFS4ERR_BADSLOT:
394                 case -NFS4ERR_BAD_HIGH_SLOT:
395                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
396                 case -NFS4ERR_DEADSESSION:
397                 case -NFS4ERR_SEQ_FALSE_RETRY:
398                 case -NFS4ERR_SEQ_MISORDERED:
399                         dprintk("%s ERROR: %d Reset session\n", __func__,
400                                 errorcode);
401                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
402                         goto wait_on_recovery;
403 #endif /* defined(CONFIG_NFS_V4_1) */
404                 case -NFS4ERR_FILE_OPEN:
405                         if (exception->timeout > HZ) {
406                                 /* We have retried a decent amount, time to
407                                  * fail
408                                  */
409                                 ret = -EBUSY;
410                                 break;
411                         }
412                 case -NFS4ERR_GRACE:
413                 case -NFS4ERR_DELAY:
414                         ret = nfs4_delay(server->client, &exception->timeout);
415                         if (ret != 0)
416                                 break;
417                 case -NFS4ERR_RETRY_UNCACHED_REP:
418                 case -NFS4ERR_OLD_STATEID:
419                         exception->retry = 1;
420                         break;
421                 case -NFS4ERR_BADOWNER:
422                         /* The following works around a Linux server bug! */
423                 case -NFS4ERR_BADNAME:
424                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
425                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
426                                 exception->retry = 1;
427                                 printk(KERN_WARNING "NFS: v4 server %s "
428                                                 "does not accept raw "
429                                                 "uid/gids. "
430                                                 "Reenabling the idmapper.\n",
431                                                 server->nfs_client->cl_hostname);
432                         }
433         }
434         /* We failed to handle the error */
435         return nfs4_map_errors(ret);
436 wait_on_recovery:
437         ret = nfs4_wait_clnt_recover(clp);
438         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
439                 return -EIO;
440         if (ret == 0)
441                 exception->retry = 1;
442         return ret;
443 }
444
445 /*
446  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
447  * or 'false' otherwise.
448  */
449 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
450 {
451         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
452
453         if (flavor == RPC_AUTH_GSS_KRB5I ||
454             flavor == RPC_AUTH_GSS_KRB5P)
455                 return true;
456
457         return false;
458 }
459
460 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
461 {
462         spin_lock(&clp->cl_lock);
463         if (time_before(clp->cl_last_renewal,timestamp))
464                 clp->cl_last_renewal = timestamp;
465         spin_unlock(&clp->cl_lock);
466 }
467
468 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
469 {
470         struct nfs_client *clp = server->nfs_client;
471
472         if (!nfs4_has_session(clp))
473                 do_renew_lease(clp, timestamp);
474 }
475
476 struct nfs4_call_sync_data {
477         const struct nfs_server *seq_server;
478         struct nfs4_sequence_args *seq_args;
479         struct nfs4_sequence_res *seq_res;
480 };
481
482 void nfs4_init_sequence(struct nfs4_sequence_args *args,
483                         struct nfs4_sequence_res *res, int cache_reply)
484 {
485         args->sa_slot = NULL;
486         args->sa_cache_this = cache_reply;
487         args->sa_privileged = 0;
488
489         res->sr_slot = NULL;
490 }
491
492 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
493 {
494         args->sa_privileged = 1;
495 }
496
497 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
498                          struct nfs4_sequence_args *args,
499                          struct nfs4_sequence_res *res,
500                          struct rpc_task *task)
501 {
502         struct nfs4_slot *slot;
503
504         /* slot already allocated? */
505         if (res->sr_slot != NULL)
506                 goto out_start;
507
508         spin_lock(&tbl->slot_tbl_lock);
509         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
510                 goto out_sleep;
511
512         slot = nfs4_alloc_slot(tbl);
513         if (IS_ERR(slot)) {
514                 if (slot == ERR_PTR(-ENOMEM))
515                         task->tk_timeout = HZ >> 2;
516                 goto out_sleep;
517         }
518         spin_unlock(&tbl->slot_tbl_lock);
519
520         args->sa_slot = slot;
521         res->sr_slot = slot;
522
523 out_start:
524         rpc_call_start(task);
525         return 0;
526
527 out_sleep:
528         if (args->sa_privileged)
529                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
530                                 NULL, RPC_PRIORITY_PRIVILEGED);
531         else
532                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
533         spin_unlock(&tbl->slot_tbl_lock);
534         return -EAGAIN;
535 }
536 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
537
538 static int nfs40_sequence_done(struct rpc_task *task,
539                                struct nfs4_sequence_res *res)
540 {
541         struct nfs4_slot *slot = res->sr_slot;
542         struct nfs4_slot_table *tbl;
543
544         if (slot == NULL)
545                 goto out;
546
547         tbl = slot->table;
548         spin_lock(&tbl->slot_tbl_lock);
549         if (!nfs41_wake_and_assign_slot(tbl, slot))
550                 nfs4_free_slot(tbl, slot);
551         spin_unlock(&tbl->slot_tbl_lock);
552
553         res->sr_slot = NULL;
554 out:
555         return 1;
556 }
557
558 #if defined(CONFIG_NFS_V4_1)
559
560 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
561 {
562         struct nfs4_session *session;
563         struct nfs4_slot_table *tbl;
564         struct nfs4_slot *slot = res->sr_slot;
565         bool send_new_highest_used_slotid = false;
566
567         tbl = slot->table;
568         session = tbl->session;
569
570         spin_lock(&tbl->slot_tbl_lock);
571         /* Be nice to the server: try to ensure that the last transmitted
572          * value for highest_user_slotid <= target_highest_slotid
573          */
574         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
575                 send_new_highest_used_slotid = true;
576
577         if (nfs41_wake_and_assign_slot(tbl, slot)) {
578                 send_new_highest_used_slotid = false;
579                 goto out_unlock;
580         }
581         nfs4_free_slot(tbl, slot);
582
583         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
584                 send_new_highest_used_slotid = false;
585 out_unlock:
586         spin_unlock(&tbl->slot_tbl_lock);
587         res->sr_slot = NULL;
588         if (send_new_highest_used_slotid)
589                 nfs41_notify_server(session->clp);
590 }
591
592 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
593 {
594         struct nfs4_session *session;
595         struct nfs4_slot *slot = res->sr_slot;
596         struct nfs_client *clp;
597         bool interrupted = false;
598         int ret = 1;
599
600         if (slot == NULL)
601                 goto out_noaction;
602         /* don't increment the sequence number if the task wasn't sent */
603         if (!RPC_WAS_SENT(task))
604                 goto out;
605
606         session = slot->table->session;
607
608         if (slot->interrupted) {
609                 slot->interrupted = 0;
610                 interrupted = true;
611         }
612
613         trace_nfs4_sequence_done(session, res);
614         /* Check the SEQUENCE operation status */
615         switch (res->sr_status) {
616         case 0:
617                 /* Update the slot's sequence and clientid lease timer */
618                 ++slot->seq_nr;
619                 clp = session->clp;
620                 do_renew_lease(clp, res->sr_timestamp);
621                 /* Check sequence flags */
622                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
623                 nfs41_update_target_slotid(slot->table, slot, res);
624                 break;
625         case 1:
626                 /*
627                  * sr_status remains 1 if an RPC level error occurred.
628                  * The server may or may not have processed the sequence
629                  * operation..
630                  * Mark the slot as having hosted an interrupted RPC call.
631                  */
632                 slot->interrupted = 1;
633                 goto out;
634         case -NFS4ERR_DELAY:
635                 /* The server detected a resend of the RPC call and
636                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
637                  * of RFC5661.
638                  */
639                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
640                         __func__,
641                         slot->slot_nr,
642                         slot->seq_nr);
643                 goto out_retry;
644         case -NFS4ERR_BADSLOT:
645                 /*
646                  * The slot id we used was probably retired. Try again
647                  * using a different slot id.
648                  */
649                 goto retry_nowait;
650         case -NFS4ERR_SEQ_MISORDERED:
651                 /*
652                  * Was the last operation on this sequence interrupted?
653                  * If so, retry after bumping the sequence number.
654                  */
655                 if (interrupted) {
656                         ++slot->seq_nr;
657                         goto retry_nowait;
658                 }
659                 /*
660                  * Could this slot have been previously retired?
661                  * If so, then the server may be expecting seq_nr = 1!
662                  */
663                 if (slot->seq_nr != 1) {
664                         slot->seq_nr = 1;
665                         goto retry_nowait;
666                 }
667                 break;
668         case -NFS4ERR_SEQ_FALSE_RETRY:
669                 ++slot->seq_nr;
670                 goto retry_nowait;
671         default:
672                 /* Just update the slot sequence no. */
673                 ++slot->seq_nr;
674         }
675 out:
676         /* The session may be reset by one of the error handlers. */
677         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
678         nfs41_sequence_free_slot(res);
679 out_noaction:
680         return ret;
681 retry_nowait:
682         if (rpc_restart_call_prepare(task)) {
683                 task->tk_status = 0;
684                 ret = 0;
685         }
686         goto out;
687 out_retry:
688         if (!rpc_restart_call(task))
689                 goto out;
690         rpc_delay(task, NFS4_POLL_RETRY_MAX);
691         return 0;
692 }
693 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
694
695 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
696 {
697         if (res->sr_slot == NULL)
698                 return 1;
699         if (!res->sr_slot->table->session)
700                 return nfs40_sequence_done(task, res);
701         return nfs41_sequence_done(task, res);
702 }
703 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
704
705 int nfs41_setup_sequence(struct nfs4_session *session,
706                                 struct nfs4_sequence_args *args,
707                                 struct nfs4_sequence_res *res,
708                                 struct rpc_task *task)
709 {
710         struct nfs4_slot *slot;
711         struct nfs4_slot_table *tbl;
712
713         dprintk("--> %s\n", __func__);
714         /* slot already allocated? */
715         if (res->sr_slot != NULL)
716                 goto out_success;
717
718         tbl = &session->fc_slot_table;
719
720         task->tk_timeout = 0;
721
722         spin_lock(&tbl->slot_tbl_lock);
723         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
724             !args->sa_privileged) {
725                 /* The state manager will wait until the slot table is empty */
726                 dprintk("%s session is draining\n", __func__);
727                 goto out_sleep;
728         }
729
730         slot = nfs4_alloc_slot(tbl);
731         if (IS_ERR(slot)) {
732                 /* If out of memory, try again in 1/4 second */
733                 if (slot == ERR_PTR(-ENOMEM))
734                         task->tk_timeout = HZ >> 2;
735                 dprintk("<-- %s: no free slots\n", __func__);
736                 goto out_sleep;
737         }
738         spin_unlock(&tbl->slot_tbl_lock);
739
740         args->sa_slot = slot;
741
742         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
743                         slot->slot_nr, slot->seq_nr);
744
745         res->sr_slot = slot;
746         res->sr_timestamp = jiffies;
747         res->sr_status_flags = 0;
748         /*
749          * sr_status is only set in decode_sequence, and so will remain
750          * set to 1 if an rpc level failure occurs.
751          */
752         res->sr_status = 1;
753         trace_nfs4_setup_sequence(session, args);
754 out_success:
755         rpc_call_start(task);
756         return 0;
757 out_sleep:
758         /* Privileged tasks are queued with top priority */
759         if (args->sa_privileged)
760                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
761                                 NULL, RPC_PRIORITY_PRIVILEGED);
762         else
763                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
764         spin_unlock(&tbl->slot_tbl_lock);
765         return -EAGAIN;
766 }
767 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
768
769 static int nfs4_setup_sequence(const struct nfs_server *server,
770                                struct nfs4_sequence_args *args,
771                                struct nfs4_sequence_res *res,
772                                struct rpc_task *task)
773 {
774         struct nfs4_session *session = nfs4_get_session(server);
775         int ret = 0;
776
777         if (!session)
778                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
779                                             args, res, task);
780
781         dprintk("--> %s clp %p session %p sr_slot %u\n",
782                 __func__, session->clp, session, res->sr_slot ?
783                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
784
785         ret = nfs41_setup_sequence(session, args, res, task);
786
787         dprintk("<-- %s status=%d\n", __func__, ret);
788         return ret;
789 }
790
791 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
792 {
793         struct nfs4_call_sync_data *data = calldata;
794         struct nfs4_session *session = nfs4_get_session(data->seq_server);
795
796         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
797
798         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
799 }
800
801 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
802 {
803         struct nfs4_call_sync_data *data = calldata;
804
805         nfs41_sequence_done(task, data->seq_res);
806 }
807
808 static const struct rpc_call_ops nfs41_call_sync_ops = {
809         .rpc_call_prepare = nfs41_call_sync_prepare,
810         .rpc_call_done = nfs41_call_sync_done,
811 };
812
813 #else   /* !CONFIG_NFS_V4_1 */
814
815 static int nfs4_setup_sequence(const struct nfs_server *server,
816                                struct nfs4_sequence_args *args,
817                                struct nfs4_sequence_res *res,
818                                struct rpc_task *task)
819 {
820         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
821                                     args, res, task);
822 }
823
824 int nfs4_sequence_done(struct rpc_task *task,
825                        struct nfs4_sequence_res *res)
826 {
827         return nfs40_sequence_done(task, res);
828 }
829 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
830
831 #endif  /* !CONFIG_NFS_V4_1 */
832
833 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
834 {
835         struct nfs4_call_sync_data *data = calldata;
836         nfs4_setup_sequence(data->seq_server,
837                                 data->seq_args, data->seq_res, task);
838 }
839
840 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
841 {
842         struct nfs4_call_sync_data *data = calldata;
843         nfs4_sequence_done(task, data->seq_res);
844 }
845
846 static const struct rpc_call_ops nfs40_call_sync_ops = {
847         .rpc_call_prepare = nfs40_call_sync_prepare,
848         .rpc_call_done = nfs40_call_sync_done,
849 };
850
851 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
852                                    struct nfs_server *server,
853                                    struct rpc_message *msg,
854                                    struct nfs4_sequence_args *args,
855                                    struct nfs4_sequence_res *res)
856 {
857         int ret;
858         struct rpc_task *task;
859         struct nfs_client *clp = server->nfs_client;
860         struct nfs4_call_sync_data data = {
861                 .seq_server = server,
862                 .seq_args = args,
863                 .seq_res = res,
864         };
865         struct rpc_task_setup task_setup = {
866                 .rpc_client = clnt,
867                 .rpc_message = msg,
868                 .callback_ops = clp->cl_mvops->call_sync_ops,
869                 .callback_data = &data
870         };
871
872         task = rpc_run_task(&task_setup);
873         if (IS_ERR(task))
874                 ret = PTR_ERR(task);
875         else {
876                 ret = task->tk_status;
877                 rpc_put_task(task);
878         }
879         return ret;
880 }
881
882 int nfs4_call_sync(struct rpc_clnt *clnt,
883                    struct nfs_server *server,
884                    struct rpc_message *msg,
885                    struct nfs4_sequence_args *args,
886                    struct nfs4_sequence_res *res,
887                    int cache_reply)
888 {
889         nfs4_init_sequence(args, res, cache_reply);
890         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
891 }
892
893 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
894 {
895         struct nfs_inode *nfsi = NFS_I(dir);
896
897         spin_lock(&dir->i_lock);
898         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
899         if (!cinfo->atomic || cinfo->before != dir->i_version)
900                 nfs_force_lookup_revalidate(dir);
901         dir->i_version = cinfo->after;
902         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
903         nfs_fscache_invalidate(dir);
904         spin_unlock(&dir->i_lock);
905 }
906
907 struct nfs4_opendata {
908         struct kref kref;
909         struct nfs_openargs o_arg;
910         struct nfs_openres o_res;
911         struct nfs_open_confirmargs c_arg;
912         struct nfs_open_confirmres c_res;
913         struct nfs4_string owner_name;
914         struct nfs4_string group_name;
915         struct nfs4_label *a_label;
916         struct nfs_fattr f_attr;
917         struct nfs4_label *f_label;
918         struct dentry *dir;
919         struct dentry *dentry;
920         struct nfs4_state_owner *owner;
921         struct nfs4_state *state;
922         struct iattr attrs;
923         unsigned long timestamp;
924         unsigned int rpc_done : 1;
925         unsigned int file_created : 1;
926         unsigned int is_recover : 1;
927         int rpc_status;
928         int cancelled;
929 };
930
931 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
932                 int err, struct nfs4_exception *exception)
933 {
934         if (err != -EINVAL)
935                 return false;
936         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
937                 return false;
938         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
939         exception->retry = 1;
940         return true;
941 }
942
943 static u32
944 nfs4_map_atomic_open_share(struct nfs_server *server,
945                 fmode_t fmode, int openflags)
946 {
947         u32 res = 0;
948
949         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
950         case FMODE_READ:
951                 res = NFS4_SHARE_ACCESS_READ;
952                 break;
953         case FMODE_WRITE:
954                 res = NFS4_SHARE_ACCESS_WRITE;
955                 break;
956         case FMODE_READ|FMODE_WRITE:
957                 res = NFS4_SHARE_ACCESS_BOTH;
958         }
959         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
960                 goto out;
961         /* Want no delegation if we're using O_DIRECT */
962         if (openflags & O_DIRECT)
963                 res |= NFS4_SHARE_WANT_NO_DELEG;
964 out:
965         return res;
966 }
967
968 static enum open_claim_type4
969 nfs4_map_atomic_open_claim(struct nfs_server *server,
970                 enum open_claim_type4 claim)
971 {
972         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
973                 return claim;
974         switch (claim) {
975         default:
976                 return claim;
977         case NFS4_OPEN_CLAIM_FH:
978                 return NFS4_OPEN_CLAIM_NULL;
979         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
980                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
981         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
982                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
983         }
984 }
985
986 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
987 {
988         p->o_res.f_attr = &p->f_attr;
989         p->o_res.f_label = p->f_label;
990         p->o_res.seqid = p->o_arg.seqid;
991         p->c_res.seqid = p->c_arg.seqid;
992         p->o_res.server = p->o_arg.server;
993         p->o_res.access_request = p->o_arg.access;
994         nfs_fattr_init(&p->f_attr);
995         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
996 }
997
998 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
999                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1000                 const struct iattr *attrs,
1001                 struct nfs4_label *label,
1002                 enum open_claim_type4 claim,
1003                 gfp_t gfp_mask)
1004 {
1005         struct dentry *parent = dget_parent(dentry);
1006         struct inode *dir = d_inode(parent);
1007         struct nfs_server *server = NFS_SERVER(dir);
1008         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1009         struct nfs4_opendata *p;
1010
1011         p = kzalloc(sizeof(*p), gfp_mask);
1012         if (p == NULL)
1013                 goto err;
1014
1015         p->f_label = nfs4_label_alloc(server, gfp_mask);
1016         if (IS_ERR(p->f_label))
1017                 goto err_free_p;
1018
1019         p->a_label = nfs4_label_alloc(server, gfp_mask);
1020         if (IS_ERR(p->a_label))
1021                 goto err_free_f;
1022
1023         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1024         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1025         if (IS_ERR(p->o_arg.seqid))
1026                 goto err_free_label;
1027         nfs_sb_active(dentry->d_sb);
1028         p->dentry = dget(dentry);
1029         p->dir = parent;
1030         p->owner = sp;
1031         atomic_inc(&sp->so_count);
1032         p->o_arg.open_flags = flags;
1033         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1034         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1035                         fmode, flags);
1036         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1037          * will return permission denied for all bits until close */
1038         if (!(flags & O_EXCL)) {
1039                 /* ask server to check for all possible rights as results
1040                  * are cached */
1041                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1042                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1043         }
1044         p->o_arg.clientid = server->nfs_client->cl_clientid;
1045         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1046         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1047         p->o_arg.name = &dentry->d_name;
1048         p->o_arg.server = server;
1049         p->o_arg.bitmask = nfs4_bitmask(server, label);
1050         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1051         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1052         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1053         switch (p->o_arg.claim) {
1054         case NFS4_OPEN_CLAIM_NULL:
1055         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1056         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1057                 p->o_arg.fh = NFS_FH(dir);
1058                 break;
1059         case NFS4_OPEN_CLAIM_PREVIOUS:
1060         case NFS4_OPEN_CLAIM_FH:
1061         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1062         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1063                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1064         }
1065         if (attrs != NULL && attrs->ia_valid != 0) {
1066                 __u32 verf[2];
1067
1068                 p->o_arg.u.attrs = &p->attrs;
1069                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1070
1071                 verf[0] = jiffies;
1072                 verf[1] = current->pid;
1073                 memcpy(p->o_arg.u.verifier.data, verf,
1074                                 sizeof(p->o_arg.u.verifier.data));
1075         }
1076         p->c_arg.fh = &p->o_res.fh;
1077         p->c_arg.stateid = &p->o_res.stateid;
1078         p->c_arg.seqid = p->o_arg.seqid;
1079         nfs4_init_opendata_res(p);
1080         kref_init(&p->kref);
1081         return p;
1082
1083 err_free_label:
1084         nfs4_label_free(p->a_label);
1085 err_free_f:
1086         nfs4_label_free(p->f_label);
1087 err_free_p:
1088         kfree(p);
1089 err:
1090         dput(parent);
1091         return NULL;
1092 }
1093
1094 static void nfs4_opendata_free(struct kref *kref)
1095 {
1096         struct nfs4_opendata *p = container_of(kref,
1097                         struct nfs4_opendata, kref);
1098         struct super_block *sb = p->dentry->d_sb;
1099
1100         nfs_free_seqid(p->o_arg.seqid);
1101         if (p->state != NULL)
1102                 nfs4_put_open_state(p->state);
1103         nfs4_put_state_owner(p->owner);
1104
1105         nfs4_label_free(p->a_label);
1106         nfs4_label_free(p->f_label);
1107
1108         dput(p->dir);
1109         dput(p->dentry);
1110         nfs_sb_deactive(sb);
1111         nfs_fattr_free_names(&p->f_attr);
1112         kfree(p->f_attr.mdsthreshold);
1113         kfree(p);
1114 }
1115
1116 static void nfs4_opendata_put(struct nfs4_opendata *p)
1117 {
1118         if (p != NULL)
1119                 kref_put(&p->kref, nfs4_opendata_free);
1120 }
1121
1122 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1123 {
1124         int ret;
1125
1126         ret = rpc_wait_for_completion_task(task);
1127         return ret;
1128 }
1129
1130 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1131                 fmode_t fmode)
1132 {
1133         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1134         case FMODE_READ|FMODE_WRITE:
1135                 return state->n_rdwr != 0;
1136         case FMODE_WRITE:
1137                 return state->n_wronly != 0;
1138         case FMODE_READ:
1139                 return state->n_rdonly != 0;
1140         }
1141         WARN_ON_ONCE(1);
1142         return false;
1143 }
1144
1145 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1146 {
1147         int ret = 0;
1148
1149         if (open_mode & (O_EXCL|O_TRUNC))
1150                 goto out;
1151         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1152                 case FMODE_READ:
1153                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1154                                 && state->n_rdonly != 0;
1155                         break;
1156                 case FMODE_WRITE:
1157                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1158                                 && state->n_wronly != 0;
1159                         break;
1160                 case FMODE_READ|FMODE_WRITE:
1161                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1162                                 && state->n_rdwr != 0;
1163         }
1164 out:
1165         return ret;
1166 }
1167
1168 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1169                 enum open_claim_type4 claim)
1170 {
1171         if (delegation == NULL)
1172                 return 0;
1173         if ((delegation->type & fmode) != fmode)
1174                 return 0;
1175         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1176                 return 0;
1177         switch (claim) {
1178         case NFS4_OPEN_CLAIM_NULL:
1179         case NFS4_OPEN_CLAIM_FH:
1180                 break;
1181         case NFS4_OPEN_CLAIM_PREVIOUS:
1182                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1183                         break;
1184         default:
1185                 return 0;
1186         }
1187         nfs_mark_delegation_referenced(delegation);
1188         return 1;
1189 }
1190
1191 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1192 {
1193         switch (fmode) {
1194                 case FMODE_WRITE:
1195                         state->n_wronly++;
1196                         break;
1197                 case FMODE_READ:
1198                         state->n_rdonly++;
1199                         break;
1200                 case FMODE_READ|FMODE_WRITE:
1201                         state->n_rdwr++;
1202         }
1203         nfs4_state_set_mode_locked(state, state->state | fmode);
1204 }
1205
1206 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1207 {
1208         struct nfs_client *clp = state->owner->so_server->nfs_client;
1209         bool need_recover = false;
1210
1211         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1212                 need_recover = true;
1213         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1214                 need_recover = true;
1215         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1216                 need_recover = true;
1217         if (need_recover)
1218                 nfs4_state_mark_reclaim_nograce(clp, state);
1219 }
1220
1221 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1222                 nfs4_stateid *stateid)
1223 {
1224         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1225                 return true;
1226         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1227                 nfs_test_and_clear_all_open_stateid(state);
1228                 return true;
1229         }
1230         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1231                 return true;
1232         return false;
1233 }
1234
1235 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1236 {
1237         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1238                 return;
1239         if (state->n_wronly)
1240                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1241         if (state->n_rdonly)
1242                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1243         if (state->n_rdwr)
1244                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1245         set_bit(NFS_OPEN_STATE, &state->flags);
1246 }
1247
1248 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1249                 nfs4_stateid *arg_stateid,
1250                 nfs4_stateid *stateid, fmode_t fmode)
1251 {
1252         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1253         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1254         case FMODE_WRITE:
1255                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1256                 break;
1257         case FMODE_READ:
1258                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1259                 break;
1260         case 0:
1261                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1262                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1263                 clear_bit(NFS_OPEN_STATE, &state->flags);
1264         }
1265         if (stateid == NULL)
1266                 return;
1267         /* Handle races with OPEN */
1268         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1269             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1270             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1271                 nfs_resync_open_stateid_locked(state);
1272                 return;
1273         }
1274         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1275                 nfs4_stateid_copy(&state->stateid, stateid);
1276         nfs4_stateid_copy(&state->open_stateid, stateid);
1277 }
1278
1279 static void nfs_clear_open_stateid(struct nfs4_state *state,
1280         nfs4_stateid *arg_stateid,
1281         nfs4_stateid *stateid, fmode_t fmode)
1282 {
1283         write_seqlock(&state->seqlock);
1284         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1285         write_sequnlock(&state->seqlock);
1286         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1287                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1288 }
1289
1290 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1291 {
1292         switch (fmode) {
1293                 case FMODE_READ:
1294                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1295                         break;
1296                 case FMODE_WRITE:
1297                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1298                         break;
1299                 case FMODE_READ|FMODE_WRITE:
1300                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1301         }
1302         if (!nfs_need_update_open_stateid(state, stateid))
1303                 return;
1304         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1305                 nfs4_stateid_copy(&state->stateid, stateid);
1306         nfs4_stateid_copy(&state->open_stateid, stateid);
1307 }
1308
1309 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1310 {
1311         /*
1312          * Protect the call to nfs4_state_set_mode_locked and
1313          * serialise the stateid update
1314          */
1315         write_seqlock(&state->seqlock);
1316         if (deleg_stateid != NULL) {
1317                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1318                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1319         }
1320         if (open_stateid != NULL)
1321                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1322         write_sequnlock(&state->seqlock);
1323         spin_lock(&state->owner->so_lock);
1324         update_open_stateflags(state, fmode);
1325         spin_unlock(&state->owner->so_lock);
1326 }
1327
1328 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1329 {
1330         struct nfs_inode *nfsi = NFS_I(state->inode);
1331         struct nfs_delegation *deleg_cur;
1332         int ret = 0;
1333
1334         fmode &= (FMODE_READ|FMODE_WRITE);
1335
1336         rcu_read_lock();
1337         deleg_cur = rcu_dereference(nfsi->delegation);
1338         if (deleg_cur == NULL)
1339                 goto no_delegation;
1340
1341         spin_lock(&deleg_cur->lock);
1342         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1343            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1344             (deleg_cur->type & fmode) != fmode)
1345                 goto no_delegation_unlock;
1346
1347         if (delegation == NULL)
1348                 delegation = &deleg_cur->stateid;
1349         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1350                 goto no_delegation_unlock;
1351
1352         nfs_mark_delegation_referenced(deleg_cur);
1353         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1354         ret = 1;
1355 no_delegation_unlock:
1356         spin_unlock(&deleg_cur->lock);
1357 no_delegation:
1358         rcu_read_unlock();
1359
1360         if (!ret && open_stateid != NULL) {
1361                 __update_open_stateid(state, open_stateid, NULL, fmode);
1362                 ret = 1;
1363         }
1364         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1365                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1366
1367         return ret;
1368 }
1369
1370 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1371                 const nfs4_stateid *stateid)
1372 {
1373         struct nfs4_state *state = lsp->ls_state;
1374         bool ret = false;
1375
1376         spin_lock(&state->state_lock);
1377         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1378                 goto out_noupdate;
1379         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1380                 goto out_noupdate;
1381         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1382         ret = true;
1383 out_noupdate:
1384         spin_unlock(&state->state_lock);
1385         return ret;
1386 }
1387
1388 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1389 {
1390         struct nfs_delegation *delegation;
1391
1392         rcu_read_lock();
1393         delegation = rcu_dereference(NFS_I(inode)->delegation);
1394         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1395                 rcu_read_unlock();
1396                 return;
1397         }
1398         rcu_read_unlock();
1399         nfs4_inode_return_delegation(inode);
1400 }
1401
1402 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1403 {
1404         struct nfs4_state *state = opendata->state;
1405         struct nfs_inode *nfsi = NFS_I(state->inode);
1406         struct nfs_delegation *delegation;
1407         int open_mode = opendata->o_arg.open_flags;
1408         fmode_t fmode = opendata->o_arg.fmode;
1409         enum open_claim_type4 claim = opendata->o_arg.claim;
1410         nfs4_stateid stateid;
1411         int ret = -EAGAIN;
1412
1413         for (;;) {
1414                 spin_lock(&state->owner->so_lock);
1415                 if (can_open_cached(state, fmode, open_mode)) {
1416                         update_open_stateflags(state, fmode);
1417                         spin_unlock(&state->owner->so_lock);
1418                         goto out_return_state;
1419                 }
1420                 spin_unlock(&state->owner->so_lock);
1421                 rcu_read_lock();
1422                 delegation = rcu_dereference(nfsi->delegation);
1423                 if (!can_open_delegated(delegation, fmode, claim)) {
1424                         rcu_read_unlock();
1425                         break;
1426                 }
1427                 /* Save the delegation */
1428                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1429                 rcu_read_unlock();
1430                 nfs_release_seqid(opendata->o_arg.seqid);
1431                 if (!opendata->is_recover) {
1432                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1433                         if (ret != 0)
1434                                 goto out;
1435                 }
1436                 ret = -EAGAIN;
1437
1438                 /* Try to update the stateid using the delegation */
1439                 if (update_open_stateid(state, NULL, &stateid, fmode))
1440                         goto out_return_state;
1441         }
1442 out:
1443         return ERR_PTR(ret);
1444 out_return_state:
1445         atomic_inc(&state->count);
1446         return state;
1447 }
1448
1449 static void
1450 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1451 {
1452         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1453         struct nfs_delegation *delegation;
1454         int delegation_flags = 0;
1455
1456         rcu_read_lock();
1457         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1458         if (delegation)
1459                 delegation_flags = delegation->flags;
1460         rcu_read_unlock();
1461         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1462                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1463                                    "returning a delegation for "
1464                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1465                                    clp->cl_hostname);
1466         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1467                 nfs_inode_set_delegation(state->inode,
1468                                          data->owner->so_cred,
1469                                          &data->o_res);
1470         else
1471                 nfs_inode_reclaim_delegation(state->inode,
1472                                              data->owner->so_cred,
1473                                              &data->o_res);
1474 }
1475
1476 /*
1477  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1478  * and update the nfs4_state.
