nfs4: client: do not send empty SETATTR after OPEN_CREATE
[cascardo/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 /* file attributes which can be mapped to nfs attributes */
78 #define NFS4_VALID_ATTRS (ATTR_MODE \
79         | ATTR_UID \
80         | ATTR_GID \
81         | ATTR_SIZE \
82         | ATTR_ATIME \
83         | ATTR_MTIME \
84         | ATTR_CTIME \
85         | ATTR_ATIME_SET \
86         | ATTR_MTIME_SET)
87
88 struct nfs4_opendata;
89 static int _nfs4_proc_open(struct nfs4_opendata *data);
90 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
91 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
92 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
93 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
95 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
96                             struct nfs_fattr *fattr, struct iattr *sattr,
97                             struct nfs4_state *state, struct nfs4_label *ilabel,
98                             struct nfs4_label *olabel);
99 #ifdef CONFIG_NFS_V4_1
100 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
101                 struct rpc_cred *);
102 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
103                 struct rpc_cred *);
104 #endif
105
106 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
107 static inline struct nfs4_label *
108 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
109         struct iattr *sattr, struct nfs4_label *label)
110 {
111         int err;
112
113         if (label == NULL)
114                 return NULL;
115
116         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
117                 return NULL;
118
119         err = security_dentry_init_security(dentry, sattr->ia_mode,
120                                 &dentry->d_name, (void **)&label->label, &label->len);
121         if (err == 0)
122                 return label;
123
124         return NULL;
125 }
126 static inline void
127 nfs4_label_release_security(struct nfs4_label *label)
128 {
129         if (label)
130                 security_release_secctx(label->label, label->len);
131 }
132 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
133 {
134         if (label)
135                 return server->attr_bitmask;
136
137         return server->attr_bitmask_nl;
138 }
139 #else
140 static inline struct nfs4_label *
141 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
142         struct iattr *sattr, struct nfs4_label *l)
143 { return NULL; }
144 static inline void
145 nfs4_label_release_security(struct nfs4_label *label)
146 { return; }
147 static inline u32 *
148 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
149 { return server->attr_bitmask; }
150 #endif
151
152 /* Prevent leaks of NFSv4 errors into userland */
153 static int nfs4_map_errors(int err)
154 {
155         if (err >= -1000)
156                 return err;
157         switch (err) {
158         case -NFS4ERR_RESOURCE:
159         case -NFS4ERR_LAYOUTTRYLATER:
160         case -NFS4ERR_RECALLCONFLICT:
161                 return -EREMOTEIO;
162         case -NFS4ERR_WRONGSEC:
163         case -NFS4ERR_WRONG_CRED:
164                 return -EPERM;
165         case -NFS4ERR_BADOWNER:
166         case -NFS4ERR_BADNAME:
167                 return -EINVAL;
168         case -NFS4ERR_SHARE_DENIED:
169                 return -EACCES;
170         case -NFS4ERR_MINOR_VERS_MISMATCH:
171                 return -EPROTONOSUPPORT;
172         case -NFS4ERR_FILE_OPEN:
173                 return -EBUSY;
174         default:
175                 dprintk("%s could not handle NFSv4 error %d\n",
176                                 __func__, -err);
177                 break;
178         }
179         return -EIO;
180 }
181
182 /*
183  * This is our standard bitmap for GETATTR requests.
184  */
185 const u32 nfs4_fattr_bitmap[3] = {
186         FATTR4_WORD0_TYPE
187         | FATTR4_WORD0_CHANGE
188         | FATTR4_WORD0_SIZE
189         | FATTR4_WORD0_FSID
190         | FATTR4_WORD0_FILEID,
191         FATTR4_WORD1_MODE
192         | FATTR4_WORD1_NUMLINKS
193         | FATTR4_WORD1_OWNER
194         | FATTR4_WORD1_OWNER_GROUP
195         | FATTR4_WORD1_RAWDEV
196         | FATTR4_WORD1_SPACE_USED
197         | FATTR4_WORD1_TIME_ACCESS
198         | FATTR4_WORD1_TIME_METADATA
199         | FATTR4_WORD1_TIME_MODIFY
200         | FATTR4_WORD1_MOUNTED_ON_FILEID,
201 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
202         FATTR4_WORD2_SECURITY_LABEL
203 #endif
204 };
205
206 static const u32 nfs4_pnfs_open_bitmap[3] = {
207         FATTR4_WORD0_TYPE
208         | FATTR4_WORD0_CHANGE
209         | FATTR4_WORD0_SIZE
210         | FATTR4_WORD0_FSID
211         | FATTR4_WORD0_FILEID,
212         FATTR4_WORD1_MODE
213         | FATTR4_WORD1_NUMLINKS
214         | FATTR4_WORD1_OWNER
215         | FATTR4_WORD1_OWNER_GROUP
216         | FATTR4_WORD1_RAWDEV
217         | FATTR4_WORD1_SPACE_USED
218         | FATTR4_WORD1_TIME_ACCESS
219         | FATTR4_WORD1_TIME_METADATA
220         | FATTR4_WORD1_TIME_MODIFY,
221         FATTR4_WORD2_MDSTHRESHOLD
222 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
223         | FATTR4_WORD2_SECURITY_LABEL
224 #endif
225 };
226
227 static const u32 nfs4_open_noattr_bitmap[3] = {
228         FATTR4_WORD0_TYPE
229         | FATTR4_WORD0_CHANGE
230         | FATTR4_WORD0_FILEID,
231 };
232
233 const u32 nfs4_statfs_bitmap[3] = {
234         FATTR4_WORD0_FILES_AVAIL
235         | FATTR4_WORD0_FILES_FREE
236         | FATTR4_WORD0_FILES_TOTAL,
237         FATTR4_WORD1_SPACE_AVAIL
238         | FATTR4_WORD1_SPACE_FREE
239         | FATTR4_WORD1_SPACE_TOTAL
240 };
241
242 const u32 nfs4_pathconf_bitmap[3] = {
243         FATTR4_WORD0_MAXLINK
244         | FATTR4_WORD0_MAXNAME,
245         0
246 };
247
248 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
249                         | FATTR4_WORD0_MAXREAD
250                         | FATTR4_WORD0_MAXWRITE
251                         | FATTR4_WORD0_LEASE_TIME,
252                         FATTR4_WORD1_TIME_DELTA
253                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
254                         FATTR4_WORD2_LAYOUT_BLKSIZE
255                         | FATTR4_WORD2_CLONE_BLKSIZE
256 };
257
258 const u32 nfs4_fs_locations_bitmap[3] = {
259         FATTR4_WORD0_TYPE
260         | FATTR4_WORD0_CHANGE
261         | FATTR4_WORD0_SIZE
262         | FATTR4_WORD0_FSID
263         | FATTR4_WORD0_FILEID
264         | FATTR4_WORD0_FS_LOCATIONS,
265         FATTR4_WORD1_MODE
266         | FATTR4_WORD1_NUMLINKS
267         | FATTR4_WORD1_OWNER
268         | FATTR4_WORD1_OWNER_GROUP
269         | FATTR4_WORD1_RAWDEV
270         | FATTR4_WORD1_SPACE_USED
271         | FATTR4_WORD1_TIME_ACCESS
272         | FATTR4_WORD1_TIME_METADATA
273         | FATTR4_WORD1_TIME_MODIFY
274         | FATTR4_WORD1_MOUNTED_ON_FILEID,
275 };
276
277 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
278                 struct nfs4_readdir_arg *readdir)
279 {
280         __be32 *start, *p;
281
282         if (cookie > 2) {
283                 readdir->cookie = cookie;
284                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
285                 return;
286         }
287
288         readdir->cookie = 0;
289         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
290         if (cookie == 2)
291                 return;
292         
293         /*
294          * NFSv4 servers do not return entries for '.' and '..'
295          * Therefore, we fake these entries here.  We let '.'
296          * have cookie 0 and '..' have cookie 1.  Note that
297          * when talking to the server, we always send cookie 0
298          * instead of 1 or 2.
299          */
300         start = p = kmap_atomic(*readdir->pages);
301         
302         if (cookie == 0) {
303                 *p++ = xdr_one;                                  /* next */
304                 *p++ = xdr_zero;                   /* cookie, first word */
305                 *p++ = xdr_one;                   /* cookie, second word */
306                 *p++ = xdr_one;                             /* entry len */
307                 memcpy(p, ".\0\0\0", 4);                        /* entry */
308                 p++;
309                 *p++ = xdr_one;                         /* bitmap length */
310                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
311                 *p++ = htonl(8);              /* attribute buffer length */
312                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
313         }
314         
315         *p++ = xdr_one;                                  /* next */
316         *p++ = xdr_zero;                   /* cookie, first word */
317         *p++ = xdr_two;                   /* cookie, second word */
318         *p++ = xdr_two;                             /* entry len */
319         memcpy(p, "..\0\0", 4);                         /* entry */
320         p++;
321         *p++ = xdr_one;                         /* bitmap length */
322         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
323         *p++ = htonl(8);              /* attribute buffer length */
324         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
325
326         readdir->pgbase = (char *)p - (char *)start;
327         readdir->count -= readdir->pgbase;
328         kunmap_atomic(start);
329 }
330
331 static long nfs4_update_delay(long *timeout)
332 {
333         long ret;
334         if (!timeout)
335                 return NFS4_POLL_RETRY_MAX;
336         if (*timeout <= 0)
337                 *timeout = NFS4_POLL_RETRY_MIN;
338         if (*timeout > NFS4_POLL_RETRY_MAX)
339                 *timeout = NFS4_POLL_RETRY_MAX;
340         ret = *timeout;
341         *timeout <<= 1;
342         return ret;
343 }
344
345 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
346 {
347         int res = 0;
348
349         might_sleep();
350
351         freezable_schedule_timeout_killable_unsafe(
352                 nfs4_update_delay(timeout));
353         if (fatal_signal_pending(current))
354                 res = -ERESTARTSYS;
355         return res;
356 }
357
358 /* This is the error handling routine for processes that are allowed
359  * to sleep.
360  */
361 static int nfs4_do_handle_exception(struct nfs_server *server,
362                 int errorcode, struct nfs4_exception *exception)
363 {
364         struct nfs_client *clp = server->nfs_client;
365         struct nfs4_state *state = exception->state;
366         struct inode *inode = exception->inode;
367         int ret = errorcode;
368
369         exception->delay = 0;
370         exception->recovering = 0;
371         exception->retry = 0;
372         switch(errorcode) {
373                 case 0:
374                         return 0;
375                 case -NFS4ERR_OPENMODE:
376                 case -NFS4ERR_DELEG_REVOKED:
377                 case -NFS4ERR_ADMIN_REVOKED:
378                 case -NFS4ERR_BAD_STATEID:
379                         if (inode && nfs_async_inode_return_delegation(inode,
380                                                 NULL) == 0)
381                                 goto wait_on_recovery;
382                         if (state == NULL)
383                                 break;
384                         ret = nfs4_schedule_stateid_recovery(server, state);
385                         if (ret < 0)
386                                 break;
387                         goto wait_on_recovery;
388                 case -NFS4ERR_EXPIRED:
389                         if (state != NULL) {
390                                 ret = nfs4_schedule_stateid_recovery(server, state);
391                                 if (ret < 0)
392                                         break;
393                         }
394                 case -NFS4ERR_STALE_STATEID:
395                 case -NFS4ERR_STALE_CLIENTID:
396                         nfs4_schedule_lease_recovery(clp);
397                         goto wait_on_recovery;
398                 case -NFS4ERR_MOVED:
399                         ret = nfs4_schedule_migration_recovery(server);
400                         if (ret < 0)
401                                 break;
402                         goto wait_on_recovery;
403                 case -NFS4ERR_LEASE_MOVED:
404                         nfs4_schedule_lease_moved_recovery(clp);
405                         goto wait_on_recovery;
406 #if defined(CONFIG_NFS_V4_1)
407                 case -NFS4ERR_BADSESSION:
408                 case -NFS4ERR_BADSLOT:
409                 case -NFS4ERR_BAD_HIGH_SLOT:
410                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
411                 case -NFS4ERR_DEADSESSION:
412                 case -NFS4ERR_SEQ_FALSE_RETRY:
413                 case -NFS4ERR_SEQ_MISORDERED:
414                         dprintk("%s ERROR: %d Reset session\n", __func__,
415                                 errorcode);
416                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
417                         goto wait_on_recovery;
418 #endif /* defined(CONFIG_NFS_V4_1) */
419                 case -NFS4ERR_FILE_OPEN:
420                         if (exception->timeout > HZ) {
421                                 /* We have retried a decent amount, time to
422                                  * fail
423                                  */
424                                 ret = -EBUSY;
425                                 break;
426                         }
427                 case -NFS4ERR_DELAY:
428                         nfs_inc_server_stats(server, NFSIOS_DELAY);
429                 case -NFS4ERR_GRACE:
430                         exception->delay = 1;
431                         return 0;
432
433                 case -NFS4ERR_RETRY_UNCACHED_REP:
434                 case -NFS4ERR_OLD_STATEID:
435                         exception->retry = 1;
436                         break;
437                 case -NFS4ERR_BADOWNER:
438                         /* The following works around a Linux server bug! */
439                 case -NFS4ERR_BADNAME:
440                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
441                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
442                                 exception->retry = 1;
443                                 printk(KERN_WARNING "NFS: v4 server %s "
444                                                 "does not accept raw "
445                                                 "uid/gids. "
446                                                 "Reenabling the idmapper.\n",
447                                                 server->nfs_client->cl_hostname);
448                         }
449         }
450         /* We failed to handle the error */
451         return nfs4_map_errors(ret);
452 wait_on_recovery:
453         exception->recovering = 1;
454         return 0;
455 }
456
457 /* This is the error handling routine for processes that are allowed
458  * to sleep.
459  */
460 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
461 {
462         struct nfs_client *clp = server->nfs_client;
463         int ret;
464
465         ret = nfs4_do_handle_exception(server, errorcode, exception);
466         if (exception->delay) {
467                 ret = nfs4_delay(server->client, &exception->timeout);
468                 goto out_retry;
469         }
470         if (exception->recovering) {
471                 ret = nfs4_wait_clnt_recover(clp);
472                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
473                         return -EIO;
474                 goto out_retry;
475         }
476         return ret;
477 out_retry:
478         if (ret == 0)
479                 exception->retry = 1;
480         return ret;
481 }
482
483 static int
484 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
485                 int errorcode, struct nfs4_exception *exception)
486 {
487         struct nfs_client *clp = server->nfs_client;
488         int ret;
489
490         ret = nfs4_do_handle_exception(server, errorcode, exception);
491         if (exception->delay) {
492                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
493                 goto out_retry;
494         }
495         if (exception->recovering) {
496                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
497                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
498                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
499                 goto out_retry;
500         }
501         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
502                 ret = -EIO;
503         return ret;
504 out_retry:
505         if (ret == 0)
506                 exception->retry = 1;
507         return ret;
508 }
509
510 static int
511 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
512                         struct nfs4_state *state, long *timeout)
513 {
514         struct nfs4_exception exception = {
515                 .state = state,
516         };
517
518         if (task->tk_status >= 0)
519                 return 0;
520         if (timeout)
521                 exception.timeout = *timeout;
522         task->tk_status = nfs4_async_handle_exception(task, server,
523                         task->tk_status,
524                         &exception);
525         if (exception.delay && timeout)
526                 *timeout = exception.timeout;
527         if (exception.retry)
528                 return -EAGAIN;
529         return 0;
530 }
531
532 /*
533  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
534  * or 'false' otherwise.
535  */
536 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
537 {
538         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
539
540         if (flavor == RPC_AUTH_GSS_KRB5I ||
541             flavor == RPC_AUTH_GSS_KRB5P)
542                 return true;
543
544         return false;
545 }
546
547 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
548 {
549         spin_lock(&clp->cl_lock);
550         if (time_before(clp->cl_last_renewal,timestamp))
551                 clp->cl_last_renewal = timestamp;
552         spin_unlock(&clp->cl_lock);
553 }
554
555 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
556 {
557         struct nfs_client *clp = server->nfs_client;
558
559         if (!nfs4_has_session(clp))
560                 do_renew_lease(clp, timestamp);
561 }
562
563 struct nfs4_call_sync_data {
564         const struct nfs_server *seq_server;
565         struct nfs4_sequence_args *seq_args;
566         struct nfs4_sequence_res *seq_res;
567 };
568
569 void nfs4_init_sequence(struct nfs4_sequence_args *args,
570                         struct nfs4_sequence_res *res, int cache_reply)
571 {
572         args->sa_slot = NULL;
573         args->sa_cache_this = cache_reply;
574         args->sa_privileged = 0;
575
576         res->sr_slot = NULL;
577 }
578
579 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
580 {
581         args->sa_privileged = 1;
582 }
583
584 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
585                          struct nfs4_sequence_args *args,
586                          struct nfs4_sequence_res *res,
587                          struct rpc_task *task)
588 {
589         struct nfs4_slot *slot;
590
591         /* slot already allocated? */
592         if (res->sr_slot != NULL)
593                 goto out_start;
594
595         spin_lock(&tbl->slot_tbl_lock);
596         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
597                 goto out_sleep;
598
599         slot = nfs4_alloc_slot(tbl);
600         if (IS_ERR(slot)) {
601                 if (slot == ERR_PTR(-ENOMEM))
602                         task->tk_timeout = HZ >> 2;
603                 goto out_sleep;
604         }
605         spin_unlock(&tbl->slot_tbl_lock);
606
607         args->sa_slot = slot;
608         res->sr_slot = slot;
609
610 out_start:
611         rpc_call_start(task);
612         return 0;
613
614 out_sleep:
615         if (args->sa_privileged)
616                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
617                                 NULL, RPC_PRIORITY_PRIVILEGED);
618         else
619                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
620         spin_unlock(&tbl->slot_tbl_lock);
621         return -EAGAIN;
622 }
623 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
624
625 static int nfs40_sequence_done(struct rpc_task *task,
626                                struct nfs4_sequence_res *res)
627 {
628         struct nfs4_slot *slot = res->sr_slot;
629         struct nfs4_slot_table *tbl;
630
631         if (slot == NULL)
632                 goto out;
633
634         tbl = slot->table;
635         spin_lock(&tbl->slot_tbl_lock);
636         if (!nfs41_wake_and_assign_slot(tbl, slot))
637                 nfs4_free_slot(tbl, slot);
638         spin_unlock(&tbl->slot_tbl_lock);
639
640         res->sr_slot = NULL;
641 out:
642         return 1;
643 }
644
645 #if defined(CONFIG_NFS_V4_1)
646
647 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
648 {
649         struct nfs4_session *session;
650         struct nfs4_slot_table *tbl;
651         struct nfs4_slot *slot = res->sr_slot;
652         bool send_new_highest_used_slotid = false;
653
654         tbl = slot->table;
655         session = tbl->session;
656
657         spin_lock(&tbl->slot_tbl_lock);
658         /* Be nice to the server: try to ensure that the last transmitted
659          * value for highest_user_slotid <= target_highest_slotid
660          */
661         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
662                 send_new_highest_used_slotid = true;
663
664         if (nfs41_wake_and_assign_slot(tbl, slot)) {
665                 send_new_highest_used_slotid = false;
666                 goto out_unlock;
667         }
668         nfs4_free_slot(tbl, slot);
669
670         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
671                 send_new_highest_used_slotid = false;
672 out_unlock:
673         spin_unlock(&tbl->slot_tbl_lock);
674         res->sr_slot = NULL;
675         if (send_new_highest_used_slotid)
676                 nfs41_notify_server(session->clp);
677 }
678
679 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
680 {
681         struct nfs4_session *session;
682         struct nfs4_slot *slot = res->sr_slot;
683         struct nfs_client *clp;
684         bool interrupted = false;
685         int ret = 1;
686
687         if (slot == NULL)
688                 goto out_noaction;
689         /* don't increment the sequence number if the task wasn't sent */
690         if (!RPC_WAS_SENT(task))
691                 goto out;
692
693         session = slot->table->session;
694
695         if (slot->interrupted) {
696                 slot->interrupted = 0;
697                 interrupted = true;
698         }
699
700         trace_nfs4_sequence_done(session, res);
701         /* Check the SEQUENCE operation status */
702         switch (res->sr_status) {
703         case 0:
704                 /* Update the slot's sequence and clientid lease timer */
705                 ++slot->seq_nr;
706                 clp = session->clp;
707                 do_renew_lease(clp, res->sr_timestamp);
708                 /* Check sequence flags */
709                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
710                 nfs41_update_target_slotid(slot->table, slot, res);
711                 break;
712         case 1:
713                 /*
714                  * sr_status remains 1 if an RPC level error occurred.
715                  * The server may or may not have processed the sequence
716                  * operation..
717                  * Mark the slot as having hosted an interrupted RPC call.
718                  */
719                 slot->interrupted = 1;
720                 goto out;
721         case -NFS4ERR_DELAY:
722                 /* The server detected a resend of the RPC call and
723                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
724                  * of RFC5661.
725                  */
726                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
727                         __func__,
728                         slot->slot_nr,
729                         slot->seq_nr);
730                 goto out_retry;
731         case -NFS4ERR_BADSLOT:
732                 /*
733                  * The slot id we used was probably retired. Try again
734                  * using a different slot id.
735                  */
736                 goto retry_nowait;
737         case -NFS4ERR_SEQ_MISORDERED:
738                 /*
739                  * Was the last operation on this sequence interrupted?
740                  * If so, retry after bumping the sequence number.
741                  */
742                 if (interrupted) {
743                         ++slot->seq_nr;
744                         goto retry_nowait;
745                 }
746                 /*
747                  * Could this slot have been previously retired?
748                  * If so, then the server may be expecting seq_nr = 1!
749                  */
750                 if (slot->seq_nr != 1) {
751                         slot->seq_nr = 1;
752                         goto retry_nowait;
753                 }
754                 break;
755         case -NFS4ERR_SEQ_FALSE_RETRY:
756                 ++slot->seq_nr;
757                 goto retry_nowait;
758         default:
759                 /* Just update the slot sequence no. */
760                 ++slot->seq_nr;
761         }
762 out:
763         /* The session may be reset by one of the error handlers. */
764         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
765         nfs41_sequence_free_slot(res);
766 out_noaction:
767         return ret;
768 retry_nowait:
769         if (rpc_restart_call_prepare(task)) {
770                 task->tk_status = 0;
771                 ret = 0;
772         }
773         goto out;
774 out_retry:
775         if (!rpc_restart_call(task))
776                 goto out;
777         rpc_delay(task, NFS4_POLL_RETRY_MAX);
778         return 0;
779 }
780 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
781
782 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
783 {
784         if (res->sr_slot == NULL)
785                 return 1;
786         if (!res->sr_slot->table->session)
787                 return nfs40_sequence_done(task, res);
788         return nfs41_sequence_done(task, res);
789 }
790 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
791
792 int nfs41_setup_sequence(struct nfs4_session *session,
793                                 struct nfs4_sequence_args *args,
794                                 struct nfs4_sequence_res *res,
795                                 struct rpc_task *task)
796 {
797         struct nfs4_slot *slot;
798         struct nfs4_slot_table *tbl;
799
800         dprintk("--> %s\n", __func__);
801         /* slot already allocated? */
802         if (res->sr_slot != NULL)
803                 goto out_success;
804
805         tbl = &session->fc_slot_table;
806
807         task->tk_timeout = 0;
808
809         spin_lock(&tbl->slot_tbl_lock);
810         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
811             !args->sa_privileged) {
812                 /* The state manager will wait until the slot table is empty */
813                 dprintk("%s session is draining\n", __func__);
814                 goto out_sleep;
815         }
816
817         slot = nfs4_alloc_slot(tbl);
818         if (IS_ERR(slot)) {
819                 /* If out of memory, try again in 1/4 second */
820                 if (slot == ERR_PTR(-ENOMEM))
821                         task->tk_timeout = HZ >> 2;
822                 dprintk("<-- %s: no free slots\n", __func__);
823                 goto out_sleep;
824         }
825         spin_unlock(&tbl->slot_tbl_lock);
826
827         args->sa_slot = slot;
828
829         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
830                         slot->slot_nr, slot->seq_nr);
831
832         res->sr_slot = slot;
833         res->sr_timestamp = jiffies;
834         res->sr_status_flags = 0;
835         /*
836          * sr_status is only set in decode_sequence, and so will remain
837          * set to 1 if an rpc level failure occurs.
838          */
839         res->sr_status = 1;
840         trace_nfs4_setup_sequence(session, args);
841 out_success:
842         rpc_call_start(task);
843         return 0;
844 out_sleep:
845         /* Privileged tasks are queued with top priority */
846         if (args->sa_privileged)
847                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
848                                 NULL, RPC_PRIORITY_PRIVILEGED);
849         else
850                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
851         spin_unlock(&tbl->slot_tbl_lock);
852         return -EAGAIN;
853 }
854 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
855
856 static int nfs4_setup_sequence(const struct nfs_server *server,
857                                struct nfs4_sequence_args *args,
858                                struct nfs4_sequence_res *res,
859                                struct rpc_task *task)
860 {
861         struct nfs4_session *session = nfs4_get_session(server);
862         int ret = 0;
863
864         if (!session)
865                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
866                                             args, res, task);
867
868         dprintk("--> %s clp %p session %p sr_slot %u\n",
869                 __func__, session->clp, session, res->sr_slot ?
