OSDN Git Service

Merge tag '5.6-rc-smb3-plugfest-patches' of git://git.samba.org/sfrench/cifs-2.6
[tomoyo/tomoyo-test1.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(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 int nfs41_sequence_process(struct rpc_task *task,
778                 struct nfs4_sequence_res *res)
779 {
780         struct nfs4_session *session;
781         struct nfs4_slot *slot = res->sr_slot;
782         struct nfs_client *clp;
783         int status;
784         int ret = 1;
785
786         if (slot == NULL)
787                 goto out_noaction;
788         /* don't increment the sequence number if the task wasn't sent */
789         if (!RPC_WAS_SENT(task) || slot->seq_done)
790                 goto out;
791
792         session = slot->table->session;
793
794         trace_nfs4_sequence_done(session, res);
795
796         status = res->sr_status;
797         if (task->tk_status == -NFS4ERR_DEADSESSION)
798                 status = -NFS4ERR_DEADSESSION;
799
800         /* Check the SEQUENCE operation status */
801         switch (status) {
802         case 0:
803                 /* Mark this sequence number as having been acked */
804                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
805                 /* Update the slot's sequence and clientid lease timer */
806                 slot->seq_done = 1;
807                 clp = session->clp;
808                 do_renew_lease(clp, res->sr_timestamp);
809                 /* Check sequence flags */
810                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
811                                 !!slot->privileged);
812                 nfs41_update_target_slotid(slot->table, slot, res);
813                 break;
814         case 1:
815                 /*
816                  * sr_status remains 1 if an RPC level error occurred.
817                  * The server may or may not have processed the sequence
818                  * operation..
819                  */
820                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
821                 slot->seq_done = 1;
822                 goto out;
823         case -NFS4ERR_DELAY:
824                 /* The server detected a resend of the RPC call and
825                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
826                  * of RFC5661.
827                  */
828                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
829                         __func__,
830                         slot->slot_nr,
831                         slot->seq_nr);
832                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
833                 goto out_retry;
834         case -NFS4ERR_RETRY_UNCACHED_REP:
835         case -NFS4ERR_SEQ_FALSE_RETRY:
836                 /*
837                  * The server thinks we tried to replay a request.
838                  * Retry the call after bumping the sequence ID.
839                  */
840                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
841                 goto retry_new_seq;
842         case -NFS4ERR_BADSLOT:
843                 /*
844                  * The slot id we used was probably retired. Try again
845                  * using a different slot id.
846                  */
847                 if (slot->slot_nr < slot->table->target_highest_slotid)
848                         goto session_recover;
849                 goto retry_nowait;
850         case -NFS4ERR_SEQ_MISORDERED:
851                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
852                 /*
853                  * Were one or more calls using this slot interrupted?
854                  * If the server never received the request, then our
855                  * transmitted slot sequence number may be too high.
856                  */
857                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
858                         slot->seq_nr--;
859                         goto retry_nowait;
860                 }
861                 /*
862                  * RFC5661:
863                  * A retry might be sent while the original request is
864                  * still in progress on the replier. The replier SHOULD
865                  * deal with the issue by returning NFS4ERR_DELAY as the
866                  * reply to SEQUENCE or CB_SEQUENCE operation, but
867                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
868                  *
869                  * Restart the search after a delay.