1479  */
1480 static struct nfs4_state *
1481 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1482 {
1483         struct inode *inode = data->state->inode;
1484         struct nfs4_state *state = data->state;
1485         int ret;
1486
1487         if (!data->rpc_done) {
1488                 if (data->rpc_status) {
1489                         ret = data->rpc_status;
1490                         goto err;
1491                 }
1492                 /* cached opens have already been processed */
1493                 goto update;
1494         }
1495
1496         ret = nfs_refresh_inode(inode, &data->f_attr);
1497         if (ret)
1498                 goto err;
1499
1500         if (data->o_res.delegation_type != 0)
1501                 nfs4_opendata_check_deleg(data, state);
1502 update:
1503         update_open_stateid(state, &data->o_res.stateid, NULL,
1504                             data->o_arg.fmode);
1505         atomic_inc(&state->count);
1506
1507         return state;
1508 err:
1509         return ERR_PTR(ret);
1510
1511 }
1512
1513 static struct nfs4_state *
1514 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1515 {
1516         struct inode *inode;
1517         struct nfs4_state *state = NULL;
1518         int ret;
1519
1520         if (!data->rpc_done) {
1521                 state = nfs4_try_open_cached(data);
1522                 goto out;
1523         }
1524
1525         ret = -EAGAIN;
1526         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1527                 goto err;
1528         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1529         ret = PTR_ERR(inode);
1530         if (IS_ERR(inode))
1531                 goto err;
1532         ret = -ENOMEM;
1533         state = nfs4_get_open_state(inode, data->owner);
1534         if (state == NULL)
1535                 goto err_put_inode;
1536         if (data->o_res.delegation_type != 0)
1537                 nfs4_opendata_check_deleg(data, state);
1538         update_open_stateid(state, &data->o_res.stateid, NULL,
1539                         data->o_arg.fmode);
1540         iput(inode);
1541 out:
1542         nfs_release_seqid(data->o_arg.seqid);
1543         return state;
1544 err_put_inode:
1545         iput(inode);
1546 err:
1547         return ERR_PTR(ret);
1548 }
1549
1550 static struct nfs4_state *
1551 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1552 {
1553         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1554                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1555         return _nfs4_opendata_to_nfs4_state(data);
1556 }
1557
1558 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1559 {
1560         struct nfs_inode *nfsi = NFS_I(state->inode);
1561         struct nfs_open_context *ctx;
1562
1563         spin_lock(&state->inode->i_lock);
1564         list_for_each_entry(ctx, &nfsi->open_files, list) {
1565                 if (ctx->state != state)
1566                         continue;
1567                 get_nfs_open_context(ctx);
1568                 spin_unlock(&state->inode->i_lock);
1569                 return ctx;
1570         }
1571         spin_unlock(&state->inode->i_lock);
1572         return ERR_PTR(-ENOENT);
1573 }
1574
1575 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1576                 struct nfs4_state *state, enum open_claim_type4 claim)
1577 {
1578         struct nfs4_opendata *opendata;
1579
1580         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1581                         NULL, NULL, claim, GFP_NOFS);
1582         if (opendata == NULL)
1583                 return ERR_PTR(-ENOMEM);
1584         opendata->state = state;
1585         atomic_inc(&state->count);
1586         return opendata;
1587 }
1588
1589 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1590                 fmode_t fmode)
1591 {
1592         struct nfs4_state *newstate;
1593         int ret;
1594
1595         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1596                 return 0;
1597         opendata->o_arg.open_flags = 0;
1598         opendata->o_arg.fmode = fmode;
1599         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1600                         NFS_SB(opendata->dentry->d_sb),
1601                         fmode, 0);
1602         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1603         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1604         nfs4_init_opendata_res(opendata);
1605         ret = _nfs4_recover_proc_open(opendata);
1606         if (ret != 0)
1607                 return ret; 
1608         newstate = nfs4_opendata_to_nfs4_state(opendata);
1609         if (IS_ERR(newstate))
1610                 return PTR_ERR(newstate);
1611         if (newstate != opendata->state)
1612                 ret = -ESTALE;
1613         nfs4_close_state(newstate, fmode);
1614         return ret;
1615 }
1616
1617 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1618 {
1619         int ret;
1620
1621         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1622         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1623         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1624         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1625         /* memory barrier prior to reading state->n_* */
1626         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1627         clear_bit(NFS_OPEN_STATE, &state->flags);
1628         smp_rmb();
1629         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1630         if (ret != 0)
1631                 return ret;
1632         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1633         if (ret != 0)
1634                 return ret;
1635         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1636         if (ret != 0)
1637                 return ret;
1638         /*
1639          * We may have performed cached opens for all three recoveries.
1640          * Check if we need to update the current stateid.
1641          */
1642         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1643             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1644                 write_seqlock(&state->seqlock);
1645                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1646                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1647                 write_sequnlock(&state->seqlock);
1648         }
1649         return 0;
1650 }
1651
1652 /*
1653  * OPEN_RECLAIM:
1654  *      reclaim state on the server after a reboot.
1655  */
1656 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1657 {
1658         struct nfs_delegation *delegation;
1659         struct nfs4_opendata *opendata;
1660         fmode_t delegation_type = 0;
1661         int status;
1662
1663         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1664                         NFS4_OPEN_CLAIM_PREVIOUS);
1665         if (IS_ERR(opendata))
1666                 return PTR_ERR(opendata);
1667         rcu_read_lock();
1668         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1669         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1670                 delegation_type = delegation->type;
1671         rcu_read_unlock();
1672         opendata->o_arg.u.delegation_type = delegation_type;
1673         status = nfs4_open_recover(opendata, state);
1674         nfs4_opendata_put(opendata);
1675         return status;
1676 }
1677
1678 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1679 {
1680         struct nfs_server *server = NFS_SERVER(state->inode);
1681         struct nfs4_exception exception = { };
1682         int err;
1683         do {
1684                 err = _nfs4_do_open_reclaim(ctx, state);
1685                 trace_nfs4_open_reclaim(ctx, 0, err);
1686                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1687                         continue;
1688                 if (err != -NFS4ERR_DELAY)
1689                         break;
1690                 nfs4_handle_exception(server, err, &exception);
1691         } while (exception.retry);
1692         return err;
1693 }
1694
1695 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1696 {
1697         struct nfs_open_context *ctx;
1698         int ret;
1699
1700         ctx = nfs4_state_find_open_context(state);
1701         if (IS_ERR(ctx))
1702                 return -EAGAIN;
1703         ret = nfs4_do_open_reclaim(ctx, state);
1704         put_nfs_open_context(ctx);
1705         return ret;
1706 }
1707
1708 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1709 {
1710         switch (err) {
1711                 default:
1712                         printk(KERN_ERR "NFS: %s: unhandled error "
1713                                         "%d.\n", __func__, err);
1714                 case 0:
1715                 case -ENOENT:
1716                 case -EAGAIN:
1717                 case -ESTALE:
1718                         break;
1719                 case -NFS4ERR_BADSESSION:
1720                 case -NFS4ERR_BADSLOT:
1721                 case -NFS4ERR_BAD_HIGH_SLOT:
1722                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1723                 case -NFS4ERR_DEADSESSION:
1724                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1725                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1726                         return -EAGAIN;
1727                 case -NFS4ERR_STALE_CLIENTID:
1728                 case -NFS4ERR_STALE_STATEID:
1729                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1730                 case -NFS4ERR_EXPIRED:
1731                         /* Don't recall a delegation if it was lost */
1732                         nfs4_schedule_lease_recovery(server->nfs_client);
1733                         return -EAGAIN;
1734                 case -NFS4ERR_MOVED:
1735                         nfs4_schedule_migration_recovery(server);
1736                         return -EAGAIN;
1737                 case -NFS4ERR_LEASE_MOVED:
1738                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1739                         return -EAGAIN;
1740                 case -NFS4ERR_DELEG_REVOKED:
1741                 case -NFS4ERR_ADMIN_REVOKED:
1742                 case -NFS4ERR_BAD_STATEID:
1743                 case -NFS4ERR_OPENMODE:
1744                         nfs_inode_find_state_and_recover(state->inode,
1745                                         stateid);
1746                         nfs4_schedule_stateid_recovery(server, state);
1747                         return -EAGAIN;
1748                 case -NFS4ERR_DELAY:
1749                 case -NFS4ERR_GRACE:
1750                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1751                         ssleep(1);
1752                         return -EAGAIN;
1753                 case -ENOMEM:
1754                 case -NFS4ERR_DENIED:
1755                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1756                         return 0;
1757         }
1758         return err;
1759 }
1760
1761 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1762                 struct nfs4_state *state, const nfs4_stateid *stateid,
1763                 fmode_t type)
1764 {
1765         struct nfs_server *server = NFS_SERVER(state->inode);
1766         struct nfs4_opendata *opendata;
1767         int err = 0;
1768
1769         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1770                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1771         if (IS_ERR(opendata))
1772                 return PTR_ERR(opendata);
1773         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1774         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1775         switch (type & (FMODE_READ|FMODE_WRITE)) {
1776         case FMODE_READ|FMODE_WRITE:
1777         case FMODE_WRITE:
1778                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1779                 if (err)
1780                         break;
1781                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1782                 if (err)
1783                         break;
1784         case FMODE_READ:
1785                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1786         }
1787         nfs4_opendata_put(opendata);
1788         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1789 }
1790
1791 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1792 {
1793         struct nfs4_opendata *data = calldata;
1794
1795         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1796                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1797 }
1798
1799 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1800 {
1801         struct nfs4_opendata *data = calldata;
1802
1803         nfs40_sequence_done(task, &data->c_res.seq_res);
1804
1805         data->rpc_status = task->tk_status;
1806         if (data->rpc_status == 0) {
1807                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1808                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1809                 renew_lease(data->o_res.server, data->timestamp);
1810                 data->rpc_done = 1;
1811         }
1812 }
1813
1814 static void nfs4_open_confirm_release(void *calldata)
1815 {
1816         struct nfs4_opendata *data = calldata;
1817         struct nfs4_state *state = NULL;
1818
1819         /* If this request hasn't been cancelled, do nothing */
1820         if (data->cancelled == 0)
1821                 goto out_free;
1822         /* In case of error, no cleanup! */
1823         if (!data->rpc_done)
1824                 goto out_free;
1825         state = nfs4_opendata_to_nfs4_state(data);
1826         if (!IS_ERR(state))
1827                 nfs4_close_state(state, data->o_arg.fmode);
1828 out_free:
1829         nfs4_opendata_put(data);
1830 }
1831
1832 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1833         .rpc_call_prepare = nfs4_open_confirm_prepare,
1834         .rpc_call_done = nfs4_open_confirm_done,
1835         .rpc_release = nfs4_open_confirm_release,
1836 };
1837
1838 /*
1839  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1840  */
1841 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1842 {
1843         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1844         struct rpc_task *task;
1845         struct  rpc_message msg = {
1846                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1847                 .rpc_argp = &data->c_arg,
1848                 .rpc_resp = &data->c_res,
1849                 .rpc_cred = data->owner->so_cred,
1850         };
1851         struct rpc_task_setup task_setup_data = {
1852                 .rpc_client = server->client,
1853                 .rpc_message = &msg,
1854                 .callback_ops = &nfs4_open_confirm_ops,
1855                 .callback_data = data,
1856                 .workqueue = nfsiod_workqueue,
1857                 .flags = RPC_TASK_ASYNC,
1858         };
1859         int status;
1860
1861         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1862         kref_get(&data->kref);
1863         data->rpc_done = 0;
1864         data->rpc_status = 0;
1865         data->timestamp = jiffies;
1866         task = rpc_run_task(&task_setup_data);
1867         if (IS_ERR(task))
1868                 return PTR_ERR(task);
1869         status = nfs4_wait_for_completion_rpc_task(task);
1870         if (status != 0) {
1871                 data->cancelled = 1;
1872                 smp_wmb();
1873         } else
1874                 status = data->rpc_status;
1875         rpc_put_task(task);
1876         return status;
1877 }
1878
1879 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1880 {
1881         struct nfs4_opendata *data = calldata;
1882         struct nfs4_state_owner *sp = data->owner;
1883         struct nfs_client *clp = sp->so_server->nfs_client;
1884         enum open_claim_type4 claim = data->o_arg.claim;
1885
1886         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1887                 goto out_wait;
1888         /*
1889          * Check if we still need to send an OPEN call, or if we can use
1890          * a delegation instead.
1891          */
1892         if (data->state != NULL) {
1893                 struct nfs_delegation *delegation;
1894
1895                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1896                         goto out_no_action;
1897                 rcu_read_lock();
1898                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1899                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1900                         goto unlock_no_action;
1901                 rcu_read_unlock();
1902         }
1903         /* Update client id. */
1904         data->o_arg.clientid = clp->cl_clientid;
1905         switch (claim) {
1906         default:
1907                 break;
1908         case NFS4_OPEN_CLAIM_PREVIOUS:
1909         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1910         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1911                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1912         case NFS4_OPEN_CLAIM_FH:
1913                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1914                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1915         }
1916         data->timestamp = jiffies;
1917         if (nfs4_setup_sequence(data->o_arg.server,
1918                                 &data->o_arg.seq_args,
1919                                 &data->o_res.seq_res,
1920                                 task) != 0)
1921                 nfs_release_seqid(data->o_arg.seqid);
1922
1923         /* Set the create mode (note dependency on the session type) */
1924         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1925         if (data->o_arg.open_flags & O_EXCL) {
1926                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1927                 if (nfs4_has_persistent_session(clp))
1928                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1929                 else if (clp->cl_mvops->minor_version > 0)
1930                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1931         }
1932         return;
1933 unlock_no_action:
1934         rcu_read_unlock();
1935 out_no_action:
1936         task->tk_action = NULL;
1937 out_wait:
1938         nfs4_sequence_done(task, &data->o_res.seq_res);
1939 }
1940
1941 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1942 {
1943         struct nfs4_opendata *data = calldata;
1944
1945         data->rpc_status = task->tk_status;
1946
1947         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1948                 return;
1949
1950         if (task->tk_status == 0) {
1951                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1952                         switch (data->o_res.f_attr->mode & S_IFMT) {
1953                         case S_IFREG:
1954                                 break;
1955                         case S_IFLNK:
1956                                 data->rpc_status = -ELOOP;
1957                                 break;
1958                         case S_IFDIR:
1959                                 data->rpc_status = -EISDIR;
1960                                 break;
1961                         default:
1962                                 data->rpc_status = -ENOTDIR;
1963                         }
1964                 }
1965                 renew_lease(data->o_res.server, data->timestamp);
1966                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1967                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1968         }
1969         data->rpc_done = 1;
1970 }
1971
1972 static void nfs4_open_release(void *calldata)
1973 {
1974         struct nfs4_opendata *data = calldata;
1975         struct nfs4_state *state = NULL;
1976
1977         /* If this request hasn't been cancelled, do nothing */
1978         if (data->cancelled == 0)
1979                 goto out_free;
1980         /* In case of error, no cleanup! */
1981         if (data->rpc_status != 0 || !data->rpc_done)
1982                 goto out_free;
1983         /* In case we need an open_confirm, no cleanup! */
1984         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1985                 goto out_free;
1986         state = nfs4_opendata_to_nfs4_state(data);
1987         if (!IS_ERR(state))
1988                 nfs4_close_state(state, data->o_arg.fmode);
1989 out_free:
1990         nfs4_opendata_put(data);
1991 }
1992
1993 static const struct rpc_call_ops nfs4_open_ops = {
1994         .rpc_call_prepare = nfs4_open_prepare,
1995         .rpc_call_done = nfs4_open_done,
1996         .rpc_release = nfs4_open_release,
1997 };
1998
1999 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2000 {
2001         struct inode *dir = d_inode(data->dir);
2002         struct nfs_server *server = NFS_SERVER(dir);
2003         struct nfs_openargs *o_arg = &data->o_arg;
2004         struct nfs_openres *o_res = &data->o_res;
2005         struct rpc_task *task;
2006         struct rpc_message msg = {
2007                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2008                 .rpc_argp = o_arg,
2009                 .rpc_resp = o_res,
2010                 .rpc_cred = data->owner->so_cred,
2011         };
2012         struct rpc_task_setup task_setup_data = {
2013                 .rpc_client = server->client,
2014                 .rpc_message = &msg,
2015                 .callback_ops = &nfs4_open_ops,
2016                 .callback_data = data,
2017                 .workqueue = nfsiod_workqueue,
2018                 .flags = RPC_TASK_ASYNC,
2019         };
2020         int status;
2021
2022         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2023         kref_get(&data->kref);
2024         data->rpc_done = 0;
2025         data->rpc_status = 0;
2026         data->cancelled = 0;
2027         data->is_recover = 0;
2028         if (isrecover) {
2029                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2030                 data->is_recover = 1;
2031         }
2032         task = rpc_run_task(&task_setup_data);
2033         if (IS_ERR(task))
2034                 return PTR_ERR(task);
2035         status = nfs4_wait_for_completion_rpc_task(task);
2036         if (status != 0) {
2037                 data->cancelled = 1;
2038                 smp_wmb();
2039         } else
2040                 status = data->rpc_status;
2041         rpc_put_task(task);
2042
2043         return status;
2044 }
2045
2046 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2047 {
2048         struct inode *dir = d_inode(data->dir);
2049         struct nfs_openres *o_res = &data->o_res;
2050         int status;
2051
2052         status = nfs4_run_open_task(data, 1);
2053         if (status != 0 || !data->rpc_done)
2054                 return status;
2055
2056         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2057
2058         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2059                 status = _nfs4_proc_open_confirm(data);
2060                 if (status != 0)
2061                         return status;
2062         }
2063
2064         return status;
2065 }
2066
2067 /*
2068  * Additional permission checks in order to distinguish between an
2069  * open for read, and an open for execute. This works around the
2070  * fact that NFSv4 OPEN treats read and execute permissions as being
2071  * the same.
2072  * Note that in the non-execute case, we want to turn off permission
2073  * checking if we just created a new file (POSIX open() semantics).
2074  */
2075 static int nfs4_opendata_access(struct rpc_cred *cred,
2076                                 struct nfs4_opendata *opendata,
2077                                 struct nfs4_state *state, fmode_t fmode,
2078                                 int openflags)
2079 {
2080         struct nfs_access_entry cache;
2081         u32 mask;
2082
2083         /* access call failed or for some reason the server doesn't
2084          * support any access modes -- defer access call until later */
2085         if (opendata->o_res.access_supported == 0)
2086                 return 0;
2087
2088         mask = 0;
2089         /*
2090          * Use openflags to check for exec, because fmode won't
2091          * always have FMODE_EXEC set when file open for exec.
2092          */
2093         if (openflags & __FMODE_EXEC) {
2094                 /* ONLY check for exec rights */
2095                 mask = MAY_EXEC;
2096         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2097                 mask = MAY_READ;
2098
2099         cache.cred = cred;
2100         cache.jiffies = jiffies;
2101         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2102         nfs_access_add_cache(state->inode, &cache);
2103
2104         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2105                 return 0;
2106
2107         /* even though OPEN succeeded, access is denied. Close the file */
2108         nfs4_close_state(state, fmode);
2109         return -EACCES;
2110 }
2111
2112 /*
2113  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2114  */
2115 static int _nfs4_proc_open(struct nfs4_opendata *data)
2116 {
2117         struct inode *dir = d_inode(data->dir);
2118         struct nfs_server *server = NFS_SERVER(dir);
2119         struct nfs_openargs *o_arg = &data->o_arg;
2120         struct nfs_openres *o_res = &data->o_res;
2121         int status;
2122
2123         status = nfs4_run_open_task(data, 0);
2124         if (!data->rpc_done)
2125                 return status;
2126         if (status != 0) {
2127                 if (status == -NFS4ERR_BADNAME &&
2128                                 !(o_arg->open_flags & O_CREAT))
2129                         return -ENOENT;
2130                 return status;
2131         }
2132
2133         nfs_fattr_map_and_free_names(server, &data->f_attr);
2134
2135         if (o_arg->open_flags & O_CREAT) {
2136                 update_changeattr(dir, &o_res->cinfo);
2137                 if (o_arg->open_flags & O_EXCL)
2138                         data->file_created = 1;
2139                 else if (o_res->cinfo.before != o_res->cinfo.after)
2140                         data->file_created = 1;
2141         }
2142         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2143                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2144         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2145                 status = _nfs4_proc_open_confirm(data);
2146                 if (status != 0)
2147                         return status;
2148         }
2149         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2150                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2151         return 0;
2152 }
2153
2154 static int nfs4_recover_expired_lease(struct nfs_server *server)
2155 {
2156         return nfs4_client_recover_expired_lease(server->nfs_client);
2157 }
2158
2159 /*
2160  * OPEN_EXPIRED:
2161  *      reclaim state on the server after a network partition.
2162  *      Assumes caller holds the appropriate lock
2163  */
2164 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2165 {
2166         struct nfs4_opendata *opendata;
2167         int ret;
2168
2169         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2170                         NFS4_OPEN_CLAIM_FH);
2171         if (IS_ERR(opendata))
2172                 return PTR_ERR(opendata);
2173         ret = nfs4_open_recover(opendata, state);
2174         if (ret == -ESTALE)
2175                 d_drop(ctx->dentry);
2176         nfs4_opendata_put(opendata);
2177         return ret;
2178 }
2179
2180 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2181 {
2182         struct nfs_server *server = NFS_SERVER(state->inode);
2183         struct nfs4_exception exception = { };
2184         int err;
2185
2186         do {
2187                 err = _nfs4_open_expired(ctx, state);
2188                 trace_nfs4_open_expired(ctx, 0, err);
2189                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2190                         continue;
2191                 switch (err) {
2192                 default:
2193                         goto out;
2194                 case -NFS4ERR_GRACE:
2195                 case -NFS4ERR_DELAY:
2196                         nfs4_handle_exception(server, err, &exception);
2197                         err = 0;
2198                 }
2199         } while (exception.retry);
2200 out:
2201         return err;
2202 }
2203
2204 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2205 {
2206         struct nfs_open_context *ctx;
2207         int ret;
2208
2209         ctx = nfs4_state_find_open_context(state);
2210         if (IS_ERR(ctx))
2211                 return -EAGAIN;
2212         ret = nfs4_do_open_expired(ctx, state);
2213         put_nfs_open_context(ctx);
2214         return ret;
2215 }
2216
2217 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2218 {
2219         nfs_remove_bad_delegation(state->inode);
2220         write_seqlock(&state->seqlock);
2221         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2222         write_sequnlock(&state->seqlock);
2223         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2224 }
2225
2226 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2227 {
2228         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2229                 nfs_finish_clear_delegation_stateid(state);
2230 }
2231
2232 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2233 {
2234         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2235         nfs40_clear_delegation_stateid(state);
2236         return nfs4_open_expired(sp, state);
2237 }
2238
2239 #if defined(CONFIG_NFS_V4_1)
2240 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2241 {
2242         struct nfs_server *server = NFS_SERVER(state->inode);
2243         nfs4_stateid stateid;
2244         struct nfs_delegation *delegation;
2245         struct rpc_cred *cred;
2246         int status;
2247
2248         /* Get the delegation credential for use by test/free_stateid */
2249         rcu_read_lock();
2250         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2251         if (delegation == NULL) {
2252                 rcu_read_unlock();
2253                 return;
2254         }
2255
2256         nfs4_stateid_copy(&stateid, &delegation->stateid);
2257         cred = get_rpccred(delegation->cred);
2258         rcu_read_unlock();
2259         status = nfs41_test_stateid(server, &stateid, cred);
2260         trace_nfs4_test_delegation_stateid(state, NULL, status);
2261
2262         if (status != NFS_OK) {
2263                 /* Free the stateid unless the server explicitly
2264                  * informs us the stateid is unrecognized. */
2265                 if (status != -NFS4ERR_BAD_STATEID)
2266                         nfs41_free_stateid(server, &stateid, cred);
2267                 nfs_finish_clear_delegation_stateid(state);
2268         }
2269
2270         put_rpccred(cred);
2271 }
2272
2273 /**
2274  * nfs41_check_open_stateid - possibly free an open stateid
2275  *
2276  * @state: NFSv4 state for an inode
2277  *
2278  * Returns NFS_OK if recovery for this stateid is now finished.
2279  * Otherwise a negative NFS4ERR value is returned.
2280  */
2281 static int nfs41_check_open_stateid(struct nfs4_state *state)
2282 {
2283         struct nfs_server *server = NFS_SERVER(state->inode);
2284         nfs4_stateid *stateid = &state->open_stateid;
2285         struct rpc_cred *cred = state->owner->so_cred;
2286         int status;
2287
2288         /* If a state reset has been done, test_stateid is unneeded */
2289         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2290             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2291             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2292                 return -NFS4ERR_BAD_STATEID;
2293
2294         status = nfs41_test_stateid(server, stateid, cred);
2295         trace_nfs4_test_open_stateid(state, NULL, status);
2296         if (status != NFS_OK) {
2297                 /* Free the stateid unless the server explicitly
2298                  * informs us the stateid is unrecognized. */
2299                 if (status != -NFS4ERR_BAD_STATEID)
2300                         nfs41_free_stateid(server, stateid, cred);
2301
2302                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2303                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2304                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2305                 clear_bit(NFS_OPEN_STATE, &state->flags);
2306         }
2307         return status;
2308 }
2309
2310 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2311 {
2312         int status;
2313
2314         nfs41_check_delegation_stateid(state);
2315         status = nfs41_check_open_stateid(state);
2316         if (status != NFS_OK)
2317                 status = nfs4_open_expired(sp, state);
2318         return status;
2319 }
2320 #endif
2321
2322 /*
2323  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2324  * fields corresponding to attributes that were used to store the verifier.
2325  * Make sure we clobber those fields in the later setattr call
2326  */
2327 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2328                                 struct iattr *sattr, struct nfs4_label **label)
2329 {
2330         const u32 *attrset = opendata->o_res.attrset;
2331
2332         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2333             !(sattr->ia_valid & ATTR_ATIME_SET))
2334                 sattr->ia_valid |= ATTR_ATIME;
2335
2336         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2337             !(sattr->ia_valid & ATTR_MTIME_SET))
2338                 sattr->ia_valid |= ATTR_MTIME;
2339
2340         /* Except MODE, it seems harmless of setting twice. */
2341         if ((attrset[1] & FATTR4_WORD1_MODE))
2342                 sattr->ia_valid &= ~ATTR_MODE;
2343
2344         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2345                 *label = NULL;
2346 }
2347
2348 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2349                 fmode_t fmode,
2350                 int flags,
2351                 struct nfs_open_context *ctx)
2352 {
2353         struct nfs4_state_owner *sp = opendata->owner;
2354         struct nfs_server *server = sp->so_server;
2355         struct dentry *dentry;
2356         struct nfs4_state *state;
2357         unsigned int seq;
2358         int ret;
2359
2360         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2361
2362         ret = _nfs4_proc_open(opendata);
2363         if (ret != 0)
2364                 goto out;
2365
2366         state = nfs4_opendata_to_nfs4_state(opendata);
2367         ret = PTR_ERR(state);
2368         if (IS_ERR(state))
2369                 goto out;
2370         if (server->caps & NFS_CAP_POSIX_LOCK)
2371                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2372
2373         dentry = opendata->dentry;
2374         if (d_really_is_negative(dentry)) {
2375                 /* FIXME: Is this d_drop() ever needed? */
2376                 d_drop(dentry);
2377                 dentry = d_add_unique(dentry, igrab(state->inode));
2378                 if (dentry == NULL) {
2379                         dentry = opendata->dentry;
2380                 } else if (dentry != ctx->dentry) {
2381                         dput(ctx->dentry);
2382                         ctx->dentry = dget(dentry);
2383                 }
2384                 nfs_set_verifier(dentry,
2385                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2386         }
2387
2388         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2389         if (ret != 0)
2390                 goto out;
2391
2392         ctx->state = state;
2393         if (d_inode(dentry) == state->inode) {
2394                 nfs_inode_attach_open_context(ctx);
2395                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2396                         nfs4_schedule_stateid_recovery(server, state);
2397         }
2398 out:
2399         return ret;
2400 }
2401
2402 /*
2403  * Returns a referenced nfs4_state
2404  */
2405 static int _nfs4_do_open(struct inode *dir,
2406                         struct nfs_open_context *ctx,
2407                         int flags,
2408                         struct iattr *sattr,
2409                         struct nfs4_label *label,
2410                         int *opened)
2411 {
2412         struct nfs4_state_owner  *sp;
2413         struct nfs4_state     *state = NULL;
2414         struct nfs_server       *server = NFS_SERVER(dir);
2415         struct nfs4_opendata *opendata;
2416         struct dentry *dentry = ctx->dentry;
2417         struct rpc_cred *cred = ctx->cred;
2418         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2419         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2420         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2421         struct nfs4_label *olabel = NULL;
2422         int status;
2423
2424         /* Protect against reboot recovery conflicts */
2425         status = -ENOMEM;
2426         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2427         if (sp == NULL) {
2428                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2429                 goto out_err;
2430         }
2431         status = nfs4_recover_expired_lease(server);
2432         if (status != 0)
2433                 goto err_put_state_owner;
2434         if (d_really_is_positive(dentry))
2435                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2436         status = -ENOMEM;
2437         if (d_really_is_positive(dentry))
2438                 claim = NFS4_OPEN_CLAIM_FH;
2439         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2440                         label, claim, GFP_KERNEL);
2441         if (opendata == NULL)
2442                 goto err_put_state_owner;
2443
2444         if (label) {
2445                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2446                 if (IS_ERR(olabel)) {
2447                         status = PTR_ERR(olabel);
2448                         goto err_opendata_put;
2449                 }
2450         }
2451
2452         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2453                 if (!opendata->f_attr.mdsthreshold) {
2454                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2455                         if (!opendata->f_attr.mdsthreshold)
2456                                 goto err_free_label;
2457                 }
2458                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2459         }
2460         if (d_really_is_positive(dentry))
2461                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2462
2463         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2464         if (status != 0)
2465                 goto err_free_label;
2466         state = ctx->state;
2467
2468         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2469             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2470                 nfs4_exclusive_attrset(opendata, sattr, &label);
2471
2472                 nfs_fattr_init(opendata->o_res.f_attr);
2473                 status = nfs4_do_setattr(state->inode, cred,
2474                                 opendata->o_res.f_attr, sattr,
2475                                 state, label, olabel);
2476                 if (status == 0) {
2477                         nfs_setattr_update_inode(state->inode, sattr,
2478                                         opendata->o_res.f_attr);
2479                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2480                 }
2481         }
2482         if (opened && opendata->file_created)
2483                 *opened |= FILE_CREATED;
2484
2485         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2486                 *ctx_th = opendata->f_attr.mdsthreshold;
2487                 opendata->f_attr.mdsthreshold = NULL;
2488         }
2489
2490         nfs4_label_free(olabel);
2491
2492         nfs4_opendata_put(opendata);
2493         nfs4_put_state_owner(sp);
2494         return 0;
2495 err_free_label:
2496         nfs4_label_free(olabel);
2497 err_opendata_put:
2498         nfs4_opendata_put(opendata);
2499 err_put_state_owner:
2500         nfs4_put_state_owner(sp);
2501 out_err:
2502         return status;
2503 }
2504
2505
2506 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2507                                         struct nfs_open_context *ctx,
2508                                         int flags,
2509                                         struct iattr *sattr,
2510                                         struct nfs4_label *label,
2511                                         int *opened)
2512 {
2513         struct nfs_server *server = NFS_SERVER(dir);
2514         struct nfs4_exception exception = { };
2515         struct nfs4_state *res;
2516         int status;
2517
2518         do {
2519                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2520                 res = ctx->state;
2521                 trace_nfs4_open_file(ctx, flags, status);
2522                 if (status == 0)
2523                         break;
2524                 /* NOTE: BAD_SEQID means the server and client disagree about the
2525                  * book-keeping w.r.t. state-changing operations
2526                  * (OPEN/CLOSE/LOCK/LOCKU...)