870                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
871
872         ret = nfs41_setup_sequence(session, args, res, task);
873
874         dprintk("<-- %s status=%d\n", __func__, ret);
875         return ret;
876 }
877
878 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
879 {
880         struct nfs4_call_sync_data *data = calldata;
881         struct nfs4_session *session = nfs4_get_session(data->seq_server);
882
883         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
884
885         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
886 }
887
888 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
889 {
890         struct nfs4_call_sync_data *data = calldata;
891
892         nfs41_sequence_done(task, data->seq_res);
893 }
894
895 static const struct rpc_call_ops nfs41_call_sync_ops = {
896         .rpc_call_prepare = nfs41_call_sync_prepare,
897         .rpc_call_done = nfs41_call_sync_done,
898 };
899
900 #else   /* !CONFIG_NFS_V4_1 */
901
902 static int nfs4_setup_sequence(const struct nfs_server *server,
903                                struct nfs4_sequence_args *args,
904                                struct nfs4_sequence_res *res,
905                                struct rpc_task *task)
906 {
907         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
908                                     args, res, task);
909 }
910
911 int nfs4_sequence_done(struct rpc_task *task,
912                        struct nfs4_sequence_res *res)
913 {
914         return nfs40_sequence_done(task, res);
915 }
916 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
917
918 #endif  /* !CONFIG_NFS_V4_1 */
919
920 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
921 {
922         struct nfs4_call_sync_data *data = calldata;
923         nfs4_setup_sequence(data->seq_server,
924                                 data->seq_args, data->seq_res, task);
925 }
926
927 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
928 {
929         struct nfs4_call_sync_data *data = calldata;
930         nfs4_sequence_done(task, data->seq_res);
931 }
932
933 static const struct rpc_call_ops nfs40_call_sync_ops = {
934         .rpc_call_prepare = nfs40_call_sync_prepare,
935         .rpc_call_done = nfs40_call_sync_done,
936 };
937
938 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
939                                    struct nfs_server *server,
940                                    struct rpc_message *msg,
941                                    struct nfs4_sequence_args *args,
942                                    struct nfs4_sequence_res *res)
943 {
944         int ret;
945         struct rpc_task *task;
946         struct nfs_client *clp = server->nfs_client;
947         struct nfs4_call_sync_data data = {
948                 .seq_server = server,
949                 .seq_args = args,
950                 .seq_res = res,
951         };
952         struct rpc_task_setup task_setup = {
953                 .rpc_client = clnt,
954                 .rpc_message = msg,
955                 .callback_ops = clp->cl_mvops->call_sync_ops,
956                 .callback_data = &data
957         };
958
959         task = rpc_run_task(&task_setup);
960         if (IS_ERR(task))
961                 ret = PTR_ERR(task);
962         else {
963                 ret = task->tk_status;
964                 rpc_put_task(task);
965         }
966         return ret;
967 }
968
969 int nfs4_call_sync(struct rpc_clnt *clnt,
970                    struct nfs_server *server,
971                    struct rpc_message *msg,
972                    struct nfs4_sequence_args *args,
973                    struct nfs4_sequence_res *res,
974                    int cache_reply)
975 {
976         nfs4_init_sequence(args, res, cache_reply);
977         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
978 }
979
980 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
981 {
982         struct nfs_inode *nfsi = NFS_I(dir);
983
984         spin_lock(&dir->i_lock);
985         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
986         if (!cinfo->atomic || cinfo->before != dir->i_version)
987                 nfs_force_lookup_revalidate(dir);
988         dir->i_version = cinfo->after;
989         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
990         nfs_fscache_invalidate(dir);
991         spin_unlock(&dir->i_lock);
992 }
993
994 struct nfs4_opendata {
995         struct kref kref;
996         struct nfs_openargs o_arg;
997         struct nfs_openres o_res;
998         struct nfs_open_confirmargs c_arg;
999         struct nfs_open_confirmres c_res;
1000         struct nfs4_string owner_name;
1001         struct nfs4_string group_name;
1002         struct nfs4_label *a_label;
1003         struct nfs_fattr f_attr;
1004         struct nfs4_label *f_label;
1005         struct dentry *dir;
1006         struct dentry *dentry;
1007         struct nfs4_state_owner *owner;
1008         struct nfs4_state *state;
1009         struct iattr attrs;
1010         unsigned long timestamp;
1011         unsigned int rpc_done : 1;
1012         unsigned int file_created : 1;
1013         unsigned int is_recover : 1;
1014         int rpc_status;
1015         int cancelled;
1016 };
1017
1018 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1019                 int err, struct nfs4_exception *exception)
1020 {
1021         if (err != -EINVAL)
1022                 return false;
1023         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1024                 return false;
1025         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1026         exception->retry = 1;
1027         return true;
1028 }
1029
1030 static u32
1031 nfs4_map_atomic_open_share(struct nfs_server *server,
1032                 fmode_t fmode, int openflags)
1033 {
1034         u32 res = 0;
1035
1036         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1037         case FMODE_READ:
1038                 res = NFS4_SHARE_ACCESS_READ;
1039                 break;
1040         case FMODE_WRITE:
1041                 res = NFS4_SHARE_ACCESS_WRITE;
1042                 break;
1043         case FMODE_READ|FMODE_WRITE:
1044                 res = NFS4_SHARE_ACCESS_BOTH;
1045         }
1046         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1047                 goto out;
1048         /* Want no delegation if we're using O_DIRECT */
1049         if (openflags & O_DIRECT)
1050                 res |= NFS4_SHARE_WANT_NO_DELEG;
1051 out:
1052         return res;
1053 }
1054
1055 static enum open_claim_type4
1056 nfs4_map_atomic_open_claim(struct nfs_server *server,
1057                 enum open_claim_type4 claim)
1058 {
1059         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1060                 return claim;
1061         switch (claim) {
1062         default:
1063                 return claim;
1064         case NFS4_OPEN_CLAIM_FH:
1065                 return NFS4_OPEN_CLAIM_NULL;
1066         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1067                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1068         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1069                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1070         }
1071 }
1072
1073 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1074 {
1075         p->o_res.f_attr = &p->f_attr;
1076         p->o_res.f_label = p->f_label;
1077         p->o_res.seqid = p->o_arg.seqid;
1078         p->c_res.seqid = p->c_arg.seqid;
1079         p->o_res.server = p->o_arg.server;
1080         p->o_res.access_request = p->o_arg.access;
1081         nfs_fattr_init(&p->f_attr);
1082         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1083 }
1084
1085 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1086                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1087                 const struct iattr *attrs,
1088                 struct nfs4_label *label,
1089                 enum open_claim_type4 claim,
1090                 gfp_t gfp_mask)
1091 {
1092         struct dentry *parent = dget_parent(dentry);
1093         struct inode *dir = d_inode(parent);
1094         struct nfs_server *server = NFS_SERVER(dir);
1095         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1096         struct nfs4_opendata *p;
1097
1098         p = kzalloc(sizeof(*p), gfp_mask);
1099         if (p == NULL)
1100                 goto err;
1101
1102         p->f_label = nfs4_label_alloc(server, gfp_mask);
1103         if (IS_ERR(p->f_label))
1104                 goto err_free_p;
1105
1106         p->a_label = nfs4_label_alloc(server, gfp_mask);
1107         if (IS_ERR(p->a_label))
1108                 goto err_free_f;
1109
1110         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1111         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1112         if (IS_ERR(p->o_arg.seqid))
1113                 goto err_free_label;
1114         nfs_sb_active(dentry->d_sb);
1115         p->dentry = dget(dentry);
1116         p->dir = parent;
1117         p->owner = sp;
1118         atomic_inc(&sp->so_count);
1119         p->o_arg.open_flags = flags;
1120         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1121         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1122                         fmode, flags);
1123         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1124          * will return permission denied for all bits until close */
1125         if (!(flags & O_EXCL)) {
1126                 /* ask server to check for all possible rights as results
1127                  * are cached */
1128                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1129                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1130         }
1131         p->o_arg.clientid = server->nfs_client->cl_clientid;
1132         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1133         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1134         p->o_arg.name = &dentry->d_name;
1135         p->o_arg.server = server;
1136         p->o_arg.bitmask = nfs4_bitmask(server, label);
1137         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1138         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1139         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1140         switch (p->o_arg.claim) {
1141         case NFS4_OPEN_CLAIM_NULL:
1142         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1143         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1144                 p->o_arg.fh = NFS_FH(dir);
1145                 break;
1146         case NFS4_OPEN_CLAIM_PREVIOUS:
1147         case NFS4_OPEN_CLAIM_FH:
1148         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1149         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1150                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1151         }
1152         if (attrs != NULL && attrs->ia_valid != 0) {
1153                 __u32 verf[2];
1154
1155                 p->o_arg.u.attrs = &p->attrs;
1156                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1157
1158                 verf[0] = jiffies;
1159                 verf[1] = current->pid;
1160                 memcpy(p->o_arg.u.verifier.data, verf,
1161                                 sizeof(p->o_arg.u.verifier.data));
1162         }
1163         p->c_arg.fh = &p->o_res.fh;
1164         p->c_arg.stateid = &p->o_res.stateid;
1165         p->c_arg.seqid = p->o_arg.seqid;
1166         nfs4_init_opendata_res(p);
1167         kref_init(&p->kref);
1168         return p;
1169
1170 err_free_label:
1171         nfs4_label_free(p->a_label);
1172 err_free_f:
1173         nfs4_label_free(p->f_label);
1174 err_free_p:
1175         kfree(p);
1176 err:
1177         dput(parent);
1178         return NULL;
1179 }
1180
1181 static void nfs4_opendata_free(struct kref *kref)
1182 {
1183         struct nfs4_opendata *p = container_of(kref,
1184                         struct nfs4_opendata, kref);
1185         struct super_block *sb = p->dentry->d_sb;
1186
1187         nfs_free_seqid(p->o_arg.seqid);
1188         if (p->state != NULL)
1189                 nfs4_put_open_state(p->state);
1190         nfs4_put_state_owner(p->owner);
1191
1192         nfs4_label_free(p->a_label);
1193         nfs4_label_free(p->f_label);
1194
1195         dput(p->dir);
1196         dput(p->dentry);
1197         nfs_sb_deactive(sb);
1198         nfs_fattr_free_names(&p->f_attr);
1199         kfree(p->f_attr.mdsthreshold);
1200         kfree(p);
1201 }
1202
1203 static void nfs4_opendata_put(struct nfs4_opendata *p)
1204 {
1205         if (p != NULL)
1206                 kref_put(&p->kref, nfs4_opendata_free);
1207 }
1208
1209 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1210 {
1211         int ret;
1212
1213         ret = rpc_wait_for_completion_task(task);
1214         return ret;
1215 }
1216
1217 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1218                 fmode_t fmode)
1219 {
1220         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1221         case FMODE_READ|FMODE_WRITE:
1222                 return state->n_rdwr != 0;
1223         case FMODE_WRITE:
1224                 return state->n_wronly != 0;
1225         case FMODE_READ:
1226                 return state->n_rdonly != 0;
1227         }
1228         WARN_ON_ONCE(1);
1229         return false;
1230 }
1231
1232 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1233 {
1234         int ret = 0;
1235
1236         if (open_mode & (O_EXCL|O_TRUNC))
1237                 goto out;
1238         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1239                 case FMODE_READ:
1240                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1241                                 && state->n_rdonly != 0;
1242                         break;
1243                 case FMODE_WRITE:
1244                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1245                                 && state->n_wronly != 0;
1246                         break;
1247                 case FMODE_READ|FMODE_WRITE:
1248                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1249                                 && state->n_rdwr != 0;
1250         }
1251 out:
1252         return ret;
1253 }
1254
1255 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1256                 enum open_claim_type4 claim)
1257 {
1258         if (delegation == NULL)
1259                 return 0;
1260         if ((delegation->type & fmode) != fmode)
1261                 return 0;
1262         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1263                 return 0;
1264         switch (claim) {
1265         case NFS4_OPEN_CLAIM_NULL:
1266         case NFS4_OPEN_CLAIM_FH:
1267                 break;
1268         case NFS4_OPEN_CLAIM_PREVIOUS:
1269                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1270                         break;
1271         default:
1272                 return 0;
1273         }
1274         nfs_mark_delegation_referenced(delegation);
1275         return 1;
1276 }
1277
1278 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1279 {
1280         switch (fmode) {
1281                 case FMODE_WRITE:
1282                         state->n_wronly++;
1283                         break;
1284                 case FMODE_READ:
1285                         state->n_rdonly++;
1286                         break;
1287                 case FMODE_READ|FMODE_WRITE:
1288                         state->n_rdwr++;
1289         }
1290         nfs4_state_set_mode_locked(state, state->state | fmode);
1291 }
1292
1293 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1294 {
1295         struct nfs_client *clp = state->owner->so_server->nfs_client;
1296         bool need_recover = false;
1297
1298         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1299                 need_recover = true;
1300         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1301                 need_recover = true;
1302         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1303                 need_recover = true;
1304         if (need_recover)
1305                 nfs4_state_mark_reclaim_nograce(clp, state);
1306 }
1307
1308 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1309                 nfs4_stateid *stateid)
1310 {
1311         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1312                 return true;
1313         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1314                 nfs_test_and_clear_all_open_stateid(state);
1315                 return true;
1316         }
1317         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1318                 return true;
1319         return false;
1320 }
1321
1322 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1323 {
1324         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1325                 return;
1326         if (state->n_wronly)
1327                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1328         if (state->n_rdonly)
1329                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1330         if (state->n_rdwr)
1331                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1332         set_bit(NFS_OPEN_STATE, &state->flags);
1333 }
1334
1335 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1336                 nfs4_stateid *arg_stateid,
1337                 nfs4_stateid *stateid, fmode_t fmode)
1338 {
1339         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1340         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1341         case FMODE_WRITE:
1342                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1343                 break;
1344         case FMODE_READ:
1345                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1346                 break;
1347         case 0:
1348                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1349                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1350                 clear_bit(NFS_OPEN_STATE, &state->flags);
1351         }
1352         if (stateid == NULL)
1353                 return;
1354         /* Handle races with OPEN */
1355         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1356             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1357             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1358                 nfs_resync_open_stateid_locked(state);
1359                 return;
1360         }
1361         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1362                 nfs4_stateid_copy(&state->stateid, stateid);
1363         nfs4_stateid_copy(&state->open_stateid, stateid);
1364 }
1365
1366 static void nfs_clear_open_stateid(struct nfs4_state *state,
1367         nfs4_stateid *arg_stateid,
1368         nfs4_stateid *stateid, fmode_t fmode)
1369 {
1370         write_seqlock(&state->seqlock);
1371         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1372         write_sequnlock(&state->seqlock);
1373         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1374                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1375 }
1376
1377 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1378 {
1379         switch (fmode) {
1380                 case FMODE_READ:
1381                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1382                         break;
1383                 case FMODE_WRITE:
1384                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1385                         break;
1386                 case FMODE_READ|FMODE_WRITE:
1387                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1388         }
1389         if (!nfs_need_update_open_stateid(state, stateid))
1390                 return;
1391         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1392                 nfs4_stateid_copy(&state->stateid, stateid);
1393         nfs4_stateid_copy(&state->open_stateid, stateid);
1394 }
1395
1396 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1397 {
1398         /*
1399          * Protect the call to nfs4_state_set_mode_locked and
1400          * serialise the stateid update
1401          */
1402         spin_lock(&state->owner->so_lock);
1403         write_seqlock(&state->seqlock);
1404         if (deleg_stateid != NULL) {
1405                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1406                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1407         }
1408         if (open_stateid != NULL)
1409                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1410         write_sequnlock(&state->seqlock);
1411         update_open_stateflags(state, fmode);
1412         spin_unlock(&state->owner->so_lock);
1413 }
1414
1415 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1416 {
1417         struct nfs_inode *nfsi = NFS_I(state->inode);
1418         struct nfs_delegation *deleg_cur;
1419         int ret = 0;
1420
1421         fmode &= (FMODE_READ|FMODE_WRITE);
1422
1423         rcu_read_lock();
1424         deleg_cur = rcu_dereference(nfsi->delegation);
1425         if (deleg_cur == NULL)
1426                 goto no_delegation;
1427
1428         spin_lock(&deleg_cur->lock);
1429         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1430            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1431             (deleg_cur->type & fmode) != fmode)
1432                 goto no_delegation_unlock;
1433
1434         if (delegation == NULL)
1435                 delegation = &deleg_cur->stateid;
1436         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1437                 goto no_delegation_unlock;
1438
1439         nfs_mark_delegation_referenced(deleg_cur);
1440         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1441         ret = 1;
1442 no_delegation_unlock:
1443         spin_unlock(&deleg_cur->lock);
1444 no_delegation:
1445         rcu_read_unlock();
1446
1447         if (!ret && open_stateid != NULL) {
1448                 __update_open_stateid(state, open_stateid, NULL, fmode);
1449                 ret = 1;
1450         }
1451         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1452                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1453
1454         return ret;
1455 }
1456
1457 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1458                 const nfs4_stateid *stateid)
1459 {
1460         struct nfs4_state *state = lsp->ls_state;
1461         bool ret = false;
1462
1463         spin_lock(&state->state_lock);
1464         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1465                 goto out_noupdate;
1466         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1467                 goto out_noupdate;
1468         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1469         ret = true;
1470 out_noupdate:
1471         spin_unlock(&state->state_lock);
1472         return ret;
1473 }
1474
1475 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1476 {
1477         struct nfs_delegation *delegation;
1478
1479         rcu_read_lock();
1480         delegation = rcu_dereference(NFS_I(inode)->delegation);
1481         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1482                 rcu_read_unlock();
1483                 return;
1484         }
1485         rcu_read_unlock();
1486         nfs4_inode_return_delegation(inode);
1487 }
1488
1489 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1490 {
1491         struct nfs4_state *state = opendata->state;
1492         struct nfs_inode *nfsi = NFS_I(state->inode);
1493         struct nfs_delegation *delegation;
1494         int open_mode = opendata->o_arg.open_flags;
1495         fmode_t fmode = opendata->o_arg.fmode;
1496         enum open_claim_type4 claim = opendata->o_arg.claim;
1497         nfs4_stateid stateid;
1498         int ret = -EAGAIN;
1499
1500         for (;;) {
1501                 spin_lock(&state->owner->so_lock);
1502                 if (can_open_cached(state, fmode, open_mode)) {
1503                         update_open_stateflags(state, fmode);
1504                         spin_unlock(&state->owner->so_lock);
1505                         goto out_return_state;
1506                 }
1507                 spin_unlock(&state->owner->so_lock);
1508                 rcu_read_lock();
1509                 delegation = rcu_dereference(nfsi->delegation);
1510                 if (!can_open_delegated(delegation, fmode, claim)) {
1511                         rcu_read_unlock();
1512                         break;
1513                 }
1514                 /* Save the delegation */
1515                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1516                 rcu_read_unlock();
1517                 nfs_release_seqid(opendata->o_arg.seqid);
1518                 if (!opendata->is_recover) {
1519                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1520                         if (ret != 0)
1521                                 goto out;
1522                 }
1523                 ret = -EAGAIN;
1524
1525                 /* Try to update the stateid using the delegation */
1526                 if (update_open_stateid(state, NULL, &stateid, fmode))
1527                         goto out_return_state;
1528         }
1529 out:
1530         return ERR_PTR(ret);
1531 out_return_state:
1532         atomic_inc(&state->count);
1533         return state;
1534 }
1535
1536 static void
1537 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1538 {
1539         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1540         struct nfs_delegation *delegation;
1541         int delegation_flags = 0;
1542
1543         rcu_read_lock();
1544         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1545         if (delegation)
1546                 delegation_flags = delegation->flags;
1547         rcu_read_unlock();
1548         switch (data->o_arg.claim) {
1549         default:
1550                 break;
1551         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1552         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1553                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1554                                    "returning a delegation for "
1555                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1556                                    clp->cl_hostname);
1557                 return;
1558         }
1559         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1560                 nfs_inode_set_delegation(state->inode,
1561                                          data->owner->so_cred,
1562                                          &data->o_res);
1563         else
1564                 nfs_inode_reclaim_delegation(state->inode,
1565                                              data->owner->so_cred,
1566                                              &data->o_res);
1567 }
1568
1569 /*
1570  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1571  * and update the nfs4_state.
1572  */
1573 static struct nfs4_state *
1574 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1575 {
1576         struct inode *inode = data->state->inode;
1577         struct nfs4_state *state = data->state;
1578         int ret;
1579
1580         if (!data->rpc_done) {
1581                 if (data->rpc_status) {
1582                         ret = data->rpc_status;
1583                         goto err;
1584                 }
1585                 /* cached opens have already been processed */
1586                 goto update;
1587         }
1588
1589         ret = nfs_refresh_inode(inode, &data->f_attr);
1590         if (ret)
1591                 goto err;
1592
1593         if (data->o_res.delegation_type != 0)
1594                 nfs4_opendata_check_deleg(data, state);
1595 update:
1596         update_open_stateid(state, &data->o_res.stateid, NULL,
1597                             data->o_arg.fmode);
1598         atomic_inc(&state->count);
1599
1600         return state;
1601 err:
1602         return ERR_PTR(ret);
1603
1604 }
1605
1606 static struct nfs4_state *
1607 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1608 {
1609         struct inode *inode;
1610         struct nfs4_state *state = NULL;
1611         int ret;
1612
1613         if (!data->rpc_done) {
1614                 state = nfs4_try_open_cached(data);
1615                 trace_nfs4_cached_open(data->state);
1616                 goto out;
1617         }
1618
1619         ret = -EAGAIN;
1620         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1621                 goto err;
1622         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1623         ret = PTR_ERR(inode);
1624         if (IS_ERR(inode))
1625                 goto err;
1626         ret = -ENOMEM;
1627         state = nfs4_get_open_state(inode, data->owner);
1628         if (state == NULL)
1629                 goto err_put_inode;
1630         if (data->o_res.delegation_type != 0)
1631                 nfs4_opendata_check_deleg(data, state);
1632         update_open_stateid(state, &data->o_res.stateid, NULL,
1633                         data->o_arg.fmode);
1634         iput(inode);
1635 out:
1636         nfs_release_seqid(data->o_arg.seqid);
1637         return state;
1638 err_put_inode:
1639         iput(inode);
1640 err:
1641         return ERR_PTR(ret);
1642 }
1643
1644 static struct nfs4_state *
1645 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1646 {
1647         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1648                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1649         return _nfs4_opendata_to_nfs4_state(data);
1650 }
1651
1652 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1653 {
1654         struct nfs_inode *nfsi = NFS_I(state->inode);
1655         struct nfs_open_context *ctx;
1656
1657         spin_lock(&state->inode->i_lock);
1658         list_for_each_entry(ctx, &nfsi->open_files, list) {
1659                 if (ctx->state != state)
1660                         continue;
1661                 get_nfs_open_context(ctx);
1662                 spin_unlock(&state->inode->i_lock);
1663                 return ctx;
1664         }
1665         spin_unlock(&state->inode->i_lock);
1666         return ERR_PTR(-ENOENT);
1667 }
1668
1669 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1670                 struct nfs4_state *state, enum open_claim_type4 claim)
1671 {
1672         struct nfs4_opendata *opendata;
1673
1674         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1675                         NULL, NULL, claim, GFP_NOFS);
1676         if (opendata == NULL)
1677                 return ERR_PTR(-ENOMEM);
1678         opendata->state = state;
1679         atomic_inc(&state->count);
1680         return opendata;
1681 }
1682
1683 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1684                 fmode_t fmode)
1685 {
1686         struct nfs4_state *newstate;
1687         int ret;
1688
1689         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1690                 return 0;
1691         opendata->o_arg.open_flags = 0;
1692         opendata->o_arg.fmode = fmode;
1693         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1694                         NFS_SB(opendata->dentry->d_sb),
1695                         fmode, 0);
1696         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1697         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1698         nfs4_init_opendata_res(opendata);
1699         ret = _nfs4_recover_proc_open(opendata);
1700         if (ret != 0)
1701                 return ret; 
1702         newstate = nfs4_opendata_to_nfs4_state(opendata);
1703         if (IS_ERR(newstate))
1704                 return PTR_ERR(newstate);
1705         if (newstate != opendata->state)
1706                 ret = -ESTALE;
1707         nfs4_close_state(newstate, fmode);
1708         return ret;
1709 }
1710
1711 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1712 {
1713         int ret;
1714
1715         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1716         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1717         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1718         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1719         /* memory barrier prior to reading state->n_* */
1720         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1721         clear_bit(NFS_OPEN_STATE, &state->flags);
1722         smp_rmb();
1723         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1724         if (ret != 0)
1725                 return ret;
1726         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1727         if (ret != 0)
1728                 return ret;
1729         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1730         if (ret != 0)
1731                 return ret;
1732         /*
1733          * We may have performed cached opens for all three recoveries.
1734          * Check if we need to update the current stateid.
1735          */
1736         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1737             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1738                 write_seqlock(&state->seqlock);
1739                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1740                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1741                 write_sequnlock(&state->seqlock);
1742         }
1743         return 0;
1744 }
1745
1746 /*
1747  * OPEN_RECLAIM:
1748  *      reclaim state on the server after a reboot.
1749  */
1750 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1751 {
1752         struct nfs_delegation *delegation;
1753         struct nfs4_opendata *opendata;
1754         fmode_t delegation_type = 0;
1755         int status;
1756
1757         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1758                         NFS4_OPEN_CLAIM_PREVIOUS);
1759         if (IS_ERR(opendata))
1760                 return PTR_ERR(opendata);
1761         rcu_read_lock();
1762         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1763         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1764                 delegation_type = delegation->type;
1765         rcu_read_unlock();
1766         opendata->o_arg.u.delegation_type = delegation_type;
1767         status = nfs4_open_recover(opendata, state);
1768         nfs4_opendata_put(opendata);
1769         return status;
1770 }
1771
1772 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1773 {
1774         struct nfs_server *server = NFS_SERVER(state->inode);
1775         struct nfs4_exception exception = { };
1776         int err;
1777         do {
1778                 err = _nfs4_do_open_reclaim(ctx, state);
1779                 trace_nfs4_open_reclaim(ctx, 0, err);
1780                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1781                         continue;
1782                 if (err != -NFS4ERR_DELAY)
1783                         break;
1784                 nfs4_handle_exception(server, err, &exception);
1785         } while (exception.retry);
1786         return err;
1787 }
1788
1789 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1790 {
1791         struct nfs_open_context *ctx;
1792         int ret;
1793
1794         ctx = nfs4_state_find_open_context(state);
1795         if (IS_ERR(ctx))
1796                 return -EAGAIN;
1797         ret = nfs4_do_open_reclaim(ctx, state);
1798         put_nfs_open_context(ctx);
1799         return ret;
1800 }
1801
1802 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1803 {
1804         switch (err) {
1805                 default:
1806                         printk(KERN_ERR "NFS: %s: unhandled error "
1807                                         "%d.\n", __func__, err);
1808                 case 0:
1809                 case -ENOENT:
1810                 case -EAGAIN:
1811                 case -ESTALE:
1812                         break;
1813                 case -NFS4ERR_BADSESSION:
1814                 case -NFS4ERR_BADSLOT:
1815                 case -NFS4ERR_BAD_HIGH_SLOT:
1816                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1817                 case -NFS4ERR_DEADSESSION:
1818                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1819                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1820                         return -EAGAIN;
1821                 case -NFS4ERR_STALE_CLIENTID:
1822                 case -NFS4ERR_STALE_STATEID:
1823                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1824                 case -NFS4ERR_EXPIRED:
1825                         /* Don't recall a delegation if it was lost */
1826                         nfs4_schedule_lease_recovery(server->nfs_client);
1827                         return -EAGAIN;
1828                 case -NFS4ERR_MOVED:
1829                         nfs4_schedule_migration_recovery(server);
1830                         return -EAGAIN;
1831                 case -NFS4ERR_LEASE_MOVED:
1832                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1833                         return -EAGAIN;
1834                 case -NFS4ERR_DELEG_REVOKED:
1835                 case -NFS4ERR_ADMIN_REVOKED:
1836                 case -NFS4ERR_BAD_STATEID:
1837                 case -NFS4ERR_OPENMODE:
1838                         nfs_inode_find_state_and_recover(state->inode,
1839                                         stateid);
1840                         nfs4_schedule_stateid_recovery(server, state);
1841                         return -EAGAIN;
1842                 case -NFS4ERR_DELAY:
1843                 case -NFS4ERR_GRACE:
1844                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1845                         ssleep(1);
1846                         return -EAGAIN;
1847                 case -ENOMEM:
1848                 case -NFS4ERR_DENIED:
1849                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1850                         return 0;
1851         }
1852         return err;
1853 }
1854
1855 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1856                 struct nfs4_state *state, const nfs4_stateid *stateid,
1857                 fmode_t type)
1858 {
1859         struct nfs_server *server = NFS_SERVER(state->inode);
1860         struct nfs4_opendata *opendata;
1861         int err = 0;
1862
1863         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1864                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1865         if (IS_ERR(opendata))
1866                 return PTR_ERR(opendata);
1867         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1868         write_seqlock(&state->seqlock);
1869         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1870         write_sequnlock(&state->seqlock);
1871         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1872         switch (type & (FMODE_READ|FMODE_WRITE)) {
1873         case FMODE_READ|FMODE_WRITE:
1874         case FMODE_WRITE:
1875                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1876                 if (err)
1877                         break;
1878                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1879                 if (err)
1880                         break;
1881         case FMODE_READ:
1882                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1883         }
1884         nfs4_opendata_put(opendata);
1885         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1886 }
1887
1888 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1889 {
1890         struct nfs4_opendata *data = calldata;
1891
1892         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1893                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1894 }
1895
1896 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1897 {
1898         struct nfs4_opendata *data = calldata;
1899
1900         nfs40_sequence_done(task, &data->c_res.seq_res);
1901
1902         data->rpc_status = task->tk_status;
1903         if (data->rpc_status == 0) {
1904                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1905                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1906                 renew_lease(data->o_res.server, data->timestamp);
1907                 data->rpc_done = 1;
1908         }
1909 }
1910
1911 static void nfs4_open_confirm_release(void *calldata)
1912 {
1913         struct nfs4_opendata *data = calldata;
1914         struct nfs4_state *state = NULL;
1915
1916         /* If this request hasn't been cancelled, do nothing */
1917         if (data->cancelled == 0)
1918                 goto out_free;
1919         /* In case of error, no cleanup! */
1920         if (!data->rpc_done)
1921                 goto out_free;
1922         state = nfs4_opendata_to_nfs4_state(data);
1923         if (!IS_ERR(state))
1924                 nfs4_close_state(state, data->o_arg.fmode);
1925 out_free:
1926         nfs4_opendata_put(data);
1927 }
1928
1929 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1930         .rpc_call_prepare = nfs4_open_confirm_prepare,
1931         .rpc_call_done = nfs4_open_confirm_done,
1932         .rpc_release = nfs4_open_confirm_release,
1933 };
1934
1935 /*
1936  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1937  */
1938 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1939 {
1940         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1941         struct rpc_task *task;
1942         struct  rpc_message msg = {
1943                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1944                 .rpc_argp = &data->c_arg,
1945                 .rpc_resp = &data->c_res,
1946                 .rpc_cred = data->owner->so_cred,
1947         };
1948         struct rpc_task_setup task_setup_data = {
1949                 .rpc_client = server->client,
1950                 .rpc_message = &msg,
1951                 .callback_ops = &nfs4_open_confirm_ops,
1952                 .callback_data = data,
1953                 .workqueue = nfsiod_workqueue,
1954                 .flags = RPC_TASK_ASYNC,
1955         };
1956         int status;
1957
1958         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1959         kref_get(&data->kref);
1960         data->rpc_done = 0;
1961         data->rpc_status = 0;
1962         data->timestamp = jiffies;
1963         if (data->is_recover)
1964                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1965         task = rpc_run_task(&task_setup_data);
1966         if (IS_ERR(task))
1967                 return PTR_ERR(task);
1968         status = nfs4_wait_for_completion_rpc_task(task);
1969         if (status != 0) {
1970                 data->cancelled = 1;
1971                 smp_wmb();
1972         } else
1973                 status = data->rpc_status;
1974         rpc_put_task(task);
1975         return status;
1976 }
1977
1978 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1979 {
1980         struct nfs4_opendata *data = calldata;
1981         struct nfs4_state_owner *sp = data->owner;
1982         struct nfs_client *clp = sp->so_server->nfs_client;
1983         enum open_claim_type4 claim = data->o_arg.claim;
1984
1985         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1986                 goto out_wait;
1987         /*
1988          * Check if we still need to send an OPEN call, or if we can use
1989          * a delegation instead.
1990          */
1991         if (data->state != NULL) {
1992                 struct nfs_delegation *delegation;
1993
1994                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1995                         goto out_no_action;
1996                 rcu_read_lock();
1997                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1998                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1999                         goto unlock_no_action;
2000                 rcu_read_unlock();
2001         }
2002         /* Update client id. */
2003         data->o_arg.clientid = clp->cl_clientid;
2004         switch (claim) {
2005         default:
2006                 break;
2007         case NFS4_OPEN_CLAIM_PREVIOUS:
2008         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2009         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2010                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2011         case NFS4_OPEN_CLAIM_FH:
2012                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2013                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2014         }
2015         data->timestamp = jiffies;
2016         if (nfs4_setup_sequence(data->o_arg.server,
2017                                 &data->o_arg.seq_args,
2018                                 &data->o_res.seq_res,
2019                                 task) != 0)
2020                 nfs_release_seqid(data->o_arg.seqid);
2021
2022         /* Set the create mode (note dependency on the session type) */
2023         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2024         if (data->o_arg.open_flags & O_EXCL) {
2025                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2026                 if (nfs4_has_persistent_session(clp))
2027                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2028                 else if (clp->cl_mvops->minor_version > 0)
2029                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2030         }
2031         return;
2032 unlock_no_action:
2033         trace_nfs4_cached_open(data->state);
2034         rcu_read_unlock();
2035 out_no_action:
2036         task->tk_action = NULL;
2037 out_wait:
2038         nfs4_sequence_done(task, &data->o_res.seq_res);
2039 }
2040
2041 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2042 {
2043         struct nfs4_opendata *data = calldata;
2044
2045         data->rpc_status = task->tk_status;
2046
2047         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2048                 return;
2049
2050         if (task->tk_status == 0) {
2051                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2052                         switch (data->o_res.f_attr->mode & S_IFMT) {
2053                         case S_IFREG:
2054                                 break;
2055                         case S_IFLNK:
2056                                 data->rpc_status = -ELOOP;
2057                                 break;
2058                         case S_IFDIR:
2059                                 data->rpc_status = -EISDIR;
2060                                 break;
2061                         default:
2062                                 data->rpc_status = -ENOTDIR;
2063                         }
2064                 }
2065                 renew_lease(data->o_res.server, data->timestamp);
2066                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2067                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2068         }
2069         data->rpc_done = 1;
2070 }
2071
2072 static void nfs4_open_release(void *calldata)
2073 {
2074         struct nfs4_opendata *data = calldata;
2075         struct nfs4_state *state = NULL;
2076
2077         /* If this request hasn't been cancelled, do nothing */
2078         if (data->cancelled == 0)
2079                 goto out_free;
2080         /* In case of error, no cleanup! */
2081         if (data->rpc_status != 0 || !data->rpc_done)
2082                 goto out_free;
2083         /* In case we need an open_confirm, no cleanup! */
2084         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2085                 goto out_free;
2086         state = nfs4_opendata_to_nfs4_state(data);
2087         if (!IS_ERR(state))
2088                 nfs4_close_state(state, data->o_arg.fmode);
2089 out_free:
2090         nfs4_opendata_put(data);
2091 }
2092
2093 static const struct rpc_call_ops nfs4_open_ops = {
2094         .rpc_call_prepare = nfs4_open_prepare,
2095         .rpc_call_done = nfs4_open_done,
2096         .rpc_release = nfs4_open_release,
2097 };
2098
2099 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2100 {
2101         struct inode *dir = d_inode(data->dir);
2102         struct nfs_server *server = NFS_SERVER(dir);
2103         struct nfs_openargs *o_arg = &data->o_arg;
2104         struct nfs_openres *o_res = &data->o_res;
2105         struct rpc_task *task;
2106         struct rpc_message msg = {
2107                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2108                 .rpc_argp = o_arg,
2109                 .rpc_resp = o_res,
2110                 .rpc_cred = data->owner->so_cred,
2111         };
2112         struct rpc_task_setup task_setup_data = {
2113                 .rpc_client = server->client,
2114                 .rpc_message = &msg,
2115                 .callback_ops = &nfs4_open_ops,
2116                 .callback_data = data,
2117                 .workqueue = nfsiod_workqueue,
2118                 .flags = RPC_TASK_ASYNC,
2119         };
2120         int status;
2121
2122         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2123         kref_get(&data->kref);
2124         data->rpc_done = 0;
2125         data->rpc_status = 0;
2126         data->cancelled = 0;
2127         data->is_recover = 0;
2128         if (isrecover) {
2129                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2130                 data->is_recover = 1;
2131         }
2132         task = rpc_run_task(&task_setup_data);
2133         if (IS_ERR(task))
2134                 return PTR_ERR(task);
2135         status = nfs4_wait_for_completion_rpc_task(task);
2136         if (status != 0) {
2137                 data->cancelled = 1;
2138                 smp_wmb();
2139         } else
2140                 status = data->rpc_status;
2141         rpc_put_task(task);
2142
2143         return status;
2144 }
2145
2146 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2147 {
2148         struct inode *dir = d_inode(data->dir);
2149         struct nfs_openres *o_res = &data->o_res;
2150         int status;
2151
2152         status = nfs4_run_open_task(data, 1);
2153         if (status != 0 || !data->rpc_done)
2154                 return status;
2155
2156         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2157
2158         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2159                 status = _nfs4_proc_open_confirm(data);
2160                 if (status != 0)
2161                         return status;
2162         }
2163
2164         return status;
2165 }
2166
2167 /*
2168  * Additional permission checks in order to distinguish between an
2169  * open for read, and an open for execute. This works around the
2170  * fact that NFSv4 OPEN treats read and execute permissions as being
2171  * the same.
2172  * Note that in the non-execute case, we want to turn off permission
2173  * checking if we just created a new file (POSIX open() semantics).
2174  */
2175 static int nfs4_opendata_access(struct rpc_cred *cred,
2176                                 struct nfs4_opendata *opendata,
2177                                 struct nfs4_state *state, fmode_t fmode,
2178                                 int openflags)
2179 {
2180         struct nfs_access_entry cache;
2181         u32 mask;
2182
2183         /* access call failed or for some reason the server doesn't
2184          * support any access modes -- defer access call until later */
2185         if (opendata->o_res.access_supported == 0)
2186                 return 0;
2187
2188         mask = 0;
2189         /*
2190          * Use openflags to check for exec, because fmode won't
2191          * always have FMODE_EXEC set when file open for exec.
2192          */
2193         if (openflags & __FMODE_EXEC) {
2194                 /* ONLY check for exec rights */
2195                 mask = MAY_EXEC;
2196         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2197                 mask = MAY_READ;
2198
2199         cache.cred = cred;
2200         cache.jiffies = jiffies;
2201         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2202         nfs_access_add_cache(state->inode, &cache);
2203
2204         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2205                 return 0;
2206
2207         /* even though OPEN succeeded, access is denied. Close the file */
2208         nfs4_close_state(state, fmode);
2209         return -EACCES;
2210 }
2211
2212 /*
2213  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2214  */
2215 static int _nfs4_proc_open(struct nfs4_opendata *data)
2216 {
2217         struct inode *dir = d_inode(data->dir);
2218         struct nfs_server *server = NFS_SERVER(dir);
2219         struct nfs_openargs *o_arg = &data->o_arg;
2220         struct nfs_openres *o_res = &data->o_res;
2221         int status;
2222
2223         status = nfs4_run_open_task(data, 0);
2224         if (!data->rpc_done)
2225                 return status;
2226         if (status != 0) {
2227                 if (status == -NFS4ERR_BADNAME &&
2228                                 !(o_arg->open_flags & O_CREAT))
2229                         return -ENOENT;
2230                 return status;
2231         }
2232
2233         nfs_fattr_map_and_free_names(server, &data->f_attr);
2234
2235         if (o_arg->open_flags & O_CREAT) {
2236                 update_changeattr(dir, &o_res->cinfo);
2237                 if (o_arg->open_flags & O_EXCL)
2238                         data->file_created = 1;
2239                 else if (o_res->cinfo.before != o_res->cinfo.after)
2240                         data->file_created = 1;
2241         }
2242         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2243                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2244         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2245                 status = _nfs4_proc_open_confirm(data);
2246                 if (status != 0)
2247                         return status;
2248         }
2249         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2250                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2251         return 0;
2252 }
2253
2254 static int nfs4_recover_expired_lease(struct nfs_server *server)
2255 {
2256         return nfs4_client_recover_expired_lease(server->nfs_client);
2257 }
2258
2259 /*
2260  * OPEN_EXPIRED:
2261  *      reclaim state on the server after a network partition.