870                  */
871                 slot->seq_nr = slot->seq_nr_highest_sent;
872                 goto out_retry;
873         case -NFS4ERR_BADSESSION:
874         case -NFS4ERR_DEADSESSION:
875         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
876                 goto session_recover;
877         default:
878                 /* Just update the slot sequence no. */
879                 slot->seq_done = 1;
880         }
881 out:
882         /* The session may be reset by one of the error handlers. */
883         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
884 out_noaction:
885         return ret;
886 session_recover:
887         nfs4_schedule_session_recovery(session, status);
888         dprintk("%s ERROR: %d Reset session\n", __func__, status);
889         nfs41_sequence_free_slot(res);
890         goto out;
891 retry_new_seq:
892         ++slot->seq_nr;
893 retry_nowait:
894         if (rpc_restart_call_prepare(task)) {
895                 nfs41_sequence_free_slot(res);
896                 task->tk_status = 0;
897                 ret = 0;
898         }
899         goto out;
900 out_retry:
901         if (!rpc_restart_call(task))
902                 goto out;
903         rpc_delay(task, NFS4_POLL_RETRY_MAX);
904         return 0;
905 }
906
907 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
908 {
909         if (!nfs41_sequence_process(task, res))
910                 return 0;
911         if (res->sr_slot != NULL)
912                 nfs41_sequence_free_slot(res);
913         return 1;
914
915 }
916 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
917
918 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
919 {
920         if (res->sr_slot == NULL)
921                 return 1;
922         if (res->sr_slot->table->session != NULL)
923                 return nfs41_sequence_process(task, res);
924         return nfs40_sequence_done(task, res);
925 }
926
927 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
928 {
929         if (res->sr_slot != NULL) {
930                 if (res->sr_slot->table->session != NULL)
931                         nfs41_sequence_free_slot(res);
932                 else
933                         nfs40_sequence_free_slot(res);
934         }
935 }
936
937 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
938 {
939         if (res->sr_slot == NULL)
940                 return 1;
941         if (!res->sr_slot->table->session)
942                 return nfs40_sequence_done(task, res);
943         return nfs41_sequence_done(task, res);
944 }
945 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
946
947 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
948 {
949         struct nfs4_call_sync_data *data = calldata;
950
951         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
952
953         nfs4_setup_sequence(data->seq_server->nfs_client,
954                             data->seq_args, data->seq_res, task);
955 }
956
957 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
958 {
959         struct nfs4_call_sync_data *data = calldata;
960
961         nfs41_sequence_done(task, data->seq_res);
962 }
963
964 static const struct rpc_call_ops nfs41_call_sync_ops = {
965         .rpc_call_prepare = nfs41_call_sync_prepare,
966         .rpc_call_done = nfs41_call_sync_done,
967 };
968
969 #else   /* !CONFIG_NFS_V4_1 */
970
971 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
972 {
973         return nfs40_sequence_done(task, res);
974 }
975
976 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
977 {
978         if (res->sr_slot != NULL)
979                 nfs40_sequence_free_slot(res);
980 }
981
982 int nfs4_sequence_done(struct rpc_task *task,
983                        struct nfs4_sequence_res *res)
984 {
985         return nfs40_sequence_done(task, res);
986 }
987 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
988
989 #endif  /* !CONFIG_NFS_V4_1 */
990
991 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
992 {
993         res->sr_timestamp = jiffies;
994         res->sr_status_flags = 0;
995         res->sr_status = 1;
996 }
997
998 static
999 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1000                 struct nfs4_sequence_res *res,
1001                 struct nfs4_slot *slot)
1002 {
1003         if (!slot)
1004                 return;
1005         slot->privileged = args->sa_privileged ? 1 : 0;
1006         args->sa_slot = slot;
1007
1008         res->sr_slot = slot;
1009 }
1010
1011 int nfs4_setup_sequence(struct nfs_client *client,
1012                         struct nfs4_sequence_args *args,
1013                         struct nfs4_sequence_res *res,
1014                         struct rpc_task *task)
1015 {
1016         struct nfs4_session *session = nfs4_get_session(client);
1017         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1018         struct nfs4_slot *slot;
1019
1020         /* slot already allocated? */
1021         if (res->sr_slot != NULL)
1022                 goto out_start;
1023
1024         if (session)
1025                 tbl = &session->fc_slot_table;
1026
1027         spin_lock(&tbl->slot_tbl_lock);
1028         /* The state manager will wait until the slot table is empty */
1029         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1030                 goto out_sleep;
1031
1032         slot = nfs4_alloc_slot(tbl);
1033         if (IS_ERR(slot)) {
1034                 if (slot == ERR_PTR(-ENOMEM))
1035                         goto out_sleep_timeout;
1036                 goto out_sleep;
1037         }
1038         spin_unlock(&tbl->slot_tbl_lock);
1039
1040         nfs4_sequence_attach_slot(args, res, slot);
1041
1042         trace_nfs4_setup_sequence(session, args);
1043 out_start:
1044         nfs41_sequence_res_init(res);
1045         rpc_call_start(task);
1046         return 0;
1047 out_sleep_timeout:
1048         /* Try again in 1/4 second */
1049         if (args->sa_privileged)