2527                  * It is actually a sign of a bug on the client or on the server.
2528                  *
2529                  * If we receive a BAD_SEQID error in the particular case of
2530                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2531                  * have unhashed the old state_owner for us, and that we can
2532                  * therefore safely retry using a new one. We should still warn
2533                  * the user though...
2534                  */
2535                 if (status == -NFS4ERR_BAD_SEQID) {
2536                         pr_warn_ratelimited("NFS: v4 server %s "
2537                                         " returned a bad sequence-id error!\n",
2538                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2539                         exception.retry = 1;
2540                         continue;
2541                 }
2542                 /*
2543                  * BAD_STATEID on OPEN means that the server cancelled our
2544                  * state before it received the OPEN_CONFIRM.
2545                  * Recover by retrying the request as per the discussion
2546                  * on Page 181 of RFC3530.
2547                  */
2548                 if (status == -NFS4ERR_BAD_STATEID) {
2549                         exception.retry = 1;
2550                         continue;
2551                 }
2552                 if (status == -EAGAIN) {
2553                         /* We must have found a delegation */
2554                         exception.retry = 1;
2555                         continue;
2556                 }
2557                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2558                         continue;
2559                 res = ERR_PTR(nfs4_handle_exception(server,
2560                                         status, &exception));
2561         } while (exception.retry);
2562         return res;
2563 }
2564
2565 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2566                             struct nfs_fattr *fattr, struct iattr *sattr,
2567                             struct nfs4_state *state, struct nfs4_label *ilabel,
2568                             struct nfs4_label *olabel)
2569 {
2570         struct nfs_server *server = NFS_SERVER(inode);
2571         struct nfs_setattrargs  arg = {
2572                 .fh             = NFS_FH(inode),
2573                 .iap            = sattr,
2574                 .server         = server,
2575                 .bitmask = server->attr_bitmask,
2576                 .label          = ilabel,
2577         };
2578         struct nfs_setattrres  res = {
2579                 .fattr          = fattr,
2580                 .label          = olabel,
2581                 .server         = server,
2582         };
2583         struct rpc_message msg = {
2584                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2585                 .rpc_argp       = &arg,
2586                 .rpc_resp       = &res,
2587                 .rpc_cred       = cred,
2588         };
2589         unsigned long timestamp = jiffies;
2590         fmode_t fmode;
2591         bool truncate;
2592         int status;
2593
2594         arg.bitmask = nfs4_bitmask(server, ilabel);
2595         if (ilabel)
2596                 arg.bitmask = nfs4_bitmask(server, olabel);
2597
2598         nfs_fattr_init(fattr);
2599
2600         /* Servers should only apply open mode checks for file size changes */
2601         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2602         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2603
2604         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2605                 /* Use that stateid */
2606         } else if (truncate && state != NULL) {
2607                 struct nfs_lockowner lockowner = {
2608                         .l_owner = current->files,
2609                         .l_pid = current->tgid,
2610                 };
2611                 if (!nfs4_valid_open_stateid(state))
2612                         return -EBADF;
2613                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2614                                 &lockowner) == -EIO)
2615                         return -EBADF;
2616         } else
2617                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2618
2619         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2620         if (status == 0 && state != NULL)
2621                 renew_lease(server, timestamp);
2622         return status;
2623 }
2624
2625 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2626                            struct nfs_fattr *fattr, struct iattr *sattr,
2627                            struct nfs4_state *state, struct nfs4_label *ilabel,
2628                            struct nfs4_label *olabel)
2629 {
2630         struct nfs_server *server = NFS_SERVER(inode);
2631         struct nfs4_exception exception = {
2632                 .state = state,
2633                 .inode = inode,
2634         };
2635         int err;
2636         do {
2637                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2638                 trace_nfs4_setattr(inode, err);
2639                 switch (err) {
2640                 case -NFS4ERR_OPENMODE:
2641                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2642                                 pr_warn_once("NFSv4: server %s is incorrectly "
2643                                                 "applying open mode checks to "
2644                                                 "a SETATTR that is not "
2645                                                 "changing file size.\n",
2646                                                 server->nfs_client->cl_hostname);
2647                         }
2648                         if (state && !(state->state & FMODE_WRITE)) {
2649                                 err = -EBADF;
2650                                 if (sattr->ia_valid & ATTR_OPEN)
2651                                         err = -EACCES;
2652                                 goto out;
2653                         }
2654                 }
2655                 err = nfs4_handle_exception(server, err, &exception);
2656         } while (exception.retry);
2657 out:
2658         return err;
2659 }
2660
2661 static bool
2662 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2663 {
2664         if (inode == NULL || !nfs_have_layout(inode))
2665                 return false;
2666
2667         return pnfs_wait_on_layoutreturn(inode, task);
2668 }
2669
2670 struct nfs4_closedata {
2671         struct inode *inode;
2672         struct nfs4_state *state;
2673         struct nfs_closeargs arg;
2674         struct nfs_closeres res;
2675         struct nfs_fattr fattr;
2676         unsigned long timestamp;
2677         bool roc;
2678         u32 roc_barrier;
2679 };
2680
2681 static void nfs4_free_closedata(void *data)
2682 {
2683         struct nfs4_closedata *calldata = data;
2684         struct nfs4_state_owner *sp = calldata->state->owner;
2685         struct super_block *sb = calldata->state->inode->i_sb;
2686
2687         if (calldata->roc)
2688                 pnfs_roc_release(calldata->state->inode);
2689         nfs4_put_open_state(calldata->state);
2690         nfs_free_seqid(calldata->arg.seqid);
2691         nfs4_put_state_owner(sp);
2692         nfs_sb_deactive(sb);
2693         kfree(calldata);
2694 }
2695
2696 static void nfs4_close_done(struct rpc_task *task, void *data)
2697 {
2698         struct nfs4_closedata *calldata = data;
2699         struct nfs4_state *state = calldata->state;
2700         struct nfs_server *server = NFS_SERVER(calldata->inode);
2701         nfs4_stateid *res_stateid = NULL;
2702
2703         dprintk("%s: begin!\n", __func__);
2704         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2705                 return;
2706         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2707         /* hmm. we are done with the inode, and in the process of freeing
2708          * the state_owner. we keep this around to process errors
2709          */
2710         switch (task->tk_status) {
2711                 case 0:
2712                         res_stateid = &calldata->res.stateid;
2713                         if (calldata->roc)
2714                                 pnfs_roc_set_barrier(state->inode,
2715                                                      calldata->roc_barrier);
2716                         renew_lease(server, calldata->timestamp);
2717                         break;
2718                 case -NFS4ERR_ADMIN_REVOKED:
2719                 case -NFS4ERR_STALE_STATEID:
2720                 case -NFS4ERR_OLD_STATEID:
2721                 case -NFS4ERR_BAD_STATEID:
2722                 case -NFS4ERR_EXPIRED:
2723                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2724                                                 &state->open_stateid)) {
2725                                 rpc_restart_call_prepare(task);
2726                                 goto out_release;
2727                         }
2728                         if (calldata->arg.fmode == 0)
2729                                 break;
2730                 default:
2731                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2732                                 rpc_restart_call_prepare(task);
2733                                 goto out_release;
2734                         }
2735         }
2736         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2737                         res_stateid, calldata->arg.fmode);
2738 out_release:
2739         nfs_release_seqid(calldata->arg.seqid);
2740         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2741         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2742 }
2743
2744 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2745 {
2746         struct nfs4_closedata *calldata = data;
2747         struct nfs4_state *state = calldata->state;
2748         struct inode *inode = calldata->inode;
2749         bool is_rdonly, is_wronly, is_rdwr;
2750         int call_close = 0;
2751
2752         dprintk("%s: begin!\n", __func__);
2753         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2754                 goto out_wait;
2755
2756         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2757         spin_lock(&state->owner->so_lock);
2758         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2759         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2760         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2761         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2762         /* Calculate the change in open mode */
2763         calldata->arg.fmode = 0;
2764         if (state->n_rdwr == 0) {
2765                 if (state->n_rdonly == 0)
2766                         call_close |= is_rdonly;
2767                 else if (is_rdonly)
2768                         calldata->arg.fmode |= FMODE_READ;
2769                 if (state->n_wronly == 0)
2770                         call_close |= is_wronly;
2771                 else if (is_wronly)
2772                         calldata->arg.fmode |= FMODE_WRITE;
2773         } else if (is_rdwr)
2774                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2775
2776         if (calldata->arg.fmode == 0)
2777                 call_close |= is_rdwr;
2778
2779         if (!nfs4_valid_open_stateid(state))
2780                 call_close = 0;
2781         spin_unlock(&state->owner->so_lock);
2782
2783         if (!call_close) {
2784                 /* Note: exit _without_ calling nfs4_close_done */
2785                 goto out_no_action;
2786         }
2787
2788         if (nfs4_wait_on_layoutreturn(inode, task)) {
2789                 nfs_release_seqid(calldata->arg.seqid);
2790                 goto out_wait;
2791         }
2792
2793         if (calldata->arg.fmode == 0)
2794                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2795         if (calldata->roc)
2796                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2797
2798         calldata->arg.share_access =
2799                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2800                                 calldata->arg.fmode, 0);
2801
2802         nfs_fattr_init(calldata->res.fattr);
2803         calldata->timestamp = jiffies;
2804         if (nfs4_setup_sequence(NFS_SERVER(inode),
2805                                 &calldata->arg.seq_args,
2806                                 &calldata->res.seq_res,
2807                                 task) != 0)
2808                 nfs_release_seqid(calldata->arg.seqid);
2809         dprintk("%s: done!\n", __func__);
2810         return;
2811 out_no_action:
2812         task->tk_action = NULL;
2813 out_wait:
2814         nfs4_sequence_done(task, &calldata->res.seq_res);
2815 }
2816
2817 static const struct rpc_call_ops nfs4_close_ops = {
2818         .rpc_call_prepare = nfs4_close_prepare,
2819         .rpc_call_done = nfs4_close_done,
2820         .rpc_release = nfs4_free_closedata,
2821 };
2822
2823 static bool nfs4_roc(struct inode *inode)
2824 {
2825         if (!nfs_have_layout(inode))
2826                 return false;
2827         return pnfs_roc(inode);
2828 }
2829
2830 /* 
2831  * It is possible for data to be read/written from a mem-mapped file 
2832  * after the sys_close call (which hits the vfs layer as a flush).
2833  * This means that we can't safely call nfsv4 close on a file until 
2834  * the inode is cleared. This in turn means that we are not good
2835  * NFSv4 citizens - we do not indicate to the server to update the file's 
2836  * share state even when we are done with one of the three share 
2837  * stateid's in the inode.
2838  *
2839  * NOTE: Caller must be holding the sp->so_owner semaphore!
2840  */
2841 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2842 {
2843         struct nfs_server *server = NFS_SERVER(state->inode);
2844         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2845         struct nfs4_closedata *calldata;
2846         struct nfs4_state_owner *sp = state->owner;
2847         struct rpc_task *task;
2848         struct rpc_message msg = {
2849                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2850                 .rpc_cred = state->owner->so_cred,
2851         };
2852         struct rpc_task_setup task_setup_data = {
2853                 .rpc_client = server->client,
2854                 .rpc_message = &msg,
2855                 .callback_ops = &nfs4_close_ops,
2856                 .workqueue = nfsiod_workqueue,
2857                 .flags = RPC_TASK_ASYNC,
2858         };
2859         int status = -ENOMEM;
2860
2861         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2862                 &task_setup_data.rpc_client, &msg);
2863
2864         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2865         if (calldata == NULL)
2866                 goto out;
2867         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2868         calldata->inode = state->inode;
2869         calldata->state = state;
2870         calldata->arg.fh = NFS_FH(state->inode);
2871         /* Serialization for the sequence id */
2872         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2873         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2874         if (IS_ERR(calldata->arg.seqid))
2875                 goto out_free_calldata;
2876         calldata->arg.fmode = 0;
2877         calldata->arg.bitmask = server->cache_consistency_bitmask;
2878         calldata->res.fattr = &calldata->fattr;
2879         calldata->res.seqid = calldata->arg.seqid;
2880         calldata->res.server = server;
2881         calldata->roc = nfs4_roc(state->inode);
2882         nfs_sb_active(calldata->inode->i_sb);
2883
2884         msg.rpc_argp = &calldata->arg;
2885         msg.rpc_resp = &calldata->res;
2886         task_setup_data.callback_data = calldata;
2887         task = rpc_run_task(&task_setup_data);
2888         if (IS_ERR(task))
2889                 return PTR_ERR(task);
2890         status = 0;
2891         if (wait)
2892                 status = rpc_wait_for_completion_task(task);
2893         rpc_put_task(task);
2894         return status;
2895 out_free_calldata:
2896         kfree(calldata);
2897 out:
2898         nfs4_put_open_state(state);
2899         nfs4_put_state_owner(sp);
2900         return status;
2901 }
2902
2903 static struct inode *
2904 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2905                 int open_flags, struct iattr *attr, int *opened)
2906 {
2907         struct nfs4_state *state;
2908         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2909
2910         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2911
2912         /* Protect against concurrent sillydeletes */
2913         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2914
2915         nfs4_label_release_security(label);
2916
2917         if (IS_ERR(state))
2918                 return ERR_CAST(state);
2919         return state->inode;
2920 }
2921
2922 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2923 {
2924         if (ctx->state == NULL)
2925                 return;
2926         if (is_sync)
2927                 nfs4_close_sync(ctx->state, ctx->mode);
2928         else
2929                 nfs4_close_state(ctx->state, ctx->mode);
2930 }
2931
2932 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2933 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2934 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2935
2936 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2937 {
2938         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
2939         struct nfs4_server_caps_arg args = {
2940                 .fhandle = fhandle,
2941                 .bitmask = bitmask,
2942         };
2943         struct nfs4_server_caps_res res = {};
2944         struct rpc_message msg = {
2945                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2946                 .rpc_argp = &args,
2947                 .rpc_resp = &res,
2948         };
2949         int status;
2950
2951         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
2952                      FATTR4_WORD0_FH_EXPIRE_TYPE |
2953                      FATTR4_WORD0_LINK_SUPPORT |
2954                      FATTR4_WORD0_SYMLINK_SUPPORT |
2955                      FATTR4_WORD0_ACLSUPPORT;
2956         if (minorversion)
2957                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
2958
2959         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2960         if (status == 0) {
2961                 /* Sanity check the server answers */
2962                 switch (minorversion) {
2963                 case 0:
2964                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2965                         res.attr_bitmask[2] = 0;
2966                         break;
2967                 case 1:
2968                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2969                         break;
2970                 case 2:
2971                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2972                 }
2973                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2974                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2975                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2976                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2977                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2978                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2979                                 NFS_CAP_SECURITY_LABEL);
2980                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2981                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2982                         server->caps |= NFS_CAP_ACLS;
2983                 if (res.has_links != 0)
2984                         server->caps |= NFS_CAP_HARDLINKS;
2985                 if (res.has_symlinks != 0)
2986                         server->caps |= NFS_CAP_SYMLINKS;
2987                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2988                         server->caps |= NFS_CAP_FILEID;
2989                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2990                         server->caps |= NFS_CAP_MODE;
2991                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2992                         server->caps |= NFS_CAP_NLINK;
2993                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2994                         server->caps |= NFS_CAP_OWNER;
2995                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2996                         server->caps |= NFS_CAP_OWNER_GROUP;
2997                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2998                         server->caps |= NFS_CAP_ATIME;
2999                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3000                         server->caps |= NFS_CAP_CTIME;
3001                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3002                         server->caps |= NFS_CAP_MTIME;
3003 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3004                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3005                         server->caps |= NFS_CAP_SECURITY_LABEL;
3006 #endif
3007                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3008                                 sizeof(server->attr_bitmask));
3009                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3010
3011                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3012                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3013                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3014                 server->cache_consistency_bitmask[2] = 0;
3015                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3016                         sizeof(server->exclcreat_bitmask));
3017                 server->acl_bitmask = res.acl_bitmask;
3018                 server->fh_expire_type = res.fh_expire_type;
3019         }
3020
3021         return status;
3022 }
3023
3024 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3025 {
3026         struct nfs4_exception exception = { };
3027         int err;
3028         do {
3029                 err = nfs4_handle_exception(server,
3030                                 _nfs4_server_capabilities(server, fhandle),
3031                                 &exception);
3032         } while (exception.retry);
3033         return err;
3034 }
3035
3036 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3037                 struct nfs_fsinfo *info)
3038 {
3039         u32 bitmask[3];
3040         struct nfs4_lookup_root_arg args = {
3041                 .bitmask = bitmask,
3042         };
3043         struct nfs4_lookup_res res = {
3044                 .server = server,
3045                 .fattr = info->fattr,
3046                 .fh = fhandle,
3047         };
3048         struct rpc_message msg = {
3049                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3050                 .rpc_argp = &args,
3051                 .rpc_resp = &res,
3052         };
3053
3054         bitmask[0] = nfs4_fattr_bitmap[0];
3055         bitmask[1] = nfs4_fattr_bitmap[1];
3056         /*
3057          * Process the label in the upcoming getfattr
3058          */
3059         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3060
3061         nfs_fattr_init(info->fattr);
3062         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3063 }
3064
3065 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3066                 struct nfs_fsinfo *info)
3067 {
3068         struct nfs4_exception exception = { };
3069         int err;
3070         do {
3071                 err = _nfs4_lookup_root(server, fhandle, info);
3072                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3073                 switch (err) {
3074                 case 0:
3075                 case -NFS4ERR_WRONGSEC:
3076                         goto out;
3077                 default:
3078                         err = nfs4_handle_exception(server, err, &exception);
3079                 }
3080         } while (exception.retry);
3081 out:
3082         return err;
3083 }
3084
3085 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3086                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3087 {
3088         struct rpc_auth_create_args auth_args = {
3089                 .pseudoflavor = flavor,
3090         };
3091         struct rpc_auth *auth;
3092         int ret;
3093
3094         auth = rpcauth_create(&auth_args, server->client);
3095         if (IS_ERR(auth)) {
3096                 ret = -EACCES;
3097                 goto out;
3098         }
3099         ret = nfs4_lookup_root(server, fhandle, info);
3100 out:
3101         return ret;
3102 }
3103
3104 /*
3105  * Retry pseudoroot lookup with various security flavors.  We do this when:
3106  *
3107  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3108  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3109  *
3110  * Returns zero on success, or a negative NFS4ERR value, or a
3111  * negative errno value.
3112  */
3113 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3114                               struct nfs_fsinfo *info)
3115 {
3116         /* Per 3530bis 15.33.5 */
3117         static const rpc_authflavor_t flav_array[] = {
3118                 RPC_AUTH_GSS_KRB5P,
3119                 RPC_AUTH_GSS_KRB5I,
3120                 RPC_AUTH_GSS_KRB5,
3121                 RPC_AUTH_UNIX,                  /* courtesy */
3122                 RPC_AUTH_NULL,
3123         };
3124         int status = -EPERM;
3125         size_t i;
3126
3127         if (server->auth_info.flavor_len > 0) {
3128                 /* try each flavor specified by user */
3129                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3130                         status = nfs4_lookup_root_sec(server, fhandle, info,
3131                                                 server->auth_info.flavors[i]);
3132                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3133                                 continue;
3134                         break;
3135                 }
3136         } else {
3137                 /* no flavors specified by user, try default list */
3138                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3139                         status = nfs4_lookup_root_sec(server, fhandle, info,
3140                                                       flav_array[i]);
3141                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3142                                 continue;
3143                         break;
3144                 }
3145         }
3146
3147         /*
3148          * -EACCESS could mean that the user doesn't have correct permissions
3149          * to access the mount.  It could also mean that we tried to mount
3150          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3151          * existing mount programs don't handle -EACCES very well so it should
3152          * be mapped to -EPERM instead.
3153          */
3154         if (status == -EACCES)
3155                 status = -EPERM;
3156         return status;
3157 }
3158
3159 static int nfs4_do_find_root_sec(struct nfs_server *server,
3160                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3161 {
3162         int mv = server->nfs_client->cl_minorversion;
3163         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3164 }
3165
3166 /**
3167  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3168  * @server: initialized nfs_server handle
3169  * @fhandle: we fill in the pseudo-fs root file handle
3170  * @info: we fill in an FSINFO struct
3171  * @auth_probe: probe the auth flavours
3172  *
3173  * Returns zero on success, or a negative errno.
3174  */
3175 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3176                          struct nfs_fsinfo *info,
3177                          bool auth_probe)
3178 {
3179         int status = 0;
3180
3181         if (!auth_probe)
3182                 status = nfs4_lookup_root(server, fhandle, info);
3183
3184         if (auth_probe || status == NFS4ERR_WRONGSEC)
3185                 status = nfs4_do_find_root_sec(server, fhandle, info);
3186
3187         if (status == 0)
3188                 status = nfs4_server_capabilities(server, fhandle);
3189         if (status == 0)
3190                 status = nfs4_do_fsinfo(server, fhandle, info);
3191
3192         return nfs4_map_errors(status);
3193 }
3194
3195 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3196                               struct nfs_fsinfo *info)
3197 {
3198         int error;
3199         struct nfs_fattr *fattr = info->fattr;
3200         struct nfs4_label *label = NULL;
3201
3202         error = nfs4_server_capabilities(server, mntfh);
3203         if (error < 0) {
3204                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3205                 return error;
3206         }
3207
3208         label = nfs4_label_alloc(server, GFP_KERNEL);
3209         if (IS_ERR(label))
3210                 return PTR_ERR(label);
3211
3212         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3213         if (error < 0) {
3214                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3215                 goto err_free_label;
3216         }
3217
3218         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3219             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3220                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3221
3222 err_free_label:
3223         nfs4_label_free(label);
3224
3225         return error;
3226 }
3227
3228 /*
3229  * Get locations and (maybe) other attributes of a referral.
3230  * Note that we'll actually follow the referral later when
3231  * we detect fsid mismatch in inode revalidation
3232  */
3233 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3234                              const struct qstr *name, struct nfs_fattr *fattr,
3235                              struct nfs_fh *fhandle)
3236 {
3237         int status = -ENOMEM;
3238         struct page *page = NULL;
3239         struct nfs4_fs_locations *locations = NULL;
3240
3241         page = alloc_page(GFP_KERNEL);
3242         if (page == NULL)
3243                 goto out;
3244         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3245         if (locations == NULL)
3246                 goto out;
3247
3248         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3249         if (status != 0)
3250                 goto out;
3251
3252         /*
3253          * If the fsid didn't change, this is a migration event, not a
3254          * referral.  Cause us to drop into the exception handler, which
3255          * will kick off migration recovery.
3256          */
3257         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3258                 dprintk("%s: server did not return a different fsid for"
3259                         " a referral at %s\n", __func__, name->name);
3260                 status = -NFS4ERR_MOVED;
3261                 goto out;
3262         }
3263         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3264         nfs_fixup_referral_attributes(&locations->fattr);
3265
3266         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3267         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3268         memset(fhandle, 0, sizeof(struct nfs_fh));
3269 out:
3270         if (page)
3271                 __free_page(page);
3272         kfree(locations);
3273         return status;
3274 }
3275
3276 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3277                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3278 {
3279         struct nfs4_getattr_arg args = {
3280                 .fh = fhandle,
3281                 .bitmask = server->attr_bitmask,
3282         };
3283         struct nfs4_getattr_res res = {
3284                 .fattr = fattr,
3285                 .label = label,
3286                 .server = server,
3287         };
3288         struct rpc_message msg = {
3289                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3290                 .rpc_argp = &args,
3291                 .rpc_resp = &res,
3292         };
3293
3294         args.bitmask = nfs4_bitmask(server, label);
3295
3296         nfs_fattr_init(fattr);
3297         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3298 }
3299
3300 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3301                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3302 {
3303         struct nfs4_exception exception = { };
3304         int err;
3305         do {
3306                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3307                 trace_nfs4_getattr(server, fhandle, fattr, err);
3308                 err = nfs4_handle_exception(server, err,
3309                                 &exception);
3310         } while (exception.retry);
3311         return err;
3312 }
3313
3314 /* 
3315  * The file is not closed if it is opened due to the a request to change
3316  * the size of the file. The open call will not be needed once the
3317  * VFS layer lookup-intents are implemented.
3318  *
3319  * Close is called when the inode is destroyed.
3320  * If we haven't opened the file for O_WRONLY, we
3321  * need to in the size_change case to obtain a stateid.
3322  *
3323  * Got race?
3324  * Because OPEN is always done by name in nfsv4, it is
3325  * possible that we opened a different file by the same
3326  * name.  We can recognize this race condition, but we
3327  * can't do anything about it besides returning an error.
3328  *
3329  * This will be fixed with VFS changes (lookup-intent).
3330  */
3331 static int
3332 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3333                   struct iattr *sattr)
3334 {
3335         struct inode *inode = d_inode(dentry);
3336         struct rpc_cred *cred = NULL;
3337         struct nfs4_state *state = NULL;
3338         struct nfs4_label *label = NULL;
3339         int status;
3340
3341         if (pnfs_ld_layoutret_on_setattr(inode) &&
3342             sattr->ia_valid & ATTR_SIZE &&
3343             sattr->ia_size < i_size_read(inode))
3344                 pnfs_commit_and_return_layout(inode);
3345
3346         nfs_fattr_init(fattr);
3347         
3348         /* Deal with open(O_TRUNC) */
3349         if (sattr->ia_valid & ATTR_OPEN)
3350                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3351
3352         /* Optimization: if the end result is no change, don't RPC */
3353         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3354                 return 0;
3355
3356         /* Search for an existing open(O_WRITE) file */
3357         if (sattr->ia_valid & ATTR_FILE) {
3358                 struct nfs_open_context *ctx;
3359
3360                 ctx = nfs_file_open_context(sattr->ia_file);
3361                 if (ctx) {
3362                         cred = ctx->cred;
3363                         state = ctx->state;
3364                 }
3365         }
3366
3367         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3368         if (IS_ERR(label))
3369                 return PTR_ERR(label);
3370
3371         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3372         if (status == 0) {
3373                 nfs_setattr_update_inode(inode, sattr, fattr);
3374                 nfs_setsecurity(inode, fattr, label);
3375         }
3376         nfs4_label_free(label);
3377         return status;
3378 }
3379
3380 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3381                 const struct qstr *name, struct nfs_fh *fhandle,
3382                 struct nfs_fattr *fattr, struct nfs4_label *label)
3383 {
3384         struct nfs_server *server = NFS_SERVER(dir);
3385         int                    status;
3386         struct nfs4_lookup_arg args = {
3387                 .bitmask = server->attr_bitmask,
3388                 .dir_fh = NFS_FH(dir),
3389                 .name = name,
3390         };
3391         struct nfs4_lookup_res res = {
3392                 .server = server,
3393                 .fattr = fattr,
3394                 .label = label,
3395                 .fh = fhandle,
3396         };
3397         struct rpc_message msg = {
3398                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3399                 .rpc_argp = &args,
3400                 .rpc_resp = &res,
3401         };
3402
3403         args.bitmask = nfs4_bitmask(server, label);
3404
3405         nfs_fattr_init(fattr);
3406
3407         dprintk("NFS call  lookup %s\n", name->name);
3408         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3409         dprintk("NFS reply lookup: %d\n", status);
3410         return status;
3411 }
3412
3413 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3414 {
3415         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3416                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3417         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3418         fattr->nlink = 2;
3419 }
3420
3421 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3422                                    struct qstr *name, struct nfs_fh *fhandle,
3423                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3424 {
3425         struct nfs4_exception exception = { };
3426         struct rpc_clnt *client = *clnt;
3427         int err;
3428         do {
3429                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3430                 trace_nfs4_lookup(dir, name, err);
3431                 switch (err) {
3432                 case -NFS4ERR_BADNAME:
3433                         err = -ENOENT;
3434                         goto out;
3435                 case -NFS4ERR_MOVED:
3436                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3437                         if (err == -NFS4ERR_MOVED)
3438                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3439                         goto out;
3440                 case -NFS4ERR_WRONGSEC:
3441                         err = -EPERM;
3442                         if (client != *clnt)
3443                                 goto out;
3444                         client = nfs4_negotiate_security(client, dir, name);
3445                         if (IS_ERR(client))
3446                                 return PTR_ERR(client);
3447
3448                         exception.retry = 1;
3449                         break;
3450                 default:
3451                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3452                 }
3453         } while (exception.retry);
3454
3455 out:
3456         if (err == 0)
3457                 *clnt = client;
3458         else if (client != *clnt)
3459                 rpc_shutdown_client(client);
3460
3461         return err;
3462 }
3463
3464 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3465                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3466                             struct nfs4_label *label)
3467 {
3468         int status;
3469         struct rpc_clnt *client = NFS_CLIENT(dir);
3470
3471         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3472         if (client != NFS_CLIENT(dir)) {
3473                 rpc_shutdown_client(client);
3474                 nfs_fixup_secinfo_attributes(fattr);
3475         }
3476         return status;
3477 }
3478
3479 struct rpc_clnt *
3480 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3481                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3482 {
3483         struct rpc_clnt *client = NFS_CLIENT(dir);
3484         int status;
3485
3486         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3487         if (status < 0)
3488                 return ERR_PTR(status);
3489         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3490 }
3491
3492 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3493 {
3494         struct nfs_server *server = NFS_SERVER(inode);
3495         struct nfs4_accessargs args = {
3496                 .fh = NFS_FH(inode),
3497                 .bitmask = server->cache_consistency_bitmask,
3498         };
3499         struct nfs4_accessres res = {
3500                 .server = server,
3501         };
3502         struct rpc_message msg = {
3503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3504                 .rpc_argp = &args,
3505                 .rpc_resp = &res,
3506                 .rpc_cred = entry->cred,
3507         };
3508         int mode = entry->mask;
3509         int status = 0;
3510
3511         /*
3512          * Determine which access bits we want to ask for...
3513          */
3514         if (mode & MAY_READ)
3515                 args.access |= NFS4_ACCESS_READ;
3516         if (S_ISDIR(inode->i_mode)) {
3517                 if (mode & MAY_WRITE)
3518                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3519                 if (mode & MAY_EXEC)
3520                         args.access |= NFS4_ACCESS_LOOKUP;
3521         } else {
3522                 if (mode & MAY_WRITE)
3523                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3524                 if (mode & MAY_EXEC)
3525                         args.access |= NFS4_ACCESS_EXECUTE;
3526         }
3527
3528         res.fattr = nfs_alloc_fattr();
3529         if (res.fattr == NULL)
3530                 return -ENOMEM;
3531
3532         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3533         if (!status) {
3534                 nfs_access_set_mask(entry, res.access);
3535                 nfs_refresh_inode(inode, res.fattr);
3536         }
3537         nfs_free_fattr(res.fattr);
3538         return status;
3539 }
3540
3541 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3542 {
3543         struct nfs4_exception exception = { };
3544         int err;
3545         do {
3546                 err = _nfs4_proc_access(inode, entry);
3547                 trace_nfs4_access(inode, err);
3548                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3549                                 &exception);
3550         } while (exception.retry);
3551         return err;
3552 }
3553
3554 /*
3555  * TODO: For the time being, we don't try to get any attributes
3556  * along with any of the zero-copy operations READ, READDIR,
3557  * READLINK, WRITE.