2262  *      Assumes caller holds the appropriate lock
2263  */
2264 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2265 {
2266         struct nfs4_opendata *opendata;
2267         int ret;
2268
2269         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2270                         NFS4_OPEN_CLAIM_FH);
2271         if (IS_ERR(opendata))
2272                 return PTR_ERR(opendata);
2273         ret = nfs4_open_recover(opendata, state);
2274         if (ret == -ESTALE)
2275                 d_drop(ctx->dentry);
2276         nfs4_opendata_put(opendata);
2277         return ret;
2278 }
2279
2280 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2281 {
2282         struct nfs_server *server = NFS_SERVER(state->inode);
2283         struct nfs4_exception exception = { };
2284         int err;
2285
2286         do {
2287                 err = _nfs4_open_expired(ctx, state);
2288                 trace_nfs4_open_expired(ctx, 0, err);
2289                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2290                         continue;
2291                 switch (err) {
2292                 default:
2293                         goto out;
2294                 case -NFS4ERR_GRACE:
2295                 case -NFS4ERR_DELAY:
2296                         nfs4_handle_exception(server, err, &exception);
2297                         err = 0;
2298                 }
2299         } while (exception.retry);
2300 out:
2301         return err;
2302 }
2303
2304 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2305 {
2306         struct nfs_open_context *ctx;
2307         int ret;
2308
2309         ctx = nfs4_state_find_open_context(state);
2310         if (IS_ERR(ctx))
2311                 return -EAGAIN;
2312         ret = nfs4_do_open_expired(ctx, state);
2313         put_nfs_open_context(ctx);
2314         return ret;
2315 }
2316
2317 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2318 {
2319         nfs_remove_bad_delegation(state->inode);
2320         write_seqlock(&state->seqlock);
2321         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2322         write_sequnlock(&state->seqlock);
2323         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2324 }
2325
2326 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2327 {
2328         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2329                 nfs_finish_clear_delegation_stateid(state);
2330 }
2331
2332 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2333 {
2334         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2335         nfs40_clear_delegation_stateid(state);
2336         return nfs4_open_expired(sp, state);
2337 }
2338
2339 #if defined(CONFIG_NFS_V4_1)
2340 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2341 {
2342         struct nfs_server *server = NFS_SERVER(state->inode);
2343         nfs4_stateid stateid;
2344         struct nfs_delegation *delegation;
2345         struct rpc_cred *cred;
2346         int status;
2347
2348         /* Get the delegation credential for use by test/free_stateid */
2349         rcu_read_lock();
2350         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2351         if (delegation == NULL) {
2352                 rcu_read_unlock();
2353                 return;
2354         }
2355
2356         nfs4_stateid_copy(&stateid, &delegation->stateid);
2357         cred = get_rpccred(delegation->cred);
2358         rcu_read_unlock();
2359         status = nfs41_test_stateid(server, &stateid, cred);
2360         trace_nfs4_test_delegation_stateid(state, NULL, status);
2361
2362         if (status != NFS_OK) {
2363                 /* Free the stateid unless the server explicitly
2364                  * informs us the stateid is unrecognized. */
2365                 if (status != -NFS4ERR_BAD_STATEID)
2366                         nfs41_free_stateid(server, &stateid, cred);
2367                 nfs_finish_clear_delegation_stateid(state);
2368         }
2369
2370         put_rpccred(cred);
2371 }
2372
2373 /**
2374  * nfs41_check_open_stateid - possibly free an open stateid
2375  *
2376  * @state: NFSv4 state for an inode
2377  *
2378  * Returns NFS_OK if recovery for this stateid is now finished.
2379  * Otherwise a negative NFS4ERR value is returned.
2380  */
2381 static int nfs41_check_open_stateid(struct nfs4_state *state)
2382 {
2383         struct nfs_server *server = NFS_SERVER(state->inode);
2384         nfs4_stateid *stateid = &state->open_stateid;
2385         struct rpc_cred *cred = state->owner->so_cred;
2386         int status;
2387
2388         /* If a state reset has been done, test_stateid is unneeded */
2389         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2390             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2391             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2392                 return -NFS4ERR_BAD_STATEID;
2393
2394         status = nfs41_test_stateid(server, stateid, cred);
2395         trace_nfs4_test_open_stateid(state, NULL, status);
2396         if (status != NFS_OK) {
2397                 /* Free the stateid unless the server explicitly
2398                  * informs us the stateid is unrecognized. */
2399                 if (status != -NFS4ERR_BAD_STATEID)
2400                         nfs41_free_stateid(server, stateid, cred);
2401
2402                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2403                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2404                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2405                 clear_bit(NFS_OPEN_STATE, &state->flags);
2406         }
2407         return status;
2408 }
2409
2410 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2411 {
2412         int status;
2413
2414         nfs41_check_delegation_stateid(state);
2415         status = nfs41_check_open_stateid(state);
2416         if (status != NFS_OK)
2417                 status = nfs4_open_expired(sp, state);
2418         return status;
2419 }
2420 #endif
2421
2422 /*
2423  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2424  * fields corresponding to attributes that were used to store the verifier.
2425  * Make sure we clobber those fields in the later setattr call
2426  */
2427 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2428                                 struct iattr *sattr, struct nfs4_label **label)
2429 {
2430         const u32 *attrset = opendata->o_res.attrset;
2431
2432         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2433             !(sattr->ia_valid & ATTR_ATIME_SET))
2434                 sattr->ia_valid |= ATTR_ATIME;
2435
2436         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2437             !(sattr->ia_valid & ATTR_MTIME_SET))
2438                 sattr->ia_valid |= ATTR_MTIME;
2439
2440         /* Except MODE, it seems harmless of setting twice. */
2441         if ((attrset[1] & FATTR4_WORD1_MODE))
2442                 sattr->ia_valid &= ~ATTR_MODE;
2443
2444         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2445                 *label = NULL;
2446 }
2447
2448 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2449                 fmode_t fmode,
2450                 int flags,
2451                 struct nfs_open_context *ctx)
2452 {
2453         struct nfs4_state_owner *sp = opendata->owner;
2454         struct nfs_server *server = sp->so_server;
2455         struct dentry *dentry;
2456         struct nfs4_state *state;
2457         unsigned int seq;
2458         int ret;
2459
2460         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2461
2462         ret = _nfs4_proc_open(opendata);
2463         if (ret != 0)
2464                 goto out;
2465
2466         state = nfs4_opendata_to_nfs4_state(opendata);
2467         ret = PTR_ERR(state);
2468         if (IS_ERR(state))
2469                 goto out;
2470         if (server->caps & NFS_CAP_POSIX_LOCK)
2471                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2472
2473         dentry = opendata->dentry;
2474         if (d_really_is_negative(dentry)) {
2475                 struct dentry *alias;
2476                 d_drop(dentry);
2477                 alias = d_exact_alias(dentry, state->inode);
2478                 if (!alias)
2479                         alias = d_splice_alias(igrab(state->inode), dentry);
2480                 /* d_splice_alias() can't fail here - it's a non-directory */
2481                 if (alias) {
2482                         dput(ctx->dentry);
2483                         ctx->dentry = dentry = alias;
2484                 }
2485                 nfs_set_verifier(dentry,
2486                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2487         }
2488
2489         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2490         if (ret != 0)
2491                 goto out;
2492
2493         ctx->state = state;
2494         if (d_inode(dentry) == state->inode) {
2495                 nfs_inode_attach_open_context(ctx);
2496                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2497                         nfs4_schedule_stateid_recovery(server, state);
2498         }
2499 out:
2500         return ret;
2501 }
2502
2503 /*
2504  * Returns a referenced nfs4_state
2505  */
2506 static int _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 nfs4_state_owner  *sp;
2514         struct nfs4_state     *state = NULL;
2515         struct nfs_server       *server = NFS_SERVER(dir);
2516         struct nfs4_opendata *opendata;
2517         struct dentry *dentry = ctx->dentry;
2518         struct rpc_cred *cred = ctx->cred;
2519         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2520         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2521         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2522         struct nfs4_label *olabel = NULL;
2523         int status;
2524
2525         /* Protect against reboot recovery conflicts */
2526         status = -ENOMEM;
2527         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2528         if (sp == NULL) {
2529                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2530                 goto out_err;
2531         }
2532         status = nfs4_recover_expired_lease(server);
2533         if (status != 0)
2534                 goto err_put_state_owner;
2535         if (d_really_is_positive(dentry))
2536                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2537         status = -ENOMEM;
2538         if (d_really_is_positive(dentry))
2539                 claim = NFS4_OPEN_CLAIM_FH;
2540         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2541                         label, claim, GFP_KERNEL);
2542         if (opendata == NULL)
2543                 goto err_put_state_owner;
2544
2545         if (label) {
2546                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2547                 if (IS_ERR(olabel)) {
2548                         status = PTR_ERR(olabel);
2549                         goto err_opendata_put;
2550                 }
2551         }
2552
2553         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2554                 if (!opendata->f_attr.mdsthreshold) {
2555                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2556                         if (!opendata->f_attr.mdsthreshold)
2557                                 goto err_free_label;
2558                 }
2559                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2560         }
2561         if (d_really_is_positive(dentry))
2562                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2563
2564         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2565         if (status != 0)
2566                 goto err_free_label;
2567         state = ctx->state;
2568
2569         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2570             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2571                 nfs4_exclusive_attrset(opendata, sattr, &label);
2572                 /*
2573                  * send create attributes which was not set by open
2574                  * with an extra setattr.
2575                  */
2576                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2577                         nfs_fattr_init(opendata->o_res.f_attr);
2578                         status = nfs4_do_setattr(state->inode, cred,
2579                                         opendata->o_res.f_attr, sattr,
2580                                         state, label, olabel);
2581                         if (status == 0) {
2582                                 nfs_setattr_update_inode(state->inode, sattr,
2583                                                 opendata->o_res.f_attr);
2584                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2585                         }
2586                 }
2587         }
2588         if (opened && opendata->file_created)
2589                 *opened |= FILE_CREATED;
2590
2591         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2592                 *ctx_th = opendata->f_attr.mdsthreshold;
2593                 opendata->f_attr.mdsthreshold = NULL;
2594         }
2595
2596         nfs4_label_free(olabel);
2597
2598         nfs4_opendata_put(opendata);
2599         nfs4_put_state_owner(sp);
2600         return 0;
2601 err_free_label:
2602         nfs4_label_free(olabel);
2603 err_opendata_put:
2604         nfs4_opendata_put(opendata);
2605 err_put_state_owner:
2606         nfs4_put_state_owner(sp);
2607 out_err:
2608         return status;
2609 }
2610
2611
2612 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2613                                         struct nfs_open_context *ctx,
2614                                         int flags,
2615                                         struct iattr *sattr,
2616                                         struct nfs4_label *label,
2617                                         int *opened)
2618 {
2619         struct nfs_server *server = NFS_SERVER(dir);
2620         struct nfs4_exception exception = { };
2621         struct nfs4_state *res;
2622         int status;
2623
2624         do {
2625                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2626                 res = ctx->state;
2627                 trace_nfs4_open_file(ctx, flags, status);
2628                 if (status == 0)
2629                         break;
2630                 /* NOTE: BAD_SEQID means the server and client disagree about the
2631                  * book-keeping w.r.t. state-changing operations
2632                  * (OPEN/CLOSE/LOCK/LOCKU...)
2633                  * It is actually a sign of a bug on the client or on the server.
2634                  *
2635                  * If we receive a BAD_SEQID error in the particular case of
2636                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2637                  * have unhashed the old state_owner for us, and that we can
2638                  * therefore safely retry using a new one. We should still warn
2639                  * the user though...
2640                  */
2641                 if (status == -NFS4ERR_BAD_SEQID) {
2642                         pr_warn_ratelimited("NFS: v4 server %s "
2643                                         " returned a bad sequence-id error!\n",
2644                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2645                         exception.retry = 1;
2646                         continue;
2647                 }
2648                 /*
2649                  * BAD_STATEID on OPEN means that the server cancelled our
2650                  * state before it received the OPEN_CONFIRM.
2651                  * Recover by retrying the request as per the discussion
2652                  * on Page 181 of RFC3530.
2653                  */
2654                 if (status == -NFS4ERR_BAD_STATEID) {
2655                         exception.retry = 1;
2656                         continue;
2657                 }
2658                 if (status == -EAGAIN) {
2659                         /* We must have found a delegation */
2660                         exception.retry = 1;
2661                         continue;
2662                 }
2663                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2664                         continue;
2665                 res = ERR_PTR(nfs4_handle_exception(server,
2666                                         status, &exception));
2667         } while (exception.retry);
2668         return res;
2669 }
2670
2671 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2672                             struct nfs_fattr *fattr, struct iattr *sattr,
2673                             struct nfs4_state *state, struct nfs4_label *ilabel,
2674                             struct nfs4_label *olabel)
2675 {
2676         struct nfs_server *server = NFS_SERVER(inode);
2677         struct nfs_setattrargs  arg = {
2678                 .fh             = NFS_FH(inode),
2679                 .iap            = sattr,
2680                 .server         = server,
2681                 .bitmask = server->attr_bitmask,
2682                 .label          = ilabel,
2683         };
2684         struct nfs_setattrres  res = {
2685                 .fattr          = fattr,
2686                 .label          = olabel,
2687                 .server         = server,
2688         };
2689         struct rpc_message msg = {
2690                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2691                 .rpc_argp       = &arg,
2692                 .rpc_resp       = &res,
2693                 .rpc_cred       = cred,
2694         };
2695         unsigned long timestamp = jiffies;
2696         fmode_t fmode;
2697         bool truncate;
2698         int status;
2699
2700         arg.bitmask = nfs4_bitmask(server, ilabel);
2701         if (ilabel)
2702                 arg.bitmask = nfs4_bitmask(server, olabel);
2703
2704         nfs_fattr_init(fattr);
2705
2706         /* Servers should only apply open mode checks for file size changes */
2707         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2708         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2709
2710         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2711                 /* Use that stateid */
2712         } else if (truncate && state != NULL) {
2713                 struct nfs_lockowner lockowner = {
2714                         .l_owner = current->files,
2715                         .l_pid = current->tgid,
2716                 };
2717                 if (!nfs4_valid_open_stateid(state))
2718                         return -EBADF;
2719                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2720                                 &lockowner) == -EIO)
2721                         return -EBADF;
2722         } else
2723                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2724
2725         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2726         if (status == 0 && state != NULL)
2727                 renew_lease(server, timestamp);
2728         trace_nfs4_setattr(inode, &arg.stateid, status);
2729         return status;
2730 }
2731
2732 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2733                            struct nfs_fattr *fattr, struct iattr *sattr,
2734                            struct nfs4_state *state, struct nfs4_label *ilabel,
2735                            struct nfs4_label *olabel)
2736 {
2737         struct nfs_server *server = NFS_SERVER(inode);
2738         struct nfs4_exception exception = {
2739                 .state = state,
2740                 .inode = inode,
2741         };
2742         int err;
2743         do {
2744                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2745                 switch (err) {
2746                 case -NFS4ERR_OPENMODE:
2747                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2748                                 pr_warn_once("NFSv4: server %s is incorrectly "
2749                                                 "applying open mode checks to "
2750                                                 "a SETATTR that is not "
2751                                                 "changing file size.\n",
2752                                                 server->nfs_client->cl_hostname);
2753                         }
2754                         if (state && !(state->state & FMODE_WRITE)) {
2755                                 err = -EBADF;
2756                                 if (sattr->ia_valid & ATTR_OPEN)
2757                                         err = -EACCES;
2758                                 goto out;
2759                         }
2760                 }
2761                 err = nfs4_handle_exception(server, err, &exception);
2762         } while (exception.retry);
2763 out:
2764         return err;
2765 }
2766
2767 static bool
2768 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2769 {
2770         if (inode == NULL || !nfs_have_layout(inode))
2771                 return false;
2772
2773         return pnfs_wait_on_layoutreturn(inode, task);
2774 }
2775
2776 struct nfs4_closedata {
2777         struct inode *inode;
2778         struct nfs4_state *state;
2779         struct nfs_closeargs arg;
2780         struct nfs_closeres res;
2781         struct nfs_fattr fattr;
2782         unsigned long timestamp;
2783         bool roc;
2784         u32 roc_barrier;
2785 };
2786
2787 static void nfs4_free_closedata(void *data)
2788 {
2789         struct nfs4_closedata *calldata = data;
2790         struct nfs4_state_owner *sp = calldata->state->owner;
2791         struct super_block *sb = calldata->state->inode->i_sb;
2792
2793         if (calldata->roc)
2794                 pnfs_roc_release(calldata->state->inode);
2795         nfs4_put_open_state(calldata->state);
2796         nfs_free_seqid(calldata->arg.seqid);
2797         nfs4_put_state_owner(sp);
2798         nfs_sb_deactive(sb);
2799         kfree(calldata);
2800 }
2801
2802 static void nfs4_close_done(struct rpc_task *task, void *data)
2803 {
2804         struct nfs4_closedata *calldata = data;
2805         struct nfs4_state *state = calldata->state;
2806         struct nfs_server *server = NFS_SERVER(calldata->inode);
2807         nfs4_stateid *res_stateid = NULL;
2808
2809         dprintk("%s: begin!\n", __func__);
2810         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2811                 return;
2812         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2813         /* hmm. we are done with the inode, and in the process of freeing
2814          * the state_owner. we keep this around to process errors
2815          */
2816         switch (task->tk_status) {
2817                 case 0:
2818                         res_stateid = &calldata->res.stateid;
2819                         if (calldata->roc)
2820                                 pnfs_roc_set_barrier(state->inode,
2821                                                      calldata->roc_barrier);
2822                         renew_lease(server, calldata->timestamp);
2823                         break;
2824                 case -NFS4ERR_ADMIN_REVOKED:
2825                 case -NFS4ERR_STALE_STATEID:
2826                 case -NFS4ERR_OLD_STATEID:
2827                 case -NFS4ERR_BAD_STATEID:
2828                 case -NFS4ERR_EXPIRED:
2829                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2830                                                 &state->open_stateid)) {
2831                                 rpc_restart_call_prepare(task);
2832                                 goto out_release;
2833                         }
2834                         if (calldata->arg.fmode == 0)
2835                                 break;
2836                 default:
2837                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2838                                 rpc_restart_call_prepare(task);
2839                                 goto out_release;
2840                         }
2841         }
2842         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2843                         res_stateid, calldata->arg.fmode);
2844 out_release:
2845         nfs_release_seqid(calldata->arg.seqid);
2846         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2847         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2848 }
2849
2850 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2851 {
2852         struct nfs4_closedata *calldata = data;
2853         struct nfs4_state *state = calldata->state;
2854         struct inode *inode = calldata->inode;
2855         bool is_rdonly, is_wronly, is_rdwr;
2856         int call_close = 0;
2857
2858         dprintk("%s: begin!\n", __func__);
2859         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2860                 goto out_wait;
2861
2862         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2863         spin_lock(&state->owner->so_lock);
2864         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2865         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2866         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2867         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2868         /* Calculate the change in open mode */
2869         calldata->arg.fmode = 0;
2870         if (state->n_rdwr == 0) {
2871                 if (state->n_rdonly == 0)
2872                         call_close |= is_rdonly;
2873                 else if (is_rdonly)
2874                         calldata->arg.fmode |= FMODE_READ;
2875                 if (state->n_wronly == 0)
2876                         call_close |= is_wronly;
2877                 else if (is_wronly)
2878                         calldata->arg.fmode |= FMODE_WRITE;
2879         } else if (is_rdwr)
2880                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2881
2882         if (calldata->arg.fmode == 0)
2883                 call_close |= is_rdwr;
2884
2885         if (!nfs4_valid_open_stateid(state))
2886                 call_close = 0;
2887         spin_unlock(&state->owner->so_lock);
2888
2889         if (!call_close) {
2890                 /* Note: exit _without_ calling nfs4_close_done */
2891                 goto out_no_action;
2892         }
2893
2894         if (nfs4_wait_on_layoutreturn(inode, task)) {
2895                 nfs_release_seqid(calldata->arg.seqid);
2896                 goto out_wait;
2897         }
2898
2899         if (calldata->arg.fmode == 0)
2900                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2901         if (calldata->roc)
2902                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2903
2904         calldata->arg.share_access =
2905                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2906                                 calldata->arg.fmode, 0);
2907
2908         nfs_fattr_init(calldata->res.fattr);
2909         calldata->timestamp = jiffies;
2910         if (nfs4_setup_sequence(NFS_SERVER(inode),
2911                                 &calldata->arg.seq_args,
2912                                 &calldata->res.seq_res,
2913                                 task) != 0)
2914                 nfs_release_seqid(calldata->arg.seqid);
2915         dprintk("%s: done!\n", __func__);
2916         return;
2917 out_no_action:
2918         task->tk_action = NULL;
2919 out_wait:
2920         nfs4_sequence_done(task, &calldata->res.seq_res);
2921 }
2922
2923 static const struct rpc_call_ops nfs4_close_ops = {
2924         .rpc_call_prepare = nfs4_close_prepare,
2925         .rpc_call_done = nfs4_close_done,
2926         .rpc_release = nfs4_free_closedata,
2927 };
2928
2929 static bool nfs4_roc(struct inode *inode)
2930 {
2931         if (!nfs_have_layout(inode))
2932                 return false;
2933         return pnfs_roc(inode);
2934 }
2935
2936 /* 
2937  * It is possible for data to be read/written from a mem-mapped file 
2938  * after the sys_close call (which hits the vfs layer as a flush).
2939  * This means that we can't safely call nfsv4 close on a file until 
2940  * the inode is cleared. This in turn means that we are not good
2941  * NFSv4 citizens - we do not indicate to the server to update the file's 
2942  * share state even when we are done with one of the three share 
2943  * stateid's in the inode.
2944  *
2945  * NOTE: Caller must be holding the sp->so_owner semaphore!
2946  */
2947 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2948 {
2949         struct nfs_server *server = NFS_SERVER(state->inode);
2950         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2951         struct nfs4_closedata *calldata;
2952         struct nfs4_state_owner *sp = state->owner;
2953         struct rpc_task *task;
2954         struct rpc_message msg = {
2955                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2956                 .rpc_cred = state->owner->so_cred,
2957         };
2958         struct rpc_task_setup task_setup_data = {
2959                 .rpc_client = server->client,
2960                 .rpc_message = &msg,
2961                 .callback_ops = &nfs4_close_ops,
2962                 .workqueue = nfsiod_workqueue,
2963                 .flags = RPC_TASK_ASYNC,
2964         };
2965         int status = -ENOMEM;
2966
2967         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2968                 &task_setup_data.rpc_client, &msg);
2969
2970         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2971         if (calldata == NULL)
2972                 goto out;
2973         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2974         calldata->inode = state->inode;
2975         calldata->state = state;
2976         calldata->arg.fh = NFS_FH(state->inode);
2977         /* Serialization for the sequence id */
2978         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2979         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2980         if (IS_ERR(calldata->arg.seqid))
2981                 goto out_free_calldata;
2982         calldata->arg.fmode = 0;
2983         calldata->arg.bitmask = server->cache_consistency_bitmask;
2984         calldata->res.fattr = &calldata->fattr;
2985         calldata->res.seqid = calldata->arg.seqid;
2986         calldata->res.server = server;
2987         calldata->roc = nfs4_roc(state->inode);
2988         nfs_sb_active(calldata->inode->i_sb);
2989
2990         msg.rpc_argp = &calldata->arg;
2991         msg.rpc_resp = &calldata->res;
2992         task_setup_data.callback_data = calldata;
2993         task = rpc_run_task(&task_setup_data);
2994         if (IS_ERR(task))
2995                 return PTR_ERR(task);
2996         status = 0;
2997         if (wait)
2998                 status = rpc_wait_for_completion_task(task);
2999         rpc_put_task(task);
3000         return status;
3001 out_free_calldata:
3002         kfree(calldata);
3003 out:
3004         nfs4_put_open_state(state);
3005         nfs4_put_state_owner(sp);
3006         return status;
3007 }
3008
3009 static struct inode *
3010 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3011                 int open_flags, struct iattr *attr, int *opened)
3012 {
3013         struct nfs4_state *state;
3014         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3015
3016         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3017
3018         /* Protect against concurrent sillydeletes */
3019         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3020
3021         nfs4_label_release_security(label);
3022
3023         if (IS_ERR(state))
3024                 return ERR_CAST(state);
3025         return state->inode;
3026 }
3027
3028 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3029 {
3030         if (ctx->state == NULL)
3031                 return;
3032         if (is_sync)
3033                 nfs4_close_sync(ctx->state, ctx->mode);
3034         else
3035                 nfs4_close_state(ctx->state, ctx->mode);
3036 }
3037
3038 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3039 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3040 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3041
3042 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3043 {
3044         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3045         struct nfs4_server_caps_arg args = {
3046                 .fhandle = fhandle,
3047                 .bitmask = bitmask,
3048         };
3049         struct nfs4_server_caps_res res = {};
3050         struct rpc_message msg = {
3051                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3052                 .rpc_argp = &args,
3053                 .rpc_resp = &res,
3054         };
3055         int status;
3056
3057         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3058                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3059                      FATTR4_WORD0_LINK_SUPPORT |
3060                      FATTR4_WORD0_SYMLINK_SUPPORT |
3061                      FATTR4_WORD0_ACLSUPPORT;
3062         if (minorversion)
3063                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3064
3065         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3066         if (status == 0) {
3067                 /* Sanity check the server answers */
3068                 switch (minorversion) {
3069                 case 0:
3070                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3071                         res.attr_bitmask[2] = 0;
3072                         break;
3073                 case 1:
3074                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3075                         break;
3076                 case 2:
3077                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3078                 }
3079                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3080                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3081                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3082                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3083                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3084                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3085                                 NFS_CAP_SECURITY_LABEL);
3086                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3087                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3088                         server->caps |= NFS_CAP_ACLS;
3089                 if (res.has_links != 0)
3090                         server->caps |= NFS_CAP_HARDLINKS;
3091                 if (res.has_symlinks != 0)
3092                         server->caps |= NFS_CAP_SYMLINKS;
3093                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3094                         server->caps |= NFS_CAP_FILEID;
3095                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3096                         server->caps |= NFS_CAP_MODE;
3097                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3098                         server->caps |= NFS_CAP_NLINK;
3099                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3100                         server->caps |= NFS_CAP_OWNER;
3101                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3102                         server->caps |= NFS_CAP_OWNER_GROUP;
3103                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3104                         server->caps |= NFS_CAP_ATIME;
3105                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3106                         server->caps |= NFS_CAP_CTIME;
3107                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3108                         server->caps |= NFS_CAP_MTIME;
3109 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3110                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3111                         server->caps |= NFS_CAP_SECURITY_LABEL;
3112 #endif
3113                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3114                                 sizeof(server->attr_bitmask));
3115                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3116
3117                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3118                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3119                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3120                 server->cache_consistency_bitmask[2] = 0;
3121                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3122                         sizeof(server->exclcreat_bitmask));
3123                 server->acl_bitmask = res.acl_bitmask;
3124                 server->fh_expire_type = res.fh_expire_type;
3125         }
3126
3127         return status;
3128 }
3129
3130 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3131 {
3132         struct nfs4_exception exception = { };
3133         int err;
3134         do {
3135                 err = nfs4_handle_exception(server,
3136                                 _nfs4_server_capabilities(server, fhandle),
3137                                 &exception);
3138         } while (exception.retry);
3139         return err;
3140 }
3141
3142 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3143                 struct nfs_fsinfo *info)
3144 {
3145         u32 bitmask[3];
3146         struct nfs4_lookup_root_arg args = {
3147                 .bitmask = bitmask,
3148         };
3149         struct nfs4_lookup_res res = {
3150                 .server = server,
3151                 .fattr = info->fattr,
3152                 .fh = fhandle,
3153         };
3154         struct rpc_message msg = {
3155                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3156                 .rpc_argp = &args,
3157                 .rpc_resp = &res,
3158         };
3159
3160         bitmask[0] = nfs4_fattr_bitmap[0];
3161         bitmask[1] = nfs4_fattr_bitmap[1];
3162         /*
3163          * Process the label in the upcoming getfattr
3164          */
3165         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3166
3167         nfs_fattr_init(info->fattr);
3168         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3169 }
3170
3171 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3172                 struct nfs_fsinfo *info)
3173 {
3174         struct nfs4_exception exception = { };
3175         int err;
3176         do {
3177                 err = _nfs4_lookup_root(server, fhandle, info);
3178                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3179                 switch (err) {
3180                 case 0:
3181                 case -NFS4ERR_WRONGSEC:
3182                         goto out;
3183                 default:
3184                         err = nfs4_handle_exception(server, err, &exception);
3185                 }
3186         } while (exception.retry);
3187 out:
3188         return err;
3189 }
3190
3191 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3192                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3193 {
3194         struct rpc_auth_create_args auth_args = {
3195                 .pseudoflavor = flavor,
3196         };
3197         struct rpc_auth *auth;
3198         int ret;
3199
3200         auth = rpcauth_create(&auth_args, server->client);
3201         if (IS_ERR(auth)) {
3202                 ret = -EACCES;
3203                 goto out;
3204         }
3205         ret = nfs4_lookup_root(server, fhandle, info);
3206 out:
3207         return ret;
3208 }
3209
3210 /*
3211  * Retry pseudoroot lookup with various security flavors.  We do this when:
3212  *
3213  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3214  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3215  *
3216  * Returns zero on success, or a negative NFS4ERR value, or a
3217  * negative errno value.
3218  */
3219 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3220                               struct nfs_fsinfo *info)
3221 {
3222         /* Per 3530bis 15.33.5 */
3223         static const rpc_authflavor_t flav_array[] = {
3224                 RPC_AUTH_GSS_KRB5P,
3225                 RPC_AUTH_GSS_KRB5I,
3226                 RPC_AUTH_GSS_KRB5,
3227                 RPC_AUTH_UNIX,                  /* courtesy */
3228                 RPC_AUTH_NULL,
3229         };
3230         int status = -EPERM;
3231         size_t i;
3232
3233         if (server->auth_info.flavor_len > 0) {
3234                 /* try each flavor specified by user */
3235                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3236                         status = nfs4_lookup_root_sec(server, fhandle, info,
3237                                                 server->auth_info.flavors[i]);
3238                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3239                                 continue;
3240                         break;
3241                 }
3242         } else {
3243                 /* no flavors specified by user, try default list */
3244                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3245                         status = nfs4_lookup_root_sec(server, fhandle, info,
3246                                                       flav_array[i]);
3247                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3248                                 continue;
3249                         break;
3250                 }
3251         }
3252
3253         /*
3254          * -EACCESS could mean that the user doesn't have correct permissions
3255          * to access the mount.  It could also mean that we tried to mount
3256          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3257          * existing mount programs don't handle -EACCES very well so it should
3258          * be mapped to -EPERM instead.
3259          */
3260         if (status == -EACCES)
3261                 status = -EPERM;
3262         return status;
3263 }
3264
3265 static int nfs4_do_find_root_sec(struct nfs_server *server,
3266                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3267 {
3268         int mv = server->nfs_client->cl_minorversion;
3269         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3270 }
3271
3272 /**
3273  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3274  * @server: initialized nfs_server handle
3275  * @fhandle: we fill in the pseudo-fs root file handle
3276  * @info: we fill in an FSINFO struct
3277  * @auth_probe: probe the auth flavours
3278  *
3279  * Returns zero on success, or a negative errno.
3280  */
3281 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3282                          struct nfs_fsinfo *info,
3283                          bool auth_probe)
3284 {
3285         int status = 0;
3286
3287         if (!auth_probe)
3288                 status = nfs4_lookup_root(server, fhandle, info);
3289
3290         if (auth_probe || status == NFS4ERR_WRONGSEC)
3291                 status = nfs4_do_find_root_sec(server, fhandle, info);
3292
3293         if (status == 0)
3294                 status = nfs4_server_capabilities(server, fhandle);
3295         if (status == 0)
3296                 status = nfs4_do_fsinfo(server, fhandle, info);
3297
3298         return nfs4_map_errors(status);
3299 }
3300
3301 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3302                               struct nfs_fsinfo *info)
3303 {
3304         int error;
3305         struct nfs_fattr *fattr = info->fattr;
3306         struct nfs4_label *label = NULL;
3307
3308         error = nfs4_server_capabilities(server, mntfh);
3309         if (error < 0) {
3310                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3311                 return error;
3312         }
3313
3314         label = nfs4_label_alloc(server, GFP_KERNEL);
3315         if (IS_ERR(label))
3316                 return PTR_ERR(label);
3317
3318         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3319         if (error < 0) {
3320                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3321                 goto err_free_label;
3322         }
3323
3324         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3325             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3326                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3327
3328 err_free_label:
3329         nfs4_label_free(label);
3330
3331         return error;
3332 }
3333
3334 /*
3335  * Get locations and (maybe) other attributes of a referral.
3336  * Note that we'll actually follow the referral later when
3337  * we detect fsid mismatch in inode revalidation
3338  */
3339 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3340                              const struct qstr *name, struct nfs_fattr *fattr,
3341                              struct nfs_fh *fhandle)
3342 {
3343         int status = -ENOMEM;
3344         struct page *page = NULL;
3345         struct nfs4_fs_locations *locations = NULL;
3346
3347         page = alloc_page(GFP_KERNEL);
3348         if (page == NULL)
3349                 goto out;
3350         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3351         if (locations == NULL)
3352                 goto out;
3353
3354         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3355         if (status != 0)
3356                 goto out;
3357
3358         /*
3359          * If the fsid didn't change, this is a migration event, not a
3360          * referral.  Cause us to drop into the exception handler, which
3361          * will kick off migration recovery.