1050                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1051                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1052         else
1053                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1054                                 NULL, jiffies + (HZ >> 2));
1055         spin_unlock(&tbl->slot_tbl_lock);
1056         return -EAGAIN;
1057 out_sleep:
1058         if (args->sa_privileged)
1059                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1060                                 RPC_PRIORITY_PRIVILEGED);
1061         else
1062                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1063         spin_unlock(&tbl->slot_tbl_lock);
1064         return -EAGAIN;
1065 }
1066 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1067
1068 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1069 {
1070         struct nfs4_call_sync_data *data = calldata;
1071         nfs4_setup_sequence(data->seq_server->nfs_client,
1072                                 data->seq_args, data->seq_res, task);
1073 }
1074
1075 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1076 {
1077         struct nfs4_call_sync_data *data = calldata;
1078         nfs4_sequence_done(task, data->seq_res);
1079 }
1080
1081 static const struct rpc_call_ops nfs40_call_sync_ops = {
1082         .rpc_call_prepare = nfs40_call_sync_prepare,
1083         .rpc_call_done = nfs40_call_sync_done,
1084 };
1085
1086 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1087 {
1088         int ret;
1089         struct rpc_task *task;
1090
1091         task = rpc_run_task(task_setup);
1092         if (IS_ERR(task))
1093                 return PTR_ERR(task);
1094
1095         ret = task->tk_status;
1096         rpc_put_task(task);
1097         return ret;
1098 }
1099
1100 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1101                              struct nfs_server *server,
1102                              struct rpc_message *msg,
1103                              struct nfs4_sequence_args *args,
1104                              struct nfs4_sequence_res *res,
1105                              unsigned short task_flags)
1106 {
1107         struct nfs_client *clp = server->nfs_client;
1108         struct nfs4_call_sync_data data = {
1109                 .seq_server = server,
1110                 .seq_args = args,
1111                 .seq_res = res,
1112         };
1113         struct rpc_task_setup task_setup = {
1114                 .rpc_client = clnt,
1115                 .rpc_message = msg,
1116                 .callback_ops = clp->cl_mvops->call_sync_ops,
1117                 .callback_data = &data,
1118                 .flags = task_flags,
1119         };
1120
1121         return nfs4_call_sync_custom(&task_setup);
1122 }
1123
1124 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1125                                    struct nfs_server *server,
1126                                    struct rpc_message *msg,
1127                                    struct nfs4_sequence_args *args,
1128                                    struct nfs4_sequence_res *res)
1129 {
1130         return nfs4_do_call_sync(clnt, server, msg, args, res, 0);
1131 }
1132
1133
1134 int nfs4_call_sync(struct rpc_clnt *clnt,
1135                    struct nfs_server *server,
1136                    struct rpc_message *msg,
1137                    struct nfs4_sequence_args *args,
1138                    struct nfs4_sequence_res *res,
1139                    int cache_reply)
1140 {
1141         nfs4_init_sequence(args, res, cache_reply, 0);
1142         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1143 }
1144
1145 static void
1146 nfs4_inc_nlink_locked(struct inode *inode)
1147 {
1148         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1149         inc_nlink(inode);
1150 }
1151
1152 static void
1153 nfs4_dec_nlink_locked(struct inode *inode)
1154 {
1155         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1156         drop_nlink(inode);
1157 }
1158
1159 static void
1160 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1161                 unsigned long timestamp, unsigned long cache_validity)
1162 {
1163         struct nfs_inode *nfsi = NFS_I(dir);
1164
1165         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1166                 | NFS_INO_INVALID_MTIME
1167                 | NFS_INO_INVALID_DATA
1168                 | cache_validity;
1169         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1170                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1171                 nfsi->attrtimeo_timestamp = jiffies;
1172         } else {
1173                 nfs_force_lookup_revalidate(dir);
1174                 if (cinfo->before != inode_peek_iversion_raw(dir))
1175                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1176                                 NFS_INO_INVALID_ACL;
1177         }
1178         inode_set_iversion_raw(dir, cinfo->after);
1179         nfsi->read_cache_jiffies = timestamp;
1180         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1181         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1182         nfs_fscache_invalidate(dir);
1183 }
1184
1185 static void
1186 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1187                 unsigned long timestamp, unsigned long cache_validity)
1188 {
1189         spin_lock(&dir->i_lock);
1190         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1191         spin_unlock(&dir->i_lock);
1192 }
1193
1194 struct nfs4_open_createattrs {
1195         struct nfs4_label *label;
1196         struct iattr *sattr;
1197         const __u32 verf[2];
1198 };
1199
1200 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1201                 int err, struct nfs4_exception *exception)
1202 {
1203         if (err != -EINVAL)
1204                 return false;
