Merge branches 'pm-sleep', 'pm-domains', 'powercap' and 'pm-tools'
[linux-2.6-microblaze.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57 #include <linux/iversion.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_BITMASK_SZ         3
75
76 #define NFS4_POLL_RETRY_MIN     (HZ/10)
77 #define NFS4_POLL_RETRY_MAX     (15*HZ)
78
79 /* file attributes which can be mapped to nfs attributes */
80 #define NFS4_VALID_ATTRS (ATTR_MODE \
81         | ATTR_UID \
82         | ATTR_GID \
83         | ATTR_SIZE \
84         | ATTR_ATIME \
85         | ATTR_MTIME \
86         | ATTR_CTIME \
87         | ATTR_ATIME_SET \
88         | ATTR_MTIME_SET)
89
90 struct nfs4_opendata;
91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
95 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
96                             struct nfs_fattr *fattr, struct iattr *sattr,
97                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
98                             struct nfs4_label *olabel);
99 #ifdef CONFIG_NFS_V4_1
100 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
101                 const struct cred *cred,
102                 struct nfs4_slot *slot,
103                 bool is_privileged);
104 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
105                 const struct cred *);
106 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
107                 const struct cred *, bool);
108 #endif
109
110 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
111 static inline struct nfs4_label *
112 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
113         struct iattr *sattr, struct nfs4_label *label)
114 {
115         int err;
116
117         if (label == NULL)
118                 return NULL;
119
120         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
121                 return NULL;
122
123         err = security_dentry_init_security(dentry, sattr->ia_mode,
124                                 &dentry->d_name, (void **)&label->label, &label->len);
125         if (err == 0)
126                 return label;
127
128         return NULL;
129 }
130 static inline void
131 nfs4_label_release_security(struct nfs4_label *label)
132 {
133         if (label)
134                 security_release_secctx(label->label, label->len);
135 }
136 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
137 {
138         if (label)
139                 return server->attr_bitmask;
140
141         return server->attr_bitmask_nl;
142 }
143 #else
144 static inline struct nfs4_label *
145 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
146         struct iattr *sattr, struct nfs4_label *l)
147 { return NULL; }
148 static inline void
149 nfs4_label_release_security(struct nfs4_label *label)
150 { return; }
151 static inline u32 *
152 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
153 { return server->attr_bitmask; }
154 #endif
155
156 /* Prevent leaks of NFSv4 errors into userland */
157 static int nfs4_map_errors(int err)
158 {
159         if (err >= -1000)
160                 return err;
161         switch (err) {
162         case -NFS4ERR_RESOURCE:
163         case -NFS4ERR_LAYOUTTRYLATER:
164         case -NFS4ERR_RECALLCONFLICT:
165                 return -EREMOTEIO;
166         case -NFS4ERR_WRONGSEC:
167         case -NFS4ERR_WRONG_CRED:
168                 return -EPERM;
169         case -NFS4ERR_BADOWNER:
170         case -NFS4ERR_BADNAME:
171                 return -EINVAL;
172         case -NFS4ERR_SHARE_DENIED:
173                 return -EACCES;
174         case -NFS4ERR_MINOR_VERS_MISMATCH:
175                 return -EPROTONOSUPPORT;
176         case -NFS4ERR_FILE_OPEN:
177                 return -EBUSY;
178         default:
179                 dprintk("%s could not handle NFSv4 error %d\n",
180                                 __func__, -err);
181                 break;
182         }
183         return -EIO;
184 }
185
186 /*
187  * This is our standard bitmap for GETATTR requests.
188  */
189 const u32 nfs4_fattr_bitmap[3] = {
190         FATTR4_WORD0_TYPE
191         | FATTR4_WORD0_CHANGE
192         | FATTR4_WORD0_SIZE
193         | FATTR4_WORD0_FSID
194         | FATTR4_WORD0_FILEID,
195         FATTR4_WORD1_MODE
196         | FATTR4_WORD1_NUMLINKS
197         | FATTR4_WORD1_OWNER
198         | FATTR4_WORD1_OWNER_GROUP
199         | FATTR4_WORD1_RAWDEV
200         | FATTR4_WORD1_SPACE_USED
201         | FATTR4_WORD1_TIME_ACCESS
202         | FATTR4_WORD1_TIME_METADATA
203         | FATTR4_WORD1_TIME_MODIFY
204         | FATTR4_WORD1_MOUNTED_ON_FILEID,
205 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
206         FATTR4_WORD2_SECURITY_LABEL
207 #endif
208 };
209
210 static const u32 nfs4_pnfs_open_bitmap[3] = {
211         FATTR4_WORD0_TYPE
212         | FATTR4_WORD0_CHANGE
213         | FATTR4_WORD0_SIZE
214         | FATTR4_WORD0_FSID
215         | FATTR4_WORD0_FILEID,
216         FATTR4_WORD1_MODE
217         | FATTR4_WORD1_NUMLINKS
218         | FATTR4_WORD1_OWNER
219         | FATTR4_WORD1_OWNER_GROUP
220         | FATTR4_WORD1_RAWDEV
221         | FATTR4_WORD1_SPACE_USED
222         | FATTR4_WORD1_TIME_ACCESS
223         | FATTR4_WORD1_TIME_METADATA
224         | FATTR4_WORD1_TIME_MODIFY,
225         FATTR4_WORD2_MDSTHRESHOLD
226 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
227         | FATTR4_WORD2_SECURITY_LABEL
228 #endif
229 };
230
231 static const u32 nfs4_open_noattr_bitmap[3] = {
232         FATTR4_WORD0_TYPE
233         | FATTR4_WORD0_FILEID,
234 };
235
236 const u32 nfs4_statfs_bitmap[3] = {
237         FATTR4_WORD0_FILES_AVAIL
238         | FATTR4_WORD0_FILES_FREE
239         | FATTR4_WORD0_FILES_TOTAL,
240         FATTR4_WORD1_SPACE_AVAIL
241         | FATTR4_WORD1_SPACE_FREE
242         | FATTR4_WORD1_SPACE_TOTAL
243 };
244
245 const u32 nfs4_pathconf_bitmap[3] = {
246         FATTR4_WORD0_MAXLINK
247         | FATTR4_WORD0_MAXNAME,
248         0
249 };
250
251 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
252                         | FATTR4_WORD0_MAXREAD
253                         | FATTR4_WORD0_MAXWRITE
254                         | FATTR4_WORD0_LEASE_TIME,
255                         FATTR4_WORD1_TIME_DELTA
256                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
257                         FATTR4_WORD2_LAYOUT_BLKSIZE
258                         | FATTR4_WORD2_CLONE_BLKSIZE
259 };
260
261 const u32 nfs4_fs_locations_bitmap[3] = {
262         FATTR4_WORD0_CHANGE
263         | FATTR4_WORD0_SIZE
264         | FATTR4_WORD0_FSID
265         | FATTR4_WORD0_FILEID
266         | FATTR4_WORD0_FS_LOCATIONS,
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_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
278                 struct inode *inode)
279 {
280         unsigned long cache_validity;
281
282         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
283         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
284                 return;
285
286         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
287         if (!(cache_validity & NFS_INO_REVAL_FORCED))
288                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
289                                 | NFS_INO_INVALID_SIZE);
290
291         if (!(cache_validity & NFS_INO_INVALID_SIZE))
292                 dst[0] &= ~FATTR4_WORD0_SIZE;
293
294         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
295                 dst[0] &= ~FATTR4_WORD0_CHANGE;
296 }
297
298 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
299                 const __u32 *src, struct inode *inode)
300 {
301         nfs4_bitmap_copy_adjust(dst, src, inode);
302 }
303
304 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
305                 struct nfs4_readdir_arg *readdir)
306 {
307         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
308         __be32 *start, *p;
309
310         if (cookie > 2) {
311                 readdir->cookie = cookie;
312                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
313                 return;
314         }
315
316         readdir->cookie = 0;
317         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
318         if (cookie == 2)
319                 return;
320         
321         /*
322          * NFSv4 servers do not return entries for '.' and '..'
323          * Therefore, we fake these entries here.  We let '.'
324          * have cookie 0 and '..' have cookie 1.  Note that
325          * when talking to the server, we always send cookie 0
326          * instead of 1 or 2.
327          */
328         start = p = kmap_atomic(*readdir->pages);
329         
330         if (cookie == 0) {
331                 *p++ = xdr_one;                                  /* next */
332                 *p++ = xdr_zero;                   /* cookie, first word */
333                 *p++ = xdr_one;                   /* cookie, second word */
334                 *p++ = xdr_one;                             /* entry len */
335                 memcpy(p, ".\0\0\0", 4);                        /* entry */
336                 p++;
337                 *p++ = xdr_one;                         /* bitmap length */
338                 *p++ = htonl(attrs);                           /* bitmap */
339                 *p++ = htonl(12);             /* attribute buffer length */
340                 *p++ = htonl(NF4DIR);
341                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
342         }
343         
344         *p++ = xdr_one;                                  /* next */
345         *p++ = xdr_zero;                   /* cookie, first word */
346         *p++ = xdr_two;                   /* cookie, second word */
347         *p++ = xdr_two;                             /* entry len */
348         memcpy(p, "..\0\0", 4);                         /* entry */
349         p++;
350         *p++ = xdr_one;                         /* bitmap length */
351         *p++ = htonl(attrs);                           /* bitmap */
352         *p++ = htonl(12);             /* attribute buffer length */
353         *p++ = htonl(NF4DIR);
354         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
355
356         readdir->pgbase = (char *)p - (char *)start;
357         readdir->count -= readdir->pgbase;
358         kunmap_atomic(start);
359 }
360
361 static void nfs4_test_and_free_stateid(struct nfs_server *server,
362                 nfs4_stateid *stateid,
363                 const struct cred *cred)
364 {
365         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
366
367         ops->test_and_free_expired(server, stateid, cred);
368 }
369
370 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
371                 nfs4_stateid *stateid,
372                 const struct cred *cred)
373 {
374         stateid->type = NFS4_REVOKED_STATEID_TYPE;
375         nfs4_test_and_free_stateid(server, stateid, cred);
376 }
377
378 static void nfs4_free_revoked_stateid(struct nfs_server *server,
379                 const nfs4_stateid *stateid,
380                 const struct cred *cred)
381 {
382         nfs4_stateid tmp;
383
384         nfs4_stateid_copy(&tmp, stateid);
385         __nfs4_free_revoked_stateid(server, &tmp, cred);
386 }
387
388 static long nfs4_update_delay(long *timeout)
389 {
390         long ret;
391         if (!timeout)
392                 return NFS4_POLL_RETRY_MAX;
393         if (*timeout <= 0)
394                 *timeout = NFS4_POLL_RETRY_MIN;
395         if (*timeout > NFS4_POLL_RETRY_MAX)
396                 *timeout = NFS4_POLL_RETRY_MAX;
397         ret = *timeout;
398         *timeout <<= 1;
399         return ret;
400 }
401
402 static int nfs4_delay_killable(long *timeout)
403 {
404         might_sleep();
405
406         freezable_schedule_timeout_killable_unsafe(
407                 nfs4_update_delay(timeout));
408         if (!__fatal_signal_pending(current))
409                 return 0;
410         return -EINTR;
411 }
412
413 static int nfs4_delay_interruptible(long *timeout)
414 {
415         might_sleep();
416
417         freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
418         if (!signal_pending(current))
419                 return 0;
420         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
421 }
422
423 static int nfs4_delay(long *timeout, bool interruptible)
424 {
425         if (interruptible)
426                 return nfs4_delay_interruptible(timeout);
427         return nfs4_delay_killable(timeout);
428 }
429
430 static const nfs4_stateid *
431 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
432 {
433         if (!stateid)
434                 return NULL;
435         switch (stateid->type) {
436         case NFS4_OPEN_STATEID_TYPE:
437         case NFS4_LOCK_STATEID_TYPE:
438         case NFS4_DELEGATION_STATEID_TYPE:
439                 return stateid;
440         default:
441                 break;
442         }
443         return NULL;
444 }
445
446 /* This is the error handling routine for processes that are allowed
447  * to sleep.
448  */
449 static int nfs4_do_handle_exception(struct nfs_server *server,
450                 int errorcode, struct nfs4_exception *exception)
451 {
452         struct nfs_client *clp = server->nfs_client;
453         struct nfs4_state *state = exception->state;
454         const nfs4_stateid *stateid;
455         struct inode *inode = exception->inode;
456         int ret = errorcode;
457
458         exception->delay = 0;
459         exception->recovering = 0;
460         exception->retry = 0;
461
462         stateid = nfs4_recoverable_stateid(exception->stateid);
463         if (stateid == NULL && state != NULL)
464                 stateid = nfs4_recoverable_stateid(&state->stateid);
465
466         switch(errorcode) {
467                 case 0:
468                         return 0;
469                 case -NFS4ERR_BADHANDLE:
470                 case -ESTALE:
471                         if (inode != NULL && S_ISREG(inode->i_mode))
472                                 pnfs_destroy_layout(NFS_I(inode));
473                         break;
474                 case -NFS4ERR_DELEG_REVOKED:
475                 case -NFS4ERR_ADMIN_REVOKED:
476                 case -NFS4ERR_EXPIRED:
477                 case -NFS4ERR_BAD_STATEID:
478                 case -NFS4ERR_PARTNER_NO_AUTH:
479                         if (inode != NULL && stateid != NULL) {
480                                 nfs_inode_find_state_and_recover(inode,
481                                                 stateid);
482                                 goto wait_on_recovery;
483                         }
484                         /* Fall through */
485                 case -NFS4ERR_OPENMODE:
486                         if (inode) {
487                                 int err;
488
489                                 err = nfs_async_inode_return_delegation(inode,
490                                                 stateid);
491                                 if (err == 0)
492                                         goto wait_on_recovery;
493                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
494                                         exception->retry = 1;
495                                         break;
496                                 }
497                         }
498                         if (state == NULL)
499                                 break;
500                         ret = nfs4_schedule_stateid_recovery(server, state);
501                         if (ret < 0)
502                                 break;
503                         goto wait_on_recovery;
504                 case -NFS4ERR_STALE_STATEID:
505                 case -NFS4ERR_STALE_CLIENTID:
506                         nfs4_schedule_lease_recovery(clp);
507                         goto wait_on_recovery;
508                 case -NFS4ERR_MOVED:
509                         ret = nfs4_schedule_migration_recovery(server);
510                         if (ret < 0)
511                                 break;
512                         goto wait_on_recovery;
513                 case -NFS4ERR_LEASE_MOVED:
514                         nfs4_schedule_lease_moved_recovery(clp);
515                         goto wait_on_recovery;
516 #if defined(CONFIG_NFS_V4_1)
517                 case -NFS4ERR_BADSESSION:
518                 case -NFS4ERR_BADSLOT:
519                 case -NFS4ERR_BAD_HIGH_SLOT:
520                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
521                 case -NFS4ERR_DEADSESSION:
522                 case -NFS4ERR_SEQ_FALSE_RETRY:
523                 case -NFS4ERR_SEQ_MISORDERED:
524                         /* Handled in nfs41_sequence_process() */
525                         goto wait_on_recovery;
526 #endif /* defined(CONFIG_NFS_V4_1) */
527                 case -NFS4ERR_FILE_OPEN:
528                         if (exception->timeout > HZ) {
529                                 /* We have retried a decent amount, time to
530                                  * fail
531                                  */
532                                 ret = -EBUSY;
533                                 break;
534                         }
535                         /* Fall through */
536                 case -NFS4ERR_DELAY:
537                         nfs_inc_server_stats(server, NFSIOS_DELAY);
538                         /* Fall through */
539                 case -NFS4ERR_GRACE:
540                 case -NFS4ERR_LAYOUTTRYLATER:
541                 case -NFS4ERR_RECALLCONFLICT:
542                         exception->delay = 1;
543                         return 0;
544
545                 case -NFS4ERR_RETRY_UNCACHED_REP:
546                 case -NFS4ERR_OLD_STATEID:
547                         exception->retry = 1;
548                         break;
549                 case -NFS4ERR_BADOWNER:
550                         /* The following works around a Linux server bug! */
551                 case -NFS4ERR_BADNAME:
552                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
553                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
554                                 exception->retry = 1;
555                                 printk(KERN_WARNING "NFS: v4 server %s "
556                                                 "does not accept raw "
557                                                 "uid/gids. "
558                                                 "Reenabling the idmapper.\n",
559                                                 server->nfs_client->cl_hostname);
560                         }
561         }
562         /* We failed to handle the error */
563         return nfs4_map_errors(ret);
564 wait_on_recovery:
565         exception->recovering = 1;
566         return 0;
567 }
568
569 /* This is the error handling routine for processes that are allowed
570  * to sleep.
571  */
572 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
573 {
574         struct nfs_client *clp = server->nfs_client;
575         int ret;
576
577         ret = nfs4_do_handle_exception(server, errorcode, exception);
578         if (exception->delay) {
579                 ret = nfs4_delay(&exception->timeout,
580                                 exception->interruptible);
581                 goto out_retry;
582         }
583         if (exception->recovering) {
584                 ret = nfs4_wait_clnt_recover(clp);
585                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
586                         return -EIO;
587                 goto out_retry;
588         }
589         return ret;
590 out_retry:
591         if (ret == 0)
592                 exception->retry = 1;
593         return ret;
594 }
595
596 static int
597 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
598                 int errorcode, struct nfs4_exception *exception)
599 {
600         struct nfs_client *clp = server->nfs_client;
601         int ret;
602
603         ret = nfs4_do_handle_exception(server, errorcode, exception);
604         if (exception->delay) {
605                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
606                 goto out_retry;
607         }
608         if (exception->recovering) {
609                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
610                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
611                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
612                 goto out_retry;
613         }
614         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
615                 ret = -EIO;
616         return ret;
617 out_retry:
618         if (ret == 0) {
619                 exception->retry = 1;
620                 /*
621                  * For NFS4ERR_MOVED, the client transport will need to
622                  * be recomputed after migration recovery has completed.
623                  */
624                 if (errorcode == -NFS4ERR_MOVED)
625                         rpc_task_release_transport(task);
626         }
627         return ret;
628 }
629
630 int
631 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
632                         struct nfs4_state *state, long *timeout)
633 {
634         struct nfs4_exception exception = {
635                 .state = state,
636         };
637
638         if (task->tk_status >= 0)
639                 return 0;
640         if (timeout)
641                 exception.timeout = *timeout;
642         task->tk_status = nfs4_async_handle_exception(task, server,
643                         task->tk_status,
644                         &exception);
645         if (exception.delay && timeout)
646                 *timeout = exception.timeout;
647         if (exception.retry)
648                 return -EAGAIN;
649         return 0;
650 }
651
652 /*
653  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
654  * or 'false' otherwise.
655  */
656 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
657 {
658         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
659         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
660 }
661
662 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
663 {
664         spin_lock(&clp->cl_lock);
665         if (time_before(clp->cl_last_renewal,timestamp))
666                 clp->cl_last_renewal = timestamp;
667         spin_unlock(&clp->cl_lock);
668 }
669
670 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
671 {
672         struct nfs_client *clp = server->nfs_client;
673
674         if (!nfs4_has_session(clp))
675                 do_renew_lease(clp, timestamp);
676 }
677
678 struct nfs4_call_sync_data {
679         const struct nfs_server *seq_server;
680         struct nfs4_sequence_args *seq_args;
681         struct nfs4_sequence_res *seq_res;
682 };
683
684 void nfs4_init_sequence(struct nfs4_sequence_args *args,
685                         struct nfs4_sequence_res *res, int cache_reply,
686                         int privileged)
687 {
688         args->sa_slot = NULL;
689         args->sa_cache_this = cache_reply;
690         args->sa_privileged = privileged;
691
692         res->sr_slot = NULL;
693 }
694
695 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
696 {
697         struct nfs4_slot *slot = res->sr_slot;
698         struct nfs4_slot_table *tbl;
699
700         tbl = slot->table;
701         spin_lock(&tbl->slot_tbl_lock);
702         if (!nfs41_wake_and_assign_slot(tbl, slot))
703                 nfs4_free_slot(tbl, slot);
704         spin_unlock(&tbl->slot_tbl_lock);
705
706         res->sr_slot = NULL;
707 }
708
709 static int nfs40_sequence_done(struct rpc_task *task,
710                                struct nfs4_sequence_res *res)
711 {
712         if (res->sr_slot != NULL)
713                 nfs40_sequence_free_slot(res);
714         return 1;
715 }
716
717 #if defined(CONFIG_NFS_V4_1)
718
719 static void nfs41_release_slot(struct nfs4_slot *slot)
720 {
721         struct nfs4_session *session;
722         struct nfs4_slot_table *tbl;
723         bool send_new_highest_used_slotid = false;
724
725         if (!slot)
726                 return;
727         tbl = slot->table;
728         session = tbl->session;
729
730         /* Bump the slot sequence number */
731         if (slot->seq_done)
732                 slot->seq_nr++;
733         slot->seq_done = 0;
734
735         spin_lock(&tbl->slot_tbl_lock);
736         /* Be nice to the server: try to ensure that the last transmitted
737          * value for highest_user_slotid <= target_highest_slotid
738          */
739         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
740                 send_new_highest_used_slotid = true;
741
742         if (nfs41_wake_and_assign_slot(tbl, slot)) {
743                 send_new_highest_used_slotid = false;
744                 goto out_unlock;
745         }
746         nfs4_free_slot(tbl, slot);
747
748         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
749                 send_new_highest_used_slotid = false;
750 out_unlock:
751         spin_unlock(&tbl->slot_tbl_lock);
752         if (send_new_highest_used_slotid)
753                 nfs41_notify_server(session->clp);
754         if (waitqueue_active(&tbl->slot_waitq))
755                 wake_up_all(&tbl->slot_waitq);
756 }
757
758 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
759 {
760         nfs41_release_slot(res->sr_slot);
761         res->sr_slot = NULL;
762 }
763
764 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
765                 u32 seqnr)
766 {
767         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
768                 slot->seq_nr_highest_sent = seqnr;
769 }
770 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
771                 u32 seqnr)
772 {
773         slot->seq_nr_highest_sent = seqnr;
774         slot->seq_nr_last_acked = seqnr;
775 }
776
777 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
778                                 struct nfs4_slot *slot)
779 {
780         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
781         if (!IS_ERR(task))
782                 rpc_put_task_async(task);
783 }
784
785 static int nfs41_sequence_process(struct rpc_task *task,
786                 struct nfs4_sequence_res *res)
787 {
788         struct nfs4_session *session;
789         struct nfs4_slot *slot = res->sr_slot;
790         struct nfs_client *clp;
791         int status;
792         int ret = 1;
793
794         if (slot == NULL)
795                 goto out_noaction;
796         /* don't increment the sequence number if the task wasn't sent */
797         if (!RPC_WAS_SENT(task) || slot->seq_done)
798                 goto out;
799
800         session = slot->table->session;
801         clp = session->clp;
802
803         trace_nfs4_sequence_done(session, res);
804
805         status = res->sr_status;
806         if (task->tk_status == -NFS4ERR_DEADSESSION)
807                 status = -NFS4ERR_DEADSESSION;
808
809         /* Check the SEQUENCE operation status */
810         switch (status) {
811         case 0:
812                 /* Mark this sequence number as having been acked */
813                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
814                 /* Update the slot's sequence and clientid lease timer */
815                 slot->seq_done = 1;
816                 do_renew_lease(clp, res->sr_timestamp);
817                 /* Check sequence flags */
818                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
819                                 !!slot->privileged);
820                 nfs41_update_target_slotid(slot->table, slot, res);
821                 break;
822         case 1:
823                 /*
824                  * sr_status remains 1 if an RPC level error occurred.
825                  * The server may or may not have processed the sequence
826                  * operation..
827                  */
828                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
829                 slot->seq_done = 1;
830                 goto out;
831         case -NFS4ERR_DELAY:
832                 /* The server detected a resend of the RPC call and
833                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
834                  * of RFC5661.
835                  */
836                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
837                         __func__,
838                         slot->slot_nr,
839                         slot->seq_nr);
840                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
841                 goto out_retry;
842         case -NFS4ERR_RETRY_UNCACHED_REP:
843         case -NFS4ERR_SEQ_FALSE_RETRY:
844                 /*
845                  * The server thinks we tried to replay a request.
846                  * Retry the call after bumping the sequence ID.
847                  */
848                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
849                 goto retry_new_seq;
850         case -NFS4ERR_BADSLOT:
851                 /*
852                  * The slot id we used was probably retired. Try again
853                  * using a different slot id.
854                  */
855                 if (slot->slot_nr < slot->table->target_highest_slotid)
856                         goto session_recover;
857                 goto retry_nowait;
858         case -NFS4ERR_SEQ_MISORDERED:
859                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
860                 /*
861                  * Were one or more calls using this slot interrupted?
862                  * If the server never received the request, then our
863                  * transmitted slot sequence number may be too high. However,
864                  * if the server did receive the request then it might
865                  * accidentally give us a reply with a mismatched operation.
866                  * We can sort this out by sending a lone sequence operation
867                  * to the server on the same slot.
868                  */
869                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
870                         slot->seq_nr--;
871                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
872                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
873                                 res->sr_slot = NULL;
874                         }
875                         goto retry_nowait;
876                 }
877                 /*
878                  * RFC5661:
879                  * A retry might be sent while the original request is
880                  * still in progress on the replier. The replier SHOULD
881                  * deal with the issue by returning NFS4ERR_DELAY as the
882                  * reply to SEQUENCE or CB_SEQUENCE operation, but
883                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
884                  *
885                  * Restart the search after a delay.
886                  */
887                 slot->seq_nr = slot->seq_nr_highest_sent;
888                 goto out_retry;
889         case -NFS4ERR_BADSESSION:
890         case -NFS4ERR_DEADSESSION:
891         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
892                 goto session_recover;
893         default:
894                 /* Just update the slot sequence no. */
895                 slot->seq_done = 1;
896         }
897 out:
898         /* The session may be reset by one of the error handlers. */
899         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
900 out_noaction:
901         return ret;
902 session_recover:
903         nfs4_schedule_session_recovery(session, status);
904         dprintk("%s ERROR: %d Reset session\n", __func__, status);
905         nfs41_sequence_free_slot(res);
906         goto out;
907 retry_new_seq:
908         ++slot->seq_nr;
909 retry_nowait:
910         if (rpc_restart_call_prepare(task)) {
911                 nfs41_sequence_free_slot(res);
912                 task->tk_status = 0;
913                 ret = 0;
914         }
915         goto out;
916 out_retry:
917         if (!rpc_restart_call(task))
918                 goto out;
919         rpc_delay(task, NFS4_POLL_RETRY_MAX);
920         return 0;
921 }
922
923 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
924 {
925         if (!nfs41_sequence_process(task, res))
926                 return 0;
927         if (res->sr_slot != NULL)
928                 nfs41_sequence_free_slot(res);
929         return 1;
930
931 }
932 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
933
934 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
935 {
936         if (res->sr_slot == NULL)
937                 return 1;
938         if (res->sr_slot->table->session != NULL)
939                 return nfs41_sequence_process(task, res);
940         return nfs40_sequence_done(task, res);
941 }
942
943 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
944 {
945         if (res->sr_slot != NULL) {
946                 if (res->sr_slot->table->session != NULL)
947                         nfs41_sequence_free_slot(res);
948                 else
949                         nfs40_sequence_free_slot(res);
950         }
951 }
952
953 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
954 {
955         if (res->sr_slot == NULL)
956                 return 1;
957         if (!res->sr_slot->table->session)
958                 return nfs40_sequence_done(task, res);
959         return nfs41_sequence_done(task, res);
960 }
961 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
962
963 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
964 {
965         struct nfs4_call_sync_data *data = calldata;
966
967         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
968
969         nfs4_setup_sequence(data->seq_server->nfs_client,
970                             data->seq_args, data->seq_res, task);
971 }
972
973 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
974 {
975         struct nfs4_call_sync_data *data = calldata;
976
977         nfs41_sequence_done(task, data->seq_res);
978 }
979
980 static const struct rpc_call_ops nfs41_call_sync_ops = {
981         .rpc_call_prepare = nfs41_call_sync_prepare,
982         .rpc_call_done = nfs41_call_sync_done,
983 };
984
985 #else   /* !CONFIG_NFS_V4_1 */
986
987 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
988 {
989         return nfs40_sequence_done(task, res);
990 }
991
992 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
993 {
994         if (res->sr_slot != NULL)
995                 nfs40_sequence_free_slot(res);
996 }
997
998 int nfs4_sequence_done(struct rpc_task *task,
999                        struct nfs4_sequence_res *res)
1000 {
1001         return nfs40_sequence_done(task, res);
1002 }
1003 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1004
1005 #endif  /* !CONFIG_NFS_V4_1 */
1006
1007 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1008 {
1009         res->sr_timestamp = jiffies;
1010         res->sr_status_flags = 0;
1011         res->sr_status = 1;
1012 }
1013
1014 static
1015 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1016                 struct nfs4_sequence_res *res,
1017                 struct nfs4_slot *slot)
1018 {
1019         if (!slot)
1020                 return;
1021         slot->privileged = args->sa_privileged ? 1 : 0;
1022         args->sa_slot = slot;
1023
1024         res->sr_slot = slot;
1025 }
1026
1027 int nfs4_setup_sequence(struct nfs_client *client,
1028                         struct nfs4_sequence_args *args,
1029                         struct nfs4_sequence_res *res,
1030                         struct rpc_task *task)
1031 {
1032         struct nfs4_session *session = nfs4_get_session(client);
1033         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1034         struct nfs4_slot *slot;
1035
1036         /* slot already allocated? */
1037         if (res->sr_slot != NULL)
1038                 goto out_start;
1039
1040         if (session)
1041                 tbl = &session->fc_slot_table;
1042
1043         spin_lock(&tbl->slot_tbl_lock);
1044         /* The state manager will wait until the slot table is empty */
1045         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1046                 goto out_sleep;
1047
1048         slot = nfs4_alloc_slot(tbl);
1049         if (IS_ERR(slot)) {
1050                 if (slot == ERR_PTR(-ENOMEM))
1051                         goto out_sleep_timeout;
1052                 goto out_sleep;
1053         }
1054         spin_unlock(&tbl->slot_tbl_lock);
1055
1056         nfs4_sequence_attach_slot(args, res, slot);
1057
1058         trace_nfs4_setup_sequence(session, args);
1059 out_start:
1060         nfs41_sequence_res_init(res);
1061         rpc_call_start(task);
1062         return 0;
1063 out_sleep_timeout:
1064         /* Try again in 1/4 second */
1065         if (args->sa_privileged)
1066                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1067                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1068         else
1069                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1070                                 NULL, jiffies + (HZ >> 2));
1071         spin_unlock(&tbl->slot_tbl_lock);
1072         return -EAGAIN;
1073 out_sleep:
1074         if (args->sa_privileged)
1075                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1076                                 RPC_PRIORITY_PRIVILEGED);
1077         else
1078                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1079         spin_unlock(&tbl->slot_tbl_lock);
1080         return -EAGAIN;
1081 }
1082 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1083
1084 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1085 {
1086         struct nfs4_call_sync_data *data = calldata;
1087         nfs4_setup_sequence(data->seq_server->nfs_client,
1088                                 data->seq_args, data->seq_res, task);
1089 }
1090
1091 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1092 {
1093         struct nfs4_call_sync_data *data = calldata;
1094         nfs4_sequence_done(task, data->seq_res);
1095 }
1096
1097 static const struct rpc_call_ops nfs40_call_sync_ops = {
1098         .rpc_call_prepare = nfs40_call_sync_prepare,
1099         .rpc_call_done = nfs40_call_sync_done,
1100 };
1101
1102 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1103 {
1104         int ret;
1105         struct rpc_task *task;
1106
1107         task = rpc_run_task(task_setup);
1108         if (IS_ERR(task))
1109                 return PTR_ERR(task);
1110
1111         ret = task->tk_status;
1112         rpc_put_task(task);
1113         return ret;
1114 }
1115
1116 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1117                              struct nfs_server *server,
1118                              struct rpc_message *msg,
1119                              struct nfs4_sequence_args *args,
1120                              struct nfs4_sequence_res *res,
1121                              unsigned short task_flags)
1122 {
1123         struct nfs_client *clp = server->nfs_client;
1124         struct nfs4_call_sync_data data = {
1125                 .seq_server = server,
1126                 .seq_args = args,
1127                 .seq_res = res,
1128         };
1129         struct rpc_task_setup task_setup = {
1130                 .rpc_client = clnt,
1131                 .rpc_message = msg,
1132                 .callback_ops = clp->cl_mvops->call_sync_ops,
1133                 .callback_data = &data,
1134                 .flags = task_flags,
1135         };
1136
1137         return nfs4_call_sync_custom(&task_setup);
1138 }
1139
1140 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1141                                    struct nfs_server *server,
1142                                    struct rpc_message *msg,
1143                                    struct nfs4_sequence_args *args,
1144                                    struct nfs4_sequence_res *res)
1145 {
1146         return nfs4_do_call_sync(clnt, server, msg, args, res, 0);
1147 }
1148
1149
1150 int nfs4_call_sync(struct rpc_clnt *clnt,
1151                    struct nfs_server *server,
1152                    struct rpc_message *msg,
1153                    struct nfs4_sequence_args *args,
1154                    struct nfs4_sequence_res *res,
1155                    int cache_reply)
1156 {
1157         nfs4_init_sequence(args, res, cache_reply, 0);
1158         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1159 }
1160
1161 static void
1162 nfs4_inc_nlink_locked(struct inode *inode)
1163 {
1164         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1165         inc_nlink(inode);
1166 }
1167
1168 static void
1169 nfs4_dec_nlink_locked(struct inode *inode)
1170 {
1171         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1172         drop_nlink(inode);
1173 }
1174
1175 static void
1176 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1177                 unsigned long timestamp, unsigned long cache_validity)
1178 {
1179         struct nfs_inode *nfsi = NFS_I(dir);
1180
1181         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1182                 | NFS_INO_INVALID_MTIME
1183                 | NFS_INO_INVALID_DATA
1184                 | cache_validity;
1185         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1186                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1187                 nfsi->attrtimeo_timestamp = jiffies;
1188         } else {
1189                 nfs_force_lookup_revalidate(dir);
1190                 if (cinfo->before != inode_peek_iversion_raw(dir))
1191                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1192                                 NFS_INO_INVALID_ACL;
1193         }
1194         inode_set_iversion_raw(dir, cinfo->after);
1195         nfsi->read_cache_jiffies = timestamp;
1196         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1197         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1198         nfs_fscache_invalidate(dir);
1199 }
1200
1201 static void
1202 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1203                 unsigned long timestamp, unsigned long cache_validity)
1204 {
1205         spin_lock(&dir->i_lock);
1206         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1207         spin_unlock(&dir->i_lock);
1208 }
1209
1210 struct nfs4_open_createattrs {
1211         struct nfs4_label *label;
1212         struct iattr *sattr;
1213         const __u32 verf[2];
1214 };
1215
1216 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1217                 int err, struct nfs4_exception *exception)
1218 {
1219         if (err != -EINVAL)
1220                 return false;
1221         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1222                 return false;
1223         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1224         exception->retry = 1;
1225         return true;
1226 }
1227
1228 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1229 {
1230          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1231 }
1232
1233 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1234 {
1235         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1236
1237         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1238 }
1239
1240 static u32
1241 nfs4_map_atomic_open_share(struct nfs_server *server,
1242                 fmode_t fmode, int openflags)
1243 {
1244         u32 res = 0;
1245
1246         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1247         case FMODE_READ:
1248                 res = NFS4_SHARE_ACCESS_READ;
1249                 break;
1250         case FMODE_WRITE:
1251                 res = NFS4_SHARE_ACCESS_WRITE;
1252                 break;
1253         case FMODE_READ|FMODE_WRITE:
1254                 res = NFS4_SHARE_ACCESS_BOTH;
1255         }
1256         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1257                 goto out;
1258         /* Want no delegation if we're using O_DIRECT */
1259         if (openflags & O_DIRECT)
1260                 res |= NFS4_SHARE_WANT_NO_DELEG;
1261 out:
1262         return res;
1263 }
1264
1265 static enum open_claim_type4
1266 nfs4_map_atomic_open_claim(struct nfs_server *server,
1267                 enum open_claim_type4 claim)
1268 {
1269         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1270                 return claim;
1271         switch (claim) {
1272         default:
1273                 return claim;
1274         case NFS4_OPEN_CLAIM_FH:
1275                 return NFS4_OPEN_CLAIM_NULL;
1276         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1277                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1278         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1279                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1280         }
1281 }
1282
1283 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1284 {
1285         p->o_res.f_attr = &p->f_attr;
1286         p->o_res.f_label = p->f_label;
1287         p->o_res.seqid = p->o_arg.seqid;
1288         p->c_res.seqid = p->c_arg.seqid;
1289         p->o_res.server = p->o_arg.server;
1290         p->o_res.access_request = p->o_arg.access;
1291         nfs_fattr_init(&p->f_attr);
1292         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1293 }
1294
1295 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1296                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1297                 const struct nfs4_open_createattrs *c,
1298                 enum open_claim_type4 claim,
1299                 gfp_t gfp_mask)
1300 {
1301         struct dentry *parent = dget_parent(dentry);
1302         struct inode *dir = d_inode(parent);
1303         struct nfs_server *server = NFS_SERVER(dir);
1304         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1305         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1306         struct nfs4_opendata *p;
1307
1308         p = kzalloc(sizeof(*p), gfp_mask);
1309         if (p == NULL)
1310                 goto err;
1311
1312         p->f_label = nfs4_label_alloc(server, gfp_mask);
1313         if (IS_ERR(p->f_label))
1314                 goto err_free_p;
1315
1316         p->a_label = nfs4_label_alloc(server, gfp_mask);
1317         if (IS_ERR(p->a_label))
1318                 goto err_free_f;
1319
1320         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1321         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1322         if (IS_ERR(p->o_arg.seqid))
1323                 goto err_free_label;
1324         nfs_sb_active(dentry->d_sb);
1325         p->dentry = dget(dentry);
1326         p->dir = parent;
1327         p->owner = sp;
1328         atomic_inc(&sp->so_count);
1329         p->o_arg.open_flags = flags;
1330         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1331         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1332         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1333                         fmode, flags);
1334         if (flags & O_CREAT) {
1335                 p->o_arg.umask = current_umask();
1336                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1337                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1338                         p->o_arg.u.attrs = &p->attrs;
1339                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1340
1341                         memcpy(p->o_arg.u.verifier.data, c->verf,
1342                                         sizeof(p->o_arg.u.verifier.data));
1343                 }
1344         }
1345         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1346          * will return permission denied for all bits until close */
1347         if (!(flags & O_EXCL)) {
1348                 /* ask server to check for all possible rights as results
1349                  * are cached */
1350                 switch (p->o_arg.claim) {
1351                 default:
1352                         break;
1353                 case NFS4_OPEN_CLAIM_NULL:
1354                 case NFS4_OPEN_CLAIM_FH:
1355                         p->o_arg.access = NFS4_ACCESS_READ |
1356                                 NFS4_ACCESS_MODIFY |
1357                                 NFS4_ACCESS_EXTEND |
1358                                 NFS4_ACCESS_EXECUTE;
1359                 }
1360         }
1361         p->o_arg.clientid = server->nfs_client->cl_clientid;
1362         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1363         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1364         p->o_arg.name = &dentry->d_name;
1365         p->o_arg.server = server;
1366         p->o_arg.bitmask = nfs4_bitmask(server, label);
1367         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1368         switch (p->o_arg.claim) {
1369         case NFS4_OPEN_CLAIM_NULL:
1370         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1371         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1372                 p->o_arg.fh = NFS_FH(dir);
1373                 break;
1374         case NFS4_OPEN_CLAIM_PREVIOUS:
1375         case NFS4_OPEN_CLAIM_FH:
1376         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1377         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1378                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1379         }
1380         p->c_arg.fh = &p->o_res.fh;
1381         p->c_arg.stateid = &p->o_res.stateid;
1382         p->c_arg.seqid = p->o_arg.seqid;
1383         nfs4_init_opendata_res(p);
1384         kref_init(&p->kref);
1385         return p;
1386
1387 err_free_label:
1388         nfs4_label_free(p->a_label);
1389 err_free_f:
1390         nfs4_label_free(p->f_label);
1391 err_free_p:
1392         kfree(p);
1393 err:
1394         dput(parent);
1395         return NULL;
1396 }
1397
1398 static void nfs4_opendata_free(struct kref *kref)
1399 {
1400         struct nfs4_opendata *p = container_of(kref,
1401                         struct nfs4_opendata, kref);
1402         struct super_block *sb = p->dentry->d_sb;
1403
1404         nfs4_lgopen_release(p->lgp);
1405         nfs_free_seqid(p->o_arg.seqid);
1406         nfs4_sequence_free_slot(&p->o_res.seq_res);
1407         if (p->state != NULL)
1408                 nfs4_put_open_state(p->state);
1409         nfs4_put_state_owner(p->owner);
1410
1411         nfs4_label_free(p->a_label);
1412         nfs4_label_free(p->f_label);
1413
1414         dput(p->dir);
1415         dput(p->dentry);
1416         nfs_sb_deactive(sb);
1417         nfs_fattr_free_names(&p->f_attr);
1418         kfree(p->f_attr.mdsthreshold);
1419         kfree(p);
1420 }
1421
1422 static void nfs4_opendata_put(struct nfs4_opendata *p)
1423 {
1424         if (p != NULL)
1425                 kref_put(&p->kref, nfs4_opendata_free);
1426 }
1427
1428 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1429                 fmode_t fmode)
1430 {
1431         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1432         case FMODE_READ|FMODE_WRITE:
1433                 return state->n_rdwr != 0;
1434         case FMODE_WRITE:
1435                 return state->n_wronly != 0;
1436         case FMODE_READ:
1437                 return state->n_rdonly != 0;
1438         }
1439         WARN_ON_ONCE(1);
1440         return false;
1441 }
1442
1443 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1444                 int open_mode, enum open_claim_type4 claim)
1445 {
1446         int ret = 0;
1447
1448         if (open_mode & (O_EXCL|O_TRUNC))
1449                 goto out;
1450         switch (claim) {
1451         case NFS4_OPEN_CLAIM_NULL:
1452         case NFS4_OPEN_CLAIM_FH:
1453                 goto out;
1454         default:
1455                 break;
1456         }
1457         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1458                 case FMODE_READ:
1459                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1460                                 && state->n_rdonly != 0;
1461                         break;
1462                 case FMODE_WRITE:
1463                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1464                                 && state->n_wronly != 0;
1465                         break;
1466                 case FMODE_READ|FMODE_WRITE:
1467                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1468                                 && state->n_rdwr != 0;
1469         }
1470 out:
1471         return ret;
1472 }
1473
1474 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1475                 enum open_claim_type4 claim)
1476 {
1477         if (delegation == NULL)
1478                 return 0;
1479         if ((delegation->type & fmode) != fmode)
1480                 return 0;
1481         switch (claim) {
1482         case NFS4_OPEN_CLAIM_NULL:
1483         case NFS4_OPEN_CLAIM_FH:
1484                 break;
1485         case NFS4_OPEN_CLAIM_PREVIOUS:
1486                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1487                         break;
1488                 /* Fall through */
1489         default:
1490                 return 0;
1491         }
1492         nfs_mark_delegation_referenced(delegation);
1493         return 1;
1494 }
1495
1496 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1497 {
1498         switch (fmode) {
1499                 case FMODE_WRITE:
1500                         state->n_wronly++;
1501                         break;
1502                 case FMODE_READ:
1503                         state->n_rdonly++;
1504                         break;
1505                 case FMODE_READ|FMODE_WRITE:
1506                         state->n_rdwr++;
1507         }
1508         nfs4_state_set_mode_locked(state, state->state | fmode);
1509 }
1510
1511 #ifdef CONFIG_NFS_V4_1
1512 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1513 {
1514         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1515                 return true;
1516         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1517                 return true;
1518         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1519                 return true;
1520         return false;
1521 }
1522 #endif /* CONFIG_NFS_V4_1 */
1523
1524 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1525 {
1526         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1527                 wake_up_all(&state->waitq);
1528 }
1529
1530 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1531                 const nfs4_stateid *stateid)
1532 {
1533         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1534         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1535
1536         if (stateid_seqid == state_seqid + 1U ||
1537             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1538                 nfs_state_log_update_open_stateid(state);
1539         else
1540                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1541 }
1542
1543 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1544 {
1545         struct nfs_client *clp = state->owner->so_server->nfs_client;
1546         bool need_recover = false;
1547
1548         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1549                 need_recover = true;
1550         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1551                 need_recover = true;
1552         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1553                 need_recover = true;
1554         if (need_recover)
1555                 nfs4_state_mark_reclaim_nograce(clp, state);
1556 }
1557
1558 /*
1559  * Check for whether or not the caller may update the open stateid
1560  * to the value passed in by stateid.