3362          */
3363         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3364                 dprintk("%s: server did not return a different fsid for"
3365                         " a referral at %s\n", __func__, name->name);
3366                 status = -NFS4ERR_MOVED;
3367                 goto out;
3368         }
3369         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3370         nfs_fixup_referral_attributes(&locations->fattr);
3371
3372         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3373         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3374         memset(fhandle, 0, sizeof(struct nfs_fh));
3375 out:
3376         if (page)
3377                 __free_page(page);
3378         kfree(locations);
3379         return status;
3380 }
3381
3382 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3383                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3384 {
3385         struct nfs4_getattr_arg args = {
3386                 .fh = fhandle,
3387                 .bitmask = server->attr_bitmask,
3388         };
3389         struct nfs4_getattr_res res = {
3390                 .fattr = fattr,
3391                 .label = label,
3392                 .server = server,
3393         };
3394         struct rpc_message msg = {
3395                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3396                 .rpc_argp = &args,
3397                 .rpc_resp = &res,
3398         };
3399
3400         args.bitmask = nfs4_bitmask(server, label);
3401
3402         nfs_fattr_init(fattr);
3403         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3404 }
3405
3406 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3407                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3408 {
3409         struct nfs4_exception exception = { };
3410         int err;
3411         do {
3412                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3413                 trace_nfs4_getattr(server, fhandle, fattr, err);
3414                 err = nfs4_handle_exception(server, err,
3415                                 &exception);
3416         } while (exception.retry);
3417         return err;
3418 }
3419
3420 /* 
3421  * The file is not closed if it is opened due to the a request to change
3422  * the size of the file. The open call will not be needed once the
3423  * VFS layer lookup-intents are implemented.
3424  *
3425  * Close is called when the inode is destroyed.
3426  * If we haven't opened the file for O_WRONLY, we
3427  * need to in the size_change case to obtain a stateid.
3428  *
3429  * Got race?
3430  * Because OPEN is always done by name in nfsv4, it is
3431  * possible that we opened a different file by the same
3432  * name.  We can recognize this race condition, but we
3433  * can't do anything about it besides returning an error.
3434  *
3435  * This will be fixed with VFS changes (lookup-intent).
3436  */
3437 static int
3438 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3439                   struct iattr *sattr)
3440 {
3441         struct inode *inode = d_inode(dentry);
3442         struct rpc_cred *cred = NULL;
3443         struct nfs4_state *state = NULL;
3444         struct nfs4_label *label = NULL;
3445         int status;
3446
3447         if (pnfs_ld_layoutret_on_setattr(inode) &&
3448             sattr->ia_valid & ATTR_SIZE &&
3449             sattr->ia_size < i_size_read(inode))
3450                 pnfs_commit_and_return_layout(inode);
3451
3452         nfs_fattr_init(fattr);
3453         
3454         /* Deal with open(O_TRUNC) */
3455         if (sattr->ia_valid & ATTR_OPEN)
3456                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3457
3458         /* Optimization: if the end result is no change, don't RPC */
3459         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3460                 return 0;
3461
3462         /* Search for an existing open(O_WRITE) file */
3463         if (sattr->ia_valid & ATTR_FILE) {
3464                 struct nfs_open_context *ctx;
3465
3466                 ctx = nfs_file_open_context(sattr->ia_file);
3467                 if (ctx) {
3468                         cred = ctx->cred;
3469                         state = ctx->state;
3470                 }
3471         }
3472
3473         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3474         if (IS_ERR(label))
3475                 return PTR_ERR(label);
3476
3477         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3478         if (status == 0) {
3479                 nfs_setattr_update_inode(inode, sattr, fattr);
3480                 nfs_setsecurity(inode, fattr, label);
3481         }
3482         nfs4_label_free(label);
3483         return status;
3484 }
3485
3486 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3487                 const struct qstr *name, struct nfs_fh *fhandle,
3488                 struct nfs_fattr *fattr, struct nfs4_label *label)
3489 {
3490         struct nfs_server *server = NFS_SERVER(dir);
3491         int                    status;
3492         struct nfs4_lookup_arg args = {
3493                 .bitmask = server->attr_bitmask,
3494                 .dir_fh = NFS_FH(dir),
3495                 .name = name,
3496         };
3497         struct nfs4_lookup_res res = {
3498                 .server = server,
3499                 .fattr = fattr,
3500                 .label = label,
3501                 .fh = fhandle,
3502         };
3503         struct rpc_message msg = {
3504                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3505                 .rpc_argp = &args,
3506                 .rpc_resp = &res,
3507         };
3508
3509         args.bitmask = nfs4_bitmask(server, label);
3510
3511         nfs_fattr_init(fattr);
3512
3513         dprintk("NFS call  lookup %s\n", name->name);
3514         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3515         dprintk("NFS reply lookup: %d\n", status);
3516         return status;
3517 }
3518
3519 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3520 {
3521         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3522                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3523         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3524         fattr->nlink = 2;
3525 }
3526
3527 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3528                                    struct qstr *name, struct nfs_fh *fhandle,
3529                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3530 {
3531         struct nfs4_exception exception = { };
3532         struct rpc_clnt *client = *clnt;
3533         int err;
3534         do {
3535                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3536                 trace_nfs4_lookup(dir, name, err);
3537                 switch (err) {
3538                 case -NFS4ERR_BADNAME:
3539                         err = -ENOENT;
3540                         goto out;
3541                 case -NFS4ERR_MOVED:
3542                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3543                         if (err == -NFS4ERR_MOVED)
3544                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3545                         goto out;
3546                 case -NFS4ERR_WRONGSEC:
3547                         err = -EPERM;
3548                         if (client != *clnt)
3549                                 goto out;
3550                         client = nfs4_negotiate_security(client, dir, name);
3551                         if (IS_ERR(client))
3552                                 return PTR_ERR(client);
3553
3554                         exception.retry = 1;
3555                         break;
3556                 default:
3557                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3558                 }
3559         } while (exception.retry);
3560
3561 out:
3562         if (err == 0)
3563                 *clnt = client;
3564         else if (client != *clnt)
3565                 rpc_shutdown_client(client);
3566
3567         return err;
3568 }
3569
3570 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3571                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3572                             struct nfs4_label *label)
3573 {
3574         int status;
3575         struct rpc_clnt *client = NFS_CLIENT(dir);
3576
3577         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3578         if (client != NFS_CLIENT(dir)) {
3579                 rpc_shutdown_client(client);
3580                 nfs_fixup_secinfo_attributes(fattr);
3581         }
3582         return status;
3583 }
3584
3585 struct rpc_clnt *
3586 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3587                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3588 {
3589         struct rpc_clnt *client = NFS_CLIENT(dir);
3590         int status;
3591
3592         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3593         if (status < 0)
3594                 return ERR_PTR(status);
3595         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3596 }
3597
3598 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3599 {
3600         struct nfs_server *server = NFS_SERVER(inode);
3601         struct nfs4_accessargs args = {
3602                 .fh = NFS_FH(inode),
3603                 .bitmask = server->cache_consistency_bitmask,
3604         };
3605         struct nfs4_accessres res = {
3606                 .server = server,
3607         };
3608         struct rpc_message msg = {
3609                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3610                 .rpc_argp = &args,
3611                 .rpc_resp = &res,
3612                 .rpc_cred = entry->cred,
3613         };
3614         int mode = entry->mask;
3615         int status = 0;
3616
3617         /*
3618          * Determine which access bits we want to ask for...
3619          */
3620         if (mode & MAY_READ)
3621                 args.access |= NFS4_ACCESS_READ;
3622         if (S_ISDIR(inode->i_mode)) {
3623                 if (mode & MAY_WRITE)
3624                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3625                 if (mode & MAY_EXEC)
3626                         args.access |= NFS4_ACCESS_LOOKUP;
3627         } else {
3628                 if (mode & MAY_WRITE)
3629                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3630                 if (mode & MAY_EXEC)
3631                         args.access |= NFS4_ACCESS_EXECUTE;
3632         }
3633
3634         res.fattr = nfs_alloc_fattr();
3635         if (res.fattr == NULL)
3636                 return -ENOMEM;
3637
3638         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3639         if (!status) {
3640                 nfs_access_set_mask(entry, res.access);
3641                 nfs_refresh_inode(inode, res.fattr);
3642         }
3643         nfs_free_fattr(res.fattr);
3644         return status;
3645 }
3646
3647 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3648 {
3649         struct nfs4_exception exception = { };
3650         int err;
3651         do {
3652                 err = _nfs4_proc_access(inode, entry);
3653                 trace_nfs4_access(inode, err);
3654                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3655                                 &exception);
3656         } while (exception.retry);
3657         return err;
3658 }
3659
3660 /*
3661  * TODO: For the time being, we don't try to get any attributes
3662  * along with any of the zero-copy operations READ, READDIR,
3663  * READLINK, WRITE.
3664  *
3665  * In the case of the first three, we want to put the GETATTR
3666  * after the read-type operation -- this is because it is hard
3667  * to predict the length of a GETATTR response in v4, and thus
3668  * align the READ data correctly.  This means that the GETATTR
3669  * may end up partially falling into the page cache, and we should
3670  * shift it into the 'tail' of the xdr_buf before processing.
3671  * To do this efficiently, we need to know the total length
3672  * of data received, which doesn't seem to be available outside
3673  * of the RPC layer.
3674  *
3675  * In the case of WRITE, we also want to put the GETATTR after
3676  * the operation -- in this case because we want to make sure
3677  * we get the post-operation mtime and size.
3678  *
3679  * Both of these changes to the XDR layer would in fact be quite
3680  * minor, but I decided to leave them for a subsequent patch.
3681  */
3682 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3683                 unsigned int pgbase, unsigned int pglen)
3684 {
3685         struct nfs4_readlink args = {
3686                 .fh       = NFS_FH(inode),
3687                 .pgbase   = pgbase,
3688                 .pglen    = pglen,
3689                 .pages    = &page,
3690         };
3691         struct nfs4_readlink_res res;
3692         struct rpc_message msg = {
3693                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3694                 .rpc_argp = &args,
3695                 .rpc_resp = &res,
3696         };
3697
3698         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3699 }
3700
3701 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3702                 unsigned int pgbase, unsigned int pglen)
3703 {
3704         struct nfs4_exception exception = { };
3705         int err;
3706         do {
3707                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3708                 trace_nfs4_readlink(inode, err);
3709                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3710                                 &exception);
3711         } while (exception.retry);
3712         return err;
3713 }
3714
3715 /*
3716  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3717  */
3718 static int
3719 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3720                  int flags)
3721 {
3722         struct nfs4_label l, *ilabel = NULL;
3723         struct nfs_open_context *ctx;
3724         struct nfs4_state *state;
3725         int status = 0;
3726
3727         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3728         if (IS_ERR(ctx))
3729                 return PTR_ERR(ctx);
3730
3731         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3732
3733         sattr->ia_mode &= ~current_umask();
3734         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3735         if (IS_ERR(state)) {
3736                 status = PTR_ERR(state);
3737                 goto out;
3738         }
3739 out:
3740         nfs4_label_release_security(ilabel);
3741         put_nfs_open_context(ctx);
3742         return status;
3743 }
3744
3745 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3746 {
3747         struct nfs_server *server = NFS_SERVER(dir);
3748         struct nfs_removeargs args = {
3749                 .fh = NFS_FH(dir),
3750                 .name = *name,
3751         };
3752         struct nfs_removeres res = {
3753                 .server = server,
3754         };
3755         struct rpc_message msg = {
3756                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3757                 .rpc_argp = &args,
3758                 .rpc_resp = &res,
3759         };
3760         int status;
3761
3762         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3763         if (status == 0)
3764                 update_changeattr(dir, &res.cinfo);
3765         return status;
3766 }
3767
3768 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3769 {
3770         struct nfs4_exception exception = { };
3771         int err;
3772         do {
3773                 err = _nfs4_proc_remove(dir, name);
3774                 trace_nfs4_remove(dir, name, err);
3775                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3776                                 &exception);
3777         } while (exception.retry);
3778         return err;
3779 }
3780
3781 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3782 {
3783         struct nfs_server *server = NFS_SERVER(dir);
3784         struct nfs_removeargs *args = msg->rpc_argp;
3785         struct nfs_removeres *res = msg->rpc_resp;
3786
3787         res->server = server;
3788         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3789         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3790
3791         nfs_fattr_init(res->dir_attr);
3792 }
3793
3794 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3795 {
3796         nfs4_setup_sequence(NFS_SERVER(data->dir),
3797                         &data->args.seq_args,
3798                         &data->res.seq_res,
3799                         task);
3800 }
3801
3802 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3803 {
3804         struct nfs_unlinkdata *data = task->tk_calldata;
3805         struct nfs_removeres *res = &data->res;
3806
3807         if (!nfs4_sequence_done(task, &res->seq_res))
3808                 return 0;
3809         if (nfs4_async_handle_error(task, res->server, NULL,
3810                                     &data->timeout) == -EAGAIN)
3811                 return 0;
3812         update_changeattr(dir, &res->cinfo);
3813         return 1;
3814 }
3815
3816 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3817 {
3818         struct nfs_server *server = NFS_SERVER(dir);
3819         struct nfs_renameargs *arg = msg->rpc_argp;
3820         struct nfs_renameres *res = msg->rpc_resp;
3821
3822         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3823         res->server = server;
3824         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3825 }
3826
3827 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3828 {
3829         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3830                         &data->args.seq_args,
3831                         &data->res.seq_res,
3832                         task);
3833 }
3834
3835 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3836                                  struct inode *new_dir)
3837 {
3838         struct nfs_renamedata *data = task->tk_calldata;
3839         struct nfs_renameres *res = &data->res;
3840
3841         if (!nfs4_sequence_done(task, &res->seq_res))
3842                 return 0;
3843         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3844                 return 0;
3845
3846         update_changeattr(old_dir, &res->old_cinfo);
3847         update_changeattr(new_dir, &res->new_cinfo);
3848         return 1;
3849 }
3850
3851 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3852 {
3853         struct nfs_server *server = NFS_SERVER(inode);
3854         struct nfs4_link_arg arg = {
3855                 .fh     = NFS_FH(inode),
3856                 .dir_fh = NFS_FH(dir),
3857                 .name   = name,
3858                 .bitmask = server->attr_bitmask,
3859         };
3860         struct nfs4_link_res res = {
3861                 .server = server,
3862                 .label = NULL,
3863         };
3864         struct rpc_message msg = {
3865                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3866                 .rpc_argp = &arg,
3867                 .rpc_resp = &res,
3868         };
3869         int status = -ENOMEM;
3870
3871         res.fattr = nfs_alloc_fattr();
3872         if (res.fattr == NULL)
3873                 goto out;
3874
3875         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3876         if (IS_ERR(res.label)) {
3877                 status = PTR_ERR(res.label);
3878                 goto out;
3879         }
3880         arg.bitmask = nfs4_bitmask(server, res.label);
3881
3882         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3883         if (!status) {
3884                 update_changeattr(dir, &res.cinfo);
3885                 status = nfs_post_op_update_inode(inode, res.fattr);
3886                 if (!status)
3887                         nfs_setsecurity(inode, res.fattr, res.label);
3888         }
3889
3890
3891         nfs4_label_free(res.label);
3892
3893 out:
3894         nfs_free_fattr(res.fattr);
3895         return status;
3896 }
3897
3898 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3899 {
3900         struct nfs4_exception exception = { };
3901         int err;
3902         do {
3903                 err = nfs4_handle_exception(NFS_SERVER(inode),
3904                                 _nfs4_proc_link(inode, dir, name),
3905                                 &exception);
3906         } while (exception.retry);
3907         return err;
3908 }
3909
3910 struct nfs4_createdata {
3911         struct rpc_message msg;
3912         struct nfs4_create_arg arg;
3913         struct nfs4_create_res res;
3914         struct nfs_fh fh;
3915         struct nfs_fattr fattr;
3916         struct nfs4_label *label;
3917 };
3918
3919 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3920                 struct qstr *name, struct iattr *sattr, u32 ftype)
3921 {
3922         struct nfs4_createdata *data;
3923
3924         data = kzalloc(sizeof(*data), GFP_KERNEL);
3925         if (data != NULL) {
3926                 struct nfs_server *server = NFS_SERVER(dir);
3927
3928                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3929                 if (IS_ERR(data->label))
3930                         goto out_free;
3931
3932                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3933                 data->msg.rpc_argp = &data->arg;
3934                 data->msg.rpc_resp = &data->res;
3935                 data->arg.dir_fh = NFS_FH(dir);
3936                 data->arg.server = server;
3937                 data->arg.name = name;
3938                 data->arg.attrs = sattr;
3939                 data->arg.ftype = ftype;
3940                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3941                 data->res.server = server;
3942                 data->res.fh = &data->fh;
3943                 data->res.fattr = &data->fattr;
3944                 data->res.label = data->label;
3945                 nfs_fattr_init(data->res.fattr);
3946         }
3947         return data;
3948 out_free:
3949         kfree(data);
3950         return NULL;
3951 }
3952
3953 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3954 {
3955         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3956                                     &data->arg.seq_args, &data->res.seq_res, 1);
3957         if (status == 0) {
3958                 update_changeattr(dir, &data->res.dir_cinfo);
3959                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3960         }
3961         return status;
3962 }
3963
3964 static void nfs4_free_createdata(struct nfs4_createdata *data)
3965 {
3966         nfs4_label_free(data->label);
3967         kfree(data);
3968 }
3969
3970 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3971                 struct page *page, unsigned int len, struct iattr *sattr,
3972                 struct nfs4_label *label)
3973 {
3974         struct nfs4_createdata *data;
3975         int status = -ENAMETOOLONG;
3976
3977         if (len > NFS4_MAXPATHLEN)
3978                 goto out;
3979
3980         status = -ENOMEM;
3981         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3982         if (data == NULL)
3983                 goto out;
3984
3985         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3986         data->arg.u.symlink.pages = &page;
3987         data->arg.u.symlink.len = len;
3988         data->arg.label = label;
3989         
3990         status = nfs4_do_create(dir, dentry, data);
3991
3992         nfs4_free_createdata(data);
3993 out:
3994         return status;
3995 }
3996
3997 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3998                 struct page *page, unsigned int len, struct iattr *sattr)
3999 {
4000         struct nfs4_exception exception = { };
4001         struct nfs4_label l, *label = NULL;
4002         int err;
4003
4004         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4005
4006         do {
4007                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4008                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4009                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4010                                 &exception);
4011         } while (exception.retry);
4012
4013         nfs4_label_release_security(label);
4014         return err;
4015 }
4016
4017 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4018                 struct iattr *sattr, struct nfs4_label *label)
4019 {
4020         struct nfs4_createdata *data;
4021         int status = -ENOMEM;
4022
4023         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4024         if (data == NULL)
4025                 goto out;
4026
4027         data->arg.label = label;
4028         status = nfs4_do_create(dir, dentry, data);
4029
4030         nfs4_free_createdata(data);
4031 out:
4032         return status;
4033 }
4034
4035 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4036                 struct iattr *sattr)
4037 {
4038         struct nfs4_exception exception = { };
4039         struct nfs4_label l, *label = NULL;
4040         int err;
4041
4042         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4043
4044         sattr->ia_mode &= ~current_umask();
4045         do {
4046                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4047                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4048                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4049                                 &exception);
4050         } while (exception.retry);
4051         nfs4_label_release_security(label);
4052
4053         return err;
4054 }
4055
4056 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4057                 u64 cookie, struct page **pages, unsigned int count, int plus)
4058 {
4059         struct inode            *dir = d_inode(dentry);
4060         struct nfs4_readdir_arg args = {
4061                 .fh = NFS_FH(dir),
4062                 .pages = pages,
4063                 .pgbase = 0,
4064                 .count = count,
4065                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4066                 .plus = plus,
4067         };
4068         struct nfs4_readdir_res res;
4069         struct rpc_message msg = {
4070                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4071                 .rpc_argp = &args,
4072                 .rpc_resp = &res,
4073                 .rpc_cred = cred,
4074         };
4075         int                     status;
4076
4077         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4078                         dentry,
4079                         (unsigned long long)cookie);
4080         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4081         res.pgbase = args.pgbase;
4082         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4083         if (status >= 0) {
4084                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4085                 status += args.pgbase;
4086         }
4087
4088         nfs_invalidate_atime(dir);
4089
4090         dprintk("%s: returns %d\n", __func__, status);
4091         return status;
4092 }
4093
4094 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4095                 u64 cookie, struct page **pages, unsigned int count, int plus)
4096 {
4097         struct nfs4_exception exception = { };
4098         int err;
4099         do {
4100                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4101                                 pages, count, plus);
4102                 trace_nfs4_readdir(d_inode(dentry), err);
4103                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4104                                 &exception);
4105         } while (exception.retry);
4106         return err;
4107 }
4108
4109 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4110                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4111 {
4112         struct nfs4_createdata *data;
4113         int mode = sattr->ia_mode;
4114         int status = -ENOMEM;
4115
4116         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4117         if (data == NULL)
4118                 goto out;
4119
4120         if (S_ISFIFO(mode))
4121                 data->arg.ftype = NF4FIFO;
4122         else if (S_ISBLK(mode)) {
4123                 data->arg.ftype = NF4BLK;
4124                 data->arg.u.device.specdata1 = MAJOR(rdev);
4125                 data->arg.u.device.specdata2 = MINOR(rdev);
4126         }
4127         else if (S_ISCHR(mode)) {
4128                 data->arg.ftype = NF4CHR;
4129                 data->arg.u.device.specdata1 = MAJOR(rdev);
4130                 data->arg.u.device.specdata2 = MINOR(rdev);
4131         } else if (!S_ISSOCK(mode)) {
4132                 status = -EINVAL;
4133                 goto out_free;
4134         }
4135
4136         data->arg.label = label;
4137         status = nfs4_do_create(dir, dentry, data);
4138 out_free:
4139         nfs4_free_createdata(data);
4140 out:
4141         return status;
4142 }
4143
4144 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4145                 struct iattr *sattr, dev_t rdev)
4146 {
4147         struct nfs4_exception exception = { };
4148         struct nfs4_label l, *label = NULL;
4149         int err;
4150
4151         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4152
4153         sattr->ia_mode &= ~current_umask();
4154         do {
4155                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4156                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4157                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4158                                 &exception);
4159         } while (exception.retry);
4160
4161         nfs4_label_release_security(label);
4162
4163         return err;
4164 }
4165
4166 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4167                  struct nfs_fsstat *fsstat)
4168 {
4169         struct nfs4_statfs_arg args = {
4170                 .fh = fhandle,
4171                 .bitmask = server->attr_bitmask,
4172         };
4173         struct nfs4_statfs_res res = {
4174                 .fsstat = fsstat,
4175         };
4176         struct rpc_message msg = {
4177                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4178                 .rpc_argp = &args,
4179                 .rpc_resp = &res,
4180         };
4181
4182         nfs_fattr_init(fsstat->fattr);
4183         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4184 }
4185
4186 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4187 {
4188         struct nfs4_exception exception = { };
4189         int err;
4190         do {
4191                 err = nfs4_handle_exception(server,
4192                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4193                                 &exception);
4194         } while (exception.retry);
4195         return err;
4196 }
4197
4198 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4199                 struct nfs_fsinfo *fsinfo)
4200 {
4201         struct nfs4_fsinfo_arg args = {
4202                 .fh = fhandle,
4203                 .bitmask = server->attr_bitmask,
4204         };
4205         struct nfs4_fsinfo_res res = {
4206                 .fsinfo = fsinfo,
4207         };
4208         struct rpc_message msg = {
4209                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4210                 .rpc_argp = &args,
4211                 .rpc_resp = &res,
4212         };
4213
4214         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4215 }
4216
4217 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4218 {
4219         struct nfs4_exception exception = { };
4220         unsigned long now = jiffies;
4221         int err;
4222
4223         do {
4224                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4225                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4226                 if (err == 0) {
4227                         struct nfs_client *clp = server->nfs_client;
4228
4229                         spin_lock(&clp->cl_lock);
4230                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4231                         clp->cl_last_renewal = now;
4232                         spin_unlock(&clp->cl_lock);
4233                         break;
4234                 }
4235                 err = nfs4_handle_exception(server, err, &exception);
4236         } while (exception.retry);
4237         return err;
4238 }
4239
4240 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4241 {
4242         int error;
4243
4244         nfs_fattr_init(fsinfo->fattr);
4245         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4246         if (error == 0) {
4247                 /* block layout checks this! */
4248                 server->pnfs_blksize = fsinfo->blksize;
4249                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4250         }
4251
4252         return error;
4253 }
4254
4255 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4256                 struct nfs_pathconf *pathconf)
4257 {
4258         struct nfs4_pathconf_arg args = {
4259                 .fh = fhandle,
4260                 .bitmask = server->attr_bitmask,
4261         };
4262         struct nfs4_pathconf_res res = {
4263                 .pathconf = pathconf,
4264         };
4265         struct rpc_message msg = {
4266                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4267                 .rpc_argp = &args,
4268                 .rpc_resp = &res,
4269         };
4270
4271         /* None of the pathconf attributes are mandatory to implement */
4272         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4273                 memset(pathconf, 0, sizeof(*pathconf));
4274                 return 0;
4275         }
4276
4277         nfs_fattr_init(pathconf->fattr);
4278         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4279 }
4280
4281 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4282                 struct nfs_pathconf *pathconf)
4283 {
4284         struct nfs4_exception exception = { };
4285         int err;
4286
4287         do {
4288                 err = nfs4_handle_exception(server,
4289                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4290                                 &exception);
4291         } while (exception.retry);
4292         return err;
4293 }
4294
4295 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4296                 const struct nfs_open_context *ctx,
4297                 const struct nfs_lock_context *l_ctx,
4298                 fmode_t fmode)
4299 {
4300         const struct nfs_lockowner *lockowner = NULL;
4301
4302         if (l_ctx != NULL)
4303                 lockowner = &l_ctx->lockowner;
4304         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4305 }
4306 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4307
4308 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4309                 const struct nfs_open_context *ctx,
4310                 const struct nfs_lock_context *l_ctx,
4311                 fmode_t fmode)
4312 {
4313         nfs4_stateid current_stateid;
4314
4315         /* If the current stateid represents a lost lock, then exit */
4316         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4317                 return true;
4318         return nfs4_stateid_match(stateid, &current_stateid);
4319 }
4320
4321 static bool nfs4_error_stateid_expired(int err)
4322 {
4323         switch (err) {
4324         case -NFS4ERR_DELEG_REVOKED:
4325         case -NFS4ERR_ADMIN_REVOKED:
4326         case -NFS4ERR_BAD_STATEID:
4327         case -NFS4ERR_STALE_STATEID:
4328         case -NFS4ERR_OLD_STATEID:
4329         case -NFS4ERR_OPENMODE:
4330         case -NFS4ERR_EXPIRED:
4331                 return true;
4332         }
4333         return false;
4334 }
4335
4336 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4337 {
4338         nfs_invalidate_atime(hdr->inode);
4339 }
4340
4341 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4342 {
4343         struct nfs_server *server = NFS_SERVER(hdr->inode);
4344
4345         trace_nfs4_read(hdr, task->tk_status);
4346         if (nfs4_async_handle_error(task, server,
4347                                     hdr->args.context->state,
4348                                     NULL) == -EAGAIN) {
4349                 rpc_restart_call_prepare(task);
4350                 return -EAGAIN;
4351         }
4352
4353         __nfs4_read_done_cb(hdr);
4354         if (task->tk_status > 0)
4355                 renew_lease(server, hdr->timestamp);
4356         return 0;
4357 }
4358
4359 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4360                 struct nfs_pgio_args *args)
4361 {
4362
4363         if (!nfs4_error_stateid_expired(task->tk_status) ||
4364                 nfs4_stateid_is_current(&args->stateid,
4365                                 args->context,
4366                                 args->lock_context,
4367                                 FMODE_READ))
4368                 return false;
4369         rpc_restart_call_prepare(task);
4370         return true;
4371 }
4372
4373 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4374 {
4375
4376         dprintk("--> %s\n", __func__);
4377
4378         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4379                 return -EAGAIN;
4380         if (nfs4_read_stateid_changed(task, &hdr->args))
4381                 return -EAGAIN;
4382         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4383                                     nfs4_read_done_cb(task, hdr);
4384 }
4385
4386 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4387                                  struct rpc_message *msg)
4388 {
4389         hdr->timestamp   = jiffies;
4390         hdr->pgio_done_cb = nfs4_read_done_cb;
4391         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4392         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4393 }
4394
4395 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4396                                       struct nfs_pgio_header *hdr)
4397 {
4398         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4399                         &hdr->args.seq_args,
4400                         &hdr->res.seq_res,
4401                         task))
4402                 return 0;
4403         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4404                                 hdr->args.lock_context,
4405                                 hdr->rw_ops->rw_mode) == -EIO)
4406                 return -EIO;
4407         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4408                 return -EIO;
4409         return 0;
4410 }
4411
4412 static int nfs4_write_done_cb(struct rpc_task *task,
4413                               struct nfs_pgio_header *hdr)
4414 {
4415         struct inode *inode = hdr->inode;
4416
4417         trace_nfs4_write(hdr, task->tk_status);
4418         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4419                                     hdr->args.context->state,
4420                                     NULL) == -EAGAIN) {
4421                 rpc_restart_call_prepare(task);
4422                 return -EAGAIN;
4423         }
4424         if (task->tk_status >= 0) {
4425                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4426                 nfs_writeback_update_inode(hdr);
4427         }
4428         return 0;
4429 }
4430
4431 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4432                 struct nfs_pgio_args *args)
4433 {
4434
4435         if (!nfs4_error_stateid_expired(task->tk_status) ||
4436                 nfs4_stateid_is_current(&args->stateid,
4437                                 args->context,
4438                                 args->lock_context,
4439                                 FMODE_WRITE))
4440                 return false;
4441         rpc_restart_call_prepare(task);
4442         return true;
4443 }
4444
4445 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4446 {
4447         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4448                 return -EAGAIN;
4449         if (nfs4_write_stateid_changed(task, &hdr->args))
4450                 return -EAGAIN;
4451         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4452                 nfs4_write_done_cb(task, hdr);
4453 }
4454
4455 static
4456 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4457 {
4458         /* Don't request attributes for pNFS or O_DIRECT writes */
4459         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4460                 return false;
4461         /* Otherwise, request attributes if and only if we don't hold
4462          * a delegation
4463          */
4464         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4465 }
4466
4467 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4468                                   struct rpc_message *msg)
4469 {
4470         struct nfs_server *server = NFS_SERVER(hdr->inode);
4471
4472         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4473                 hdr->args.bitmask = NULL;
4474                 hdr->res.fattr = NULL;
4475         } else
4476                 hdr->args.bitmask = server->cache_consistency_bitmask;
4477
4478         if (!hdr->pgio_done_cb)
4479                 hdr->pgio_done_cb = nfs4_write_done_cb;
4480         hdr->res.server = server;
4481         hdr->timestamp   = jiffies;
4482
4483         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4484         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4485 }
4486
4487 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4488 {
4489         nfs4_setup_sequence(NFS_SERVER(data->inode),
4490                         &data->args.seq_args,
4491                         &data->res.seq_res,
4492                         task);
4493 }
4494
4495 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4496 {
4497         struct inode *inode = data->inode;
4498
4499         trace_nfs4_commit(data, task->tk_status);
4500         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4501                                     NULL, NULL) == -EAGAIN) {
4502                 rpc_restart_call_prepare(task);
4503                 return -EAGAIN;
4504         }
4505         return 0;
4506 }
4507
4508 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4509 {
4510         if (!nfs4_sequence_done(task, &data->res.seq_res))
4511                 return -EAGAIN;
4512         return data->commit_done_cb(task, data);
4513 }
4514
4515 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4516 {
4517         struct nfs_server *server = NFS_SERVER(data->inode);
4518
4519         if (data->commit_done_cb == NULL)
4520                 data->commit_done_cb = nfs4_commit_done_cb;
4521         data->res.server = server;
4522         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4523         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4524 }
4525
4526 struct nfs4_renewdata {
4527         struct nfs_client       *client;
4528         unsigned long           timestamp;
4529 };
4530
4531 /*
4532  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4533  * standalone procedure for queueing an asynchronous RENEW.