3558  *
3559  * In the case of the first three, we want to put the GETATTR
3560  * after the read-type operation -- this is because it is hard
3561  * to predict the length of a GETATTR response in v4, and thus
3562  * align the READ data correctly.  This means that the GETATTR
3563  * may end up partially falling into the page cache, and we should
3564  * shift it into the 'tail' of the xdr_buf before processing.
3565  * To do this efficiently, we need to know the total length
3566  * of data received, which doesn't seem to be available outside
3567  * of the RPC layer.
3568  *
3569  * In the case of WRITE, we also want to put the GETATTR after
3570  * the operation -- in this case because we want to make sure
3571  * we get the post-operation mtime and size.
3572  *
3573  * Both of these changes to the XDR layer would in fact be quite
3574  * minor, but I decided to leave them for a subsequent patch.
3575  */
3576 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3577                 unsigned int pgbase, unsigned int pglen)
3578 {
3579         struct nfs4_readlink args = {
3580                 .fh       = NFS_FH(inode),
3581                 .pgbase   = pgbase,
3582                 .pglen    = pglen,
3583                 .pages    = &page,
3584         };
3585         struct nfs4_readlink_res res;
3586         struct rpc_message msg = {
3587                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3588                 .rpc_argp = &args,
3589                 .rpc_resp = &res,
3590         };
3591
3592         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3593 }
3594
3595 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3596                 unsigned int pgbase, unsigned int pglen)
3597 {
3598         struct nfs4_exception exception = { };
3599         int err;
3600         do {
3601                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3602                 trace_nfs4_readlink(inode, err);
3603                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3604                                 &exception);
3605         } while (exception.retry);
3606         return err;
3607 }
3608
3609 /*
3610  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3611  */
3612 static int
3613 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3614                  int flags)
3615 {
3616         struct nfs4_label l, *ilabel = NULL;
3617         struct nfs_open_context *ctx;
3618         struct nfs4_state *state;
3619         int status = 0;
3620
3621         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3622         if (IS_ERR(ctx))
3623                 return PTR_ERR(ctx);
3624
3625         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3626
3627         sattr->ia_mode &= ~current_umask();
3628         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3629         if (IS_ERR(state)) {
3630                 status = PTR_ERR(state);
3631                 goto out;
3632         }
3633 out:
3634         nfs4_label_release_security(ilabel);
3635         put_nfs_open_context(ctx);
3636         return status;
3637 }
3638
3639 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3640 {
3641         struct nfs_server *server = NFS_SERVER(dir);
3642         struct nfs_removeargs args = {
3643                 .fh = NFS_FH(dir),
3644                 .name = *name,
3645         };
3646         struct nfs_removeres res = {
3647                 .server = server,
3648         };
3649         struct rpc_message msg = {
3650                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3651                 .rpc_argp = &args,
3652                 .rpc_resp = &res,
3653         };
3654         int status;
3655
3656         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3657         if (status == 0)
3658                 update_changeattr(dir, &res.cinfo);
3659         return status;
3660 }
3661
3662 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3663 {
3664         struct nfs4_exception exception = { };
3665         int err;
3666         do {
3667                 err = _nfs4_proc_remove(dir, name);
3668                 trace_nfs4_remove(dir, name, err);
3669                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3670                                 &exception);
3671         } while (exception.retry);
3672         return err;
3673 }
3674
3675 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3676 {
3677         struct nfs_server *server = NFS_SERVER(dir);
3678         struct nfs_removeargs *args = msg->rpc_argp;
3679         struct nfs_removeres *res = msg->rpc_resp;
3680
3681         res->server = server;
3682         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3683         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3684
3685         nfs_fattr_init(res->dir_attr);
3686 }
3687
3688 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3689 {
3690         nfs4_setup_sequence(NFS_SERVER(data->dir),
3691                         &data->args.seq_args,
3692                         &data->res.seq_res,
3693                         task);
3694 }
3695
3696 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3697 {
3698         struct nfs_unlinkdata *data = task->tk_calldata;
3699         struct nfs_removeres *res = &data->res;
3700
3701         if (!nfs4_sequence_done(task, &res->seq_res))
3702                 return 0;
3703         if (nfs4_async_handle_error(task, res->server, NULL,
3704                                     &data->timeout) == -EAGAIN)
3705                 return 0;
3706         update_changeattr(dir, &res->cinfo);
3707         return 1;
3708 }
3709
3710 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3711 {
3712         struct nfs_server *server = NFS_SERVER(dir);
3713         struct nfs_renameargs *arg = msg->rpc_argp;
3714         struct nfs_renameres *res = msg->rpc_resp;
3715
3716         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3717         res->server = server;
3718         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3719 }
3720
3721 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3722 {
3723         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3724                         &data->args.seq_args,
3725                         &data->res.seq_res,
3726                         task);
3727 }
3728
3729 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3730                                  struct inode *new_dir)
3731 {
3732         struct nfs_renamedata *data = task->tk_calldata;
3733         struct nfs_renameres *res = &data->res;
3734
3735         if (!nfs4_sequence_done(task, &res->seq_res))
3736                 return 0;
3737         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3738                 return 0;
3739
3740         update_changeattr(old_dir, &res->old_cinfo);
3741         update_changeattr(new_dir, &res->new_cinfo);
3742         return 1;
3743 }
3744
3745 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3746 {
3747         struct nfs_server *server = NFS_SERVER(inode);
3748         struct nfs4_link_arg arg = {
3749                 .fh     = NFS_FH(inode),
3750                 .dir_fh = NFS_FH(dir),
3751                 .name   = name,
3752                 .bitmask = server->attr_bitmask,
3753         };
3754         struct nfs4_link_res res = {
3755                 .server = server,
3756                 .label = NULL,
3757         };
3758         struct rpc_message msg = {
3759                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3760                 .rpc_argp = &arg,
3761                 .rpc_resp = &res,
3762         };
3763         int status = -ENOMEM;
3764
3765         res.fattr = nfs_alloc_fattr();
3766         if (res.fattr == NULL)
3767                 goto out;
3768
3769         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3770         if (IS_ERR(res.label)) {
3771                 status = PTR_ERR(res.label);
3772                 goto out;
3773         }
3774         arg.bitmask = nfs4_bitmask(server, res.label);
3775
3776         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3777         if (!status) {
3778                 update_changeattr(dir, &res.cinfo);
3779                 status = nfs_post_op_update_inode(inode, res.fattr);
3780                 if (!status)
3781                         nfs_setsecurity(inode, res.fattr, res.label);
3782         }
3783
3784
3785         nfs4_label_free(res.label);
3786
3787 out:
3788         nfs_free_fattr(res.fattr);
3789         return status;
3790 }
3791
3792 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3793 {
3794         struct nfs4_exception exception = { };
3795         int err;
3796         do {
3797                 err = nfs4_handle_exception(NFS_SERVER(inode),
3798                                 _nfs4_proc_link(inode, dir, name),
3799                                 &exception);
3800         } while (exception.retry);
3801         return err;
3802 }
3803
3804 struct nfs4_createdata {
3805         struct rpc_message msg;
3806         struct nfs4_create_arg arg;
3807         struct nfs4_create_res res;
3808         struct nfs_fh fh;
3809         struct nfs_fattr fattr;
3810         struct nfs4_label *label;
3811 };
3812
3813 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3814                 struct qstr *name, struct iattr *sattr, u32 ftype)
3815 {
3816         struct nfs4_createdata *data;
3817
3818         data = kzalloc(sizeof(*data), GFP_KERNEL);
3819         if (data != NULL) {
3820                 struct nfs_server *server = NFS_SERVER(dir);
3821
3822                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3823                 if (IS_ERR(data->label))
3824                         goto out_free;
3825
3826                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3827                 data->msg.rpc_argp = &data->arg;
3828                 data->msg.rpc_resp = &data->res;
3829                 data->arg.dir_fh = NFS_FH(dir);
3830                 data->arg.server = server;
3831                 data->arg.name = name;
3832                 data->arg.attrs = sattr;
3833                 data->arg.ftype = ftype;
3834                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3835                 data->res.server = server;
3836                 data->res.fh = &data->fh;
3837                 data->res.fattr = &data->fattr;
3838                 data->res.label = data->label;
3839                 nfs_fattr_init(data->res.fattr);
3840         }
3841         return data;
3842 out_free:
3843         kfree(data);
3844         return NULL;
3845 }
3846
3847 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3848 {
3849         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3850                                     &data->arg.seq_args, &data->res.seq_res, 1);
3851         if (status == 0) {
3852                 update_changeattr(dir, &data->res.dir_cinfo);
3853                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3854         }
3855         return status;
3856 }
3857
3858 static void nfs4_free_createdata(struct nfs4_createdata *data)
3859 {
3860         nfs4_label_free(data->label);
3861         kfree(data);
3862 }
3863
3864 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3865                 struct page *page, unsigned int len, struct iattr *sattr,
3866                 struct nfs4_label *label)
3867 {
3868         struct nfs4_createdata *data;
3869         int status = -ENAMETOOLONG;
3870
3871         if (len > NFS4_MAXPATHLEN)
3872                 goto out;
3873
3874         status = -ENOMEM;
3875         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3876         if (data == NULL)
3877                 goto out;
3878
3879         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3880         data->arg.u.symlink.pages = &page;
3881         data->arg.u.symlink.len = len;
3882         data->arg.label = label;
3883         
3884         status = nfs4_do_create(dir, dentry, data);
3885
3886         nfs4_free_createdata(data);
3887 out:
3888         return status;
3889 }
3890
3891 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3892                 struct page *page, unsigned int len, struct iattr *sattr)
3893 {
3894         struct nfs4_exception exception = { };
3895         struct nfs4_label l, *label = NULL;
3896         int err;
3897
3898         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3899
3900         do {
3901                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3902                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3903                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3904                                 &exception);
3905         } while (exception.retry);
3906
3907         nfs4_label_release_security(label);
3908         return err;
3909 }
3910
3911 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3912                 struct iattr *sattr, struct nfs4_label *label)
3913 {
3914         struct nfs4_createdata *data;
3915         int status = -ENOMEM;
3916
3917         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3918         if (data == NULL)
3919                 goto out;
3920
3921         data->arg.label = label;
3922         status = nfs4_do_create(dir, dentry, data);
3923
3924         nfs4_free_createdata(data);
3925 out:
3926         return status;
3927 }
3928
3929 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3930                 struct iattr *sattr)
3931 {
3932         struct nfs4_exception exception = { };
3933         struct nfs4_label l, *label = NULL;
3934         int err;
3935
3936         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3937
3938         sattr->ia_mode &= ~current_umask();
3939         do {
3940                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3941                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3942                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3943                                 &exception);
3944         } while (exception.retry);
3945         nfs4_label_release_security(label);
3946
3947         return err;
3948 }
3949
3950 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3951                 u64 cookie, struct page **pages, unsigned int count, int plus)
3952 {
3953         struct inode            *dir = d_inode(dentry);
3954         struct nfs4_readdir_arg args = {
3955                 .fh = NFS_FH(dir),
3956                 .pages = pages,
3957                 .pgbase = 0,
3958                 .count = count,
3959                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3960                 .plus = plus,
3961         };
3962         struct nfs4_readdir_res res;
3963         struct rpc_message msg = {
3964                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3965                 .rpc_argp = &args,
3966                 .rpc_resp = &res,
3967                 .rpc_cred = cred,
3968         };
3969         int                     status;
3970
3971         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3972                         dentry,
3973                         (unsigned long long)cookie);
3974         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3975         res.pgbase = args.pgbase;
3976         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3977         if (status >= 0) {
3978                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3979                 status += args.pgbase;
3980         }
3981
3982         nfs_invalidate_atime(dir);
3983
3984         dprintk("%s: returns %d\n", __func__, status);
3985         return status;
3986 }
3987
3988 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3989                 u64 cookie, struct page **pages, unsigned int count, int plus)
3990 {
3991         struct nfs4_exception exception = { };
3992         int err;
3993         do {
3994                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3995                                 pages, count, plus);
3996                 trace_nfs4_readdir(d_inode(dentry), err);
3997                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3998                                 &exception);
3999         } while (exception.retry);
4000         return err;
4001 }
4002
4003 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4004                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4005 {
4006         struct nfs4_createdata *data;
4007         int mode = sattr->ia_mode;
4008         int status = -ENOMEM;
4009
4010         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4011         if (data == NULL)
4012                 goto out;
4013
4014         if (S_ISFIFO(mode))
4015                 data->arg.ftype = NF4FIFO;
4016         else if (S_ISBLK(mode)) {
4017                 data->arg.ftype = NF4BLK;
4018                 data->arg.u.device.specdata1 = MAJOR(rdev);
4019                 data->arg.u.device.specdata2 = MINOR(rdev);
4020         }
4021         else if (S_ISCHR(mode)) {
4022                 data->arg.ftype = NF4CHR;
4023                 data->arg.u.device.specdata1 = MAJOR(rdev);
4024                 data->arg.u.device.specdata2 = MINOR(rdev);
4025         } else if (!S_ISSOCK(mode)) {
4026                 status = -EINVAL;
4027                 goto out_free;
4028         }
4029
4030         data->arg.label = label;
4031         status = nfs4_do_create(dir, dentry, data);
4032 out_free:
4033         nfs4_free_createdata(data);
4034 out:
4035         return status;
4036 }
4037
4038 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4039                 struct iattr *sattr, dev_t rdev)
4040 {
4041         struct nfs4_exception exception = { };
4042         struct nfs4_label l, *label = NULL;
4043         int err;
4044
4045         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4046
4047         sattr->ia_mode &= ~current_umask();
4048         do {
4049                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4050                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4051                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4052                                 &exception);
4053         } while (exception.retry);
4054
4055         nfs4_label_release_security(label);
4056
4057         return err;
4058 }
4059
4060 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4061                  struct nfs_fsstat *fsstat)
4062 {
4063         struct nfs4_statfs_arg args = {
4064                 .fh = fhandle,
4065                 .bitmask = server->attr_bitmask,
4066         };
4067         struct nfs4_statfs_res res = {
4068                 .fsstat = fsstat,
4069         };
4070         struct rpc_message msg = {
4071                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4072                 .rpc_argp = &args,
4073                 .rpc_resp = &res,
4074         };
4075
4076         nfs_fattr_init(fsstat->fattr);
4077         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4078 }
4079
4080 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4081 {
4082         struct nfs4_exception exception = { };
4083         int err;
4084         do {
4085                 err = nfs4_handle_exception(server,
4086                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4087                                 &exception);
4088         } while (exception.retry);
4089         return err;
4090 }
4091
4092 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4093                 struct nfs_fsinfo *fsinfo)
4094 {
4095         struct nfs4_fsinfo_arg args = {
4096                 .fh = fhandle,
4097                 .bitmask = server->attr_bitmask,
4098         };
4099         struct nfs4_fsinfo_res res = {
4100                 .fsinfo = fsinfo,
4101         };
4102         struct rpc_message msg = {
4103                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4104                 .rpc_argp = &args,
4105                 .rpc_resp = &res,
4106         };
4107
4108         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4109 }
4110
4111 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4112 {
4113         struct nfs4_exception exception = { };
4114         unsigned long now = jiffies;
4115         int err;
4116
4117         do {
4118                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4119                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4120                 if (err == 0) {
4121                         struct nfs_client *clp = server->nfs_client;
4122
4123                         spin_lock(&clp->cl_lock);
4124                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4125                         clp->cl_last_renewal = now;
4126                         spin_unlock(&clp->cl_lock);
4127                         break;
4128                 }
4129                 err = nfs4_handle_exception(server, err, &exception);
4130         } while (exception.retry);
4131         return err;
4132 }
4133
4134 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4135 {
4136         int error;
4137
4138         nfs_fattr_init(fsinfo->fattr);
4139         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4140         if (error == 0) {
4141                 /* block layout checks this! */
4142                 server->pnfs_blksize = fsinfo->blksize;
4143                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4144         }
4145
4146         return error;
4147 }
4148
4149 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4150                 struct nfs_pathconf *pathconf)
4151 {
4152         struct nfs4_pathconf_arg args = {
4153                 .fh = fhandle,
4154                 .bitmask = server->attr_bitmask,
4155         };
4156         struct nfs4_pathconf_res res = {
4157                 .pathconf = pathconf,
4158         };
4159         struct rpc_message msg = {
4160                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4161                 .rpc_argp = &args,
4162                 .rpc_resp = &res,
4163         };
4164
4165         /* None of the pathconf attributes are mandatory to implement */
4166         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4167                 memset(pathconf, 0, sizeof(*pathconf));
4168                 return 0;
4169         }
4170
4171         nfs_fattr_init(pathconf->fattr);
4172         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4173 }
4174
4175 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4176                 struct nfs_pathconf *pathconf)
4177 {
4178         struct nfs4_exception exception = { };
4179         int err;
4180
4181         do {
4182                 err = nfs4_handle_exception(server,
4183                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4184                                 &exception);
4185         } while (exception.retry);
4186         return err;
4187 }
4188
4189 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4190                 const struct nfs_open_context *ctx,
4191                 const struct nfs_lock_context *l_ctx,
4192                 fmode_t fmode)
4193 {
4194         const struct nfs_lockowner *lockowner = NULL;
4195
4196         if (l_ctx != NULL)
4197                 lockowner = &l_ctx->lockowner;
4198         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4199 }
4200 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4201
4202 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4203                 const struct nfs_open_context *ctx,
4204                 const struct nfs_lock_context *l_ctx,
4205                 fmode_t fmode)
4206 {
4207         nfs4_stateid current_stateid;
4208
4209         /* If the current stateid represents a lost lock, then exit */
4210         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4211                 return true;
4212         return nfs4_stateid_match(stateid, &current_stateid);
4213 }
4214
4215 static bool nfs4_error_stateid_expired(int err)
4216 {
4217         switch (err) {
4218         case -NFS4ERR_DELEG_REVOKED:
4219         case -NFS4ERR_ADMIN_REVOKED:
4220         case -NFS4ERR_BAD_STATEID:
4221         case -NFS4ERR_STALE_STATEID:
4222         case -NFS4ERR_OLD_STATEID:
4223         case -NFS4ERR_OPENMODE:
4224         case -NFS4ERR_EXPIRED:
4225                 return true;
4226         }
4227         return false;
4228 }
4229
4230 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4231 {
4232         nfs_invalidate_atime(hdr->inode);
4233 }
4234
4235 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4236 {
4237         struct nfs_server *server = NFS_SERVER(hdr->inode);
4238
4239         trace_nfs4_read(hdr, task->tk_status);
4240         if (nfs4_async_handle_error(task, server,
4241                                     hdr->args.context->state,
4242                                     NULL) == -EAGAIN) {
4243                 rpc_restart_call_prepare(task);
4244                 return -EAGAIN;
4245         }
4246
4247         __nfs4_read_done_cb(hdr);
4248         if (task->tk_status > 0)
4249                 renew_lease(server, hdr->timestamp);
4250         return 0;
4251 }
4252
4253 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4254                 struct nfs_pgio_args *args)
4255 {
4256
4257         if (!nfs4_error_stateid_expired(task->tk_status) ||
4258                 nfs4_stateid_is_current(&args->stateid,
4259                                 args->context,
4260                                 args->lock_context,
4261                                 FMODE_READ))
4262                 return false;
4263         rpc_restart_call_prepare(task);
4264         return true;
4265 }
4266
4267 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4268 {
4269
4270         dprintk("--> %s\n", __func__);
4271
4272         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4273                 return -EAGAIN;
4274         if (nfs4_read_stateid_changed(task, &hdr->args))
4275                 return -EAGAIN;
4276         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4277                                     nfs4_read_done_cb(task, hdr);
4278 }
4279
4280 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4281                                  struct rpc_message *msg)
4282 {
4283         hdr->timestamp   = jiffies;
4284         hdr->pgio_done_cb = nfs4_read_done_cb;
4285         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4286         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4287 }
4288
4289 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4290                                       struct nfs_pgio_header *hdr)
4291 {
4292         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4293                         &hdr->args.seq_args,
4294                         &hdr->res.seq_res,
4295                         task))
4296                 return 0;
4297         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4298                                 hdr->args.lock_context,
4299                                 hdr->rw_ops->rw_mode) == -EIO)
4300                 return -EIO;
4301         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4302                 return -EIO;
4303         return 0;
4304 }
4305
4306 static int nfs4_write_done_cb(struct rpc_task *task,
4307                               struct nfs_pgio_header *hdr)
4308 {
4309         struct inode *inode = hdr->inode;
4310
4311         trace_nfs4_write(hdr, task->tk_status);
4312         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4313                                     hdr->args.context->state,
4314                                     NULL) == -EAGAIN) {
4315                 rpc_restart_call_prepare(task);
4316                 return -EAGAIN;
4317         }
4318         if (task->tk_status >= 0) {
4319                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4320                 nfs_writeback_update_inode(hdr);
4321         }
4322         return 0;
4323 }
4324
4325 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4326                 struct nfs_pgio_args *args)
4327 {
4328
4329         if (!nfs4_error_stateid_expired(task->tk_status) ||
4330                 nfs4_stateid_is_current(&args->stateid,
4331                                 args->context,
4332                                 args->lock_context,
4333                                 FMODE_WRITE))
4334                 return false;
4335         rpc_restart_call_prepare(task);
4336         return true;
4337 }
4338
4339 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4340 {
4341         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4342                 return -EAGAIN;
4343         if (nfs4_write_stateid_changed(task, &hdr->args))
4344                 return -EAGAIN;
4345         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4346                 nfs4_write_done_cb(task, hdr);
4347 }
4348
4349 static
4350 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4351 {
4352         /* Don't request attributes for pNFS or O_DIRECT writes */
4353         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4354                 return false;
4355         /* Otherwise, request attributes if and only if we don't hold
4356          * a delegation
4357          */
4358         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4359 }
4360
4361 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4362                                   struct rpc_message *msg)
4363 {
4364         struct nfs_server *server = NFS_SERVER(hdr->inode);
4365
4366         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4367                 hdr->args.bitmask = NULL;
4368                 hdr->res.fattr = NULL;
4369         } else
4370                 hdr->args.bitmask = server->cache_consistency_bitmask;
4371
4372         if (!hdr->pgio_done_cb)
4373                 hdr->pgio_done_cb = nfs4_write_done_cb;
4374         hdr->res.server = server;
4375         hdr->timestamp   = jiffies;
4376
4377         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4378         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4379 }
4380
4381 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4382 {
4383         nfs4_setup_sequence(NFS_SERVER(data->inode),
4384                         &data->args.seq_args,
4385                         &data->res.seq_res,
4386                         task);
4387 }
4388
4389 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4390 {
4391         struct inode *inode = data->inode;
4392
4393         trace_nfs4_commit(data, task->tk_status);
4394         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4395                                     NULL, NULL) == -EAGAIN) {
4396                 rpc_restart_call_prepare(task);
4397                 return -EAGAIN;
4398         }
4399         return 0;
4400 }
4401
4402 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4403 {
4404         if (!nfs4_sequence_done(task, &data->res.seq_res))
4405                 return -EAGAIN;
4406         return data->commit_done_cb(task, data);
4407 }
4408
4409 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4410 {
4411         struct nfs_server *server = NFS_SERVER(data->inode);
4412
4413         if (data->commit_done_cb == NULL)
4414                 data->commit_done_cb = nfs4_commit_done_cb;
4415         data->res.server = server;
4416         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4417         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4418 }
4419
4420 struct nfs4_renewdata {
4421         struct nfs_client       *client;
4422         unsigned long           timestamp;
4423 };
4424
4425 /*
4426  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4427  * standalone procedure for queueing an asynchronous RENEW.
4428  */
4429 static void nfs4_renew_release(void *calldata)
4430 {
4431         struct nfs4_renewdata *data = calldata;
4432         struct nfs_client *clp = data->client;
4433
4434         if (atomic_read(&clp->cl_count) > 1)
4435                 nfs4_schedule_state_renewal(clp);
4436         nfs_put_client(clp);
4437         kfree(data);
4438 }
4439
4440 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4441 {
4442         struct nfs4_renewdata *data = calldata;
4443         struct nfs_client *clp = data->client;
4444         unsigned long timestamp = data->timestamp;
4445
4446         trace_nfs4_renew_async(clp, task->tk_status);
4447         switch (task->tk_status) {
4448         case 0:
4449                 break;
4450         case -NFS4ERR_LEASE_MOVED:
4451                 nfs4_schedule_lease_moved_recovery(clp);
4452                 break;
4453         default:
4454                 /* Unless we're shutting down, schedule state recovery! */
4455                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4456                         return;
4457                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4458                         nfs4_schedule_lease_recovery(clp);
4459                         return;
4460                 }
4461                 nfs4_schedule_path_down_recovery(clp);
4462         }
4463         do_renew_lease(clp, timestamp);
4464 }
4465
4466 static const struct rpc_call_ops nfs4_renew_ops = {
4467         .rpc_call_done = nfs4_renew_done,
4468         .rpc_release = nfs4_renew_release,
4469 };
4470
4471 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4472 {
4473         struct rpc_message msg = {
4474                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4475                 .rpc_argp       = clp,
4476                 .rpc_cred       = cred,
4477         };
4478         struct nfs4_renewdata *data;
4479
4480         if (renew_flags == 0)
4481                 return 0;
4482         if (!atomic_inc_not_zero(&clp->cl_count))
4483                 return -EIO;
4484         data = kmalloc(sizeof(*data), GFP_NOFS);
4485         if (data == NULL)
4486                 return -ENOMEM;
4487         data->client = clp;
4488         data->timestamp = jiffies;
4489         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4490                         &nfs4_renew_ops, data);
4491 }
4492
4493 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4494 {
4495         struct rpc_message msg = {
4496                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4497                 .rpc_argp       = clp,
4498                 .rpc_cred       = cred,
4499         };
4500         unsigned long now = jiffies;
4501         int status;
4502
4503         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4504         if (status < 0)
4505                 return status;
4506         do_renew_lease(clp, now);
4507         return 0;
4508 }
4509
4510 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4511 {
4512         return server->caps & NFS_CAP_ACLS;
4513 }
4514
4515 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4516  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4517  * the stack.
4518  */
4519 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4520
4521 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4522                 struct page **pages, unsigned int *pgbase)
4523 {
4524         struct page *newpage, **spages;
4525         int rc = 0;
4526         size_t len;
4527         spages = pages;
4528
4529         do {
4530                 len = min_t(size_t, PAGE_SIZE, buflen);
4531                 newpage = alloc_page(GFP_KERNEL);
4532
4533                 if (newpage == NULL)
4534                         goto unwind;
4535                 memcpy(page_address(newpage), buf, len);
4536                 buf += len;
4537                 buflen -= len;
4538                 *pages++ = newpage;
4539                 rc++;
4540         } while (buflen != 0);
4541
4542         return rc;
4543
4544 unwind:
4545         for(; rc > 0; rc--)
4546                 __free_page(spages[rc-1]);
4547         return -ENOMEM;
4548 }
4549
4550 struct nfs4_cached_acl {
4551         int cached;
4552         size_t len;
4553         char data[0];
4554 };
4555
4556 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4557 {
4558         struct nfs_inode *nfsi = NFS_I(inode);
4559
4560         spin_lock(&inode->i_lock);
4561         kfree(nfsi->nfs4_acl);
4562         nfsi->nfs4_acl = acl;
4563         spin_unlock(&inode->i_lock);
4564 }
4565
4566 static void nfs4_zap_acl_attr(struct inode *inode)
4567 {
4568         nfs4_set_cached_acl(inode, NULL);
4569 }
4570
4571 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4572 {
4573         struct nfs_inode *nfsi = NFS_I(inode);
4574         struct nfs4_cached_acl *acl;
4575         int ret = -ENOENT;
4576
4577         spin_lock(&inode->i_lock);
4578         acl = nfsi->nfs4_acl;
4579         if (acl == NULL)
4580                 goto out;
4581         if (buf == NULL) /* user is just asking for length */
4582                 goto out_len;
4583         if (acl->cached == 0)
4584                 goto out;
4585         ret = -ERANGE; /* see getxattr(2) man page */
4586         if (acl->len > buflen)
4587                 goto out;
4588         memcpy(buf, acl->data, acl->len);
4589 out_len:
4590         ret = acl->len;
4591 out:
4592         spin_unlock(&inode->i_lock);
4593         return ret;
4594 }
4595
4596 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4597 {
4598         struct nfs4_cached_acl *acl;
4599         size_t buflen = sizeof(*acl) + acl_len;
4600
4601         if (buflen <= PAGE_SIZE) {
4602                 acl = kmalloc(buflen, GFP_KERNEL);
4603                 if (acl == NULL)
4604                         goto out;
4605                 acl->cached = 1;
4606                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4607         } else {
4608                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4609                 if (acl == NULL)
4610                         goto out;
4611                 acl->cached = 0;
4612         }
4613         acl->len = acl_len;
4614 out:
4615         nfs4_set_cached_acl(inode, acl);
4616 }
4617
4618 /*
4619  * The getxattr API returns the required buffer length when called with a
4620  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4621  * the required buf.  On a NULL buf, we send a page of data to the server
4622  * guessing that the ACL request can be serviced by a page. If so, we cache
4623  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4624  * the cache. If not so, we throw away the page, and cache the required
4625  * length. The next getxattr call will then produce another round trip to
4626  * the server, this time with the input buf of the required size.
4627  */
4628 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4629 {
4630         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4631         struct nfs_getaclargs args = {
4632                 .fh = NFS_FH(inode),
4633                 .acl_pages = pages,
4634                 .acl_len = buflen,
4635         };
4636         struct nfs_getaclres res = {
4637                 .acl_len = buflen,
4638         };
4639         struct rpc_message msg = {
4640                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4641                 .rpc_argp = &args,
4642                 .rpc_resp = &res,
4643         };
4644         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4645         int ret = -ENOMEM, i;
4646
4647         /* As long as we're doing a round trip to the server anyway,
4648          * let's be prepared for a page of acl data. */
4649         if (npages == 0)
4650                 npages = 1;
4651         if (npages > ARRAY_SIZE(pages))
4652                 return -ERANGE;
4653
4654         for (i = 0; i < npages; i++) {
4655                 pages[i] = alloc_page(GFP_KERNEL);
4656                 if (!pages[i])
4657                         goto out_free;
4658         }
4659
4660         /* for decoding across pages */
4661         res.acl_scratch = alloc_page(GFP_KERNEL);
4662         if (!res.acl_scratch)
4663                 goto out_free;
4664
4665         args.acl_len = npages * PAGE_SIZE;
4666         args.acl_pgbase = 0;
4667
4668         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4669                 __func__, buf, buflen, npages, args.acl_len);
4670         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4671                              &msg, &args.seq_args, &res.seq_res, 0);
4672         if (ret)
4673                 goto out_free;
4674
4675         /* Handle the case where the passed-in buffer is too short */
4676         if (res.acl_flags & NFS4_ACL_TRUNC) {
4677                 /* Did the user only issue a request for the acl length? */
4678                 if (buf == NULL)
4679                         goto out_ok;
4680                 ret = -ERANGE;
4681                 goto out_free;
4682         }
4683         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4684         if (buf) {
4685                 if (res.acl_len > buflen) {
4686                         ret = -ERANGE;
4687                         goto out_free;
4688                 }
4689                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4690         }
4691 out_ok:
4692         ret = res.acl_len;
4693 out_free:
4694         for (i = 0; i < npages; i++)
4695                 if (pages[i])
4696                         __free_page(pages[i]);
4697         if (res.acl_scratch)
4698                 __free_page(res.acl_scratch);
4699         return ret;
4700 }
4701
4702 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4703 {
4704         struct nfs4_exception exception = { };
4705         ssize_t ret;
4706         do {
4707                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4708                 trace_nfs4_get_acl(inode, ret);
4709                 if (ret >= 0)
4710                         break;
4711                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4712         } while (exception.retry);
4713         return ret;
4714 }
4715
4716 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4717 {
4718         struct nfs_server *server = NFS_SERVER(inode);
4719         int ret;
4720
4721         if (!nfs4_server_supports_acls(server))
4722                 return -EOPNOTSUPP;
4723         ret = nfs_revalidate_inode(server, inode);
4724         if (ret < 0)
4725                 return ret;
4726         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4727                 nfs_zap_acl_cache(inode);
4728         ret = nfs4_read_cached_acl(inode, buf, buflen);
4729         if (ret != -ENOENT)
4730                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4731                  * but no cached acl data, just the acl length */
4732                 return ret;
4733         return nfs4_get_acl_uncached(inode, buf, buflen);
4734 }
4735
4736 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4737 {
4738         struct nfs_server *server = NFS_SERVER(inode);
4739         struct page *pages[NFS4ACL_MAXPAGES];
4740         struct nfs_setaclargs arg = {
4741                 .fh             = NFS_FH(inode),
4742                 .acl_pages      = pages,
4743                 .acl_len        = buflen,
4744         };
4745         struct nfs_setaclres res;
4746         struct rpc_message msg = {
4747                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4748                 .rpc_argp       = &arg,
4749                 .rpc_resp       = &res,
4750         };
4751         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4752         int ret, i;
4753
4754         if (!nfs4_server_supports_acls(server))
4755                 return -EOPNOTSUPP;
4756         if (npages > ARRAY_SIZE(pages))
4757                 return -ERANGE;
4758         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4759         if (i < 0)
4760                 return i;
4761         nfs4_inode_return_delegation(inode);
4762         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4763
4764         /*
4765          * Free each page after tx, so the only ref left is
4766          * held by the network stack
4767          */
4768         for (; i > 0; i--)
4769                 put_page(pages[i-1]);
4770
4771         /*
4772          * Acl update can result in inode attribute update.