1561  *
1562  * Note: This function relies heavily on the server implementing
1563  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1564  * correctly.
1565  * i.e. The stateid seqids have to be initialised to 1, and
1566  * are then incremented on every state transition.
1567  */
1568 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1569                 const nfs4_stateid *stateid)
1570 {
1571         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1572             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1573                 if (stateid->seqid == cpu_to_be32(1))
1574                         nfs_state_log_update_open_stateid(state);
1575                 else
1576                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1577                 return true;
1578         }
1579
1580         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1581                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1582                 return true;
1583         }
1584         return false;
1585 }
1586
1587 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1588 {
1589         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1590                 return;
1591         if (state->n_wronly)
1592                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1593         if (state->n_rdonly)
1594                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1595         if (state->n_rdwr)
1596                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1597         set_bit(NFS_OPEN_STATE, &state->flags);
1598 }
1599
1600 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1601                 nfs4_stateid *stateid, fmode_t fmode)
1602 {
1603         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1604         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1605         case FMODE_WRITE:
1606                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1607                 break;
1608         case FMODE_READ:
1609                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1610                 break;
1611         case 0:
1612                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1613                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1614                 clear_bit(NFS_OPEN_STATE, &state->flags);
1615         }
1616         if (stateid == NULL)
1617                 return;
1618         /* Handle OPEN+OPEN_DOWNGRADE races */
1619         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1620             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1621                 nfs_resync_open_stateid_locked(state);
1622                 goto out;
1623         }
1624         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1625                 nfs4_stateid_copy(&state->stateid, stateid);
1626         nfs4_stateid_copy(&state->open_stateid, stateid);
1627         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1628 out:
1629         nfs_state_log_update_open_stateid(state);
1630 }
1631
1632 static void nfs_clear_open_stateid(struct nfs4_state *state,
1633         nfs4_stateid *arg_stateid,
1634         nfs4_stateid *stateid, fmode_t fmode)
1635 {
1636         write_seqlock(&state->seqlock);
1637         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1638         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1639                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1640         write_sequnlock(&state->seqlock);
1641         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1642                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1643 }
1644
1645 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1646                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1647         __must_hold(&state->owner->so_lock)
1648         __must_hold(&state->seqlock)
1649         __must_hold(RCU)
1650
1651 {
1652         DEFINE_WAIT(wait);
1653         int status = 0;
1654         for (;;) {
1655
1656                 if (!nfs_need_update_open_stateid(state, stateid))
1657                         return;
1658                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1659                         break;
1660                 if (status)
1661                         break;
1662                 /* Rely on seqids for serialisation with NFSv4.0 */
1663                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1664                         break;
1665
1666                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1667                 /*
1668                  * Ensure we process the state changes in the same order
1669                  * in which the server processed them by delaying the
1670                  * update of the stateid until we are in sequence.
1671                  */
1672                 write_sequnlock(&state->seqlock);
1673                 spin_unlock(&state->owner->so_lock);
1674                 rcu_read_unlock();
1675                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1676                 if (!signal_pending(current)) {
1677                         if (schedule_timeout(5*HZ) == 0)
1678                                 status = -EAGAIN;
1679                         else
1680                                 status = 0;
1681                 } else
1682                         status = -EINTR;
1683                 finish_wait(&state->waitq, &wait);
1684                 rcu_read_lock();
1685                 spin_lock(&state->owner->so_lock);
1686                 write_seqlock(&state->seqlock);
1687         }
1688
1689         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1690             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1691                 nfs4_stateid_copy(freeme, &state->open_stateid);
1692                 nfs_test_and_clear_all_open_stateid(state);
1693         }
1694
1695         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1696                 nfs4_stateid_copy(&state->stateid, stateid);
1697         nfs4_stateid_copy(&state->open_stateid, stateid);
1698         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1699         nfs_state_log_update_open_stateid(state);
1700 }
1701
1702 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1703                 const nfs4_stateid *open_stateid,
1704                 fmode_t fmode,
1705                 nfs4_stateid *freeme)
1706 {
1707         /*
1708          * Protect the call to nfs4_state_set_mode_locked and
1709          * serialise the stateid update
1710          */
1711         write_seqlock(&state->seqlock);
1712         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1713         switch (fmode) {
1714         case FMODE_READ:
1715                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1716                 break;
1717         case FMODE_WRITE:
1718                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1719                 break;
1720         case FMODE_READ|FMODE_WRITE:
1721                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1722         }
1723         set_bit(NFS_OPEN_STATE, &state->flags);
1724         write_sequnlock(&state->seqlock);
1725 }
1726
1727 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1728 {
1729         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1730         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1731         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1732         clear_bit(NFS_OPEN_STATE, &state->flags);
1733 }
1734
1735 static void nfs_state_set_delegation(struct nfs4_state *state,
1736                 const nfs4_stateid *deleg_stateid,
1737                 fmode_t fmode)
1738 {
1739         /*
1740          * Protect the call to nfs4_state_set_mode_locked and
1741          * serialise the stateid update
1742          */
1743         write_seqlock(&state->seqlock);
1744         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1745         set_bit(NFS_DELEGATED_STATE, &state->flags);
1746         write_sequnlock(&state->seqlock);
1747 }
1748
1749 static void nfs_state_clear_delegation(struct nfs4_state *state)
1750 {
1751         write_seqlock(&state->seqlock);
1752         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1753         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1754         write_sequnlock(&state->seqlock);
1755 }
1756
1757 int update_open_stateid(struct nfs4_state *state,
1758                 const nfs4_stateid *open_stateid,
1759                 const nfs4_stateid *delegation,
1760                 fmode_t fmode)
1761 {
1762         struct nfs_server *server = NFS_SERVER(state->inode);
1763         struct nfs_client *clp = server->nfs_client;
1764         struct nfs_inode *nfsi = NFS_I(state->inode);
1765         struct nfs_delegation *deleg_cur;
1766         nfs4_stateid freeme = { };
1767         int ret = 0;
1768
1769         fmode &= (FMODE_READ|FMODE_WRITE);
1770
1771         rcu_read_lock();
1772         spin_lock(&state->owner->so_lock);
1773         if (open_stateid != NULL) {
1774                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1775                 ret = 1;
1776         }
1777
1778         deleg_cur = nfs4_get_valid_delegation(state->inode);
1779         if (deleg_cur == NULL)
1780                 goto no_delegation;
1781
1782         spin_lock(&deleg_cur->lock);
1783         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1784            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1785             (deleg_cur->type & fmode) != fmode)
1786                 goto no_delegation_unlock;
1787
1788         if (delegation == NULL)
1789                 delegation = &deleg_cur->stateid;
1790         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1791                 goto no_delegation_unlock;
1792
1793         nfs_mark_delegation_referenced(deleg_cur);
1794         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1795         ret = 1;
1796 no_delegation_unlock:
1797         spin_unlock(&deleg_cur->lock);
1798 no_delegation:
1799         if (ret)
1800                 update_open_stateflags(state, fmode);
1801         spin_unlock(&state->owner->so_lock);
1802         rcu_read_unlock();
1803
1804         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1805                 nfs4_schedule_state_manager(clp);
1806         if (freeme.type != 0)
1807                 nfs4_test_and_free_stateid(server, &freeme,
1808                                 state->owner->so_cred);
1809
1810         return ret;
1811 }
1812
1813 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1814                 const nfs4_stateid *stateid)
1815 {
1816         struct nfs4_state *state = lsp->ls_state;
1817         bool ret = false;
1818
1819         spin_lock(&state->state_lock);
1820         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1821                 goto out_noupdate;
1822         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1823                 goto out_noupdate;
1824         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1825         ret = true;
1826 out_noupdate:
1827         spin_unlock(&state->state_lock);
1828         return ret;
1829 }
1830
1831 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1832 {
1833         struct nfs_delegation *delegation;
1834
1835         fmode &= FMODE_READ|FMODE_WRITE;
1836         rcu_read_lock();
1837         delegation = nfs4_get_valid_delegation(inode);
1838         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1839                 rcu_read_unlock();
1840                 return;
1841         }
1842         rcu_read_unlock();
1843         nfs4_inode_return_delegation(inode);
1844 }
1845
1846 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1847 {
1848         struct nfs4_state *state = opendata->state;
1849         struct nfs_delegation *delegation;
1850         int open_mode = opendata->o_arg.open_flags;
1851         fmode_t fmode = opendata->o_arg.fmode;
1852         enum open_claim_type4 claim = opendata->o_arg.claim;
1853         nfs4_stateid stateid;
1854         int ret = -EAGAIN;
1855
1856         for (;;) {
1857                 spin_lock(&state->owner->so_lock);
1858                 if (can_open_cached(state, fmode, open_mode, claim)) {
1859                         update_open_stateflags(state, fmode);
1860                         spin_unlock(&state->owner->so_lock);
1861                         goto out_return_state;
1862                 }
1863                 spin_unlock(&state->owner->so_lock);
1864                 rcu_read_lock();
1865                 delegation = nfs4_get_valid_delegation(state->inode);
1866                 if (!can_open_delegated(delegation, fmode, claim)) {
1867                         rcu_read_unlock();
1868                         break;
1869                 }
1870                 /* Save the delegation */
1871                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1872                 rcu_read_unlock();
1873                 nfs_release_seqid(opendata->o_arg.seqid);
1874                 if (!opendata->is_recover) {
1875                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1876                         if (ret != 0)
1877                                 goto out;
1878                 }
1879                 ret = -EAGAIN;
1880
1881                 /* Try to update the stateid using the delegation */
1882                 if (update_open_stateid(state, NULL, &stateid, fmode))
1883                         goto out_return_state;
1884         }
1885 out:
1886         return ERR_PTR(ret);
1887 out_return_state:
1888         refcount_inc(&state->count);
1889         return state;
1890 }
1891
1892 static void
1893 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1894 {
1895         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1896         struct nfs_delegation *delegation;
1897         int delegation_flags = 0;
1898
1899         rcu_read_lock();
1900         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1901         if (delegation)
1902                 delegation_flags = delegation->flags;
1903         rcu_read_unlock();
1904         switch (data->o_arg.claim) {
1905         default:
1906                 break;
1907         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1908         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1909                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1910                                    "returning a delegation for "
1911                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1912                                    clp->cl_hostname);
1913                 return;
1914         }
1915         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1916                 nfs_inode_set_delegation(state->inode,
1917                                 data->owner->so_cred,
1918                                 data->o_res.delegation_type,
1919                                 &data->o_res.delegation,
1920                                 data->o_res.pagemod_limit);
1921         else
1922                 nfs_inode_reclaim_delegation(state->inode,
1923                                 data->owner->so_cred,
1924                                 data->o_res.delegation_type,
1925                                 &data->o_res.delegation,
1926                                 data->o_res.pagemod_limit);
1927
1928         if (data->o_res.do_recall)
1929                 nfs_async_inode_return_delegation(state->inode,
1930                                                   &data->o_res.delegation);
1931 }
1932
1933 /*
1934  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1935  * and update the nfs4_state.
1936  */
1937 static struct nfs4_state *
1938 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1939 {
1940         struct inode *inode = data->state->inode;
1941         struct nfs4_state *state = data->state;
1942         int ret;
1943
1944         if (!data->rpc_done) {
1945                 if (data->rpc_status)
1946                         return ERR_PTR(data->rpc_status);
1947                 /* cached opens have already been processed */
1948                 goto update;
1949         }
1950
1951         ret = nfs_refresh_inode(inode, &data->f_attr);
1952         if (ret)
1953                 return ERR_PTR(ret);
1954
1955         if (data->o_res.delegation_type != 0)
1956                 nfs4_opendata_check_deleg(data, state);
1957 update:
1958         if (!update_open_stateid(state, &data->o_res.stateid,
1959                                 NULL, data->o_arg.fmode))
1960                 return ERR_PTR(-EAGAIN);
1961         refcount_inc(&state->count);
1962
1963         return state;
1964 }
1965
1966 static struct inode *
1967 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1968 {
1969         struct inode *inode;
1970
1971         switch (data->o_arg.claim) {
1972         case NFS4_OPEN_CLAIM_NULL:
1973         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1974         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1975                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1976                         return ERR_PTR(-EAGAIN);
1977                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1978                                 &data->f_attr, data->f_label);
1979                 break;
1980         default:
1981                 inode = d_inode(data->dentry);
1982                 ihold(inode);
1983                 nfs_refresh_inode(inode, &data->f_attr);
1984         }
1985         return inode;
1986 }
1987
1988 static struct nfs4_state *
1989 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1990 {
1991         struct nfs4_state *state;
1992         struct inode *inode;
1993
1994         inode = nfs4_opendata_get_inode(data);
1995         if (IS_ERR(inode))
1996                 return ERR_CAST(inode);
1997         if (data->state != NULL && data->state->inode == inode) {
1998                 state = data->state;
1999                 refcount_inc(&state->count);
2000         } else
2001                 state = nfs4_get_open_state(inode, data->owner);
2002         iput(inode);
2003         if (state == NULL)
2004                 state = ERR_PTR(-ENOMEM);
2005         return state;
2006 }
2007
2008 static struct nfs4_state *
2009 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2010 {
2011         struct nfs4_state *state;
2012
2013         if (!data->rpc_done) {
2014                 state = nfs4_try_open_cached(data);
2015                 trace_nfs4_cached_open(data->state);
2016                 goto out;
2017         }
2018
2019         state = nfs4_opendata_find_nfs4_state(data);
2020         if (IS_ERR(state))
2021                 goto out;
2022
2023         if (data->o_res.delegation_type != 0)
2024                 nfs4_opendata_check_deleg(data, state);
2025         if (!update_open_stateid(state, &data->o_res.stateid,
2026                                 NULL, data->o_arg.fmode)) {
2027                 nfs4_put_open_state(state);
2028                 state = ERR_PTR(-EAGAIN);
2029         }
2030 out:
2031         nfs_release_seqid(data->o_arg.seqid);
2032         return state;
2033 }
2034
2035 static struct nfs4_state *
2036 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2037 {
2038         struct nfs4_state *ret;
2039
2040         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2041                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2042         else
2043                 ret = _nfs4_opendata_to_nfs4_state(data);
2044         nfs4_sequence_free_slot(&data->o_res.seq_res);
2045         return ret;
2046 }
2047
2048 static struct nfs_open_context *
2049 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2050 {
2051         struct nfs_inode *nfsi = NFS_I(state->inode);
2052         struct nfs_open_context *ctx;
2053
2054         rcu_read_lock();
2055         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2056                 if (ctx->state != state)
2057                         continue;
2058                 if ((ctx->mode & mode) != mode)
2059                         continue;
2060                 if (!get_nfs_open_context(ctx))
2061                         continue;
2062                 rcu_read_unlock();
2063                 return ctx;
2064         }
2065         rcu_read_unlock();
2066         return ERR_PTR(-ENOENT);
2067 }
2068
2069 static struct nfs_open_context *
2070 nfs4_state_find_open_context(struct nfs4_state *state)
2071 {
2072         struct nfs_open_context *ctx;
2073
2074         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2075         if (!IS_ERR(ctx))
2076                 return ctx;
2077         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2078         if (!IS_ERR(ctx))
2079                 return ctx;
2080         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2081 }
2082
2083 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2084                 struct nfs4_state *state, enum open_claim_type4 claim)
2085 {
2086         struct nfs4_opendata *opendata;
2087
2088         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2089                         NULL, claim, GFP_NOFS);
2090         if (opendata == NULL)
2091                 return ERR_PTR(-ENOMEM);
2092         opendata->state = state;
2093         refcount_inc(&state->count);
2094         return opendata;
2095 }
2096
2097 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2098                 fmode_t fmode)
2099 {
2100         struct nfs4_state *newstate;
2101         int ret;
2102
2103         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2104                 return 0;
2105         opendata->o_arg.open_flags = 0;
2106         opendata->o_arg.fmode = fmode;
2107         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2108                         NFS_SB(opendata->dentry->d_sb),
2109                         fmode, 0);
2110         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2111         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2112         nfs4_init_opendata_res(opendata);
2113         ret = _nfs4_recover_proc_open(opendata);
2114         if (ret != 0)
2115                 return ret; 
2116         newstate = nfs4_opendata_to_nfs4_state(opendata);
2117         if (IS_ERR(newstate))
2118                 return PTR_ERR(newstate);
2119         if (newstate != opendata->state)
2120                 ret = -ESTALE;
2121         nfs4_close_state(newstate, fmode);
2122         return ret;
2123 }
2124
2125 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2126 {
2127         int ret;
2128
2129         /* memory barrier prior to reading state->n_* */
2130         smp_rmb();
2131         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2132         if (ret != 0)
2133                 return ret;
2134         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2135         if (ret != 0)
2136                 return ret;
2137         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2138         if (ret != 0)
2139                 return ret;
2140         /*
2141          * We may have performed cached opens for all three recoveries.
2142          * Check if we need to update the current stateid.
2143          */
2144         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2145             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2146                 write_seqlock(&state->seqlock);
2147                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2148                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2149                 write_sequnlock(&state->seqlock);
2150         }
2151         return 0;
2152 }
2153
2154 /*
2155  * OPEN_RECLAIM:
2156  *      reclaim state on the server after a reboot.
2157  */
2158 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2159 {
2160         struct nfs_delegation *delegation;
2161         struct nfs4_opendata *opendata;
2162         fmode_t delegation_type = 0;
2163         int status;
2164
2165         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2166                         NFS4_OPEN_CLAIM_PREVIOUS);
2167         if (IS_ERR(opendata))
2168                 return PTR_ERR(opendata);
2169         rcu_read_lock();
2170         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2171         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2172                 delegation_type = delegation->type;
2173         rcu_read_unlock();
2174         opendata->o_arg.u.delegation_type = delegation_type;
2175         status = nfs4_open_recover(opendata, state);
2176         nfs4_opendata_put(opendata);
2177         return status;
2178 }
2179
2180 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2181 {
2182         struct nfs_server *server = NFS_SERVER(state->inode);
2183         struct nfs4_exception exception = { };
2184         int err;
2185         do {
2186                 err = _nfs4_do_open_reclaim(ctx, state);
2187                 trace_nfs4_open_reclaim(ctx, 0, err);
2188                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2189                         continue;
2190                 if (err != -NFS4ERR_DELAY)
2191                         break;
2192                 nfs4_handle_exception(server, err, &exception);
2193         } while (exception.retry);
2194         return err;
2195 }
2196
2197 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2198 {
2199         struct nfs_open_context *ctx;
2200         int ret;
2201
2202         ctx = nfs4_state_find_open_context(state);
2203         if (IS_ERR(ctx))
2204                 return -EAGAIN;
2205         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2206         nfs_state_clear_open_state_flags(state);
2207         ret = nfs4_do_open_reclaim(ctx, state);
2208         put_nfs_open_context(ctx);
2209         return ret;
2210 }
2211
2212 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, struct file_lock *fl, int err)
2213 {
2214         switch (err) {
2215                 default:
2216                         printk(KERN_ERR "NFS: %s: unhandled error "
2217                                         "%d.\n", __func__, err);
2218                 case 0:
2219                 case -ENOENT:
2220                 case -EAGAIN:
2221                 case -ESTALE:
2222                 case -ETIMEDOUT:
2223                         break;
2224                 case -NFS4ERR_BADSESSION:
2225                 case -NFS4ERR_BADSLOT:
2226                 case -NFS4ERR_BAD_HIGH_SLOT:
2227                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2228                 case -NFS4ERR_DEADSESSION:
2229                         return -EAGAIN;
2230                 case -NFS4ERR_STALE_CLIENTID:
2231                 case -NFS4ERR_STALE_STATEID:
2232                         /* Don't recall a delegation if it was lost */
2233                         nfs4_schedule_lease_recovery(server->nfs_client);
2234                         return -EAGAIN;
2235                 case -NFS4ERR_MOVED:
2236                         nfs4_schedule_migration_recovery(server);
2237                         return -EAGAIN;
2238                 case -NFS4ERR_LEASE_MOVED:
2239                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2240                         return -EAGAIN;
2241                 case -NFS4ERR_DELEG_REVOKED:
2242                 case -NFS4ERR_ADMIN_REVOKED:
2243                 case -NFS4ERR_EXPIRED:
2244                 case -NFS4ERR_BAD_STATEID:
2245                 case -NFS4ERR_OPENMODE:
2246                         nfs_inode_find_state_and_recover(state->inode,
2247                                         stateid);
2248                         nfs4_schedule_stateid_recovery(server, state);
2249                         return -EAGAIN;
2250                 case -NFS4ERR_DELAY:
2251                 case -NFS4ERR_GRACE:
2252                         ssleep(1);
2253                         return -EAGAIN;
2254                 case -ENOMEM:
2255                 case -NFS4ERR_DENIED:
2256                         if (fl) {
2257                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2258                                 if (lsp)
2259                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2260                         }
2261                         return 0;
2262         }
2263         return err;
2264 }
2265
2266 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2267                 struct nfs4_state *state, const nfs4_stateid *stateid)
2268 {
2269         struct nfs_server *server = NFS_SERVER(state->inode);
2270         struct nfs4_opendata *opendata;
2271         int err = 0;
2272
2273         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2274                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2275         if (IS_ERR(opendata))
2276                 return PTR_ERR(opendata);
2277         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2278         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2279                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2280                 if (err)
2281                         goto out;
2282         }
2283         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2284                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2285                 if (err)
2286                         goto out;
2287         }
2288         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2289                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2290                 if (err)
2291                         goto out;
2292         }
2293         nfs_state_clear_delegation(state);
2294 out:
2295         nfs4_opendata_put(opendata);
2296         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2297 }
2298
2299 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2300 {
2301         struct nfs4_opendata *data = calldata;
2302
2303         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2304                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2305 }
2306
2307 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2308 {
2309         struct nfs4_opendata *data = calldata;
2310
2311         nfs40_sequence_done(task, &data->c_res.seq_res);
2312
2313         data->rpc_status = task->tk_status;
2314         if (data->rpc_status == 0) {
2315                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2316                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2317                 renew_lease(data->o_res.server, data->timestamp);
2318                 data->rpc_done = true;
2319         }
2320 }
2321
2322 static void nfs4_open_confirm_release(void *calldata)
2323 {
2324         struct nfs4_opendata *data = calldata;
2325         struct nfs4_state *state = NULL;
2326
2327         /* If this request hasn't been cancelled, do nothing */
2328         if (!data->cancelled)
2329                 goto out_free;
2330         /* In case of error, no cleanup! */
2331         if (!data->rpc_done)
2332                 goto out_free;
2333         state = nfs4_opendata_to_nfs4_state(data);
2334         if (!IS_ERR(state))
2335                 nfs4_close_state(state, data->o_arg.fmode);
2336 out_free:
2337         nfs4_opendata_put(data);
2338 }
2339
2340 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2341         .rpc_call_prepare = nfs4_open_confirm_prepare,
2342         .rpc_call_done = nfs4_open_confirm_done,
2343         .rpc_release = nfs4_open_confirm_release,
2344 };
2345
2346 /*
2347  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2348  */
2349 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2350 {
2351         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2352         struct rpc_task *task;
2353         struct  rpc_message msg = {
2354                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2355                 .rpc_argp = &data->c_arg,
2356                 .rpc_resp = &data->c_res,
2357                 .rpc_cred = data->owner->so_cred,
2358         };
2359         struct rpc_task_setup task_setup_data = {
2360                 .rpc_client = server->client,
2361                 .rpc_message = &msg,
2362                 .callback_ops = &nfs4_open_confirm_ops,
2363                 .callback_data = data,
2364                 .workqueue = nfsiod_workqueue,
2365                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2366         };
2367         int status;
2368
2369         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2370                                 data->is_recover);
2371         kref_get(&data->kref);
2372         data->rpc_done = false;
2373         data->rpc_status = 0;
2374         data->timestamp = jiffies;
2375         task = rpc_run_task(&task_setup_data);
2376         if (IS_ERR(task))
2377                 return PTR_ERR(task);
2378         status = rpc_wait_for_completion_task(task);
2379         if (status != 0) {
2380                 data->cancelled = true;
2381                 smp_wmb();
2382         } else
2383                 status = data->rpc_status;
2384         rpc_put_task(task);
2385         return status;
2386 }
2387
2388 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2389 {
2390         struct nfs4_opendata *data = calldata;
2391         struct nfs4_state_owner *sp = data->owner;
2392         struct nfs_client *clp = sp->so_server->nfs_client;
2393         enum open_claim_type4 claim = data->o_arg.claim;
2394
2395         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2396                 goto out_wait;
2397         /*
2398          * Check if we still need to send an OPEN call, or if we can use
2399          * a delegation instead.
2400          */
2401         if (data->state != NULL) {
2402                 struct nfs_delegation *delegation;
2403
2404                 if (can_open_cached(data->state, data->o_arg.fmode,
2405                                         data->o_arg.open_flags, claim))
2406                         goto out_no_action;
2407                 rcu_read_lock();
2408                 delegation = nfs4_get_valid_delegation(data->state->inode);
2409                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2410                         goto unlock_no_action;
2411                 rcu_read_unlock();
2412         }
2413         /* Update client id. */
2414         data->o_arg.clientid = clp->cl_clientid;
2415         switch (claim) {
2416         default:
2417                 break;
2418         case NFS4_OPEN_CLAIM_PREVIOUS:
2419         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2420         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2421                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2422                 /* Fall through */
2423         case NFS4_OPEN_CLAIM_FH:
2424                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2425         }
2426         data->timestamp = jiffies;
2427         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2428                                 &data->o_arg.seq_args,
2429                                 &data->o_res.seq_res,
2430                                 task) != 0)
2431                 nfs_release_seqid(data->o_arg.seqid);
2432
2433         /* Set the create mode (note dependency on the session type) */
2434         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2435         if (data->o_arg.open_flags & O_EXCL) {
2436                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2437                 if (nfs4_has_persistent_session(clp))
2438                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2439                 else if (clp->cl_mvops->minor_version > 0)
2440                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2441         }
2442         return;
2443 unlock_no_action:
2444         trace_nfs4_cached_open(data->state);
2445         rcu_read_unlock();
2446 out_no_action:
2447         task->tk_action = NULL;
2448 out_wait:
2449         nfs4_sequence_done(task, &data->o_res.seq_res);
2450 }
2451
2452 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2453 {
2454         struct nfs4_opendata *data = calldata;
2455
2456         data->rpc_status = task->tk_status;
2457
2458         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2459                 return;
2460
2461         if (task->tk_status == 0) {
2462                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2463                         switch (data->o_res.f_attr->mode & S_IFMT) {
2464                         case S_IFREG:
2465                                 break;
2466                         case S_IFLNK:
2467                                 data->rpc_status = -ELOOP;
2468                                 break;
2469                         case S_IFDIR:
2470                                 data->rpc_status = -EISDIR;
2471                                 break;
2472                         default:
2473                                 data->rpc_status = -ENOTDIR;
2474                         }
2475                 }
2476                 renew_lease(data->o_res.server, data->timestamp);
2477                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2478                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2479         }
2480         data->rpc_done = true;
2481 }
2482
2483 static void nfs4_open_release(void *calldata)
2484 {
2485         struct nfs4_opendata *data = calldata;
2486         struct nfs4_state *state = NULL;
2487
2488         /* If this request hasn't been cancelled, do nothing */
2489         if (!data->cancelled)
2490                 goto out_free;
2491         /* In case of error, no cleanup! */
2492         if (data->rpc_status != 0 || !data->rpc_done)
2493                 goto out_free;
2494         /* In case we need an open_confirm, no cleanup! */
2495         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2496                 goto out_free;
2497         state = nfs4_opendata_to_nfs4_state(data);
2498         if (!IS_ERR(state))
2499                 nfs4_close_state(state, data->o_arg.fmode);
2500 out_free:
2501         nfs4_opendata_put(data);
2502 }
2503
2504 static const struct rpc_call_ops nfs4_open_ops = {
2505         .rpc_call_prepare = nfs4_open_prepare,
2506         .rpc_call_done = nfs4_open_done,
2507         .rpc_release = nfs4_open_release,
2508 };
2509
2510 static int nfs4_run_open_task(struct nfs4_opendata *data,
2511                               struct nfs_open_context *ctx)
2512 {
2513         struct inode *dir = d_inode(data->dir);
2514         struct nfs_server *server = NFS_SERVER(dir);
2515         struct nfs_openargs *o_arg = &data->o_arg;
2516         struct nfs_openres *o_res = &data->o_res;
2517         struct rpc_task *task;
2518         struct rpc_message msg = {
2519                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2520                 .rpc_argp = o_arg,
2521                 .rpc_resp = o_res,
2522                 .rpc_cred = data->owner->so_cred,
2523         };
2524         struct rpc_task_setup task_setup_data = {
2525                 .rpc_client = server->client,
2526                 .rpc_message = &msg,
2527                 .callback_ops = &nfs4_open_ops,
2528                 .callback_data = data,
2529                 .workqueue = nfsiod_workqueue,
2530                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2531         };
2532         int status;
2533
2534         kref_get(&data->kref);
2535         data->rpc_done = false;
2536         data->rpc_status = 0;
2537         data->cancelled = false;
2538         data->is_recover = false;
2539         if (!ctx) {
2540                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2541                 data->is_recover = true;
2542                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2543         } else {
2544                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2545                 pnfs_lgopen_prepare(data, ctx);
2546         }
2547         task = rpc_run_task(&task_setup_data);
2548         if (IS_ERR(task))
2549                 return PTR_ERR(task);
2550         status = rpc_wait_for_completion_task(task);
2551         if (status != 0) {
2552                 data->cancelled = true;
2553                 smp_wmb();
2554         } else
2555                 status = data->rpc_status;
2556         rpc_put_task(task);
2557
2558         return status;
2559 }
2560
2561 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2562 {
2563         struct inode *dir = d_inode(data->dir);
2564         struct nfs_openres *o_res = &data->o_res;
2565         int status;
2566
2567         status = nfs4_run_open_task(data, NULL);
2568         if (status != 0 || !data->rpc_done)
2569                 return status;
2570
2571         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2572
2573         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2574                 status = _nfs4_proc_open_confirm(data);
2575
2576         return status;
2577 }
2578
2579 /*
2580  * Additional permission checks in order to distinguish between an
2581  * open for read, and an open for execute. This works around the
2582  * fact that NFSv4 OPEN treats read and execute permissions as being
2583  * the same.
2584  * Note that in the non-execute case, we want to turn off permission
2585  * checking if we just created a new file (POSIX open() semantics).