4534  */
4535 static void nfs4_renew_release(void *calldata)
4536 {
4537         struct nfs4_renewdata *data = calldata;
4538         struct nfs_client *clp = data->client;
4539
4540         if (atomic_read(&clp->cl_count) > 1)
4541                 nfs4_schedule_state_renewal(clp);
4542         nfs_put_client(clp);
4543         kfree(data);
4544 }
4545
4546 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4547 {
4548         struct nfs4_renewdata *data = calldata;
4549         struct nfs_client *clp = data->client;
4550         unsigned long timestamp = data->timestamp;
4551
4552         trace_nfs4_renew_async(clp, task->tk_status);
4553         switch (task->tk_status) {
4554         case 0:
4555                 break;
4556         case -NFS4ERR_LEASE_MOVED:
4557                 nfs4_schedule_lease_moved_recovery(clp);
4558                 break;
4559         default:
4560                 /* Unless we're shutting down, schedule state recovery! */
4561                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4562                         return;
4563                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4564                         nfs4_schedule_lease_recovery(clp);
4565                         return;
4566                 }
4567                 nfs4_schedule_path_down_recovery(clp);
4568         }
4569         do_renew_lease(clp, timestamp);
4570 }
4571
4572 static const struct rpc_call_ops nfs4_renew_ops = {
4573         .rpc_call_done = nfs4_renew_done,
4574         .rpc_release = nfs4_renew_release,
4575 };
4576
4577 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4578 {
4579         struct rpc_message msg = {
4580                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4581                 .rpc_argp       = clp,
4582                 .rpc_cred       = cred,
4583         };
4584         struct nfs4_renewdata *data;
4585
4586         if (renew_flags == 0)
4587                 return 0;
4588         if (!atomic_inc_not_zero(&clp->cl_count))
4589                 return -EIO;
4590         data = kmalloc(sizeof(*data), GFP_NOFS);
4591         if (data == NULL)
4592                 return -ENOMEM;
4593         data->client = clp;
4594         data->timestamp = jiffies;
4595         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4596                         &nfs4_renew_ops, data);
4597 }
4598
4599 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4600 {
4601         struct rpc_message msg = {
4602                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4603                 .rpc_argp       = clp,
4604                 .rpc_cred       = cred,
4605         };
4606         unsigned long now = jiffies;
4607         int status;
4608
4609         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4610         if (status < 0)
4611                 return status;
4612         do_renew_lease(clp, now);
4613         return 0;
4614 }
4615
4616 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4617 {
4618         return server->caps & NFS_CAP_ACLS;
4619 }
4620
4621 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4622  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4623  * the stack.
4624  */
4625 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4626
4627 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4628                 struct page **pages)
4629 {
4630         struct page *newpage, **spages;
4631         int rc = 0;
4632         size_t len;
4633         spages = pages;
4634
4635         do {
4636                 len = min_t(size_t, PAGE_SIZE, buflen);
4637                 newpage = alloc_page(GFP_KERNEL);
4638
4639                 if (newpage == NULL)
4640                         goto unwind;
4641                 memcpy(page_address(newpage), buf, len);
4642                 buf += len;
4643                 buflen -= len;
4644                 *pages++ = newpage;
4645                 rc++;
4646         } while (buflen != 0);
4647
4648         return rc;
4649
4650 unwind:
4651         for(; rc > 0; rc--)
4652                 __free_page(spages[rc-1]);
4653         return -ENOMEM;
4654 }
4655
4656 struct nfs4_cached_acl {
4657         int cached;
4658         size_t len;
4659         char data[0];
4660 };
4661
4662 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4663 {
4664         struct nfs_inode *nfsi = NFS_I(inode);
4665
4666         spin_lock(&inode->i_lock);
4667         kfree(nfsi->nfs4_acl);
4668         nfsi->nfs4_acl = acl;
4669         spin_unlock(&inode->i_lock);
4670 }
4671
4672 static void nfs4_zap_acl_attr(struct inode *inode)
4673 {
4674         nfs4_set_cached_acl(inode, NULL);
4675 }
4676
4677 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4678 {
4679         struct nfs_inode *nfsi = NFS_I(inode);
4680         struct nfs4_cached_acl *acl;
4681         int ret = -ENOENT;
4682
4683         spin_lock(&inode->i_lock);
4684         acl = nfsi->nfs4_acl;
4685         if (acl == NULL)
4686                 goto out;
4687         if (buf == NULL) /* user is just asking for length */
4688                 goto out_len;
4689         if (acl->cached == 0)
4690                 goto out;
4691         ret = -ERANGE; /* see getxattr(2) man page */
4692         if (acl->len > buflen)
4693                 goto out;
4694         memcpy(buf, acl->data, acl->len);
4695 out_len:
4696         ret = acl->len;
4697 out:
4698         spin_unlock(&inode->i_lock);
4699         return ret;
4700 }
4701
4702 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4703 {
4704         struct nfs4_cached_acl *acl;
4705         size_t buflen = sizeof(*acl) + acl_len;
4706
4707         if (buflen <= PAGE_SIZE) {
4708                 acl = kmalloc(buflen, GFP_KERNEL);
4709                 if (acl == NULL)
4710                         goto out;
4711                 acl->cached = 1;
4712                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4713         } else {
4714                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4715                 if (acl == NULL)
4716                         goto out;
4717                 acl->cached = 0;
4718         }
4719         acl->len = acl_len;
4720 out:
4721         nfs4_set_cached_acl(inode, acl);
4722 }
4723
4724 /*
4725  * The getxattr API returns the required buffer length when called with a
4726  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4727  * the required buf.  On a NULL buf, we send a page of data to the server
4728  * guessing that the ACL request can be serviced by a page. If so, we cache
4729  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4730  * the cache. If not so, we throw away the page, and cache the required
4731  * length. The next getxattr call will then produce another round trip to
4732  * the server, this time with the input buf of the required size.
4733  */
4734 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4735 {
4736         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4737         struct nfs_getaclargs args = {
4738                 .fh = NFS_FH(inode),
4739                 .acl_pages = pages,
4740                 .acl_len = buflen,
4741         };
4742         struct nfs_getaclres res = {
4743                 .acl_len = buflen,
4744         };
4745         struct rpc_message msg = {
4746                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4747                 .rpc_argp = &args,
4748                 .rpc_resp = &res,
4749         };
4750         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4751         int ret = -ENOMEM, i;
4752
4753         /* As long as we're doing a round trip to the server anyway,
4754          * let's be prepared for a page of acl data. */
4755         if (npages == 0)
4756                 npages = 1;
4757         if (npages > ARRAY_SIZE(pages))
4758                 return -ERANGE;
4759
4760         for (i = 0; i < npages; i++) {
4761                 pages[i] = alloc_page(GFP_KERNEL);
4762                 if (!pages[i])
4763                         goto out_free;
4764         }
4765
4766         /* for decoding across pages */
4767         res.acl_scratch = alloc_page(GFP_KERNEL);
4768         if (!res.acl_scratch)
4769                 goto out_free;
4770
4771         args.acl_len = npages * PAGE_SIZE;
4772
4773         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4774                 __func__, buf, buflen, npages, args.acl_len);
4775         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4776                              &msg, &args.seq_args, &res.seq_res, 0);
4777         if (ret)
4778                 goto out_free;
4779
4780         /* Handle the case where the passed-in buffer is too short */
4781         if (res.acl_flags & NFS4_ACL_TRUNC) {
4782                 /* Did the user only issue a request for the acl length? */
4783                 if (buf == NULL)
4784                         goto out_ok;
4785                 ret = -ERANGE;
4786                 goto out_free;
4787         }
4788         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4789         if (buf) {
4790                 if (res.acl_len > buflen) {
4791                         ret = -ERANGE;
4792                         goto out_free;
4793                 }
4794                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4795         }
4796 out_ok:
4797         ret = res.acl_len;
4798 out_free:
4799         for (i = 0; i < npages; i++)
4800                 if (pages[i])
4801                         __free_page(pages[i]);
4802         if (res.acl_scratch)
4803                 __free_page(res.acl_scratch);
4804         return ret;
4805 }
4806
4807 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4808 {
4809         struct nfs4_exception exception = { };
4810         ssize_t ret;
4811         do {
4812                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4813                 trace_nfs4_get_acl(inode, ret);
4814                 if (ret >= 0)
4815                         break;
4816                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4817         } while (exception.retry);
4818         return ret;
4819 }
4820
4821 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4822 {
4823         struct nfs_server *server = NFS_SERVER(inode);
4824         int ret;
4825
4826         if (!nfs4_server_supports_acls(server))
4827                 return -EOPNOTSUPP;
4828         ret = nfs_revalidate_inode(server, inode);
4829         if (ret < 0)
4830                 return ret;
4831         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4832                 nfs_zap_acl_cache(inode);
4833         ret = nfs4_read_cached_acl(inode, buf, buflen);
4834         if (ret != -ENOENT)
4835                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4836                  * but no cached acl data, just the acl length */
4837                 return ret;
4838         return nfs4_get_acl_uncached(inode, buf, buflen);
4839 }
4840
4841 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4842 {
4843         struct nfs_server *server = NFS_SERVER(inode);
4844         struct page *pages[NFS4ACL_MAXPAGES];
4845         struct nfs_setaclargs arg = {
4846                 .fh             = NFS_FH(inode),
4847                 .acl_pages      = pages,
4848                 .acl_len        = buflen,
4849         };
4850         struct nfs_setaclres res;
4851         struct rpc_message msg = {
4852                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4853                 .rpc_argp       = &arg,
4854                 .rpc_resp       = &res,
4855         };
4856         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4857         int ret, i;
4858
4859         if (!nfs4_server_supports_acls(server))
4860                 return -EOPNOTSUPP;
4861         if (npages > ARRAY_SIZE(pages))
4862                 return -ERANGE;
4863         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4864         if (i < 0)
4865                 return i;
4866         nfs4_inode_return_delegation(inode);
4867         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4868
4869         /*
4870          * Free each page after tx, so the only ref left is
4871          * held by the network stack
4872          */
4873         for (; i > 0; i--)
4874                 put_page(pages[i-1]);
4875
4876         /*
4877          * Acl update can result in inode attribute update.
4878          * so mark the attribute cache invalid.
4879          */
4880         spin_lock(&inode->i_lock);
4881         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4882         spin_unlock(&inode->i_lock);
4883         nfs_access_zap_cache(inode);
4884         nfs_zap_acl_cache(inode);
4885         return ret;
4886 }
4887
4888 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4889 {
4890         struct nfs4_exception exception = { };
4891         int err;
4892         do {
4893                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4894                 trace_nfs4_set_acl(inode, err);
4895                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4896                                 &exception);
4897         } while (exception.retry);
4898         return err;
4899 }
4900
4901 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4902 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4903                                         size_t buflen)
4904 {
4905         struct nfs_server *server = NFS_SERVER(inode);
4906         struct nfs_fattr fattr;
4907         struct nfs4_label label = {0, 0, buflen, buf};
4908
4909         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4910         struct nfs4_getattr_arg arg = {
4911                 .fh             = NFS_FH(inode),
4912                 .bitmask        = bitmask,
4913         };
4914         struct nfs4_getattr_res res = {
4915                 .fattr          = &fattr,
4916                 .label          = &label,
4917                 .server         = server,
4918         };
4919         struct rpc_message msg = {
4920                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4921                 .rpc_argp       = &arg,
4922                 .rpc_resp       = &res,
4923         };
4924         int ret;
4925
4926         nfs_fattr_init(&fattr);
4927
4928         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4929         if (ret)
4930                 return ret;
4931         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4932                 return -ENOENT;
4933         if (buflen < label.len)
4934                 return -ERANGE;
4935         return 0;
4936 }
4937
4938 static int nfs4_get_security_label(struct inode *inode, void *buf,
4939                                         size_t buflen)
4940 {
4941         struct nfs4_exception exception = { };
4942         int err;
4943
4944         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4945                 return -EOPNOTSUPP;
4946
4947         do {
4948                 err = _nfs4_get_security_label(inode, buf, buflen);
4949                 trace_nfs4_get_security_label(inode, err);
4950                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4951                                 &exception);
4952         } while (exception.retry);
4953         return err;
4954 }
4955
4956 static int _nfs4_do_set_security_label(struct inode *inode,
4957                 struct nfs4_label *ilabel,
4958                 struct nfs_fattr *fattr,
4959                 struct nfs4_label *olabel)
4960 {
4961
4962         struct iattr sattr = {0};
4963         struct nfs_server *server = NFS_SERVER(inode);
4964         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4965         struct nfs_setattrargs arg = {
4966                 .fh             = NFS_FH(inode),
4967                 .iap            = &sattr,
4968                 .server         = server,
4969                 .bitmask        = bitmask,
4970                 .label          = ilabel,
4971         };
4972         struct nfs_setattrres res = {
4973                 .fattr          = fattr,
4974                 .label          = olabel,
4975                 .server         = server,
4976         };
4977         struct rpc_message msg = {
4978                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4979                 .rpc_argp       = &arg,
4980                 .rpc_resp       = &res,
4981         };
4982         int status;
4983
4984         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4985
4986         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4987         if (status)
4988                 dprintk("%s failed: %d\n", __func__, status);
4989
4990         return status;
4991 }
4992
4993 static int nfs4_do_set_security_label(struct inode *inode,
4994                 struct nfs4_label *ilabel,
4995                 struct nfs_fattr *fattr,
4996                 struct nfs4_label *olabel)
4997 {
4998         struct nfs4_exception exception = { };
4999         int err;
5000
5001         do {
5002                 err = _nfs4_do_set_security_label(inode, ilabel,
5003                                 fattr, olabel);
5004                 trace_nfs4_set_security_label(inode, err);
5005                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5006                                 &exception);
5007         } while (exception.retry);
5008         return err;
5009 }
5010
5011 static int
5012 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
5013 {
5014         struct nfs4_label ilabel, *olabel = NULL;
5015         struct nfs_fattr fattr;
5016         struct rpc_cred *cred;
5017         struct inode *inode = d_inode(dentry);
5018         int status;
5019
5020         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5021                 return -EOPNOTSUPP;
5022
5023         nfs_fattr_init(&fattr);
5024
5025         ilabel.pi = 0;
5026         ilabel.lfs = 0;
5027         ilabel.label = (char *)buf;
5028         ilabel.len = buflen;
5029
5030         cred = rpc_lookup_cred();
5031         if (IS_ERR(cred))
5032                 return PTR_ERR(cred);
5033
5034         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5035         if (IS_ERR(olabel)) {
5036                 status = -PTR_ERR(olabel);
5037                 goto out;
5038         }
5039
5040         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5041         if (status == 0)
5042                 nfs_setsecurity(inode, &fattr, olabel);
5043
5044         nfs4_label_free(olabel);
5045 out:
5046         put_rpccred(cred);
5047         return status;
5048 }
5049 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5050
5051
5052 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5053                                     nfs4_verifier *bootverf)
5054 {
5055         __be32 verf[2];
5056
5057         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5058                 /* An impossible timestamp guarantees this value
5059                  * will never match a generated boot time. */
5060                 verf[0] = 0;
5061                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5062         } else {
5063                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5064                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5065                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5066         }
5067         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5068 }
5069
5070 static int
5071 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5072 {
5073         size_t len;
5074         char *str;
5075
5076         if (clp->cl_owner_id != NULL)
5077                 return 0;
5078
5079         rcu_read_lock();
5080         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5081                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5082                 1 +
5083                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5084                 1;
5085         rcu_read_unlock();
5086
5087         if (len > NFS4_OPAQUE_LIMIT + 1)
5088                 return -EINVAL;
5089
5090         /*
5091          * Since this string is allocated at mount time, and held until the
5092          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5093          * about a memory-reclaim deadlock.
5094          */
5095         str = kmalloc(len, GFP_KERNEL);
5096         if (!str)
5097                 return -ENOMEM;
5098
5099         rcu_read_lock();
5100         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5101                         clp->cl_ipaddr,
5102                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5103                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5104         rcu_read_unlock();
5105
5106         clp->cl_owner_id = str;
5107         return 0;
5108 }
5109
5110 static int
5111 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5112 {
5113         size_t len;
5114         char *str;
5115
5116         len = 10 + 10 + 1 + 10 + 1 +
5117                 strlen(nfs4_client_id_uniquifier) + 1 +
5118                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5119
5120         if (len > NFS4_OPAQUE_LIMIT + 1)
5121                 return -EINVAL;
5122
5123         /*
5124          * Since this string is allocated at mount time, and held until the
5125          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5126          * about a memory-reclaim deadlock.
5127          */
5128         str = kmalloc(len, GFP_KERNEL);
5129         if (!str)
5130                 return -ENOMEM;
5131
5132         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5133                         clp->rpc_ops->version, clp->cl_minorversion,
5134                         nfs4_client_id_uniquifier,
5135                         clp->cl_rpcclient->cl_nodename);
5136         clp->cl_owner_id = str;
5137         return 0;
5138 }
5139
5140 static int
5141 nfs4_init_uniform_client_string(struct nfs_client *clp)
5142 {
5143         size_t len;
5144         char *str;
5145
5146         if (clp->cl_owner_id != NULL)
5147                 return 0;
5148
5149         if (nfs4_client_id_uniquifier[0] != '\0')
5150                 return nfs4_init_uniquifier_client_string(clp);
5151
5152         len = 10 + 10 + 1 + 10 + 1 +
5153                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5154
5155         if (len > NFS4_OPAQUE_LIMIT + 1)
5156                 return -EINVAL;
5157
5158         /*
5159          * Since this string is allocated at mount time, and held until the
5160          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5161          * about a memory-reclaim deadlock.
5162          */
5163         str = kmalloc(len, GFP_KERNEL);
5164         if (!str)
5165                 return -ENOMEM;
5166
5167         scnprintf(str, len, "Linux NFSv%u.%u %s",
5168                         clp->rpc_ops->version, clp->cl_minorversion,
5169                         clp->cl_rpcclient->cl_nodename);
5170         clp->cl_owner_id = str;
5171         return 0;
5172 }
5173
5174 /*
5175  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5176  * services.  Advertise one based on the address family of the
5177  * clientaddr.
5178  */
5179 static unsigned int
5180 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5181 {
5182         if (strchr(clp->cl_ipaddr, ':') != NULL)
5183                 return scnprintf(buf, len, "tcp6");
5184         else
5185                 return scnprintf(buf, len, "tcp");
5186 }
5187
5188 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5189 {
5190         struct nfs4_setclientid *sc = calldata;
5191
5192         if (task->tk_status == 0)
5193                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5194 }
5195
5196 static const struct rpc_call_ops nfs4_setclientid_ops = {
5197         .rpc_call_done = nfs4_setclientid_done,
5198 };
5199
5200 /**
5201  * nfs4_proc_setclientid - Negotiate client ID
5202  * @clp: state data structure
5203  * @program: RPC program for NFSv4 callback service
5204  * @port: IP port number for NFS4 callback service
5205  * @cred: RPC credential to use for this call
5206  * @res: where to place the result
5207  *
5208  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5209  */
5210 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5211                 unsigned short port, struct rpc_cred *cred,
5212                 struct nfs4_setclientid_res *res)
5213 {
5214         nfs4_verifier sc_verifier;
5215         struct nfs4_setclientid setclientid = {
5216                 .sc_verifier = &sc_verifier,
5217                 .sc_prog = program,
5218                 .sc_clnt = clp,
5219         };
5220         struct rpc_message msg = {
5221                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5222                 .rpc_argp = &setclientid,
5223                 .rpc_resp = res,
5224                 .rpc_cred = cred,
5225         };
5226         struct rpc_task *task;
5227         struct rpc_task_setup task_setup_data = {
5228                 .rpc_client = clp->cl_rpcclient,
5229                 .rpc_message = &msg,
5230                 .callback_ops = &nfs4_setclientid_ops,
5231                 .callback_data = &setclientid,
5232                 .flags = RPC_TASK_TIMEOUT,
5233         };
5234         int status;
5235
5236         /* nfs_client_id4 */
5237         nfs4_init_boot_verifier(clp, &sc_verifier);
5238
5239         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5240                 status = nfs4_init_uniform_client_string(clp);
5241         else
5242                 status = nfs4_init_nonuniform_client_string(clp);
5243
5244         if (status)
5245                 goto out;
5246
5247         /* cb_client4 */
5248         setclientid.sc_netid_len =
5249                                 nfs4_init_callback_netid(clp,
5250                                                 setclientid.sc_netid,
5251                                                 sizeof(setclientid.sc_netid));
5252         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5253                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5254                                 clp->cl_ipaddr, port >> 8, port & 255);
5255
5256         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5257                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5258                 clp->cl_owner_id);
5259         task = rpc_run_task(&task_setup_data);
5260         if (IS_ERR(task)) {
5261                 status = PTR_ERR(task);
5262                 goto out;
5263         }
5264         status = task->tk_status;
5265         if (setclientid.sc_cred) {
5266                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5267                 put_rpccred(setclientid.sc_cred);
5268         }
5269         rpc_put_task(task);
5270 out:
5271         trace_nfs4_setclientid(clp, status);
5272         dprintk("NFS reply setclientid: %d\n", status);
5273         return status;
5274 }
5275
5276 /**
5277  * nfs4_proc_setclientid_confirm - Confirm client ID
5278  * @clp: state data structure
5279  * @res: result of a previous SETCLIENTID
5280  * @cred: RPC credential to use for this call
5281  *
5282  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5283  */
5284 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5285                 struct nfs4_setclientid_res *arg,
5286                 struct rpc_cred *cred)
5287 {
5288         struct rpc_message msg = {
5289                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5290                 .rpc_argp = arg,
5291                 .rpc_cred = cred,
5292         };
5293         int status;
5294
5295         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5296                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5297                 clp->cl_clientid);
5298         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5299         trace_nfs4_setclientid_confirm(clp, status);
5300         dprintk("NFS reply setclientid_confirm: %d\n", status);
5301         return status;
5302 }
5303
5304 struct nfs4_delegreturndata {
5305         struct nfs4_delegreturnargs args;
5306         struct nfs4_delegreturnres res;
5307         struct nfs_fh fh;
5308         nfs4_stateid stateid;
5309         unsigned long timestamp;
5310         struct nfs_fattr fattr;
5311         int rpc_status;
5312         struct inode *inode;
5313         bool roc;
5314         u32 roc_barrier;
5315 };
5316
5317 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5318 {
5319         struct nfs4_delegreturndata *data = calldata;
5320
5321         if (!nfs4_sequence_done(task, &data->res.seq_res))
5322                 return;
5323
5324         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5325         switch (task->tk_status) {
5326         case 0:
5327                 renew_lease(data->res.server, data->timestamp);
5328         case -NFS4ERR_ADMIN_REVOKED:
5329         case -NFS4ERR_DELEG_REVOKED:
5330         case -NFS4ERR_BAD_STATEID:
5331         case -NFS4ERR_OLD_STATEID:
5332         case -NFS4ERR_STALE_STATEID:
5333         case -NFS4ERR_EXPIRED:
5334                 task->tk_status = 0;
5335                 if (data->roc)
5336                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5337                 break;
5338         default:
5339                 if (nfs4_async_handle_error(task, data->res.server,
5340                                             NULL, NULL) == -EAGAIN) {
5341                         rpc_restart_call_prepare(task);
5342                         return;
5343                 }
5344         }
5345         data->rpc_status = task->tk_status;
5346 }
5347
5348 static void nfs4_delegreturn_release(void *calldata)
5349 {
5350         struct nfs4_delegreturndata *data = calldata;
5351         struct inode *inode = data->inode;
5352
5353         if (inode) {
5354                 if (data->roc)
5355                         pnfs_roc_release(inode);
5356                 nfs_iput_and_deactive(inode);
5357         }
5358         kfree(calldata);
5359 }
5360
5361 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5362 {
5363         struct nfs4_delegreturndata *d_data;
5364
5365         d_data = (struct nfs4_delegreturndata *)data;
5366
5367         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5368                 return;
5369
5370         if (d_data->roc)
5371                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5372
5373         nfs4_setup_sequence(d_data->res.server,
5374                         &d_data->args.seq_args,
5375                         &d_data->res.seq_res,
5376                         task);
5377 }
5378
5379 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5380         .rpc_call_prepare = nfs4_delegreturn_prepare,
5381         .rpc_call_done = nfs4_delegreturn_done,
5382         .rpc_release = nfs4_delegreturn_release,
5383 };
5384
5385 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5386 {
5387         struct nfs4_delegreturndata *data;
5388         struct nfs_server *server = NFS_SERVER(inode);
5389         struct rpc_task *task;
5390         struct rpc_message msg = {
5391                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5392                 .rpc_cred = cred,
5393         };
5394         struct rpc_task_setup task_setup_data = {
5395                 .rpc_client = server->client,
5396                 .rpc_message = &msg,
5397                 .callback_ops = &nfs4_delegreturn_ops,
5398                 .flags = RPC_TASK_ASYNC,
5399         };
5400         int status = 0;
5401
5402         data = kzalloc(sizeof(*data), GFP_NOFS);
5403         if (data == NULL)
5404                 return -ENOMEM;
5405         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5406
5407         nfs4_state_protect(server->nfs_client,
5408                         NFS_SP4_MACH_CRED_CLEANUP,
5409                         &task_setup_data.rpc_client, &msg);
5410
5411         data->args.fhandle = &data->fh;
5412         data->args.stateid = &data->stateid;
5413         data->args.bitmask = server->cache_consistency_bitmask;
5414         nfs_copy_fh(&data->fh, NFS_FH(inode));
5415         nfs4_stateid_copy(&data->stateid, stateid);
5416         data->res.fattr = &data->fattr;
5417         data->res.server = server;
5418         nfs_fattr_init(data->res.fattr);
5419         data->timestamp = jiffies;
5420         data->rpc_status = 0;
5421         data->inode = nfs_igrab_and_active(inode);
5422         if (data->inode)
5423                 data->roc = nfs4_roc(inode);
5424
5425         task_setup_data.callback_data = data;
5426         msg.rpc_argp = &data->args;
5427         msg.rpc_resp = &data->res;
5428         task = rpc_run_task(&task_setup_data);
5429         if (IS_ERR(task))
5430                 return PTR_ERR(task);
5431         if (!issync)
5432                 goto out;
5433         status = nfs4_wait_for_completion_rpc_task(task);
5434         if (status != 0)
5435                 goto out;
5436         status = data->rpc_status;
5437         if (status == 0)
5438                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5439         else
5440                 nfs_refresh_inode(inode, &data->fattr);
5441 out:
5442         rpc_put_task(task);
5443         return status;
5444 }
5445
5446 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5447 {
5448         struct nfs_server *server = NFS_SERVER(inode);
5449         struct nfs4_exception exception = { };
5450         int err;
5451         do {
5452                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5453                 trace_nfs4_delegreturn(inode, stateid, err);
5454                 switch (err) {
5455                         case -NFS4ERR_STALE_STATEID:
5456                         case -NFS4ERR_EXPIRED:
5457                         case 0:
5458                                 return 0;
5459                 }
5460                 err = nfs4_handle_exception(server, err, &exception);
5461         } while (exception.retry);
5462         return err;
5463 }
5464
5465 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5466 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5467
5468 /* 
5469  * sleep, with exponential backoff, and retry the LOCK operation. 
5470  */
5471 static unsigned long
5472 nfs4_set_lock_task_retry(unsigned long timeout)
5473 {
5474         freezable_schedule_timeout_killable_unsafe(timeout);
5475         timeout <<= 1;
5476         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5477                 return NFS4_LOCK_MAXTIMEOUT;
5478         return timeout;
5479 }
5480
5481 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5482 {
5483         struct inode *inode = state->inode;
5484         struct nfs_server *server = NFS_SERVER(inode);
5485         struct nfs_client *clp = server->nfs_client;
5486         struct nfs_lockt_args arg = {
5487                 .fh = NFS_FH(inode),
5488                 .fl = request,
5489         };
5490         struct nfs_lockt_res res = {
5491                 .denied = request,
5492         };
5493         struct rpc_message msg = {
5494                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5495                 .rpc_argp       = &arg,
5496                 .rpc_resp       = &res,
5497                 .rpc_cred       = state->owner->so_cred,
5498         };
5499         struct nfs4_lock_state *lsp;
5500         int status;
5501
5502         arg.lock_owner.clientid = clp->cl_clientid;
5503         status = nfs4_set_lock_state(state, request);
5504         if (status != 0)
5505                 goto out;
5506         lsp = request->fl_u.nfs4_fl.owner;
5507         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5508         arg.lock_owner.s_dev = server->s_dev;
5509         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5510         switch (status) {
5511                 case 0:
5512                         request->fl_type = F_UNLCK;
5513                         break;
5514                 case -NFS4ERR_DENIED:
5515                         status = 0;
5516         }
5517         request->fl_ops->fl_release_private(request);
5518         request->fl_ops = NULL;
5519 out:
5520         return status;
5521 }
5522
5523 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5524 {
5525         struct nfs4_exception exception = { };
5526         int err;
5527
5528         do {
5529                 err = _nfs4_proc_getlk(state, cmd, request);
5530                 trace_nfs4_get_lock(request, state, cmd, err);
5531                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5532                                 &exception);
5533         } while (exception.retry);
5534         return err;
5535 }
5536
5537 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5538 {
5539         return locks_lock_inode_wait(inode, fl);
5540 }
5541
5542 struct nfs4_unlockdata {
5543         struct nfs_locku_args arg;
5544         struct nfs_locku_res res;
5545         struct nfs4_lock_state *lsp;
5546         struct nfs_open_context *ctx;
5547         struct file_lock fl;
5548         struct nfs_server *server;
5549         unsigned long timestamp;
5550 };
5551
5552 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5553                 struct nfs_open_context *ctx,
5554                 struct nfs4_lock_state *lsp,
5555                 struct nfs_seqid *seqid)
5556 {
5557         struct nfs4_unlockdata *p;
5558         struct inode *inode = lsp->ls_state->inode;
5559
5560         p = kzalloc(sizeof(*p), GFP_NOFS);
5561         if (p == NULL)
5562                 return NULL;
5563         p->arg.fh = NFS_FH(inode);
5564         p->arg.fl = &p->fl;
5565         p->arg.seqid = seqid;
5566         p->res.seqid = seqid;
5567         p->lsp = lsp;
5568         atomic_inc(&lsp->ls_count);
5569         /* Ensure we don't close file until we're done freeing locks! */
5570         p->ctx = get_nfs_open_context(ctx);
5571         memcpy(&p->fl, fl, sizeof(p->fl));
5572         p->server = NFS_SERVER(inode);
5573         return p;
5574 }
5575
5576 static void nfs4_locku_release_calldata(void *data)
5577 {
5578         struct nfs4_unlockdata *calldata = data;
5579         nfs_free_seqid(calldata->arg.seqid);
5580         nfs4_put_lock_state(calldata->lsp);
5581         put_nfs_open_context(calldata->ctx);
5582         kfree(calldata);
5583 }
5584
5585 static void nfs4_locku_done(struct rpc_task *task, void *data)
5586 {
5587         struct nfs4_unlockdata *calldata = data;
5588
5589         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5590                 return;
5591         switch (task->tk_status) {
5592                 case 0:
5593                         renew_lease(calldata->server, calldata->timestamp);
5594                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5595                         if (nfs4_update_lock_stateid(calldata->lsp,
5596                                         &calldata->res.stateid))
5597                                 break;
5598                 case -NFS4ERR_BAD_STATEID:
5599                 case -NFS4ERR_OLD_STATEID:
5600                 case -NFS4ERR_STALE_STATEID:
5601                 case -NFS4ERR_EXPIRED:
5602                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5603                                                 &calldata->lsp->ls_stateid))
5604                                 rpc_restart_call_prepare(task);
5605                         break;
5606                 default:
5607                         if (nfs4_async_handle_error(task, calldata->server,
5608                                                     NULL, NULL) == -EAGAIN)
5609                                 rpc_restart_call_prepare(task);
5610         }
5611         nfs_release_seqid(calldata->arg.seqid);
5612 }
5613
5614 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5615 {
5616         struct nfs4_unlockdata *calldata = data;
5617
5618         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5619                 goto out_wait;
5620         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5621         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5622                 /* Note: exit _without_ running nfs4_locku_done */
5623                 goto out_no_action;
5624         }
5625         calldata->timestamp = jiffies;
5626         if (nfs4_setup_sequence(calldata->server,
5627                                 &calldata->arg.seq_args,
5628                                 &calldata->res.seq_res,
5629                                 task) != 0)
5630                 nfs_release_seqid(calldata->arg.seqid);
5631         return;
5632 out_no_action:
5633         task->tk_action = NULL;
5634 out_wait:
5635         nfs4_sequence_done(task, &calldata->res.seq_res);
5636 }
5637
5638 static const struct rpc_call_ops nfs4_locku_ops = {
5639         .rpc_call_prepare = nfs4_locku_prepare,
5640         .rpc_call_done = nfs4_locku_done,
5641         .rpc_release = nfs4_locku_release_calldata,
5642 };
5643
5644 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5645                 struct nfs_open_context *ctx,
5646                 struct nfs4_lock_state *lsp,
5647                 struct nfs_seqid *seqid)
5648 {
5649         struct nfs4_unlockdata *data;
5650         struct rpc_message msg = {
5651                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5652                 .rpc_cred = ctx->cred,
5653         };
5654         struct rpc_task_setup task_setup_data = {
5655                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5656                 .rpc_message = &msg,
5657                 .callback_ops = &nfs4_locku_ops,
5658                 .workqueue = nfsiod_workqueue,
5659                 .flags = RPC_TASK_ASYNC,
5660         };
5661
5662         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5663                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5664
5665         /* Ensure this is an unlock - when canceling a lock, the
5666          * canceled lock is passed in, and it won't be an unlock.