4773          * so mark the attribute cache invalid.
4774          */
4775         spin_lock(&inode->i_lock);
4776         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4777         spin_unlock(&inode->i_lock);
4778         nfs_access_zap_cache(inode);
4779         nfs_zap_acl_cache(inode);
4780         return ret;
4781 }
4782
4783 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4784 {
4785         struct nfs4_exception exception = { };
4786         int err;
4787         do {
4788                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4789                 trace_nfs4_set_acl(inode, err);
4790                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4791                                 &exception);
4792         } while (exception.retry);
4793         return err;
4794 }
4795
4796 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4797 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4798                                         size_t buflen)
4799 {
4800         struct nfs_server *server = NFS_SERVER(inode);
4801         struct nfs_fattr fattr;
4802         struct nfs4_label label = {0, 0, buflen, buf};
4803
4804         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4805         struct nfs4_getattr_arg arg = {
4806                 .fh             = NFS_FH(inode),
4807                 .bitmask        = bitmask,
4808         };
4809         struct nfs4_getattr_res res = {
4810                 .fattr          = &fattr,
4811                 .label          = &label,
4812                 .server         = server,
4813         };
4814         struct rpc_message msg = {
4815                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4816                 .rpc_argp       = &arg,
4817                 .rpc_resp       = &res,
4818         };
4819         int ret;
4820
4821         nfs_fattr_init(&fattr);
4822
4823         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4824         if (ret)
4825                 return ret;
4826         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4827                 return -ENOENT;
4828         if (buflen < label.len)
4829                 return -ERANGE;
4830         return 0;
4831 }
4832
4833 static int nfs4_get_security_label(struct inode *inode, void *buf,
4834                                         size_t buflen)
4835 {
4836         struct nfs4_exception exception = { };
4837         int err;
4838
4839         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4840                 return -EOPNOTSUPP;
4841
4842         do {
4843                 err = _nfs4_get_security_label(inode, buf, buflen);
4844                 trace_nfs4_get_security_label(inode, err);
4845                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4846                                 &exception);
4847         } while (exception.retry);
4848         return err;
4849 }
4850
4851 static int _nfs4_do_set_security_label(struct inode *inode,
4852                 struct nfs4_label *ilabel,
4853                 struct nfs_fattr *fattr,
4854                 struct nfs4_label *olabel)
4855 {
4856
4857         struct iattr sattr = {0};
4858         struct nfs_server *server = NFS_SERVER(inode);
4859         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4860         struct nfs_setattrargs arg = {
4861                 .fh             = NFS_FH(inode),
4862                 .iap            = &sattr,
4863                 .server         = server,
4864                 .bitmask        = bitmask,
4865                 .label          = ilabel,
4866         };
4867         struct nfs_setattrres res = {
4868                 .fattr          = fattr,
4869                 .label          = olabel,
4870                 .server         = server,
4871         };
4872         struct rpc_message msg = {
4873                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4874                 .rpc_argp       = &arg,
4875                 .rpc_resp       = &res,
4876         };
4877         int status;
4878
4879         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4880
4881         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4882         if (status)
4883                 dprintk("%s failed: %d\n", __func__, status);
4884
4885         return status;
4886 }
4887
4888 static int nfs4_do_set_security_label(struct inode *inode,
4889                 struct nfs4_label *ilabel,
4890                 struct nfs_fattr *fattr,
4891                 struct nfs4_label *olabel)
4892 {
4893         struct nfs4_exception exception = { };
4894         int err;
4895
4896         do {
4897                 err = _nfs4_do_set_security_label(inode, ilabel,
4898                                 fattr, olabel);
4899                 trace_nfs4_set_security_label(inode, err);
4900                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4901                                 &exception);
4902         } while (exception.retry);
4903         return err;
4904 }
4905
4906 static int
4907 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4908 {
4909         struct nfs4_label ilabel, *olabel = NULL;
4910         struct nfs_fattr fattr;
4911         struct rpc_cred *cred;
4912         struct inode *inode = d_inode(dentry);
4913         int status;
4914
4915         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4916                 return -EOPNOTSUPP;
4917
4918         nfs_fattr_init(&fattr);
4919
4920         ilabel.pi = 0;
4921         ilabel.lfs = 0;
4922         ilabel.label = (char *)buf;
4923         ilabel.len = buflen;
4924
4925         cred = rpc_lookup_cred();
4926         if (IS_ERR(cred))
4927                 return PTR_ERR(cred);
4928
4929         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4930         if (IS_ERR(olabel)) {
4931                 status = -PTR_ERR(olabel);
4932                 goto out;
4933         }
4934
4935         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4936         if (status == 0)
4937                 nfs_setsecurity(inode, &fattr, olabel);
4938
4939         nfs4_label_free(olabel);
4940 out:
4941         put_rpccred(cred);
4942         return status;
4943 }
4944 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4945
4946
4947 static int
4948 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4949                         struct nfs4_state *state, long *timeout)
4950 {
4951         struct nfs_client *clp = server->nfs_client;
4952
4953         if (task->tk_status >= 0)
4954                 return 0;
4955         switch(task->tk_status) {
4956                 case -NFS4ERR_DELEG_REVOKED:
4957                 case -NFS4ERR_ADMIN_REVOKED:
4958                 case -NFS4ERR_BAD_STATEID:
4959                 case -NFS4ERR_OPENMODE:
4960                         if (state == NULL)
4961                                 break;
4962                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4963                                 goto recovery_failed;
4964                         goto wait_on_recovery;
4965                 case -NFS4ERR_EXPIRED:
4966                         if (state != NULL) {
4967                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4968                                         goto recovery_failed;
4969                         }
4970                 case -NFS4ERR_STALE_STATEID:
4971                 case -NFS4ERR_STALE_CLIENTID:
4972                         nfs4_schedule_lease_recovery(clp);
4973                         goto wait_on_recovery;
4974                 case -NFS4ERR_MOVED:
4975                         if (nfs4_schedule_migration_recovery(server) < 0)
4976                                 goto recovery_failed;
4977                         goto wait_on_recovery;
4978                 case -NFS4ERR_LEASE_MOVED:
4979                         nfs4_schedule_lease_moved_recovery(clp);
4980                         goto wait_on_recovery;
4981 #if defined(CONFIG_NFS_V4_1)
4982                 case -NFS4ERR_BADSESSION:
4983                 case -NFS4ERR_BADSLOT:
4984                 case -NFS4ERR_BAD_HIGH_SLOT:
4985                 case -NFS4ERR_DEADSESSION:
4986                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4987                 case -NFS4ERR_SEQ_FALSE_RETRY:
4988                 case -NFS4ERR_SEQ_MISORDERED:
4989                         dprintk("%s ERROR %d, Reset session\n", __func__,
4990                                 task->tk_status);
4991                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4992                         goto wait_on_recovery;
4993 #endif /* CONFIG_NFS_V4_1 */
4994                 case -NFS4ERR_DELAY:
4995                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4996                         rpc_delay(task, nfs4_update_delay(timeout));
4997                         goto restart_call;
4998                 case -NFS4ERR_GRACE:
4999                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
5000                 case -NFS4ERR_RETRY_UNCACHED_REP:
5001                 case -NFS4ERR_OLD_STATEID:
5002                         goto restart_call;
5003         }
5004         task->tk_status = nfs4_map_errors(task->tk_status);
5005         return 0;
5006 recovery_failed:
5007         task->tk_status = -EIO;
5008         return 0;
5009 wait_on_recovery:
5010         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
5011         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
5012                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
5013         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
5014                 goto recovery_failed;
5015 restart_call:
5016         task->tk_status = 0;
5017         return -EAGAIN;
5018 }
5019
5020 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5021                                     nfs4_verifier *bootverf)
5022 {
5023         __be32 verf[2];
5024
5025         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5026                 /* An impossible timestamp guarantees this value
5027                  * will never match a generated boot time. */
5028                 verf[0] = 0;
5029                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5030         } else {
5031                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5032                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5033                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5034         }
5035         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5036 }
5037
5038 static int
5039 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5040 {
5041         int result;
5042         size_t len;
5043         char *str;
5044
5045         if (clp->cl_owner_id != NULL)
5046                 return 0;
5047
5048         rcu_read_lock();
5049         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5050                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5051                 1 +
5052                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5053                 1;
5054         rcu_read_unlock();
5055
5056         if (len > NFS4_OPAQUE_LIMIT + 1)
5057                 return -EINVAL;
5058
5059         /*
5060          * Since this string is allocated at mount time, and held until the
5061          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5062          * about a memory-reclaim deadlock.
5063          */
5064         str = kmalloc(len, GFP_KERNEL);
5065         if (!str)
5066                 return -ENOMEM;
5067
5068         rcu_read_lock();
5069         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5070                         clp->cl_ipaddr,
5071                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5072                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5073         rcu_read_unlock();
5074
5075         clp->cl_owner_id = str;
5076         return 0;
5077 }
5078
5079 static int
5080 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5081 {
5082         int result;
5083         size_t len;
5084         char *str;
5085
5086         len = 10 + 10 + 1 + 10 + 1 +
5087                 strlen(nfs4_client_id_uniquifier) + 1 +
5088                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5089
5090         if (len > NFS4_OPAQUE_LIMIT + 1)
5091                 return -EINVAL;
5092
5093         /*
5094          * Since this string is allocated at mount time, and held until the
5095          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5096          * about a memory-reclaim deadlock.
5097          */
5098         str = kmalloc(len, GFP_KERNEL);
5099         if (!str)
5100                 return -ENOMEM;
5101
5102         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5103                         clp->rpc_ops->version, clp->cl_minorversion,
5104                         nfs4_client_id_uniquifier,
5105                         clp->cl_rpcclient->cl_nodename);
5106         clp->cl_owner_id = str;
5107         return 0;
5108 }
5109
5110 static int
5111 nfs4_init_uniform_client_string(struct nfs_client *clp)
5112 {
5113         int result;
5114         size_t len;
5115         char *str;
5116
5117         if (clp->cl_owner_id != NULL)
5118                 return 0;
5119
5120         if (nfs4_client_id_uniquifier[0] != '\0')
5121                 return nfs4_init_uniquifier_client_string(clp);
5122
5123         len = 10 + 10 + 1 + 10 + 1 +
5124                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5125
5126         if (len > NFS4_OPAQUE_LIMIT + 1)
5127                 return -EINVAL;
5128
5129         /*
5130          * Since this string is allocated at mount time, and held until the
5131          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5132          * about a memory-reclaim deadlock.
5133          */
5134         str = kmalloc(len, GFP_KERNEL);
5135         if (!str)
5136                 return -ENOMEM;
5137
5138         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5139                         clp->rpc_ops->version, clp->cl_minorversion,
5140                         clp->cl_rpcclient->cl_nodename);
5141         clp->cl_owner_id = str;
5142         return 0;
5143 }
5144
5145 /*
5146  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5147  * services.  Advertise one based on the address family of the
5148  * clientaddr.
5149  */
5150 static unsigned int
5151 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5152 {
5153         if (strchr(clp->cl_ipaddr, ':') != NULL)
5154                 return scnprintf(buf, len, "tcp6");
5155         else
5156                 return scnprintf(buf, len, "tcp");
5157 }
5158
5159 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5160 {
5161         struct nfs4_setclientid *sc = calldata;
5162
5163         if (task->tk_status == 0)
5164                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5165 }
5166
5167 static const struct rpc_call_ops nfs4_setclientid_ops = {
5168         .rpc_call_done = nfs4_setclientid_done,
5169 };
5170
5171 /**
5172  * nfs4_proc_setclientid - Negotiate client ID
5173  * @clp: state data structure
5174  * @program: RPC program for NFSv4 callback service
5175  * @port: IP port number for NFS4 callback service
5176  * @cred: RPC credential to use for this call
5177  * @res: where to place the result
5178  *
5179  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5180  */
5181 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5182                 unsigned short port, struct rpc_cred *cred,
5183                 struct nfs4_setclientid_res *res)
5184 {
5185         nfs4_verifier sc_verifier;
5186         struct nfs4_setclientid setclientid = {
5187                 .sc_verifier = &sc_verifier,
5188                 .sc_prog = program,
5189                 .sc_clnt = clp,
5190         };
5191         struct rpc_message msg = {
5192                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5193                 .rpc_argp = &setclientid,
5194                 .rpc_resp = res,
5195                 .rpc_cred = cred,
5196         };
5197         struct rpc_task *task;
5198         struct rpc_task_setup task_setup_data = {
5199                 .rpc_client = clp->cl_rpcclient,
5200                 .rpc_message = &msg,
5201                 .callback_ops = &nfs4_setclientid_ops,
5202                 .callback_data = &setclientid,
5203                 .flags = RPC_TASK_TIMEOUT,
5204         };
5205         int status;
5206
5207         /* nfs_client_id4 */
5208         nfs4_init_boot_verifier(clp, &sc_verifier);
5209
5210         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5211                 status = nfs4_init_uniform_client_string(clp);
5212         else
5213                 status = nfs4_init_nonuniform_client_string(clp);
5214
5215         if (status)
5216                 goto out;
5217
5218         /* cb_client4 */
5219         setclientid.sc_netid_len =
5220                                 nfs4_init_callback_netid(clp,
5221                                                 setclientid.sc_netid,
5222                                                 sizeof(setclientid.sc_netid));
5223         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5224                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5225                                 clp->cl_ipaddr, port >> 8, port & 255);
5226
5227         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5228                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5229                 clp->cl_owner_id);
5230         task = rpc_run_task(&task_setup_data);
5231         if (IS_ERR(task)) {
5232                 status = PTR_ERR(task);
5233                 goto out;
5234         }
5235         status = task->tk_status;
5236         if (setclientid.sc_cred) {
5237                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5238                 put_rpccred(setclientid.sc_cred);
5239         }
5240         rpc_put_task(task);
5241 out:
5242         trace_nfs4_setclientid(clp, status);
5243         dprintk("NFS reply setclientid: %d\n", status);
5244         return status;
5245 }
5246
5247 /**
5248  * nfs4_proc_setclientid_confirm - Confirm client ID
5249  * @clp: state data structure
5250  * @res: result of a previous SETCLIENTID
5251  * @cred: RPC credential to use for this call
5252  *
5253  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5254  */
5255 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5256                 struct nfs4_setclientid_res *arg,
5257                 struct rpc_cred *cred)
5258 {
5259         struct rpc_message msg = {
5260                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5261                 .rpc_argp = arg,
5262                 .rpc_cred = cred,
5263         };
5264         int status;
5265
5266         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5267                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5268                 clp->cl_clientid);
5269         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5270         trace_nfs4_setclientid_confirm(clp, status);
5271         dprintk("NFS reply setclientid_confirm: %d\n", status);
5272         return status;
5273 }
5274
5275 struct nfs4_delegreturndata {
5276         struct nfs4_delegreturnargs args;
5277         struct nfs4_delegreturnres res;
5278         struct nfs_fh fh;
5279         nfs4_stateid stateid;
5280         unsigned long timestamp;
5281         struct nfs_fattr fattr;
5282         int rpc_status;
5283         struct inode *inode;
5284         bool roc;
5285         u32 roc_barrier;
5286 };
5287
5288 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5289 {
5290         struct nfs4_delegreturndata *data = calldata;
5291
5292         if (!nfs4_sequence_done(task, &data->res.seq_res))
5293                 return;
5294
5295         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5296         switch (task->tk_status) {
5297         case 0:
5298                 renew_lease(data->res.server, data->timestamp);
5299         case -NFS4ERR_ADMIN_REVOKED:
5300         case -NFS4ERR_DELEG_REVOKED:
5301         case -NFS4ERR_BAD_STATEID:
5302         case -NFS4ERR_OLD_STATEID:
5303         case -NFS4ERR_STALE_STATEID:
5304         case -NFS4ERR_EXPIRED:
5305                 task->tk_status = 0;
5306                 if (data->roc)
5307                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5308                 break;
5309         default:
5310                 if (nfs4_async_handle_error(task, data->res.server,
5311                                             NULL, NULL) == -EAGAIN) {
5312                         rpc_restart_call_prepare(task);
5313                         return;
5314                 }
5315         }
5316         data->rpc_status = task->tk_status;
5317 }
5318
5319 static void nfs4_delegreturn_release(void *calldata)
5320 {
5321         struct nfs4_delegreturndata *data = calldata;
5322         struct inode *inode = data->inode;
5323
5324         if (inode) {
5325                 if (data->roc)
5326                         pnfs_roc_release(inode);
5327                 nfs_iput_and_deactive(inode);
5328         }
5329         kfree(calldata);
5330 }
5331
5332 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5333 {
5334         struct nfs4_delegreturndata *d_data;
5335
5336         d_data = (struct nfs4_delegreturndata *)data;
5337
5338         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5339                 return;
5340
5341         if (d_data->roc)
5342                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5343
5344         nfs4_setup_sequence(d_data->res.server,
5345                         &d_data->args.seq_args,
5346                         &d_data->res.seq_res,
5347                         task);
5348 }
5349
5350 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5351         .rpc_call_prepare = nfs4_delegreturn_prepare,
5352         .rpc_call_done = nfs4_delegreturn_done,
5353         .rpc_release = nfs4_delegreturn_release,
5354 };
5355
5356 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5357 {
5358         struct nfs4_delegreturndata *data;
5359         struct nfs_server *server = NFS_SERVER(inode);
5360         struct rpc_task *task;
5361         struct rpc_message msg = {
5362                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5363                 .rpc_cred = cred,
5364         };
5365         struct rpc_task_setup task_setup_data = {
5366                 .rpc_client = server->client,
5367                 .rpc_message = &msg,
5368                 .callback_ops = &nfs4_delegreturn_ops,
5369                 .flags = RPC_TASK_ASYNC,
5370         };
5371         int status = 0;
5372
5373         data = kzalloc(sizeof(*data), GFP_NOFS);
5374         if (data == NULL)
5375                 return -ENOMEM;
5376         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5377         data->args.fhandle = &data->fh;
5378         data->args.stateid = &data->stateid;
5379         data->args.bitmask = server->cache_consistency_bitmask;
5380         nfs_copy_fh(&data->fh, NFS_FH(inode));
5381         nfs4_stateid_copy(&data->stateid, stateid);
5382         data->res.fattr = &data->fattr;
5383         data->res.server = server;
5384         nfs_fattr_init(data->res.fattr);
5385         data->timestamp = jiffies;
5386         data->rpc_status = 0;
5387         data->inode = nfs_igrab_and_active(inode);
5388         if (data->inode)
5389                 data->roc = nfs4_roc(inode);
5390
5391         task_setup_data.callback_data = data;
5392         msg.rpc_argp = &data->args;
5393         msg.rpc_resp = &data->res;
5394         task = rpc_run_task(&task_setup_data);
5395         if (IS_ERR(task))
5396                 return PTR_ERR(task);
5397         if (!issync)
5398                 goto out;
5399         status = nfs4_wait_for_completion_rpc_task(task);
5400         if (status != 0)
5401                 goto out;
5402         status = data->rpc_status;
5403         if (status == 0)
5404                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5405         else
5406                 nfs_refresh_inode(inode, &data->fattr);
5407 out:
5408         rpc_put_task(task);
5409         return status;
5410 }
5411
5412 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5413 {
5414         struct nfs_server *server = NFS_SERVER(inode);
5415         struct nfs4_exception exception = { };
5416         int err;
5417         do {
5418                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5419                 trace_nfs4_delegreturn(inode, err);
5420                 switch (err) {
5421                         case -NFS4ERR_STALE_STATEID:
5422                         case -NFS4ERR_EXPIRED:
5423                         case 0:
5424                                 return 0;
5425                 }
5426                 err = nfs4_handle_exception(server, err, &exception);
5427         } while (exception.retry);
5428         return err;
5429 }
5430
5431 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5432 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5433
5434 /* 
5435  * sleep, with exponential backoff, and retry the LOCK operation. 
5436  */
5437 static unsigned long
5438 nfs4_set_lock_task_retry(unsigned long timeout)
5439 {
5440         freezable_schedule_timeout_killable_unsafe(timeout);
5441         timeout <<= 1;
5442         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5443                 return NFS4_LOCK_MAXTIMEOUT;
5444         return timeout;
5445 }
5446
5447 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5448 {
5449         struct inode *inode = state->inode;
5450         struct nfs_server *server = NFS_SERVER(inode);
5451         struct nfs_client *clp = server->nfs_client;
5452         struct nfs_lockt_args arg = {
5453                 .fh = NFS_FH(inode),
5454                 .fl = request,
5455         };
5456         struct nfs_lockt_res res = {
5457                 .denied = request,
5458         };
5459         struct rpc_message msg = {
5460                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5461                 .rpc_argp       = &arg,
5462                 .rpc_resp       = &res,
5463                 .rpc_cred       = state->owner->so_cred,
5464         };
5465         struct nfs4_lock_state *lsp;
5466         int status;
5467
5468         arg.lock_owner.clientid = clp->cl_clientid;
5469         status = nfs4_set_lock_state(state, request);
5470         if (status != 0)
5471                 goto out;
5472         lsp = request->fl_u.nfs4_fl.owner;
5473         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5474         arg.lock_owner.s_dev = server->s_dev;
5475         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5476         switch (status) {
5477                 case 0:
5478                         request->fl_type = F_UNLCK;
5479                         break;
5480                 case -NFS4ERR_DENIED:
5481                         status = 0;
5482         }
5483         request->fl_ops->fl_release_private(request);
5484         request->fl_ops = NULL;
5485 out:
5486         return status;
5487 }
5488
5489 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5490 {
5491         struct nfs4_exception exception = { };
5492         int err;
5493
5494         do {
5495                 err = _nfs4_proc_getlk(state, cmd, request);
5496                 trace_nfs4_get_lock(request, state, cmd, err);
5497                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5498                                 &exception);
5499         } while (exception.retry);
5500         return err;
5501 }
5502
5503 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5504 {
5505         int res = 0;
5506         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5507                 case FL_POSIX:
5508                         res = posix_lock_inode_wait(inode, fl);
5509                         break;
5510                 case FL_FLOCK:
5511                         res = flock_lock_inode_wait(inode, fl);
5512                         break;
5513                 default:
5514                         BUG();
5515         }
5516         return res;
5517 }
5518
5519 struct nfs4_unlockdata {
5520         struct nfs_locku_args arg;
5521         struct nfs_locku_res res;
5522         struct nfs4_lock_state *lsp;
5523         struct nfs_open_context *ctx;
5524         struct file_lock fl;
5525         const struct nfs_server *server;
5526         unsigned long timestamp;
5527 };
5528
5529 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5530                 struct nfs_open_context *ctx,
5531                 struct nfs4_lock_state *lsp,
5532                 struct nfs_seqid *seqid)
5533 {
5534         struct nfs4_unlockdata *p;
5535         struct inode *inode = lsp->ls_state->inode;
5536
5537         p = kzalloc(sizeof(*p), GFP_NOFS);
5538         if (p == NULL)
5539                 return NULL;
5540         p->arg.fh = NFS_FH(inode);
5541         p->arg.fl = &p->fl;
5542         p->arg.seqid = seqid;
5543         p->res.seqid = seqid;
5544         p->lsp = lsp;
5545         atomic_inc(&lsp->ls_count);
5546         /* Ensure we don't close file until we're done freeing locks! */
5547         p->ctx = get_nfs_open_context(ctx);
5548         memcpy(&p->fl, fl, sizeof(p->fl));
5549         p->server = NFS_SERVER(inode);
5550         return p;
5551 }
5552
5553 static void nfs4_locku_release_calldata(void *data)
5554 {
5555         struct nfs4_unlockdata *calldata = data;
5556         nfs_free_seqid(calldata->arg.seqid);
5557         nfs4_put_lock_state(calldata->lsp);
5558         put_nfs_open_context(calldata->ctx);
5559         kfree(calldata);
5560 }
5561
5562 static void nfs4_locku_done(struct rpc_task *task, void *data)
5563 {
5564         struct nfs4_unlockdata *calldata = data;
5565
5566         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5567                 return;
5568         switch (task->tk_status) {
5569                 case 0:
5570                         renew_lease(calldata->server, calldata->timestamp);
5571                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5572                         if (nfs4_update_lock_stateid(calldata->lsp,
5573                                         &calldata->res.stateid))
5574                                 break;
5575                 case -NFS4ERR_BAD_STATEID:
5576                 case -NFS4ERR_OLD_STATEID:
5577                 case -NFS4ERR_STALE_STATEID:
5578                 case -NFS4ERR_EXPIRED:
5579                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5580                                                 &calldata->lsp->ls_stateid))
5581                                 rpc_restart_call_prepare(task);
5582                         break;
5583                 default:
5584                         if (nfs4_async_handle_error(task, calldata->server,
5585                                                     NULL, NULL) == -EAGAIN)
5586                                 rpc_restart_call_prepare(task);
5587         }
5588         nfs_release_seqid(calldata->arg.seqid);
5589 }
5590
5591 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5592 {
5593         struct nfs4_unlockdata *calldata = data;
5594
5595         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5596                 goto out_wait;
5597         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5598         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5599                 /* Note: exit _without_ running nfs4_locku_done */
5600                 goto out_no_action;
5601         }
5602         calldata->timestamp = jiffies;
5603         if (nfs4_setup_sequence(calldata->server,
5604                                 &calldata->arg.seq_args,
5605                                 &calldata->res.seq_res,
5606                                 task) != 0)
5607                 nfs_release_seqid(calldata->arg.seqid);
5608         return;
5609 out_no_action:
5610         task->tk_action = NULL;
5611 out_wait:
5612         nfs4_sequence_done(task, &calldata->res.seq_res);
5613 }
5614
5615 static const struct rpc_call_ops nfs4_locku_ops = {
5616         .rpc_call_prepare = nfs4_locku_prepare,
5617         .rpc_call_done = nfs4_locku_done,
5618         .rpc_release = nfs4_locku_release_calldata,
5619 };
5620
5621 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5622                 struct nfs_open_context *ctx,
5623                 struct nfs4_lock_state *lsp,
5624                 struct nfs_seqid *seqid)
5625 {
5626         struct nfs4_unlockdata *data;
5627         struct rpc_message msg = {
5628                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5629                 .rpc_cred = ctx->cred,
5630         };
5631         struct rpc_task_setup task_setup_data = {
5632                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5633                 .rpc_message = &msg,
5634                 .callback_ops = &nfs4_locku_ops,
5635                 .workqueue = nfsiod_workqueue,
5636                 .flags = RPC_TASK_ASYNC,
5637         };
5638
5639         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5640                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5641
5642         /* Ensure this is an unlock - when canceling a lock, the
5643          * canceled lock is passed in, and it won't be an unlock.