2586  */
2587 static int nfs4_opendata_access(const struct cred *cred,
2588                                 struct nfs4_opendata *opendata,
2589                                 struct nfs4_state *state, fmode_t fmode,
2590                                 int openflags)
2591 {
2592         struct nfs_access_entry cache;
2593         u32 mask, flags;
2594
2595         /* access call failed or for some reason the server doesn't
2596          * support any access modes -- defer access call until later */
2597         if (opendata->o_res.access_supported == 0)
2598                 return 0;
2599
2600         mask = 0;
2601         /*
2602          * Use openflags to check for exec, because fmode won't
2603          * always have FMODE_EXEC set when file open for exec.
2604          */
2605         if (openflags & __FMODE_EXEC) {
2606                 /* ONLY check for exec rights */
2607                 if (S_ISDIR(state->inode->i_mode))
2608                         mask = NFS4_ACCESS_LOOKUP;
2609                 else
2610                         mask = NFS4_ACCESS_EXECUTE;
2611         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2612                 mask = NFS4_ACCESS_READ;
2613
2614         cache.cred = cred;
2615         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2616         nfs_access_add_cache(state->inode, &cache);
2617
2618         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2619         if ((mask & ~cache.mask & flags) == 0)
2620                 return 0;
2621
2622         return -EACCES;
2623 }
2624
2625 /*
2626  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2627  */
2628 static int _nfs4_proc_open(struct nfs4_opendata *data,
2629                            struct nfs_open_context *ctx)
2630 {
2631         struct inode *dir = d_inode(data->dir);
2632         struct nfs_server *server = NFS_SERVER(dir);
2633         struct nfs_openargs *o_arg = &data->o_arg;
2634         struct nfs_openres *o_res = &data->o_res;
2635         int status;
2636
2637         status = nfs4_run_open_task(data, ctx);
2638         if (!data->rpc_done)
2639                 return status;
2640         if (status != 0) {
2641                 if (status == -NFS4ERR_BADNAME &&
2642                                 !(o_arg->open_flags & O_CREAT))
2643                         return -ENOENT;
2644                 return status;
2645         }
2646
2647         nfs_fattr_map_and_free_names(server, &data->f_attr);
2648
2649         if (o_arg->open_flags & O_CREAT) {
2650                 if (o_arg->open_flags & O_EXCL)
2651                         data->file_created = true;
2652                 else if (o_res->cinfo.before != o_res->cinfo.after)
2653                         data->file_created = true;
2654                 if (data->file_created ||
2655                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2656                         update_changeattr(dir, &o_res->cinfo,
2657                                         o_res->f_attr->time_start, 0);
2658         }
2659         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2660                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2661         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2662                 status = _nfs4_proc_open_confirm(data);
2663                 if (status != 0)
2664                         return status;
2665         }
2666         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2667                 nfs4_sequence_free_slot(&o_res->seq_res);
2668                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2669                                 o_res->f_label, NULL);
2670         }
2671         return 0;
2672 }
2673
2674 /*
2675  * OPEN_EXPIRED:
2676  *      reclaim state on the server after a network partition.
2677  *      Assumes caller holds the appropriate lock
2678  */
2679 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2680 {
2681         struct nfs4_opendata *opendata;
2682         int ret;
2683
2684         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2685                         NFS4_OPEN_CLAIM_FH);
2686         if (IS_ERR(opendata))
2687                 return PTR_ERR(opendata);
2688         ret = nfs4_open_recover(opendata, state);
2689         if (ret == -ESTALE)
2690                 d_drop(ctx->dentry);
2691         nfs4_opendata_put(opendata);
2692         return ret;
2693 }
2694
2695 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2696 {
2697         struct nfs_server *server = NFS_SERVER(state->inode);
2698         struct nfs4_exception exception = { };
2699         int err;
2700
2701         do {
2702                 err = _nfs4_open_expired(ctx, state);
2703                 trace_nfs4_open_expired(ctx, 0, err);
2704                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2705                         continue;
2706                 switch (err) {
2707                 default:
2708                         goto out;
2709                 case -NFS4ERR_GRACE:
2710                 case -NFS4ERR_DELAY:
2711                         nfs4_handle_exception(server, err, &exception);
2712                         err = 0;
2713                 }
2714         } while (exception.retry);
2715 out:
2716         return err;
2717 }
2718
2719 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2720 {
2721         struct nfs_open_context *ctx;
2722         int ret;
2723
2724         ctx = nfs4_state_find_open_context(state);
2725         if (IS_ERR(ctx))
2726                 return -EAGAIN;
2727         ret = nfs4_do_open_expired(ctx, state);
2728         put_nfs_open_context(ctx);
2729         return ret;
2730 }
2731
2732 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2733                 const nfs4_stateid *stateid)
2734 {
2735         nfs_remove_bad_delegation(state->inode, stateid);
2736         nfs_state_clear_delegation(state);
2737 }
2738
2739 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2740 {
2741         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2742                 nfs_finish_clear_delegation_stateid(state, NULL);
2743 }
2744
2745 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2746 {
2747         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2748         nfs40_clear_delegation_stateid(state);
2749         nfs_state_clear_open_state_flags(state);
2750         return nfs4_open_expired(sp, state);
2751 }
2752
2753 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2754                 nfs4_stateid *stateid,
2755                 const struct cred *cred)
2756 {
2757         return -NFS4ERR_BAD_STATEID;
2758 }
2759
2760 #if defined(CONFIG_NFS_V4_1)
2761 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2762                 nfs4_stateid *stateid,
2763                 const struct cred *cred)
2764 {
2765         int status;
2766
2767         switch (stateid->type) {
2768         default:
2769                 break;
2770         case NFS4_INVALID_STATEID_TYPE:
2771         case NFS4_SPECIAL_STATEID_TYPE:
2772                 return -NFS4ERR_BAD_STATEID;
2773         case NFS4_REVOKED_STATEID_TYPE:
2774                 goto out_free;
2775         }
2776
2777         status = nfs41_test_stateid(server, stateid, cred);
2778         switch (status) {
2779         case -NFS4ERR_EXPIRED:
2780         case -NFS4ERR_ADMIN_REVOKED:
2781         case -NFS4ERR_DELEG_REVOKED:
2782                 break;
2783         default:
2784                 return status;
2785         }
2786 out_free:
2787         /* Ack the revoked state to the server */
2788         nfs41_free_stateid(server, stateid, cred, true);
2789         return -NFS4ERR_EXPIRED;
2790 }
2791
2792 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2793 {
2794         struct nfs_server *server = NFS_SERVER(state->inode);
2795         nfs4_stateid stateid;
2796         struct nfs_delegation *delegation;
2797         const struct cred *cred = NULL;
2798         int status, ret = NFS_OK;
2799
2800         /* Get the delegation credential for use by test/free_stateid */
2801         rcu_read_lock();
2802         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2803         if (delegation == NULL) {
2804                 rcu_read_unlock();
2805                 nfs_state_clear_delegation(state);
2806                 return NFS_OK;
2807         }
2808
2809         spin_lock(&delegation->lock);
2810         nfs4_stateid_copy(&stateid, &delegation->stateid);
2811
2812         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2813                                 &delegation->flags)) {
2814                 spin_unlock(&delegation->lock);
2815                 rcu_read_unlock();
2816                 return NFS_OK;
2817         }
2818
2819         if (delegation->cred)
2820                 cred = get_cred(delegation->cred);
2821         spin_unlock(&delegation->lock);
2822         rcu_read_unlock();
2823         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2824         trace_nfs4_test_delegation_stateid(state, NULL, status);
2825         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2826                 nfs_finish_clear_delegation_stateid(state, &stateid);
2827         else
2828                 ret = status;
2829
2830         put_cred(cred);
2831         return ret;
2832 }
2833
2834 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2835 {
2836         nfs4_stateid tmp;
2837
2838         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2839             nfs4_copy_delegation_stateid(state->inode, state->state,
2840                                 &tmp, NULL) &&
2841             nfs4_stateid_match_other(&state->stateid, &tmp))
2842                 nfs_state_set_delegation(state, &tmp, state->state);
2843         else
2844                 nfs_state_clear_delegation(state);
2845 }
2846
2847 /**
2848  * nfs41_check_expired_locks - possibly free a lock stateid
2849  *
2850  * @state: NFSv4 state for an inode
2851  *
2852  * Returns NFS_OK if recovery for this stateid is now finished.
2853  * Otherwise a negative NFS4ERR value is returned.
2854  */
2855 static int nfs41_check_expired_locks(struct nfs4_state *state)
2856 {
2857         int status, ret = NFS_OK;
2858         struct nfs4_lock_state *lsp, *prev = NULL;
2859         struct nfs_server *server = NFS_SERVER(state->inode);
2860
2861         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2862                 goto out;
2863
2864         spin_lock(&state->state_lock);
2865         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2866                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2867                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2868
2869                         refcount_inc(&lsp->ls_count);
2870                         spin_unlock(&state->state_lock);
2871
2872                         nfs4_put_lock_state(prev);
2873                         prev = lsp;
2874
2875                         status = nfs41_test_and_free_expired_stateid(server,
2876                                         &lsp->ls_stateid,
2877                                         cred);
2878                         trace_nfs4_test_lock_stateid(state, lsp, status);
2879                         if (status == -NFS4ERR_EXPIRED ||
2880                             status == -NFS4ERR_BAD_STATEID) {
2881                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2882                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2883                                 if (!recover_lost_locks)
2884                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2885                         } else if (status != NFS_OK) {
2886                                 ret = status;
2887                                 nfs4_put_lock_state(prev);
2888                                 goto out;
2889                         }
2890                         spin_lock(&state->state_lock);
2891                 }
2892         }
2893         spin_unlock(&state->state_lock);
2894         nfs4_put_lock_state(prev);
2895 out:
2896         return ret;
2897 }
2898
2899 /**
2900  * nfs41_check_open_stateid - possibly free an open stateid
2901  *
2902  * @state: NFSv4 state for an inode
2903  *
2904  * Returns NFS_OK if recovery for this stateid is now finished.
2905  * Otherwise a negative NFS4ERR value is returned.
2906  */
2907 static int nfs41_check_open_stateid(struct nfs4_state *state)
2908 {
2909         struct nfs_server *server = NFS_SERVER(state->inode);
2910         nfs4_stateid *stateid = &state->open_stateid;
2911         const struct cred *cred = state->owner->so_cred;
2912         int status;
2913
2914         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2915                 return -NFS4ERR_BAD_STATEID;
2916         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2917         trace_nfs4_test_open_stateid(state, NULL, status);
2918         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2919                 nfs_state_clear_open_state_flags(state);
2920                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2921                 return status;
2922         }
2923         if (nfs_open_stateid_recover_openmode(state))
2924                 return -NFS4ERR_OPENMODE;
2925         return NFS_OK;
2926 }
2927
2928 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2929 {
2930         int status;
2931
2932         status = nfs41_check_delegation_stateid(state);
2933         if (status != NFS_OK)
2934                 return status;
2935         nfs41_delegation_recover_stateid(state);
2936
2937         status = nfs41_check_expired_locks(state);
2938         if (status != NFS_OK)
2939                 return status;
2940         status = nfs41_check_open_stateid(state);
2941         if (status != NFS_OK)
2942                 status = nfs4_open_expired(sp, state);
2943         return status;
2944 }
2945 #endif
2946
2947 /*
2948  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2949  * fields corresponding to attributes that were used to store the verifier.
2950  * Make sure we clobber those fields in the later setattr call
2951  */
2952 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2953                                 struct iattr *sattr, struct nfs4_label **label)
2954 {
2955         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2956         __u32 attrset[3];
2957         unsigned ret;
2958         unsigned i;
2959
2960         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2961                 attrset[i] = opendata->o_res.attrset[i];
2962                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2963                         attrset[i] &= ~bitmask[i];
2964         }
2965
2966         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2967                 sattr->ia_valid : 0;
2968
2969         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2970                 if (sattr->ia_valid & ATTR_ATIME_SET)
2971                         ret |= ATTR_ATIME_SET;
2972                 else
2973                         ret |= ATTR_ATIME;
2974         }
2975
2976         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2977                 if (sattr->ia_valid & ATTR_MTIME_SET)
2978                         ret |= ATTR_MTIME_SET;
2979                 else
2980                         ret |= ATTR_MTIME;
2981         }
2982
2983         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2984                 *label = NULL;
2985         return ret;
2986 }
2987
2988 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2989                 int flags, struct nfs_open_context *ctx)
2990 {
2991         struct nfs4_state_owner *sp = opendata->owner;
2992         struct nfs_server *server = sp->so_server;
2993         struct dentry *dentry;
2994         struct nfs4_state *state;
2995         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
2996         struct inode *dir = d_inode(opendata->dir);
2997         unsigned long dir_verifier;
2998         unsigned int seq;
2999         int ret;
3000
3001         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3002         dir_verifier = nfs_save_change_attribute(dir);
3003
3004         ret = _nfs4_proc_open(opendata, ctx);
3005         if (ret != 0)
3006                 goto out;
3007
3008         state = _nfs4_opendata_to_nfs4_state(opendata);
3009         ret = PTR_ERR(state);
3010         if (IS_ERR(state))
3011                 goto out;
3012         ctx->state = state;
3013         if (server->caps & NFS_CAP_POSIX_LOCK)
3014                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3015         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3016                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3017
3018         dentry = opendata->dentry;
3019         if (d_really_is_negative(dentry)) {
3020                 struct dentry *alias;
3021                 d_drop(dentry);
3022                 alias = d_exact_alias(dentry, state->inode);
3023                 if (!alias)
3024                         alias = d_splice_alias(igrab(state->inode), dentry);
3025                 /* d_splice_alias() can't fail here - it's a non-directory */
3026                 if (alias) {
3027                         dput(ctx->dentry);
3028                         ctx->dentry = dentry = alias;
3029                 }
3030         }
3031
3032         switch(opendata->o_arg.claim) {
3033         default:
3034                 break;
3035         case NFS4_OPEN_CLAIM_NULL:
3036         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3037         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3038                 if (!opendata->rpc_done)
3039                         break;
3040                 if (opendata->o_res.delegation_type != 0)
3041                         dir_verifier = nfs_save_change_attribute(dir);
3042                 nfs_set_verifier(dentry, dir_verifier);
3043         }
3044
3045         /* Parse layoutget results before we check for access */
3046         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3047
3048         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3049                         acc_mode, flags);
3050         if (ret != 0)
3051                 goto out;
3052
3053         if (d_inode(dentry) == state->inode) {
3054                 nfs_inode_attach_open_context(ctx);
3055                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3056                         nfs4_schedule_stateid_recovery(server, state);
3057         }
3058
3059 out:
3060         if (!opendata->cancelled)
3061                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3062         return ret;
3063 }
3064
3065 /*
3066  * Returns a referenced nfs4_state
3067  */
3068 static int _nfs4_do_open(struct inode *dir,
3069                         struct nfs_open_context *ctx,
3070                         int flags,
3071                         const struct nfs4_open_createattrs *c,
3072                         int *opened)
3073 {
3074         struct nfs4_state_owner  *sp;
3075         struct nfs4_state     *state = NULL;
3076         struct nfs_server       *server = NFS_SERVER(dir);
3077         struct nfs4_opendata *opendata;
3078         struct dentry *dentry = ctx->dentry;
3079         const struct cred *cred = ctx->cred;
3080         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3081         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3082         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3083         struct iattr *sattr = c->sattr;
3084         struct nfs4_label *label = c->label;
3085         struct nfs4_label *olabel = NULL;
3086         int status;
3087
3088         /* Protect against reboot recovery conflicts */
3089         status = -ENOMEM;
3090         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3091         if (sp == NULL) {
3092                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3093                 goto out_err;
3094         }
3095         status = nfs4_client_recover_expired_lease(server->nfs_client);
3096         if (status != 0)
3097                 goto err_put_state_owner;
3098         if (d_really_is_positive(dentry))
3099                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3100         status = -ENOMEM;
3101         if (d_really_is_positive(dentry))
3102                 claim = NFS4_OPEN_CLAIM_FH;
3103         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3104                         c, claim, GFP_KERNEL);
3105         if (opendata == NULL)
3106                 goto err_put_state_owner;
3107
3108         if (label) {
3109                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3110                 if (IS_ERR(olabel)) {
3111                         status = PTR_ERR(olabel);
3112                         goto err_opendata_put;
3113                 }
3114         }
3115
3116         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3117                 if (!opendata->f_attr.mdsthreshold) {
3118                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3119                         if (!opendata->f_attr.mdsthreshold)
3120                                 goto err_free_label;
3121                 }
3122                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3123         }
3124         if (d_really_is_positive(dentry))
3125                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3126
3127         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3128         if (status != 0)
3129                 goto err_free_label;
3130         state = ctx->state;
3131
3132         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3133             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3134                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3135                 /*
3136                  * send create attributes which was not set by open
3137                  * with an extra setattr.
3138                  */
3139                 if (attrs || label) {
3140                         unsigned ia_old = sattr->ia_valid;
3141
3142                         sattr->ia_valid = attrs;
3143                         nfs_fattr_init(opendata->o_res.f_attr);
3144                         status = nfs4_do_setattr(state->inode, cred,
3145                                         opendata->o_res.f_attr, sattr,
3146                                         ctx, label, olabel);
3147                         if (status == 0) {
3148                                 nfs_setattr_update_inode(state->inode, sattr,
3149                                                 opendata->o_res.f_attr);
3150                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3151                         }
3152                         sattr->ia_valid = ia_old;
3153                 }
3154         }
3155         if (opened && opendata->file_created)
3156                 *opened = 1;
3157
3158         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3159                 *ctx_th = opendata->f_attr.mdsthreshold;
3160                 opendata->f_attr.mdsthreshold = NULL;
3161         }
3162
3163         nfs4_label_free(olabel);
3164
3165         nfs4_opendata_put(opendata);
3166         nfs4_put_state_owner(sp);
3167         return 0;
3168 err_free_label:
3169         nfs4_label_free(olabel);
3170 err_opendata_put:
3171         nfs4_opendata_put(opendata);
3172 err_put_state_owner:
3173         nfs4_put_state_owner(sp);
3174 out_err:
3175         return status;
3176 }
3177
3178
3179 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3180                                         struct nfs_open_context *ctx,
3181                                         int flags,
3182                                         struct iattr *sattr,
3183                                         struct nfs4_label *label,
3184                                         int *opened)
3185 {
3186         struct nfs_server *server = NFS_SERVER(dir);
3187         struct nfs4_exception exception = {
3188                 .interruptible = true,
3189         };
3190         struct nfs4_state *res;
3191         struct nfs4_open_createattrs c = {
3192                 .label = label,
3193                 .sattr = sattr,
3194                 .verf = {
3195                         [0] = (__u32)jiffies,
3196                         [1] = (__u32)current->pid,
3197                 },
3198         };
3199         int status;
3200
3201         do {
3202                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3203                 res = ctx->state;
3204                 trace_nfs4_open_file(ctx, flags, status);
3205                 if (status == 0)
3206                         break;
3207                 /* NOTE: BAD_SEQID means the server and client disagree about the
3208                  * book-keeping w.r.t. state-changing operations
3209                  * (OPEN/CLOSE/LOCK/LOCKU...)
3210                  * It is actually a sign of a bug on the client or on the server.
3211                  *
3212                  * If we receive a BAD_SEQID error in the particular case of
3213                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3214                  * have unhashed the old state_owner for us, and that we can
3215                  * therefore safely retry using a new one. We should still warn
3216                  * the user though...
3217                  */
3218                 if (status == -NFS4ERR_BAD_SEQID) {
3219                         pr_warn_ratelimited("NFS: v4 server %s "
3220                                         " returned a bad sequence-id error!\n",
3221                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3222                         exception.retry = 1;
3223                         continue;
3224                 }
3225                 /*
3226                  * BAD_STATEID on OPEN means that the server cancelled our
3227                  * state before it received the OPEN_CONFIRM.
3228                  * Recover by retrying the request as per the discussion
3229                  * on Page 181 of RFC3530.
3230                  */
3231                 if (status == -NFS4ERR_BAD_STATEID) {
3232                         exception.retry = 1;
3233                         continue;
3234                 }
3235                 if (status == -NFS4ERR_EXPIRED) {
3236                         nfs4_schedule_lease_recovery(server->nfs_client);
3237                         exception.retry = 1;
3238                         continue;
3239                 }
3240                 if (status == -EAGAIN) {
3241                         /* We must have found a delegation */
3242                         exception.retry = 1;
3243                         continue;
3244                 }
3245                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3246                         continue;
3247                 res = ERR_PTR(nfs4_handle_exception(server,
3248                                         status, &exception));
3249         } while (exception.retry);
3250         return res;
3251 }
3252
3253 static int _nfs4_do_setattr(struct inode *inode,
3254                             struct nfs_setattrargs *arg,
3255                             struct nfs_setattrres *res,
3256                             const struct cred *cred,
3257                             struct nfs_open_context *ctx)
3258 {
3259         struct nfs_server *server = NFS_SERVER(inode);
3260         struct rpc_message msg = {
3261                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3262                 .rpc_argp       = arg,
3263                 .rpc_resp       = res,
3264                 .rpc_cred       = cred,
3265         };
3266         const struct cred *delegation_cred = NULL;
3267         unsigned long timestamp = jiffies;
3268         bool truncate;
3269         int status;
3270
3271         nfs_fattr_init(res->fattr);
3272
3273         /* Servers should only apply open mode checks for file size changes */
3274         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3275         if (!truncate)
3276                 goto zero_stateid;
3277
3278         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3279                 /* Use that stateid */
3280         } else if (ctx != NULL && ctx->state) {
3281                 struct nfs_lock_context *l_ctx;
3282                 if (!nfs4_valid_open_stateid(ctx->state))
3283                         return -EBADF;
3284                 l_ctx = nfs_get_lock_context(ctx);
3285                 if (IS_ERR(l_ctx))
3286                         return PTR_ERR(l_ctx);
3287                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3288                                                 &arg->stateid, &delegation_cred);
3289                 nfs_put_lock_context(l_ctx);
3290                 if (status == -EIO)
3291                         return -EBADF;
3292                 else if (status == -EAGAIN)
3293                         goto zero_stateid;
3294         } else {
3295 zero_stateid:
3296                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3297         }
3298         if (delegation_cred)
3299                 msg.rpc_cred = delegation_cred;
3300
3301         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3302
3303         put_cred(delegation_cred);
3304         if (status == 0 && ctx != NULL)
3305                 renew_lease(server, timestamp);
3306         trace_nfs4_setattr(inode, &arg->stateid, status);
3307         return status;
3308 }
3309
3310 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3311                            struct nfs_fattr *fattr, struct iattr *sattr,
3312                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3313                            struct nfs4_label *olabel)
3314 {
3315         struct nfs_server *server = NFS_SERVER(inode);
3316         __u32 bitmask[NFS4_BITMASK_SZ];
3317         struct nfs4_state *state = ctx ? ctx->state : NULL;
3318         struct nfs_setattrargs  arg = {
3319                 .fh             = NFS_FH(inode),
3320                 .iap            = sattr,
3321                 .server         = server,
3322                 .bitmask = bitmask,
3323                 .label          = ilabel,
3324         };
3325         struct nfs_setattrres  res = {
3326                 .fattr          = fattr,
3327                 .label          = olabel,
3328                 .server         = server,
3329         };
3330         struct nfs4_exception exception = {
3331                 .state = state,
3332                 .inode = inode,
3333                 .stateid = &arg.stateid,
3334         };
3335         int err;
3336
3337         do {
3338                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3339                                 nfs4_bitmask(server, olabel),
3340                                 inode);
3341
3342                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3343                 switch (err) {
3344                 case -NFS4ERR_OPENMODE:
3345                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3346                                 pr_warn_once("NFSv4: server %s is incorrectly "
3347                                                 "applying open mode checks to "
3348                                                 "a SETATTR that is not "
3349                                                 "changing file size.\n",
3350                                                 server->nfs_client->cl_hostname);
3351                         }
3352                         if (state && !(state->state & FMODE_WRITE)) {
3353                                 err = -EBADF;
3354                                 if (sattr->ia_valid & ATTR_OPEN)
3355                                         err = -EACCES;
3356                                 goto out;
3357                         }
3358                 }
3359                 err = nfs4_handle_exception(server, err, &exception);
3360         } while (exception.retry);
3361 out:
3362         return err;
3363 }
3364
3365 static bool
3366 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3367 {
3368         if (inode == NULL || !nfs_have_layout(inode))
3369                 return false;
3370
3371         return pnfs_wait_on_layoutreturn(inode, task);
3372 }
3373
3374 /*
3375  * Update the seqid of an open stateid
3376  */
3377 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3378                 struct nfs4_state *state)
3379 {
3380         __be32 seqid_open;
3381         u32 dst_seqid;
3382         int seq;
3383
3384         for (;;) {
3385                 if (!nfs4_valid_open_stateid(state))
3386                         break;
3387                 seq = read_seqbegin(&state->seqlock);
3388                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3389                         nfs4_stateid_copy(dst, &state->open_stateid);
3390                         if (read_seqretry(&state->seqlock, seq))
3391                                 continue;
3392                         break;
3393                 }
3394                 seqid_open = state->open_stateid.seqid;
3395                 if (read_seqretry(&state->seqlock, seq))
3396                         continue;
3397
3398                 dst_seqid = be32_to_cpu(dst->seqid);
3399                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3400                         dst->seqid = seqid_open;
3401                 break;
3402         }
3403 }
3404
3405 /*
3406  * Update the seqid of an open stateid after receiving
3407  * NFS4ERR_OLD_STATEID
3408  */
3409 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3410                 struct nfs4_state *state)
3411 {
3412         __be32 seqid_open;
3413         u32 dst_seqid;
3414         bool ret;
3415         int seq;
3416
3417         for (;;) {
3418                 ret = false;
3419                 if (!nfs4_valid_open_stateid(state))
3420                         break;
3421                 seq = read_seqbegin(&state->seqlock);
3422                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3423                         if (read_seqretry(&state->seqlock, seq))
3424                                 continue;
3425                         break;
3426                 }
3427                 seqid_open = state->open_stateid.seqid;
3428                 if (read_seqretry(&state->seqlock, seq))
3429                         continue;
3430
3431                 dst_seqid = be32_to_cpu(dst->seqid);
3432                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) >= 0)
3433                         dst->seqid = cpu_to_be32(dst_seqid + 1);
3434                 else
3435                         dst->seqid = seqid_open;
3436                 ret = true;
3437                 break;
3438         }
3439
3440         return ret;
3441 }
3442
3443 struct nfs4_closedata {
3444         struct inode *inode;
3445         struct nfs4_state *state;
3446         struct nfs_closeargs arg;
3447         struct nfs_closeres res;
3448         struct {
3449                 struct nfs4_layoutreturn_args arg;
3450                 struct nfs4_layoutreturn_res res;
3451                 struct nfs4_xdr_opaque_data ld_private;
3452                 u32 roc_barrier;
3453                 bool roc;
3454         } lr;
3455         struct nfs_fattr fattr;
3456         unsigned long timestamp;
3457 };
3458
3459 static void nfs4_free_closedata(void *data)
3460 {
3461         struct nfs4_closedata *calldata = data;
3462         struct nfs4_state_owner *sp = calldata->state->owner;
3463         struct super_block *sb = calldata->state->inode->i_sb;
3464
3465         if (calldata->lr.roc)
3466                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3467                                 calldata->res.lr_ret);
3468         nfs4_put_open_state(calldata->state);
3469         nfs_free_seqid(calldata->arg.seqid);
3470         nfs4_put_state_owner(sp);
3471         nfs_sb_deactive(sb);
3472         kfree(calldata);
3473 }
3474
3475 static void nfs4_close_done(struct rpc_task *task, void *data)
3476 {
3477         struct nfs4_closedata *calldata = data;
3478         struct nfs4_state *state = calldata->state;
3479         struct nfs_server *server = NFS_SERVER(calldata->inode);
3480         nfs4_stateid *res_stateid = NULL;
3481         struct nfs4_exception exception = {
3482                 .state = state,
3483                 .inode = calldata->inode,
3484                 .stateid = &calldata->arg.stateid,
3485         };
3486
3487         dprintk("%s: begin!\n", __func__);
3488         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3489                 return;
3490         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3491
3492         /* Handle Layoutreturn errors */
3493         if (pnfs_roc_done(task, calldata->inode,
3494                                 &calldata->arg.lr_args,
3495                                 &calldata->res.lr_res,
3496                                 &calldata->res.lr_ret) == -EAGAIN)
3497                 goto out_restart;
3498
3499         /* hmm. we are done with the inode, and in the process of freeing
3500          * the state_owner. we keep this around to process errors
3501          */
3502         switch (task->tk_status) {
3503                 case 0:
3504                         res_stateid = &calldata->res.stateid;
3505                         renew_lease(server, calldata->timestamp);
3506                         break;
3507                 case -NFS4ERR_ACCESS:
3508                         if (calldata->arg.bitmask != NULL) {
3509                                 calldata->arg.bitmask = NULL;
3510                                 calldata->res.fattr = NULL;
3511                                 goto out_restart;
3512
3513                         }
3514                         break;
3515                 case -NFS4ERR_OLD_STATEID:
3516                         /* Did we race with OPEN? */
3517                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3518                                                 state))
3519                                 goto out_restart;
3520                         goto out_release;
3521                 case -NFS4ERR_ADMIN_REVOKED:
3522                 case -NFS4ERR_STALE_STATEID:
3523                 case -NFS4ERR_EXPIRED:
3524                         nfs4_free_revoked_stateid(server,
3525                                         &calldata->arg.stateid,
3526                                         task->tk_msg.rpc_cred);
3527                         /* Fallthrough */
3528                 case -NFS4ERR_BAD_STATEID:
3529                         if (calldata->arg.fmode == 0)
3530                                 break;
3531                         /* Fallthrough */
3532                 default:
3533                         task->tk_status = nfs4_async_handle_exception(task,
3534                                         server, task->tk_status, &exception);
3535                         if (exception.retry)
3536                                 goto out_restart;
3537         }
3538         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3539                         res_stateid, calldata->arg.fmode);
3540 out_release:
3541         task->tk_status = 0;
3542         nfs_release_seqid(calldata->arg.seqid);
3543         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3544         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3545         return;
3546 out_restart:
3547         task->tk_status = 0;
3548         rpc_restart_call_prepare(task);
3549         goto out_release;
3550 }
3551
3552 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3553 {
3554         struct nfs4_closedata *calldata = data;
3555         struct nfs4_state *state = calldata->state;
3556         struct inode *inode = calldata->inode;
3557         struct pnfs_layout_hdr *lo;
3558         bool is_rdonly, is_wronly, is_rdwr;
3559         int call_close = 0;
3560
3561         dprintk("%s: begin!\n", __func__);
3562         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3563                 goto out_wait;
3564
3565         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3566         spin_lock(&state->owner->so_lock);
3567         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3568         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3569         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3570         /* Calculate the change in open mode */
3571         calldata->arg.fmode = 0;
3572         if (state->n_rdwr == 0) {
3573                 if (state->n_rdonly == 0)
3574                         call_close |= is_rdonly;
3575                 else if (is_rdonly)
3576                         calldata->arg.fmode |= FMODE_READ;
3577                 if (state->n_wronly == 0)
3578                         call_close |= is_wronly;
3579                 else if (is_wronly)
3580                         calldata->arg.fmode |= FMODE_WRITE;
3581                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3582                         call_close |= is_rdwr;
3583         } else if (is_rdwr)
3584                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3585
3586         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3587         if (!nfs4_valid_open_stateid(state))
3588                 call_close = 0;
3589         spin_unlock(&state->owner->so_lock);
3590
3591         if (!call_close) {
3592                 /* Note: exit _without_ calling nfs4_close_done */
3593                 goto out_no_action;
3594         }
3595
3596         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3597                 nfs_release_seqid(calldata->arg.seqid);
3598                 goto out_wait;
3599         }
3600
3601         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3602         if (lo && !pnfs_layout_is_valid(lo)) {
3603                 calldata->arg.lr_args = NULL;
3604                 calldata->res.lr_res = NULL;
3605         }
3606
3607         if (calldata->arg.fmode == 0)
3608                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3609
3610         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3611                 /* Close-to-open cache consistency revalidation */
3612                 if (!nfs4_have_delegation(inode, FMODE_READ))
3613                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3614                 else
3615                         calldata->arg.bitmask = NULL;
3616         }
3617
3618         calldata->arg.share_access =
3619                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3620                                 calldata->arg.fmode, 0);
3621
3622         if (calldata->res.fattr == NULL)
3623                 calldata->arg.bitmask = NULL;
3624         else if (calldata->arg.bitmask == NULL)
3625                 calldata->res.fattr = NULL;
3626         calldata->timestamp = jiffies;
3627         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3628                                 &calldata->arg.seq_args,
3629                                 &calldata->res.seq_res,
3630                                 task) != 0)
3631                 nfs_release_seqid(calldata->arg.seqid);
3632         dprintk("%s: done!\n", __func__);
3633         return;
3634 out_no_action:
3635         task->tk_action = NULL;
3636 out_wait:
3637         nfs4_sequence_done(task, &calldata->res.seq_res);
3638 }
3639
3640 static const struct rpc_call_ops nfs4_close_ops = {
3641         .rpc_call_prepare = nfs4_close_prepare,
3642         .rpc_call_done = nfs4_close_done,
3643         .rpc_release = nfs4_free_closedata,
3644 };
3645
3646 /* 
3647  * It is possible for data to be read/written from a mem-mapped file 
3648  * after the sys_close call (which hits the vfs layer as a flush).
3649  * This means that we can't safely call nfsv4 close on a file until 
3650  * the inode is cleared. This in turn means that we are not good
3651  * NFSv4 citizens - we do not indicate to the server to update the file's 
3652  * share state even when we are done with one of the three share 
3653  * stateid's in the inode.
3654  *
3655  * NOTE: Caller must be holding the sp->so_owner semaphore!