5667          */
5668         fl->fl_type = F_UNLCK;
5669
5670         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5671         if (data == NULL) {
5672                 nfs_free_seqid(seqid);
5673                 return ERR_PTR(-ENOMEM);
5674         }
5675
5676         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5677         msg.rpc_argp = &data->arg;
5678         msg.rpc_resp = &data->res;
5679         task_setup_data.callback_data = data;
5680         return rpc_run_task(&task_setup_data);
5681 }
5682
5683 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5684 {
5685         struct inode *inode = state->inode;
5686         struct nfs4_state_owner *sp = state->owner;
5687         struct nfs_inode *nfsi = NFS_I(inode);
5688         struct nfs_seqid *seqid;
5689         struct nfs4_lock_state *lsp;
5690         struct rpc_task *task;
5691         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5692         int status = 0;
5693         unsigned char fl_flags = request->fl_flags;
5694
5695         status = nfs4_set_lock_state(state, request);
5696         /* Unlock _before_ we do the RPC call */
5697         request->fl_flags |= FL_EXISTS;
5698         /* Exclude nfs_delegation_claim_locks() */
5699         mutex_lock(&sp->so_delegreturn_mutex);
5700         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5701         down_read(&nfsi->rwsem);
5702         if (do_vfs_lock(inode, request) == -ENOENT) {
5703                 up_read(&nfsi->rwsem);
5704                 mutex_unlock(&sp->so_delegreturn_mutex);
5705                 goto out;
5706         }
5707         up_read(&nfsi->rwsem);
5708         mutex_unlock(&sp->so_delegreturn_mutex);
5709         if (status != 0)
5710                 goto out;
5711         /* Is this a delegated lock? */
5712         lsp = request->fl_u.nfs4_fl.owner;
5713         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5714                 goto out;
5715         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5716         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5717         status = -ENOMEM;
5718         if (IS_ERR(seqid))
5719                 goto out;
5720         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5721         status = PTR_ERR(task);
5722         if (IS_ERR(task))
5723                 goto out;
5724         status = nfs4_wait_for_completion_rpc_task(task);
5725         rpc_put_task(task);
5726 out:
5727         request->fl_flags = fl_flags;
5728         trace_nfs4_unlock(request, state, F_SETLK, status);
5729         return status;
5730 }
5731
5732 struct nfs4_lockdata {
5733         struct nfs_lock_args arg;
5734         struct nfs_lock_res res;
5735         struct nfs4_lock_state *lsp;
5736         struct nfs_open_context *ctx;
5737         struct file_lock fl;
5738         unsigned long timestamp;
5739         int rpc_status;
5740         int cancelled;
5741         struct nfs_server *server;
5742 };
5743
5744 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5745                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5746                 gfp_t gfp_mask)
5747 {
5748         struct nfs4_lockdata *p;
5749         struct inode *inode = lsp->ls_state->inode;
5750         struct nfs_server *server = NFS_SERVER(inode);
5751         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5752
5753         p = kzalloc(sizeof(*p), gfp_mask);
5754         if (p == NULL)
5755                 return NULL;
5756
5757         p->arg.fh = NFS_FH(inode);
5758         p->arg.fl = &p->fl;
5759         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5760         if (IS_ERR(p->arg.open_seqid))
5761                 goto out_free;
5762         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5763         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5764         if (IS_ERR(p->arg.lock_seqid))
5765                 goto out_free_seqid;
5766         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5767         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5768         p->arg.lock_owner.s_dev = server->s_dev;
5769         p->res.lock_seqid = p->arg.lock_seqid;
5770         p->lsp = lsp;
5771         p->server = server;
5772         atomic_inc(&lsp->ls_count);
5773         p->ctx = get_nfs_open_context(ctx);
5774         get_file(fl->fl_file);
5775         memcpy(&p->fl, fl, sizeof(p->fl));
5776         return p;
5777 out_free_seqid:
5778         nfs_free_seqid(p->arg.open_seqid);
5779 out_free:
5780         kfree(p);
5781         return NULL;
5782 }
5783
5784 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5785 {
5786         struct nfs4_lockdata *data = calldata;
5787         struct nfs4_state *state = data->lsp->ls_state;
5788
5789         dprintk("%s: begin!\n", __func__);
5790         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5791                 goto out_wait;
5792         /* Do we need to do an open_to_lock_owner? */
5793         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5794                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5795                         goto out_release_lock_seqid;
5796                 }
5797                 nfs4_stateid_copy(&data->arg.open_stateid,
5798                                 &state->open_stateid);
5799                 data->arg.new_lock_owner = 1;
5800                 data->res.open_seqid = data->arg.open_seqid;
5801         } else {
5802                 data->arg.new_lock_owner = 0;
5803                 nfs4_stateid_copy(&data->arg.lock_stateid,
5804                                 &data->lsp->ls_stateid);
5805         }
5806         if (!nfs4_valid_open_stateid(state)) {
5807                 data->rpc_status = -EBADF;
5808                 task->tk_action = NULL;
5809                 goto out_release_open_seqid;
5810         }
5811         data->timestamp = jiffies;
5812         if (nfs4_setup_sequence(data->server,
5813                                 &data->arg.seq_args,
5814                                 &data->res.seq_res,
5815                                 task) == 0)
5816                 return;
5817 out_release_open_seqid:
5818         nfs_release_seqid(data->arg.open_seqid);
5819 out_release_lock_seqid:
5820         nfs_release_seqid(data->arg.lock_seqid);
5821 out_wait:
5822         nfs4_sequence_done(task, &data->res.seq_res);
5823         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5824 }
5825
5826 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5827 {
5828         struct nfs4_lockdata *data = calldata;
5829         struct nfs4_lock_state *lsp = data->lsp;
5830
5831         dprintk("%s: begin!\n", __func__);
5832
5833         if (!nfs4_sequence_done(task, &data->res.seq_res))
5834                 return;
5835
5836         data->rpc_status = task->tk_status;
5837         switch (task->tk_status) {
5838         case 0:
5839                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5840                                 data->timestamp);
5841                 if (data->arg.new_lock) {
5842                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5843                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5844                                 rpc_restart_call_prepare(task);
5845                                 break;
5846                         }
5847                 }
5848                 if (data->arg.new_lock_owner != 0) {
5849                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5850                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5851                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5852                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5853                         rpc_restart_call_prepare(task);
5854                 break;
5855         case -NFS4ERR_BAD_STATEID:
5856         case -NFS4ERR_OLD_STATEID:
5857         case -NFS4ERR_STALE_STATEID:
5858         case -NFS4ERR_EXPIRED:
5859                 if (data->arg.new_lock_owner != 0) {
5860                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5861                                                 &lsp->ls_state->open_stateid))
5862                                 rpc_restart_call_prepare(task);
5863                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5864                                                 &lsp->ls_stateid))
5865                                 rpc_restart_call_prepare(task);
5866         }
5867         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5868 }
5869
5870 static void nfs4_lock_release(void *calldata)
5871 {
5872         struct nfs4_lockdata *data = calldata;
5873
5874         dprintk("%s: begin!\n", __func__);
5875         nfs_free_seqid(data->arg.open_seqid);
5876         if (data->cancelled != 0) {
5877                 struct rpc_task *task;
5878                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5879                                 data->arg.lock_seqid);
5880                 if (!IS_ERR(task))
5881                         rpc_put_task_async(task);
5882                 dprintk("%s: cancelling lock!\n", __func__);
5883         } else
5884                 nfs_free_seqid(data->arg.lock_seqid);
5885         nfs4_put_lock_state(data->lsp);
5886         put_nfs_open_context(data->ctx);
5887         fput(data->fl.fl_file);
5888         kfree(data);
5889         dprintk("%s: done!\n", __func__);
5890 }
5891
5892 static const struct rpc_call_ops nfs4_lock_ops = {
5893         .rpc_call_prepare = nfs4_lock_prepare,
5894         .rpc_call_done = nfs4_lock_done,
5895         .rpc_release = nfs4_lock_release,
5896 };
5897
5898 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5899 {
5900         switch (error) {
5901         case -NFS4ERR_ADMIN_REVOKED:
5902         case -NFS4ERR_BAD_STATEID:
5903                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5904                 if (new_lock_owner != 0 ||
5905                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5906                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5907                 break;
5908         case -NFS4ERR_STALE_STATEID:
5909                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5910         case -NFS4ERR_EXPIRED:
5911                 nfs4_schedule_lease_recovery(server->nfs_client);
5912         };
5913 }
5914
5915 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5916 {
5917         struct nfs4_lockdata *data;
5918         struct rpc_task *task;
5919         struct rpc_message msg = {
5920                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5921                 .rpc_cred = state->owner->so_cred,
5922         };
5923         struct rpc_task_setup task_setup_data = {
5924                 .rpc_client = NFS_CLIENT(state->inode),
5925                 .rpc_message = &msg,
5926                 .callback_ops = &nfs4_lock_ops,
5927                 .workqueue = nfsiod_workqueue,
5928                 .flags = RPC_TASK_ASYNC,
5929         };
5930         int ret;
5931
5932         dprintk("%s: begin!\n", __func__);
5933         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5934                         fl->fl_u.nfs4_fl.owner,
5935                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5936         if (data == NULL)
5937                 return -ENOMEM;
5938         if (IS_SETLKW(cmd))
5939                 data->arg.block = 1;
5940         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5941         msg.rpc_argp = &data->arg;
5942         msg.rpc_resp = &data->res;
5943         task_setup_data.callback_data = data;
5944         if (recovery_type > NFS_LOCK_NEW) {
5945                 if (recovery_type == NFS_LOCK_RECLAIM)
5946                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5947                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5948         } else
5949                 data->arg.new_lock = 1;
5950         task = rpc_run_task(&task_setup_data);
5951         if (IS_ERR(task))
5952                 return PTR_ERR(task);
5953         ret = nfs4_wait_for_completion_rpc_task(task);
5954         if (ret == 0) {
5955                 ret = data->rpc_status;
5956                 if (ret)
5957                         nfs4_handle_setlk_error(data->server, data->lsp,
5958                                         data->arg.new_lock_owner, ret);
5959         } else
5960                 data->cancelled = 1;
5961         rpc_put_task(task);
5962         dprintk("%s: done, ret = %d!\n", __func__, ret);
5963         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
5964         return ret;
5965 }
5966
5967 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5968 {
5969         struct nfs_server *server = NFS_SERVER(state->inode);
5970         struct nfs4_exception exception = {
5971                 .inode = state->inode,
5972         };
5973         int err;
5974
5975         do {
5976                 /* Cache the lock if possible... */
5977                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5978                         return 0;
5979                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5980                 if (err != -NFS4ERR_DELAY)
5981                         break;
5982                 nfs4_handle_exception(server, err, &exception);
5983         } while (exception.retry);
5984         return err;
5985 }
5986
5987 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5988 {
5989         struct nfs_server *server = NFS_SERVER(state->inode);
5990         struct nfs4_exception exception = {
5991                 .inode = state->inode,
5992         };
5993         int err;
5994
5995         err = nfs4_set_lock_state(state, request);
5996         if (err != 0)
5997                 return err;
5998         if (!recover_lost_locks) {
5999                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6000                 return 0;
6001         }
6002         do {
6003                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6004                         return 0;
6005                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6006                 switch (err) {
6007                 default:
6008                         goto out;
6009                 case -NFS4ERR_GRACE:
6010                 case -NFS4ERR_DELAY:
6011                         nfs4_handle_exception(server, err, &exception);
6012                         err = 0;
6013                 }
6014         } while (exception.retry);
6015 out:
6016         return err;
6017 }
6018
6019 #if defined(CONFIG_NFS_V4_1)
6020 /**
6021  * nfs41_check_expired_locks - possibly free a lock stateid
6022  *
6023  * @state: NFSv4 state for an inode
6024  *
6025  * Returns NFS_OK if recovery for this stateid is now finished.
6026  * Otherwise a negative NFS4ERR value is returned.
6027  */
6028 static int nfs41_check_expired_locks(struct nfs4_state *state)
6029 {
6030         int status, ret = -NFS4ERR_BAD_STATEID;
6031         struct nfs4_lock_state *lsp;
6032         struct nfs_server *server = NFS_SERVER(state->inode);
6033
6034         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6035                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6036                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6037
6038                         status = nfs41_test_stateid(server,
6039                                         &lsp->ls_stateid,
6040                                         cred);
6041                         trace_nfs4_test_lock_stateid(state, lsp, status);
6042                         if (status != NFS_OK) {
6043                                 /* Free the stateid unless the server
6044                                  * informs us the stateid is unrecognized. */
6045                                 if (status != -NFS4ERR_BAD_STATEID)
6046                                         nfs41_free_stateid(server,
6047                                                         &lsp->ls_stateid,
6048                                                         cred);
6049                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6050                                 ret = status;
6051                         }
6052                 }
6053         };
6054
6055         return ret;
6056 }
6057
6058 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6059 {
6060         int status = NFS_OK;
6061
6062         if (test_bit(LK_STATE_IN_USE, &state->flags))
6063                 status = nfs41_check_expired_locks(state);
6064         if (status != NFS_OK)
6065                 status = nfs4_lock_expired(state, request);
6066         return status;
6067 }
6068 #endif
6069
6070 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6071 {
6072         struct nfs_inode *nfsi = NFS_I(state->inode);
6073         struct nfs4_state_owner *sp = state->owner;
6074         unsigned char fl_flags = request->fl_flags;
6075         int status = -ENOLCK;
6076
6077         if ((fl_flags & FL_POSIX) &&
6078                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6079                 goto out;
6080         /* Is this a delegated open? */
6081         status = nfs4_set_lock_state(state, request);
6082         if (status != 0)
6083                 goto out;
6084         request->fl_flags |= FL_ACCESS;
6085         status = do_vfs_lock(state->inode, request);
6086         if (status < 0)
6087                 goto out;
6088         mutex_lock(&sp->so_delegreturn_mutex);
6089         down_read(&nfsi->rwsem);
6090         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6091                 /* Yes: cache locks! */
6092                 /* ...but avoid races with delegation recall... */
6093                 request->fl_flags = fl_flags & ~FL_SLEEP;
6094                 status = do_vfs_lock(state->inode, request);
6095                 up_read(&nfsi->rwsem);
6096                 mutex_unlock(&sp->so_delegreturn_mutex);
6097                 goto out;
6098         }
6099         up_read(&nfsi->rwsem);
6100         mutex_unlock(&sp->so_delegreturn_mutex);
6101         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6102 out:
6103         request->fl_flags = fl_flags;
6104         return status;
6105 }
6106
6107 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6108 {
6109         struct nfs4_exception exception = {
6110                 .state = state,
6111                 .inode = state->inode,
6112         };
6113         int err;
6114
6115         do {
6116                 err = _nfs4_proc_setlk(state, cmd, request);
6117                 if (err == -NFS4ERR_DENIED)
6118                         err = -EAGAIN;
6119                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6120                                 err, &exception);
6121         } while (exception.retry);
6122         return err;
6123 }
6124
6125 static int
6126 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6127 {
6128         struct nfs_open_context *ctx;
6129         struct nfs4_state *state;
6130         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6131         int status;
6132
6133         /* verify open state */
6134         ctx = nfs_file_open_context(filp);
6135         state = ctx->state;
6136
6137         if (request->fl_start < 0 || request->fl_end < 0)
6138                 return -EINVAL;
6139
6140         if (IS_GETLK(cmd)) {
6141                 if (state != NULL)
6142                         return nfs4_proc_getlk(state, F_GETLK, request);
6143                 return 0;
6144         }
6145
6146         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6147                 return -EINVAL;
6148
6149         if (request->fl_type == F_UNLCK) {
6150                 if (state != NULL)
6151                         return nfs4_proc_unlck(state, cmd, request);
6152                 return 0;
6153         }
6154
6155         if (state == NULL)
6156                 return -ENOLCK;
6157         /*
6158          * Don't rely on the VFS having checked the file open mode,
6159          * since it won't do this for flock() locks.
6160          */
6161         switch (request->fl_type) {
6162         case F_RDLCK:
6163                 if (!(filp->f_mode & FMODE_READ))
6164                         return -EBADF;
6165                 break;
6166         case F_WRLCK:
6167                 if (!(filp->f_mode & FMODE_WRITE))
6168                         return -EBADF;
6169         }
6170
6171         do {
6172                 status = nfs4_proc_setlk(state, cmd, request);
6173                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6174                         break;
6175                 timeout = nfs4_set_lock_task_retry(timeout);
6176                 status = -ERESTARTSYS;
6177                 if (signalled())
6178                         break;
6179         } while(status < 0);
6180         return status;
6181 }
6182
6183 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6184 {
6185         struct nfs_server *server = NFS_SERVER(state->inode);
6186         int err;
6187
6188         err = nfs4_set_lock_state(state, fl);
6189         if (err != 0)
6190                 return err;
6191         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6192         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6193 }
6194
6195 struct nfs_release_lockowner_data {
6196         struct nfs4_lock_state *lsp;
6197         struct nfs_server *server;
6198         struct nfs_release_lockowner_args args;
6199         struct nfs_release_lockowner_res res;
6200         unsigned long timestamp;
6201 };
6202
6203 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6204 {
6205         struct nfs_release_lockowner_data *data = calldata;
6206         struct nfs_server *server = data->server;
6207         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6208                              &data->args.seq_args, &data->res.seq_res, task);
6209         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6210         data->timestamp = jiffies;
6211 }
6212
6213 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6214 {
6215         struct nfs_release_lockowner_data *data = calldata;
6216         struct nfs_server *server = data->server;
6217
6218         nfs40_sequence_done(task, &data->res.seq_res);
6219
6220         switch (task->tk_status) {
6221         case 0:
6222                 renew_lease(server, data->timestamp);
6223                 break;
6224         case -NFS4ERR_STALE_CLIENTID:
6225         case -NFS4ERR_EXPIRED:
6226                 nfs4_schedule_lease_recovery(server->nfs_client);
6227                 break;
6228         case -NFS4ERR_LEASE_MOVED:
6229         case -NFS4ERR_DELAY:
6230                 if (nfs4_async_handle_error(task, server,
6231                                             NULL, NULL) == -EAGAIN)
6232                         rpc_restart_call_prepare(task);
6233         }
6234 }
6235
6236 static void nfs4_release_lockowner_release(void *calldata)
6237 {
6238         struct nfs_release_lockowner_data *data = calldata;
6239         nfs4_free_lock_state(data->server, data->lsp);
6240         kfree(calldata);
6241 }
6242
6243 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6244         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6245         .rpc_call_done = nfs4_release_lockowner_done,
6246         .rpc_release = nfs4_release_lockowner_release,
6247 };
6248
6249 static void
6250 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6251 {
6252         struct nfs_release_lockowner_data *data;
6253         struct rpc_message msg = {
6254                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6255         };
6256
6257         if (server->nfs_client->cl_mvops->minor_version != 0)
6258                 return;
6259
6260         data = kmalloc(sizeof(*data), GFP_NOFS);
6261         if (!data)
6262                 return;
6263         data->lsp = lsp;
6264         data->server = server;
6265         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6266         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6267         data->args.lock_owner.s_dev = server->s_dev;
6268
6269         msg.rpc_argp = &data->args;
6270         msg.rpc_resp = &data->res;
6271         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6272         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6273 }
6274
6275 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6276
6277 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6278                                    struct dentry *dentry, const char *key,
6279                                    const void *buf, size_t buflen,
6280                                    int flags)
6281 {
6282         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6283 }
6284
6285 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6286                                    struct dentry *dentry, const char *key,
6287                                    void *buf, size_t buflen)
6288 {
6289         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6290 }
6291
6292 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6293 {
6294         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6295 }
6296
6297 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6298
6299 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6300                                      struct dentry *dentry, const char *key,
6301                                      const void *buf, size_t buflen,
6302                                      int flags)
6303 {
6304         if (security_ismaclabel(key))
6305                 return nfs4_set_security_label(dentry, buf, buflen);
6306
6307         return -EOPNOTSUPP;
6308 }
6309
6310 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6311                                      struct dentry *dentry, const char *key,
6312                                      void *buf, size_t buflen)
6313 {
6314         if (security_ismaclabel(key))
6315                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6316         return -EOPNOTSUPP;
6317 }
6318
6319 static ssize_t
6320 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6321 {
6322         int len = 0;
6323
6324         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6325                 len = security_inode_listsecurity(inode, list, list_len);
6326                 if (list_len && len > list_len)
6327                         return -ERANGE;
6328         }
6329         return len;
6330 }
6331
6332 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6333         .prefix = XATTR_SECURITY_PREFIX,
6334         .get    = nfs4_xattr_get_nfs4_label,
6335         .set    = nfs4_xattr_set_nfs4_label,
6336 };
6337
6338 #else
6339
6340 static ssize_t
6341 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6342 {
6343         return 0;
6344 }
6345
6346 #endif
6347
6348 /*
6349  * nfs_fhget will use either the mounted_on_fileid or the fileid
6350  */
6351 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6352 {
6353         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6354                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6355               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6356               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6357                 return;
6358
6359         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6360                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6361         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6362         fattr->nlink = 2;
6363 }
6364
6365 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6366                                    const struct qstr *name,
6367                                    struct nfs4_fs_locations *fs_locations,
6368                                    struct page *page)
6369 {
6370         struct nfs_server *server = NFS_SERVER(dir);
6371         u32 bitmask[3] = {
6372                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6373         };
6374         struct nfs4_fs_locations_arg args = {
6375                 .dir_fh = NFS_FH(dir),
6376                 .name = name,
6377                 .page = page,
6378                 .bitmask = bitmask,
6379         };
6380         struct nfs4_fs_locations_res res = {
6381                 .fs_locations = fs_locations,
6382         };
6383         struct rpc_message msg = {
6384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6385                 .rpc_argp = &args,
6386                 .rpc_resp = &res,
6387         };
6388         int status;
6389
6390         dprintk("%s: start\n", __func__);
6391
6392         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6393          * is not supported */
6394         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6395                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6396         else
6397                 bitmask[0] |= FATTR4_WORD0_FILEID;
6398
6399         nfs_fattr_init(&fs_locations->fattr);
6400         fs_locations->server = server;
6401         fs_locations->nlocations = 0;
6402         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6403         dprintk("%s: returned status = %d\n", __func__, status);
6404         return status;
6405 }
6406
6407 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6408                            const struct qstr *name,
6409                            struct nfs4_fs_locations *fs_locations,
6410                            struct page *page)
6411 {
6412         struct nfs4_exception exception = { };
6413         int err;
6414         do {
6415                 err = _nfs4_proc_fs_locations(client, dir, name,
6416                                 fs_locations, page);
6417                 trace_nfs4_get_fs_locations(dir, name, err);
6418                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6419                                 &exception);
6420         } while (exception.retry);
6421         return err;
6422 }
6423
6424 /*
6425  * This operation also signals the server that this client is
6426  * performing migration recovery.  The server can stop returning
6427  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6428  * appended to this compound to identify the client ID which is
6429  * performing recovery.
6430  */
6431 static int _nfs40_proc_get_locations(struct inode *inode,
6432                                      struct nfs4_fs_locations *locations,
6433                                      struct page *page, struct rpc_cred *cred)
6434 {
6435         struct nfs_server *server = NFS_SERVER(inode);
6436         struct rpc_clnt *clnt = server->client;
6437         u32 bitmask[2] = {
6438                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6439         };
6440         struct nfs4_fs_locations_arg args = {
6441                 .clientid       = server->nfs_client->cl_clientid,
6442                 .fh             = NFS_FH(inode),
6443                 .page           = page,
6444                 .bitmask        = bitmask,
6445                 .migration      = 1,            /* skip LOOKUP */
6446                 .renew          = 1,            /* append RENEW */
6447         };
6448         struct nfs4_fs_locations_res res = {
6449                 .fs_locations   = locations,
6450                 .migration      = 1,
6451                 .renew          = 1,
6452         };
6453         struct rpc_message msg = {
6454                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6455                 .rpc_argp       = &args,
6456                 .rpc_resp       = &res,
6457                 .rpc_cred       = cred,
6458         };
6459         unsigned long now = jiffies;
6460         int status;
6461
6462         nfs_fattr_init(&locations->fattr);
6463         locations->server = server;
6464         locations->nlocations = 0;
6465
6466         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6467         nfs4_set_sequence_privileged(&args.seq_args);
6468         status = nfs4_call_sync_sequence(clnt, server, &msg,
6469                                         &args.seq_args, &res.seq_res);
6470         if (status)
6471                 return status;
6472
6473         renew_lease(server, now);
6474         return 0;
6475 }
6476
6477 #ifdef CONFIG_NFS_V4_1
6478
6479 /*
6480  * This operation also signals the server that this client is
6481  * performing migration recovery.  The server can stop asserting
6482  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6483  * performing this operation is identified in the SEQUENCE
6484  * operation in this compound.
6485  *
6486  * When the client supports GETATTR(fs_locations_info), it can
6487  * be plumbed in here.
6488  */
6489 static int _nfs41_proc_get_locations(struct inode *inode,
6490                                      struct nfs4_fs_locations *locations,
6491                                      struct page *page, struct rpc_cred *cred)
6492 {
6493         struct nfs_server *server = NFS_SERVER(inode);
6494         struct rpc_clnt *clnt = server->client;
6495         u32 bitmask[2] = {
6496                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6497         };
6498         struct nfs4_fs_locations_arg args = {
6499                 .fh             = NFS_FH(inode),
6500                 .page           = page,
6501                 .bitmask        = bitmask,
6502                 .migration      = 1,            /* skip LOOKUP */
6503         };
6504         struct nfs4_fs_locations_res res = {
6505                 .fs_locations   = locations,
6506                 .migration      = 1,
6507         };
6508         struct rpc_message msg = {
6509                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6510                 .rpc_argp       = &args,
6511                 .rpc_resp       = &res,
6512                 .rpc_cred       = cred,
6513         };
6514         int status;
6515
6516         nfs_fattr_init(&locations->fattr);
6517         locations->server = server;
6518         locations->nlocations = 0;
6519
6520         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6521         nfs4_set_sequence_privileged(&args.seq_args);
6522         status = nfs4_call_sync_sequence(clnt, server, &msg,
6523                                         &args.seq_args, &res.seq_res);
6524         if (status == NFS4_OK &&
6525             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6526                 status = -NFS4ERR_LEASE_MOVED;
6527         return status;
6528 }
6529
6530 #endif  /* CONFIG_NFS_V4_1 */
6531
6532 /**
6533  * nfs4_proc_get_locations - discover locations for a migrated FSID
6534  * @inode: inode on FSID that is migrating
6535  * @locations: result of query
6536  * @page: buffer
6537  * @cred: credential to use for this operation
6538  *
6539  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6540  * operation failed, or a negative errno if a local error occurred.
6541  *
6542  * On success, "locations" is filled in, but if the server has
6543  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6544  * asserted.
6545  *
6546  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6547  * from this client that require migration recovery.
6548  */
6549 int nfs4_proc_get_locations(struct inode *inode,
6550                             struct nfs4_fs_locations *locations,
6551                             struct page *page, struct rpc_cred *cred)
6552 {
6553         struct nfs_server *server = NFS_SERVER(inode);
6554         struct nfs_client *clp = server->nfs_client;
6555         const struct nfs4_mig_recovery_ops *ops =
6556                                         clp->cl_mvops->mig_recovery_ops;
6557         struct nfs4_exception exception = { };
6558         int status;
6559
6560         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6561                 (unsigned long long)server->fsid.major,
6562                 (unsigned long long)server->fsid.minor,
6563                 clp->cl_hostname);
6564         nfs_display_fhandle(NFS_FH(inode), __func__);
6565
6566         do {
6567                 status = ops->get_locations(inode, locations, page, cred);
6568                 if (status != -NFS4ERR_DELAY)
6569                         break;
6570                 nfs4_handle_exception(server, status, &exception);
6571         } while (exception.retry);
6572         return status;
6573 }
6574
6575 /*
6576  * This operation also signals the server that this client is
6577  * performing "lease moved" recovery.  The server can stop
6578  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6579  * is appended to this compound to identify the client ID which is
6580  * performing recovery.
6581  */
6582 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6583 {
6584         struct nfs_server *server = NFS_SERVER(inode);
6585         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6586         struct rpc_clnt *clnt = server->client;
6587         struct nfs4_fsid_present_arg args = {
6588                 .fh             = NFS_FH(inode),
6589                 .clientid       = clp->cl_clientid,
6590                 .renew          = 1,            /* append RENEW */
6591         };
6592         struct nfs4_fsid_present_res res = {
6593                 .renew          = 1,
6594         };
6595         struct rpc_message msg = {
6596                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6597                 .rpc_argp       = &args,
6598                 .rpc_resp       = &res,
6599                 .rpc_cred       = cred,
6600         };
6601         unsigned long now = jiffies;
6602         int status;
6603
6604         res.fh = nfs_alloc_fhandle();
6605         if (res.fh == NULL)
6606                 return -ENOMEM;
6607
6608         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6609         nfs4_set_sequence_privileged(&args.seq_args);
6610         status = nfs4_call_sync_sequence(clnt, server, &msg,
6611                                                 &args.seq_args, &res.seq_res);
6612         nfs_free_fhandle(res.fh);
6613         if (status)
6614                 return status;
6615
6616         do_renew_lease(clp, now);
6617         return 0;
6618 }
6619
6620 #ifdef CONFIG_NFS_V4_1
6621
6622 /*
6623  * This operation also signals the server that this client is
6624  * performing "lease moved" recovery.  The server can stop asserting
6625  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6626  * this operation is identified in the SEQUENCE operation in this
6627  * compound.
6628  */
6629 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6630 {
6631         struct nfs_server *server = NFS_SERVER(inode);
6632         struct rpc_clnt *clnt = server->client;
6633         struct nfs4_fsid_present_arg args = {
6634                 .fh             = NFS_FH(inode),
6635         };
6636         struct nfs4_fsid_present_res res = {
6637         };
6638         struct rpc_message msg = {
6639                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6640                 .rpc_argp       = &args,
6641                 .rpc_resp       = &res,
6642                 .rpc_cred       = cred,
6643         };
6644         int status;
6645
6646         res.fh = nfs_alloc_fhandle();
6647         if (res.fh == NULL)
6648                 return -ENOMEM;
6649
6650         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6651         nfs4_set_sequence_privileged(&args.seq_args);
6652         status = nfs4_call_sync_sequence(clnt, server, &msg,
6653                                                 &args.seq_args, &res.seq_res);
6654         nfs_free_fhandle(res.fh);
6655         if (status == NFS4_OK &&
6656             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6657                 status = -NFS4ERR_LEASE_MOVED;
6658         return status;
6659 }
6660
6661 #endif  /* CONFIG_NFS_V4_1 */
6662
6663 /**
6664  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6665  * @inode: inode on FSID to check
6666  * @cred: credential to use for this operation
6667  *
6668  * Server indicates whether the FSID is present, moved, or not
6669  * recognized.  This operation is necessary to clear a LEASE_MOVED
6670  * condition for this client ID.
6671  *
6672  * Returns NFS4_OK if the FSID is present on this server,
6673  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6674  *  NFS4ERR code if some error occurred on the server, or a
6675  *  negative errno if a local failure occurred.
6676  */
6677 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6678 {
6679         struct nfs_server *server = NFS_SERVER(inode);
6680         struct nfs_client *clp = server->nfs_client;
6681         const struct nfs4_mig_recovery_ops *ops =
6682                                         clp->cl_mvops->mig_recovery_ops;
6683         struct nfs4_exception exception = { };
6684         int status;
6685
6686         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6687                 (unsigned long long)server->fsid.major,
6688                 (unsigned long long)server->fsid.minor,
6689                 clp->cl_hostname);
6690         nfs_display_fhandle(NFS_FH(inode), __func__);
6691
6692         do {
6693                 status = ops->fsid_present(inode, cred);
6694                 if (status != -NFS4ERR_DELAY)
6695                         break;
6696                 nfs4_handle_exception(server, status, &exception);
6697         } while (exception.retry);
6698         return status;
6699 }
6700
6701 /**
6702  * If 'use_integrity' is true and the state managment nfs_client
6703  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6704  * and the machine credential as per RFC3530bis and RFC5661 Security
6705  * Considerations sections. Otherwise, just use the user cred with the
6706  * filesystem's rpc_client.
6707  */
6708 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6709 {
6710         int status;
6711         struct nfs4_secinfo_arg args = {
6712                 .dir_fh = NFS_FH(dir),
6713                 .name   = name,
6714         };
6715         struct nfs4_secinfo_res res = {
6716                 .flavors     = flavors,
6717         };
6718         struct rpc_message msg = {
6719                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6720                 .rpc_argp = &args,
6721                 .rpc_resp = &res,
6722         };
6723         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6724         struct rpc_cred *cred = NULL;
6725
6726         if (use_integrity) {
6727                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6728                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6729                 msg.rpc_cred = cred;
6730         }
6731
6732         dprintk("NFS call  secinfo %s\n", name->name);
6733
6734         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6735                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6736
6737         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6738                                 &res.seq_res, 0);
6739         dprintk("NFS reply  secinfo: %d\n", status);
6740
6741         if (cred)
6742                 put_rpccred(cred);
6743
6744         return status;
6745 }
6746
6747 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6748                       struct nfs4_secinfo_flavors *flavors)
6749 {
6750         struct nfs4_exception exception = { };
6751         int err;
6752         do {
6753                 err = -NFS4ERR_WRONGSEC;
6754
6755                 /* try to use integrity protection with machine cred */
6756                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6757                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6758
6759                 /*
6760                  * if unable to use integrity protection, or SECINFO with
6761                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6762                  * disallowed by spec, but exists in deployed servers) use
6763                  * the current filesystem's rpc_client and the user cred.