5644          */
5645         fl->fl_type = F_UNLCK;
5646
5647         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5648         if (data == NULL) {
5649                 nfs_free_seqid(seqid);
5650                 return ERR_PTR(-ENOMEM);
5651         }
5652
5653         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5654         msg.rpc_argp = &data->arg;
5655         msg.rpc_resp = &data->res;
5656         task_setup_data.callback_data = data;
5657         return rpc_run_task(&task_setup_data);
5658 }
5659
5660 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5661 {
5662         struct inode *inode = state->inode;
5663         struct nfs4_state_owner *sp = state->owner;
5664         struct nfs_inode *nfsi = NFS_I(inode);
5665         struct nfs_seqid *seqid;
5666         struct nfs4_lock_state *lsp;
5667         struct rpc_task *task;
5668         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5669         int status = 0;
5670         unsigned char fl_flags = request->fl_flags;
5671
5672         status = nfs4_set_lock_state(state, request);
5673         /* Unlock _before_ we do the RPC call */
5674         request->fl_flags |= FL_EXISTS;
5675         /* Exclude nfs_delegation_claim_locks() */
5676         mutex_lock(&sp->so_delegreturn_mutex);
5677         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5678         down_read(&nfsi->rwsem);
5679         if (do_vfs_lock(inode, request) == -ENOENT) {
5680                 up_read(&nfsi->rwsem);
5681                 mutex_unlock(&sp->so_delegreturn_mutex);
5682                 goto out;
5683         }
5684         up_read(&nfsi->rwsem);
5685         mutex_unlock(&sp->so_delegreturn_mutex);
5686         if (status != 0)
5687                 goto out;
5688         /* Is this a delegated lock? */
5689         lsp = request->fl_u.nfs4_fl.owner;
5690         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5691                 goto out;
5692         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5693         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5694         status = -ENOMEM;
5695         if (IS_ERR(seqid))
5696                 goto out;
5697         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5698         status = PTR_ERR(task);
5699         if (IS_ERR(task))
5700                 goto out;
5701         status = nfs4_wait_for_completion_rpc_task(task);
5702         rpc_put_task(task);
5703 out:
5704         request->fl_flags = fl_flags;
5705         trace_nfs4_unlock(request, state, F_SETLK, status);
5706         return status;
5707 }
5708
5709 struct nfs4_lockdata {
5710         struct nfs_lock_args arg;
5711         struct nfs_lock_res res;
5712         struct nfs4_lock_state *lsp;
5713         struct nfs_open_context *ctx;
5714         struct file_lock fl;
5715         unsigned long timestamp;
5716         int rpc_status;
5717         int cancelled;
5718         struct nfs_server *server;
5719 };
5720
5721 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5722                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5723                 gfp_t gfp_mask)
5724 {
5725         struct nfs4_lockdata *p;
5726         struct inode *inode = lsp->ls_state->inode;
5727         struct nfs_server *server = NFS_SERVER(inode);
5728         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5729
5730         p = kzalloc(sizeof(*p), gfp_mask);
5731         if (p == NULL)
5732                 return NULL;
5733
5734         p->arg.fh = NFS_FH(inode);
5735         p->arg.fl = &p->fl;
5736         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5737         if (IS_ERR(p->arg.open_seqid))
5738                 goto out_free;
5739         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5740         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5741         if (IS_ERR(p->arg.lock_seqid))
5742                 goto out_free_seqid;
5743         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5744         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5745         p->arg.lock_owner.s_dev = server->s_dev;
5746         p->res.lock_seqid = p->arg.lock_seqid;
5747         p->lsp = lsp;
5748         p->server = server;
5749         atomic_inc(&lsp->ls_count);
5750         p->ctx = get_nfs_open_context(ctx);
5751         get_file(fl->fl_file);
5752         memcpy(&p->fl, fl, sizeof(p->fl));
5753         return p;
5754 out_free_seqid:
5755         nfs_free_seqid(p->arg.open_seqid);
5756 out_free:
5757         kfree(p);
5758         return NULL;
5759 }
5760
5761 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5762 {
5763         struct nfs4_lockdata *data = calldata;
5764         struct nfs4_state *state = data->lsp->ls_state;
5765
5766         dprintk("%s: begin!\n", __func__);
5767         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5768                 goto out_wait;
5769         /* Do we need to do an open_to_lock_owner? */
5770         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5771                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5772                         goto out_release_lock_seqid;
5773                 }
5774                 nfs4_stateid_copy(&data->arg.open_stateid,
5775                                 &state->open_stateid);
5776                 data->arg.new_lock_owner = 1;
5777                 data->res.open_seqid = data->arg.open_seqid;
5778         } else {
5779                 data->arg.new_lock_owner = 0;
5780                 nfs4_stateid_copy(&data->arg.lock_stateid,
5781                                 &data->lsp->ls_stateid);
5782         }
5783         if (!nfs4_valid_open_stateid(state)) {
5784                 data->rpc_status = -EBADF;
5785                 task->tk_action = NULL;
5786                 goto out_release_open_seqid;
5787         }
5788         data->timestamp = jiffies;
5789         if (nfs4_setup_sequence(data->server,
5790                                 &data->arg.seq_args,
5791                                 &data->res.seq_res,
5792                                 task) == 0)
5793                 return;
5794 out_release_open_seqid:
5795         nfs_release_seqid(data->arg.open_seqid);
5796 out_release_lock_seqid:
5797         nfs_release_seqid(data->arg.lock_seqid);
5798 out_wait:
5799         nfs4_sequence_done(task, &data->res.seq_res);
5800         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5801 }
5802
5803 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5804 {
5805         struct nfs4_lockdata *data = calldata;
5806         struct nfs4_lock_state *lsp = data->lsp;
5807
5808         dprintk("%s: begin!\n", __func__);
5809
5810         if (!nfs4_sequence_done(task, &data->res.seq_res))
5811                 return;
5812
5813         data->rpc_status = task->tk_status;
5814         switch (task->tk_status) {
5815         case 0:
5816                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5817                                 data->timestamp);
5818                 if (data->arg.new_lock) {
5819                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5820                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5821                                 rpc_restart_call_prepare(task);
5822                                 break;
5823                         }
5824                 }
5825                 if (data->arg.new_lock_owner != 0) {
5826                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5827                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5828                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5829                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5830                         rpc_restart_call_prepare(task);
5831                 break;
5832         case -NFS4ERR_BAD_STATEID:
5833         case -NFS4ERR_OLD_STATEID:
5834         case -NFS4ERR_STALE_STATEID:
5835         case -NFS4ERR_EXPIRED:
5836                 if (data->arg.new_lock_owner != 0) {
5837                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5838                                                 &lsp->ls_state->open_stateid))
5839                                 rpc_restart_call_prepare(task);
5840                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5841                                                 &lsp->ls_stateid))
5842                                 rpc_restart_call_prepare(task);
5843         }
5844         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5845 }
5846
5847 static void nfs4_lock_release(void *calldata)
5848 {
5849         struct nfs4_lockdata *data = calldata;
5850
5851         dprintk("%s: begin!\n", __func__);
5852         nfs_free_seqid(data->arg.open_seqid);
5853         if (data->cancelled != 0) {
5854                 struct rpc_task *task;
5855                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5856                                 data->arg.lock_seqid);
5857                 if (!IS_ERR(task))
5858                         rpc_put_task_async(task);
5859                 dprintk("%s: cancelling lock!\n", __func__);
5860         } else
5861                 nfs_free_seqid(data->arg.lock_seqid);
5862         nfs4_put_lock_state(data->lsp);
5863         put_nfs_open_context(data->ctx);
5864         fput(data->fl.fl_file);
5865         kfree(data);
5866         dprintk("%s: done!\n", __func__);
5867 }
5868
5869 static const struct rpc_call_ops nfs4_lock_ops = {
5870         .rpc_call_prepare = nfs4_lock_prepare,
5871         .rpc_call_done = nfs4_lock_done,
5872         .rpc_release = nfs4_lock_release,
5873 };
5874
5875 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5876 {
5877         switch (error) {
5878         case -NFS4ERR_ADMIN_REVOKED:
5879         case -NFS4ERR_BAD_STATEID:
5880                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5881                 if (new_lock_owner != 0 ||
5882                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5883                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5884                 break;
5885         case -NFS4ERR_STALE_STATEID:
5886                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5887         case -NFS4ERR_EXPIRED:
5888                 nfs4_schedule_lease_recovery(server->nfs_client);
5889         };
5890 }
5891
5892 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5893 {
5894         struct nfs4_lockdata *data;
5895         struct rpc_task *task;
5896         struct rpc_message msg = {
5897                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5898                 .rpc_cred = state->owner->so_cred,
5899         };
5900         struct rpc_task_setup task_setup_data = {
5901                 .rpc_client = NFS_CLIENT(state->inode),
5902                 .rpc_message = &msg,
5903                 .callback_ops = &nfs4_lock_ops,
5904                 .workqueue = nfsiod_workqueue,
5905                 .flags = RPC_TASK_ASYNC,
5906         };
5907         int ret;
5908
5909         dprintk("%s: begin!\n", __func__);
5910         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5911                         fl->fl_u.nfs4_fl.owner,
5912                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5913         if (data == NULL)
5914                 return -ENOMEM;
5915         if (IS_SETLKW(cmd))
5916                 data->arg.block = 1;
5917         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5918         msg.rpc_argp = &data->arg;
5919         msg.rpc_resp = &data->res;
5920         task_setup_data.callback_data = data;
5921         if (recovery_type > NFS_LOCK_NEW) {
5922                 if (recovery_type == NFS_LOCK_RECLAIM)
5923                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5924                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5925         } else
5926                 data->arg.new_lock = 1;
5927         task = rpc_run_task(&task_setup_data);
5928         if (IS_ERR(task))
5929                 return PTR_ERR(task);
5930         ret = nfs4_wait_for_completion_rpc_task(task);
5931         if (ret == 0) {
5932                 ret = data->rpc_status;
5933                 if (ret)
5934                         nfs4_handle_setlk_error(data->server, data->lsp,
5935                                         data->arg.new_lock_owner, ret);
5936         } else
5937                 data->cancelled = 1;
5938         rpc_put_task(task);
5939         dprintk("%s: done, ret = %d!\n", __func__, ret);
5940         return ret;
5941 }
5942
5943 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5944 {
5945         struct nfs_server *server = NFS_SERVER(state->inode);
5946         struct nfs4_exception exception = {
5947                 .inode = state->inode,
5948         };
5949         int err;
5950
5951         do {
5952                 /* Cache the lock if possible... */
5953                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5954                         return 0;
5955                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5956                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5957                 if (err != -NFS4ERR_DELAY)
5958                         break;
5959                 nfs4_handle_exception(server, err, &exception);
5960         } while (exception.retry);
5961         return err;
5962 }
5963
5964 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5965 {
5966         struct nfs_server *server = NFS_SERVER(state->inode);
5967         struct nfs4_exception exception = {
5968                 .inode = state->inode,
5969         };
5970         int err;
5971
5972         err = nfs4_set_lock_state(state, request);
5973         if (err != 0)
5974                 return err;
5975         if (!recover_lost_locks) {
5976                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5977                 return 0;
5978         }
5979         do {
5980                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5981                         return 0;
5982                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5983                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5984                 switch (err) {
5985                 default:
5986                         goto out;
5987                 case -NFS4ERR_GRACE:
5988                 case -NFS4ERR_DELAY:
5989                         nfs4_handle_exception(server, err, &exception);
5990                         err = 0;
5991                 }
5992         } while (exception.retry);
5993 out:
5994         return err;
5995 }
5996
5997 #if defined(CONFIG_NFS_V4_1)
5998 /**
5999  * nfs41_check_expired_locks - possibly free a lock stateid
6000  *
6001  * @state: NFSv4 state for an inode
6002  *
6003  * Returns NFS_OK if recovery for this stateid is now finished.
6004  * Otherwise a negative NFS4ERR value is returned.
6005  */
6006 static int nfs41_check_expired_locks(struct nfs4_state *state)
6007 {
6008         int status, ret = -NFS4ERR_BAD_STATEID;
6009         struct nfs4_lock_state *lsp;
6010         struct nfs_server *server = NFS_SERVER(state->inode);
6011
6012         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6013                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6014                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6015
6016                         status = nfs41_test_stateid(server,
6017                                         &lsp->ls_stateid,
6018                                         cred);
6019                         trace_nfs4_test_lock_stateid(state, lsp, status);
6020                         if (status != NFS_OK) {
6021                                 /* Free the stateid unless the server
6022                                  * informs us the stateid is unrecognized. */
6023                                 if (status != -NFS4ERR_BAD_STATEID)
6024                                         nfs41_free_stateid(server,
6025                                                         &lsp->ls_stateid,
6026                                                         cred);
6027                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6028                                 ret = status;
6029                         }
6030                 }
6031         };
6032
6033         return ret;
6034 }
6035
6036 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6037 {
6038         int status = NFS_OK;
6039
6040         if (test_bit(LK_STATE_IN_USE, &state->flags))
6041                 status = nfs41_check_expired_locks(state);
6042         if (status != NFS_OK)
6043                 status = nfs4_lock_expired(state, request);
6044         return status;
6045 }
6046 #endif
6047
6048 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6049 {
6050         struct nfs_inode *nfsi = NFS_I(state->inode);
6051         unsigned char fl_flags = request->fl_flags;
6052         int status = -ENOLCK;
6053
6054         if ((fl_flags & FL_POSIX) &&
6055                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6056                 goto out;
6057         /* Is this a delegated open? */
6058         status = nfs4_set_lock_state(state, request);
6059         if (status != 0)
6060                 goto out;
6061         request->fl_flags |= FL_ACCESS;
6062         status = do_vfs_lock(state->inode, request);
6063         if (status < 0)
6064                 goto out;
6065         down_read(&nfsi->rwsem);
6066         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6067                 /* Yes: cache locks! */
6068                 /* ...but avoid races with delegation recall... */
6069                 request->fl_flags = fl_flags & ~FL_SLEEP;
6070                 status = do_vfs_lock(state->inode, request);
6071                 up_read(&nfsi->rwsem);
6072                 goto out;
6073         }
6074         up_read(&nfsi->rwsem);
6075         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6076 out:
6077         request->fl_flags = fl_flags;
6078         return status;
6079 }
6080
6081 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6082 {
6083         struct nfs4_exception exception = {
6084                 .state = state,
6085                 .inode = state->inode,
6086         };
6087         int err;
6088
6089         do {
6090                 err = _nfs4_proc_setlk(state, cmd, request);
6091                 trace_nfs4_set_lock(request, state, cmd, err);
6092                 if (err == -NFS4ERR_DENIED)
6093                         err = -EAGAIN;
6094                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6095                                 err, &exception);
6096         } while (exception.retry);
6097         return err;
6098 }
6099
6100 static int
6101 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6102 {
6103         struct nfs_open_context *ctx;
6104         struct nfs4_state *state;
6105         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6106         int status;
6107
6108         /* verify open state */
6109         ctx = nfs_file_open_context(filp);
6110         state = ctx->state;
6111
6112         if (request->fl_start < 0 || request->fl_end < 0)
6113                 return -EINVAL;
6114
6115         if (IS_GETLK(cmd)) {
6116                 if (state != NULL)
6117                         return nfs4_proc_getlk(state, F_GETLK, request);
6118                 return 0;
6119         }
6120
6121         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6122                 return -EINVAL;
6123
6124         if (request->fl_type == F_UNLCK) {
6125                 if (state != NULL)
6126                         return nfs4_proc_unlck(state, cmd, request);
6127                 return 0;
6128         }
6129
6130         if (state == NULL)
6131                 return -ENOLCK;
6132         /*
6133          * Don't rely on the VFS having checked the file open mode,
6134          * since it won't do this for flock() locks.
6135          */
6136         switch (request->fl_type) {
6137         case F_RDLCK:
6138                 if (!(filp->f_mode & FMODE_READ))
6139                         return -EBADF;
6140                 break;
6141         case F_WRLCK:
6142                 if (!(filp->f_mode & FMODE_WRITE))
6143                         return -EBADF;
6144         }
6145
6146         do {
6147                 status = nfs4_proc_setlk(state, cmd, request);
6148                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6149                         break;
6150                 timeout = nfs4_set_lock_task_retry(timeout);
6151                 status = -ERESTARTSYS;
6152                 if (signalled())
6153                         break;
6154         } while(status < 0);
6155         return status;
6156 }
6157
6158 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6159 {
6160         struct nfs_server *server = NFS_SERVER(state->inode);
6161         int err;
6162
6163         err = nfs4_set_lock_state(state, fl);
6164         if (err != 0)
6165                 return err;
6166         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6167         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6168 }
6169
6170 struct nfs_release_lockowner_data {
6171         struct nfs4_lock_state *lsp;
6172         struct nfs_server *server;
6173         struct nfs_release_lockowner_args args;
6174         struct nfs_release_lockowner_res res;
6175         unsigned long timestamp;
6176 };
6177
6178 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6179 {
6180         struct nfs_release_lockowner_data *data = calldata;
6181         struct nfs_server *server = data->server;
6182         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6183                              &data->args.seq_args, &data->res.seq_res, task);
6184         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6185         data->timestamp = jiffies;
6186 }
6187
6188 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6189 {
6190         struct nfs_release_lockowner_data *data = calldata;
6191         struct nfs_server *server = data->server;
6192
6193         nfs40_sequence_done(task, &data->res.seq_res);
6194
6195         switch (task->tk_status) {
6196         case 0:
6197                 renew_lease(server, data->timestamp);
6198                 break;
6199         case -NFS4ERR_STALE_CLIENTID:
6200         case -NFS4ERR_EXPIRED:
6201                 nfs4_schedule_lease_recovery(server->nfs_client);
6202                 break;
6203         case -NFS4ERR_LEASE_MOVED:
6204         case -NFS4ERR_DELAY:
6205                 if (nfs4_async_handle_error(task, server,
6206                                             NULL, NULL) == -EAGAIN)
6207                         rpc_restart_call_prepare(task);
6208         }
6209 }
6210
6211 static void nfs4_release_lockowner_release(void *calldata)
6212 {
6213         struct nfs_release_lockowner_data *data = calldata;
6214         nfs4_free_lock_state(data->server, data->lsp);
6215         kfree(calldata);
6216 }
6217
6218 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6219         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6220         .rpc_call_done = nfs4_release_lockowner_done,
6221         .rpc_release = nfs4_release_lockowner_release,
6222 };
6223
6224 static void
6225 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6226 {
6227         struct nfs_release_lockowner_data *data;
6228         struct rpc_message msg = {
6229                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6230         };
6231
6232         if (server->nfs_client->cl_mvops->minor_version != 0)
6233                 return;
6234
6235         data = kmalloc(sizeof(*data), GFP_NOFS);
6236         if (!data)
6237                 return;
6238         data->lsp = lsp;
6239         data->server = server;
6240         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6241         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6242         data->args.lock_owner.s_dev = server->s_dev;
6243
6244         msg.rpc_argp = &data->args;
6245         msg.rpc_resp = &data->res;
6246         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6247         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6248 }
6249
6250 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6251
6252 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6253                                    const void *buf, size_t buflen,
6254                                    int flags, int type)
6255 {
6256         if (strcmp(key, "") != 0)
6257                 return -EINVAL;
6258
6259         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6260 }
6261
6262 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6263                                    void *buf, size_t buflen, int type)
6264 {
6265         if (strcmp(key, "") != 0)
6266                 return -EINVAL;
6267
6268         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6269 }
6270
6271 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6272                                        size_t list_len, const char *name,
6273                                        size_t name_len, int type)
6274 {
6275         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6276
6277         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6278                 return 0;
6279
6280         if (list && len <= list_len)
6281                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6282         return len;
6283 }
6284
6285 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6286 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6287 {
6288         return server->caps & NFS_CAP_SECURITY_LABEL;
6289 }
6290
6291 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6292                                    const void *buf, size_t buflen,
6293                                    int flags, int type)
6294 {
6295         if (security_ismaclabel(key))
6296                 return nfs4_set_security_label(dentry, buf, buflen);
6297
6298         return -EOPNOTSUPP;
6299 }
6300
6301 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6302                                    void *buf, size_t buflen, int type)
6303 {
6304         if (security_ismaclabel(key))
6305                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6306         return -EOPNOTSUPP;
6307 }
6308
6309 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6310                                        size_t list_len, const char *name,
6311                                        size_t name_len, int type)
6312 {
6313         size_t len = 0;
6314
6315         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6316                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6317                 if (list && len <= list_len)
6318                         security_inode_listsecurity(d_inode(dentry), list, len);
6319         }
6320         return len;
6321 }
6322
6323 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6324         .prefix = XATTR_SECURITY_PREFIX,
6325         .list   = nfs4_xattr_list_nfs4_label,
6326         .get    = nfs4_xattr_get_nfs4_label,
6327         .set    = nfs4_xattr_set_nfs4_label,
6328 };
6329 #endif
6330
6331
6332 /*
6333  * nfs_fhget will use either the mounted_on_fileid or the fileid
6334  */
6335 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6336 {
6337         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6338                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6339               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6340               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6341                 return;
6342
6343         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6344                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6345         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6346         fattr->nlink = 2;
6347 }
6348
6349 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6350                                    const struct qstr *name,
6351                                    struct nfs4_fs_locations *fs_locations,
6352                                    struct page *page)
6353 {
6354         struct nfs_server *server = NFS_SERVER(dir);
6355         u32 bitmask[3] = {
6356                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6357         };
6358         struct nfs4_fs_locations_arg args = {
6359                 .dir_fh = NFS_FH(dir),
6360                 .name = name,
6361                 .page = page,
6362                 .bitmask = bitmask,
6363         };
6364         struct nfs4_fs_locations_res res = {
6365                 .fs_locations = fs_locations,
6366         };
6367         struct rpc_message msg = {
6368                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6369                 .rpc_argp = &args,
6370                 .rpc_resp = &res,
6371         };
6372         int status;
6373
6374         dprintk("%s: start\n", __func__);
6375
6376         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6377          * is not supported */
6378         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6379                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6380         else
6381                 bitmask[0] |= FATTR4_WORD0_FILEID;
6382
6383         nfs_fattr_init(&fs_locations->fattr);
6384         fs_locations->server = server;
6385         fs_locations->nlocations = 0;
6386         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6387         dprintk("%s: returned status = %d\n", __func__, status);
6388         return status;
6389 }
6390
6391 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6392                            const struct qstr *name,
6393                            struct nfs4_fs_locations *fs_locations,
6394                            struct page *page)
6395 {
6396         struct nfs4_exception exception = { };
6397         int err;
6398         do {
6399                 err = _nfs4_proc_fs_locations(client, dir, name,
6400                                 fs_locations, page);
6401                 trace_nfs4_get_fs_locations(dir, name, err);
6402                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6403                                 &exception);
6404         } while (exception.retry);
6405         return err;
6406 }
6407
6408 /*
6409  * This operation also signals the server that this client is
6410  * performing migration recovery.  The server can stop returning
6411  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6412  * appended to this compound to identify the client ID which is
6413  * performing recovery.
6414  */
6415 static int _nfs40_proc_get_locations(struct inode *inode,
6416                                      struct nfs4_fs_locations *locations,
6417                                      struct page *page, struct rpc_cred *cred)
6418 {
6419         struct nfs_server *server = NFS_SERVER(inode);
6420         struct rpc_clnt *clnt = server->client;
6421         u32 bitmask[2] = {
6422                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6423         };
6424         struct nfs4_fs_locations_arg args = {
6425                 .clientid       = server->nfs_client->cl_clientid,
6426                 .fh             = NFS_FH(inode),
6427                 .page           = page,
6428                 .bitmask        = bitmask,
6429                 .migration      = 1,            /* skip LOOKUP */
6430                 .renew          = 1,            /* append RENEW */
6431         };
6432         struct nfs4_fs_locations_res res = {
6433                 .fs_locations   = locations,
6434                 .migration      = 1,
6435                 .renew          = 1,
6436         };
6437         struct rpc_message msg = {
6438                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6439                 .rpc_argp       = &args,
6440                 .rpc_resp       = &res,
6441                 .rpc_cred       = cred,
6442         };
6443         unsigned long now = jiffies;
6444         int status;
6445
6446         nfs_fattr_init(&locations->fattr);
6447         locations->server = server;
6448         locations->nlocations = 0;
6449
6450         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6451         nfs4_set_sequence_privileged(&args.seq_args);
6452         status = nfs4_call_sync_sequence(clnt, server, &msg,
6453                                         &args.seq_args, &res.seq_res);
6454         if (status)
6455                 return status;
6456
6457         renew_lease(server, now);
6458         return 0;
6459 }
6460
6461 #ifdef CONFIG_NFS_V4_1
6462
6463 /*
6464  * This operation also signals the server that this client is
6465  * performing migration recovery.  The server can stop asserting
6466  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6467  * performing this operation is identified in the SEQUENCE
6468  * operation in this compound.
6469  *
6470  * When the client supports GETATTR(fs_locations_info), it can
6471  * be plumbed in here.
6472  */
6473 static int _nfs41_proc_get_locations(struct inode *inode,
6474                                      struct nfs4_fs_locations *locations,
6475                                      struct page *page, struct rpc_cred *cred)
6476 {
6477         struct nfs_server *server = NFS_SERVER(inode);
6478         struct rpc_clnt *clnt = server->client;
6479         u32 bitmask[2] = {
6480                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6481         };
6482         struct nfs4_fs_locations_arg args = {
6483                 .fh             = NFS_FH(inode),
6484                 .page           = page,
6485                 .bitmask        = bitmask,
6486                 .migration      = 1,            /* skip LOOKUP */
6487         };
6488         struct nfs4_fs_locations_res res = {
6489                 .fs_locations   = locations,
6490                 .migration      = 1,
6491         };
6492         struct rpc_message msg = {
6493                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6494                 .rpc_argp       = &args,
6495                 .rpc_resp       = &res,
6496                 .rpc_cred       = cred,
6497         };
6498         int status;
6499
6500         nfs_fattr_init(&locations->fattr);
6501         locations->server = server;
6502         locations->nlocations = 0;
6503
6504         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6505         nfs4_set_sequence_privileged(&args.seq_args);
6506         status = nfs4_call_sync_sequence(clnt, server, &msg,
6507                                         &args.seq_args, &res.seq_res);
6508         if (status == NFS4_OK &&
6509             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6510                 status = -NFS4ERR_LEASE_MOVED;
6511         return status;
6512 }
6513
6514 #endif  /* CONFIG_NFS_V4_1 */
6515
6516 /**
6517  * nfs4_proc_get_locations - discover locations for a migrated FSID
6518  * @inode: inode on FSID that is migrating
6519  * @locations: result of query
6520  * @page: buffer
6521  * @cred: credential to use for this operation
6522  *
6523  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6524  * operation failed, or a negative errno if a local error occurred.
6525  *
6526  * On success, "locations" is filled in, but if the server has
6527  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6528  * asserted.
6529  *
6530  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6531  * from this client that require migration recovery.
6532  */
6533 int nfs4_proc_get_locations(struct inode *inode,
6534                             struct nfs4_fs_locations *locations,
6535                             struct page *page, struct rpc_cred *cred)
6536 {
6537         struct nfs_server *server = NFS_SERVER(inode);
6538         struct nfs_client *clp = server->nfs_client;
6539         const struct nfs4_mig_recovery_ops *ops =
6540                                         clp->cl_mvops->mig_recovery_ops;
6541         struct nfs4_exception exception = { };
6542         int status;
6543
6544         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6545                 (unsigned long long)server->fsid.major,
6546                 (unsigned long long)server->fsid.minor,
6547                 clp->cl_hostname);
6548         nfs_display_fhandle(NFS_FH(inode), __func__);
6549
6550         do {
6551                 status = ops->get_locations(inode, locations, page, cred);
6552                 if (status != -NFS4ERR_DELAY)
6553                         break;
6554                 nfs4_handle_exception(server, status, &exception);
6555         } while (exception.retry);
6556         return status;
6557 }
6558
6559 /*
6560  * This operation also signals the server that this client is
6561  * performing "lease moved" recovery.  The server can stop
6562  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6563  * is appended to this compound to identify the client ID which is
6564  * performing recovery.
6565  */
6566 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6567 {
6568         struct nfs_server *server = NFS_SERVER(inode);
6569         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6570         struct rpc_clnt *clnt = server->client;
6571         struct nfs4_fsid_present_arg args = {
6572                 .fh             = NFS_FH(inode),
6573                 .clientid       = clp->cl_clientid,
6574                 .renew          = 1,            /* append RENEW */
6575         };
6576         struct nfs4_fsid_present_res res = {
6577                 .renew          = 1,
6578         };
6579         struct rpc_message msg = {
6580                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6581                 .rpc_argp       = &args,
6582                 .rpc_resp       = &res,
6583                 .rpc_cred       = cred,
6584         };
6585         unsigned long now = jiffies;
6586         int status;
6587
6588         res.fh = nfs_alloc_fhandle();
6589         if (res.fh == NULL)
6590                 return -ENOMEM;
6591
6592         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6593         nfs4_set_sequence_privileged(&args.seq_args);
6594         status = nfs4_call_sync_sequence(clnt, server, &msg,
6595                                                 &args.seq_args, &res.seq_res);
6596         nfs_free_fhandle(res.fh);
6597         if (status)
6598                 return status;
6599
6600         do_renew_lease(clp, now);
6601         return 0;
6602 }
6603
6604 #ifdef CONFIG_NFS_V4_1
6605
6606 /*
6607  * This operation also signals the server that this client is
6608  * performing "lease moved" recovery.  The server can stop asserting
6609  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6610  * this operation is identified in the SEQUENCE operation in this
6611  * compound.
6612  */
6613 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6614 {
6615         struct nfs_server *server = NFS_SERVER(inode);
6616         struct rpc_clnt *clnt = server->client;
6617         struct nfs4_fsid_present_arg args = {
6618                 .fh             = NFS_FH(inode),
6619         };
6620         struct nfs4_fsid_present_res res = {
6621         };
6622         struct rpc_message msg = {
6623                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6624                 .rpc_argp       = &args,
6625                 .rpc_resp       = &res,
6626                 .rpc_cred       = cred,
6627         };
6628         int status;
6629
6630         res.fh = nfs_alloc_fhandle();
6631         if (res.fh == NULL)
6632                 return -ENOMEM;
6633
6634         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6635         nfs4_set_sequence_privileged(&args.seq_args);
6636         status = nfs4_call_sync_sequence(clnt, server, &msg,
6637                                                 &args.seq_args, &res.seq_res);
6638         nfs_free_fhandle(res.fh);
6639         if (status == NFS4_OK &&
6640             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6641                 status = -NFS4ERR_LEASE_MOVED;
6642         return status;
6643 }
6644
6645 #endif  /* CONFIG_NFS_V4_1 */
6646
6647 /**
6648  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6649  * @inode: inode on FSID to check
6650  * @cred: credential to use for this operation
6651  *
6652  * Server indicates whether the FSID is present, moved, or not
6653  * recognized.  This operation is necessary to clear a LEASE_MOVED
6654  * condition for this client ID.
6655  *
6656  * Returns NFS4_OK if the FSID is present on this server,
6657  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6658  *  NFS4ERR code if some error occurred on the server, or a
6659  *  negative errno if a local failure occurred.
6660  */
6661 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6662 {
6663         struct nfs_server *server = NFS_SERVER(inode);
6664         struct nfs_client *clp = server->nfs_client;
6665         const struct nfs4_mig_recovery_ops *ops =
6666                                         clp->cl_mvops->mig_recovery_ops;
6667         struct nfs4_exception exception = { };
6668         int status;
6669
6670         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6671                 (unsigned long long)server->fsid.major,
6672                 (unsigned long long)server->fsid.minor,
6673                 clp->cl_hostname);
6674         nfs_display_fhandle(NFS_FH(inode), __func__);
6675
6676         do {
6677                 status = ops->fsid_present(inode, cred);
6678                 if (status != -NFS4ERR_DELAY)
6679                         break;
6680                 nfs4_handle_exception(server, status, &exception);
6681         } while (exception.retry);
6682         return status;
6683 }
6684
6685 /**
6686  * If 'use_integrity' is true and the state managment nfs_client
6687  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6688  * and the machine credential as per RFC3530bis and RFC5661 Security
6689  * Considerations sections. Otherwise, just use the user cred with the
6690  * filesystem's rpc_client.
6691  */
6692 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6693 {
6694         int status;
6695         struct nfs4_secinfo_arg args = {
6696                 .dir_fh = NFS_FH(dir),
6697                 .name   = name,
6698         };
6699         struct nfs4_secinfo_res res = {
6700                 .flavors     = flavors,
6701         };
6702         struct rpc_message msg = {
6703                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6704                 .rpc_argp = &args,
6705                 .rpc_resp = &res,
6706         };
6707         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6708         struct rpc_cred *cred = NULL;
6709
6710         if (use_integrity) {
6711                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6712                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6713                 msg.rpc_cred = cred;
6714         }
6715
6716         dprintk("NFS call  secinfo %s\n", name->name);
6717
6718         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6719                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6720
6721         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6722                                 &res.seq_res, 0);
6723         dprintk("NFS reply  secinfo: %d\n", status);
6724
6725         if (cred)
6726                 put_rpccred(cred);
6727
6728         return status;
6729 }
6730
6731 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6732                       struct nfs4_secinfo_flavors *flavors)
6733 {
6734         struct nfs4_exception exception = { };
6735         int err;
6736         do {
6737                 err = -NFS4ERR_WRONGSEC;
6738
6739                 /* try to use integrity protection with machine cred */
6740                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6741                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6742
6743                 /*
6744                  * if unable to use integrity protection, or SECINFO with
6745                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6746                  * disallowed by spec, but exists in deployed servers) use
6747                  * the current filesystem's rpc_client and the user cred.
6748                  */
6749                 if (err == -NFS4ERR_WRONGSEC)
6750                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6751
6752                 trace_nfs4_secinfo(dir, name, err);
6753                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6754                                 &exception);
6755         } while (exception.retry);
6756         return err;
6757 }
6758
6759 #ifdef CONFIG_NFS_V4_1
6760 /*
6761  * Check the exchange flags returned by the server for invalid flags, having
6762  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6763  * DS flags set.