3656  */
3657 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3658 {
3659         struct nfs_server *server = NFS_SERVER(state->inode);
3660         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3661         struct nfs4_closedata *calldata;
3662         struct nfs4_state_owner *sp = state->owner;
3663         struct rpc_task *task;
3664         struct rpc_message msg = {
3665                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3666                 .rpc_cred = state->owner->so_cred,
3667         };
3668         struct rpc_task_setup task_setup_data = {
3669                 .rpc_client = server->client,
3670                 .rpc_message = &msg,
3671                 .callback_ops = &nfs4_close_ops,
3672                 .workqueue = nfsiod_workqueue,
3673                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3674         };
3675         int status = -ENOMEM;
3676
3677         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3678                 &task_setup_data.rpc_client, &msg);
3679
3680         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3681         if (calldata == NULL)
3682                 goto out;
3683         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3684         calldata->inode = state->inode;
3685         calldata->state = state;
3686         calldata->arg.fh = NFS_FH(state->inode);
3687         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3688                 goto out_free_calldata;
3689         /* Serialization for the sequence id */
3690         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3691         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3692         if (IS_ERR(calldata->arg.seqid))
3693                 goto out_free_calldata;
3694         nfs_fattr_init(&calldata->fattr);
3695         calldata->arg.fmode = 0;
3696         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3697         calldata->res.fattr = &calldata->fattr;
3698         calldata->res.seqid = calldata->arg.seqid;
3699         calldata->res.server = server;
3700         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3701         calldata->lr.roc = pnfs_roc(state->inode,
3702                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3703         if (calldata->lr.roc) {
3704                 calldata->arg.lr_args = &calldata->lr.arg;
3705                 calldata->res.lr_res = &calldata->lr.res;
3706         }
3707         nfs_sb_active(calldata->inode->i_sb);
3708
3709         msg.rpc_argp = &calldata->arg;
3710         msg.rpc_resp = &calldata->res;
3711         task_setup_data.callback_data = calldata;
3712         task = rpc_run_task(&task_setup_data);
3713         if (IS_ERR(task))
3714                 return PTR_ERR(task);
3715         status = 0;
3716         if (wait)
3717                 status = rpc_wait_for_completion_task(task);
3718         rpc_put_task(task);
3719         return status;
3720 out_free_calldata:
3721         kfree(calldata);
3722 out:
3723         nfs4_put_open_state(state);
3724         nfs4_put_state_owner(sp);
3725         return status;
3726 }
3727
3728 static struct inode *
3729 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3730                 int open_flags, struct iattr *attr, int *opened)
3731 {
3732         struct nfs4_state *state;
3733         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3734
3735         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3736
3737         /* Protect against concurrent sillydeletes */
3738         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3739
3740         nfs4_label_release_security(label);
3741
3742         if (IS_ERR(state))
3743                 return ERR_CAST(state);
3744         return state->inode;
3745 }
3746
3747 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3748 {
3749         if (ctx->state == NULL)
3750                 return;
3751         if (is_sync)
3752                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3753         else
3754                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3755 }
3756
3757 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3758 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3759 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3760
3761 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3762 {
3763         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3764         struct nfs4_server_caps_arg args = {
3765                 .fhandle = fhandle,
3766                 .bitmask = bitmask,
3767         };
3768         struct nfs4_server_caps_res res = {};
3769         struct rpc_message msg = {
3770                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3771                 .rpc_argp = &args,
3772                 .rpc_resp = &res,
3773         };
3774         int status;
3775         int i;
3776
3777         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3778                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3779                      FATTR4_WORD0_LINK_SUPPORT |
3780                      FATTR4_WORD0_SYMLINK_SUPPORT |
3781                      FATTR4_WORD0_ACLSUPPORT;
3782         if (minorversion)
3783                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3784
3785         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3786         if (status == 0) {
3787                 /* Sanity check the server answers */
3788                 switch (minorversion) {
3789                 case 0:
3790                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3791                         res.attr_bitmask[2] = 0;
3792                         break;
3793                 case 1:
3794                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3795                         break;
3796                 case 2:
3797                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3798                 }
3799                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3800                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3801                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3802                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3803                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3804                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3805                                 NFS_CAP_SECURITY_LABEL);
3806                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3807                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3808                         server->caps |= NFS_CAP_ACLS;
3809                 if (res.has_links != 0)
3810                         server->caps |= NFS_CAP_HARDLINKS;
3811                 if (res.has_symlinks != 0)
3812                         server->caps |= NFS_CAP_SYMLINKS;
3813                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3814                         server->caps |= NFS_CAP_FILEID;
3815                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3816                         server->caps |= NFS_CAP_MODE;
3817                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3818                         server->caps |= NFS_CAP_NLINK;
3819                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3820                         server->caps |= NFS_CAP_OWNER;
3821                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3822                         server->caps |= NFS_CAP_OWNER_GROUP;
3823                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3824                         server->caps |= NFS_CAP_ATIME;
3825                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3826                         server->caps |= NFS_CAP_CTIME;
3827                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3828                         server->caps |= NFS_CAP_MTIME;
3829 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3830                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3831                         server->caps |= NFS_CAP_SECURITY_LABEL;
3832 #endif
3833                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3834                                 sizeof(server->attr_bitmask));
3835                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3836
3837                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3838                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3839                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3840                 server->cache_consistency_bitmask[2] = 0;
3841
3842                 /* Avoid a regression due to buggy server */
3843                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3844                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3845                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3846                         sizeof(server->exclcreat_bitmask));
3847
3848                 server->acl_bitmask = res.acl_bitmask;
3849                 server->fh_expire_type = res.fh_expire_type;
3850         }
3851
3852         return status;
3853 }
3854
3855 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3856 {
3857         struct nfs4_exception exception = {
3858                 .interruptible = true,
3859         };
3860         int err;
3861         do {
3862                 err = nfs4_handle_exception(server,
3863                                 _nfs4_server_capabilities(server, fhandle),
3864                                 &exception);
3865         } while (exception.retry);
3866         return err;
3867 }
3868
3869 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3870                 struct nfs_fsinfo *info)
3871 {
3872         u32 bitmask[3];
3873         struct nfs4_lookup_root_arg args = {
3874                 .bitmask = bitmask,
3875         };
3876         struct nfs4_lookup_res res = {
3877                 .server = server,
3878                 .fattr = info->fattr,
3879                 .fh = fhandle,
3880         };
3881         struct rpc_message msg = {
3882                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3883                 .rpc_argp = &args,
3884                 .rpc_resp = &res,
3885         };
3886
3887         bitmask[0] = nfs4_fattr_bitmap[0];
3888         bitmask[1] = nfs4_fattr_bitmap[1];
3889         /*
3890          * Process the label in the upcoming getfattr
3891          */
3892         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3893
3894         nfs_fattr_init(info->fattr);
3895         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3896 }
3897
3898 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3899                 struct nfs_fsinfo *info)
3900 {
3901         struct nfs4_exception exception = {
3902                 .interruptible = true,
3903         };
3904         int err;
3905         do {
3906                 err = _nfs4_lookup_root(server, fhandle, info);
3907                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3908                 switch (err) {
3909                 case 0:
3910                 case -NFS4ERR_WRONGSEC:
3911                         goto out;
3912                 default:
3913                         err = nfs4_handle_exception(server, err, &exception);
3914                 }
3915         } while (exception.retry);
3916 out:
3917         return err;
3918 }
3919
3920 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3921                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3922 {
3923         struct rpc_auth_create_args auth_args = {
3924                 .pseudoflavor = flavor,
3925         };
3926         struct rpc_auth *auth;
3927
3928         auth = rpcauth_create(&auth_args, server->client);
3929         if (IS_ERR(auth))
3930                 return -EACCES;
3931         return nfs4_lookup_root(server, fhandle, info);
3932 }
3933
3934 /*
3935  * Retry pseudoroot lookup with various security flavors.  We do this when:
3936  *
3937  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3938  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3939  *
3940  * Returns zero on success, or a negative NFS4ERR value, or a
3941  * negative errno value.
3942  */
3943 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3944                               struct nfs_fsinfo *info)
3945 {
3946         /* Per 3530bis 15.33.5 */
3947         static const rpc_authflavor_t flav_array[] = {
3948                 RPC_AUTH_GSS_KRB5P,
3949                 RPC_AUTH_GSS_KRB5I,
3950                 RPC_AUTH_GSS_KRB5,
3951                 RPC_AUTH_UNIX,                  /* courtesy */
3952                 RPC_AUTH_NULL,
3953         };
3954         int status = -EPERM;
3955         size_t i;
3956
3957         if (server->auth_info.flavor_len > 0) {
3958                 /* try each flavor specified by user */
3959                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3960                         status = nfs4_lookup_root_sec(server, fhandle, info,
3961                                                 server->auth_info.flavors[i]);
3962                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3963                                 continue;
3964                         break;
3965                 }
3966         } else {
3967                 /* no flavors specified by user, try default list */
3968                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3969                         status = nfs4_lookup_root_sec(server, fhandle, info,
3970                                                       flav_array[i]);
3971                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3972                                 continue;
3973                         break;
3974                 }
3975         }
3976
3977         /*
3978          * -EACCES could mean that the user doesn't have correct permissions
3979          * to access the mount.  It could also mean that we tried to mount
3980          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3981          * existing mount programs don't handle -EACCES very well so it should
3982          * be mapped to -EPERM instead.
3983          */
3984         if (status == -EACCES)
3985                 status = -EPERM;
3986         return status;
3987 }
3988
3989 /**
3990  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3991  * @server: initialized nfs_server handle
3992  * @fhandle: we fill in the pseudo-fs root file handle
3993  * @info: we fill in an FSINFO struct
3994  * @auth_probe: probe the auth flavours
3995  *
3996  * Returns zero on success, or a negative errno.
3997  */
3998 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3999                          struct nfs_fsinfo *info,
4000                          bool auth_probe)
4001 {
4002         int status = 0;
4003
4004         if (!auth_probe)
4005                 status = nfs4_lookup_root(server, fhandle, info);
4006
4007         if (auth_probe || status == NFS4ERR_WRONGSEC)
4008                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4009                                 fhandle, info);
4010
4011         if (status == 0)
4012                 status = nfs4_server_capabilities(server, fhandle);
4013         if (status == 0)
4014                 status = nfs4_do_fsinfo(server, fhandle, info);
4015
4016         return nfs4_map_errors(status);
4017 }
4018
4019 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4020                               struct nfs_fsinfo *info)
4021 {
4022         int error;
4023         struct nfs_fattr *fattr = info->fattr;
4024         struct nfs4_label *label = fattr->label;
4025
4026         error = nfs4_server_capabilities(server, mntfh);
4027         if (error < 0) {
4028                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4029                 return error;
4030         }
4031
4032         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4033         if (error < 0) {
4034                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4035                 goto out;
4036         }
4037
4038         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4039             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4040                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4041
4042 out:
4043         return error;
4044 }
4045
4046 /*
4047  * Get locations and (maybe) other attributes of a referral.
4048  * Note that we'll actually follow the referral later when
4049  * we detect fsid mismatch in inode revalidation
4050  */
4051 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4052                              const struct qstr *name, struct nfs_fattr *fattr,
4053                              struct nfs_fh *fhandle)
4054 {
4055         int status = -ENOMEM;
4056         struct page *page = NULL;
4057         struct nfs4_fs_locations *locations = NULL;
4058
4059         page = alloc_page(GFP_KERNEL);
4060         if (page == NULL)
4061                 goto out;
4062         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4063         if (locations == NULL)
4064                 goto out;
4065
4066         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4067         if (status != 0)
4068                 goto out;
4069
4070         /*
4071          * If the fsid didn't change, this is a migration event, not a
4072          * referral.  Cause us to drop into the exception handler, which
4073          * will kick off migration recovery.
4074          */
4075         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4076                 dprintk("%s: server did not return a different fsid for"
4077                         " a referral at %s\n", __func__, name->name);
4078                 status = -NFS4ERR_MOVED;
4079                 goto out;
4080         }
4081         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4082         nfs_fixup_referral_attributes(&locations->fattr);
4083
4084         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4085         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4086         memset(fhandle, 0, sizeof(struct nfs_fh));
4087 out:
4088         if (page)
4089                 __free_page(page);
4090         kfree(locations);
4091         return status;
4092 }
4093
4094 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4095                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4096                                 struct inode *inode)
4097 {
4098         __u32 bitmask[NFS4_BITMASK_SZ];
4099         struct nfs4_getattr_arg args = {
4100                 .fh = fhandle,
4101                 .bitmask = bitmask,
4102         };
4103         struct nfs4_getattr_res res = {
4104                 .fattr = fattr,
4105                 .label = label,
4106                 .server = server,
4107         };
4108         struct rpc_message msg = {
4109                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4110                 .rpc_argp = &args,
4111                 .rpc_resp = &res,
4112         };
4113         unsigned short task_flags = 0;
4114
4115         /* Is this is an attribute revalidation, subject to softreval? */
4116         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4117                 task_flags |= RPC_TASK_TIMEOUT;
4118
4119         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4120
4121         nfs_fattr_init(fattr);
4122         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4123         return nfs4_do_call_sync(server->client, server, &msg,
4124                         &args.seq_args, &res.seq_res, task_flags);
4125 }
4126
4127 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4128                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4129                                 struct inode *inode)
4130 {
4131         struct nfs4_exception exception = {
4132                 .interruptible = true,
4133         };
4134         int err;
4135         do {
4136                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4137                 trace_nfs4_getattr(server, fhandle, fattr, err);
4138                 err = nfs4_handle_exception(server, err,
4139                                 &exception);
4140         } while (exception.retry);
4141         return err;
4142 }
4143
4144 /* 
4145  * The file is not closed if it is opened due to the a request to change
4146  * the size of the file. The open call will not be needed once the
4147  * VFS layer lookup-intents are implemented.
4148  *
4149  * Close is called when the inode is destroyed.
4150  * If we haven't opened the file for O_WRONLY, we
4151  * need to in the size_change case to obtain a stateid.
4152  *
4153  * Got race?
4154  * Because OPEN is always done by name in nfsv4, it is
4155  * possible that we opened a different file by the same
4156  * name.  We can recognize this race condition, but we
4157  * can't do anything about it besides returning an error.
4158  *
4159  * This will be fixed with VFS changes (lookup-intent).
4160  */
4161 static int
4162 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4163                   struct iattr *sattr)
4164 {
4165         struct inode *inode = d_inode(dentry);
4166         const struct cred *cred = NULL;
4167         struct nfs_open_context *ctx = NULL;
4168         struct nfs4_label *label = NULL;
4169         int status;
4170
4171         if (pnfs_ld_layoutret_on_setattr(inode) &&
4172             sattr->ia_valid & ATTR_SIZE &&
4173             sattr->ia_size < i_size_read(inode))
4174                 pnfs_commit_and_return_layout(inode);
4175
4176         nfs_fattr_init(fattr);
4177         
4178         /* Deal with open(O_TRUNC) */
4179         if (sattr->ia_valid & ATTR_OPEN)
4180                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4181
4182         /* Optimization: if the end result is no change, don't RPC */
4183         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4184                 return 0;
4185
4186         /* Search for an existing open(O_WRITE) file */
4187         if (sattr->ia_valid & ATTR_FILE) {
4188
4189                 ctx = nfs_file_open_context(sattr->ia_file);
4190                 if (ctx)
4191                         cred = ctx->cred;
4192         }
4193
4194         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4195         if (IS_ERR(label))
4196                 return PTR_ERR(label);
4197
4198         /* Return any delegations if we're going to change ACLs */
4199         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4200                 nfs4_inode_make_writeable(inode);
4201
4202         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4203         if (status == 0) {
4204                 nfs_setattr_update_inode(inode, sattr, fattr);
4205                 nfs_setsecurity(inode, fattr, label);
4206         }
4207         nfs4_label_free(label);
4208         return status;
4209 }
4210
4211 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4212                 struct dentry *dentry, struct nfs_fh *fhandle,
4213                 struct nfs_fattr *fattr, struct nfs4_label *label)
4214 {
4215         struct nfs_server *server = NFS_SERVER(dir);
4216         int                    status;
4217         struct nfs4_lookup_arg args = {
4218                 .bitmask = server->attr_bitmask,
4219                 .dir_fh = NFS_FH(dir),
4220                 .name = &dentry->d_name,
4221         };
4222         struct nfs4_lookup_res res = {
4223                 .server = server,
4224                 .fattr = fattr,
4225                 .label = label,
4226                 .fh = fhandle,
4227         };
4228         struct rpc_message msg = {
4229                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4230                 .rpc_argp = &args,
4231                 .rpc_resp = &res,
4232         };
4233         unsigned short task_flags = 0;
4234
4235         /* Is this is an attribute revalidation, subject to softreval? */
4236         if (nfs_lookup_is_soft_revalidate(dentry))
4237                 task_flags |= RPC_TASK_TIMEOUT;
4238
4239         args.bitmask = nfs4_bitmask(server, label);
4240
4241         nfs_fattr_init(fattr);
4242
4243         dprintk("NFS call  lookup %pd2\n", dentry);
4244         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4245         status = nfs4_do_call_sync(clnt, server, &msg,
4246                         &args.seq_args, &res.seq_res, task_flags);
4247         dprintk("NFS reply lookup: %d\n", status);
4248         return status;
4249 }
4250
4251 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4252 {
4253         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4254                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4255         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4256         fattr->nlink = 2;
4257 }
4258
4259 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4260                                    struct dentry *dentry, struct nfs_fh *fhandle,
4261                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4262 {
4263         struct nfs4_exception exception = {
4264                 .interruptible = true,
4265         };
4266         struct rpc_clnt *client = *clnt;
4267         const struct qstr *name = &dentry->d_name;
4268         int err;
4269         do {
4270                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4271                 trace_nfs4_lookup(dir, name, err);
4272                 switch (err) {
4273                 case -NFS4ERR_BADNAME:
4274                         err = -ENOENT;
4275                         goto out;
4276                 case -NFS4ERR_MOVED:
4277                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4278                         if (err == -NFS4ERR_MOVED)
4279                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4280                         goto out;
4281                 case -NFS4ERR_WRONGSEC:
4282                         err = -EPERM;
4283                         if (client != *clnt)
4284                                 goto out;
4285                         client = nfs4_negotiate_security(client, dir, name);
4286                         if (IS_ERR(client))
4287                                 return PTR_ERR(client);
4288
4289                         exception.retry = 1;
4290                         break;
4291                 default:
4292                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4293                 }
4294         } while (exception.retry);
4295
4296 out:
4297         if (err == 0)
4298                 *clnt = client;
4299         else if (client != *clnt)
4300                 rpc_shutdown_client(client);
4301
4302         return err;
4303 }
4304
4305 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4306                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4307                             struct nfs4_label *label)
4308 {
4309         int status;
4310         struct rpc_clnt *client = NFS_CLIENT(dir);
4311
4312         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4313         if (client != NFS_CLIENT(dir)) {
4314                 rpc_shutdown_client(client);
4315                 nfs_fixup_secinfo_attributes(fattr);
4316         }
4317         return status;
4318 }
4319
4320 struct rpc_clnt *
4321 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4322                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4323 {
4324         struct rpc_clnt *client = NFS_CLIENT(dir);
4325         int status;
4326
4327         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4328         if (status < 0)
4329                 return ERR_PTR(status);
4330         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4331 }
4332
4333 static int _nfs4_proc_lookupp(struct inode *inode,
4334                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4335                 struct nfs4_label *label)
4336 {
4337         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4338         struct nfs_server *server = NFS_SERVER(inode);
4339         int                    status;
4340         struct nfs4_lookupp_arg args = {
4341                 .bitmask = server->attr_bitmask,
4342                 .fh = NFS_FH(inode),
4343         };
4344         struct nfs4_lookupp_res res = {
4345                 .server = server,
4346                 .fattr = fattr,
4347                 .label = label,
4348                 .fh = fhandle,
4349         };
4350         struct rpc_message msg = {
4351                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4352                 .rpc_argp = &args,
4353                 .rpc_resp = &res,
4354         };
4355
4356         args.bitmask = nfs4_bitmask(server, label);
4357
4358         nfs_fattr_init(fattr);
4359
4360         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4361         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4362                                 &res.seq_res, 0);
4363         dprintk("NFS reply lookupp: %d\n", status);
4364         return status;
4365 }
4366
4367 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4368                              struct nfs_fattr *fattr, struct nfs4_label *label)
4369 {
4370         struct nfs4_exception exception = {
4371                 .interruptible = true,
4372         };
4373         int err;
4374         do {
4375                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4376                 trace_nfs4_lookupp(inode, err);
4377                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4378                                 &exception);
4379         } while (exception.retry);
4380         return err;
4381 }
4382
4383 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4384 {
4385         struct nfs_server *server = NFS_SERVER(inode);
4386         struct nfs4_accessargs args = {
4387                 .fh = NFS_FH(inode),
4388                 .access = entry->mask,
4389         };
4390         struct nfs4_accessres res = {
4391                 .server = server,
4392         };
4393         struct rpc_message msg = {
4394                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4395                 .rpc_argp = &args,
4396                 .rpc_resp = &res,
4397                 .rpc_cred = entry->cred,
4398         };
4399         int status = 0;
4400
4401         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4402                 res.fattr = nfs_alloc_fattr();
4403                 if (res.fattr == NULL)
4404                         return -ENOMEM;
4405                 args.bitmask = server->cache_consistency_bitmask;
4406         }
4407         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4408         if (!status) {
4409                 nfs_access_set_mask(entry, res.access);
4410                 if (res.fattr)
4411                         nfs_refresh_inode(inode, res.fattr);
4412         }
4413         nfs_free_fattr(res.fattr);
4414         return status;
4415 }
4416
4417 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4418 {
4419         struct nfs4_exception exception = {
4420                 .interruptible = true,
4421         };
4422         int err;
4423         do {
4424                 err = _nfs4_proc_access(inode, entry);
4425                 trace_nfs4_access(inode, err);
4426                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4427                                 &exception);
4428         } while (exception.retry);
4429         return err;
4430 }
4431
4432 /*
4433  * TODO: For the time being, we don't try to get any attributes
4434  * along with any of the zero-copy operations READ, READDIR,
4435  * READLINK, WRITE.
4436  *
4437  * In the case of the first three, we want to put the GETATTR
4438  * after the read-type operation -- this is because it is hard
4439  * to predict the length of a GETATTR response in v4, and thus
4440  * align the READ data correctly.  This means that the GETATTR
4441  * may end up partially falling into the page cache, and we should
4442  * shift it into the 'tail' of the xdr_buf before processing.
4443  * To do this efficiently, we need to know the total length
4444  * of data received, which doesn't seem to be available outside
4445  * of the RPC layer.
4446  *
4447  * In the case of WRITE, we also want to put the GETATTR after
4448  * the operation -- in this case because we want to make sure
4449  * we get the post-operation mtime and size.
4450  *
4451  * Both of these changes to the XDR layer would in fact be quite
4452  * minor, but I decided to leave them for a subsequent patch.
4453  */
4454 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4455                 unsigned int pgbase, unsigned int pglen)
4456 {
4457         struct nfs4_readlink args = {
4458                 .fh       = NFS_FH(inode),
4459                 .pgbase   = pgbase,
4460                 .pglen    = pglen,
4461                 .pages    = &page,
4462         };
4463         struct nfs4_readlink_res res;
4464         struct rpc_message msg = {
4465                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4466                 .rpc_argp = &args,
4467                 .rpc_resp = &res,
4468         };
4469
4470         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4471 }
4472
4473 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4474                 unsigned int pgbase, unsigned int pglen)
4475 {
4476         struct nfs4_exception exception = {
4477                 .interruptible = true,
4478         };
4479         int err;
4480         do {
4481                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4482                 trace_nfs4_readlink(inode, err);
4483                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4484                                 &exception);
4485         } while (exception.retry);
4486         return err;
4487 }
4488
4489 /*
4490  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4491  */
4492 static int
4493 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4494                  int flags)
4495 {
4496         struct nfs_server *server = NFS_SERVER(dir);
4497         struct nfs4_label l, *ilabel = NULL;
4498         struct nfs_open_context *ctx;
4499         struct nfs4_state *state;
4500         int status = 0;
4501
4502         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4503         if (IS_ERR(ctx))
4504                 return PTR_ERR(ctx);
4505
4506         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4507
4508         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4509                 sattr->ia_mode &= ~current_umask();
4510         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4511         if (IS_ERR(state)) {
4512                 status = PTR_ERR(state);
4513                 goto out;
4514         }
4515 out:
4516         nfs4_label_release_security(ilabel);
4517         put_nfs_open_context(ctx);
4518         return status;
4519 }
4520
4521 static int
4522 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4523 {
4524         struct nfs_server *server = NFS_SERVER(dir);
4525         struct nfs_removeargs args = {
4526                 .fh = NFS_FH(dir),
4527                 .name = *name,
4528         };
4529         struct nfs_removeres res = {
4530                 .server = server,
4531         };
4532         struct rpc_message msg = {
4533                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4534                 .rpc_argp = &args,
4535                 .rpc_resp = &res,
4536         };
4537         unsigned long timestamp = jiffies;
4538         int status;
4539
4540         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4541         if (status == 0) {
4542                 spin_lock(&dir->i_lock);
4543                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4544                 /* Removing a directory decrements nlink in the parent */
4545                 if (ftype == NF4DIR && dir->i_nlink > 2)
4546                         nfs4_dec_nlink_locked(dir);
4547                 spin_unlock(&dir->i_lock);
4548         }
4549         return status;
4550 }
4551
4552 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4553 {
4554         struct nfs4_exception exception = {
4555                 .interruptible = true,
4556         };
4557         struct inode *inode = d_inode(dentry);
4558         int err;
4559
4560         if (inode) {
4561                 if (inode->i_nlink == 1)
4562                         nfs4_inode_return_delegation(inode);
4563                 else
4564                         nfs4_inode_make_writeable(inode);
4565         }
4566         do {
4567                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4568                 trace_nfs4_remove(dir, &dentry->d_name, err);
4569                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4570                                 &exception);
4571         } while (exception.retry);
4572         return err;
4573 }
4574
4575 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4576 {
4577         struct nfs4_exception exception = {
4578                 .interruptible = true,
4579         };
4580         int err;
4581
4582         do {
4583                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4584                 trace_nfs4_remove(dir, name, err);
4585                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4586                                 &exception);
4587         } while (exception.retry);
4588         return err;
4589 }
4590
4591 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4592                 struct dentry *dentry,
4593                 struct inode *inode)
4594 {
4595         struct nfs_removeargs *args = msg->rpc_argp;
4596         struct nfs_removeres *res = msg->rpc_resp;
4597
4598         res->server = NFS_SB(dentry->d_sb);
4599         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4600         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4601
4602         nfs_fattr_init(res->dir_attr);
4603
4604         if (inode)
4605                 nfs4_inode_return_delegation(inode);
4606 }
4607
4608 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4609 {
4610         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4611                         &data->args.seq_args,
4612                         &data->res.seq_res,
4613                         task);
4614 }
4615
4616 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4617 {
4618         struct nfs_unlinkdata *data = task->tk_calldata;
4619         struct nfs_removeres *res = &data->res;
4620
4621         if (!nfs4_sequence_done(task, &res->seq_res))
4622                 return 0;
4623         if (nfs4_async_handle_error(task, res->server, NULL,
4624                                     &data->timeout) == -EAGAIN)
4625                 return 0;
4626         if (task->tk_status == 0)
4627                 update_changeattr(dir, &res->cinfo,
4628                                 res->dir_attr->time_start, 0);
4629         return 1;
4630 }
4631
4632 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4633                 struct dentry *old_dentry,
4634                 struct dentry *new_dentry)
4635 {
4636         struct nfs_renameargs *arg = msg->rpc_argp;
4637         struct nfs_renameres *res = msg->rpc_resp;
4638         struct inode *old_inode = d_inode(old_dentry);
4639         struct inode *new_inode = d_inode(new_dentry);
4640
4641         if (old_inode)
4642                 nfs4_inode_make_writeable(old_inode);
4643         if (new_inode)
4644                 nfs4_inode_return_delegation(new_inode);
4645         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4646         res->server = NFS_SB(old_dentry->d_sb);
4647         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4648 }
4649
4650 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4651 {
4652         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4653                         &data->args.seq_args,
4654                         &data->res.seq_res,
4655                         task);
4656 }
4657
4658 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4659                                  struct inode *new_dir)
4660 {
4661         struct nfs_renamedata *data = task->tk_calldata;
4662         struct nfs_renameres *res = &data->res;
4663
4664         if (!nfs4_sequence_done(task, &res->seq_res))
4665                 return 0;
4666         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4667                 return 0;
4668
4669         if (task->tk_status == 0) {
4670                 if (new_dir != old_dir) {
4671                         /* Note: If we moved a directory, nlink will change */
4672                         update_changeattr(old_dir, &res->old_cinfo,
4673                                         res->old_fattr->time_start,
4674                                         NFS_INO_INVALID_OTHER);
4675                         update_changeattr(new_dir, &res->new_cinfo,
4676                                         res->new_fattr->time_start,
4677                                         NFS_INO_INVALID_OTHER);
4678                 } else
4679                         update_changeattr(old_dir, &res->old_cinfo,
4680                                         res->old_fattr->time_start,
4681                                         0);
4682         }
4683         return 1;
4684 }
4685
4686 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4687 {
4688         struct nfs_server *server = NFS_SERVER(inode);
4689         __u32 bitmask[NFS4_BITMASK_SZ];
4690         struct nfs4_link_arg arg = {
4691                 .fh     = NFS_FH(inode),
4692                 .dir_fh = NFS_FH(dir),
4693                 .name   = name,
4694                 .bitmask = bitmask,
4695         };
4696         struct nfs4_link_res res = {
4697                 .server = server,
4698                 .label = NULL,
4699         };
4700         struct rpc_message msg = {
4701                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4702                 .rpc_argp = &arg,
4703                 .rpc_resp = &res,
4704         };
4705         int status = -ENOMEM;
4706
4707         res.fattr = nfs_alloc_fattr();
4708         if (res.fattr == NULL)
4709                 goto out;
4710
4711         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4712         if (IS_ERR(res.label)) {
4713                 status = PTR_ERR(res.label);
4714                 goto out;
4715         }
4716
4717         nfs4_inode_make_writeable(inode);
4718         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4719
4720         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4721         if (!status) {
4722                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4723                 status = nfs_post_op_update_inode(inode, res.fattr);
4724                 if (!status)
4725                         nfs_setsecurity(inode, res.fattr, res.label);
4726         }
4727
4728
4729         nfs4_label_free(res.label);
4730
4731 out:
4732         nfs_free_fattr(res.fattr);
4733         return status;
4734 }
4735
4736 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4737 {
4738         struct nfs4_exception exception = {
4739                 .interruptible = true,
4740         };
4741         int err;
4742         do {
4743                 err = nfs4_handle_exception(NFS_SERVER(inode),
4744                                 _nfs4_proc_link(inode, dir, name),
4745                                 &exception);
4746         } while (exception.retry);
4747         return err;
4748 }
4749
4750 struct nfs4_createdata {
4751         struct rpc_message msg;
4752         struct nfs4_create_arg arg;
4753         struct nfs4_create_res res;
4754         struct nfs_fh fh;
4755         struct nfs_fattr fattr;
4756         struct nfs4_label *label;
4757 };
4758
4759 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4760                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4761 {
4762         struct nfs4_createdata *data;
4763
4764         data = kzalloc(sizeof(*data), GFP_KERNEL);
4765         if (data != NULL) {
4766                 struct nfs_server *server = NFS_SERVER(dir);
4767
4768                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4769                 if (IS_ERR(data->label))
4770                         goto out_free;
4771
4772                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4773                 data->msg.rpc_argp = &data->arg;
4774                 data->msg.rpc_resp = &data->res;
4775                 data->arg.dir_fh = NFS_FH(dir);
4776                 data->arg.server = server;
4777                 data->arg.name = name;
4778                 data->arg.attrs = sattr;
4779                 data->arg.ftype = ftype;
4780                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4781                 data->arg.umask = current_umask();
4782                 data->res.server = server;
4783                 data->res.fh = &data->fh;
4784                 data->res.fattr = &data->fattr;
4785                 data->res.label = data->label;
4786                 nfs_fattr_init(data->res.fattr);
4787         }
4788         return data;
4789 out_free:
4790         kfree(data);
4791         return NULL;
4792 }
4793
4794 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4795 {
4796         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4797                                     &data->arg.seq_args, &data->res.seq_res, 1);
4798         if (status == 0) {
4799                 spin_lock(&dir->i_lock);
4800                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4801                                 data->res.fattr->time_start, 0);
4802                 /* Creating a directory bumps nlink in the parent */
4803                 if (data->arg.ftype == NF4DIR)
4804                         nfs4_inc_nlink_locked(dir);
4805                 spin_unlock(&dir->i_lock);
4806                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4807         }
4808         return status;
4809 }
4810
4811 static void nfs4_free_createdata(struct nfs4_createdata *data)
4812 {
4813         nfs4_label_free(data->label);
4814         kfree(data);
4815 }
4816
4817 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4818                 struct page *page, unsigned int len, struct iattr *sattr,
4819                 struct nfs4_label *label)
4820 {
4821         struct nfs4_createdata *data;
4822         int status = -ENAMETOOLONG;
4823
4824         if (len > NFS4_MAXPATHLEN)
4825                 goto out;
4826
4827         status = -ENOMEM;
4828         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4829         if (data == NULL)
4830                 goto out;
4831
4832         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4833         data->arg.u.symlink.pages = &page;
4834         data->arg.u.symlink.len = len;
4835         data->arg.label = label;
4836         
4837         status = nfs4_do_create(dir, dentry, data);
4838
4839         nfs4_free_createdata(data);
4840 out:
4841         return status;
4842 }
4843
4844 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4845                 struct page *page, unsigned int len, struct iattr *sattr)
4846 {
4847         struct nfs4_exception exception = {
4848                 .interruptible = true,
4849         };
4850         struct nfs4_label l, *label = NULL;
4851         int err;
4852
4853         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4854
4855         do {
4856                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4857                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4858                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4859                                 &exception);
4860         } while (exception.retry);
4861
4862         nfs4_label_release_security(label);
4863         return err;
4864 }
4865
4866 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4867                 struct iattr *sattr, struct nfs4_label *label)
4868 {
4869         struct nfs4_createdata *data;
4870         int status = -ENOMEM;
4871
4872         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4873         if (data == NULL)
4874                 goto out;
4875
4876         data->arg.label = label;
4877         status = nfs4_do_create(dir, dentry, data);
4878
4879         nfs4_free_createdata(data);
4880 out:
4881         return status;
4882 }
4883
4884 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4885                 struct iattr *sattr)
4886 {
4887         struct nfs_server *server = NFS_SERVER(dir);
4888         struct nfs4_exception exception = {
4889                 .interruptible = true,
4890         };
4891         struct nfs4_label l, *label = NULL;
4892         int err;
4893
4894         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4895
4896         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4897                 sattr->ia_mode &= ~current_umask();
4898         do {
4899                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4900                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4901                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4902                                 &exception);
4903         } while (exception.retry);
4904         nfs4_label_release_security(label);
4905
4906         return err;
4907 }
4908
4909 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4910                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4911 {
4912         struct inode            *dir = d_inode(dentry);
4913         struct nfs4_readdir_arg args = {
4914                 .fh = NFS_FH(dir),
4915                 .pages = pages,
4916                 .pgbase = 0,
4917                 .count = count,
4918                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4919                 .plus = plus,
4920         };
4921         struct nfs4_readdir_res res;
4922         struct rpc_message msg = {
4923                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4924                 .rpc_argp = &args,
4925                 .rpc_resp = &res,
4926                 .rpc_cred = cred,
4927         };
4928         int                     status;
4929
4930         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4931                         dentry,
4932                         (unsigned long long)cookie);
4933         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4934         res.pgbase = args.pgbase;
4935         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4936         if (status >= 0) {
4937                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4938                 status += args.pgbase;
4939         }
4940
4941         nfs_invalidate_atime(dir);
4942
4943         dprintk("%s: returns %d\n", __func__, status);
4944         return status;
4945 }
4946
4947 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4948                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4949 {
4950         struct nfs4_exception exception = {
4951                 .interruptible = true,
4952         };
4953         int err;
4954         do {
4955                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4956                                 pages, count, plus);
4957                 trace_nfs4_readdir(d_inode(dentry), err);
4958                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4959                                 &exception);
4960         } while (exception.retry);
4961         return err;
4962 }
4963
4964 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4965                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4966 {
4967         struct nfs4_createdata *data;
4968         int mode = sattr->ia_mode;
4969         int status = -ENOMEM;
4970
4971         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4972         if (data == NULL)
4973                 goto out;
4974
4975         if (S_ISFIFO(mode))
4976                 data->arg.ftype = NF4FIFO;
4977         else if (S_ISBLK(mode)) {
4978                 data->arg.ftype = NF4BLK;
4979                 data->arg.u.device.specdata1 = MAJOR(rdev);
4980                 data->arg.u.device.specdata2 = MINOR(rdev);
4981         }
4982         else if (S_ISCHR(mode)) {
4983                 data->arg.ftype = NF4CHR;
4984                 data->arg.u.device.specdata1 = MAJOR(rdev);
4985                 data->arg.u.device.specdata2 = MINOR(rdev);
4986         } else if (!S_ISSOCK(mode)) {
4987                 status = -EINVAL;
4988                 goto out_free;
4989         }
4990
4991         data->arg.label = label;
4992         status = nfs4_do_create(dir, dentry, data);
4993 out_free:
4994         nfs4_free_createdata(data);
4995 out:
4996         return status;
4997 }
4998
4999 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5000                 struct iattr *sattr, dev_t rdev)
5001 {
5002         struct nfs_server *server = NFS_SERVER(dir);
5003         struct nfs4_exception exception = {
5004                 .interruptible = true,
5005         };
5006         struct nfs4_label l, *label = NULL;
5007         int err;
5008
5009         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5010
5011         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5012                 sattr->ia_mode &= ~current_umask();
5013         do {
5014                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5015                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5016                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5017                                 &exception);
5018         } while (exception.retry);
5019
5020         nfs4_label_release_security(label);
5021
5022         return err;
5023 }
5024
5025 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5026                  struct nfs_fsstat *fsstat)
5027 {
5028         struct nfs4_statfs_arg args = {
5029                 .fh = fhandle,
5030                 .bitmask = server->attr_bitmask,
5031         };
5032         struct nfs4_statfs_res res = {
5033                 .fsstat = fsstat,
5034         };
5035         struct rpc_message msg = {
5036                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5037                 .rpc_argp = &args,
5038                 .rpc_resp = &res,
5039         };
5040
5041         nfs_fattr_init(fsstat->fattr);
5042         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5043 }
5044
5045 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5046 {
5047         struct nfs4_exception exception = {
5048                 .interruptible = true,
5049         };
5050         int err;
5051         do {
5052                 err = nfs4_handle_exception(server,
5053                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5054                                 &exception);
5055         } while (exception.retry);
5056         return err;
5057 }
5058
5059 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5060                 struct nfs_fsinfo *fsinfo)
5061 {
5062         struct nfs4_fsinfo_arg args = {
5063                 .fh = fhandle,
5064                 .bitmask = server->attr_bitmask,
5065         };
5066         struct nfs4_fsinfo_res res = {
5067                 .fsinfo = fsinfo,
5068         };
5069         struct rpc_message msg = {
5070                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5071                 .rpc_argp = &args,
5072                 .rpc_resp = &res,
5073         };
5074
5075         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5076 }
5077
5078 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5079 {
5080         struct nfs4_exception exception = {
5081                 .interruptible = true,
5082         };
5083         int err;
5084
5085         do {
5086                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5087                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5088                 if (err == 0) {
5089                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5090                         break;
5091                 }
5092                 err = nfs4_handle_exception(server, err, &exception);
5093         } while (exception.retry);
5094         return err;
5095 }
5096
5097 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5098 {
5099         int error;
5100
5101         nfs_fattr_init(fsinfo->fattr);
5102         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5103         if (error == 0) {
5104                 /* block layout checks this! */
5105                 server->pnfs_blksize = fsinfo->blksize;
5106                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5107         }
5108
5109         return error;
5110 }
5111
5112 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5113                 struct nfs_pathconf *pathconf)
5114 {
5115         struct nfs4_pathconf_arg args = {
5116                 .fh = fhandle,
5117                 .bitmask = server->attr_bitmask,
5118         };
5119         struct nfs4_pathconf_res res = {
5120                 .pathconf = pathconf,
5121         };
5122         struct rpc_message msg = {
5123                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5124                 .rpc_argp = &args,
5125                 .rpc_resp = &res,
5126         };
5127
5128         /* None of the pathconf attributes are mandatory to implement */
5129         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5130                 memset(pathconf, 0, sizeof(*pathconf));
5131                 return 0;
5132         }
5133
5134         nfs_fattr_init(pathconf->fattr);
5135         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5136 }
5137
5138 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5139                 struct nfs_pathconf *pathconf)
5140 {
5141         struct nfs4_exception exception = {
5142                 .interruptible = true,
5143         };
5144         int err;
5145
5146         do {
5147                 err = nfs4_handle_exception(server,
5148                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5149                                 &exception);
5150         } while (exception.retry);
5151         return err;
5152 }
5153
5154 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5155                 const struct nfs_open_context *ctx,
5156                 const struct nfs_lock_context *l_ctx,
5157                 fmode_t fmode)
5158 {
5159         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5160 }
5161 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5162
5163 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5164                 const struct nfs_open_context *ctx,
5165                 const struct nfs_lock_context *l_ctx,
5166                 fmode_t fmode)
5167 {
5168         nfs4_stateid _current_stateid;
5169
5170         /* If the current stateid represents a lost lock, then exit */
5171         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5172                 return true;
5173         return nfs4_stateid_match(stateid, &_current_stateid);
5174 }
5175
5176 static bool nfs4_error_stateid_expired(int err)
5177 {
5178         switch (err) {
5179         case -NFS4ERR_DELEG_REVOKED:
5180         case -NFS4ERR_ADMIN_REVOKED:
5181         case -NFS4ERR_BAD_STATEID:
5182         case -NFS4ERR_STALE_STATEID:
5183         case -NFS4ERR_OLD_STATEID:
5184         case -NFS4ERR_OPENMODE:
5185         case -NFS4ERR_EXPIRED:
5186                 return true;
5187         }
5188         return false;
5189 }
5190
5191 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5192 {
5193         struct nfs_server *server = NFS_SERVER(hdr->inode);
5194
5195         trace_nfs4_read(hdr, task->tk_status);
5196         if (task->tk_status < 0) {
5197                 struct nfs4_exception exception = {
5198                         .inode = hdr->inode,
5199                         .state = hdr->args.context->state,
5200                         .stateid = &hdr->args.stateid,
5201                 };
5202                 task->tk_status = nfs4_async_handle_exception(task,
5203                                 server, task->tk_status, &exception);
5204                 if (exception.retry) {
5205                         rpc_restart_call_prepare(task);
5206                         return -EAGAIN;
5207                 }
5208         }
5209
5210         if (task->tk_status > 0)
5211                 renew_lease(server, hdr->timestamp);
5212         return 0;
5213 }
5214
5215 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5216                 struct nfs_pgio_args *args)
5217 {
5218
5219         if (!nfs4_error_stateid_expired(task->tk_status) ||
5220                 nfs4_stateid_is_current(&args->stateid,
5221                                 args->context,
5222                                 args->lock_context,
5223                                 FMODE_READ))
5224                 return false;
5225         rpc_restart_call_prepare(task);
5226         return true;
5227 }
5228
5229 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5230 {
5231
5232         dprintk("--> %s\n", __func__);
5233
5234         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5235                 return -EAGAIN;
5236         if (nfs4_read_stateid_changed(task, &hdr->args))
5237                 return -EAGAIN;
5238         if (task->tk_status > 0)
5239                 nfs_invalidate_atime(hdr->inode);
5240         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5241                                     nfs4_read_done_cb(task, hdr);
5242 }
5243
5244 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5245                                  struct rpc_message *msg)
5246 {
5247         hdr->timestamp   = jiffies;
5248         if (!hdr->pgio_done_cb)
5249                 hdr->pgio_done_cb = nfs4_read_done_cb;
5250         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5251         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5252 }
5253
5254 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5255                                       struct nfs_pgio_header *hdr)
5256 {
5257         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5258                         &hdr->args.seq_args,
5259                         &hdr->res.seq_res,
5260                         task))
5261                 return 0;
5262         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5263                                 hdr->args.lock_context,
5264                                 hdr->rw_mode) == -EIO)
5265                 return -EIO;
5266         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5267                 return -EIO;
5268         return 0;
5269 }
5270
5271 static int nfs4_write_done_cb(struct rpc_task *task,
5272                               struct nfs_pgio_header *hdr)
5273 {
5274         struct inode *inode = hdr->inode;
5275
5276         trace_nfs4_write(hdr, task->tk_status);
5277         if (task->tk_status < 0) {
5278                 struct nfs4_exception exception = {
5279                         .inode = hdr->inode,
5280                         .state = hdr->args.context->state,
5281                         .stateid = &hdr->args.stateid,
5282                 };
5283                 task->tk_status = nfs4_async_handle_exception(task,
5284                                 NFS_SERVER(inode), task->tk_status,
5285                                 &exception);
5286                 if (exception.retry) {
5287                         rpc_restart_call_prepare(task);
5288                         return -EAGAIN;
5289                 }
5290         }
5291         if (task->tk_status >= 0) {
5292                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5293                 nfs_writeback_update_inode(hdr);
5294         }
5295         return 0;
5296 }
5297
5298 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5299                 struct nfs_pgio_args *args)
5300 {
5301
5302         if (!nfs4_error_stateid_expired(task->tk_status) ||
5303                 nfs4_stateid_is_current(&args->stateid,
5304                                 args->context,
5305                                 args->lock_context,
5306                                 FMODE_WRITE))
5307                 return false;
5308         rpc_restart_call_prepare(task);
5309         return true;
5310 }
5311
5312 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5313 {
5314         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5315                 return -EAGAIN;
5316         if (nfs4_write_stateid_changed(task, &hdr->args))
5317                 return -EAGAIN;
5318         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5319                 nfs4_write_done_cb(task, hdr);
5320 }
5321
5322 static
5323 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5324 {
5325         /* Don't request attributes for pNFS or O_DIRECT writes */
5326         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5327                 return false;
5328         /* Otherwise, request attributes if and only if we don't hold
5329          * a delegation
5330          */
5331         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5332 }
5333
5334 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5335                                   struct rpc_message *msg,
5336                                   struct rpc_clnt **clnt)
5337 {
5338         struct nfs_server *server = NFS_SERVER(hdr->inode);
5339
5340         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5341                 hdr->args.bitmask = NULL;
5342                 hdr->res.fattr = NULL;
5343         } else
5344                 hdr->args.bitmask = server->cache_consistency_bitmask;
5345
5346         if (!hdr->pgio_done_cb)
5347                 hdr->pgio_done_cb = nfs4_write_done_cb;
5348         hdr->res.server = server;
5349         hdr->timestamp   = jiffies;
5350
5351         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5352         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5353         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5354 }
5355
5356 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5357 {
5358         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5359                         &data->args.seq_args,
5360                         &data->res.seq_res,
5361                         task);
5362 }
5363
5364 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5365 {
5366         struct inode *inode = data->inode;
5367
5368         trace_nfs4_commit(data, task->tk_status);
5369         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5370                                     NULL, NULL) == -EAGAIN) {
5371                 rpc_restart_call_prepare(task);
5372                 return -EAGAIN;
5373         }
5374         return 0;
5375 }
5376
5377 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5378 {
5379         if (!nfs4_sequence_done(task, &data->res.seq_res))
5380                 return -EAGAIN;
5381         return data->commit_done_cb(task, data);
5382 }
5383
5384 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5385                                    struct rpc_clnt **clnt)
5386 {
5387         struct nfs_server *server = NFS_SERVER(data->inode);
5388
5389         if (data->commit_done_cb == NULL)
5390                 data->commit_done_cb = nfs4_commit_done_cb;
5391         data->res.server = server;
5392         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5393         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5394         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5395 }
5396
5397 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5398                                 struct nfs_commitres *res)
5399 {
5400         struct inode *dst_inode = file_inode(dst);
5401         struct nfs_server *server = NFS_SERVER(dst_inode);
5402         struct rpc_message msg = {
5403                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5404                 .rpc_argp = args,
5405                 .rpc_resp = res,
5406         };
5407
5408         args->fh = NFS_FH(dst_inode);
5409         return nfs4_call_sync(server->client, server, &msg,
5410                         &args->seq_args, &res->seq_res, 1);
5411 }
5412
5413 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5414 {
5415         struct nfs_commitargs args = {
5416                 .offset = offset,
5417                 .count = count,
5418         };
5419         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5420         struct nfs4_exception exception = { };
5421         int status;
5422
5423         do {
5424                 status = _nfs4_proc_commit(dst, &args, res);
5425                 status = nfs4_handle_exception(dst_server, status, &exception);
5426         } while (exception.retry);
5427
5428         return status;
5429 }
5430
5431 struct nfs4_renewdata {
5432         struct nfs_client       *client;
5433         unsigned long           timestamp;
5434 };
5435
5436 /*
5437  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5438  * standalone procedure for queueing an asynchronous RENEW.