6764                  */
6765                 if (err == -NFS4ERR_WRONGSEC)
6766                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6767
6768                 trace_nfs4_secinfo(dir, name, err);
6769                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6770                                 &exception);
6771         } while (exception.retry);
6772         return err;
6773 }
6774
6775 #ifdef CONFIG_NFS_V4_1
6776 /*
6777  * Check the exchange flags returned by the server for invalid flags, having
6778  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6779  * DS flags set.
6780  */
6781 static int nfs4_check_cl_exchange_flags(u32 flags)
6782 {
6783         if (flags & ~EXCHGID4_FLAG_MASK_R)
6784                 goto out_inval;
6785         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6786             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6787                 goto out_inval;
6788         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6789                 goto out_inval;
6790         return NFS_OK;
6791 out_inval:
6792         return -NFS4ERR_INVAL;
6793 }
6794
6795 static bool
6796 nfs41_same_server_scope(struct nfs41_server_scope *a,
6797                         struct nfs41_server_scope *b)
6798 {
6799         if (a->server_scope_sz == b->server_scope_sz &&
6800             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6801                 return true;
6802
6803         return false;
6804 }
6805
6806 static void
6807 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
6808 {
6809 }
6810
6811 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
6812         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
6813 };
6814
6815 /*
6816  * nfs4_proc_bind_one_conn_to_session()
6817  *
6818  * The 4.1 client currently uses the same TCP connection for the
6819  * fore and backchannel.
6820  */
6821 static
6822 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
6823                 struct rpc_xprt *xprt,
6824                 struct nfs_client *clp,
6825                 struct rpc_cred *cred)
6826 {
6827         int status;
6828         struct nfs41_bind_conn_to_session_args args = {
6829                 .client = clp,
6830                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6831         };
6832         struct nfs41_bind_conn_to_session_res res;
6833         struct rpc_message msg = {
6834                 .rpc_proc =
6835                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6836                 .rpc_argp = &args,
6837                 .rpc_resp = &res,
6838                 .rpc_cred = cred,
6839         };
6840         struct rpc_task_setup task_setup_data = {
6841                 .rpc_client = clnt,
6842                 .rpc_xprt = xprt,
6843                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
6844                 .rpc_message = &msg,
6845                 .flags = RPC_TASK_TIMEOUT,
6846         };
6847         struct rpc_task *task;
6848
6849         dprintk("--> %s\n", __func__);
6850
6851         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6852         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6853                 args.dir = NFS4_CDFC4_FORE;
6854
6855         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
6856         if (xprt != rcu_access_pointer(clnt->cl_xprt))
6857                 args.dir = NFS4_CDFC4_FORE;
6858
6859         task = rpc_run_task(&task_setup_data);
6860         if (!IS_ERR(task)) {
6861                 status = task->tk_status;
6862                 rpc_put_task(task);
6863         } else
6864                 status = PTR_ERR(task);
6865         trace_nfs4_bind_conn_to_session(clp, status);
6866         if (status == 0) {
6867                 if (memcmp(res.sessionid.data,
6868                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6869                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6870                         status = -EIO;
6871                         goto out;
6872                 }
6873                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6874                         dprintk("NFS: %s: Unexpected direction from server\n",
6875                                 __func__);
6876                         status = -EIO;
6877                         goto out;
6878                 }
6879                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6880                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6881                                 __func__);
6882                         status = -EIO;
6883                         goto out;
6884                 }
6885         }
6886 out:
6887         dprintk("<-- %s status= %d\n", __func__, status);
6888         return status;
6889 }
6890
6891 struct rpc_bind_conn_calldata {
6892         struct nfs_client *clp;
6893         struct rpc_cred *cred;
6894 };
6895
6896 static int
6897 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
6898                 struct rpc_xprt *xprt,
6899                 void *calldata)
6900 {
6901         struct rpc_bind_conn_calldata *p = calldata;
6902
6903         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
6904 }
6905
6906 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6907 {
6908         struct rpc_bind_conn_calldata data = {
6909                 .clp = clp,
6910                 .cred = cred,
6911         };
6912         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
6913                         nfs4_proc_bind_conn_to_session_callback, &data);
6914 }
6915
6916 /*
6917  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6918  * and operations we'd like to see to enable certain features in the allow map
6919  */
6920 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6921         .how = SP4_MACH_CRED,
6922         .enforce.u.words = {
6923                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6924                       1 << (OP_EXCHANGE_ID - 32) |
6925                       1 << (OP_CREATE_SESSION - 32) |
6926                       1 << (OP_DESTROY_SESSION - 32) |
6927                       1 << (OP_DESTROY_CLIENTID - 32)
6928         },
6929         .allow.u.words = {
6930                 [0] = 1 << (OP_CLOSE) |
6931                       1 << (OP_OPEN_DOWNGRADE) |
6932                       1 << (OP_LOCKU) |
6933                       1 << (OP_DELEGRETURN) |
6934                       1 << (OP_COMMIT),
6935                 [1] = 1 << (OP_SECINFO - 32) |
6936                       1 << (OP_SECINFO_NO_NAME - 32) |
6937                       1 << (OP_LAYOUTRETURN - 32) |
6938                       1 << (OP_TEST_STATEID - 32) |
6939                       1 << (OP_FREE_STATEID - 32) |
6940                       1 << (OP_WRITE - 32)
6941         }
6942 };
6943
6944 /*
6945  * Select the state protection mode for client `clp' given the server results
6946  * from exchange_id in `sp'.
6947  *
6948  * Returns 0 on success, negative errno otherwise.
6949  */
6950 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6951                                  struct nfs41_state_protection *sp)
6952 {
6953         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6954                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6955                       1 << (OP_EXCHANGE_ID - 32) |
6956                       1 << (OP_CREATE_SESSION - 32) |
6957                       1 << (OP_DESTROY_SESSION - 32) |
6958                       1 << (OP_DESTROY_CLIENTID - 32)
6959         };
6960         unsigned int i;
6961
6962         if (sp->how == SP4_MACH_CRED) {
6963                 /* Print state protect result */
6964                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6965                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6966                         if (test_bit(i, sp->enforce.u.longs))
6967                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6968                         if (test_bit(i, sp->allow.u.longs))
6969                                 dfprintk(MOUNT, "  allow op %d\n", i);
6970                 }
6971
6972                 /* make sure nothing is on enforce list that isn't supported */
6973                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6974                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6975                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6976                                 return -EINVAL;
6977                         }
6978                 }
6979
6980                 /*
6981                  * Minimal mode - state operations are allowed to use machine
6982                  * credential.  Note this already happens by default, so the
6983                  * client doesn't have to do anything more than the negotiation.
6984                  *
6985                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6986                  *       we're already using the machine cred for exchange_id
6987                  *       and will never use a different cred.
6988                  */
6989                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6990                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6991                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6992                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6993                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6994                         dfprintk(MOUNT, "  minimal mode enabled\n");
6995                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6996                 } else {
6997                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6998                         return -EINVAL;
6999                 }
7000
7001                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7002                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7003                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7004                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7005                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7006                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7007                 }
7008
7009                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7010                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7011                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7012                                 &clp->cl_sp4_flags);
7013                 }
7014
7015                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7016                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7017                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7018                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7019                 }
7020
7021                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7022                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7023                         dfprintk(MOUNT, "  stateid mode enabled\n");
7024                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7025                 }
7026
7027                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7028                         dfprintk(MOUNT, "  write mode enabled\n");
7029                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7030                 }
7031
7032                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7033                         dfprintk(MOUNT, "  commit mode enabled\n");
7034                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7035                 }
7036         }
7037
7038         return 0;
7039 }
7040
7041 /*
7042  * _nfs4_proc_exchange_id()
7043  *
7044  * Wrapper for EXCHANGE_ID operation.
7045  */
7046 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7047         u32 sp4_how)
7048 {
7049         nfs4_verifier verifier;
7050         struct nfs41_exchange_id_args args = {
7051                 .verifier = &verifier,
7052                 .client = clp,
7053 #ifdef CONFIG_NFS_V4_1_MIGRATION
7054                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7055                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
7056                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7057 #else
7058                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7059                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
7060 #endif
7061         };
7062         struct nfs41_exchange_id_res res = {
7063                 0
7064         };
7065         int status;
7066         struct rpc_message msg = {
7067                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7068                 .rpc_argp = &args,
7069                 .rpc_resp = &res,
7070                 .rpc_cred = cred,
7071         };
7072
7073         nfs4_init_boot_verifier(clp, &verifier);
7074
7075         status = nfs4_init_uniform_client_string(clp);
7076         if (status)
7077                 goto out;
7078
7079         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7080                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7081                 clp->cl_owner_id);
7082
7083         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7084                                         GFP_NOFS);
7085         if (unlikely(res.server_owner == NULL)) {
7086                 status = -ENOMEM;
7087                 goto out;
7088         }
7089
7090         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7091                                         GFP_NOFS);
7092         if (unlikely(res.server_scope == NULL)) {
7093                 status = -ENOMEM;
7094                 goto out_server_owner;
7095         }
7096
7097         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7098         if (unlikely(res.impl_id == NULL)) {
7099                 status = -ENOMEM;
7100                 goto out_server_scope;
7101         }
7102
7103         switch (sp4_how) {
7104         case SP4_NONE:
7105                 args.state_protect.how = SP4_NONE;
7106                 break;
7107
7108         case SP4_MACH_CRED:
7109                 args.state_protect = nfs4_sp4_mach_cred_request;
7110                 break;
7111
7112         default:
7113                 /* unsupported! */
7114                 WARN_ON_ONCE(1);
7115                 status = -EINVAL;
7116                 goto out_impl_id;
7117         }
7118
7119         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7120         trace_nfs4_exchange_id(clp, status);
7121         if (status == 0)
7122                 status = nfs4_check_cl_exchange_flags(res.flags);
7123
7124         if (status == 0)
7125                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7126
7127         if (status == 0) {
7128                 clp->cl_clientid = res.clientid;
7129                 clp->cl_exchange_flags = res.flags;
7130                 /* Client ID is not confirmed */
7131                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7132                         clear_bit(NFS4_SESSION_ESTABLISHED,
7133                                         &clp->cl_session->session_state);
7134                         clp->cl_seqid = res.seqid;
7135                 }
7136
7137                 kfree(clp->cl_serverowner);
7138                 clp->cl_serverowner = res.server_owner;
7139                 res.server_owner = NULL;
7140
7141                 /* use the most recent implementation id */
7142                 kfree(clp->cl_implid);
7143                 clp->cl_implid = res.impl_id;
7144                 res.impl_id = NULL;
7145
7146                 if (clp->cl_serverscope != NULL &&
7147                     !nfs41_same_server_scope(clp->cl_serverscope,
7148                                              res.server_scope)) {
7149                         dprintk("%s: server_scope mismatch detected\n",
7150                                 __func__);
7151                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7152                         kfree(clp->cl_serverscope);
7153                         clp->cl_serverscope = NULL;
7154                 }
7155
7156                 if (clp->cl_serverscope == NULL) {
7157                         clp->cl_serverscope = res.server_scope;
7158                         res.server_scope = NULL;
7159                 }
7160         }
7161
7162 out_impl_id:
7163         kfree(res.impl_id);
7164 out_server_scope:
7165         kfree(res.server_scope);
7166 out_server_owner:
7167         kfree(res.server_owner);
7168 out:
7169         if (clp->cl_implid != NULL)
7170                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7171                         "domain: %s, name: %s, date: %llu,%u\n",
7172                         clp->cl_implid->domain, clp->cl_implid->name,
7173                         clp->cl_implid->date.seconds,
7174                         clp->cl_implid->date.nseconds);
7175         dprintk("NFS reply exchange_id: %d\n", status);
7176         return status;
7177 }
7178
7179 /*
7180  * nfs4_proc_exchange_id()
7181  *
7182  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7183  *
7184  * Since the clientid has expired, all compounds using sessions
7185  * associated with the stale clientid will be returning
7186  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7187  * be in some phase of session reset.
7188  *
7189  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7190  */
7191 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7192 {
7193         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7194         int status;
7195
7196         /* try SP4_MACH_CRED if krb5i/p */
7197         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7198             authflavor == RPC_AUTH_GSS_KRB5P) {
7199                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7200                 if (!status)
7201                         return 0;
7202         }
7203
7204         /* try SP4_NONE */
7205         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7206 }
7207
7208 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7209                 struct rpc_cred *cred)
7210 {
7211         struct rpc_message msg = {
7212                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7213                 .rpc_argp = clp,
7214                 .rpc_cred = cred,
7215         };
7216         int status;
7217
7218         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7219         trace_nfs4_destroy_clientid(clp, status);
7220         if (status)
7221                 dprintk("NFS: Got error %d from the server %s on "
7222                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7223         return status;
7224 }
7225
7226 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7227                 struct rpc_cred *cred)
7228 {
7229         unsigned int loop;
7230         int ret;
7231
7232         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7233                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7234                 switch (ret) {
7235                 case -NFS4ERR_DELAY:
7236                 case -NFS4ERR_CLIENTID_BUSY:
7237                         ssleep(1);
7238                         break;
7239                 default:
7240                         return ret;
7241                 }
7242         }
7243         return 0;
7244 }
7245
7246 int nfs4_destroy_clientid(struct nfs_client *clp)
7247 {
7248         struct rpc_cred *cred;
7249         int ret = 0;
7250
7251         if (clp->cl_mvops->minor_version < 1)
7252                 goto out;
7253         if (clp->cl_exchange_flags == 0)
7254                 goto out;
7255         if (clp->cl_preserve_clid)
7256                 goto out;
7257         cred = nfs4_get_clid_cred(clp);
7258         ret = nfs4_proc_destroy_clientid(clp, cred);
7259         if (cred)
7260                 put_rpccred(cred);
7261         switch (ret) {
7262         case 0:
7263         case -NFS4ERR_STALE_CLIENTID:
7264                 clp->cl_exchange_flags = 0;
7265         }
7266 out:
7267         return ret;
7268 }
7269
7270 struct nfs4_get_lease_time_data {
7271         struct nfs4_get_lease_time_args *args;
7272         struct nfs4_get_lease_time_res *res;
7273         struct nfs_client *clp;
7274 };
7275
7276 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7277                                         void *calldata)
7278 {
7279         struct nfs4_get_lease_time_data *data =
7280                         (struct nfs4_get_lease_time_data *)calldata;
7281
7282         dprintk("--> %s\n", __func__);
7283         /* just setup sequence, do not trigger session recovery
7284            since we're invoked within one */
7285         nfs41_setup_sequence(data->clp->cl_session,
7286                         &data->args->la_seq_args,
7287                         &data->res->lr_seq_res,
7288                         task);
7289         dprintk("<-- %s\n", __func__);
7290 }
7291
7292 /*
7293  * Called from nfs4_state_manager thread for session setup, so don't recover
7294  * from sequence operation or clientid errors.
7295  */
7296 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7297 {
7298         struct nfs4_get_lease_time_data *data =
7299                         (struct nfs4_get_lease_time_data *)calldata;
7300
7301         dprintk("--> %s\n", __func__);
7302         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7303                 return;
7304         switch (task->tk_status) {
7305         case -NFS4ERR_DELAY:
7306         case -NFS4ERR_GRACE:
7307                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7308                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7309                 task->tk_status = 0;
7310                 /* fall through */
7311         case -NFS4ERR_RETRY_UNCACHED_REP:
7312                 rpc_restart_call_prepare(task);
7313                 return;
7314         }
7315         dprintk("<-- %s\n", __func__);
7316 }
7317
7318 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7319         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7320         .rpc_call_done = nfs4_get_lease_time_done,
7321 };
7322
7323 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7324 {
7325         struct rpc_task *task;
7326         struct nfs4_get_lease_time_args args;
7327         struct nfs4_get_lease_time_res res = {
7328                 .lr_fsinfo = fsinfo,
7329         };
7330         struct nfs4_get_lease_time_data data = {
7331                 .args = &args,
7332                 .res = &res,
7333                 .clp = clp,
7334         };
7335         struct rpc_message msg = {
7336                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7337                 .rpc_argp = &args,
7338                 .rpc_resp = &res,
7339         };
7340         struct rpc_task_setup task_setup = {
7341                 .rpc_client = clp->cl_rpcclient,
7342                 .rpc_message = &msg,
7343                 .callback_ops = &nfs4_get_lease_time_ops,
7344                 .callback_data = &data,
7345                 .flags = RPC_TASK_TIMEOUT,
7346         };
7347         int status;
7348
7349         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7350         nfs4_set_sequence_privileged(&args.la_seq_args);
7351         dprintk("--> %s\n", __func__);
7352         task = rpc_run_task(&task_setup);
7353
7354         if (IS_ERR(task))
7355                 status = PTR_ERR(task);
7356         else {
7357                 status = task->tk_status;
7358                 rpc_put_task(task);
7359         }
7360         dprintk("<-- %s return %d\n", __func__, status);
7361
7362         return status;
7363 }
7364
7365 /*
7366  * Initialize the values to be used by the client in CREATE_SESSION
7367  * If nfs4_init_session set the fore channel request and response sizes,
7368  * use them.
7369  *
7370  * Set the back channel max_resp_sz_cached to zero to force the client to
7371  * always set csa_cachethis to FALSE because the current implementation
7372  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7373  */
7374 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7375 {
7376         unsigned int max_rqst_sz, max_resp_sz;
7377
7378         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7379         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7380
7381         /* Fore channel attributes */
7382         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7383         args->fc_attrs.max_resp_sz = max_resp_sz;
7384         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7385         args->fc_attrs.max_reqs = max_session_slots;
7386
7387         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7388                 "max_ops=%u max_reqs=%u\n",
7389                 __func__,
7390                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7391                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7392
7393         /* Back channel attributes */
7394         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7395         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7396         args->bc_attrs.max_resp_sz_cached = 0;
7397         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7398         args->bc_attrs.max_reqs = NFS41_BC_MAX_CALLBACKS;
7399
7400         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7401                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7402                 __func__,
7403                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7404                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7405                 args->bc_attrs.max_reqs);
7406 }
7407
7408 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7409                 struct nfs41_create_session_res *res)
7410 {
7411         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7412         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7413
7414         if (rcvd->max_resp_sz > sent->max_resp_sz)
7415                 return -EINVAL;
7416         /*
7417          * Our requested max_ops is the minimum we need; we're not
7418          * prepared to break up compounds into smaller pieces than that.
7419          * So, no point even trying to continue if the server won't
7420          * cooperate:
7421          */
7422         if (rcvd->max_ops < sent->max_ops)
7423                 return -EINVAL;
7424         if (rcvd->max_reqs == 0)
7425                 return -EINVAL;
7426         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7427                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7428         return 0;
7429 }
7430
7431 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7432                 struct nfs41_create_session_res *res)
7433 {
7434         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7435         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7436
7437         if (!(res->flags & SESSION4_BACK_CHAN))
7438                 goto out;
7439         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7440                 return -EINVAL;
7441         if (rcvd->max_resp_sz < sent->max_resp_sz)
7442                 return -EINVAL;
7443         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7444                 return -EINVAL;
7445         /* These would render the backchannel useless: */
7446         if (rcvd->max_ops != sent->max_ops)
7447                 return -EINVAL;
7448         if (rcvd->max_reqs != sent->max_reqs)
7449                 return -EINVAL;
7450 out:
7451         return 0;
7452 }
7453
7454 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7455                                      struct nfs41_create_session_res *res)
7456 {
7457         int ret;
7458
7459         ret = nfs4_verify_fore_channel_attrs(args, res);
7460         if (ret)
7461                 return ret;
7462         return nfs4_verify_back_channel_attrs(args, res);
7463 }
7464
7465 static void nfs4_update_session(struct nfs4_session *session,
7466                 struct nfs41_create_session_res *res)
7467 {
7468         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7469         /* Mark client id and session as being confirmed */
7470         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7471         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7472         session->flags = res->flags;
7473         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7474         if (res->flags & SESSION4_BACK_CHAN)
7475                 memcpy(&session->bc_attrs, &res->bc_attrs,
7476                                 sizeof(session->bc_attrs));
7477 }
7478
7479 static int _nfs4_proc_create_session(struct nfs_client *clp,
7480                 struct rpc_cred *cred)
7481 {
7482         struct nfs4_session *session = clp->cl_session;
7483         struct nfs41_create_session_args args = {
7484                 .client = clp,
7485                 .clientid = clp->cl_clientid,
7486                 .seqid = clp->cl_seqid,
7487                 .cb_program = NFS4_CALLBACK,
7488         };
7489         struct nfs41_create_session_res res;
7490
7491         struct rpc_message msg = {
7492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7493                 .rpc_argp = &args,
7494                 .rpc_resp = &res,
7495                 .rpc_cred = cred,
7496         };
7497         int status;
7498
7499         nfs4_init_channel_attrs(&args);
7500         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7501
7502         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7503         trace_nfs4_create_session(clp, status);
7504
7505         if (!status) {
7506                 /* Verify the session's negotiated channel_attrs values */
7507                 status = nfs4_verify_channel_attrs(&args, &res);
7508                 /* Increment the clientid slot sequence id */
7509                 if (clp->cl_seqid == res.seqid)
7510                         clp->cl_seqid++;
7511                 if (status)
7512                         goto out;
7513                 nfs4_update_session(session, &res);
7514         }
7515 out:
7516         return status;
7517 }
7518
7519 /*
7520  * Issues a CREATE_SESSION operation to the server.
7521  * It is the responsibility of the caller to verify the session is
7522  * expired before calling this routine.
7523  */
7524 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7525 {
7526         int status;
7527         unsigned *ptr;
7528         struct nfs4_session *session = clp->cl_session;
7529
7530         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7531
7532         status = _nfs4_proc_create_session(clp, cred);
7533         if (status)
7534                 goto out;
7535
7536         /* Init or reset the session slot tables */
7537         status = nfs4_setup_session_slot_tables(session);
7538         dprintk("slot table setup returned %d\n", status);
7539         if (status)
7540                 goto out;
7541
7542         ptr = (unsigned *)&session->sess_id.data[0];
7543         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7544                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7545 out:
7546         dprintk("<-- %s\n", __func__);
7547         return status;
7548 }
7549
7550 /*
7551  * Issue the over-the-wire RPC DESTROY_SESSION.
7552  * The caller must serialize access to this routine.
7553  */
7554 int nfs4_proc_destroy_session(struct nfs4_session *session,
7555                 struct rpc_cred *cred)
7556 {
7557         struct rpc_message msg = {
7558                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7559                 .rpc_argp = session,
7560                 .rpc_cred = cred,
7561         };
7562         int status = 0;
7563
7564         dprintk("--> nfs4_proc_destroy_session\n");
7565
7566         /* session is still being setup */
7567         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7568                 return 0;
7569
7570         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7571         trace_nfs4_destroy_session(session->clp, status);
7572
7573         if (status)
7574                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7575                         "Session has been destroyed regardless...\n", status);
7576
7577         dprintk("<-- nfs4_proc_destroy_session\n");
7578         return status;
7579 }
7580
7581 /*
7582  * Renew the cl_session lease.
7583  */
7584 struct nfs4_sequence_data {
7585         struct nfs_client *clp;
7586         struct nfs4_sequence_args args;
7587         struct nfs4_sequence_res res;
7588 };
7589
7590 static void nfs41_sequence_release(void *data)
7591 {
7592         struct nfs4_sequence_data *calldata = data;
7593         struct nfs_client *clp = calldata->clp;
7594
7595         if (atomic_read(&clp->cl_count) > 1)
7596                 nfs4_schedule_state_renewal(clp);
7597         nfs_put_client(clp);
7598         kfree(calldata);
7599 }
7600
7601 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7602 {
7603         switch(task->tk_status) {
7604         case -NFS4ERR_DELAY:
7605                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7606                 return -EAGAIN;
7607         default:
7608                 nfs4_schedule_lease_recovery(clp);
7609         }
7610         return 0;
7611 }
7612
7613 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7614 {
7615         struct nfs4_sequence_data *calldata = data;
7616         struct nfs_client *clp = calldata->clp;
7617
7618         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7619                 return;
7620
7621         trace_nfs4_sequence(clp, task->tk_status);
7622         if (task->tk_status < 0) {
7623                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7624                 if (atomic_read(&clp->cl_count) == 1)
7625                         goto out;
7626
7627                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7628                         rpc_restart_call_prepare(task);
7629                         return;
7630                 }
7631         }
7632         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7633 out:
7634         dprintk("<-- %s\n", __func__);
7635 }
7636
7637 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7638 {
7639         struct nfs4_sequence_data *calldata = data;
7640         struct nfs_client *clp = calldata->clp;
7641         struct nfs4_sequence_args *args;
7642         struct nfs4_sequence_res *res;
7643
7644         args = task->tk_msg.rpc_argp;
7645         res = task->tk_msg.rpc_resp;
7646
7647         nfs41_setup_sequence(clp->cl_session, args, res, task);
7648 }
7649
7650 static const struct rpc_call_ops nfs41_sequence_ops = {
7651         .rpc_call_done = nfs41_sequence_call_done,
7652         .rpc_call_prepare = nfs41_sequence_prepare,
7653         .rpc_release = nfs41_sequence_release,
7654 };
7655
7656 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7657                 struct rpc_cred *cred,
7658                 bool is_privileged)
7659 {
7660         struct nfs4_sequence_data *calldata;
7661         struct rpc_message msg = {
7662                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7663                 .rpc_cred = cred,
7664         };
7665         struct rpc_task_setup task_setup_data = {
7666                 .rpc_client = clp->cl_rpcclient,
7667                 .rpc_message = &msg,
7668                 .callback_ops = &nfs41_sequence_ops,
7669                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7670         };
7671
7672         if (!atomic_inc_not_zero(&clp->cl_count))
7673                 return ERR_PTR(-EIO);
7674         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7675         if (calldata == NULL) {
7676                 nfs_put_client(clp);
7677                 return ERR_PTR(-ENOMEM);
7678         }
7679         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7680         if (is_privileged)
7681                 nfs4_set_sequence_privileged(&calldata->args);
7682         msg.rpc_argp = &calldata->args;
7683         msg.rpc_resp = &calldata->res;
7684         calldata->clp = clp;
7685         task_setup_data.callback_data = calldata;
7686
7687         return rpc_run_task(&task_setup_data);
7688 }
7689
7690 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7691 {
7692         struct rpc_task *task;
7693         int ret = 0;
7694
7695         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7696                 return -EAGAIN;
7697         task = _nfs41_proc_sequence(clp, cred, false);
7698         if (IS_ERR(task))
7699                 ret = PTR_ERR(task);
7700         else
7701                 rpc_put_task_async(task);
7702         dprintk("<-- %s status=%d\n", __func__, ret);
7703         return ret;
7704 }
7705
7706 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7707 {
7708         struct rpc_task *task;
7709         int ret;
7710
7711         task = _nfs41_proc_sequence(clp, cred, true);
7712         if (IS_ERR(task)) {
7713                 ret = PTR_ERR(task);
7714                 goto out;
7715         }
7716         ret = rpc_wait_for_completion_task(task);
7717         if (!ret)
7718                 ret = task->tk_status;
7719         rpc_put_task(task);
7720 out:
7721         dprintk("<-- %s status=%d\n", __func__, ret);
7722         return ret;
7723 }
7724
7725 struct nfs4_reclaim_complete_data {
7726         struct nfs_client *clp;
7727         struct nfs41_reclaim_complete_args arg;
7728         struct nfs41_reclaim_complete_res res;
7729 };
7730
7731 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7732 {
7733         struct nfs4_reclaim_complete_data *calldata = data;
7734
7735         nfs41_setup_sequence(calldata->clp->cl_session,
7736                         &calldata->arg.seq_args,
7737                         &calldata->res.seq_res,
7738                         task);
7739 }
7740
7741 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7742 {
7743         switch(task->tk_status) {
7744         case 0:
7745         case -NFS4ERR_COMPLETE_ALREADY:
7746         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7747                 break;
7748         case -NFS4ERR_DELAY:
7749                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7750                 /* fall through */
7751         case -NFS4ERR_RETRY_UNCACHED_REP:
7752                 return -EAGAIN;
7753         default:
7754                 nfs4_schedule_lease_recovery(clp);
7755         }
7756         return 0;
7757 }
7758
7759 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7760 {
7761         struct nfs4_reclaim_complete_data *calldata = data;
7762         struct nfs_client *clp = calldata->clp;
7763         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7764
7765         dprintk("--> %s\n", __func__);
7766         if (!nfs41_sequence_done(task, res))
7767                 return;
7768
7769         trace_nfs4_reclaim_complete(clp, task->tk_status);
7770         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7771                 rpc_restart_call_prepare(task);
7772                 return;
7773         }
7774         dprintk("<-- %s\n", __func__);
7775 }
7776
7777 static void nfs4_free_reclaim_complete_data(void *data)
7778 {
7779         struct nfs4_reclaim_complete_data *calldata = data;
7780
7781         kfree(calldata);
7782 }
7783
7784 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7785         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7786         .rpc_call_done = nfs4_reclaim_complete_done,
7787         .rpc_release = nfs4_free_reclaim_complete_data,
7788 };
7789
7790 /*
7791  * Issue a global reclaim complete.
7792  */
7793 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7794                 struct rpc_cred *cred)
7795 {
7796         struct nfs4_reclaim_complete_data *calldata;
7797         struct rpc_task *task;
7798         struct rpc_message msg = {
7799                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7800                 .rpc_cred = cred,
7801         };
7802         struct rpc_task_setup task_setup_data = {
7803                 .rpc_client = clp->cl_rpcclient,
7804                 .rpc_message = &msg,
7805                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7806                 .flags = RPC_TASK_ASYNC,
7807         };
7808         int status = -ENOMEM;
7809
7810         dprintk("--> %s\n", __func__);
7811         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7812         if (calldata == NULL)
7813                 goto out;
7814         calldata->clp = clp;
7815         calldata->arg.one_fs = 0;
7816
7817         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7818         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7819         msg.rpc_argp = &calldata->arg;
7820         msg.rpc_resp = &calldata->res;
7821         task_setup_data.callback_data = calldata;
7822         task = rpc_run_task(&task_setup_data);
7823         if (IS_ERR(task)) {
7824                 status = PTR_ERR(task);
7825                 goto out;
7826         }
7827         status = nfs4_wait_for_completion_rpc_task(task);
7828         if (status == 0)
7829                 status = task->tk_status;
7830         rpc_put_task(task);
7831         return 0;
7832 out:
7833         dprintk("<-- %s status=%d\n", __func__, status);
7834         return status;
7835 }
7836
7837 static void
7838 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7839 {
7840         struct nfs4_layoutget *lgp = calldata;
7841         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7842         struct nfs4_session *session = nfs4_get_session(server);
7843         int ret;
7844
7845         dprintk("--> %s\n", __func__);
7846         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7847          * right now covering the LAYOUTGET we are about to send.
7848          * However, that is not so catastrophic, and there seems
7849          * to be no way to prevent it completely.
7850          */
7851         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7852                                 &lgp->res.seq_res, task))
7853                 return;
7854         ret = pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7855                                           NFS_I(lgp->args.inode)->layout,
7856                                           &lgp->args.range,
7857                                           lgp->args.ctx->state);
7858         if (ret < 0)
7859                 rpc_exit(task, ret);
7860 }
7861
7862 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7863 {
7864         struct nfs4_layoutget *lgp = calldata;
7865         struct inode *inode = lgp->args.inode;
7866         struct nfs_server *server = NFS_SERVER(inode);
7867         struct pnfs_layout_hdr *lo;
7868         struct nfs4_state *state = NULL;
7869         unsigned long timeo, now, giveup;
7870
7871         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7872
7873         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7874                 goto out;
7875
7876         switch (task->tk_status) {
7877         case 0:
7878                 goto out;
7879
7880         /*
7881          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
7882          * on the file. set tk_status to -ENODATA to tell upper layer to
7883          * retry go inband.
7884          */
7885         case -NFS4ERR_LAYOUTUNAVAILABLE:
7886                 task->tk_status = -ENODATA;
7887                 goto out;
7888         /*
7889          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7890          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7891          */
7892         case -NFS4ERR_BADLAYOUT:
7893                 goto out_overflow;
7894         /*
7895          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7896          * (or clients) writing to the same RAID stripe except when
7897          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7898          */
7899         case -NFS4ERR_LAYOUTTRYLATER:
7900                 if (lgp->args.minlength == 0)
7901                         goto out_overflow;
7902         /*
7903          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7904          * existing layout before getting a new one).
7905          */
7906         case -NFS4ERR_RECALLCONFLICT:
7907                 timeo = rpc_get_timeout(task->tk_client);
7908                 giveup = lgp->args.timestamp + timeo;
7909                 now = jiffies;
7910                 if (time_after(giveup, now)) {
7911                         unsigned long delay;
7912
7913                         /* Delay for:
7914                          * - Not less then NFS4_POLL_RETRY_MIN.