1205         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1206                 return false;
1207         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1208         exception->retry = 1;
1209         return true;
1210 }
1211
1212 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1213 {
1214          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1215 }
1216
1217 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1218 {
1219         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1220
1221         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1222 }
1223
1224 static u32
1225 nfs4_map_atomic_open_share(struct nfs_server *server,
1226                 fmode_t fmode, int openflags)
1227 {
1228         u32 res = 0;
1229
1230         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1231         case FMODE_READ:
1232                 res = NFS4_SHARE_ACCESS_READ;
1233                 break;
1234         case FMODE_WRITE:
1235                 res = NFS4_SHARE_ACCESS_WRITE;
1236                 break;
1237         case FMODE_READ|FMODE_WRITE:
1238                 res = NFS4_SHARE_ACCESS_BOTH;
1239         }
1240         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1241                 goto out;
1242         /* Want no delegation if we're using O_DIRECT */
1243         if (openflags & O_DIRECT)
1244                 res |= NFS4_SHARE_WANT_NO_DELEG;
1245 out:
1246         return res;
1247 }
1248
1249 static enum open_claim_type4
1250 nfs4_map_atomic_open_claim(struct nfs_server *server,
1251                 enum open_claim_type4 claim)
1252 {
1253         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1254                 return claim;
1255         switch (claim) {
1256         default:
1257                 return claim;
1258         case NFS4_OPEN_CLAIM_FH:
1259                 return NFS4_OPEN_CLAIM_NULL;
1260         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1261                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1262         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1263                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1264         }
1265 }
1266
1267 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1268 {
1269         p->o_res.f_attr = &p->f_attr;
1270         p->o_res.f_label = p->f_label;
1271         p->o_res.seqid = p->o_arg.seqid;
1272         p->c_res.seqid = p->c_arg.seqid;
1273         p->o_res.server = p->o_arg.server;
1274         p->o_res.access_request = p->o_arg.access;
1275         nfs_fattr_init(&p->f_attr);
1276         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1277 }
1278
1279 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1280                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1281                 const struct nfs4_open_createattrs *c,
1282                 enum open_claim_type4 claim,
1283                 gfp_t gfp_mask)
1284 {
1285         struct dentry *parent = dget_parent(dentry);
1286         struct inode *dir = d_inode(parent);
1287         struct nfs_server *server = NFS_SERVER(dir);
1288         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1289         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1290         struct nfs4_opendata *p;
1291
1292         p = kzalloc(sizeof(*p), gfp_mask);
1293         if (p == NULL)
1294                 goto err;
1295
1296         p->f_label = nfs4_label_alloc(server, gfp_mask);
1297         if (IS_ERR(p->f_label))
1298                 goto err_free_p;
1299
1300         p->a_label = nfs4_label_alloc(server, gfp_mask);
1301         if (IS_ERR(p->a_label))
1302                 goto err_free_f;
1303
1304         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1305         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1306         if (IS_ERR(p->o_arg.seqid))
1307                 goto err_free_label;
1308         nfs_sb_active(dentry->d_sb);
1309         p->dentry = dget(dentry);
1310         p->dir = parent;
1311         p->owner = sp;
1312         atomic_inc(&sp->so_count);
1313         p->o_arg.open_flags = flags;
1314         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1315         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1316         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1317                         fmode, flags);
1318         if (flags & O_CREAT) {
1319                 p->o_arg.umask = current_umask();
1320                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1321                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1322                         p->o_arg.u.attrs = &p->attrs;
1323                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1324
1325                         memcpy(p->o_arg.u.verifier.data, c->verf,
1326                                         sizeof(p->o_arg.u.verifier.data));
1327                 }
1328         }
1329         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1330          * will return permission denied for all bits until close */
1331         if (!(flags & O_EXCL)) {
1332                 /* ask server to check for all possible rights as results
1333                  * are cached */
1334                 switch (p->o_arg.claim) {
1335                 default:
1336                         break;
1337                 case NFS4_OPEN_CLAIM_NULL:
1338                 case NFS4_OPEN_CLAIM_FH:
1339                         p->o_arg.access = NFS4_ACCESS_READ |
1340                                 NFS4_ACCESS_MODIFY |
1341                                 NFS4_ACCESS_EXTEND |
1342                                 NFS4_ACCESS_EXECUTE;
1343                 }
1344         }
1345         p->o_arg.clientid = server->nfs_client->cl_clientid;
1346         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1347         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1348         p->o_arg.name = &dentry->d_name;
1349         p->o_arg.server = server;