5439  */
5440 static void nfs4_renew_release(void *calldata)
5441 {
5442         struct nfs4_renewdata *data = calldata;
5443         struct nfs_client *clp = data->client;
5444
5445         if (refcount_read(&clp->cl_count) > 1)
5446                 nfs4_schedule_state_renewal(clp);
5447         nfs_put_client(clp);
5448         kfree(data);
5449 }
5450
5451 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5452 {
5453         struct nfs4_renewdata *data = calldata;
5454         struct nfs_client *clp = data->client;
5455         unsigned long timestamp = data->timestamp;
5456
5457         trace_nfs4_renew_async(clp, task->tk_status);
5458         switch (task->tk_status) {
5459         case 0:
5460                 break;
5461         case -NFS4ERR_LEASE_MOVED:
5462                 nfs4_schedule_lease_moved_recovery(clp);
5463                 break;
5464         default:
5465                 /* Unless we're shutting down, schedule state recovery! */
5466                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5467                         return;
5468                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5469                         nfs4_schedule_lease_recovery(clp);
5470                         return;
5471                 }
5472                 nfs4_schedule_path_down_recovery(clp);
5473         }
5474         do_renew_lease(clp, timestamp);
5475 }
5476
5477 static const struct rpc_call_ops nfs4_renew_ops = {
5478         .rpc_call_done = nfs4_renew_done,
5479         .rpc_release = nfs4_renew_release,
5480 };
5481
5482 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5483 {
5484         struct rpc_message msg = {
5485                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5486                 .rpc_argp       = clp,
5487                 .rpc_cred       = cred,
5488         };
5489         struct nfs4_renewdata *data;
5490
5491         if (renew_flags == 0)
5492                 return 0;
5493         if (!refcount_inc_not_zero(&clp->cl_count))
5494                 return -EIO;
5495         data = kmalloc(sizeof(*data), GFP_NOFS);
5496         if (data == NULL) {
5497                 nfs_put_client(clp);
5498                 return -ENOMEM;
5499         }
5500         data->client = clp;
5501         data->timestamp = jiffies;
5502         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5503                         &nfs4_renew_ops, data);
5504 }
5505
5506 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5507 {
5508         struct rpc_message msg = {
5509                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5510                 .rpc_argp       = clp,
5511                 .rpc_cred       = cred,
5512         };
5513         unsigned long now = jiffies;
5514         int status;
5515
5516         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5517         if (status < 0)
5518                 return status;
5519         do_renew_lease(clp, now);
5520         return 0;
5521 }
5522
5523 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5524 {
5525         return server->caps & NFS_CAP_ACLS;
5526 }
5527
5528 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5529  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5530  * the stack.
5531  */
5532 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5533
5534 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5535                 struct page **pages)
5536 {
5537         struct page *newpage, **spages;
5538         int rc = 0;
5539         size_t len;
5540         spages = pages;
5541
5542         do {
5543                 len = min_t(size_t, PAGE_SIZE, buflen);
5544                 newpage = alloc_page(GFP_KERNEL);
5545
5546                 if (newpage == NULL)
5547                         goto unwind;
5548                 memcpy(page_address(newpage), buf, len);
5549                 buf += len;
5550                 buflen -= len;
5551                 *pages++ = newpage;
5552                 rc++;
5553         } while (buflen != 0);
5554
5555         return rc;
5556
5557 unwind:
5558         for(; rc > 0; rc--)
5559                 __free_page(spages[rc-1]);
5560         return -ENOMEM;
5561 }
5562
5563 struct nfs4_cached_acl {
5564         int cached;
5565         size_t len;
5566         char data[];
5567 };
5568
5569 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5570 {
5571         struct nfs_inode *nfsi = NFS_I(inode);
5572
5573         spin_lock(&inode->i_lock);
5574         kfree(nfsi->nfs4_acl);
5575         nfsi->nfs4_acl = acl;
5576         spin_unlock(&inode->i_lock);
5577 }
5578
5579 static void nfs4_zap_acl_attr(struct inode *inode)
5580 {
5581         nfs4_set_cached_acl(inode, NULL);
5582 }
5583
5584 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5585 {
5586         struct nfs_inode *nfsi = NFS_I(inode);
5587         struct nfs4_cached_acl *acl;
5588         int ret = -ENOENT;
5589
5590         spin_lock(&inode->i_lock);
5591         acl = nfsi->nfs4_acl;
5592         if (acl == NULL)
5593                 goto out;
5594         if (buf == NULL) /* user is just asking for length */
5595                 goto out_len;
5596         if (acl->cached == 0)
5597                 goto out;
5598         ret = -ERANGE; /* see getxattr(2) man page */
5599         if (acl->len > buflen)
5600                 goto out;
5601         memcpy(buf, acl->data, acl->len);
5602 out_len:
5603         ret = acl->len;
5604 out:
5605         spin_unlock(&inode->i_lock);
5606         return ret;
5607 }
5608
5609 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5610 {
5611         struct nfs4_cached_acl *acl;
5612         size_t buflen = sizeof(*acl) + acl_len;
5613
5614         if (buflen <= PAGE_SIZE) {
5615                 acl = kmalloc(buflen, GFP_KERNEL);
5616                 if (acl == NULL)
5617                         goto out;
5618                 acl->cached = 1;
5619                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5620         } else {
5621                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5622                 if (acl == NULL)
5623                         goto out;
5624                 acl->cached = 0;
5625         }
5626         acl->len = acl_len;
5627 out:
5628         nfs4_set_cached_acl(inode, acl);
5629 }
5630
5631 /*
5632  * The getxattr API returns the required buffer length when called with a
5633  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5634  * the required buf.  On a NULL buf, we send a page of data to the server
5635  * guessing that the ACL request can be serviced by a page. If so, we cache
5636  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5637  * the cache. If not so, we throw away the page, and cache the required
5638  * length. The next getxattr call will then produce another round trip to
5639  * the server, this time with the input buf of the required size.
5640  */
5641 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5642 {
5643         struct page **pages;
5644         struct nfs_getaclargs args = {
5645                 .fh = NFS_FH(inode),
5646                 .acl_len = buflen,
5647         };
5648         struct nfs_getaclres res = {
5649                 .acl_len = buflen,
5650         };
5651         struct rpc_message msg = {
5652                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5653                 .rpc_argp = &args,
5654                 .rpc_resp = &res,
5655         };
5656         unsigned int npages;
5657         int ret = -ENOMEM, i;
5658         struct nfs_server *server = NFS_SERVER(inode);
5659
5660         if (buflen == 0)
5661                 buflen = server->rsize;
5662
5663         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5664         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5665         if (!pages)
5666                 return -ENOMEM;
5667
5668         args.acl_pages = pages;
5669
5670         for (i = 0; i < npages; i++) {
5671                 pages[i] = alloc_page(GFP_KERNEL);
5672                 if (!pages[i])
5673                         goto out_free;
5674         }
5675
5676         /* for decoding across pages */
5677         res.acl_scratch = alloc_page(GFP_KERNEL);
5678         if (!res.acl_scratch)
5679                 goto out_free;
5680
5681         args.acl_len = npages * PAGE_SIZE;
5682
5683         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5684                 __func__, buf, buflen, npages, args.acl_len);
5685         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5686                              &msg, &args.seq_args, &res.seq_res, 0);
5687         if (ret)
5688                 goto out_free;
5689
5690         /* Handle the case where the passed-in buffer is too short */
5691         if (res.acl_flags & NFS4_ACL_TRUNC) {
5692                 /* Did the user only issue a request for the acl length? */
5693                 if (buf == NULL)
5694                         goto out_ok;
5695                 ret = -ERANGE;
5696                 goto out_free;
5697         }
5698         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5699         if (buf) {
5700                 if (res.acl_len > buflen) {
5701                         ret = -ERANGE;
5702                         goto out_free;
5703                 }
5704                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5705         }
5706 out_ok:
5707         ret = res.acl_len;
5708 out_free:
5709         for (i = 0; i < npages; i++)
5710                 if (pages[i])
5711                         __free_page(pages[i]);
5712         if (res.acl_scratch)
5713                 __free_page(res.acl_scratch);
5714         kfree(pages);
5715         return ret;
5716 }
5717
5718 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5719 {
5720         struct nfs4_exception exception = {
5721                 .interruptible = true,
5722         };
5723         ssize_t ret;
5724         do {
5725                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5726                 trace_nfs4_get_acl(inode, ret);
5727                 if (ret >= 0)
5728                         break;
5729                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5730         } while (exception.retry);
5731         return ret;
5732 }
5733
5734 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5735 {
5736         struct nfs_server *server = NFS_SERVER(inode);
5737         int ret;
5738
5739         if (!nfs4_server_supports_acls(server))
5740                 return -EOPNOTSUPP;
5741         ret = nfs_revalidate_inode(server, inode);
5742         if (ret < 0)
5743                 return ret;
5744         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5745                 nfs_zap_acl_cache(inode);
5746         ret = nfs4_read_cached_acl(inode, buf, buflen);
5747         if (ret != -ENOENT)
5748                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5749                  * but no cached acl data, just the acl length */
5750                 return ret;
5751         return nfs4_get_acl_uncached(inode, buf, buflen);
5752 }
5753
5754 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5755 {
5756         struct nfs_server *server = NFS_SERVER(inode);
5757         struct page *pages[NFS4ACL_MAXPAGES];
5758         struct nfs_setaclargs arg = {
5759                 .fh             = NFS_FH(inode),
5760                 .acl_pages      = pages,
5761                 .acl_len        = buflen,
5762         };
5763         struct nfs_setaclres res;
5764         struct rpc_message msg = {
5765                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5766                 .rpc_argp       = &arg,
5767                 .rpc_resp       = &res,
5768         };
5769         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5770         int ret, i;
5771
5772         if (!nfs4_server_supports_acls(server))
5773                 return -EOPNOTSUPP;
5774         if (npages > ARRAY_SIZE(pages))
5775                 return -ERANGE;
5776         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5777         if (i < 0)
5778                 return i;
5779         nfs4_inode_make_writeable(inode);
5780         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5781
5782         /*
5783          * Free each page after tx, so the only ref left is
5784          * held by the network stack
5785          */
5786         for (; i > 0; i--)
5787                 put_page(pages[i-1]);
5788
5789         /*
5790          * Acl update can result in inode attribute update.
5791          * so mark the attribute cache invalid.
5792          */
5793         spin_lock(&inode->i_lock);
5794         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5795                 | NFS_INO_INVALID_CTIME
5796                 | NFS_INO_REVAL_FORCED;
5797         spin_unlock(&inode->i_lock);
5798         nfs_access_zap_cache(inode);
5799         nfs_zap_acl_cache(inode);
5800         return ret;
5801 }
5802
5803 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5804 {
5805         struct nfs4_exception exception = { };
5806         int err;
5807         do {
5808                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5809                 trace_nfs4_set_acl(inode, err);
5810                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5811                                 &exception);
5812         } while (exception.retry);
5813         return err;
5814 }
5815
5816 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5817 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5818                                         size_t buflen)
5819 {
5820         struct nfs_server *server = NFS_SERVER(inode);
5821         struct nfs_fattr fattr;
5822         struct nfs4_label label = {0, 0, buflen, buf};
5823
5824         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5825         struct nfs4_getattr_arg arg = {
5826                 .fh             = NFS_FH(inode),
5827                 .bitmask        = bitmask,
5828         };
5829         struct nfs4_getattr_res res = {
5830                 .fattr          = &fattr,
5831                 .label          = &label,
5832                 .server         = server,
5833         };
5834         struct rpc_message msg = {
5835                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5836                 .rpc_argp       = &arg,
5837                 .rpc_resp       = &res,
5838         };
5839         int ret;
5840
5841         nfs_fattr_init(&fattr);
5842
5843         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5844         if (ret)
5845                 return ret;
5846         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5847                 return -ENOENT;
5848         if (buflen < label.len)
5849                 return -ERANGE;
5850         return 0;
5851 }
5852
5853 static int nfs4_get_security_label(struct inode *inode, void *buf,
5854                                         size_t buflen)
5855 {
5856         struct nfs4_exception exception = {
5857                 .interruptible = true,
5858         };
5859         int err;
5860
5861         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5862                 return -EOPNOTSUPP;
5863
5864         do {
5865                 err = _nfs4_get_security_label(inode, buf, buflen);
5866                 trace_nfs4_get_security_label(inode, err);
5867                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5868                                 &exception);
5869         } while (exception.retry);
5870         return err;
5871 }
5872
5873 static int _nfs4_do_set_security_label(struct inode *inode,
5874                 struct nfs4_label *ilabel,
5875                 struct nfs_fattr *fattr,
5876                 struct nfs4_label *olabel)
5877 {
5878
5879         struct iattr sattr = {0};
5880         struct nfs_server *server = NFS_SERVER(inode);
5881         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5882         struct nfs_setattrargs arg = {
5883                 .fh             = NFS_FH(inode),
5884                 .iap            = &sattr,
5885                 .server         = server,
5886                 .bitmask        = bitmask,
5887                 .label          = ilabel,
5888         };
5889         struct nfs_setattrres res = {
5890                 .fattr          = fattr,
5891                 .label          = olabel,
5892                 .server         = server,
5893         };
5894         struct rpc_message msg = {
5895                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5896                 .rpc_argp       = &arg,
5897                 .rpc_resp       = &res,
5898         };
5899         int status;
5900
5901         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5902
5903         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5904         if (status)
5905                 dprintk("%s failed: %d\n", __func__, status);
5906
5907         return status;
5908 }
5909
5910 static int nfs4_do_set_security_label(struct inode *inode,
5911                 struct nfs4_label *ilabel,
5912                 struct nfs_fattr *fattr,
5913                 struct nfs4_label *olabel)
5914 {
5915         struct nfs4_exception exception = { };
5916         int err;
5917
5918         do {
5919                 err = _nfs4_do_set_security_label(inode, ilabel,
5920                                 fattr, olabel);
5921                 trace_nfs4_set_security_label(inode, err);
5922                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5923                                 &exception);
5924         } while (exception.retry);
5925         return err;
5926 }
5927
5928 static int
5929 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5930 {
5931         struct nfs4_label ilabel, *olabel = NULL;
5932         struct nfs_fattr fattr;
5933         int status;
5934
5935         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5936                 return -EOPNOTSUPP;
5937
5938         nfs_fattr_init(&fattr);
5939
5940         ilabel.pi = 0;
5941         ilabel.lfs = 0;
5942         ilabel.label = (char *)buf;
5943         ilabel.len = buflen;
5944
5945         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5946         if (IS_ERR(olabel)) {
5947                 status = -PTR_ERR(olabel);
5948                 goto out;
5949         }
5950
5951         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5952         if (status == 0)
5953                 nfs_setsecurity(inode, &fattr, olabel);
5954
5955         nfs4_label_free(olabel);
5956 out:
5957         return status;
5958 }
5959 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5960
5961
5962 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5963                                     nfs4_verifier *bootverf)
5964 {
5965         __be32 verf[2];
5966
5967         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5968                 /* An impossible timestamp guarantees this value
5969                  * will never match a generated boot time. */
5970                 verf[0] = cpu_to_be32(U32_MAX);
5971                 verf[1] = cpu_to_be32(U32_MAX);
5972         } else {
5973                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5974                 u64 ns = ktime_to_ns(nn->boot_time);
5975
5976                 verf[0] = cpu_to_be32(ns >> 32);
5977                 verf[1] = cpu_to_be32(ns);
5978         }
5979         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5980 }
5981
5982 static int
5983 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5984 {
5985         size_t len;
5986         char *str;
5987
5988         if (clp->cl_owner_id != NULL)
5989                 return 0;
5990
5991         rcu_read_lock();
5992         len = 14 +
5993                 strlen(clp->cl_rpcclient->cl_nodename) +
5994                 1 +
5995                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5996                 1;
5997         rcu_read_unlock();
5998         if (nfs4_client_id_uniquifier[0] != '\0')
5999                 len += strlen(nfs4_client_id_uniquifier) + 1;
6000         if (len > NFS4_OPAQUE_LIMIT + 1)
6001                 return -EINVAL;
6002
6003         /*
6004          * Since this string is allocated at mount time, and held until the
6005          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6006          * about a memory-reclaim deadlock.
6007          */
6008         str = kmalloc(len, GFP_KERNEL);
6009         if (!str)
6010                 return -ENOMEM;
6011
6012         rcu_read_lock();
6013         if (nfs4_client_id_uniquifier[0] != '\0')
6014                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6015                           clp->cl_rpcclient->cl_nodename,
6016                           nfs4_client_id_uniquifier,
6017                           rpc_peeraddr2str(clp->cl_rpcclient,
6018                                            RPC_DISPLAY_ADDR));
6019         else
6020                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6021                           clp->cl_rpcclient->cl_nodename,
6022                           rpc_peeraddr2str(clp->cl_rpcclient,
6023                                            RPC_DISPLAY_ADDR));
6024         rcu_read_unlock();
6025
6026         clp->cl_owner_id = str;
6027         return 0;
6028 }
6029
6030 static int
6031 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
6032 {
6033         size_t len;
6034         char *str;
6035
6036         len = 10 + 10 + 1 + 10 + 1 +
6037                 strlen(nfs4_client_id_uniquifier) + 1 +
6038                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6039
6040         if (len > NFS4_OPAQUE_LIMIT + 1)
6041                 return -EINVAL;
6042
6043         /*
6044          * Since this string is allocated at mount time, and held until the
6045          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6046          * about a memory-reclaim deadlock.
6047          */
6048         str = kmalloc(len, GFP_KERNEL);
6049         if (!str)
6050                 return -ENOMEM;
6051
6052         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6053                         clp->rpc_ops->version, clp->cl_minorversion,
6054                         nfs4_client_id_uniquifier,
6055                         clp->cl_rpcclient->cl_nodename);
6056         clp->cl_owner_id = str;
6057         return 0;
6058 }
6059
6060 static int
6061 nfs4_init_uniform_client_string(struct nfs_client *clp)
6062 {
6063         size_t len;
6064         char *str;
6065
6066         if (clp->cl_owner_id != NULL)
6067                 return 0;
6068
6069         if (nfs4_client_id_uniquifier[0] != '\0')
6070                 return nfs4_init_uniquifier_client_string(clp);
6071
6072         len = 10 + 10 + 1 + 10 + 1 +
6073                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6074
6075         if (len > NFS4_OPAQUE_LIMIT + 1)
6076                 return -EINVAL;
6077
6078         /*
6079          * Since this string is allocated at mount time, and held until the
6080          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6081          * about a memory-reclaim deadlock.
6082          */
6083         str = kmalloc(len, GFP_KERNEL);
6084         if (!str)
6085                 return -ENOMEM;
6086
6087         scnprintf(str, len, "Linux NFSv%u.%u %s",
6088                         clp->rpc_ops->version, clp->cl_minorversion,
6089                         clp->cl_rpcclient->cl_nodename);
6090         clp->cl_owner_id = str;
6091         return 0;
6092 }
6093
6094 /*
6095  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6096  * services.  Advertise one based on the address family of the
6097  * clientaddr.
6098  */
6099 static unsigned int
6100 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6101 {
6102         if (strchr(clp->cl_ipaddr, ':') != NULL)
6103                 return scnprintf(buf, len, "tcp6");
6104         else
6105                 return scnprintf(buf, len, "tcp");
6106 }
6107
6108 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6109 {
6110         struct nfs4_setclientid *sc = calldata;
6111
6112         if (task->tk_status == 0)
6113                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6114 }
6115
6116 static const struct rpc_call_ops nfs4_setclientid_ops = {
6117         .rpc_call_done = nfs4_setclientid_done,
6118 };
6119
6120 /**
6121  * nfs4_proc_setclientid - Negotiate client ID
6122  * @clp: state data structure
6123  * @program: RPC program for NFSv4 callback service
6124  * @port: IP port number for NFS4 callback service
6125  * @cred: credential to use for this call
6126  * @res: where to place the result
6127  *
6128  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6129  */
6130 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6131                 unsigned short port, const struct cred *cred,
6132                 struct nfs4_setclientid_res *res)
6133 {
6134         nfs4_verifier sc_verifier;
6135         struct nfs4_setclientid setclientid = {
6136                 .sc_verifier = &sc_verifier,
6137                 .sc_prog = program,
6138                 .sc_clnt = clp,
6139         };
6140         struct rpc_message msg = {
6141                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6142                 .rpc_argp = &setclientid,
6143                 .rpc_resp = res,
6144                 .rpc_cred = cred,
6145         };
6146         struct rpc_task_setup task_setup_data = {
6147                 .rpc_client = clp->cl_rpcclient,
6148                 .rpc_message = &msg,
6149                 .callback_ops = &nfs4_setclientid_ops,
6150                 .callback_data = &setclientid,
6151                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6152         };
6153         unsigned long now = jiffies;
6154         int status;
6155
6156         /* nfs_client_id4 */
6157         nfs4_init_boot_verifier(clp, &sc_verifier);
6158
6159         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6160                 status = nfs4_init_uniform_client_string(clp);
6161         else
6162                 status = nfs4_init_nonuniform_client_string(clp);
6163
6164         if (status)
6165                 goto out;
6166
6167         /* cb_client4 */
6168         setclientid.sc_netid_len =
6169                                 nfs4_init_callback_netid(clp,
6170                                                 setclientid.sc_netid,
6171                                                 sizeof(setclientid.sc_netid));
6172         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6173                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6174                                 clp->cl_ipaddr, port >> 8, port & 255);
6175
6176         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6177                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6178                 clp->cl_owner_id);
6179
6180         status = nfs4_call_sync_custom(&task_setup_data);
6181         if (setclientid.sc_cred) {
6182                 kfree(clp->cl_acceptor);
6183                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6184                 put_rpccred(setclientid.sc_cred);
6185         }
6186
6187         if (status == 0)
6188                 do_renew_lease(clp, now);
6189 out:
6190         trace_nfs4_setclientid(clp, status);
6191         dprintk("NFS reply setclientid: %d\n", status);
6192         return status;
6193 }
6194
6195 /**
6196  * nfs4_proc_setclientid_confirm - Confirm client ID
6197  * @clp: state data structure
6198  * @arg: result of a previous SETCLIENTID
6199  * @cred: credential to use for this call
6200  *
6201  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6202  */
6203 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6204                 struct nfs4_setclientid_res *arg,
6205                 const struct cred *cred)
6206 {
6207         struct rpc_message msg = {
6208                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6209                 .rpc_argp = arg,
6210                 .rpc_cred = cred,
6211         };
6212         int status;
6213
6214         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6215                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6216                 clp->cl_clientid);
6217         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6218                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6219         trace_nfs4_setclientid_confirm(clp, status);
6220         dprintk("NFS reply setclientid_confirm: %d\n", status);
6221         return status;
6222 }
6223
6224 struct nfs4_delegreturndata {
6225         struct nfs4_delegreturnargs args;
6226         struct nfs4_delegreturnres res;
6227         struct nfs_fh fh;
6228         nfs4_stateid stateid;
6229         unsigned long timestamp;
6230         struct {
6231                 struct nfs4_layoutreturn_args arg;
6232                 struct nfs4_layoutreturn_res res;
6233                 struct nfs4_xdr_opaque_data ld_private;
6234                 u32 roc_barrier;
6235                 bool roc;
6236         } lr;
6237         struct nfs_fattr fattr;
6238         int rpc_status;
6239         struct inode *inode;
6240 };
6241
6242 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6243 {
6244         struct nfs4_delegreturndata *data = calldata;
6245         struct nfs4_exception exception = {
6246                 .inode = data->inode,
6247                 .stateid = &data->stateid,
6248         };
6249
6250         if (!nfs4_sequence_done(task, &data->res.seq_res))
6251                 return;
6252
6253         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6254
6255         /* Handle Layoutreturn errors */
6256         if (pnfs_roc_done(task, data->inode,
6257                                 &data->args.lr_args,
6258                                 &data->res.lr_res,
6259                                 &data->res.lr_ret) == -EAGAIN)
6260                 goto out_restart;
6261
6262         switch (task->tk_status) {
6263         case 0:
6264                 renew_lease(data->res.server, data->timestamp);
6265                 break;
6266         case -NFS4ERR_ADMIN_REVOKED:
6267         case -NFS4ERR_DELEG_REVOKED:
6268         case -NFS4ERR_EXPIRED:
6269                 nfs4_free_revoked_stateid(data->res.server,
6270                                 data->args.stateid,
6271                                 task->tk_msg.rpc_cred);
6272                 /* Fallthrough */
6273         case -NFS4ERR_BAD_STATEID:
6274         case -NFS4ERR_STALE_STATEID:
6275         case -ETIMEDOUT:
6276                 task->tk_status = 0;
6277                 break;
6278         case -NFS4ERR_OLD_STATEID:
6279                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6280                         nfs4_stateid_seqid_inc(&data->stateid);
6281                 if (data->args.bitmask) {
6282                         data->args.bitmask = NULL;
6283                         data->res.fattr = NULL;
6284                 }
6285                 goto out_restart;
6286         case -NFS4ERR_ACCESS:
6287                 if (data->args.bitmask) {
6288                         data->args.bitmask = NULL;
6289                         data->res.fattr = NULL;
6290                         goto out_restart;
6291                 }
6292                 /* Fallthrough */
6293         default:
6294                 task->tk_status = nfs4_async_handle_exception(task,
6295                                 data->res.server, task->tk_status,
6296                                 &exception);
6297                 if (exception.retry)
6298                         goto out_restart;
6299         }
6300         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6301         data->rpc_status = task->tk_status;
6302         return;
6303 out_restart:
6304         task->tk_status = 0;
6305         rpc_restart_call_prepare(task);
6306 }
6307
6308 static void nfs4_delegreturn_release(void *calldata)
6309 {
6310         struct nfs4_delegreturndata *data = calldata;
6311         struct inode *inode = data->inode;
6312
6313         if (inode) {
6314                 if (data->lr.roc)
6315                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6316                                         data->res.lr_ret);
6317                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6318                 nfs_iput_and_deactive(inode);
6319         }
6320         kfree(calldata);
6321 }
6322
6323 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6324 {
6325         struct nfs4_delegreturndata *d_data;
6326         struct pnfs_layout_hdr *lo;
6327
6328         d_data = (struct nfs4_delegreturndata *)data;
6329
6330         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6331                 nfs4_sequence_done(task, &d_data->res.seq_res);
6332                 return;
6333         }
6334
6335         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6336         if (lo && !pnfs_layout_is_valid(lo)) {
6337                 d_data->args.lr_args = NULL;
6338                 d_data->res.lr_res = NULL;
6339         }
6340
6341         nfs4_setup_sequence(d_data->res.server->nfs_client,
6342                         &d_data->args.seq_args,
6343                         &d_data->res.seq_res,
6344                         task);
6345 }
6346
6347 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6348         .rpc_call_prepare = nfs4_delegreturn_prepare,
6349         .rpc_call_done = nfs4_delegreturn_done,
6350         .rpc_release = nfs4_delegreturn_release,
6351 };
6352
6353 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6354 {
6355         struct nfs4_delegreturndata *data;
6356         struct nfs_server *server = NFS_SERVER(inode);
6357         struct rpc_task *task;
6358         struct rpc_message msg = {
6359                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6360                 .rpc_cred = cred,
6361         };
6362         struct rpc_task_setup task_setup_data = {
6363                 .rpc_client = server->client,
6364                 .rpc_message = &msg,
6365                 .callback_ops = &nfs4_delegreturn_ops,
6366                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6367         };
6368         int status = 0;
6369
6370         data = kzalloc(sizeof(*data), GFP_NOFS);
6371         if (data == NULL)
6372                 return -ENOMEM;
6373         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6374
6375         nfs4_state_protect(server->nfs_client,
6376                         NFS_SP4_MACH_CRED_CLEANUP,
6377                         &task_setup_data.rpc_client, &msg);
6378
6379         data->args.fhandle = &data->fh;
6380         data->args.stateid = &data->stateid;
6381         data->args.bitmask = server->cache_consistency_bitmask;
6382         nfs_copy_fh(&data->fh, NFS_FH(inode));
6383         nfs4_stateid_copy(&data->stateid, stateid);
6384         data->res.fattr = &data->fattr;
6385         data->res.server = server;
6386         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6387         data->lr.arg.ld_private = &data->lr.ld_private;
6388         nfs_fattr_init(data->res.fattr);
6389         data->timestamp = jiffies;
6390         data->rpc_status = 0;
6391         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6392         data->inode = nfs_igrab_and_active(inode);
6393         if (data->inode) {
6394                 if (data->lr.roc) {
6395                         data->args.lr_args = &data->lr.arg;
6396                         data->res.lr_res = &data->lr.res;
6397                 }
6398         } else if (data->lr.roc) {
6399                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6400                 data->lr.roc = false;
6401         }
6402
6403         task_setup_data.callback_data = data;
6404         msg.rpc_argp = &data->args;
6405         msg.rpc_resp = &data->res;
6406         task = rpc_run_task(&task_setup_data);
6407         if (IS_ERR(task))
6408                 return PTR_ERR(task);
6409         if (!issync)
6410                 goto out;
6411         status = rpc_wait_for_completion_task(task);
6412         if (status != 0)
6413                 goto out;
6414         status = data->rpc_status;
6415 out:
6416         rpc_put_task(task);
6417         return status;
6418 }
6419
6420 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6421 {
6422         struct nfs_server *server = NFS_SERVER(inode);
6423         struct nfs4_exception exception = { };
6424         int err;
6425         do {
6426                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6427                 trace_nfs4_delegreturn(inode, stateid, err);
6428                 switch (err) {
6429                         case -NFS4ERR_STALE_STATEID:
6430                         case -NFS4ERR_EXPIRED:
6431                         case 0:
6432                                 return 0;
6433                 }
6434                 err = nfs4_handle_exception(server, err, &exception);
6435         } while (exception.retry);
6436         return err;
6437 }
6438
6439 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6440 {
6441         struct inode *inode = state->inode;
6442         struct nfs_server *server = NFS_SERVER(inode);
6443         struct nfs_client *clp = server->nfs_client;
6444         struct nfs_lockt_args arg = {
6445                 .fh = NFS_FH(inode),
6446                 .fl = request,
6447         };
6448         struct nfs_lockt_res res = {
6449                 .denied = request,
6450         };
6451         struct rpc_message msg = {
6452                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6453                 .rpc_argp       = &arg,
6454                 .rpc_resp       = &res,
6455                 .rpc_cred       = state->owner->so_cred,
6456         };
6457         struct nfs4_lock_state *lsp;
6458         int status;
6459
6460         arg.lock_owner.clientid = clp->cl_clientid;
6461         status = nfs4_set_lock_state(state, request);
6462         if (status != 0)
6463                 goto out;
6464         lsp = request->fl_u.nfs4_fl.owner;
6465         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6466         arg.lock_owner.s_dev = server->s_dev;
6467         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6468         switch (status) {
6469                 case 0:
6470                         request->fl_type = F_UNLCK;
6471                         break;
6472                 case -NFS4ERR_DENIED:
6473                         status = 0;
6474         }
6475         request->fl_ops->fl_release_private(request);
6476         request->fl_ops = NULL;
6477 out:
6478         return status;
6479 }
6480
6481 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6482 {
6483         struct nfs4_exception exception = {
6484                 .interruptible = true,
6485         };
6486         int err;
6487
6488         do {
6489                 err = _nfs4_proc_getlk(state, cmd, request);
6490                 trace_nfs4_get_lock(request, state, cmd, err);
6491                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6492                                 &exception);
6493         } while (exception.retry);
6494         return err;
6495 }
6496
6497 /*
6498  * Update the seqid of a lock stateid after receiving
6499  * NFS4ERR_OLD_STATEID
6500  */
6501 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6502                 struct nfs4_lock_state *lsp)
6503 {
6504         struct nfs4_state *state = lsp->ls_state;
6505         bool ret = false;
6506
6507         spin_lock(&state->state_lock);
6508         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6509                 goto out;
6510         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6511                 nfs4_stateid_seqid_inc(dst);
6512         else
6513                 dst->seqid = lsp->ls_stateid.seqid;
6514         ret = true;
6515 out:
6516         spin_unlock(&state->state_lock);
6517         return ret;
6518 }
6519
6520 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6521                 struct nfs4_lock_state *lsp)
6522 {
6523         struct nfs4_state *state = lsp->ls_state;
6524         bool ret;
6525
6526         spin_lock(&state->state_lock);
6527         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6528         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6529         spin_unlock(&state->state_lock);
6530         return ret;
6531 }
6532
6533 struct nfs4_unlockdata {
6534         struct nfs_locku_args arg;
6535         struct nfs_locku_res res;
6536         struct nfs4_lock_state *lsp;
6537         struct nfs_open_context *ctx;
6538         struct nfs_lock_context *l_ctx;
6539         struct file_lock fl;
6540         struct nfs_server *server;
6541         unsigned long timestamp;
6542 };
6543
6544 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6545                 struct nfs_open_context *ctx,
6546                 struct nfs4_lock_state *lsp,
6547                 struct nfs_seqid *seqid)
6548 {
6549         struct nfs4_unlockdata *p;
6550         struct nfs4_state *state = lsp->ls_state;
6551         struct inode *inode = state->inode;
6552
6553         p = kzalloc(sizeof(*p), GFP_NOFS);
6554         if (p == NULL)
6555                 return NULL;
6556         p->arg.fh = NFS_FH(inode);
6557         p->arg.fl = &p->fl;
6558         p->arg.seqid = seqid;
6559         p->res.seqid = seqid;
6560         p->lsp = lsp;
6561         /* Ensure we don't close file until we're done freeing locks! */
6562         p->ctx = get_nfs_open_context(ctx);
6563         p->l_ctx = nfs_get_lock_context(ctx);
6564         locks_init_lock(&p->fl);
6565         locks_copy_lock(&p->fl, fl);
6566         p->server = NFS_SERVER(inode);
6567         spin_lock(&state->state_lock);
6568         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6569         spin_unlock(&state->state_lock);
6570         return p;
6571 }
6572
6573 static void nfs4_locku_release_calldata(void *data)
6574 {
6575         struct nfs4_unlockdata *calldata = data;
6576         nfs_free_seqid(calldata->arg.seqid);
6577         nfs4_put_lock_state(calldata->lsp);
6578         nfs_put_lock_context(calldata->l_ctx);
6579         put_nfs_open_context(calldata->ctx);
6580         kfree(calldata);
6581 }
6582
6583 static void nfs4_locku_done(struct rpc_task *task, void *data)
6584 {
6585         struct nfs4_unlockdata *calldata = data;
6586         struct nfs4_exception exception = {
6587                 .inode = calldata->lsp->ls_state->inode,
6588                 .stateid = &calldata->arg.stateid,
6589         };
6590
6591         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6592                 return;
6593         switch (task->tk_status) {
6594                 case 0:
6595                         renew_lease(calldata->server, calldata->timestamp);
6596                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6597                         if (nfs4_update_lock_stateid(calldata->lsp,
6598                                         &calldata->res.stateid))
6599                                 break;
6600                         /* Fall through */
6601                 case -NFS4ERR_ADMIN_REVOKED:
6602                 case -NFS4ERR_EXPIRED:
6603                         nfs4_free_revoked_stateid(calldata->server,
6604                                         &calldata->arg.stateid,
6605                                         task->tk_msg.rpc_cred);
6606                         /* Fall through */
6607                 case -NFS4ERR_BAD_STATEID:
6608                 case -NFS4ERR_STALE_STATEID:
6609                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6610                                                 calldata->lsp))
6611                                 rpc_restart_call_prepare(task);
6612                         break;
6613                 case -NFS4ERR_OLD_STATEID:
6614                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6615                                                 calldata->lsp))
6616                                 rpc_restart_call_prepare(task);
6617                         break;
6618                 default:
6619                         task->tk_status = nfs4_async_handle_exception(task,
6620                                         calldata->server, task->tk_status,
6621                                         &exception);
6622                         if (exception.retry)
6623                                 rpc_restart_call_prepare(task);
6624         }
6625         nfs_release_seqid(calldata->arg.seqid);
6626 }
6627
6628 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6629 {
6630         struct nfs4_unlockdata *calldata = data;
6631
6632         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6633                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6634                 return;
6635
6636         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6637                 goto out_wait;
6638         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6639                 /* Note: exit _without_ running nfs4_locku_done */
6640                 goto out_no_action;
6641         }
6642         calldata->timestamp = jiffies;
6643         if (nfs4_setup_sequence(calldata->server->nfs_client,
6644                                 &calldata->arg.seq_args,
6645                                 &calldata->res.seq_res,
6646                                 task) != 0)
6647                 nfs_release_seqid(calldata->arg.seqid);
6648         return;
6649 out_no_action:
6650         task->tk_action = NULL;
6651 out_wait:
6652         nfs4_sequence_done(task, &calldata->res.seq_res);
6653 }
6654
6655 static const struct rpc_call_ops nfs4_locku_ops = {
6656         .rpc_call_prepare = nfs4_locku_prepare,
6657         .rpc_call_done = nfs4_locku_done,
6658         .rpc_release = nfs4_locku_release_calldata,
6659 };
6660
6661 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6662                 struct nfs_open_context *ctx,
6663                 struct nfs4_lock_state *lsp,
6664                 struct nfs_seqid *seqid)
6665 {
6666         struct nfs4_unlockdata *data;
6667         struct rpc_message msg = {
6668                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6669                 .rpc_cred = ctx->cred,
6670         };
6671         struct rpc_task_setup task_setup_data = {
6672                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6673                 .rpc_message = &msg,
6674                 .callback_ops = &nfs4_locku_ops,
6675                 .workqueue = nfsiod_workqueue,
6676                 .flags = RPC_TASK_ASYNC,
6677         };
6678
6679         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6680                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6681
6682         /* Ensure this is an unlock - when canceling a lock, the
6683          * canceled lock is passed in, and it won't be an unlock.