7915                          * - One last time a jiffie before we give up
7916                          * - exponential backoff (time_now minus start_attempt)
7917                          */
7918                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7919                                     min((giveup - now - 1),
7920                                         now - lgp->args.timestamp));
7921
7922                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7923                                 __func__, delay);
7924                         rpc_delay(task, delay);
7925                         /* Do not call nfs4_async_handle_error() */
7926                         goto out_restart;
7927                 }
7928                 break;
7929         case -NFS4ERR_EXPIRED:
7930         case -NFS4ERR_BAD_STATEID:
7931                 spin_lock(&inode->i_lock);
7932                 if (nfs4_stateid_match(&lgp->args.stateid,
7933                                         &lgp->args.ctx->state->stateid)) {
7934                         spin_unlock(&inode->i_lock);
7935                         /* If the open stateid was bad, then recover it. */
7936                         state = lgp->args.ctx->state;
7937                         break;
7938                 }
7939                 lo = NFS_I(inode)->layout;
7940                 if (lo && nfs4_stateid_match(&lgp->args.stateid,
7941                                         &lo->plh_stateid)) {
7942                         LIST_HEAD(head);
7943
7944                         /*
7945                          * Mark the bad layout state as invalid, then retry
7946                          * with the current stateid.
7947                          */
7948                         set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7949                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7950                         spin_unlock(&inode->i_lock);
7951                         pnfs_free_lseg_list(&head);
7952                 } else
7953                         spin_unlock(&inode->i_lock);
7954                 goto out_restart;
7955         }
7956         if (nfs4_async_handle_error(task, server, state, &lgp->timeout) == -EAGAIN)
7957                 goto out_restart;
7958 out:
7959         dprintk("<-- %s\n", __func__);
7960         return;
7961 out_restart:
7962         task->tk_status = 0;
7963         rpc_restart_call_prepare(task);
7964         return;
7965 out_overflow:
7966         task->tk_status = -EOVERFLOW;
7967         goto out;
7968 }
7969
7970 static size_t max_response_pages(struct nfs_server *server)
7971 {
7972         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7973         return nfs_page_array_len(0, max_resp_sz);
7974 }
7975
7976 static void nfs4_free_pages(struct page **pages, size_t size)
7977 {
7978         int i;
7979
7980         if (!pages)
7981                 return;
7982
7983         for (i = 0; i < size; i++) {
7984                 if (!pages[i])
7985                         break;
7986                 __free_page(pages[i]);
7987         }
7988         kfree(pages);
7989 }
7990
7991 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7992 {
7993         struct page **pages;
7994         int i;
7995
7996         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7997         if (!pages) {
7998                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7999                 return NULL;
8000         }
8001
8002         for (i = 0; i < size; i++) {
8003                 pages[i] = alloc_page(gfp_flags);
8004                 if (!pages[i]) {
8005                         dprintk("%s: failed to allocate page\n", __func__);
8006                         nfs4_free_pages(pages, size);
8007                         return NULL;
8008                 }
8009         }
8010
8011         return pages;
8012 }
8013
8014 static void nfs4_layoutget_release(void *calldata)
8015 {
8016         struct nfs4_layoutget *lgp = calldata;
8017         struct inode *inode = lgp->args.inode;
8018         struct nfs_server *server = NFS_SERVER(inode);
8019         size_t max_pages = max_response_pages(server);
8020
8021         dprintk("--> %s\n", __func__);
8022         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8023         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8024         put_nfs_open_context(lgp->args.ctx);
8025         kfree(calldata);
8026         dprintk("<-- %s\n", __func__);
8027 }
8028
8029 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8030         .rpc_call_prepare = nfs4_layoutget_prepare,
8031         .rpc_call_done = nfs4_layoutget_done,
8032         .rpc_release = nfs4_layoutget_release,
8033 };
8034
8035 struct pnfs_layout_segment *
8036 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
8037 {
8038         struct inode *inode = lgp->args.inode;
8039         struct nfs_server *server = NFS_SERVER(inode);
8040         size_t max_pages = max_response_pages(server);
8041         struct rpc_task *task;
8042         struct rpc_message msg = {
8043                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8044                 .rpc_argp = &lgp->args,
8045                 .rpc_resp = &lgp->res,
8046                 .rpc_cred = lgp->cred,
8047         };
8048         struct rpc_task_setup task_setup_data = {
8049                 .rpc_client = server->client,
8050                 .rpc_message = &msg,
8051                 .callback_ops = &nfs4_layoutget_call_ops,
8052                 .callback_data = lgp,
8053                 .flags = RPC_TASK_ASYNC,
8054         };
8055         struct pnfs_layout_segment *lseg = NULL;
8056         int status = 0;
8057
8058         dprintk("--> %s\n", __func__);
8059
8060         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8061         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8062
8063         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8064         if (!lgp->args.layout.pages) {
8065                 nfs4_layoutget_release(lgp);
8066                 return ERR_PTR(-ENOMEM);
8067         }
8068         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8069         lgp->args.timestamp = jiffies;
8070
8071         lgp->res.layoutp = &lgp->args.layout;
8072         lgp->res.seq_res.sr_slot = NULL;
8073         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8074
8075         task = rpc_run_task(&task_setup_data);
8076         if (IS_ERR(task))
8077                 return ERR_CAST(task);
8078         status = nfs4_wait_for_completion_rpc_task(task);
8079         if (status == 0)
8080                 status = task->tk_status;
8081         trace_nfs4_layoutget(lgp->args.ctx,
8082                         &lgp->args.range,
8083                         &lgp->res.range,
8084                         &lgp->res.stateid,
8085                         status);
8086         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8087         if (status == 0 && lgp->res.layoutp->len)
8088                 lseg = pnfs_layout_process(lgp);
8089         rpc_put_task(task);
8090         dprintk("<-- %s status=%d\n", __func__, status);
8091         if (status)
8092                 return ERR_PTR(status);
8093         return lseg;
8094 }
8095
8096 static void
8097 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8098 {
8099         struct nfs4_layoutreturn *lrp = calldata;
8100
8101         dprintk("--> %s\n", __func__);
8102         nfs41_setup_sequence(lrp->clp->cl_session,
8103                         &lrp->args.seq_args,
8104                         &lrp->res.seq_res,
8105                         task);
8106 }
8107
8108 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8109 {
8110         struct nfs4_layoutreturn *lrp = calldata;
8111         struct nfs_server *server;
8112
8113         dprintk("--> %s\n", __func__);
8114
8115         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8116                 return;
8117
8118         server = NFS_SERVER(lrp->args.inode);
8119         switch (task->tk_status) {
8120         default:
8121                 task->tk_status = 0;
8122         case 0:
8123                 break;
8124         case -NFS4ERR_DELAY:
8125                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8126                         break;
8127                 rpc_restart_call_prepare(task);
8128                 return;
8129         }
8130         dprintk("<-- %s\n", __func__);
8131 }
8132
8133 static void nfs4_layoutreturn_release(void *calldata)
8134 {
8135         struct nfs4_layoutreturn *lrp = calldata;
8136         struct pnfs_layout_hdr *lo = lrp->args.layout;
8137         LIST_HEAD(freeme);
8138
8139         dprintk("--> %s\n", __func__);
8140         spin_lock(&lo->plh_inode->i_lock);
8141         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
8142         pnfs_mark_layout_returned_if_empty(lo);
8143         if (lrp->res.lrs_present)
8144                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8145         pnfs_clear_layoutreturn_waitbit(lo);
8146         spin_unlock(&lo->plh_inode->i_lock);
8147         pnfs_free_lseg_list(&freeme);
8148         pnfs_put_layout_hdr(lrp->args.layout);
8149         nfs_iput_and_deactive(lrp->inode);
8150         kfree(calldata);
8151         dprintk("<-- %s\n", __func__);
8152 }
8153
8154 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8155         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8156         .rpc_call_done = nfs4_layoutreturn_done,
8157         .rpc_release = nfs4_layoutreturn_release,
8158 };
8159
8160 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8161 {
8162         struct rpc_task *task;
8163         struct rpc_message msg = {
8164                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8165                 .rpc_argp = &lrp->args,
8166                 .rpc_resp = &lrp->res,
8167                 .rpc_cred = lrp->cred,
8168         };
8169         struct rpc_task_setup task_setup_data = {
8170                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8171                 .rpc_message = &msg,
8172                 .callback_ops = &nfs4_layoutreturn_call_ops,
8173                 .callback_data = lrp,
8174         };
8175         int status = 0;
8176
8177         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8178                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8179                         &task_setup_data.rpc_client, &msg);
8180
8181         dprintk("--> %s\n", __func__);
8182         if (!sync) {
8183                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8184                 if (!lrp->inode) {
8185                         nfs4_layoutreturn_release(lrp);
8186                         return -EAGAIN;
8187                 }
8188                 task_setup_data.flags |= RPC_TASK_ASYNC;
8189         }
8190         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8191         task = rpc_run_task(&task_setup_data);
8192         if (IS_ERR(task))
8193                 return PTR_ERR(task);
8194         if (sync)
8195                 status = task->tk_status;
8196         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8197         dprintk("<-- %s status=%d\n", __func__, status);
8198         rpc_put_task(task);
8199         return status;
8200 }
8201
8202 static int
8203 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8204                 struct pnfs_device *pdev,
8205                 struct rpc_cred *cred)
8206 {
8207         struct nfs4_getdeviceinfo_args args = {
8208                 .pdev = pdev,
8209                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8210                         NOTIFY_DEVICEID4_DELETE,
8211         };
8212         struct nfs4_getdeviceinfo_res res = {
8213                 .pdev = pdev,
8214         };
8215         struct rpc_message msg = {
8216                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8217                 .rpc_argp = &args,
8218                 .rpc_resp = &res,
8219                 .rpc_cred = cred,
8220         };
8221         int status;
8222
8223         dprintk("--> %s\n", __func__);
8224         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8225         if (res.notification & ~args.notify_types)
8226                 dprintk("%s: unsupported notification\n", __func__);
8227         if (res.notification != args.notify_types)
8228                 pdev->nocache = 1;
8229
8230         dprintk("<-- %s status=%d\n", __func__, status);
8231
8232         return status;
8233 }
8234
8235 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8236                 struct pnfs_device *pdev,
8237                 struct rpc_cred *cred)
8238 {
8239         struct nfs4_exception exception = { };
8240         int err;
8241
8242         do {
8243                 err = nfs4_handle_exception(server,
8244                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8245                                         &exception);
8246         } while (exception.retry);
8247         return err;
8248 }
8249 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8250
8251 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8252 {
8253         struct nfs4_layoutcommit_data *data = calldata;
8254         struct nfs_server *server = NFS_SERVER(data->args.inode);
8255         struct nfs4_session *session = nfs4_get_session(server);
8256
8257         nfs41_setup_sequence(session,
8258                         &data->args.seq_args,
8259                         &data->res.seq_res,
8260                         task);
8261 }
8262
8263 static void
8264 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8265 {
8266         struct nfs4_layoutcommit_data *data = calldata;
8267         struct nfs_server *server = NFS_SERVER(data->args.inode);
8268
8269         if (!nfs41_sequence_done(task, &data->res.seq_res))
8270                 return;
8271
8272         switch (task->tk_status) { /* Just ignore these failures */
8273         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8274         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8275         case -NFS4ERR_BADLAYOUT:     /* no layout */
8276         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8277                 task->tk_status = 0;
8278         case 0:
8279                 break;
8280         default:
8281                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8282                         rpc_restart_call_prepare(task);
8283                         return;
8284                 }
8285         }
8286 }
8287
8288 static void nfs4_layoutcommit_release(void *calldata)
8289 {
8290         struct nfs4_layoutcommit_data *data = calldata;
8291
8292         pnfs_cleanup_layoutcommit(data);
8293         nfs_post_op_update_inode_force_wcc(data->args.inode,
8294                                            data->res.fattr);
8295         put_rpccred(data->cred);
8296         nfs_iput_and_deactive(data->inode);
8297         kfree(data);
8298 }
8299
8300 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8301         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8302         .rpc_call_done = nfs4_layoutcommit_done,
8303         .rpc_release = nfs4_layoutcommit_release,
8304 };
8305
8306 int
8307 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8308 {
8309         struct rpc_message msg = {
8310                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8311                 .rpc_argp = &data->args,
8312                 .rpc_resp = &data->res,
8313                 .rpc_cred = data->cred,
8314         };
8315         struct rpc_task_setup task_setup_data = {
8316                 .task = &data->task,
8317                 .rpc_client = NFS_CLIENT(data->args.inode),
8318                 .rpc_message = &msg,
8319                 .callback_ops = &nfs4_layoutcommit_ops,
8320                 .callback_data = data,
8321         };
8322         struct rpc_task *task;
8323         int status = 0;
8324
8325         dprintk("NFS: initiating layoutcommit call. sync %d "
8326                 "lbw: %llu inode %lu\n", sync,
8327                 data->args.lastbytewritten,
8328                 data->args.inode->i_ino);
8329
8330         if (!sync) {
8331                 data->inode = nfs_igrab_and_active(data->args.inode);
8332                 if (data->inode == NULL) {
8333                         nfs4_layoutcommit_release(data);
8334                         return -EAGAIN;
8335                 }
8336                 task_setup_data.flags = RPC_TASK_ASYNC;
8337         }
8338         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8339         task = rpc_run_task(&task_setup_data);
8340         if (IS_ERR(task))
8341                 return PTR_ERR(task);
8342         if (sync)
8343                 status = task->tk_status;
8344         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8345         dprintk("%s: status %d\n", __func__, status);
8346         rpc_put_task(task);
8347         return status;
8348 }
8349
8350 /**
8351  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8352  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8353  */
8354 static int
8355 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8356                     struct nfs_fsinfo *info,
8357                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8358 {
8359         struct nfs41_secinfo_no_name_args args = {
8360                 .style = SECINFO_STYLE_CURRENT_FH,
8361         };
8362         struct nfs4_secinfo_res res = {
8363                 .flavors = flavors,
8364         };
8365         struct rpc_message msg = {
8366                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8367                 .rpc_argp = &args,
8368                 .rpc_resp = &res,
8369         };
8370         struct rpc_clnt *clnt = server->client;
8371         struct rpc_cred *cred = NULL;
8372         int status;
8373
8374         if (use_integrity) {
8375                 clnt = server->nfs_client->cl_rpcclient;
8376                 cred = nfs4_get_clid_cred(server->nfs_client);
8377                 msg.rpc_cred = cred;
8378         }
8379
8380         dprintk("--> %s\n", __func__);
8381         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8382                                 &res.seq_res, 0);
8383         dprintk("<-- %s status=%d\n", __func__, status);
8384
8385         if (cred)
8386                 put_rpccred(cred);
8387
8388         return status;
8389 }
8390
8391 static int
8392 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8393                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8394 {
8395         struct nfs4_exception exception = { };
8396         int err;
8397         do {
8398                 /* first try using integrity protection */
8399                 err = -NFS4ERR_WRONGSEC;
8400
8401                 /* try to use integrity protection with machine cred */
8402                 if (_nfs4_is_integrity_protected(server->nfs_client))
8403                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8404                                                           flavors, true);
8405
8406                 /*
8407                  * if unable to use integrity protection, or SECINFO with
8408                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8409                  * disallowed by spec, but exists in deployed servers) use
8410                  * the current filesystem's rpc_client and the user cred.
8411                  */
8412                 if (err == -NFS4ERR_WRONGSEC)
8413                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8414                                                           flavors, false);
8415
8416                 switch (err) {
8417                 case 0:
8418                 case -NFS4ERR_WRONGSEC:
8419                 case -ENOTSUPP:
8420                         goto out;
8421                 default:
8422                         err = nfs4_handle_exception(server, err, &exception);
8423                 }
8424         } while (exception.retry);
8425 out:
8426         return err;
8427 }
8428
8429 static int
8430 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8431                     struct nfs_fsinfo *info)
8432 {
8433         int err;
8434         struct page *page;
8435         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8436         struct nfs4_secinfo_flavors *flavors;
8437         struct nfs4_secinfo4 *secinfo;
8438         int i;
8439
8440         page = alloc_page(GFP_KERNEL);
8441         if (!page) {
8442                 err = -ENOMEM;
8443                 goto out;
8444         }
8445
8446         flavors = page_address(page);
8447         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8448
8449         /*
8450          * Fall back on "guess and check" method if
8451          * the server doesn't support SECINFO_NO_NAME
8452          */
8453         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8454                 err = nfs4_find_root_sec(server, fhandle, info);
8455                 goto out_freepage;
8456         }
8457         if (err)
8458                 goto out_freepage;
8459
8460         for (i = 0; i < flavors->num_flavors; i++) {
8461                 secinfo = &flavors->flavors[i];
8462
8463                 switch (secinfo->flavor) {
8464                 case RPC_AUTH_NULL:
8465                 case RPC_AUTH_UNIX:
8466                 case RPC_AUTH_GSS:
8467                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8468                                         &secinfo->flavor_info);
8469                         break;
8470                 default:
8471                         flavor = RPC_AUTH_MAXFLAVOR;
8472                         break;
8473                 }
8474
8475                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8476                         flavor = RPC_AUTH_MAXFLAVOR;
8477
8478                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8479                         err = nfs4_lookup_root_sec(server, fhandle,
8480                                                    info, flavor);
8481                         if (!err)
8482                                 break;
8483                 }
8484         }
8485
8486         if (flavor == RPC_AUTH_MAXFLAVOR)
8487                 err = -EPERM;
8488
8489 out_freepage:
8490         put_page(page);
8491         if (err == -EACCES)
8492                 return -EPERM;
8493 out:
8494         return err;
8495 }
8496
8497 static int _nfs41_test_stateid(struct nfs_server *server,
8498                 nfs4_stateid *stateid,
8499                 struct rpc_cred *cred)
8500 {
8501         int status;
8502         struct nfs41_test_stateid_args args = {
8503                 .stateid = stateid,
8504         };
8505         struct nfs41_test_stateid_res res;
8506         struct rpc_message msg = {
8507                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8508                 .rpc_argp = &args,
8509                 .rpc_resp = &res,
8510                 .rpc_cred = cred,
8511         };
8512         struct rpc_clnt *rpc_client = server->client;
8513
8514         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8515                 &rpc_client, &msg);
8516
8517         dprintk("NFS call  test_stateid %p\n", stateid);
8518         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8519         nfs4_set_sequence_privileged(&args.seq_args);
8520         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8521                         &args.seq_args, &res.seq_res);
8522         if (status != NFS_OK) {
8523                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8524                 return status;
8525         }
8526         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8527         return -res.status;
8528 }
8529
8530 /**
8531  * nfs41_test_stateid - perform a TEST_STATEID operation
8532  *
8533  * @server: server / transport on which to perform the operation
8534  * @stateid: state ID to test
8535  * @cred: credential
8536  *
8537  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8538  * Otherwise a negative NFS4ERR value is returned if the operation
8539  * failed or the state ID is not currently valid.
8540  */
8541 static int nfs41_test_stateid(struct nfs_server *server,
8542                 nfs4_stateid *stateid,
8543                 struct rpc_cred *cred)
8544 {
8545         struct nfs4_exception exception = { };
8546         int err;
8547         do {
8548                 err = _nfs41_test_stateid(server, stateid, cred);
8549                 if (err != -NFS4ERR_DELAY)
8550                         break;
8551                 nfs4_handle_exception(server, err, &exception);
8552         } while (exception.retry);
8553         return err;
8554 }
8555
8556 struct nfs_free_stateid_data {
8557         struct nfs_server *server;
8558         struct nfs41_free_stateid_args args;
8559         struct nfs41_free_stateid_res res;
8560 };
8561
8562 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8563 {
8564         struct nfs_free_stateid_data *data = calldata;
8565         nfs41_setup_sequence(nfs4_get_session(data->server),
8566                         &data->args.seq_args,
8567                         &data->res.seq_res,
8568                         task);
8569 }
8570
8571 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8572 {
8573         struct nfs_free_stateid_data *data = calldata;
8574
8575         nfs41_sequence_done(task, &data->res.seq_res);
8576
8577         switch (task->tk_status) {
8578         case -NFS4ERR_DELAY:
8579                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8580                         rpc_restart_call_prepare(task);
8581         }
8582 }
8583
8584 static void nfs41_free_stateid_release(void *calldata)
8585 {
8586         kfree(calldata);
8587 }
8588
8589 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8590         .rpc_call_prepare = nfs41_free_stateid_prepare,
8591         .rpc_call_done = nfs41_free_stateid_done,
8592         .rpc_release = nfs41_free_stateid_release,
8593 };
8594
8595 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8596                 nfs4_stateid *stateid,
8597                 struct rpc_cred *cred,
8598                 bool privileged)
8599 {
8600         struct rpc_message msg = {
8601                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8602                 .rpc_cred = cred,
8603         };
8604         struct rpc_task_setup task_setup = {
8605                 .rpc_client = server->client,
8606                 .rpc_message = &msg,
8607                 .callback_ops = &nfs41_free_stateid_ops,
8608                 .flags = RPC_TASK_ASYNC,
8609         };
8610         struct nfs_free_stateid_data *data;
8611
8612         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8613                 &task_setup.rpc_client, &msg);
8614
8615         dprintk("NFS call  free_stateid %p\n", stateid);
8616         data = kmalloc(sizeof(*data), GFP_NOFS);
8617         if (!data)
8618                 return ERR_PTR(-ENOMEM);
8619         data->server = server;
8620         nfs4_stateid_copy(&data->args.stateid, stateid);
8621
8622         task_setup.callback_data = data;
8623
8624         msg.rpc_argp = &data->args;
8625         msg.rpc_resp = &data->res;
8626         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8627         if (privileged)
8628                 nfs4_set_sequence_privileged(&data->args.seq_args);
8629
8630         return rpc_run_task(&task_setup);
8631 }
8632
8633 /**
8634  * nfs41_free_stateid - perform a FREE_STATEID operation
8635  *
8636  * @server: server / transport on which to perform the operation
8637  * @stateid: state ID to release
8638  * @cred: credential
8639  *
8640  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8641  * negative NFS4ERR value is returned.
8642  */
8643 static int nfs41_free_stateid(struct nfs_server *server,
8644                 nfs4_stateid *stateid,
8645                 struct rpc_cred *cred)
8646 {
8647         struct rpc_task *task;
8648         int ret;
8649
8650         task = _nfs41_free_stateid(server, stateid, cred, true);
8651         if (IS_ERR(task))
8652                 return PTR_ERR(task);
8653         ret = rpc_wait_for_completion_task(task);
8654         if (!ret)
8655                 ret = task->tk_status;
8656         rpc_put_task(task);
8657         return ret;
8658 }
8659
8660 static void
8661 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8662 {
8663         struct rpc_task *task;
8664         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8665
8666         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8667         nfs4_free_lock_state(server, lsp);
8668         if (IS_ERR(task))
8669                 return;
8670         rpc_put_task(task);
8671 }
8672
8673 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8674                 const nfs4_stateid *s2)
8675 {
8676         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8677                 return false;
8678
8679         if (s1->seqid == s2->seqid)
8680                 return true;
8681         if (s1->seqid == 0 || s2->seqid == 0)
8682                 return true;
8683
8684         return false;
8685 }
8686
8687 #endif /* CONFIG_NFS_V4_1 */
8688
8689 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8690                 const nfs4_stateid *s2)
8691 {
8692         return nfs4_stateid_match(s1, s2);
8693 }
8694
8695
8696 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8697         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8698         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8699         .recover_open   = nfs4_open_reclaim,
8700         .recover_lock   = nfs4_lock_reclaim,
8701         .establish_clid = nfs4_init_clientid,
8702         .detect_trunking = nfs40_discover_server_trunking,
8703 };
8704
8705 #if defined(CONFIG_NFS_V4_1)
8706 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8707         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8708         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8709         .recover_open   = nfs4_open_reclaim,
8710         .recover_lock   = nfs4_lock_reclaim,
8711         .establish_clid = nfs41_init_clientid,
8712         .reclaim_complete = nfs41_proc_reclaim_complete,
8713         .detect_trunking = nfs41_discover_server_trunking,
8714 };
8715 #endif /* CONFIG_NFS_V4_1 */
8716
8717 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8718         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8719         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8720         .recover_open   = nfs40_open_expired,
8721         .recover_lock   = nfs4_lock_expired,
8722         .establish_clid = nfs4_init_clientid,
8723 };
8724
8725 #if defined(CONFIG_NFS_V4_1)
8726 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8727         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8728         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8729         .recover_open   = nfs41_open_expired,
8730         .recover_lock   = nfs41_lock_expired,
8731         .establish_clid = nfs41_init_clientid,
8732 };
8733 #endif /* CONFIG_NFS_V4_1 */
8734
8735 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8736         .sched_state_renewal = nfs4_proc_async_renew,
8737         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8738         .renew_lease = nfs4_proc_renew,
8739 };
8740
8741 #if defined(CONFIG_NFS_V4_1)
8742 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8743         .sched_state_renewal = nfs41_proc_async_sequence,
8744         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8745         .renew_lease = nfs4_proc_sequence,
8746 };
8747 #endif
8748
8749 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8750         .get_locations = _nfs40_proc_get_locations,
8751         .fsid_present = _nfs40_proc_fsid_present,
8752 };
8753
8754 #if defined(CONFIG_NFS_V4_1)
8755 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8756         .get_locations = _nfs41_proc_get_locations,
8757         .fsid_present = _nfs41_proc_fsid_present,
8758 };
8759 #endif  /* CONFIG_NFS_V4_1 */
8760
8761 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8762         .minor_version = 0,
8763         .init_caps = NFS_CAP_READDIRPLUS
8764                 | NFS_CAP_ATOMIC_OPEN
8765                 | NFS_CAP_POSIX_LOCK,
8766         .init_client = nfs40_init_client,
8767         .shutdown_client = nfs40_shutdown_client,
8768         .match_stateid = nfs4_match_stateid,
8769         .find_root_sec = nfs4_find_root_sec,
8770         .free_lock_state = nfs4_release_lockowner,
8771         .alloc_seqid = nfs_alloc_seqid,
8772         .call_sync_ops = &nfs40_call_sync_ops,
8773         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8774         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8775         .state_renewal_ops = &nfs40_state_renewal_ops,
8776         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8777 };
8778
8779 #if defined(CONFIG_NFS_V4_1)
8780 static struct nfs_seqid *
8781 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8782 {
8783         return NULL;
8784 }
8785
8786 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8787         .minor_version = 1,
8788         .init_caps = NFS_CAP_READDIRPLUS
8789                 | NFS_CAP_ATOMIC_OPEN
8790                 | NFS_CAP_POSIX_LOCK
8791                 | NFS_CAP_STATEID_NFSV41
8792                 | NFS_CAP_ATOMIC_OPEN_V1,
8793         .init_client = nfs41_init_client,
8794         .shutdown_client = nfs41_shutdown_client,
8795         .match_stateid = nfs41_match_stateid,
8796         .find_root_sec = nfs41_find_root_sec,
8797         .free_lock_state = nfs41_free_lock_state,
8798         .alloc_seqid = nfs_alloc_no_seqid,
8799         .call_sync_ops = &nfs41_call_sync_ops,
8800         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8801         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8802         .state_renewal_ops = &nfs41_state_renewal_ops,
8803         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8804 };
8805 #endif
8806
8807 #if defined(CONFIG_NFS_V4_2)
8808 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8809         .minor_version = 2,
8810         .init_caps = NFS_CAP_READDIRPLUS
8811                 | NFS_CAP_ATOMIC_OPEN
8812                 | NFS_CAP_POSIX_LOCK
8813                 | NFS_CAP_STATEID_NFSV41
8814                 | NFS_CAP_ATOMIC_OPEN_V1
8815                 | NFS_CAP_ALLOCATE
8816                 | NFS_CAP_COPY
8817                 | NFS_CAP_DEALLOCATE
8818                 | NFS_CAP_SEEK
8819                 | NFS_CAP_LAYOUTSTATS
8820                 | NFS_CAP_CLONE,
8821         .init_client = nfs41_init_client,
8822         .shutdown_client = nfs41_shutdown_client,
8823         .match_stateid = nfs41_match_stateid,
8824         .find_root_sec = nfs41_find_root_sec,
8825         .free_lock_state = nfs41_free_lock_state,
8826         .call_sync_ops = &nfs41_call_sync_ops,
8827         .alloc_seqid = nfs_alloc_no_seqid,
8828         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8829         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8830         .state_renewal_ops = &nfs41_state_renewal_ops,
8831         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8832 };
8833 #endif
8834
8835 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8836         [0] = &nfs_v4_0_minor_ops,
8837 #if defined(CONFIG_NFS_V4_1)
8838         [1] = &nfs_v4_1_minor_ops,
8839 #endif
8840 #if defined(CONFIG_NFS_V4_2)
8841         [2] = &nfs_v4_2_minor_ops,
8842 #endif
8843 };
8844
8845 ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
8846 {
8847         ssize_t error, error2;
8848
8849         error = generic_listxattr(dentry, list, size);
8850         if (error < 0)
8851                 return error;
8852         if (list) {
8853                 list += error;
8854                 size -= error;
8855         }
8856
8857         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
8858         if (error2 < 0)
8859                 return error2;
8860         return error + error2;
8861 }
8862
8863 static const struct inode_operations nfs4_dir_inode_operations = {
8864         .create         = nfs_create,
8865         .lookup         = nfs_lookup,
8866         .atomic_open    = nfs_atomic_open,
8867         .link           = nfs_link,
8868         .unlink         = nfs_unlink,
8869         .symlink        = nfs_symlink,
8870         .mkdir          = nfs_mkdir,
8871         .rmdir          = nfs_rmdir,
8872         .mknod          = nfs_mknod,
8873         .rename         = nfs_rename,
8874         .permission     = nfs_permission,
8875         .getattr        = nfs_getattr,
8876         .setattr        = nfs_setattr,
8877         .getxattr       = generic_getxattr,
8878         .setxattr       = generic_setxattr,
8879         .listxattr      = nfs4_listxattr,
8880         .removexattr    = generic_removexattr,
8881 };
8882
8883 static const struct inode_operations nfs4_file_inode_operations = {
8884         .permission     = nfs_permission,
8885         .getattr        = nfs_getattr,
8886         .setattr        = nfs_setattr,
8887         .getxattr       = generic_getxattr,
8888         .setxattr       = generic_setxattr,
8889         .listxattr      = nfs4_listxattr,
8890         .removexattr    = generic_removexattr,
8891 };
8892
8893 const struct nfs_rpc_ops nfs_v4_clientops = {
8894         .version        = 4,                    /* protocol version */
8895         .dentry_ops     = &nfs4_dentry_operations,
8896         .dir_inode_ops  = &nfs4_dir_inode_operations,
8897         .file_inode_ops = &nfs4_file_inode_operations,
8898         .file_ops       = &nfs4_file_operations,
8899         .getroot        = nfs4_proc_get_root,
8900         .submount       = nfs4_submount,
8901         .try_mount      = nfs4_try_mount,
8902         .getattr        = nfs4_proc_getattr,
8903         .setattr        = nfs4_proc_setattr,
8904         .lookup         = nfs4_proc_lookup,
8905         .access         = nfs4_proc_access,
8906         .readlink       = nfs4_proc_readlink,
8907         .create         = nfs4_proc_create,
8908         .remove         = nfs4_proc_remove,
8909         .unlink_setup   = nfs4_proc_unlink_setup,
8910         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8911         .unlink_done    = nfs4_proc_unlink_done,
8912         .rename_setup   = nfs4_proc_rename_setup,
8913         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8914         .rename_done    = nfs4_proc_rename_done,
8915         .link           = nfs4_proc_link,
8916         .symlink        = nfs4_proc_symlink,
8917         .mkdir          = nfs4_proc_mkdir,
8918         .rmdir          = nfs4_proc_remove,
8919         .readdir        = nfs4_proc_readdir,
8920         .mknod          = nfs4_proc_mknod,
8921         .statfs         = nfs4_proc_statfs,
8922         .fsinfo         = nfs4_proc_fsinfo,
8923         .pathconf       = nfs4_proc_pathconf,
8924         .set_capabilities = nfs4_server_capabilities,
8925         .decode_dirent  = nfs4_decode_dirent,
8926         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8927         .read_setup     = nfs4_proc_read_setup,
8928         .read_done      = nfs4_read_done,
8929         .write_setup    = nfs4_proc_write_setup,
8930         .write_done     = nfs4_write_done,
8931         .commit_setup   = nfs4_proc_commit_setup,
8932         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8933         .commit_done    = nfs4_commit_done,
8934         .lock           = nfs4_proc_lock,
8935         .clear_acl_cache = nfs4_zap_acl_attr,
8936         .close_context  = nfs4_close_context,
8937         .open_context   = nfs4_atomic_open,
8938         .have_delegation = nfs4_have_delegation,
8939         .return_delegation = nfs4_inode_return_delegation,
8940         .alloc_client   = nfs4_alloc_client,
8941         .init_client    = nfs4_init_client,
8942         .free_client    = nfs4_free_client,
8943         .create_server  = nfs4_create_server,
8944         .clone_server   = nfs_clone_server,
8945 };
8946
8947 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8948         .name   = XATTR_NAME_NFSV4_ACL,
8949         .list   = nfs4_xattr_list_nfs4_acl,
8950         .get    = nfs4_xattr_get_nfs4_acl,
8951         .set    = nfs4_xattr_set_nfs4_acl,
8952 };
8953
8954 const struct xattr_handler *nfs4_xattr_handlers[] = {
8955         &nfs4_xattr_nfs4_acl_handler,
8956 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8957         &nfs4_xattr_nfs4_label_handler,
8958 #endif
8959         NULL
8960 };
8961
8962 /*
8963  * Local variables:
8964  *  c-basic-offset: 8
8965  * End:
8966  */