1350         p->o_arg.bitmask = nfs4_bitmask(server, label);
1351         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1352         switch (p->o_arg.claim) {
1353         case NFS4_OPEN_CLAIM_NULL:
1354         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1355         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1356                 p->o_arg.fh = NFS_FH(dir);
1357                 break;
1358         case NFS4_OPEN_CLAIM_PREVIOUS:
1359         case NFS4_OPEN_CLAIM_FH:
1360         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1361         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1362                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1363         }
1364         p->c_arg.fh = &p->o_res.fh;
1365         p->c_arg.stateid = &p->o_res.stateid;
1366         p->c_arg.seqid = p->o_arg.seqid;
1367         nfs4_init_opendata_res(p);
1368         kref_init(&p->kref);
1369         return p;
1370
1371 err_free_label:
1372         nfs4_label_free(p->a_label);
1373 err_free_f:
1374         nfs4_label_free(p->f_label);
1375 err_free_p:
1376         kfree(p);
1377 err:
1378         dput(parent);
1379         return NULL;
1380 }
1381
1382 static void nfs4_opendata_free(struct kref *kref)
1383 {
1384         struct nfs4_opendata *p = container_of(kref,
1385                         struct nfs4_opendata, kref);
1386         struct super_block *sb = p->dentry->d_sb;
1387
1388         nfs4_lgopen_release(p->lgp);
1389         nfs_free_seqid(p->o_arg.seqid);
1390         nfs4_sequence_free_slot(&p->o_res.seq_res);
1391         if (p->state != NULL)
1392                 nfs4_put_open_state(p->state);
1393         nfs4_put_state_owner(p->owner);
1394
1395         nfs4_label_free(p->a_label);
1396         nfs4_label_free(p->f_label);
1397
1398         dput(p->dir);
1399         dput(p->dentry);
1400         nfs_sb_deactive(sb);
1401         nfs_fattr_free_names(&p->f_attr);
1402         kfree(p->f_attr.mdsthreshold);
1403         kfree(p);
1404 }
1405
1406 static void nfs4_opendata_put(struct nfs4_opendata *p)
1407 {
1408         if (p != NULL)
1409                 kref_put(&p->kref, nfs4_opendata_free);
1410 }
1411
1412 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1413                 fmode_t fmode)
1414 {
1415         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1416         case FMODE_READ|FMODE_WRITE:
1417                 return state->n_rdwr != 0;
1418         case FMODE_WRITE:
1419                 return state->n_wronly != 0;
1420         case FMODE_READ:
1421                 return state->n_rdonly != 0;
1422         }
1423         WARN_ON_ONCE(1);
1424         return false;
1425 }
1426
1427 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1428                 int open_mode, enum open_claim_type4 claim)
1429 {
1430         int ret = 0;
1431
1432         if (open_mode & (O_EXCL|O_TRUNC))
1433                 goto out;
1434         switch (claim) {
1435         case NFS4_OPEN_CLAIM_NULL:
1436         case NFS4_OPEN_CLAIM_FH:
1437                 goto out;
1438         default:
1439                 break;
1440         }
1441         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1442                 case FMODE_READ:
1443                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1444                                 && state->n_rdonly != 0;
1445                         break;
1446                 case FMODE_WRITE:
1447                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1448                                 && state->n_wronly != 0;
1449                         break;
1450                 case FMODE_READ|FMODE_WRITE:
1451                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1452                                 && state->n_rdwr != 0;
1453         }
1454 out:
1455         return ret;
1456 }
1457
1458 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1459                 enum open_claim_type4 claim)
1460 {
1461         if (delegation == NULL)
1462                 return 0;
1463         if ((delegation->type & fmode) != fmode)
1464                 return 0;
1465         switch (claim) {
1466         case NFS4_OPEN_CLAIM_NULL:
1467         case NFS4_OPEN_CLAIM_FH:
1468                 break;
1469         case NFS4_OPEN_CLAIM_PREVIOUS:
1470                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1471                         break;
1472                 /* Fall through */
1473         default:
1474                 return 0;
1475         }
1476         nfs_mark_delegation_referenced(delegation);
1477         return 1;
1478 }
1479
1480 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1481 {
1482         switch (fmode) {
1483                 case FMODE_WRITE:
1484                         state->n_wronly++;
1485                         break;
1486                 case FMODE_READ:
1487                         state->n_rdonly++;
1488                         break;
1489                 case FMODE_READ|FMODE_WRITE:
1490                         state->n_rdwr++;
1491         }
1492         nfs4_state_set_mode_locked(state, state->state | fmode);
1493 }
1494
1495 #ifdef CONFIG_NFS_V4_1
1496 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1497 {
1498         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1499                 return true;
1500         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1501                 return true;
1502         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1503                 return true;
1504         return false;
1505 }
1506 #endif /* CONFIG_NFS_V4_1 */
1507
1508 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1509 {
1510         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1511                 wake_up_all(&state->waitq);
1512 }
1513
1514 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1515                 const nfs4_stateid *stateid)
1516 {