6764  */
6765 static int nfs4_check_cl_exchange_flags(u32 flags)
6766 {
6767         if (flags & ~EXCHGID4_FLAG_MASK_R)
6768                 goto out_inval;
6769         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6770             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6771                 goto out_inval;
6772         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6773                 goto out_inval;
6774         return NFS_OK;
6775 out_inval:
6776         return -NFS4ERR_INVAL;
6777 }
6778
6779 static bool
6780 nfs41_same_server_scope(struct nfs41_server_scope *a,
6781                         struct nfs41_server_scope *b)
6782 {
6783         if (a->server_scope_sz == b->server_scope_sz &&
6784             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6785                 return true;
6786
6787         return false;
6788 }
6789
6790 /*
6791  * nfs4_proc_bind_conn_to_session()
6792  *
6793  * The 4.1 client currently uses the same TCP connection for the
6794  * fore and backchannel.
6795  */
6796 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6797 {
6798         int status;
6799         struct nfs41_bind_conn_to_session_args args = {
6800                 .client = clp,
6801                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6802         };
6803         struct nfs41_bind_conn_to_session_res res;
6804         struct rpc_message msg = {
6805                 .rpc_proc =
6806                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6807                 .rpc_argp = &args,
6808                 .rpc_resp = &res,
6809                 .rpc_cred = cred,
6810         };
6811
6812         dprintk("--> %s\n", __func__);
6813
6814         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6815         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6816                 args.dir = NFS4_CDFC4_FORE;
6817
6818         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6819         trace_nfs4_bind_conn_to_session(clp, status);
6820         if (status == 0) {
6821                 if (memcmp(res.sessionid.data,
6822                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6823                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6824                         status = -EIO;
6825                         goto out;
6826                 }
6827                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6828                         dprintk("NFS: %s: Unexpected direction from server\n",
6829                                 __func__);
6830                         status = -EIO;
6831                         goto out;
6832                 }
6833                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6834                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6835                                 __func__);
6836                         status = -EIO;
6837                         goto out;
6838                 }
6839         }
6840 out:
6841         dprintk("<-- %s status= %d\n", __func__, status);
6842         return status;
6843 }
6844
6845 /*
6846  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6847  * and operations we'd like to see to enable certain features in the allow map
6848  */
6849 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6850         .how = SP4_MACH_CRED,
6851         .enforce.u.words = {
6852                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6853                       1 << (OP_EXCHANGE_ID - 32) |
6854                       1 << (OP_CREATE_SESSION - 32) |
6855                       1 << (OP_DESTROY_SESSION - 32) |
6856                       1 << (OP_DESTROY_CLIENTID - 32)
6857         },
6858         .allow.u.words = {
6859                 [0] = 1 << (OP_CLOSE) |
6860                       1 << (OP_LOCKU) |
6861                       1 << (OP_COMMIT),
6862                 [1] = 1 << (OP_SECINFO - 32) |
6863                       1 << (OP_SECINFO_NO_NAME - 32) |
6864                       1 << (OP_TEST_STATEID - 32) |
6865                       1 << (OP_FREE_STATEID - 32) |
6866                       1 << (OP_WRITE - 32)
6867         }
6868 };
6869
6870 /*
6871  * Select the state protection mode for client `clp' given the server results
6872  * from exchange_id in `sp'.
6873  *
6874  * Returns 0 on success, negative errno otherwise.
6875  */
6876 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6877                                  struct nfs41_state_protection *sp)
6878 {
6879         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6880                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6881                       1 << (OP_EXCHANGE_ID - 32) |
6882                       1 << (OP_CREATE_SESSION - 32) |
6883                       1 << (OP_DESTROY_SESSION - 32) |
6884                       1 << (OP_DESTROY_CLIENTID - 32)
6885         };
6886         unsigned int i;
6887
6888         if (sp->how == SP4_MACH_CRED) {
6889                 /* Print state protect result */
6890                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6891                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6892                         if (test_bit(i, sp->enforce.u.longs))
6893                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6894                         if (test_bit(i, sp->allow.u.longs))
6895                                 dfprintk(MOUNT, "  allow op %d\n", i);
6896                 }
6897
6898                 /* make sure nothing is on enforce list that isn't supported */
6899                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6900                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6901                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6902                                 return -EINVAL;
6903                         }
6904                 }
6905
6906                 /*
6907                  * Minimal mode - state operations are allowed to use machine
6908                  * credential.  Note this already happens by default, so the
6909                  * client doesn't have to do anything more than the negotiation.
6910                  *
6911                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6912                  *       we're already using the machine cred for exchange_id
6913                  *       and will never use a different cred.
6914                  */
6915                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6916                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6917                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6918                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6919                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6920                         dfprintk(MOUNT, "  minimal mode enabled\n");
6921                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6922                 } else {
6923                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6924                         return -EINVAL;
6925                 }
6926
6927                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6928                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6929                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6930                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6931                 }
6932
6933                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6934                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6935                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6936                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6937                 }
6938
6939                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6940                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6941                         dfprintk(MOUNT, "  stateid mode enabled\n");
6942                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6943                 }
6944
6945                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6946                         dfprintk(MOUNT, "  write mode enabled\n");
6947                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6948                 }
6949
6950                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6951                         dfprintk(MOUNT, "  commit mode enabled\n");
6952                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6953                 }
6954         }
6955
6956         return 0;
6957 }
6958
6959 /*
6960  * _nfs4_proc_exchange_id()
6961  *
6962  * Wrapper for EXCHANGE_ID operation.
6963  */
6964 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6965         u32 sp4_how)
6966 {
6967         nfs4_verifier verifier;
6968         struct nfs41_exchange_id_args args = {
6969                 .verifier = &verifier,
6970                 .client = clp,
6971 #ifdef CONFIG_NFS_V4_1_MIGRATION
6972                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6973                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6974                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6975 #else
6976                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6977                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6978 #endif
6979         };
6980         struct nfs41_exchange_id_res res = {
6981                 0
6982         };
6983         int status;
6984         struct rpc_message msg = {
6985                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6986                 .rpc_argp = &args,
6987                 .rpc_resp = &res,
6988                 .rpc_cred = cred,
6989         };
6990
6991         nfs4_init_boot_verifier(clp, &verifier);
6992
6993         status = nfs4_init_uniform_client_string(clp);
6994         if (status)
6995                 goto out;
6996
6997         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
6998                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6999                 clp->cl_owner_id);
7000
7001         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7002                                         GFP_NOFS);
7003         if (unlikely(res.server_owner == NULL)) {
7004                 status = -ENOMEM;
7005                 goto out;
7006         }
7007
7008         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7009                                         GFP_NOFS);
7010         if (unlikely(res.server_scope == NULL)) {
7011                 status = -ENOMEM;
7012                 goto out_server_owner;
7013         }
7014
7015         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7016         if (unlikely(res.impl_id == NULL)) {
7017                 status = -ENOMEM;
7018                 goto out_server_scope;
7019         }
7020
7021         switch (sp4_how) {
7022         case SP4_NONE:
7023                 args.state_protect.how = SP4_NONE;
7024                 break;
7025
7026         case SP4_MACH_CRED:
7027                 args.state_protect = nfs4_sp4_mach_cred_request;
7028                 break;
7029
7030         default:
7031                 /* unsupported! */
7032                 WARN_ON_ONCE(1);
7033                 status = -EINVAL;
7034                 goto out_impl_id;
7035         }
7036
7037         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7038         trace_nfs4_exchange_id(clp, status);
7039         if (status == 0)
7040                 status = nfs4_check_cl_exchange_flags(res.flags);
7041
7042         if (status == 0)
7043                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7044
7045         if (status == 0) {
7046                 clp->cl_clientid = res.clientid;
7047                 clp->cl_exchange_flags = res.flags;
7048                 /* Client ID is not confirmed */
7049                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7050                         clear_bit(NFS4_SESSION_ESTABLISHED,
7051                                         &clp->cl_session->session_state);
7052                         clp->cl_seqid = res.seqid;
7053                 }
7054
7055                 kfree(clp->cl_serverowner);
7056                 clp->cl_serverowner = res.server_owner;
7057                 res.server_owner = NULL;
7058
7059                 /* use the most recent implementation id */
7060                 kfree(clp->cl_implid);
7061                 clp->cl_implid = res.impl_id;
7062                 res.impl_id = NULL;
7063
7064                 if (clp->cl_serverscope != NULL &&
7065                     !nfs41_same_server_scope(clp->cl_serverscope,
7066                                              res.server_scope)) {
7067                         dprintk("%s: server_scope mismatch detected\n",
7068                                 __func__);
7069                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7070                         kfree(clp->cl_serverscope);
7071                         clp->cl_serverscope = NULL;
7072                 }
7073
7074                 if (clp->cl_serverscope == NULL) {
7075                         clp->cl_serverscope = res.server_scope;
7076                         res.server_scope = NULL;
7077                 }
7078         }
7079
7080 out_impl_id:
7081         kfree(res.impl_id);
7082 out_server_scope:
7083         kfree(res.server_scope);
7084 out_server_owner:
7085         kfree(res.server_owner);
7086 out:
7087         if (clp->cl_implid != NULL)
7088                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7089                         "domain: %s, name: %s, date: %llu,%u\n",
7090                         clp->cl_implid->domain, clp->cl_implid->name,
7091                         clp->cl_implid->date.seconds,
7092                         clp->cl_implid->date.nseconds);
7093         dprintk("NFS reply exchange_id: %d\n", status);
7094         return status;
7095 }
7096
7097 /*
7098  * nfs4_proc_exchange_id()
7099  *
7100  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7101  *
7102  * Since the clientid has expired, all compounds using sessions
7103  * associated with the stale clientid will be returning
7104  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7105  * be in some phase of session reset.
7106  *
7107  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7108  */
7109 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7110 {
7111         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7112         int status;
7113
7114         /* try SP4_MACH_CRED if krb5i/p */
7115         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7116             authflavor == RPC_AUTH_GSS_KRB5P) {
7117                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7118                 if (!status)
7119                         return 0;
7120         }
7121
7122         /* try SP4_NONE */
7123         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7124 }
7125
7126 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7127                 struct rpc_cred *cred)
7128 {
7129         struct rpc_message msg = {
7130                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7131                 .rpc_argp = clp,
7132                 .rpc_cred = cred,
7133         };
7134         int status;
7135
7136         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7137         trace_nfs4_destroy_clientid(clp, status);
7138         if (status)
7139                 dprintk("NFS: Got error %d from the server %s on "
7140                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7141         return status;
7142 }
7143
7144 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7145                 struct rpc_cred *cred)
7146 {
7147         unsigned int loop;
7148         int ret;
7149
7150         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7151                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7152                 switch (ret) {
7153                 case -NFS4ERR_DELAY:
7154                 case -NFS4ERR_CLIENTID_BUSY:
7155                         ssleep(1);
7156                         break;
7157                 default:
7158                         return ret;
7159                 }
7160         }
7161         return 0;
7162 }
7163
7164 int nfs4_destroy_clientid(struct nfs_client *clp)
7165 {
7166         struct rpc_cred *cred;
7167         int ret = 0;
7168
7169         if (clp->cl_mvops->minor_version < 1)
7170                 goto out;
7171         if (clp->cl_exchange_flags == 0)
7172                 goto out;
7173         if (clp->cl_preserve_clid)
7174                 goto out;
7175         cred = nfs4_get_clid_cred(clp);
7176         ret = nfs4_proc_destroy_clientid(clp, cred);
7177         if (cred)
7178                 put_rpccred(cred);
7179         switch (ret) {
7180         case 0:
7181         case -NFS4ERR_STALE_CLIENTID:
7182                 clp->cl_exchange_flags = 0;
7183         }
7184 out:
7185         return ret;
7186 }
7187
7188 struct nfs4_get_lease_time_data {
7189         struct nfs4_get_lease_time_args *args;
7190         struct nfs4_get_lease_time_res *res;
7191         struct nfs_client *clp;
7192 };
7193
7194 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7195                                         void *calldata)
7196 {
7197         struct nfs4_get_lease_time_data *data =
7198                         (struct nfs4_get_lease_time_data *)calldata;
7199
7200         dprintk("--> %s\n", __func__);
7201         /* just setup sequence, do not trigger session recovery
7202            since we're invoked within one */
7203         nfs41_setup_sequence(data->clp->cl_session,
7204                         &data->args->la_seq_args,
7205                         &data->res->lr_seq_res,
7206                         task);
7207         dprintk("<-- %s\n", __func__);
7208 }
7209
7210 /*
7211  * Called from nfs4_state_manager thread for session setup, so don't recover
7212  * from sequence operation or clientid errors.
7213  */
7214 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7215 {
7216         struct nfs4_get_lease_time_data *data =
7217                         (struct nfs4_get_lease_time_data *)calldata;
7218
7219         dprintk("--> %s\n", __func__);
7220         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7221                 return;
7222         switch (task->tk_status) {
7223         case -NFS4ERR_DELAY:
7224         case -NFS4ERR_GRACE:
7225                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7226                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7227                 task->tk_status = 0;
7228                 /* fall through */
7229         case -NFS4ERR_RETRY_UNCACHED_REP:
7230                 rpc_restart_call_prepare(task);
7231                 return;
7232         }
7233         dprintk("<-- %s\n", __func__);
7234 }
7235
7236 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7237         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7238         .rpc_call_done = nfs4_get_lease_time_done,
7239 };
7240
7241 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7242 {
7243         struct rpc_task *task;
7244         struct nfs4_get_lease_time_args args;
7245         struct nfs4_get_lease_time_res res = {
7246                 .lr_fsinfo = fsinfo,
7247         };
7248         struct nfs4_get_lease_time_data data = {
7249                 .args = &args,
7250                 .res = &res,
7251                 .clp = clp,
7252         };
7253         struct rpc_message msg = {
7254                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7255                 .rpc_argp = &args,
7256                 .rpc_resp = &res,
7257         };
7258         struct rpc_task_setup task_setup = {
7259                 .rpc_client = clp->cl_rpcclient,
7260                 .rpc_message = &msg,
7261                 .callback_ops = &nfs4_get_lease_time_ops,
7262                 .callback_data = &data,
7263                 .flags = RPC_TASK_TIMEOUT,
7264         };
7265         int status;
7266
7267         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7268         nfs4_set_sequence_privileged(&args.la_seq_args);
7269         dprintk("--> %s\n", __func__);
7270         task = rpc_run_task(&task_setup);
7271
7272         if (IS_ERR(task))
7273                 status = PTR_ERR(task);
7274         else {
7275                 status = task->tk_status;
7276                 rpc_put_task(task);
7277         }
7278         dprintk("<-- %s return %d\n", __func__, status);
7279
7280         return status;
7281 }
7282
7283 /*
7284  * Initialize the values to be used by the client in CREATE_SESSION
7285  * If nfs4_init_session set the fore channel request and response sizes,
7286  * use them.
7287  *
7288  * Set the back channel max_resp_sz_cached to zero to force the client to
7289  * always set csa_cachethis to FALSE because the current implementation
7290  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7291  */
7292 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7293 {
7294         unsigned int max_rqst_sz, max_resp_sz;
7295
7296         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7297         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7298
7299         /* Fore channel attributes */
7300         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7301         args->fc_attrs.max_resp_sz = max_resp_sz;
7302         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7303         args->fc_attrs.max_reqs = max_session_slots;
7304
7305         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7306                 "max_ops=%u max_reqs=%u\n",
7307                 __func__,
7308                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7309                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7310
7311         /* Back channel attributes */
7312         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7313         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7314         args->bc_attrs.max_resp_sz_cached = 0;
7315         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7316         args->bc_attrs.max_reqs = 1;
7317
7318         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7319                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7320                 __func__,
7321                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7322                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7323                 args->bc_attrs.max_reqs);
7324 }
7325
7326 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7327                 struct nfs41_create_session_res *res)
7328 {
7329         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7330         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7331
7332         if (rcvd->max_resp_sz > sent->max_resp_sz)
7333                 return -EINVAL;
7334         /*
7335          * Our requested max_ops is the minimum we need; we're not
7336          * prepared to break up compounds into smaller pieces than that.
7337          * So, no point even trying to continue if the server won't
7338          * cooperate:
7339          */
7340         if (rcvd->max_ops < sent->max_ops)
7341                 return -EINVAL;
7342         if (rcvd->max_reqs == 0)
7343                 return -EINVAL;
7344         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7345                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7346         return 0;
7347 }
7348
7349 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7350                 struct nfs41_create_session_res *res)
7351 {
7352         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7353         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7354
7355         if (!(res->flags & SESSION4_BACK_CHAN))
7356                 goto out;
7357         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7358                 return -EINVAL;
7359         if (rcvd->max_resp_sz < sent->max_resp_sz)
7360                 return -EINVAL;
7361         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7362                 return -EINVAL;
7363         /* These would render the backchannel useless: */
7364         if (rcvd->max_ops != sent->max_ops)
7365                 return -EINVAL;
7366         if (rcvd->max_reqs != sent->max_reqs)
7367                 return -EINVAL;
7368 out:
7369         return 0;
7370 }
7371
7372 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7373                                      struct nfs41_create_session_res *res)
7374 {
7375         int ret;
7376
7377         ret = nfs4_verify_fore_channel_attrs(args, res);
7378         if (ret)
7379                 return ret;
7380         return nfs4_verify_back_channel_attrs(args, res);
7381 }
7382
7383 static void nfs4_update_session(struct nfs4_session *session,
7384                 struct nfs41_create_session_res *res)
7385 {
7386         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7387         /* Mark client id and session as being confirmed */
7388         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7389         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7390         session->flags = res->flags;
7391         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7392         if (res->flags & SESSION4_BACK_CHAN)
7393                 memcpy(&session->bc_attrs, &res->bc_attrs,
7394                                 sizeof(session->bc_attrs));
7395 }
7396
7397 static int _nfs4_proc_create_session(struct nfs_client *clp,
7398                 struct rpc_cred *cred)
7399 {
7400         struct nfs4_session *session = clp->cl_session;
7401         struct nfs41_create_session_args args = {
7402                 .client = clp,
7403                 .clientid = clp->cl_clientid,
7404                 .seqid = clp->cl_seqid,
7405                 .cb_program = NFS4_CALLBACK,
7406         };
7407         struct nfs41_create_session_res res;
7408
7409         struct rpc_message msg = {
7410                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7411                 .rpc_argp = &args,
7412                 .rpc_resp = &res,
7413                 .rpc_cred = cred,
7414         };
7415         int status;
7416
7417         nfs4_init_channel_attrs(&args);
7418         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7419
7420         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7421         trace_nfs4_create_session(clp, status);
7422
7423         if (!status) {
7424                 /* Verify the session's negotiated channel_attrs values */
7425                 status = nfs4_verify_channel_attrs(&args, &res);
7426                 /* Increment the clientid slot sequence id */
7427                 if (clp->cl_seqid == res.seqid)
7428                         clp->cl_seqid++;
7429                 if (status)
7430                         goto out;
7431                 nfs4_update_session(session, &res);
7432         }
7433 out:
7434         return status;
7435 }
7436
7437 /*
7438  * Issues a CREATE_SESSION operation to the server.
7439  * It is the responsibility of the caller to verify the session is
7440  * expired before calling this routine.
7441  */
7442 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7443 {
7444         int status;
7445         unsigned *ptr;
7446         struct nfs4_session *session = clp->cl_session;
7447
7448         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7449
7450         status = _nfs4_proc_create_session(clp, cred);
7451         if (status)
7452                 goto out;
7453
7454         /* Init or reset the session slot tables */
7455         status = nfs4_setup_session_slot_tables(session);
7456         dprintk("slot table setup returned %d\n", status);
7457         if (status)
7458                 goto out;
7459
7460         ptr = (unsigned *)&session->sess_id.data[0];
7461         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7462                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7463 out:
7464         dprintk("<-- %s\n", __func__);
7465         return status;
7466 }
7467
7468 /*
7469  * Issue the over-the-wire RPC DESTROY_SESSION.
7470  * The caller must serialize access to this routine.
7471  */
7472 int nfs4_proc_destroy_session(struct nfs4_session *session,
7473                 struct rpc_cred *cred)
7474 {
7475         struct rpc_message msg = {
7476                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7477                 .rpc_argp = session,
7478                 .rpc_cred = cred,
7479         };
7480         int status = 0;
7481
7482         dprintk("--> nfs4_proc_destroy_session\n");
7483
7484         /* session is still being setup */
7485         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7486                 return 0;
7487
7488         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7489         trace_nfs4_destroy_session(session->clp, status);
7490
7491         if (status)
7492                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7493                         "Session has been destroyed regardless...\n", status);
7494
7495         dprintk("<-- nfs4_proc_destroy_session\n");
7496         return status;
7497 }
7498
7499 /*
7500  * Renew the cl_session lease.
7501  */
7502 struct nfs4_sequence_data {
7503         struct nfs_client *clp;
7504         struct nfs4_sequence_args args;
7505         struct nfs4_sequence_res res;
7506 };
7507
7508 static void nfs41_sequence_release(void *data)
7509 {
7510         struct nfs4_sequence_data *calldata = data;
7511         struct nfs_client *clp = calldata->clp;
7512
7513         if (atomic_read(&clp->cl_count) > 1)
7514                 nfs4_schedule_state_renewal(clp);
7515         nfs_put_client(clp);
7516         kfree(calldata);
7517 }
7518
7519 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7520 {
7521         switch(task->tk_status) {
7522         case -NFS4ERR_DELAY:
7523                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7524                 return -EAGAIN;
7525         default:
7526                 nfs4_schedule_lease_recovery(clp);
7527         }
7528         return 0;
7529 }
7530
7531 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7532 {
7533         struct nfs4_sequence_data *calldata = data;
7534         struct nfs_client *clp = calldata->clp;
7535
7536         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7537                 return;
7538
7539         trace_nfs4_sequence(clp, task->tk_status);
7540         if (task->tk_status < 0) {
7541                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7542                 if (atomic_read(&clp->cl_count) == 1)
7543                         goto out;
7544
7545                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7546                         rpc_restart_call_prepare(task);
7547                         return;
7548                 }
7549         }
7550         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7551 out:
7552         dprintk("<-- %s\n", __func__);
7553 }
7554
7555 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7556 {
7557         struct nfs4_sequence_data *calldata = data;
7558         struct nfs_client *clp = calldata->clp;
7559         struct nfs4_sequence_args *args;
7560         struct nfs4_sequence_res *res;
7561
7562         args = task->tk_msg.rpc_argp;
7563         res = task->tk_msg.rpc_resp;
7564
7565         nfs41_setup_sequence(clp->cl_session, args, res, task);
7566 }
7567
7568 static const struct rpc_call_ops nfs41_sequence_ops = {
7569         .rpc_call_done = nfs41_sequence_call_done,
7570         .rpc_call_prepare = nfs41_sequence_prepare,
7571         .rpc_release = nfs41_sequence_release,
7572 };
7573
7574 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7575                 struct rpc_cred *cred,
7576                 bool is_privileged)
7577 {
7578         struct nfs4_sequence_data *calldata;
7579         struct rpc_message msg = {
7580                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7581                 .rpc_cred = cred,
7582         };
7583         struct rpc_task_setup task_setup_data = {
7584                 .rpc_client = clp->cl_rpcclient,
7585                 .rpc_message = &msg,
7586                 .callback_ops = &nfs41_sequence_ops,
7587                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7588         };
7589
7590         if (!atomic_inc_not_zero(&clp->cl_count))
7591                 return ERR_PTR(-EIO);
7592         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7593         if (calldata == NULL) {
7594                 nfs_put_client(clp);
7595                 return ERR_PTR(-ENOMEM);
7596         }
7597         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7598         if (is_privileged)
7599                 nfs4_set_sequence_privileged(&calldata->args);
7600         msg.rpc_argp = &calldata->args;
7601         msg.rpc_resp = &calldata->res;
7602         calldata->clp = clp;
7603         task_setup_data.callback_data = calldata;
7604
7605         return rpc_run_task(&task_setup_data);
7606 }
7607
7608 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7609 {
7610         struct rpc_task *task;
7611         int ret = 0;
7612
7613         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7614                 return -EAGAIN;
7615         task = _nfs41_proc_sequence(clp, cred, false);
7616         if (IS_ERR(task))
7617                 ret = PTR_ERR(task);
7618         else
7619                 rpc_put_task_async(task);
7620         dprintk("<-- %s status=%d\n", __func__, ret);
7621         return ret;
7622 }
7623
7624 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7625 {
7626         struct rpc_task *task;
7627         int ret;
7628
7629         task = _nfs41_proc_sequence(clp, cred, true);
7630         if (IS_ERR(task)) {
7631                 ret = PTR_ERR(task);
7632                 goto out;
7633         }
7634         ret = rpc_wait_for_completion_task(task);
7635         if (!ret)
7636                 ret = task->tk_status;
7637         rpc_put_task(task);
7638 out:
7639         dprintk("<-- %s status=%d\n", __func__, ret);
7640         return ret;
7641 }
7642
7643 struct nfs4_reclaim_complete_data {
7644         struct nfs_client *clp;
7645         struct nfs41_reclaim_complete_args arg;
7646         struct nfs41_reclaim_complete_res res;
7647 };
7648
7649 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7650 {
7651         struct nfs4_reclaim_complete_data *calldata = data;
7652
7653         nfs41_setup_sequence(calldata->clp->cl_session,
7654                         &calldata->arg.seq_args,
7655                         &calldata->res.seq_res,
7656                         task);
7657 }
7658
7659 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7660 {
7661         switch(task->tk_status) {
7662         case 0:
7663         case -NFS4ERR_COMPLETE_ALREADY:
7664         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7665                 break;
7666         case -NFS4ERR_DELAY:
7667                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7668                 /* fall through */
7669         case -NFS4ERR_RETRY_UNCACHED_REP:
7670                 return -EAGAIN;
7671         default:
7672                 nfs4_schedule_lease_recovery(clp);
7673         }
7674         return 0;
7675 }
7676
7677 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7678 {
7679         struct nfs4_reclaim_complete_data *calldata = data;
7680         struct nfs_client *clp = calldata->clp;
7681         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7682
7683         dprintk("--> %s\n", __func__);
7684         if (!nfs41_sequence_done(task, res))
7685                 return;
7686
7687         trace_nfs4_reclaim_complete(clp, task->tk_status);
7688         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7689                 rpc_restart_call_prepare(task);
7690                 return;
7691         }
7692         dprintk("<-- %s\n", __func__);
7693 }
7694
7695 static void nfs4_free_reclaim_complete_data(void *data)
7696 {
7697         struct nfs4_reclaim_complete_data *calldata = data;
7698
7699         kfree(calldata);
7700 }
7701
7702 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7703         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7704         .rpc_call_done = nfs4_reclaim_complete_done,
7705         .rpc_release = nfs4_free_reclaim_complete_data,
7706 };
7707
7708 /*
7709  * Issue a global reclaim complete.
7710  */
7711 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7712                 struct rpc_cred *cred)
7713 {
7714         struct nfs4_reclaim_complete_data *calldata;
7715         struct rpc_task *task;
7716         struct rpc_message msg = {
7717                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7718                 .rpc_cred = cred,
7719         };
7720         struct rpc_task_setup task_setup_data = {
7721                 .rpc_client = clp->cl_rpcclient,
7722                 .rpc_message = &msg,
7723                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7724                 .flags = RPC_TASK_ASYNC,
7725         };
7726         int status = -ENOMEM;
7727
7728         dprintk("--> %s\n", __func__);
7729         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7730         if (calldata == NULL)
7731                 goto out;
7732         calldata->clp = clp;
7733         calldata->arg.one_fs = 0;
7734
7735         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7736         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7737         msg.rpc_argp = &calldata->arg;
7738         msg.rpc_resp = &calldata->res;
7739         task_setup_data.callback_data = calldata;
7740         task = rpc_run_task(&task_setup_data);
7741         if (IS_ERR(task)) {
7742                 status = PTR_ERR(task);
7743                 goto out;
7744         }
7745         status = nfs4_wait_for_completion_rpc_task(task);
7746         if (status == 0)
7747                 status = task->tk_status;
7748         rpc_put_task(task);
7749         return 0;
7750 out:
7751         dprintk("<-- %s status=%d\n", __func__, status);
7752         return status;
7753 }
7754
7755 static void
7756 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7757 {
7758         struct nfs4_layoutget *lgp = calldata;
7759         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7760         struct nfs4_session *session = nfs4_get_session(server);
7761
7762         dprintk("--> %s\n", __func__);
7763         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7764          * right now covering the LAYOUTGET we are about to send.
7765          * However, that is not so catastrophic, and there seems
7766          * to be no way to prevent it completely.
7767          */
7768         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7769                                 &lgp->res.seq_res, task))
7770                 return;
7771         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7772                                           NFS_I(lgp->args.inode)->layout,
7773                                           &lgp->args.range,
7774                                           lgp->args.ctx->state)) {
7775                 rpc_exit(task, NFS4_OK);
7776         }
7777 }
7778
7779 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7780 {
7781         struct nfs4_layoutget *lgp = calldata;
7782         struct inode *inode = lgp->args.inode;
7783         struct nfs_server *server = NFS_SERVER(inode);
7784         struct pnfs_layout_hdr *lo;
7785         struct nfs4_state *state = NULL;
7786         unsigned long timeo, now, giveup;
7787
7788         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7789
7790         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7791                 goto out;
7792
7793         switch (task->tk_status) {
7794         case 0:
7795                 goto out;
7796         /*
7797          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7798          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7799          */
7800         case -NFS4ERR_BADLAYOUT:
7801                 goto out_overflow;
7802         /*
7803          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7804          * (or clients) writing to the same RAID stripe except when
7805          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7806          */
7807         case -NFS4ERR_LAYOUTTRYLATER:
7808                 if (lgp->args.minlength == 0)
7809                         goto out_overflow;
7810         /*
7811          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7812          * existing layout before getting a new one).
7813          */
7814         case -NFS4ERR_RECALLCONFLICT:
7815                 timeo = rpc_get_timeout(task->tk_client);
7816                 giveup = lgp->args.timestamp + timeo;
7817                 now = jiffies;
7818                 if (time_after(giveup, now)) {
7819                         unsigned long delay;
7820
7821                         /* Delay for:
7822                          * - Not less then NFS4_POLL_RETRY_MIN.
7823                          * - One last time a jiffie before we give up
7824                          * - exponential backoff (time_now minus start_attempt)
7825                          */
7826                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7827                                     min((giveup - now - 1),
7828                                         now - lgp->args.timestamp));
7829
7830                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7831                                 __func__, delay);
7832                         rpc_delay(task, delay);
7833                         /* Do not call nfs4_async_handle_error() */
7834                         goto out_restart;
7835                 }
7836                 break;
7837         case -NFS4ERR_EXPIRED:
7838         case -NFS4ERR_BAD_STATEID:
7839                 spin_lock(&inode->i_lock);
7840                 if (nfs4_stateid_match(&lgp->args.stateid,
7841                                         &lgp->args.ctx->state->stateid)) {
7842                         spin_unlock(&inode->i_lock);
7843                         /* If the open stateid was bad, then recover it. */
7844                         state = lgp->args.ctx->state;
7845                         break;
7846                 }
7847                 lo = NFS_I(inode)->layout;
7848                 if (lo && nfs4_stateid_match(&lgp->args.stateid,
7849                                         &lo->plh_stateid)) {
7850                         LIST_HEAD(head);
7851
7852                         /*
7853                          * Mark the bad layout state as invalid, then retry
7854                          * with the current stateid.