6684          */
6685         fl->fl_type = F_UNLCK;
6686         if (fl->fl_flags & FL_CLOSE)
6687                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6688
6689         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6690         if (data == NULL) {
6691                 nfs_free_seqid(seqid);
6692                 return ERR_PTR(-ENOMEM);
6693         }
6694
6695         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6696         msg.rpc_argp = &data->arg;
6697         msg.rpc_resp = &data->res;
6698         task_setup_data.callback_data = data;
6699         return rpc_run_task(&task_setup_data);
6700 }
6701
6702 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6703 {
6704         struct inode *inode = state->inode;
6705         struct nfs4_state_owner *sp = state->owner;
6706         struct nfs_inode *nfsi = NFS_I(inode);
6707         struct nfs_seqid *seqid;
6708         struct nfs4_lock_state *lsp;
6709         struct rpc_task *task;
6710         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6711         int status = 0;
6712         unsigned char fl_flags = request->fl_flags;
6713
6714         status = nfs4_set_lock_state(state, request);
6715         /* Unlock _before_ we do the RPC call */
6716         request->fl_flags |= FL_EXISTS;
6717         /* Exclude nfs_delegation_claim_locks() */
6718         mutex_lock(&sp->so_delegreturn_mutex);
6719         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6720         down_read(&nfsi->rwsem);
6721         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6722                 up_read(&nfsi->rwsem);
6723                 mutex_unlock(&sp->so_delegreturn_mutex);
6724                 goto out;
6725         }
6726         up_read(&nfsi->rwsem);
6727         mutex_unlock(&sp->so_delegreturn_mutex);
6728         if (status != 0)
6729                 goto out;
6730         /* Is this a delegated lock? */
6731         lsp = request->fl_u.nfs4_fl.owner;
6732         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6733                 goto out;
6734         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6735         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6736         status = -ENOMEM;
6737         if (IS_ERR(seqid))
6738                 goto out;
6739         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6740         status = PTR_ERR(task);
6741         if (IS_ERR(task))
6742                 goto out;
6743         status = rpc_wait_for_completion_task(task);
6744         rpc_put_task(task);
6745 out:
6746         request->fl_flags = fl_flags;
6747         trace_nfs4_unlock(request, state, F_SETLK, status);
6748         return status;
6749 }
6750
6751 struct nfs4_lockdata {
6752         struct nfs_lock_args arg;
6753         struct nfs_lock_res res;
6754         struct nfs4_lock_state *lsp;
6755         struct nfs_open_context *ctx;
6756         struct file_lock fl;
6757         unsigned long timestamp;
6758         int rpc_status;
6759         int cancelled;
6760         struct nfs_server *server;
6761 };
6762
6763 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6764                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6765                 gfp_t gfp_mask)
6766 {
6767         struct nfs4_lockdata *p;
6768         struct inode *inode = lsp->ls_state->inode;
6769         struct nfs_server *server = NFS_SERVER(inode);
6770         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6771
6772         p = kzalloc(sizeof(*p), gfp_mask);
6773         if (p == NULL)
6774                 return NULL;
6775
6776         p->arg.fh = NFS_FH(inode);
6777         p->arg.fl = &p->fl;
6778         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6779         if (IS_ERR(p->arg.open_seqid))
6780                 goto out_free;
6781         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6782         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6783         if (IS_ERR(p->arg.lock_seqid))
6784                 goto out_free_seqid;
6785         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6786         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6787         p->arg.lock_owner.s_dev = server->s_dev;
6788         p->res.lock_seqid = p->arg.lock_seqid;
6789         p->lsp = lsp;
6790         p->server = server;
6791         p->ctx = get_nfs_open_context(ctx);
6792         locks_init_lock(&p->fl);
6793         locks_copy_lock(&p->fl, fl);
6794         return p;
6795 out_free_seqid:
6796         nfs_free_seqid(p->arg.open_seqid);
6797 out_free:
6798         kfree(p);
6799         return NULL;
6800 }
6801
6802 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6803 {
6804         struct nfs4_lockdata *data = calldata;
6805         struct nfs4_state *state = data->lsp->ls_state;
6806
6807         dprintk("%s: begin!\n", __func__);
6808         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6809                 goto out_wait;
6810         /* Do we need to do an open_to_lock_owner? */
6811         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6812                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6813                         goto out_release_lock_seqid;
6814                 }
6815                 nfs4_stateid_copy(&data->arg.open_stateid,
6816                                 &state->open_stateid);
6817                 data->arg.new_lock_owner = 1;
6818                 data->res.open_seqid = data->arg.open_seqid;
6819         } else {
6820                 data->arg.new_lock_owner = 0;
6821                 nfs4_stateid_copy(&data->arg.lock_stateid,
6822                                 &data->lsp->ls_stateid);
6823         }
6824         if (!nfs4_valid_open_stateid(state)) {
6825                 data->rpc_status = -EBADF;
6826                 task->tk_action = NULL;
6827                 goto out_release_open_seqid;
6828         }
6829         data->timestamp = jiffies;
6830         if (nfs4_setup_sequence(data->server->nfs_client,
6831                                 &data->arg.seq_args,
6832                                 &data->res.seq_res,
6833                                 task) == 0)
6834                 return;
6835 out_release_open_seqid:
6836         nfs_release_seqid(data->arg.open_seqid);
6837 out_release_lock_seqid:
6838         nfs_release_seqid(data->arg.lock_seqid);
6839 out_wait:
6840         nfs4_sequence_done(task, &data->res.seq_res);
6841         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6842 }
6843
6844 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6845 {
6846         struct nfs4_lockdata *data = calldata;
6847         struct nfs4_lock_state *lsp = data->lsp;
6848
6849         dprintk("%s: begin!\n", __func__);
6850
6851         if (!nfs4_sequence_done(task, &data->res.seq_res))
6852                 return;
6853
6854         data->rpc_status = task->tk_status;
6855         switch (task->tk_status) {
6856         case 0:
6857                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6858                                 data->timestamp);
6859                 if (data->arg.new_lock && !data->cancelled) {
6860                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6861                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6862                                 goto out_restart;
6863                 }
6864                 if (data->arg.new_lock_owner != 0) {
6865                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6866                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6867                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6868                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6869                         goto out_restart;
6870                 break;
6871         case -NFS4ERR_BAD_STATEID:
6872         case -NFS4ERR_OLD_STATEID:
6873         case -NFS4ERR_STALE_STATEID:
6874         case -NFS4ERR_EXPIRED:
6875                 if (data->arg.new_lock_owner != 0) {
6876                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6877                                                 &lsp->ls_state->open_stateid))
6878                                 goto out_restart;
6879                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6880                                                 &lsp->ls_stateid))
6881                                 goto out_restart;
6882         }
6883 out_done:
6884         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6885         return;
6886 out_restart:
6887         if (!data->cancelled)
6888                 rpc_restart_call_prepare(task);
6889         goto out_done;
6890 }
6891
6892 static void nfs4_lock_release(void *calldata)
6893 {
6894         struct nfs4_lockdata *data = calldata;
6895
6896         dprintk("%s: begin!\n", __func__);
6897         nfs_free_seqid(data->arg.open_seqid);
6898         if (data->cancelled && data->rpc_status == 0) {
6899                 struct rpc_task *task;
6900                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6901                                 data->arg.lock_seqid);
6902                 if (!IS_ERR(task))
6903                         rpc_put_task_async(task);
6904                 dprintk("%s: cancelling lock!\n", __func__);
6905         } else
6906                 nfs_free_seqid(data->arg.lock_seqid);
6907         nfs4_put_lock_state(data->lsp);
6908         put_nfs_open_context(data->ctx);
6909         kfree(data);
6910         dprintk("%s: done!\n", __func__);
6911 }
6912
6913 static const struct rpc_call_ops nfs4_lock_ops = {
6914         .rpc_call_prepare = nfs4_lock_prepare,
6915         .rpc_call_done = nfs4_lock_done,
6916         .rpc_release = nfs4_lock_release,
6917 };
6918
6919 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6920 {
6921         switch (error) {
6922         case -NFS4ERR_ADMIN_REVOKED:
6923         case -NFS4ERR_EXPIRED:
6924         case -NFS4ERR_BAD_STATEID:
6925                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6926                 if (new_lock_owner != 0 ||
6927                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6928                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6929                 break;
6930         case -NFS4ERR_STALE_STATEID:
6931                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6932                 nfs4_schedule_lease_recovery(server->nfs_client);
6933         }
6934 }
6935
6936 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6937 {
6938         struct nfs4_lockdata *data;
6939         struct rpc_task *task;
6940         struct rpc_message msg = {
6941                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6942                 .rpc_cred = state->owner->so_cred,
6943         };
6944         struct rpc_task_setup task_setup_data = {
6945                 .rpc_client = NFS_CLIENT(state->inode),
6946                 .rpc_message = &msg,
6947                 .callback_ops = &nfs4_lock_ops,
6948                 .workqueue = nfsiod_workqueue,
6949                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
6950         };
6951         int ret;
6952
6953         dprintk("%s: begin!\n", __func__);
6954         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6955                         fl->fl_u.nfs4_fl.owner,
6956                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6957         if (data == NULL)
6958                 return -ENOMEM;
6959         if (IS_SETLKW(cmd))
6960                 data->arg.block = 1;
6961         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6962                                 recovery_type > NFS_LOCK_NEW);
6963         msg.rpc_argp = &data->arg;
6964         msg.rpc_resp = &data->res;
6965         task_setup_data.callback_data = data;
6966         if (recovery_type > NFS_LOCK_NEW) {
6967                 if (recovery_type == NFS_LOCK_RECLAIM)
6968                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6969         } else
6970                 data->arg.new_lock = 1;
6971         task = rpc_run_task(&task_setup_data);
6972         if (IS_ERR(task))
6973                 return PTR_ERR(task);
6974         ret = rpc_wait_for_completion_task(task);
6975         if (ret == 0) {
6976                 ret = data->rpc_status;
6977                 if (ret)
6978                         nfs4_handle_setlk_error(data->server, data->lsp,
6979                                         data->arg.new_lock_owner, ret);
6980         } else
6981                 data->cancelled = true;
6982         rpc_put_task(task);
6983         dprintk("%s: done, ret = %d!\n", __func__, ret);
6984         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6985         return ret;
6986 }
6987
6988 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6989 {
6990         struct nfs_server *server = NFS_SERVER(state->inode);
6991         struct nfs4_exception exception = {
6992                 .inode = state->inode,
6993         };
6994         int err;
6995
6996         do {
6997                 /* Cache the lock if possible... */
6998                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6999                         return 0;
7000                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7001                 if (err != -NFS4ERR_DELAY)
7002                         break;
7003                 nfs4_handle_exception(server, err, &exception);
7004         } while (exception.retry);
7005         return err;
7006 }
7007
7008 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7009 {
7010         struct nfs_server *server = NFS_SERVER(state->inode);
7011         struct nfs4_exception exception = {
7012                 .inode = state->inode,
7013         };
7014         int err;
7015
7016         err = nfs4_set_lock_state(state, request);
7017         if (err != 0)
7018                 return err;
7019         if (!recover_lost_locks) {
7020                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7021                 return 0;
7022         }
7023         do {
7024                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7025                         return 0;
7026                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7027                 switch (err) {
7028                 default:
7029                         goto out;
7030                 case -NFS4ERR_GRACE:
7031                 case -NFS4ERR_DELAY:
7032                         nfs4_handle_exception(server, err, &exception);
7033                         err = 0;
7034                 }
7035         } while (exception.retry);
7036 out:
7037         return err;
7038 }
7039
7040 #if defined(CONFIG_NFS_V4_1)
7041 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7042 {
7043         struct nfs4_lock_state *lsp;
7044         int status;
7045
7046         status = nfs4_set_lock_state(state, request);
7047         if (status != 0)
7048                 return status;
7049         lsp = request->fl_u.nfs4_fl.owner;
7050         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7051             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7052                 return 0;
7053         return nfs4_lock_expired(state, request);
7054 }
7055 #endif
7056
7057 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7058 {
7059         struct nfs_inode *nfsi = NFS_I(state->inode);
7060         struct nfs4_state_owner *sp = state->owner;
7061         unsigned char fl_flags = request->fl_flags;
7062         int status;
7063
7064         request->fl_flags |= FL_ACCESS;
7065         status = locks_lock_inode_wait(state->inode, request);
7066         if (status < 0)
7067                 goto out;
7068         mutex_lock(&sp->so_delegreturn_mutex);
7069         down_read(&nfsi->rwsem);
7070         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7071                 /* Yes: cache locks! */
7072                 /* ...but avoid races with delegation recall... */
7073                 request->fl_flags = fl_flags & ~FL_SLEEP;
7074                 status = locks_lock_inode_wait(state->inode, request);
7075                 up_read(&nfsi->rwsem);
7076                 mutex_unlock(&sp->so_delegreturn_mutex);
7077                 goto out;
7078         }
7079         up_read(&nfsi->rwsem);
7080         mutex_unlock(&sp->so_delegreturn_mutex);
7081         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7082 out:
7083         request->fl_flags = fl_flags;
7084         return status;
7085 }
7086
7087 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7088 {
7089         struct nfs4_exception exception = {
7090                 .state = state,
7091                 .inode = state->inode,
7092                 .interruptible = true,
7093         };
7094         int err;
7095
7096         do {
7097                 err = _nfs4_proc_setlk(state, cmd, request);
7098                 if (err == -NFS4ERR_DENIED)
7099                         err = -EAGAIN;
7100                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7101                                 err, &exception);
7102         } while (exception.retry);
7103         return err;
7104 }
7105
7106 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7107 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7108
7109 static int
7110 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7111                         struct file_lock *request)
7112 {
7113         int             status = -ERESTARTSYS;
7114         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7115
7116         while(!signalled()) {
7117                 status = nfs4_proc_setlk(state, cmd, request);
7118                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7119                         break;
7120                 freezable_schedule_timeout_interruptible(timeout);
7121                 timeout *= 2;
7122                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7123                 status = -ERESTARTSYS;
7124         }
7125         return status;
7126 }
7127
7128 #ifdef CONFIG_NFS_V4_1
7129 struct nfs4_lock_waiter {
7130         struct task_struct      *task;
7131         struct inode            *inode;
7132         struct nfs_lowner       *owner;
7133 };
7134
7135 static int
7136 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7137 {
7138         int ret;
7139         struct nfs4_lock_waiter *waiter = wait->private;
7140
7141         /* NULL key means to wake up everyone */
7142         if (key) {
7143                 struct cb_notify_lock_args      *cbnl = key;
7144                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7145                                                 *wowner = waiter->owner;
7146
7147                 /* Only wake if the callback was for the same owner. */
7148                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7149                         return 0;
7150
7151                 /* Make sure it's for the right inode */
7152                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7153                         return 0;
7154         }
7155
7156         /* override "private" so we can use default_wake_function */
7157         wait->private = waiter->task;
7158         ret = woken_wake_function(wait, mode, flags, key);
7159         if (ret)
7160                 list_del_init(&wait->entry);
7161         wait->private = waiter;
7162         return ret;
7163 }
7164
7165 static int
7166 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7167 {
7168         int status = -ERESTARTSYS;
7169         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7170         struct nfs_server *server = NFS_SERVER(state->inode);
7171         struct nfs_client *clp = server->nfs_client;
7172         wait_queue_head_t *q = &clp->cl_lock_waitq;
7173         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7174                                     .id = lsp->ls_seqid.owner_id,
7175                                     .s_dev = server->s_dev };
7176         struct nfs4_lock_waiter waiter = { .task  = current,
7177                                            .inode = state->inode,
7178                                            .owner = &owner};
7179         wait_queue_entry_t wait;
7180
7181         /* Don't bother with waitqueue if we don't expect a callback */
7182         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7183                 return nfs4_retry_setlk_simple(state, cmd, request);
7184
7185         init_wait(&wait);
7186         wait.private = &waiter;
7187         wait.func = nfs4_wake_lock_waiter;
7188
7189         while(!signalled()) {
7190                 add_wait_queue(q, &wait);
7191                 status = nfs4_proc_setlk(state, cmd, request);
7192                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7193                         finish_wait(q, &wait);
7194                         break;
7195                 }
7196
7197                 status = -ERESTARTSYS;
7198                 freezer_do_not_count();
7199                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7200                 freezer_count();
7201                 finish_wait(q, &wait);
7202         }
7203
7204         return status;
7205 }
7206 #else /* !CONFIG_NFS_V4_1 */
7207 static inline int
7208 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7209 {
7210         return nfs4_retry_setlk_simple(state, cmd, request);
7211 }
7212 #endif
7213
7214 static int
7215 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7216 {
7217         struct nfs_open_context *ctx;
7218         struct nfs4_state *state;
7219         int status;
7220
7221         /* verify open state */
7222         ctx = nfs_file_open_context(filp);
7223         state = ctx->state;
7224
7225         if (IS_GETLK(cmd)) {
7226                 if (state != NULL)
7227                         return nfs4_proc_getlk(state, F_GETLK, request);
7228                 return 0;
7229         }
7230
7231         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7232                 return -EINVAL;
7233
7234         if (request->fl_type == F_UNLCK) {
7235                 if (state != NULL)
7236                         return nfs4_proc_unlck(state, cmd, request);
7237                 return 0;
7238         }
7239
7240         if (state == NULL)
7241                 return -ENOLCK;
7242
7243         if ((request->fl_flags & FL_POSIX) &&
7244             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7245                 return -ENOLCK;
7246
7247         /*
7248          * Don't rely on the VFS having checked the file open mode,
7249          * since it won't do this for flock() locks.
7250          */
7251         switch (request->fl_type) {
7252         case F_RDLCK:
7253                 if (!(filp->f_mode & FMODE_READ))
7254                         return -EBADF;
7255                 break;
7256         case F_WRLCK:
7257                 if (!(filp->f_mode & FMODE_WRITE))
7258                         return -EBADF;
7259         }
7260
7261         status = nfs4_set_lock_state(state, request);
7262         if (status != 0)
7263                 return status;
7264
7265         return nfs4_retry_setlk(state, cmd, request);
7266 }
7267
7268 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7269 {
7270         struct nfs_server *server = NFS_SERVER(state->inode);
7271         int err;
7272
7273         err = nfs4_set_lock_state(state, fl);
7274         if (err != 0)
7275                 return err;
7276         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7277         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7278 }
7279
7280 struct nfs_release_lockowner_data {
7281         struct nfs4_lock_state *lsp;
7282         struct nfs_server *server;
7283         struct nfs_release_lockowner_args args;
7284         struct nfs_release_lockowner_res res;
7285         unsigned long timestamp;
7286 };
7287
7288 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7289 {
7290         struct nfs_release_lockowner_data *data = calldata;
7291         struct nfs_server *server = data->server;
7292         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7293                            &data->res.seq_res, task);
7294         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7295         data->timestamp = jiffies;
7296 }
7297
7298 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7299 {
7300         struct nfs_release_lockowner_data *data = calldata;
7301         struct nfs_server *server = data->server;
7302
7303         nfs40_sequence_done(task, &data->res.seq_res);
7304
7305         switch (task->tk_status) {
7306         case 0:
7307                 renew_lease(server, data->timestamp);
7308                 break;
7309         case -NFS4ERR_STALE_CLIENTID:
7310         case -NFS4ERR_EXPIRED:
7311                 nfs4_schedule_lease_recovery(server->nfs_client);
7312                 break;
7313         case -NFS4ERR_LEASE_MOVED:
7314         case -NFS4ERR_DELAY:
7315                 if (nfs4_async_handle_error(task, server,
7316                                             NULL, NULL) == -EAGAIN)
7317                         rpc_restart_call_prepare(task);
7318         }
7319 }
7320
7321 static void nfs4_release_lockowner_release(void *calldata)
7322 {
7323         struct nfs_release_lockowner_data *data = calldata;
7324         nfs4_free_lock_state(data->server, data->lsp);
7325         kfree(calldata);
7326 }
7327
7328 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7329         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7330         .rpc_call_done = nfs4_release_lockowner_done,
7331         .rpc_release = nfs4_release_lockowner_release,
7332 };
7333
7334 static void
7335 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7336 {
7337         struct nfs_release_lockowner_data *data;
7338         struct rpc_message msg = {
7339                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7340         };
7341
7342         if (server->nfs_client->cl_mvops->minor_version != 0)
7343                 return;
7344
7345         data = kmalloc(sizeof(*data), GFP_NOFS);
7346         if (!data)
7347                 return;
7348         data->lsp = lsp;
7349         data->server = server;
7350         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7351         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7352         data->args.lock_owner.s_dev = server->s_dev;
7353
7354         msg.rpc_argp = &data->args;
7355         msg.rpc_resp = &data->res;
7356         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7357         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7358 }
7359
7360 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7361
7362 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7363                                    struct dentry *unused, struct inode *inode,
7364                                    const char *key, const void *buf,
7365                                    size_t buflen, int flags)
7366 {
7367         return nfs4_proc_set_acl(inode, buf, buflen);
7368 }
7369
7370 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7371                                    struct dentry *unused, struct inode *inode,
7372                                    const char *key, void *buf, size_t buflen)
7373 {
7374         return nfs4_proc_get_acl(inode, buf, buflen);
7375 }
7376
7377 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7378 {
7379         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7380 }
7381
7382 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7383
7384 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7385                                      struct dentry *unused, struct inode *inode,
7386                                      const char *key, const void *buf,
7387                                      size_t buflen, int flags)
7388 {
7389         if (security_ismaclabel(key))
7390                 return nfs4_set_security_label(inode, buf, buflen);
7391
7392         return -EOPNOTSUPP;
7393 }
7394
7395 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7396                                      struct dentry *unused, struct inode *inode,
7397                                      const char *key, void *buf, size_t buflen)
7398 {
7399         if (security_ismaclabel(key))
7400                 return nfs4_get_security_label(inode, buf, buflen);
7401         return -EOPNOTSUPP;
7402 }
7403
7404 static ssize_t
7405 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7406 {
7407         int len = 0;
7408
7409         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7410                 len = security_inode_listsecurity(inode, list, list_len);
7411                 if (list_len && len > list_len)
7412                         return -ERANGE;
7413         }
7414         return len;
7415 }
7416
7417 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7418         .prefix = XATTR_SECURITY_PREFIX,
7419         .get    = nfs4_xattr_get_nfs4_label,
7420         .set    = nfs4_xattr_set_nfs4_label,
7421 };
7422
7423 #else
7424
7425 static ssize_t
7426 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7427 {
7428         return 0;
7429 }
7430
7431 #endif
7432
7433 /*
7434  * nfs_fhget will use either the mounted_on_fileid or the fileid
7435  */
7436 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7437 {
7438         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7439                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7440               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7441               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7442                 return;
7443
7444         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7445                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7446         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7447         fattr->nlink = 2;
7448 }
7449
7450 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7451                                    const struct qstr *name,
7452                                    struct nfs4_fs_locations *fs_locations,
7453                                    struct page *page)
7454 {
7455         struct nfs_server *server = NFS_SERVER(dir);
7456         u32 bitmask[3];
7457         struct nfs4_fs_locations_arg args = {
7458                 .dir_fh = NFS_FH(dir),
7459                 .name = name,
7460                 .page = page,
7461                 .bitmask = bitmask,
7462         };
7463         struct nfs4_fs_locations_res res = {
7464                 .fs_locations = fs_locations,
7465         };
7466         struct rpc_message msg = {
7467                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7468                 .rpc_argp = &args,
7469                 .rpc_resp = &res,
7470         };
7471         int status;
7472
7473         dprintk("%s: start\n", __func__);
7474
7475         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7476         bitmask[1] = nfs4_fattr_bitmap[1];
7477
7478         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7479          * is not supported */
7480         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7481                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7482         else
7483                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7484
7485         nfs_fattr_init(&fs_locations->fattr);
7486         fs_locations->server = server;
7487         fs_locations->nlocations = 0;
7488         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7489         dprintk("%s: returned status = %d\n", __func__, status);
7490         return status;
7491 }
7492
7493 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7494                            const struct qstr *name,
7495                            struct nfs4_fs_locations *fs_locations,
7496                            struct page *page)
7497 {
7498         struct nfs4_exception exception = {
7499                 .interruptible = true,
7500         };
7501         int err;
7502         do {
7503                 err = _nfs4_proc_fs_locations(client, dir, name,
7504                                 fs_locations, page);
7505                 trace_nfs4_get_fs_locations(dir, name, err);
7506                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7507                                 &exception);
7508         } while (exception.retry);
7509         return err;
7510 }
7511
7512 /*
7513  * This operation also signals the server that this client is
7514  * performing migration recovery.  The server can stop returning
7515  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7516  * appended to this compound to identify the client ID which is
7517  * performing recovery.
7518  */
7519 static int _nfs40_proc_get_locations(struct inode *inode,
7520                                      struct nfs4_fs_locations *locations,
7521                                      struct page *page, const struct cred *cred)
7522 {
7523         struct nfs_server *server = NFS_SERVER(inode);
7524         struct rpc_clnt *clnt = server->client;
7525         u32 bitmask[2] = {
7526                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7527         };
7528         struct nfs4_fs_locations_arg args = {
7529                 .clientid       = server->nfs_client->cl_clientid,
7530                 .fh             = NFS_FH(inode),
7531                 .page           = page,
7532                 .bitmask        = bitmask,
7533                 .migration      = 1,            /* skip LOOKUP */
7534                 .renew          = 1,            /* append RENEW */
7535         };
7536         struct nfs4_fs_locations_res res = {
7537                 .fs_locations   = locations,
7538                 .migration      = 1,
7539                 .renew          = 1,
7540         };
7541         struct rpc_message msg = {
7542                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7543                 .rpc_argp       = &args,
7544                 .rpc_resp       = &res,
7545                 .rpc_cred       = cred,
7546         };
7547         unsigned long now = jiffies;
7548         int status;
7549
7550         nfs_fattr_init(&locations->fattr);
7551         locations->server = server;
7552         locations->nlocations = 0;
7553
7554         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7555         status = nfs4_call_sync_sequence(clnt, server, &msg,
7556                                         &args.seq_args, &res.seq_res);
7557         if (status)
7558                 return status;
7559
7560         renew_lease(server, now);
7561         return 0;
7562 }
7563
7564 #ifdef CONFIG_NFS_V4_1
7565
7566 /*
7567  * This operation also signals the server that this client is
7568  * performing migration recovery.  The server can stop asserting
7569  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7570  * performing this operation is identified in the SEQUENCE
7571  * operation in this compound.
7572  *
7573  * When the client supports GETATTR(fs_locations_info), it can
7574  * be plumbed in here.
7575  */
7576 static int _nfs41_proc_get_locations(struct inode *inode,
7577                                      struct nfs4_fs_locations *locations,
7578                                      struct page *page, const struct cred *cred)
7579 {
7580         struct nfs_server *server = NFS_SERVER(inode);
7581         struct rpc_clnt *clnt = server->client;
7582         u32 bitmask[2] = {
7583                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7584         };
7585         struct nfs4_fs_locations_arg args = {
7586                 .fh             = NFS_FH(inode),
7587                 .page           = page,
7588                 .bitmask        = bitmask,
7589                 .migration      = 1,            /* skip LOOKUP */
7590         };
7591         struct nfs4_fs_locations_res res = {
7592                 .fs_locations   = locations,
7593                 .migration      = 1,
7594         };
7595         struct rpc_message msg = {
7596                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7597                 .rpc_argp       = &args,
7598                 .rpc_resp       = &res,
7599                 .rpc_cred       = cred,
7600         };
7601         int status;
7602
7603         nfs_fattr_init(&locations->fattr);
7604         locations->server = server;
7605         locations->nlocations = 0;
7606
7607         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7608         status = nfs4_call_sync_sequence(clnt, server, &msg,
7609                                         &args.seq_args, &res.seq_res);
7610         if (status == NFS4_OK &&
7611             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7612                 status = -NFS4ERR_LEASE_MOVED;
7613         return status;
7614 }
7615
7616 #endif  /* CONFIG_NFS_V4_1 */
7617
7618 /**
7619  * nfs4_proc_get_locations - discover locations for a migrated FSID
7620  * @inode: inode on FSID that is migrating
7621  * @locations: result of query
7622  * @page: buffer
7623  * @cred: credential to use for this operation
7624  *
7625  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7626  * operation failed, or a negative errno if a local error occurred.
7627  *
7628  * On success, "locations" is filled in, but if the server has
7629  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7630  * asserted.
7631  *
7632  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7633  * from this client that require migration recovery.
7634  */
7635 int nfs4_proc_get_locations(struct inode *inode,
7636                             struct nfs4_fs_locations *locations,
7637                             struct page *page, const struct cred *cred)
7638 {
7639         struct nfs_server *server = NFS_SERVER(inode);
7640         struct nfs_client *clp = server->nfs_client;
7641         const struct nfs4_mig_recovery_ops *ops =
7642                                         clp->cl_mvops->mig_recovery_ops;
7643         struct nfs4_exception exception = {
7644                 .interruptible = true,
7645         };
7646         int status;
7647
7648         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7649                 (unsigned long long)server->fsid.major,
7650                 (unsigned long long)server->fsid.minor,
7651                 clp->cl_hostname);
7652         nfs_display_fhandle(NFS_FH(inode), __func__);
7653
7654         do {
7655                 status = ops->get_locations(inode, locations, page, cred);
7656                 if (status != -NFS4ERR_DELAY)
7657                         break;
7658                 nfs4_handle_exception(server, status, &exception);
7659         } while (exception.retry);
7660         return status;
7661 }
7662
7663 /*
7664  * This operation also signals the server that this client is
7665  * performing "lease moved" recovery.  The server can stop
7666  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7667  * is appended to this compound to identify the client ID which is
7668  * performing recovery.