1517         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1518         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1519
1520         if (stateid_seqid == state_seqid + 1U ||
1521             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1522                 nfs_state_log_update_open_stateid(state);
1523         else
1524                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1525 }
1526
1527 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1528 {
1529         struct nfs_client *clp = state->owner->so_server->nfs_client;
1530         bool need_recover = false;
1531
1532         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1533                 need_recover = true;
1534         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1535                 need_recover = true;
1536         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1537                 need_recover = true;
1538         if (need_recover)
1539                 nfs4_state_mark_reclaim_nograce(clp, state);
1540 }
1541
1542 /*
1543  * Check for whether or not the caller may update the open stateid
1544  * to the value passed in by stateid.
1545  *
1546  * Note: This function relies heavily on the server implementing
1547  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1548  * correctly.
1549  * i.e. The stateid seqids have to be initialised to 1, and
1550  * are then incremented on every state transition.
1551  */
1552 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1553                 const nfs4_stateid *stateid)
1554 {
1555         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1556             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1557                 if (stateid->seqid == cpu_to_be32(1))
1558                         nfs_state_log_update_open_stateid(state);
1559                 else
1560                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1561                 return true;
1562         }
1563
1564         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1565                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1566                 return true;
1567         }
1568         return false;
1569 }
1570
1571 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1572 {
1573         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1574                 return;
1575         if (state->n_wronly)
1576                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1577         if (state->n_rdonly)
1578                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1579         if (state->n_rdwr)
1580                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1581         set_bit(NFS_OPEN_STATE, &state->flags);
1582 }
1583
1584 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1585                 nfs4_stateid *stateid, fmode_t fmode)
1586 {
1587         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1588         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1589         case FMODE_WRITE:
1590                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1591                 break;
1592         case FMODE_READ:
1593                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1594                 break;
1595         case 0:
1596                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1597                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1598                 clear_bit(NFS_OPEN_STATE, &state->flags);
1599         }
1600         if (stateid == NULL)
1601                 return;
1602         /* Handle OPEN+OPEN_DOWNGRADE races */
1603         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1604             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1605                 nfs_resync_open_stateid_locked(state);
1606                 goto out;
1607         }
1608         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1609                 nfs4_stateid_copy(&state->stateid, stateid);
1610         nfs4_stateid_copy(&state->open_stateid, stateid);
1611         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1612 out:
1613         nfs_state_log_update_open_stateid(state);
1614 }
1615
1616 static void nfs_clear_open_stateid(struct nfs4_state *state,
1617         nfs4_stateid *arg_stateid,
1618         nfs4_stateid *stateid, fmode_t fmode)
1619 {
1620         write_seqlock(&state->seqlock);
1621         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1622         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1623                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1624         write_sequnlock(&state->seqlock);
1625         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1626                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1627 }
1628
1629 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1630                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1631         __must_hold(&state->owner->so_lock)
1632         __must_hold(&state->seqlock)
1633         __must_hold(RCU)
1634
1635 {
1636         DEFINE_WAIT(wait);
1637         int status = 0;
1638         for (;;) {
1639
1640                 if (!nfs_need_update_open_stateid(state, stateid))
1641                         return;
1642                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1643                         break;
1644                 if (status)
1645                         break;
1646                 /* Rely on seqids for serialisation with NFSv4.0 */
1647                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1648                         break;
1649
1650                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1651                 /*
1652                  * Ensure we process the state changes in the same order
1653                  * in which the server processed them by delaying the
1654                  * update of the stateid until we are in sequence.