7855                          */
7856                         set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7857                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7858                         spin_unlock(&inode->i_lock);
7859                         pnfs_free_lseg_list(&head);
7860                 } else
7861                         spin_unlock(&inode->i_lock);
7862                 goto out_restart;
7863         }
7864         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7865                 goto out_restart;
7866 out:
7867         dprintk("<-- %s\n", __func__);
7868         return;
7869 out_restart:
7870         task->tk_status = 0;
7871         rpc_restart_call_prepare(task);
7872         return;
7873 out_overflow:
7874         task->tk_status = -EOVERFLOW;
7875         goto out;
7876 }
7877
7878 static size_t max_response_pages(struct nfs_server *server)
7879 {
7880         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7881         return nfs_page_array_len(0, max_resp_sz);
7882 }
7883
7884 static void nfs4_free_pages(struct page **pages, size_t size)
7885 {
7886         int i;
7887
7888         if (!pages)
7889                 return;
7890
7891         for (i = 0; i < size; i++) {
7892                 if (!pages[i])
7893                         break;
7894                 __free_page(pages[i]);
7895         }
7896         kfree(pages);
7897 }
7898
7899 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7900 {
7901         struct page **pages;
7902         int i;
7903
7904         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7905         if (!pages) {
7906                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7907                 return NULL;
7908         }
7909
7910         for (i = 0; i < size; i++) {
7911                 pages[i] = alloc_page(gfp_flags);
7912                 if (!pages[i]) {
7913                         dprintk("%s: failed to allocate page\n", __func__);
7914                         nfs4_free_pages(pages, size);
7915                         return NULL;
7916                 }
7917         }
7918
7919         return pages;
7920 }
7921
7922 static void nfs4_layoutget_release(void *calldata)
7923 {
7924         struct nfs4_layoutget *lgp = calldata;
7925         struct inode *inode = lgp->args.inode;
7926         struct nfs_server *server = NFS_SERVER(inode);
7927         size_t max_pages = max_response_pages(server);
7928
7929         dprintk("--> %s\n", __func__);
7930         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7931         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7932         put_nfs_open_context(lgp->args.ctx);
7933         kfree(calldata);
7934         dprintk("<-- %s\n", __func__);
7935 }
7936
7937 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7938         .rpc_call_prepare = nfs4_layoutget_prepare,
7939         .rpc_call_done = nfs4_layoutget_done,
7940         .rpc_release = nfs4_layoutget_release,
7941 };
7942
7943 struct pnfs_layout_segment *
7944 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7945 {
7946         struct inode *inode = lgp->args.inode;
7947         struct nfs_server *server = NFS_SERVER(inode);
7948         size_t max_pages = max_response_pages(server);
7949         struct rpc_task *task;
7950         struct rpc_message msg = {
7951                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7952                 .rpc_argp = &lgp->args,
7953                 .rpc_resp = &lgp->res,
7954                 .rpc_cred = lgp->cred,
7955         };
7956         struct rpc_task_setup task_setup_data = {
7957                 .rpc_client = server->client,
7958                 .rpc_message = &msg,
7959                 .callback_ops = &nfs4_layoutget_call_ops,
7960                 .callback_data = lgp,
7961                 .flags = RPC_TASK_ASYNC,
7962         };
7963         struct pnfs_layout_segment *lseg = NULL;
7964         int status = 0;
7965
7966         dprintk("--> %s\n", __func__);
7967
7968         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7969         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7970
7971         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7972         if (!lgp->args.layout.pages) {
7973                 nfs4_layoutget_release(lgp);
7974                 return ERR_PTR(-ENOMEM);
7975         }
7976         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7977         lgp->args.timestamp = jiffies;
7978
7979         lgp->res.layoutp = &lgp->args.layout;
7980         lgp->res.seq_res.sr_slot = NULL;
7981         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7982
7983         task = rpc_run_task(&task_setup_data);
7984         if (IS_ERR(task))
7985                 return ERR_CAST(task);
7986         status = nfs4_wait_for_completion_rpc_task(task);
7987         if (status == 0)
7988                 status = task->tk_status;
7989         trace_nfs4_layoutget(lgp->args.ctx,
7990                         &lgp->args.range,
7991                         &lgp->res.range,
7992                         status);
7993         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7994         if (status == 0 && lgp->res.layoutp->len)
7995                 lseg = pnfs_layout_process(lgp);
7996         rpc_put_task(task);
7997         dprintk("<-- %s status=%d\n", __func__, status);
7998         if (status)
7999                 return ERR_PTR(status);
8000         return lseg;
8001 }
8002
8003 static void
8004 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8005 {
8006         struct nfs4_layoutreturn *lrp = calldata;
8007
8008         dprintk("--> %s\n", __func__);
8009         nfs41_setup_sequence(lrp->clp->cl_session,
8010                         &lrp->args.seq_args,
8011                         &lrp->res.seq_res,
8012                         task);
8013 }
8014
8015 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8016 {
8017         struct nfs4_layoutreturn *lrp = calldata;
8018         struct nfs_server *server;
8019
8020         dprintk("--> %s\n", __func__);
8021
8022         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8023                 return;
8024
8025         server = NFS_SERVER(lrp->args.inode);
8026         switch (task->tk_status) {
8027         default:
8028                 task->tk_status = 0;
8029         case 0:
8030                 break;
8031         case -NFS4ERR_DELAY:
8032                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8033                         break;
8034                 rpc_restart_call_prepare(task);
8035                 return;
8036         }
8037         dprintk("<-- %s\n", __func__);
8038 }
8039
8040 static void nfs4_layoutreturn_release(void *calldata)
8041 {
8042         struct nfs4_layoutreturn *lrp = calldata;
8043         struct pnfs_layout_hdr *lo = lrp->args.layout;
8044         LIST_HEAD(freeme);
8045
8046         dprintk("--> %s\n", __func__);
8047         spin_lock(&lo->plh_inode->i_lock);
8048         if (lrp->res.lrs_present)
8049                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8050         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
8051         pnfs_clear_layoutreturn_waitbit(lo);
8052         lo->plh_block_lgets--;
8053         spin_unlock(&lo->plh_inode->i_lock);
8054         pnfs_free_lseg_list(&freeme);
8055         pnfs_put_layout_hdr(lrp->args.layout);
8056         nfs_iput_and_deactive(lrp->inode);
8057         kfree(calldata);
8058         dprintk("<-- %s\n", __func__);
8059 }
8060
8061 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8062         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8063         .rpc_call_done = nfs4_layoutreturn_done,
8064         .rpc_release = nfs4_layoutreturn_release,
8065 };
8066
8067 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8068 {
8069         struct rpc_task *task;
8070         struct rpc_message msg = {
8071                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8072                 .rpc_argp = &lrp->args,
8073                 .rpc_resp = &lrp->res,
8074                 .rpc_cred = lrp->cred,
8075         };
8076         struct rpc_task_setup task_setup_data = {
8077                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8078                 .rpc_message = &msg,
8079                 .callback_ops = &nfs4_layoutreturn_call_ops,
8080                 .callback_data = lrp,
8081         };
8082         int status = 0;
8083
8084         dprintk("--> %s\n", __func__);
8085         if (!sync) {
8086                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8087                 if (!lrp->inode) {
8088                         nfs4_layoutreturn_release(lrp);
8089                         return -EAGAIN;
8090                 }
8091                 task_setup_data.flags |= RPC_TASK_ASYNC;
8092         }
8093         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8094         task = rpc_run_task(&task_setup_data);
8095         if (IS_ERR(task))
8096                 return PTR_ERR(task);
8097         if (sync)
8098                 status = task->tk_status;
8099         trace_nfs4_layoutreturn(lrp->args.inode, status);
8100         dprintk("<-- %s status=%d\n", __func__, status);
8101         rpc_put_task(task);
8102         return status;
8103 }
8104
8105 static int
8106 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8107                 struct pnfs_device *pdev,
8108                 struct rpc_cred *cred)
8109 {
8110         struct nfs4_getdeviceinfo_args args = {
8111                 .pdev = pdev,
8112                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8113                         NOTIFY_DEVICEID4_DELETE,
8114         };
8115         struct nfs4_getdeviceinfo_res res = {
8116                 .pdev = pdev,
8117         };
8118         struct rpc_message msg = {
8119                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8120                 .rpc_argp = &args,
8121                 .rpc_resp = &res,
8122                 .rpc_cred = cred,
8123         };
8124         int status;
8125
8126         dprintk("--> %s\n", __func__);
8127         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8128         if (res.notification & ~args.notify_types)
8129                 dprintk("%s: unsupported notification\n", __func__);
8130         if (res.notification != args.notify_types)
8131                 pdev->nocache = 1;
8132
8133         dprintk("<-- %s status=%d\n", __func__, status);
8134
8135         return status;
8136 }
8137
8138 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8139                 struct pnfs_device *pdev,
8140                 struct rpc_cred *cred)
8141 {
8142         struct nfs4_exception exception = { };
8143         int err;
8144
8145         do {
8146                 err = nfs4_handle_exception(server,
8147                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8148                                         &exception);
8149         } while (exception.retry);
8150         return err;
8151 }
8152 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8153
8154 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8155 {
8156         struct nfs4_layoutcommit_data *data = calldata;
8157         struct nfs_server *server = NFS_SERVER(data->args.inode);
8158         struct nfs4_session *session = nfs4_get_session(server);
8159
8160         nfs41_setup_sequence(session,
8161                         &data->args.seq_args,
8162                         &data->res.seq_res,
8163                         task);
8164 }
8165
8166 static void
8167 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8168 {
8169         struct nfs4_layoutcommit_data *data = calldata;
8170         struct nfs_server *server = NFS_SERVER(data->args.inode);
8171
8172         if (!nfs41_sequence_done(task, &data->res.seq_res))
8173                 return;
8174
8175         switch (task->tk_status) { /* Just ignore these failures */
8176         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8177         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8178         case -NFS4ERR_BADLAYOUT:     /* no layout */
8179         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8180                 task->tk_status = 0;
8181         case 0:
8182                 break;
8183         default:
8184                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8185                         rpc_restart_call_prepare(task);
8186                         return;
8187                 }
8188         }
8189 }
8190
8191 static void nfs4_layoutcommit_release(void *calldata)
8192 {
8193         struct nfs4_layoutcommit_data *data = calldata;
8194
8195         pnfs_cleanup_layoutcommit(data);
8196         nfs_post_op_update_inode_force_wcc(data->args.inode,
8197                                            data->res.fattr);
8198         put_rpccred(data->cred);
8199         nfs_iput_and_deactive(data->inode);
8200         kfree(data);
8201 }
8202
8203 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8204         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8205         .rpc_call_done = nfs4_layoutcommit_done,
8206         .rpc_release = nfs4_layoutcommit_release,
8207 };
8208
8209 int
8210 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8211 {
8212         struct rpc_message msg = {
8213                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8214                 .rpc_argp = &data->args,
8215                 .rpc_resp = &data->res,
8216                 .rpc_cred = data->cred,
8217         };
8218         struct rpc_task_setup task_setup_data = {
8219                 .task = &data->task,
8220                 .rpc_client = NFS_CLIENT(data->args.inode),
8221                 .rpc_message = &msg,
8222                 .callback_ops = &nfs4_layoutcommit_ops,
8223                 .callback_data = data,
8224         };
8225         struct rpc_task *task;
8226         int status = 0;
8227
8228         dprintk("NFS: initiating layoutcommit call. sync %d "
8229                 "lbw: %llu inode %lu\n", sync,
8230                 data->args.lastbytewritten,
8231                 data->args.inode->i_ino);
8232
8233         if (!sync) {
8234                 data->inode = nfs_igrab_and_active(data->args.inode);
8235                 if (data->inode == NULL) {
8236                         nfs4_layoutcommit_release(data);
8237                         return -EAGAIN;
8238                 }
8239                 task_setup_data.flags = RPC_TASK_ASYNC;
8240         }
8241         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8242         task = rpc_run_task(&task_setup_data);
8243         if (IS_ERR(task))
8244                 return PTR_ERR(task);
8245         if (sync)
8246                 status = task->tk_status;
8247         trace_nfs4_layoutcommit(data->args.inode, status);
8248         dprintk("%s: status %d\n", __func__, status);
8249         rpc_put_task(task);
8250         return status;
8251 }
8252
8253 /**
8254  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8255  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8256  */
8257 static int
8258 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8259                     struct nfs_fsinfo *info,
8260                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8261 {
8262         struct nfs41_secinfo_no_name_args args = {
8263                 .style = SECINFO_STYLE_CURRENT_FH,
8264         };
8265         struct nfs4_secinfo_res res = {
8266                 .flavors = flavors,
8267         };
8268         struct rpc_message msg = {
8269                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8270                 .rpc_argp = &args,
8271                 .rpc_resp = &res,
8272         };
8273         struct rpc_clnt *clnt = server->client;
8274         struct rpc_cred *cred = NULL;
8275         int status;
8276
8277         if (use_integrity) {
8278                 clnt = server->nfs_client->cl_rpcclient;
8279                 cred = nfs4_get_clid_cred(server->nfs_client);
8280                 msg.rpc_cred = cred;
8281         }
8282
8283         dprintk("--> %s\n", __func__);
8284         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8285                                 &res.seq_res, 0);
8286         dprintk("<-- %s status=%d\n", __func__, status);
8287
8288         if (cred)
8289                 put_rpccred(cred);
8290
8291         return status;
8292 }
8293
8294 static int
8295 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8296                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8297 {
8298         struct nfs4_exception exception = { };
8299         int err;
8300         do {
8301                 /* first try using integrity protection */
8302                 err = -NFS4ERR_WRONGSEC;
8303
8304                 /* try to use integrity protection with machine cred */
8305                 if (_nfs4_is_integrity_protected(server->nfs_client))
8306                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8307                                                           flavors, true);
8308
8309                 /*
8310                  * if unable to use integrity protection, or SECINFO with
8311                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8312                  * disallowed by spec, but exists in deployed servers) use
8313                  * the current filesystem's rpc_client and the user cred.
8314                  */
8315                 if (err == -NFS4ERR_WRONGSEC)
8316                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8317                                                           flavors, false);
8318
8319                 switch (err) {
8320                 case 0:
8321                 case -NFS4ERR_WRONGSEC:
8322                 case -ENOTSUPP:
8323                         goto out;
8324                 default:
8325                         err = nfs4_handle_exception(server, err, &exception);
8326                 }
8327         } while (exception.retry);
8328 out:
8329         return err;
8330 }
8331
8332 static int
8333 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8334                     struct nfs_fsinfo *info)
8335 {
8336         int err;
8337         struct page *page;
8338         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8339         struct nfs4_secinfo_flavors *flavors;
8340         struct nfs4_secinfo4 *secinfo;
8341         int i;
8342
8343         page = alloc_page(GFP_KERNEL);
8344         if (!page) {
8345                 err = -ENOMEM;
8346                 goto out;
8347         }
8348
8349         flavors = page_address(page);
8350         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8351
8352         /*
8353          * Fall back on "guess and check" method if
8354          * the server doesn't support SECINFO_NO_NAME
8355          */
8356         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8357                 err = nfs4_find_root_sec(server, fhandle, info);
8358                 goto out_freepage;
8359         }
8360         if (err)
8361                 goto out_freepage;
8362
8363         for (i = 0; i < flavors->num_flavors; i++) {
8364                 secinfo = &flavors->flavors[i];
8365
8366                 switch (secinfo->flavor) {
8367                 case RPC_AUTH_NULL:
8368                 case RPC_AUTH_UNIX:
8369                 case RPC_AUTH_GSS:
8370                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8371                                         &secinfo->flavor_info);
8372                         break;
8373                 default:
8374                         flavor = RPC_AUTH_MAXFLAVOR;
8375                         break;
8376                 }
8377
8378                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8379                         flavor = RPC_AUTH_MAXFLAVOR;
8380
8381                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8382                         err = nfs4_lookup_root_sec(server, fhandle,
8383                                                    info, flavor);
8384                         if (!err)
8385                                 break;
8386                 }
8387         }
8388
8389         if (flavor == RPC_AUTH_MAXFLAVOR)
8390                 err = -EPERM;
8391
8392 out_freepage:
8393         put_page(page);
8394         if (err == -EACCES)
8395                 return -EPERM;
8396 out:
8397         return err;
8398 }
8399
8400 static int _nfs41_test_stateid(struct nfs_server *server,
8401                 nfs4_stateid *stateid,
8402                 struct rpc_cred *cred)
8403 {
8404         int status;
8405         struct nfs41_test_stateid_args args = {
8406                 .stateid = stateid,
8407         };
8408         struct nfs41_test_stateid_res res;
8409         struct rpc_message msg = {
8410                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8411                 .rpc_argp = &args,
8412                 .rpc_resp = &res,
8413                 .rpc_cred = cred,
8414         };
8415         struct rpc_clnt *rpc_client = server->client;
8416
8417         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8418                 &rpc_client, &msg);
8419
8420         dprintk("NFS call  test_stateid %p\n", stateid);
8421         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8422         nfs4_set_sequence_privileged(&args.seq_args);
8423         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8424                         &args.seq_args, &res.seq_res);
8425         if (status != NFS_OK) {
8426                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8427                 return status;
8428         }
8429         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8430         return -res.status;
8431 }
8432
8433 /**
8434  * nfs41_test_stateid - perform a TEST_STATEID operation
8435  *
8436  * @server: server / transport on which to perform the operation
8437  * @stateid: state ID to test
8438  * @cred: credential
8439  *
8440  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8441  * Otherwise a negative NFS4ERR value is returned if the operation
8442  * failed or the state ID is not currently valid.
8443  */
8444 static int nfs41_test_stateid(struct nfs_server *server,
8445                 nfs4_stateid *stateid,
8446                 struct rpc_cred *cred)
8447 {
8448         struct nfs4_exception exception = { };
8449         int err;
8450         do {
8451                 err = _nfs41_test_stateid(server, stateid, cred);
8452                 if (err != -NFS4ERR_DELAY)
8453                         break;
8454                 nfs4_handle_exception(server, err, &exception);
8455         } while (exception.retry);
8456         return err;
8457 }
8458
8459 struct nfs_free_stateid_data {
8460         struct nfs_server *server;
8461         struct nfs41_free_stateid_args args;
8462         struct nfs41_free_stateid_res res;
8463 };
8464
8465 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8466 {
8467         struct nfs_free_stateid_data *data = calldata;
8468         nfs41_setup_sequence(nfs4_get_session(data->server),
8469                         &data->args.seq_args,
8470                         &data->res.seq_res,
8471                         task);
8472 }
8473
8474 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8475 {
8476         struct nfs_free_stateid_data *data = calldata;
8477
8478         nfs41_sequence_done(task, &data->res.seq_res);
8479
8480         switch (task->tk_status) {
8481         case -NFS4ERR_DELAY:
8482                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8483                         rpc_restart_call_prepare(task);
8484         }
8485 }
8486
8487 static void nfs41_free_stateid_release(void *calldata)
8488 {
8489         kfree(calldata);
8490 }
8491
8492 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8493         .rpc_call_prepare = nfs41_free_stateid_prepare,
8494         .rpc_call_done = nfs41_free_stateid_done,
8495         .rpc_release = nfs41_free_stateid_release,
8496 };
8497
8498 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8499                 nfs4_stateid *stateid,
8500                 struct rpc_cred *cred,
8501                 bool privileged)
8502 {
8503         struct rpc_message msg = {
8504                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8505                 .rpc_cred = cred,
8506         };
8507         struct rpc_task_setup task_setup = {
8508                 .rpc_client = server->client,
8509                 .rpc_message = &msg,
8510                 .callback_ops = &nfs41_free_stateid_ops,
8511                 .flags = RPC_TASK_ASYNC,
8512         };
8513         struct nfs_free_stateid_data *data;
8514
8515         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8516                 &task_setup.rpc_client, &msg);
8517
8518         dprintk("NFS call  free_stateid %p\n", stateid);
8519         data = kmalloc(sizeof(*data), GFP_NOFS);
8520         if (!data)
8521                 return ERR_PTR(-ENOMEM);
8522         data->server = server;
8523         nfs4_stateid_copy(&data->args.stateid, stateid);
8524
8525         task_setup.callback_data = data;
8526
8527         msg.rpc_argp = &data->args;
8528         msg.rpc_resp = &data->res;
8529         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8530         if (privileged)
8531                 nfs4_set_sequence_privileged(&data->args.seq_args);
8532
8533         return rpc_run_task(&task_setup);
8534 }
8535
8536 /**
8537  * nfs41_free_stateid - perform a FREE_STATEID operation
8538  *
8539  * @server: server / transport on which to perform the operation
8540  * @stateid: state ID to release
8541  * @cred: credential
8542  *
8543  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8544  * negative NFS4ERR value is returned.
8545  */
8546 static int nfs41_free_stateid(struct nfs_server *server,
8547                 nfs4_stateid *stateid,
8548                 struct rpc_cred *cred)
8549 {
8550         struct rpc_task *task;
8551         int ret;
8552
8553         task = _nfs41_free_stateid(server, stateid, cred, true);
8554         if (IS_ERR(task))
8555                 return PTR_ERR(task);
8556         ret = rpc_wait_for_completion_task(task);
8557         if (!ret)
8558                 ret = task->tk_status;
8559         rpc_put_task(task);
8560         return ret;
8561 }
8562
8563 static void
8564 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8565 {
8566         struct rpc_task *task;
8567         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8568
8569         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8570         nfs4_free_lock_state(server, lsp);
8571         if (IS_ERR(task))
8572                 return;
8573         rpc_put_task(task);
8574 }
8575
8576 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8577                 const nfs4_stateid *s2)
8578 {
8579         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8580                 return false;
8581
8582         if (s1->seqid == s2->seqid)
8583                 return true;
8584         if (s1->seqid == 0 || s2->seqid == 0)
8585                 return true;
8586
8587         return false;
8588 }
8589
8590 #endif /* CONFIG_NFS_V4_1 */
8591
8592 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8593                 const nfs4_stateid *s2)
8594 {
8595         return nfs4_stateid_match(s1, s2);
8596 }
8597
8598
8599 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8600         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8601         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8602         .recover_open   = nfs4_open_reclaim,
8603         .recover_lock   = nfs4_lock_reclaim,
8604         .establish_clid = nfs4_init_clientid,
8605         .detect_trunking = nfs40_discover_server_trunking,
8606 };
8607
8608 #if defined(CONFIG_NFS_V4_1)
8609 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8610         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8611         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8612         .recover_open   = nfs4_open_reclaim,
8613         .recover_lock   = nfs4_lock_reclaim,
8614         .establish_clid = nfs41_init_clientid,
8615         .reclaim_complete = nfs41_proc_reclaim_complete,
8616         .detect_trunking = nfs41_discover_server_trunking,
8617 };
8618 #endif /* CONFIG_NFS_V4_1 */
8619
8620 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8621         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8622         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8623         .recover_open   = nfs40_open_expired,
8624         .recover_lock   = nfs4_lock_expired,
8625         .establish_clid = nfs4_init_clientid,
8626 };
8627
8628 #if defined(CONFIG_NFS_V4_1)
8629 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8630         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8631         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8632         .recover_open   = nfs41_open_expired,
8633         .recover_lock   = nfs41_lock_expired,
8634         .establish_clid = nfs41_init_clientid,
8635 };
8636 #endif /* CONFIG_NFS_V4_1 */
8637
8638 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8639         .sched_state_renewal = nfs4_proc_async_renew,
8640         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8641         .renew_lease = nfs4_proc_renew,
8642 };
8643
8644 #if defined(CONFIG_NFS_V4_1)
8645 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8646         .sched_state_renewal = nfs41_proc_async_sequence,
8647         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8648         .renew_lease = nfs4_proc_sequence,
8649 };
8650 #endif
8651
8652 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8653         .get_locations = _nfs40_proc_get_locations,
8654         .fsid_present = _nfs40_proc_fsid_present,
8655 };
8656
8657 #if defined(CONFIG_NFS_V4_1)
8658 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8659         .get_locations = _nfs41_proc_get_locations,
8660         .fsid_present = _nfs41_proc_fsid_present,
8661 };
8662 #endif  /* CONFIG_NFS_V4_1 */
8663
8664 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8665         .minor_version = 0,
8666         .init_caps = NFS_CAP_READDIRPLUS
8667                 | NFS_CAP_ATOMIC_OPEN
8668                 | NFS_CAP_POSIX_LOCK,
8669         .init_client = nfs40_init_client,
8670         .shutdown_client = nfs40_shutdown_client,
8671         .match_stateid = nfs4_match_stateid,
8672         .find_root_sec = nfs4_find_root_sec,
8673         .free_lock_state = nfs4_release_lockowner,
8674         .alloc_seqid = nfs_alloc_seqid,
8675         .call_sync_ops = &nfs40_call_sync_ops,
8676         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8677         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8678         .state_renewal_ops = &nfs40_state_renewal_ops,
8679         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8680 };
8681
8682 #if defined(CONFIG_NFS_V4_1)
8683 static struct nfs_seqid *
8684 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8685 {
8686         return NULL;
8687 }
8688
8689 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8690         .minor_version = 1,
8691         .init_caps = NFS_CAP_READDIRPLUS
8692                 | NFS_CAP_ATOMIC_OPEN
8693                 | NFS_CAP_POSIX_LOCK
8694                 | NFS_CAP_STATEID_NFSV41
8695                 | NFS_CAP_ATOMIC_OPEN_V1,
8696         .init_client = nfs41_init_client,
8697         .shutdown_client = nfs41_shutdown_client,
8698         .match_stateid = nfs41_match_stateid,
8699         .find_root_sec = nfs41_find_root_sec,
8700         .free_lock_state = nfs41_free_lock_state,
8701         .alloc_seqid = nfs_alloc_no_seqid,
8702         .call_sync_ops = &nfs41_call_sync_ops,
8703         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8704         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8705         .state_renewal_ops = &nfs41_state_renewal_ops,
8706         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8707 };
8708 #endif
8709
8710 #if defined(CONFIG_NFS_V4_2)
8711 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8712         .minor_version = 2,
8713         .init_caps = NFS_CAP_READDIRPLUS
8714                 | NFS_CAP_ATOMIC_OPEN
8715                 | NFS_CAP_POSIX_LOCK
8716                 | NFS_CAP_STATEID_NFSV41
8717                 | NFS_CAP_ATOMIC_OPEN_V1
8718                 | NFS_CAP_ALLOCATE
8719                 | NFS_CAP_DEALLOCATE
8720                 | NFS_CAP_SEEK
8721                 | NFS_CAP_LAYOUTSTATS,
8722         .init_client = nfs41_init_client,
8723         .shutdown_client = nfs41_shutdown_client,
8724         .match_stateid = nfs41_match_stateid,
8725         .find_root_sec = nfs41_find_root_sec,
8726         .free_lock_state = nfs41_free_lock_state,
8727         .call_sync_ops = &nfs41_call_sync_ops,
8728         .alloc_seqid = nfs_alloc_no_seqid,
8729         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8730         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8731         .state_renewal_ops = &nfs41_state_renewal_ops,
8732         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8733 };
8734 #endif
8735
8736 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8737         [0] = &nfs_v4_0_minor_ops,
8738 #if defined(CONFIG_NFS_V4_1)
8739         [1] = &nfs_v4_1_minor_ops,
8740 #endif
8741 #if defined(CONFIG_NFS_V4_2)
8742         [2] = &nfs_v4_2_minor_ops,
8743 #endif
8744 };
8745
8746 static const struct inode_operations nfs4_dir_inode_operations = {
8747         .create         = nfs_create,
8748         .lookup         = nfs_lookup,
8749         .atomic_open    = nfs_atomic_open,
8750         .link           = nfs_link,
8751         .unlink         = nfs_unlink,
8752         .symlink        = nfs_symlink,
8753         .mkdir          = nfs_mkdir,
8754         .rmdir          = nfs_rmdir,
8755         .mknod          = nfs_mknod,
8756         .rename         = nfs_rename,
8757         .permission     = nfs_permission,
8758         .getattr        = nfs_getattr,
8759         .setattr        = nfs_setattr,
8760         .getxattr       = generic_getxattr,
8761         .setxattr       = generic_setxattr,
8762         .listxattr      = generic_listxattr,
8763         .removexattr    = generic_removexattr,
8764 };
8765
8766 static const struct inode_operations nfs4_file_inode_operations = {
8767         .permission     = nfs_permission,
8768         .getattr        = nfs_getattr,
8769         .setattr        = nfs_setattr,
8770         .getxattr       = generic_getxattr,
8771         .setxattr       = generic_setxattr,
8772         .listxattr      = generic_listxattr,
8773         .removexattr    = generic_removexattr,
8774 };
8775
8776 const struct nfs_rpc_ops nfs_v4_clientops = {
8777         .version        = 4,                    /* protocol version */
8778         .dentry_ops     = &nfs4_dentry_operations,
8779         .dir_inode_ops  = &nfs4_dir_inode_operations,
8780         .file_inode_ops = &nfs4_file_inode_operations,
8781         .file_ops       = &nfs4_file_operations,
8782         .getroot        = nfs4_proc_get_root,
8783         .submount       = nfs4_submount,
8784         .try_mount      = nfs4_try_mount,
8785         .getattr        = nfs4_proc_getattr,
8786         .setattr        = nfs4_proc_setattr,
8787         .lookup         = nfs4_proc_lookup,
8788         .access         = nfs4_proc_access,
8789         .readlink       = nfs4_proc_readlink,
8790         .create         = nfs4_proc_create,
8791         .remove         = nfs4_proc_remove,
8792         .unlink_setup   = nfs4_proc_unlink_setup,
8793         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8794         .unlink_done    = nfs4_proc_unlink_done,
8795         .rename_setup   = nfs4_proc_rename_setup,
8796         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8797         .rename_done    = nfs4_proc_rename_done,
8798         .link           = nfs4_proc_link,
8799         .symlink        = nfs4_proc_symlink,
8800         .mkdir          = nfs4_proc_mkdir,
8801         .rmdir          = nfs4_proc_remove,
8802         .readdir        = nfs4_proc_readdir,
8803         .mknod          = nfs4_proc_mknod,
8804         .statfs         = nfs4_proc_statfs,
8805         .fsinfo         = nfs4_proc_fsinfo,
8806         .pathconf       = nfs4_proc_pathconf,
8807         .set_capabilities = nfs4_server_capabilities,
8808         .decode_dirent  = nfs4_decode_dirent,
8809         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8810         .read_setup     = nfs4_proc_read_setup,
8811         .read_done      = nfs4_read_done,
8812         .write_setup    = nfs4_proc_write_setup,
8813         .write_done     = nfs4_write_done,
8814         .commit_setup   = nfs4_proc_commit_setup,
8815         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8816         .commit_done    = nfs4_commit_done,
8817         .lock           = nfs4_proc_lock,
8818         .clear_acl_cache = nfs4_zap_acl_attr,
8819         .close_context  = nfs4_close_context,
8820         .open_context   = nfs4_atomic_open,
8821         .have_delegation = nfs4_have_delegation,
8822         .return_delegation = nfs4_inode_return_delegation,
8823         .alloc_client   = nfs4_alloc_client,
8824         .init_client    = nfs4_init_client,
8825         .free_client    = nfs4_free_client,
8826         .create_server  = nfs4_create_server,
8827         .clone_server   = nfs_clone_server,
8828 };
8829
8830 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8831         .prefix = XATTR_NAME_NFSV4_ACL,
8832         .list   = nfs4_xattr_list_nfs4_acl,
8833         .get    = nfs4_xattr_get_nfs4_acl,
8834         .set    = nfs4_xattr_set_nfs4_acl,
8835 };
8836
8837 const struct xattr_handler *nfs4_xattr_handlers[] = {
8838         &nfs4_xattr_nfs4_acl_handler,
8839 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8840         &nfs4_xattr_nfs4_label_handler,
8841 #endif
8842         NULL
8843 };
8844
8845 /*
8846  * Local variables:
8847  *  c-basic-offset: 8
8848  * End:
8849  */