7669  */
7670 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7671 {
7672         struct nfs_server *server = NFS_SERVER(inode);
7673         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7674         struct rpc_clnt *clnt = server->client;
7675         struct nfs4_fsid_present_arg args = {
7676                 .fh             = NFS_FH(inode),
7677                 .clientid       = clp->cl_clientid,
7678                 .renew          = 1,            /* append RENEW */
7679         };
7680         struct nfs4_fsid_present_res res = {
7681                 .renew          = 1,
7682         };
7683         struct rpc_message msg = {
7684                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7685                 .rpc_argp       = &args,
7686                 .rpc_resp       = &res,
7687                 .rpc_cred       = cred,
7688         };
7689         unsigned long now = jiffies;
7690         int status;
7691
7692         res.fh = nfs_alloc_fhandle();
7693         if (res.fh == NULL)
7694                 return -ENOMEM;
7695
7696         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7697         status = nfs4_call_sync_sequence(clnt, server, &msg,
7698                                                 &args.seq_args, &res.seq_res);
7699         nfs_free_fhandle(res.fh);
7700         if (status)
7701                 return status;
7702
7703         do_renew_lease(clp, now);
7704         return 0;
7705 }
7706
7707 #ifdef CONFIG_NFS_V4_1
7708
7709 /*
7710  * This operation also signals the server that this client is
7711  * performing "lease moved" recovery.  The server can stop asserting
7712  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7713  * this operation is identified in the SEQUENCE operation in this
7714  * compound.
7715  */
7716 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7717 {
7718         struct nfs_server *server = NFS_SERVER(inode);
7719         struct rpc_clnt *clnt = server->client;
7720         struct nfs4_fsid_present_arg args = {
7721                 .fh             = NFS_FH(inode),
7722         };
7723         struct nfs4_fsid_present_res res = {
7724         };
7725         struct rpc_message msg = {
7726                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7727                 .rpc_argp       = &args,
7728                 .rpc_resp       = &res,
7729                 .rpc_cred       = cred,
7730         };
7731         int status;
7732
7733         res.fh = nfs_alloc_fhandle();
7734         if (res.fh == NULL)
7735                 return -ENOMEM;
7736
7737         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7738         status = nfs4_call_sync_sequence(clnt, server, &msg,
7739                                                 &args.seq_args, &res.seq_res);
7740         nfs_free_fhandle(res.fh);
7741         if (status == NFS4_OK &&
7742             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7743                 status = -NFS4ERR_LEASE_MOVED;
7744         return status;
7745 }
7746
7747 #endif  /* CONFIG_NFS_V4_1 */
7748
7749 /**
7750  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7751  * @inode: inode on FSID to check
7752  * @cred: credential to use for this operation
7753  *
7754  * Server indicates whether the FSID is present, moved, or not
7755  * recognized.  This operation is necessary to clear a LEASE_MOVED
7756  * condition for this client ID.
7757  *
7758  * Returns NFS4_OK if the FSID is present on this server,
7759  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7760  *  NFS4ERR code if some error occurred on the server, or a
7761  *  negative errno if a local failure occurred.
7762  */
7763 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7764 {
7765         struct nfs_server *server = NFS_SERVER(inode);
7766         struct nfs_client *clp = server->nfs_client;
7767         const struct nfs4_mig_recovery_ops *ops =
7768                                         clp->cl_mvops->mig_recovery_ops;
7769         struct nfs4_exception exception = {
7770                 .interruptible = true,
7771         };
7772         int status;
7773
7774         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7775                 (unsigned long long)server->fsid.major,
7776                 (unsigned long long)server->fsid.minor,
7777                 clp->cl_hostname);
7778         nfs_display_fhandle(NFS_FH(inode), __func__);
7779
7780         do {
7781                 status = ops->fsid_present(inode, cred);
7782                 if (status != -NFS4ERR_DELAY)
7783                         break;
7784                 nfs4_handle_exception(server, status, &exception);
7785         } while (exception.retry);
7786         return status;
7787 }
7788
7789 /*
7790  * If 'use_integrity' is true and the state managment nfs_client
7791  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7792  * and the machine credential as per RFC3530bis and RFC5661 Security
7793  * Considerations sections. Otherwise, just use the user cred with the
7794  * filesystem's rpc_client.
7795  */
7796 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7797 {
7798         int status;
7799         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7800         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
7801         struct nfs4_secinfo_arg args = {
7802                 .dir_fh = NFS_FH(dir),
7803                 .name   = name,
7804         };
7805         struct nfs4_secinfo_res res = {
7806                 .flavors     = flavors,
7807         };
7808         struct rpc_message msg = {
7809                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7810                 .rpc_argp = &args,
7811                 .rpc_resp = &res,
7812         };
7813         struct nfs4_call_sync_data data = {
7814                 .seq_server = NFS_SERVER(dir),
7815                 .seq_args = &args.seq_args,
7816                 .seq_res = &res.seq_res,
7817         };
7818         struct rpc_task_setup task_setup = {
7819                 .rpc_client = clnt,
7820                 .rpc_message = &msg,
7821                 .callback_ops = clp->cl_mvops->call_sync_ops,
7822                 .callback_data = &data,
7823                 .flags = RPC_TASK_NO_ROUND_ROBIN,
7824         };
7825         const struct cred *cred = NULL;
7826
7827         if (use_integrity) {
7828                 clnt = clp->cl_rpcclient;
7829                 task_setup.rpc_client = clnt;
7830
7831                 cred = nfs4_get_clid_cred(clp);
7832                 msg.rpc_cred = cred;
7833         }
7834
7835         dprintk("NFS call  secinfo %s\n", name->name);
7836
7837         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7838         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
7839         status = nfs4_call_sync_custom(&task_setup);
7840
7841         dprintk("NFS reply  secinfo: %d\n", status);
7842
7843         put_cred(cred);
7844         return status;
7845 }
7846
7847 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7848                       struct nfs4_secinfo_flavors *flavors)
7849 {
7850         struct nfs4_exception exception = {
7851                 .interruptible = true,
7852         };
7853         int err;
7854         do {
7855                 err = -NFS4ERR_WRONGSEC;
7856
7857                 /* try to use integrity protection with machine cred */
7858                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7859                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7860
7861                 /*
7862                  * if unable to use integrity protection, or SECINFO with
7863                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7864                  * disallowed by spec, but exists in deployed servers) use
7865                  * the current filesystem's rpc_client and the user cred.
7866                  */
7867                 if (err == -NFS4ERR_WRONGSEC)
7868                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7869
7870                 trace_nfs4_secinfo(dir, name, err);
7871                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7872                                 &exception);
7873         } while (exception.retry);
7874         return err;
7875 }
7876
7877 #ifdef CONFIG_NFS_V4_1
7878 /*
7879  * Check the exchange flags returned by the server for invalid flags, having
7880  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7881  * DS flags set.
7882  */
7883 static int nfs4_check_cl_exchange_flags(u32 flags)
7884 {
7885         if (flags & ~EXCHGID4_FLAG_MASK_R)
7886                 goto out_inval;
7887         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7888             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7889                 goto out_inval;
7890         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7891                 goto out_inval;
7892         return NFS_OK;
7893 out_inval:
7894         return -NFS4ERR_INVAL;
7895 }
7896
7897 static bool
7898 nfs41_same_server_scope(struct nfs41_server_scope *a,
7899                         struct nfs41_server_scope *b)
7900 {
7901         if (a->server_scope_sz != b->server_scope_sz)
7902                 return false;
7903         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7904 }
7905
7906 static void
7907 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7908 {
7909         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
7910         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
7911         struct nfs_client *clp = args->client;
7912
7913         switch (task->tk_status) {
7914         case -NFS4ERR_BADSESSION:
7915         case -NFS4ERR_DEADSESSION:
7916                 nfs4_schedule_session_recovery(clp->cl_session,
7917                                 task->tk_status);
7918         }
7919         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
7920                         res->dir != NFS4_CDFS4_BOTH) {
7921                 rpc_task_close_connection(task);
7922                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
7923                         rpc_restart_call(task);
7924         }
7925 }
7926
7927 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7928         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
7929 };
7930
7931 /*
7932  * nfs4_proc_bind_one_conn_to_session()
7933  *
7934  * The 4.1 client currently uses the same TCP connection for the
7935  * fore and backchannel.
7936  */
7937 static
7938 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7939                 struct rpc_xprt *xprt,
7940                 struct nfs_client *clp,
7941                 const struct cred *cred)
7942 {
7943         int status;
7944         struct nfs41_bind_conn_to_session_args args = {
7945                 .client = clp,
7946                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7947                 .retries = 0,
7948         };
7949         struct nfs41_bind_conn_to_session_res res;
7950         struct rpc_message msg = {
7951                 .rpc_proc =
7952                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7953                 .rpc_argp = &args,
7954                 .rpc_resp = &res,
7955                 .rpc_cred = cred,
7956         };
7957         struct rpc_task_setup task_setup_data = {
7958                 .rpc_client = clnt,
7959                 .rpc_xprt = xprt,
7960                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7961                 .rpc_message = &msg,
7962                 .flags = RPC_TASK_TIMEOUT,
7963         };
7964         struct rpc_task *task;
7965
7966         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7967         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7968                 args.dir = NFS4_CDFC4_FORE;
7969
7970         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7971         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7972                 args.dir = NFS4_CDFC4_FORE;
7973
7974         task = rpc_run_task(&task_setup_data);
7975         if (!IS_ERR(task)) {
7976                 status = task->tk_status;
7977                 rpc_put_task(task);
7978         } else
7979                 status = PTR_ERR(task);
7980         trace_nfs4_bind_conn_to_session(clp, status);
7981         if (status == 0) {
7982                 if (memcmp(res.sessionid.data,
7983                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7984                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7985                         return -EIO;
7986                 }
7987                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7988                         dprintk("NFS: %s: Unexpected direction from server\n",
7989                                 __func__);
7990                         return -EIO;
7991                 }
7992                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7993                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7994                                 __func__);
7995                         return -EIO;
7996                 }
7997         }
7998
7999         return status;
8000 }
8001
8002 struct rpc_bind_conn_calldata {
8003         struct nfs_client *clp;
8004         const struct cred *cred;
8005 };
8006
8007 static int
8008 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8009                 struct rpc_xprt *xprt,
8010                 void *calldata)
8011 {
8012         struct rpc_bind_conn_calldata *p = calldata;
8013
8014         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8015 }
8016
8017 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8018 {
8019         struct rpc_bind_conn_calldata data = {
8020                 .clp = clp,
8021                 .cred = cred,
8022         };
8023         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8024                         nfs4_proc_bind_conn_to_session_callback, &data);
8025 }
8026
8027 /*
8028  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8029  * and operations we'd like to see to enable certain features in the allow map
8030  */
8031 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8032         .how = SP4_MACH_CRED,
8033         .enforce.u.words = {
8034                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8035                       1 << (OP_EXCHANGE_ID - 32) |
8036                       1 << (OP_CREATE_SESSION - 32) |
8037                       1 << (OP_DESTROY_SESSION - 32) |
8038                       1 << (OP_DESTROY_CLIENTID - 32)
8039         },
8040         .allow.u.words = {
8041                 [0] = 1 << (OP_CLOSE) |
8042                       1 << (OP_OPEN_DOWNGRADE) |
8043                       1 << (OP_LOCKU) |
8044                       1 << (OP_DELEGRETURN) |
8045                       1 << (OP_COMMIT),
8046                 [1] = 1 << (OP_SECINFO - 32) |
8047                       1 << (OP_SECINFO_NO_NAME - 32) |
8048                       1 << (OP_LAYOUTRETURN - 32) |
8049                       1 << (OP_TEST_STATEID - 32) |
8050                       1 << (OP_FREE_STATEID - 32) |
8051                       1 << (OP_WRITE - 32)
8052         }
8053 };
8054
8055 /*
8056  * Select the state protection mode for client `clp' given the server results
8057  * from exchange_id in `sp'.
8058  *
8059  * Returns 0 on success, negative errno otherwise.
8060  */
8061 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8062                                  struct nfs41_state_protection *sp)
8063 {
8064         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8065                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8066                       1 << (OP_EXCHANGE_ID - 32) |
8067                       1 << (OP_CREATE_SESSION - 32) |
8068                       1 << (OP_DESTROY_SESSION - 32) |
8069                       1 << (OP_DESTROY_CLIENTID - 32)
8070         };
8071         unsigned long flags = 0;
8072         unsigned int i;
8073         int ret = 0;
8074
8075         if (sp->how == SP4_MACH_CRED) {
8076                 /* Print state protect result */
8077                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8078                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8079                         if (test_bit(i, sp->enforce.u.longs))
8080                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8081                         if (test_bit(i, sp->allow.u.longs))
8082                                 dfprintk(MOUNT, "  allow op %d\n", i);
8083                 }
8084
8085                 /* make sure nothing is on enforce list that isn't supported */
8086                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8087                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8088                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8089                                 ret = -EINVAL;
8090                                 goto out;
8091                         }
8092                 }
8093
8094                 /*
8095                  * Minimal mode - state operations are allowed to use machine
8096                  * credential.  Note this already happens by default, so the
8097                  * client doesn't have to do anything more than the negotiation.
8098                  *
8099                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8100                  *       we're already using the machine cred for exchange_id
8101                  *       and will never use a different cred.
8102                  */
8103                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8104                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8105                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8106                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8107                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8108                         dfprintk(MOUNT, "  minimal mode enabled\n");
8109                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8110                 } else {
8111                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8112                         ret = -EINVAL;
8113                         goto out;
8114                 }
8115
8116                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8117                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8118                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8119                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8120                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8121                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8122                 }
8123
8124                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8125                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8126                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8127                 }
8128
8129                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8130                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8131                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8132                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8133                 }
8134
8135                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8136                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8137                         dfprintk(MOUNT, "  stateid mode enabled\n");
8138                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8139                 }
8140
8141                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8142                         dfprintk(MOUNT, "  write mode enabled\n");
8143                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8144                 }
8145
8146                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8147                         dfprintk(MOUNT, "  commit mode enabled\n");
8148                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8149                 }
8150         }
8151 out:
8152         clp->cl_sp4_flags = flags;
8153         return ret;
8154 }
8155
8156 struct nfs41_exchange_id_data {
8157         struct nfs41_exchange_id_res res;
8158         struct nfs41_exchange_id_args args;
8159 };
8160
8161 static void nfs4_exchange_id_release(void *data)
8162 {
8163         struct nfs41_exchange_id_data *cdata =
8164                                         (struct nfs41_exchange_id_data *)data;
8165
8166         nfs_put_client(cdata->args.client);
8167         kfree(cdata->res.impl_id);
8168         kfree(cdata->res.server_scope);
8169         kfree(cdata->res.server_owner);
8170         kfree(cdata);
8171 }
8172
8173 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8174         .rpc_release = nfs4_exchange_id_release,
8175 };
8176
8177 /*
8178  * _nfs4_proc_exchange_id()
8179  *
8180  * Wrapper for EXCHANGE_ID operation.
8181  */
8182 static struct rpc_task *
8183 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8184                         u32 sp4_how, struct rpc_xprt *xprt)
8185 {
8186         struct rpc_message msg = {
8187                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8188                 .rpc_cred = cred,
8189         };
8190         struct rpc_task_setup task_setup_data = {
8191                 .rpc_client = clp->cl_rpcclient,
8192                 .callback_ops = &nfs4_exchange_id_call_ops,
8193                 .rpc_message = &msg,
8194                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8195         };
8196         struct nfs41_exchange_id_data *calldata;
8197         int status;
8198
8199         if (!refcount_inc_not_zero(&clp->cl_count))
8200                 return ERR_PTR(-EIO);
8201
8202         status = -ENOMEM;
8203         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8204         if (!calldata)
8205                 goto out;
8206
8207         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8208
8209         status = nfs4_init_uniform_client_string(clp);
8210         if (status)
8211                 goto out_calldata;
8212
8213         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8214                                                 GFP_NOFS);
8215         status = -ENOMEM;
8216         if (unlikely(calldata->res.server_owner == NULL))
8217                 goto out_calldata;
8218
8219         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8220                                         GFP_NOFS);
8221         if (unlikely(calldata->res.server_scope == NULL))
8222                 goto out_server_owner;
8223
8224         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8225         if (unlikely(calldata->res.impl_id == NULL))
8226                 goto out_server_scope;
8227
8228         switch (sp4_how) {
8229         case SP4_NONE:
8230                 calldata->args.state_protect.how = SP4_NONE;
8231                 break;
8232
8233         case SP4_MACH_CRED:
8234                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8235                 break;
8236
8237         default:
8238                 /* unsupported! */
8239                 WARN_ON_ONCE(1);
8240                 status = -EINVAL;
8241                 goto out_impl_id;
8242         }
8243         if (xprt) {
8244                 task_setup_data.rpc_xprt = xprt;
8245                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8246                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8247                                 sizeof(calldata->args.verifier.data));
8248         }
8249         calldata->args.client = clp;
8250         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8251         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8252 #ifdef CONFIG_NFS_V4_1_MIGRATION
8253         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8254 #endif
8255         msg.rpc_argp = &calldata->args;
8256         msg.rpc_resp = &calldata->res;
8257         task_setup_data.callback_data = calldata;
8258
8259         return rpc_run_task(&task_setup_data);
8260
8261 out_impl_id:
8262         kfree(calldata->res.impl_id);
8263 out_server_scope:
8264         kfree(calldata->res.server_scope);
8265 out_server_owner:
8266         kfree(calldata->res.server_owner);
8267 out_calldata:
8268         kfree(calldata);
8269 out:
8270         nfs_put_client(clp);
8271         return ERR_PTR(status);
8272 }
8273
8274 /*
8275  * _nfs4_proc_exchange_id()
8276  *
8277  * Wrapper for EXCHANGE_ID operation.
8278  */
8279 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8280                         u32 sp4_how)
8281 {
8282         struct rpc_task *task;
8283         struct nfs41_exchange_id_args *argp;
8284         struct nfs41_exchange_id_res *resp;
8285         unsigned long now = jiffies;
8286         int status;
8287
8288         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8289         if (IS_ERR(task))
8290                 return PTR_ERR(task);
8291
8292         argp = task->tk_msg.rpc_argp;
8293         resp = task->tk_msg.rpc_resp;
8294         status = task->tk_status;
8295         if (status  != 0)
8296                 goto out;
8297
8298         status = nfs4_check_cl_exchange_flags(resp->flags);
8299         if (status  != 0)
8300                 goto out;
8301
8302         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8303         if (status != 0)
8304                 goto out;
8305
8306         do_renew_lease(clp, now);
8307
8308         clp->cl_clientid = resp->clientid;
8309         clp->cl_exchange_flags = resp->flags;
8310         clp->cl_seqid = resp->seqid;
8311         /* Client ID is not confirmed */
8312         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8313                 clear_bit(NFS4_SESSION_ESTABLISHED,
8314                           &clp->cl_session->session_state);
8315
8316         if (clp->cl_serverscope != NULL &&
8317             !nfs41_same_server_scope(clp->cl_serverscope,
8318                                 resp->server_scope)) {
8319                 dprintk("%s: server_scope mismatch detected\n",
8320                         __func__);
8321                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8322         }
8323
8324         swap(clp->cl_serverowner, resp->server_owner);
8325         swap(clp->cl_serverscope, resp->server_scope);
8326         swap(clp->cl_implid, resp->impl_id);
8327
8328         /* Save the EXCHANGE_ID verifier session trunk tests */
8329         memcpy(clp->cl_confirm.data, argp->verifier.data,
8330                sizeof(clp->cl_confirm.data));
8331 out:
8332         trace_nfs4_exchange_id(clp, status);
8333         rpc_put_task(task);
8334         return status;
8335 }
8336
8337 /*
8338  * nfs4_proc_exchange_id()
8339  *
8340  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8341  *
8342  * Since the clientid has expired, all compounds using sessions
8343  * associated with the stale clientid will be returning
8344  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8345  * be in some phase of session reset.
8346  *
8347  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8348  */
8349 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8350 {
8351         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8352         int status;
8353
8354         /* try SP4_MACH_CRED if krb5i/p */
8355         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8356             authflavor == RPC_AUTH_GSS_KRB5P) {
8357                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8358                 if (!status)
8359                         return 0;
8360         }
8361
8362         /* try SP4_NONE */
8363         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8364 }
8365
8366 /**
8367  * nfs4_test_session_trunk
8368  *
8369  * This is an add_xprt_test() test function called from
8370  * rpc_clnt_setup_test_and_add_xprt.
8371  *
8372  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8373  * and is dereferrenced in nfs4_exchange_id_release
8374  *
8375  * Upon success, add the new transport to the rpc_clnt
8376  *
8377  * @clnt: struct rpc_clnt to get new transport
8378  * @xprt: the rpc_xprt to test
8379  * @data: call data for _nfs4_proc_exchange_id.
8380  */
8381 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8382                             void *data)
8383 {
8384         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8385         struct rpc_task *task;
8386         int status;
8387
8388         u32 sp4_how;
8389
8390         dprintk("--> %s try %s\n", __func__,
8391                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8392
8393         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8394
8395         /* Test connection for session trunking. Async exchange_id call */
8396         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8397         if (IS_ERR(task))
8398                 return;
8399
8400         status = task->tk_status;
8401         if (status == 0)
8402                 status = nfs4_detect_session_trunking(adata->clp,
8403                                 task->tk_msg.rpc_resp, xprt);
8404
8405         if (status == 0)
8406                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8407
8408         rpc_put_task(task);
8409 }
8410 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8411
8412 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8413                 const struct cred *cred)
8414 {
8415         struct rpc_message msg = {
8416                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8417                 .rpc_argp = clp,
8418                 .rpc_cred = cred,
8419         };
8420         int status;
8421
8422         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8423                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8424         trace_nfs4_destroy_clientid(clp, status);
8425         if (status)
8426                 dprintk("NFS: Got error %d from the server %s on "
8427                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8428         return status;
8429 }
8430
8431 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8432                 const struct cred *cred)
8433 {
8434         unsigned int loop;
8435         int ret;
8436
8437         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8438                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8439                 switch (ret) {
8440                 case -NFS4ERR_DELAY:
8441                 case -NFS4ERR_CLIENTID_BUSY:
8442                         ssleep(1);
8443                         break;
8444                 default:
8445                         return ret;
8446                 }
8447         }
8448         return 0;
8449 }
8450
8451 int nfs4_destroy_clientid(struct nfs_client *clp)
8452 {
8453         const struct cred *cred;
8454         int ret = 0;
8455
8456         if (clp->cl_mvops->minor_version < 1)
8457                 goto out;
8458         if (clp->cl_exchange_flags == 0)
8459                 goto out;
8460         if (clp->cl_preserve_clid)
8461                 goto out;
8462         cred = nfs4_get_clid_cred(clp);
8463         ret = nfs4_proc_destroy_clientid(clp, cred);
8464         put_cred(cred);
8465         switch (ret) {
8466         case 0:
8467         case -NFS4ERR_STALE_CLIENTID:
8468                 clp->cl_exchange_flags = 0;
8469         }
8470 out:
8471         return ret;
8472 }
8473
8474 #endif /* CONFIG_NFS_V4_1 */
8475
8476 struct nfs4_get_lease_time_data {
8477         struct nfs4_get_lease_time_args *args;
8478         struct nfs4_get_lease_time_res *res;
8479         struct nfs_client *clp;
8480 };
8481
8482 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8483                                         void *calldata)
8484 {
8485         struct nfs4_get_lease_time_data *data =
8486                         (struct nfs4_get_lease_time_data *)calldata;
8487
8488         dprintk("--> %s\n", __func__);
8489         /* just setup sequence, do not trigger session recovery
8490            since we're invoked within one */
8491         nfs4_setup_sequence(data->clp,
8492                         &data->args->la_seq_args,
8493                         &data->res->lr_seq_res,
8494                         task);
8495         dprintk("<-- %s\n", __func__);
8496 }
8497
8498 /*
8499  * Called from nfs4_state_manager thread for session setup, so don't recover
8500  * from sequence operation or clientid errors.
8501  */
8502 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8503 {
8504         struct nfs4_get_lease_time_data *data =
8505                         (struct nfs4_get_lease_time_data *)calldata;
8506
8507         dprintk("--> %s\n", __func__);
8508         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8509                 return;
8510         switch (task->tk_status) {
8511         case -NFS4ERR_DELAY:
8512         case -NFS4ERR_GRACE:
8513                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8514                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8515                 task->tk_status = 0;
8516                 /* fall through */
8517         case -NFS4ERR_RETRY_UNCACHED_REP:
8518                 rpc_restart_call_prepare(task);
8519                 return;
8520         }
8521         dprintk("<-- %s\n", __func__);
8522 }
8523
8524 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8525         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8526         .rpc_call_done = nfs4_get_lease_time_done,
8527 };
8528
8529 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8530 {
8531         struct nfs4_get_lease_time_args args;
8532         struct nfs4_get_lease_time_res res = {
8533                 .lr_fsinfo = fsinfo,
8534         };
8535         struct nfs4_get_lease_time_data data = {
8536                 .args = &args,
8537                 .res = &res,
8538                 .clp = clp,
8539         };
8540         struct rpc_message msg = {
8541                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8542                 .rpc_argp = &args,
8543                 .rpc_resp = &res,
8544         };
8545         struct rpc_task_setup task_setup = {
8546                 .rpc_client = clp->cl_rpcclient,
8547                 .rpc_message = &msg,
8548                 .callback_ops = &nfs4_get_lease_time_ops,
8549                 .callback_data = &data,
8550                 .flags = RPC_TASK_TIMEOUT,
8551         };
8552
8553         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8554         return nfs4_call_sync_custom(&task_setup);
8555 }
8556
8557 #ifdef CONFIG_NFS_V4_1
8558
8559 /*
8560  * Initialize the values to be used by the client in CREATE_SESSION
8561  * If nfs4_init_session set the fore channel request and response sizes,
8562  * use them.
8563  *
8564  * Set the back channel max_resp_sz_cached to zero to force the client to
8565  * always set csa_cachethis to FALSE because the current implementation
8566  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8567  */
8568 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8569                                     struct rpc_clnt *clnt)
8570 {
8571         unsigned int max_rqst_sz, max_resp_sz;
8572         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8573         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8574
8575         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8576         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8577
8578         /* Fore channel attributes */
8579         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8580         args->fc_attrs.max_resp_sz = max_resp_sz;
8581         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8582         args->fc_attrs.max_reqs = max_session_slots;
8583
8584         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8585                 "max_ops=%u max_reqs=%u\n",
8586                 __func__,
8587                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8588                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8589
8590         /* Back channel attributes */
8591         args->bc_attrs.max_rqst_sz = max_bc_payload;
8592         args->bc_attrs.max_resp_sz = max_bc_payload;
8593         args->bc_attrs.max_resp_sz_cached = 0;
8594         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8595         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8596         if (args->bc_attrs.max_reqs > max_bc_slots)
8597                 args->bc_attrs.max_reqs = max_bc_slots;
8598
8599         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8600                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8601                 __func__,
8602                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8603                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8604                 args->bc_attrs.max_reqs);
8605 }
8606
8607 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8608                 struct nfs41_create_session_res *res)
8609 {
8610         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8611         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8612
8613         if (rcvd->max_resp_sz > sent->max_resp_sz)
8614                 return -EINVAL;
8615         /*
8616          * Our requested max_ops is the minimum we need; we're not
8617          * prepared to break up compounds into smaller pieces than that.
8618          * So, no point even trying to continue if the server won't
8619          * cooperate:
8620          */
8621         if (rcvd->max_ops < sent->max_ops)
8622                 return -EINVAL;
8623         if (rcvd->max_reqs == 0)
8624                 return -EINVAL;
8625         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8626                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8627         return 0;
8628 }
8629
8630 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8631                 struct nfs41_create_session_res *res)
8632 {
8633         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8634         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8635
8636         if (!(res->flags & SESSION4_BACK_CHAN))
8637                 goto out;
8638         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8639                 return -EINVAL;
8640         if (rcvd->max_resp_sz < sent->max_resp_sz)
8641                 return -EINVAL;
8642         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8643                 return -EINVAL;
8644         if (rcvd->max_ops > sent->max_ops)
8645                 return -EINVAL;
8646         if (rcvd->max_reqs > sent->max_reqs)
8647                 return -EINVAL;
8648 out:
8649         return 0;
8650 }
8651
8652 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8653                                      struct nfs41_create_session_res *res)
8654 {
8655         int ret;
8656
8657         ret = nfs4_verify_fore_channel_attrs(args, res);
8658         if (ret)
8659                 return ret;
8660         return nfs4_verify_back_channel_attrs(args, res);
8661 }
8662
8663 static void nfs4_update_session(struct nfs4_session *session,
8664                 struct nfs41_create_session_res *res)
8665 {
8666         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8667         /* Mark client id and session as being confirmed */
8668         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8669         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8670         session->flags = res->flags;
8671         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8672         if (res->flags & SESSION4_BACK_CHAN)
8673                 memcpy(&session->bc_attrs, &res->bc_attrs,
8674                                 sizeof(session->bc_attrs));
8675 }
8676
8677 static int _nfs4_proc_create_session(struct nfs_client *clp,
8678                 const struct cred *cred)
8679 {
8680         struct nfs4_session *session = clp->cl_session;
8681         struct nfs41_create_session_args args = {
8682                 .client = clp,
8683                 .clientid = clp->cl_clientid,
8684                 .seqid = clp->cl_seqid,
8685                 .cb_program = NFS4_CALLBACK,
8686         };
8687         struct nfs41_create_session_res res;
8688
8689         struct rpc_message msg = {
8690                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8691                 .rpc_argp = &args,
8692                 .rpc_resp = &res,
8693                 .rpc_cred = cred,
8694         };
8695         int status;
8696
8697         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8698         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8699
8700         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8701                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8702         trace_nfs4_create_session(clp, status);
8703
8704         switch (status) {
8705         case -NFS4ERR_STALE_CLIENTID:
8706         case -NFS4ERR_DELAY:
8707         case -ETIMEDOUT:
8708         case -EACCES:
8709         case -EAGAIN:
8710                 goto out;
8711         }
8712
8713         clp->cl_seqid++;
8714         if (!status) {
8715                 /* Verify the session's negotiated channel_attrs values */
8716                 status = nfs4_verify_channel_attrs(&args, &res);
8717                 /* Increment the clientid slot sequence id */
8718                 if (status)
8719                         goto out;
8720                 nfs4_update_session(session, &res);
8721         }
8722 out:
8723         return status;
8724 }
8725
8726 /*
8727  * Issues a CREATE_SESSION operation to the server.
8728  * It is the responsibility of the caller to verify the session is
8729  * expired before calling this routine.
8730  */
8731 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8732 {
8733         int status;
8734         unsigned *ptr;
8735         struct nfs4_session *session = clp->cl_session;
8736
8737         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8738
8739         status = _nfs4_proc_create_session(clp, cred);
8740         if (status)
8741                 goto out;
8742
8743         /* Init or reset the session slot tables */
8744         status = nfs4_setup_session_slot_tables(session);
8745         dprintk("slot table setup returned %d\n", status);
8746         if (status)
8747                 goto out;
8748
8749         ptr = (unsigned *)&session->sess_id.data[0];
8750         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8751                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8752 out:
8753         dprintk("<-- %s\n", __func__);
8754         return status;
8755 }
8756
8757 /*
8758  * Issue the over-the-wire RPC DESTROY_SESSION.
8759  * The caller must serialize access to this routine.
8760  */
8761 int nfs4_proc_destroy_session(struct nfs4_session *session,
8762                 const struct cred *cred)
8763 {
8764         struct rpc_message msg = {
8765                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8766                 .rpc_argp = session,
8767                 .rpc_cred = cred,
8768         };
8769         int status = 0;
8770
8771         dprintk("--> nfs4_proc_destroy_session\n");
8772
8773         /* session is still being setup */
8774         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8775                 return 0;
8776
8777         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8778                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8779         trace_nfs4_destroy_session(session->clp, status);
8780
8781         if (status)
8782                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8783                         "Session has been destroyed regardless...\n", status);
8784
8785         dprintk("<-- nfs4_proc_destroy_session\n");
8786         return status;
8787 }
8788
8789 /*
8790  * Renew the cl_session lease.
8791  */
8792 struct nfs4_sequence_data {
8793         struct nfs_client *clp;
8794         struct nfs4_sequence_args args;
8795         struct nfs4_sequence_res res;
8796 };
8797
8798 static void nfs41_sequence_release(void *data)
8799 {
8800         struct nfs4_sequence_data *calldata = data;
8801         struct nfs_client *clp = calldata->clp;
8802
8803         if (refcount_read(&clp->cl_count) > 1)
8804                 nfs4_schedule_state_renewal(clp);
8805         nfs_put_client(clp);
8806         kfree(calldata);
8807 }
8808
8809 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8810 {
8811         switch(task->tk_status) {
8812         case -NFS4ERR_DELAY:
8813                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8814                 return -EAGAIN;
8815         default:
8816                 nfs4_schedule_lease_recovery(clp);
8817         }
8818         return 0;
8819 }
8820
8821 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8822 {
8823         struct nfs4_sequence_data *calldata = data;
8824         struct nfs_client *clp = calldata->clp;
8825
8826         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8827                 return;
8828
8829         trace_nfs4_sequence(clp, task->tk_status);
8830         if (task->tk_status < 0) {
8831                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8832                 if (refcount_read(&clp->cl_count) == 1)
8833                         goto out;
8834
8835                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8836                         rpc_restart_call_prepare(task);
8837                         return;
8838                 }
8839         }
8840         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8841 out:
8842         dprintk("<-- %s\n", __func__);
8843 }
8844
8845 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8846 {
8847         struct nfs4_sequence_data *calldata = data;
8848         struct nfs_client *clp = calldata->clp;
8849         struct nfs4_sequence_args *args;
8850         struct nfs4_sequence_res *res;
8851
8852         args = task->tk_msg.rpc_argp;
8853         res = task->tk_msg.rpc_resp;
8854
8855         nfs4_setup_sequence(clp, args, res, task);
8856 }
8857
8858 static const struct rpc_call_ops nfs41_sequence_ops = {
8859         .rpc_call_done = nfs41_sequence_call_done,
8860         .rpc_call_prepare = nfs41_sequence_prepare,
8861         .rpc_release = nfs41_sequence_release,
8862 };
8863
8864 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8865                 const struct cred *cred,
8866                 struct nfs4_slot *slot,
8867                 bool is_privileged)
8868 {
8869         struct nfs4_sequence_data *calldata;
8870         struct rpc_message msg = {
8871                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8872                 .rpc_cred = cred,
8873         };
8874         struct rpc_task_setup task_setup_data = {
8875                 .rpc_client = clp->cl_rpcclient,
8876                 .rpc_message = &msg,
8877                 .callback_ops = &nfs41_sequence_ops,
8878                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8879         };
8880         struct rpc_task *ret;
8881
8882         ret = ERR_PTR(-EIO);
8883         if (!refcount_inc_not_zero(&clp->cl_count))
8884                 goto out_err;
8885
8886         ret = ERR_PTR(-ENOMEM);
8887         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8888         if (calldata == NULL)
8889                 goto out_put_clp;
8890         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8891         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8892         msg.rpc_argp = &calldata->args;
8893         msg.rpc_resp = &calldata->res;
8894         calldata->clp = clp;
8895         task_setup_data.callback_data = calldata;
8896
8897         ret = rpc_run_task(&task_setup_data);
8898         if (IS_ERR(ret))
8899                 goto out_err;
8900         return ret;
8901 out_put_clp:
8902         nfs_put_client(clp);
8903 out_err:
8904         nfs41_release_slot(slot);
8905         return ret;
8906 }
8907
8908 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8909 {
8910         struct rpc_task *task;
8911         int ret = 0;
8912
8913         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8914                 return -EAGAIN;
8915         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8916         if (IS_ERR(task))
8917                 ret = PTR_ERR(task);
8918         else
8919                 rpc_put_task_async(task);
8920         dprintk("<-- %s status=%d\n", __func__, ret);
8921         return ret;
8922 }
8923
8924 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8925 {
8926         struct rpc_task *task;
8927         int ret;
8928
8929         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8930         if (IS_ERR(task)) {
8931                 ret = PTR_ERR(task);
8932                 goto out;
8933         }
8934         ret = rpc_wait_for_completion_task(task);
8935         if (!ret)
8936                 ret = task->tk_status;
8937         rpc_put_task(task);
8938 out:
8939         dprintk("<-- %s status=%d\n", __func__, ret);
8940         return ret;
8941 }
8942
8943 struct nfs4_reclaim_complete_data {
8944         struct nfs_client *clp;
8945         struct nfs41_reclaim_complete_args arg;
8946         struct nfs41_reclaim_complete_res res;
8947 };
8948
8949 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8950 {
8951         struct nfs4_reclaim_complete_data *calldata = data;
8952
8953         nfs4_setup_sequence(calldata->clp,
8954                         &calldata->arg.seq_args,
8955                         &calldata->res.seq_res,
8956                         task);
8957 }
8958
8959 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8960 {
8961         switch(task->tk_status) {
8962         case 0:
8963                 wake_up_all(&clp->cl_lock_waitq);
8964                 /* Fallthrough */
8965         case -NFS4ERR_COMPLETE_ALREADY:
8966         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8967                 break;
8968         case -NFS4ERR_DELAY:
8969                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8970                 /* fall through */
8971         case -NFS4ERR_RETRY_UNCACHED_REP:
8972                 return -EAGAIN;
8973         case -NFS4ERR_BADSESSION:
8974         case -NFS4ERR_DEADSESSION:
8975         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8976                 break;
8977         default:
8978                 nfs4_schedule_lease_recovery(clp);
8979         }
8980         return 0;
8981 }
8982
8983 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8984 {
8985         struct nfs4_reclaim_complete_data *calldata = data;
8986         struct nfs_client *clp = calldata->clp;
8987         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8988
8989         dprintk("--> %s\n", __func__);
8990         if (!nfs41_sequence_done(task, res))
8991                 return;
8992
8993         trace_nfs4_reclaim_complete(clp, task->tk_status);
8994         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8995                 rpc_restart_call_prepare(task);
8996                 return;
8997         }
8998         dprintk("<-- %s\n", __func__);
8999 }
9000
9001 static void nfs4_free_reclaim_complete_data(void *data)
9002 {
9003         struct nfs4_reclaim_complete_data *calldata = data;
9004
9005         kfree(calldata);
9006 }
9007
9008 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9009         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9010         .rpc_call_done = nfs4_reclaim_complete_done,
9011         .rpc_release = nfs4_free_reclaim_complete_data,
9012 };
9013
9014 /*
9015  * Issue a global reclaim complete.
9016  */
9017 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9018                 const struct cred *cred)
9019 {
9020         struct nfs4_reclaim_complete_data *calldata;
9021         struct rpc_message msg = {
9022                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9023                 .rpc_cred = cred,
9024         };
9025         struct rpc_task_setup task_setup_data = {
9026                 .rpc_client = clp->cl_rpcclient,
9027                 .rpc_message = &msg,
9028                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9029                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9030         };
9031         int status = -ENOMEM;
9032
9033         dprintk("--> %s\n", __func__);
9034         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9035         if (calldata == NULL)
9036                 goto out;
9037         calldata->clp = clp;
9038         calldata->arg.one_fs = 0;
9039
9040         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9041         msg.rpc_argp = &calldata->arg;
9042         msg.rpc_resp = &calldata->res;
9043         task_setup_data.callback_data = calldata;
9044         status = nfs4_call_sync_custom(&task_setup_data);
9045 out:
9046         dprintk("<-- %s status=%d\n", __func__, status);
9047         return status;
9048 }
9049
9050 static void
9051 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9052 {
9053         struct nfs4_layoutget *lgp = calldata;
9054         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9055
9056         dprintk("--> %s\n", __func__);
9057         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9058                                 &lgp->res.seq_res, task);
9059         dprintk("<-- %s\n", __func__);
9060 }
9061
9062 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9063 {
9064         struct nfs4_layoutget *lgp = calldata;
9065
9066         dprintk("--> %s\n", __func__);
9067         nfs41_sequence_process(task, &lgp->res.seq_res);
9068         dprintk("<-- %s\n", __func__);
9069 }
9070
9071 static int
9072 nfs4_layoutget_handle_exception(struct rpc_task *task,
9073                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9074 {
9075         struct inode *inode = lgp->args.inode;
9076         struct nfs_server *server = NFS_SERVER(inode);
9077         struct pnfs_layout_hdr *lo;
9078         int nfs4err = task->tk_status;
9079         int err, status = 0;
9080         LIST_HEAD(head);
9081
9082         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9083
9084         nfs4_sequence_free_slot(&lgp->res.seq_res);
9085
9086         switch (nfs4err) {
9087         case 0:
9088                 goto out;
9089
9090         /*
9091          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9092          * on the file. set tk_status to -ENODATA to tell upper layer to
9093          * retry go inband.
9094          */
9095         case -NFS4ERR_LAYOUTUNAVAILABLE:
9096                 status = -ENODATA;
9097                 goto out;
9098         /*
9099          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9100          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9101          */
9102         case -NFS4ERR_BADLAYOUT:
9103                 status = -EOVERFLOW;
9104                 goto out;
9105         /*
9106          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9107          * (or clients) writing to the same RAID stripe except when
9108          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9109          *
9110          * Treat it like we would RECALLCONFLICT -- we retry for a little
9111          * while, and then eventually give up.
9112          */
9113         case -NFS4ERR_LAYOUTTRYLATER:
9114                 if (lgp->args.minlength == 0) {
9115                         status = -EOVERFLOW;
9116                         goto out;
9117                 }
9118                 status = -EBUSY;
9119                 break;
9120         case -NFS4ERR_RECALLCONFLICT:
9121                 status = -ERECALLCONFLICT;
9122                 break;
9123         case -NFS4ERR_DELEG_REVOKED:
9124         case -NFS4ERR_ADMIN_REVOKED:
9125         case -NFS4ERR_EXPIRED:
9126         case -NFS4ERR_BAD_STATEID:
9127                 exception->timeout = 0;
9128                 spin_lock(&inode->i_lock);
9129                 lo = NFS_I(inode)->layout;
9130                 /* If the open stateid was bad, then recover it. */
9131                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9132                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9133                         spin_unlock(&inode->i_lock);
9134                         exception->state = lgp->args.ctx->state;
9135                         exception->stateid = &lgp->args.stateid;
9136                         break;
9137                 }
9138
9139                 /*
9140                  * Mark the bad layout state as invalid, then retry
9141                  */
9142                 pnfs_mark_layout_stateid_invalid(lo, &head);
9143                 spin_unlock(&inode->i_lock);
9144                 nfs_commit_inode(inode, 0);
9145                 pnfs_free_lseg_list(&head);
9146                 status = -EAGAIN;
9147                 goto out;
9148         }
9149
9150         err = nfs4_handle_exception(server, nfs4err, exception);
9151         if (!status) {
9152                 if (exception->retry)
9153                         status = -EAGAIN;
9154                 else
9155                         status = err;
9156         }
9157 out:
9158         dprintk("<-- %s\n", __func__);
9159         return status;
9160 }
9161
9162 size_t max_response_pages(struct nfs_server *server)
9163 {
9164         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9165         return nfs_page_array_len(0, max_resp_sz);
9166 }
9167
9168 static void nfs4_layoutget_release(void *calldata)
9169 {
9170         struct nfs4_layoutget *lgp = calldata;
9171
9172         dprintk("--> %s\n", __func__);
9173         nfs4_sequence_free_slot(&lgp->res.seq_res);
9174         pnfs_layoutget_free(lgp);
9175         dprintk("<-- %s\n", __func__);
9176 }
9177
9178 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9179         .rpc_call_prepare = nfs4_layoutget_prepare,
9180         .rpc_call_done = nfs4_layoutget_done,
9181         .rpc_release = nfs4_layoutget_release,
9182 };
9183
9184 struct pnfs_layout_segment *
9185 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9186 {
9187         struct inode *inode = lgp->args.inode;
9188         struct nfs_server *server = NFS_SERVER(inode);
9189         struct rpc_task *task;
9190         struct rpc_message msg = {
9191                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9192                 .rpc_argp = &lgp->args,
9193                 .rpc_resp = &lgp->res,
9194                 .rpc_cred = lgp->cred,
9195         };
9196         struct rpc_task_setup task_setup_data = {
9197                 .rpc_client = server->client,
9198                 .rpc_message = &msg,
9199                 .callback_ops = &nfs4_layoutget_call_ops,
9200                 .callback_data = lgp,
9201                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
9202         };
9203         struct pnfs_layout_segment *lseg = NULL;
9204         struct nfs4_exception exception = {
9205                 .inode = inode,
9206                 .timeout = *timeout,
9207         };
9208         int status = 0;
9209
9210         dprintk("--> %s\n", __func__);
9211
9212         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9213         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9214
9215         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9216
9217         task = rpc_run_task(&task_setup_data);
9218
9219         status = rpc_wait_for_completion_task(task);
9220         if (status != 0)
9221                 goto out;
9222
9223         if (task->tk_status < 0) {
9224                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9225                 *timeout = exception.timeout;
9226         } else if (lgp->res.layoutp->len == 0) {
9227                 status = -EAGAIN;
9228                 *timeout = nfs4_update_delay(&exception.timeout);
9229         } else
9230                 lseg = pnfs_layout_process(lgp);
9231 out:
9232         trace_nfs4_layoutget(lgp->args.ctx,
9233                         &lgp->args.range,
9234                         &lgp->res.range,
9235                         &lgp->res.stateid,
9236                         status);
9237
9238         rpc_put_task(task);
9239         dprintk("<-- %s status=%d\n", __func__, status);
9240         if (status)
9241                 return ERR_PTR(status);
9242         return lseg;
9243 }
9244
9245 static void
9246 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9247 {
9248         struct nfs4_layoutreturn *lrp = calldata;
9249
9250         dprintk("--> %s\n", __func__);
9251         nfs4_setup_sequence(lrp->clp,
9252                         &lrp->args.seq_args,
9253                         &lrp->res.seq_res,
9254                         task);
9255         if (!pnfs_layout_is_valid(lrp->args.layout))
9256                 rpc_exit(task, 0);
9257 }
9258
9259 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9260 {
9261         struct nfs4_layoutreturn *lrp = calldata;
9262         struct nfs_server *server;
9263
9264         dprintk("--> %s\n", __func__);
9265
9266         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9267                 return;
9268
9269         /*
9270          * Was there an RPC level error? Assume the call succeeded,
9271          * and that we need to release the layout
9272          */
9273         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9274                 lrp->res.lrs_present = 0;
9275                 return;
9276         }
9277
9278         server = NFS_SERVER(lrp->args.inode);
9279         switch (task->tk_status) {
9280         case -NFS4ERR_OLD_STATEID:
9281                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9282                                         &lrp->args.range,
9283                                         lrp->args.inode))
9284                         goto out_restart;
9285                 /* Fallthrough */
9286         default:
9287                 task->tk_status = 0;
9288                 /* Fallthrough */
9289         case 0:
9290                 break;
9291         case -NFS4ERR_DELAY:
9292                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9293                         break;
9294                 goto out_restart;
9295         }
9296         dprintk("<-- %s\n", __func__);
9297         return;
9298 out_restart:
9299         task->tk_status = 0;
9300         nfs4_sequence_free_slot(&lrp->res.seq_res);
9301         rpc_restart_call_prepare(task);
9302 }
9303
9304 static void nfs4_layoutreturn_release(void *calldata)
9305 {
9306         struct nfs4_layoutreturn *lrp = calldata;
9307         struct pnfs_layout_hdr *lo = lrp->args.layout;
9308
9309         dprintk("--> %s\n", __func__);
9310         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9311                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9312         nfs4_sequence_free_slot(&lrp->res.seq_res);
9313         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9314                 lrp->ld_private.ops->free(&lrp->ld_private);
9315         pnfs_put_layout_hdr(lrp->args.layout);
9316         nfs_iput_and_deactive(lrp->inode);
9317         put_cred(lrp->cred);
9318         kfree(calldata);
9319         dprintk("<-- %s\n", __func__);
9320 }
9321
9322 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9323         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9324         .rpc_call_done = nfs4_layoutreturn_done,
9325         .rpc_release = nfs4_layoutreturn_release,
9326 };
9327
9328 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9329 {
9330         struct rpc_task *task;
9331         struct rpc_message msg = {
9332                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9333                 .rpc_argp = &lrp->args,
9334                 .rpc_resp = &lrp->res,
9335                 .rpc_cred = lrp->cred,
9336         };
9337         struct rpc_task_setup task_setup_data = {
9338                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9339                 .rpc_message = &msg,
9340                 .callback_ops = &nfs4_layoutreturn_call_ops,
9341                 .callback_data = lrp,
9342         };
9343         int status = 0;
9344
9345         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9346                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9347                         &task_setup_data.rpc_client, &msg);
9348
9349         dprintk("--> %s\n", __func__);
9350         if (!sync) {
9351                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9352                 if (!lrp->inode) {
9353                         nfs4_layoutreturn_release(lrp);
9354                         return -EAGAIN;
9355                 }
9356                 task_setup_data.flags |= RPC_TASK_ASYNC;
9357         }
9358         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9359         task = rpc_run_task(&task_setup_data);
9360         if (IS_ERR(task))
9361                 return PTR_ERR(task);
9362         if (sync)
9363                 status = task->tk_status;
9364         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9365         dprintk("<-- %s status=%d\n", __func__, status);
9366         rpc_put_task(task);
9367         return status;
9368 }
9369
9370 static int
9371 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9372                 struct pnfs_device *pdev,
9373                 const struct cred *cred)
9374 {
9375         struct nfs4_getdeviceinfo_args args = {
9376                 .pdev = pdev,
9377                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9378                         NOTIFY_DEVICEID4_DELETE,
9379         };
9380         struct nfs4_getdeviceinfo_res res = {
9381                 .pdev = pdev,
9382         };
9383         struct rpc_message msg = {
9384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9385                 .rpc_argp = &args,
9386                 .rpc_resp = &res,
9387                 .rpc_cred = cred,
9388         };
9389         int status;
9390
9391         dprintk("--> %s\n", __func__);
9392         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9393         if (res.notification & ~args.notify_types)
9394                 dprintk("%s: unsupported notification\n", __func__);
9395         if (res.notification != args.notify_types)
9396                 pdev->nocache = 1;
9397
9398         dprintk("<-- %s status=%d\n", __func__, status);
9399
9400         return status;
9401 }
9402
9403 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9404                 struct pnfs_device *pdev,
9405                 const struct cred *cred)
9406 {
9407         struct nfs4_exception exception = { };
9408         int err;
9409
9410         do {
9411                 err = nfs4_handle_exception(server,
9412                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9413                                         &exception);
9414         } while (exception.retry);
9415         return err;
9416 }
9417 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9418
9419 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9420 {
9421         struct nfs4_layoutcommit_data *data = calldata;
9422         struct nfs_server *server = NFS_SERVER(data->args.inode);
9423
9424         nfs4_setup_sequence(server->nfs_client,
9425                         &data->args.seq_args,
9426                         &data->res.seq_res,
9427                         task);
9428 }
9429
9430 static void
9431 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9432 {
9433         struct nfs4_layoutcommit_data *data = calldata;
9434         struct nfs_server *server = NFS_SERVER(data->args.inode);
9435
9436         if (!nfs41_sequence_done(task, &data->res.seq_res))
9437                 return;
9438
9439         switch (task->tk_status) { /* Just ignore these failures */
9440         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9441         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9442         case -NFS4ERR_BADLAYOUT:     /* no layout */
9443         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9444                 task->tk_status = 0;
9445         case 0:
9446                 break;
9447         default:
9448                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9449                         rpc_restart_call_prepare(task);
9450                         return;
9451                 }
9452         }
9453 }
9454
9455 static void nfs4_layoutcommit_release(void *calldata)
9456 {
9457         struct nfs4_layoutcommit_data *data = calldata;
9458
9459         pnfs_cleanup_layoutcommit(data);
9460         nfs_post_op_update_inode_force_wcc(data->args.inode,
9461                                            data->res.fattr);
9462         put_cred(data->cred);
9463         nfs_iput_and_deactive(data->inode);
9464         kfree(data);
9465 }
9466
9467 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9468         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9469         .rpc_call_done = nfs4_layoutcommit_done,
9470         .rpc_release = nfs4_layoutcommit_release,
9471 };
9472
9473 int
9474 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9475 {
9476         struct rpc_message msg = {
9477                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9478                 .rpc_argp = &data->args,
9479                 .rpc_resp = &data->res,
9480                 .rpc_cred = data->cred,
9481         };
9482         struct rpc_task_setup task_setup_data = {
9483                 .task = &data->task,
9484                 .rpc_client = NFS_CLIENT(data->args.inode),
9485                 .rpc_message = &msg,
9486                 .callback_ops = &nfs4_layoutcommit_ops,
9487                 .callback_data = data,
9488         };
9489         struct rpc_task *task;
9490         int status = 0;
9491
9492         dprintk("NFS: initiating layoutcommit call. sync %d "
9493                 "lbw: %llu inode %lu\n", sync,
9494                 data->args.lastbytewritten,
9495                 data->args.inode->i_ino);
9496
9497         if (!sync) {
9498                 data->inode = nfs_igrab_and_active(data->args.inode);
9499                 if (data->inode == NULL) {
9500                         nfs4_layoutcommit_release(data);
9501                         return -EAGAIN;
9502                 }
9503                 task_setup_data.flags = RPC_TASK_ASYNC;
9504         }
9505         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9506         task = rpc_run_task(&task_setup_data);
9507         if (IS_ERR(task))
9508                 return PTR_ERR(task);
9509         if (sync)
9510                 status = task->tk_status;
9511         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9512         dprintk("%s: status %d\n", __func__, status);
9513         rpc_put_task(task);
9514         return status;
9515 }
9516
9517 /*
9518  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9519  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9520  */
9521 static int
9522 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9523                     struct nfs_fsinfo *info,
9524                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9525 {
9526         struct nfs41_secinfo_no_name_args args = {
9527                 .style = SECINFO_STYLE_CURRENT_FH,
9528         };
9529         struct nfs4_secinfo_res res = {
9530                 .flavors = flavors,
9531         };
9532         struct rpc_message msg = {
9533                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9534                 .rpc_argp = &args,
9535                 .rpc_resp = &res,
9536         };
9537         struct rpc_clnt *clnt = server->client;
9538         struct nfs4_call_sync_data data = {
9539                 .seq_server = server,
9540                 .seq_args = &args.seq_args,
9541                 .seq_res = &res.seq_res,
9542         };
9543         struct rpc_task_setup task_setup = {
9544                 .rpc_client = server->client,
9545                 .rpc_message = &msg,
9546                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9547                 .callback_data = &data,
9548                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9549         };
9550         const struct cred *cred = NULL;
9551         int status;
9552
9553         if (use_integrity) {
9554                 clnt = server->nfs_client->cl_rpcclient;
9555                 task_setup.rpc_client = clnt;
9556
9557                 cred = nfs4_get_clid_cred(server->nfs_client);
9558                 msg.rpc_cred = cred;
9559         }
9560
9561         dprintk("--> %s\n", __func__);
9562         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9563         status = nfs4_call_sync_custom(&task_setup);
9564         dprintk("<-- %s status=%d\n", __func__, status);
9565
9566         put_cred(cred);
9567
9568         return status;
9569 }
9570
9571 static int
9572 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9573                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9574 {
9575         struct nfs4_exception exception = {
9576                 .interruptible = true,
9577         };
9578         int err;
9579         do {
9580                 /* first try using integrity protection */
9581                 err = -NFS4ERR_WRONGSEC;
9582
9583                 /* try to use integrity protection with machine cred */
9584                 if (_nfs4_is_integrity_protected(server->nfs_client))
9585                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9586                                                           flavors, true);
9587
9588                 /*
9589                  * if unable to use integrity protection, or SECINFO with
9590                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9591                  * disallowed by spec, but exists in deployed servers) use
9592                  * the current filesystem's rpc_client and the user cred.
9593                  */
9594                 if (err == -NFS4ERR_WRONGSEC)
9595                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9596                                                           flavors, false);
9597
9598                 switch (err) {
9599                 case 0:
9600                 case -NFS4ERR_WRONGSEC:
9601                 case -ENOTSUPP:
9602                         goto out;
9603                 default:
9604                         err = nfs4_handle_exception(server, err, &exception);
9605                 }
9606         } while (exception.retry);
9607 out:
9608         return err;
9609 }
9610
9611 static int
9612 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9613                     struct nfs_fsinfo *info)
9614 {
9615         int err;
9616         struct page *page;
9617         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9618         struct nfs4_secinfo_flavors *flavors;
9619         struct nfs4_secinfo4 *secinfo;
9620         int i;
9621
9622         page = alloc_page(GFP_KERNEL);
9623         if (!page) {
9624                 err = -ENOMEM;
9625                 goto out;
9626         }
9627
9628         flavors = page_address(page);
9629         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9630
9631         /*
9632          * Fall back on "guess and check" method if
9633          * the server doesn't support SECINFO_NO_NAME
9634          */
9635         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9636                 err = nfs4_find_root_sec(server, fhandle, info);
9637                 goto out_freepage;
9638         }
9639         if (err)
9640                 goto out_freepage;
9641
9642         for (i = 0; i < flavors->num_flavors; i++) {
9643                 secinfo = &flavors->flavors[i];
9644
9645                 switch (secinfo->flavor) {
9646                 case RPC_AUTH_NULL:
9647                 case RPC_AUTH_UNIX:
9648                 case RPC_AUTH_GSS:
9649                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9650                                         &secinfo->flavor_info);
9651                         break;
9652                 default:
9653                         flavor = RPC_AUTH_MAXFLAVOR;
9654                         break;
9655                 }
9656
9657                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9658                         flavor = RPC_AUTH_MAXFLAVOR;
9659
9660                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9661                         err = nfs4_lookup_root_sec(server, fhandle,
9662                                                    info, flavor);
9663                         if (!err)
9664                                 break;
9665                 }
9666         }
9667
9668         if (flavor == RPC_AUTH_MAXFLAVOR)
9669                 err = -EPERM;
9670
9671 out_freepage:
9672         put_page(page);
9673         if (err == -EACCES)
9674                 return -EPERM;
9675 out:
9676         return err;
9677 }
9678
9679 static int _nfs41_test_stateid(struct nfs_server *server,
9680                 nfs4_stateid *stateid,
9681                 const struct cred *cred)
9682 {
9683         int status;
9684         struct nfs41_test_stateid_args args = {
9685                 .stateid = stateid,
9686         };
9687         struct nfs41_test_stateid_res res;
9688         struct rpc_message msg = {
9689                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9690                 .rpc_argp = &args,
9691                 .rpc_resp = &res,
9692                 .rpc_cred = cred,
9693         };
9694         struct rpc_clnt *rpc_client = server->client;
9695
9696         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9697                 &rpc_client, &msg);
9698
9699         dprintk("NFS call  test_stateid %p\n", stateid);
9700         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9701         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9702                         &args.seq_args, &res.seq_res);
9703         if (status != NFS_OK) {
9704                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9705                 return status;
9706         }
9707         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9708         return -res.status;
9709 }
9710
9711 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9712                 int err, struct nfs4_exception *exception)
9713 {
9714         exception->retry = 0;
9715         switch(err) {
9716         case -NFS4ERR_DELAY:
9717         case -NFS4ERR_RETRY_UNCACHED_REP:
9718                 nfs4_handle_exception(server, err, exception);
9719                 break;
9720         case -NFS4ERR_BADSESSION:
9721         case -NFS4ERR_BADSLOT:
9722         case -NFS4ERR_BAD_HIGH_SLOT:
9723         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9724         case -NFS4ERR_DEADSESSION:
9725                 nfs4_do_handle_exception(server, err, exception);
9726         }
9727 }
9728
9729 /**
9730  * nfs41_test_stateid - perform a TEST_STATEID operation
9731  *
9732  * @server: server / transport on which to perform the operation
9733  * @stateid: state ID to test
9734  * @cred: credential
9735  *
9736  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9737  * Otherwise a negative NFS4ERR value is returned if the operation
9738  * failed or the state ID is not currently valid.
9739  */
9740 static int nfs41_test_stateid(struct nfs_server *server,
9741                 nfs4_stateid *stateid,
9742                 const struct cred *cred)
9743 {
9744         struct nfs4_exception exception = {
9745                 .interruptible = true,
9746         };
9747         int err;
9748         do {
9749                 err = _nfs41_test_stateid(server, stateid, cred);
9750                 nfs4_handle_delay_or_session_error(server, err, &exception);
9751         } while (exception.retry);
9752         return err;
9753 }
9754
9755 struct nfs_free_stateid_data {
9756         struct nfs_server *server;
9757         struct nfs41_free_stateid_args args;
9758         struct nfs41_free_stateid_res res;
9759 };
9760
9761 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9762 {
9763         struct nfs_free_stateid_data *data = calldata;
9764         nfs4_setup_sequence(data->server->nfs_client,
9765                         &data->args.seq_args,
9766                         &data->res.seq_res,
9767                         task);
9768 }
9769
9770 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9771 {
9772         struct nfs_free_stateid_data *data = calldata;
9773
9774         nfs41_sequence_done(task, &data->res.seq_res);
9775
9776         switch (task->tk_status) {
9777         case -NFS4ERR_DELAY:
9778                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9779                         rpc_restart_call_prepare(task);
9780         }
9781 }
9782
9783 static void nfs41_free_stateid_release(void *calldata)
9784 {
9785         kfree(calldata);
9786 }
9787
9788 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9789         .rpc_call_prepare = nfs41_free_stateid_prepare,
9790         .rpc_call_done = nfs41_free_stateid_done,
9791         .rpc_release = nfs41_free_stateid_release,
9792 };
9793
9794 /**
9795  * nfs41_free_stateid - perform a FREE_STATEID operation
9796  *
9797  * @server: server / transport on which to perform the operation
9798  * @stateid: state ID to release
9799  * @cred: credential
9800  * @privileged: set to true if this call needs to be privileged
9801  *
9802  * Note: this function is always asynchronous.
9803  */
9804 static int nfs41_free_stateid(struct nfs_server *server,
9805                 const nfs4_stateid *stateid,
9806                 const struct cred *cred,
9807                 bool privileged)
9808 {
9809         struct rpc_message msg = {
9810                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9811                 .rpc_cred = cred,
9812         };
9813         struct rpc_task_setup task_setup = {
9814                 .rpc_client = server->client,
9815                 .rpc_message = &msg,
9816                 .callback_ops = &nfs41_free_stateid_ops,
9817                 .flags = RPC_TASK_ASYNC,
9818         };
9819         struct nfs_free_stateid_data *data;
9820         struct rpc_task *task;
9821
9822         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9823                 &task_setup.rpc_client, &msg);
9824
9825         dprintk("NFS call  free_stateid %p\n", stateid);
9826         data = kmalloc(sizeof(*data), GFP_NOFS);
9827         if (!data)
9828                 return -ENOMEM;
9829         data->server = server;
9830         nfs4_stateid_copy(&data->args.stateid, stateid);
9831
9832         task_setup.callback_data = data;
9833
9834         msg.rpc_argp = &data->args;
9835         msg.rpc_resp = &data->res;
9836         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9837         task = rpc_run_task(&task_setup);
9838         if (IS_ERR(task))
9839                 return PTR_ERR(task);
9840         rpc_put_task(task);
9841         return 0;
9842 }
9843
9844 static void
9845 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9846 {
9847         const struct cred *cred = lsp->ls_state->owner->so_cred;
9848
9849         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9850         nfs4_free_lock_state(server, lsp);
9851 }
9852
9853 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9854                 const nfs4_stateid *s2)
9855 {
9856         if (s1->type != s2->type)
9857                 return false;
9858
9859         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9860                 return false;
9861
9862         if (s1->seqid == s2->seqid)
9863                 return true;
9864
9865         return s1->seqid == 0 || s2->seqid == 0;
9866 }
9867
9868 #endif /* CONFIG_NFS_V4_1 */
9869
9870 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9871                 const nfs4_stateid *s2)
9872 {
9873         return nfs4_stateid_match(s1, s2);
9874 }
9875
9876
9877 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9878         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9879         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9880         .recover_open   = nfs4_open_reclaim,
9881         .recover_lock   = nfs4_lock_reclaim,
9882         .establish_clid = nfs4_init_clientid,
9883         .detect_trunking = nfs40_discover_server_trunking,
9884 };
9885
9886 #if defined(CONFIG_NFS_V4_1)
9887 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9888         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9889         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9890         .recover_open   = nfs4_open_reclaim,
9891         .recover_lock   = nfs4_lock_reclaim,
9892         .establish_clid = nfs41_init_clientid,
9893         .reclaim_complete = nfs41_proc_reclaim_complete,
9894         .detect_trunking = nfs41_discover_server_trunking,
9895 };
9896 #endif /* CONFIG_NFS_V4_1 */
9897
9898 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9899         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9900         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9901         .recover_open   = nfs40_open_expired,
9902         .recover_lock   = nfs4_lock_expired,
9903         .establish_clid = nfs4_init_clientid,
9904 };
9905
9906 #if defined(CONFIG_NFS_V4_1)
9907 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9908         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9909         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9910         .recover_open   = nfs41_open_expired,
9911         .recover_lock   = nfs41_lock_expired,
9912         .establish_clid = nfs41_init_clientid,
9913 };
9914 #endif /* CONFIG_NFS_V4_1 */
9915
9916 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9917         .sched_state_renewal = nfs4_proc_async_renew,
9918         .get_state_renewal_cred = nfs4_get_renew_cred,
9919         .renew_lease = nfs4_proc_renew,
9920 };
9921
9922 #if defined(CONFIG_NFS_V4_1)
9923 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9924         .sched_state_renewal = nfs41_proc_async_sequence,
9925         .get_state_renewal_cred = nfs4_get_machine_cred,
9926         .renew_lease = nfs4_proc_sequence,
9927 };
9928 #endif
9929
9930 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9931         .get_locations = _nfs40_proc_get_locations,
9932         .fsid_present = _nfs40_proc_fsid_present,
9933 };
9934
9935 #if defined(CONFIG_NFS_V4_1)
9936 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9937         .get_locations = _nfs41_proc_get_locations,
9938         .fsid_present = _nfs41_proc_fsid_present,
9939 };
9940 #endif  /* CONFIG_NFS_V4_1 */
9941
9942 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9943         .minor_version = 0,
9944         .init_caps = NFS_CAP_READDIRPLUS
9945                 | NFS_CAP_ATOMIC_OPEN
9946                 | NFS_CAP_POSIX_LOCK,
9947         .init_client = nfs40_init_client,
9948         .shutdown_client = nfs40_shutdown_client,
9949         .match_stateid = nfs4_match_stateid,
9950         .find_root_sec = nfs4_find_root_sec,
9951         .free_lock_state = nfs4_release_lockowner,
9952         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9953         .alloc_seqid = nfs_alloc_seqid,
9954         .call_sync_ops = &nfs40_call_sync_ops,
9955         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9956         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9957         .state_renewal_ops = &nfs40_state_renewal_ops,
9958         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9959 };
9960
9961 #if defined(CONFIG_NFS_V4_1)
9962 static struct nfs_seqid *
9963 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9964 {
9965         return NULL;
9966 }
9967
9968 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9969         .minor_version = 1,
9970         .init_caps = NFS_CAP_READDIRPLUS
9971                 | NFS_CAP_ATOMIC_OPEN
9972                 | NFS_CAP_POSIX_LOCK
9973                 | NFS_CAP_STATEID_NFSV41
9974                 | NFS_CAP_ATOMIC_OPEN_V1
9975                 | NFS_CAP_LGOPEN,
9976         .init_client = nfs41_init_client,
9977         .shutdown_client = nfs41_shutdown_client,
9978         .match_stateid = nfs41_match_stateid,
9979         .find_root_sec = nfs41_find_root_sec,
9980         .free_lock_state = nfs41_free_lock_state,
9981         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9982         .alloc_seqid = nfs_alloc_no_seqid,
9983         .session_trunk = nfs4_test_session_trunk,
9984         .call_sync_ops = &nfs41_call_sync_ops,
9985         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9986         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9987         .state_renewal_ops = &nfs41_state_renewal_ops,
9988         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9989 };
9990 #endif
9991
9992 #if defined(CONFIG_NFS_V4_2)
9993 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9994         .minor_version = 2,
9995         .init_caps = NFS_CAP_READDIRPLUS
9996                 | NFS_CAP_ATOMIC_OPEN
9997                 | NFS_CAP_POSIX_LOCK
9998                 | NFS_CAP_STATEID_NFSV41
9999                 | NFS_CAP_ATOMIC_OPEN_V1
10000                 | NFS_CAP_LGOPEN
10001                 | NFS_CAP_ALLOCATE
10002                 | NFS_CAP_COPY
10003                 | NFS_CAP_OFFLOAD_CANCEL
10004                 | NFS_CAP_COPY_NOTIFY
10005                 | NFS_CAP_DEALLOCATE
10006                 | NFS_CAP_SEEK
10007                 | NFS_CAP_LAYOUTSTATS
10008                 | NFS_CAP_CLONE
10009                 | NFS_CAP_LAYOUTERROR,
10010         .init_client = nfs41_init_client,
10011         .shutdown_client = nfs41_shutdown_client,
10012         .match_stateid = nfs41_match_stateid,
10013         .find_root_sec = nfs41_find_root_sec,
10014         .free_lock_state = nfs41_free_lock_state,
10015         .call_sync_ops = &nfs41_call_sync_ops,
10016         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10017         .alloc_seqid = nfs_alloc_no_seqid,
10018         .session_trunk = nfs4_test_session_trunk,
10019         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10020         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10021         .state_renewal_ops = &nfs41_state_renewal_ops,
10022         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10023 };
10024 #endif
10025
10026 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10027         [0] = &nfs_v4_0_minor_ops,
10028 #if defined(CONFIG_NFS_V4_1)
10029         [1] = &nfs_v4_1_minor_ops,
10030 #endif
10031 #if defined(CONFIG_NFS_V4_2)
10032         [2] = &nfs_v4_2_minor_ops,
10033 #endif
10034 };
10035
10036 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10037 {
10038         ssize_t error, error2;
10039
10040         error = generic_listxattr(dentry, list, size);
10041         if (error < 0)
10042                 return error;
10043         if (list) {
10044                 list += error;
10045                 size -= error;
10046         }
10047
10048         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10049         if (error2 < 0)
10050                 return error2;
10051         return error + error2;
10052 }
10053
10054 static const struct inode_operations nfs4_dir_inode_operations = {
10055         .create         = nfs_create,
10056         .lookup         = nfs_lookup,
10057         .atomic_open    = nfs_atomic_open,
10058         .link           = nfs_link,
10059         .unlink         = nfs_unlink,
10060         .symlink        = nfs_symlink,
10061         .mkdir          = nfs_mkdir,
10062         .rmdir          = nfs_rmdir,
10063         .mknod          = nfs_mknod,
10064         .rename         = nfs_rename,
10065         .permission     = nfs_permission,
10066         .getattr        = nfs_getattr,
10067         .setattr        = nfs_setattr,
10068         .listxattr      = nfs4_listxattr,
10069 };
10070
10071 static const struct inode_operations nfs4_file_inode_operations = {
10072         .permission     = nfs_permission,
10073         .getattr        = nfs_getattr,
10074         .setattr        = nfs_setattr,
10075         .listxattr      = nfs4_listxattr,
10076 };
10077
10078 const struct nfs_rpc_ops nfs_v4_clientops = {
10079         .version        = 4,                    /* protocol version */
10080         .dentry_ops     = &nfs4_dentry_operations,
10081         .dir_inode_ops  = &nfs4_dir_inode_operations,
10082         .file_inode_ops = &nfs4_file_inode_operations,
10083         .file_ops       = &nfs4_file_operations,
10084         .getroot        = nfs4_proc_get_root,
10085         .submount       = nfs4_submount,
10086         .try_get_tree   = nfs4_try_get_tree,
10087         .getattr        = nfs4_proc_getattr,
10088         .setattr        = nfs4_proc_setattr,
10089         .lookup         = nfs4_proc_lookup,
10090         .lookupp        = nfs4_proc_lookupp,
10091         .access         = nfs4_proc_access,
10092         .readlink       = nfs4_proc_readlink,
10093         .create         = nfs4_proc_create,
10094         .remove         = nfs4_proc_remove,
10095         .unlink_setup   = nfs4_proc_unlink_setup,
10096         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10097         .unlink_done    = nfs4_proc_unlink_done,
10098         .rename_setup   = nfs4_proc_rename_setup,
10099         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10100         .rename_done    = nfs4_proc_rename_done,
10101         .link           = nfs4_proc_link,
10102         .symlink        = nfs4_proc_symlink,
10103         .mkdir          = nfs4_proc_mkdir,
10104         .rmdir          = nfs4_proc_rmdir,
10105         .readdir        = nfs4_proc_readdir,
10106         .mknod          = nfs4_proc_mknod,
10107         .statfs         = nfs4_proc_statfs,
10108         .fsinfo         = nfs4_proc_fsinfo,
10109         .pathconf       = nfs4_proc_pathconf,
10110         .set_capabilities = nfs4_server_capabilities,
10111         .decode_dirent  = nfs4_decode_dirent,
10112         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10113         .read_setup     = nfs4_proc_read_setup,
10114         .read_done      = nfs4_read_done,
10115         .write_setup    = nfs4_proc_write_setup,
10116         .write_done     = nfs4_write_done,
10117         .commit_setup   = nfs4_proc_commit_setup,
10118         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10119         .commit_done    = nfs4_commit_done,
10120         .lock           = nfs4_proc_lock,
10121         .clear_acl_cache = nfs4_zap_acl_attr,
10122         .close_context  = nfs4_close_context,
10123         .open_context   = nfs4_atomic_open,
10124         .have_delegation = nfs4_have_delegation,
10125         .alloc_client   = nfs4_alloc_client,
10126         .init_client    = nfs4_init_client,
10127         .free_client    = nfs4_free_client,
10128         .create_server  = nfs4_create_server,
10129         .clone_server   = nfs_clone_server,
10130 };
10131
10132 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10133         .name   = XATTR_NAME_NFSV4_ACL,
10134         .list   = nfs4_xattr_list_nfs4_acl,
10135         .get    = nfs4_xattr_get_nfs4_acl,
10136         .set    = nfs4_xattr_set_nfs4_acl,
10137 };
10138
10139 const struct xattr_handler *nfs4_xattr_handlers[] = {
10140         &nfs4_xattr_nfs4_acl_handler,
10141 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10142         &nfs4_xattr_nfs4_label_handler,
10143 #endif
10144         NULL
10145 };
10146
10147 /*
10148  * Local variables:
10149  *  c-basic-offset: 8
10150  * End:
10151  */