1655                  */
1656                 write_sequnlock(&state->seqlock);
1657                 spin_unlock(&state->owner->so_lock);
1658                 rcu_read_unlock();
1659                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1660                 if (!signal_pending(current)) {
1661                         if (schedule_timeout(5*HZ) == 0)
1662                                 status = -EAGAIN;
1663                         else
1664                                 status = 0;
1665                 } else
1666                         status = -EINTR;
1667                 finish_wait(&state->waitq, &wait);
1668                 rcu_read_lock();
1669                 spin_lock(&state->owner->so_lock);
1670                 write_seqlock(&state->seqlock);
1671         }
1672
1673         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1674             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1675                 nfs4_stateid_copy(freeme, &state->open_stateid);
1676                 nfs_test_and_clear_all_open_stateid(state);
1677         }
1678
1679         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1680                 nfs4_stateid_copy(&state->stateid, stateid);
1681         nfs4_stateid_copy(&state->open_stateid, stateid);
1682         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1683         nfs_state_log_update_open_stateid(state);
1684 }
1685
1686 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1687                 const nfs4_stateid *open_stateid,
1688                 fmode_t fmode,
1689                 nfs4_stateid *freeme)
1690 {
1691         /*
1692          * Protect the call to nfs4_state_set_mode_locked and
1693          * serialise the stateid update
1694          */
1695         write_seqlock(&state->seqlock);
1696         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1697         switch (fmode) {
1698         case FMODE_READ:
1699                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1700                 break;
1701         case FMODE_WRITE:
1702                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1703                 break;
1704         case FMODE_READ|FMODE_WRITE:
1705                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1706         }
1707         set_bit(NFS_OPEN_STATE, &state->flags);
1708         write_sequnlock(&state->seqlock);
1709 }
1710
1711 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1712 {
1713         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1714         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1715         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1716         clear_bit(NFS_OPEN_STATE, &state->flags);
1717 }
1718
1719 static void nfs_state_set_delegation(struct nfs4_state *state,
1720                 const nfs4_stateid *deleg_stateid,
1721                 fmode_t fmode)
1722 {
1723         /*
1724          * Protect the call to nfs4_state_set_mode_locked and
1725          * serialise the stateid update
1726          */
1727         write_seqlock(&state->seqlock);
1728         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1729         set_bit(NFS_DELEGATED_STATE, &state->flags);
1730         write_sequnlock(&state->seqlock);
1731 }
1732
1733 static void nfs_state_clear_delegation(struct nfs4_state *state)
1734 {
1735         write_seqlock(&state->seqlock);
1736         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1737         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1738         write_sequnlock(&state->seqlock);
1739 }
1740
1741 int update_open_stateid(struct nfs4_state *state,
1742                 const nfs4_stateid *open_stateid,
1743                 const nfs4_stateid *delegation,
1744                 fmode_t fmode)
1745 {
1746         struct nfs_server *server = NFS_SERVER(state->inode);
1747         struct nfs_client *clp = server->nfs_client;
1748         struct nfs_inode *nfsi = NFS_I(state->inode);
1749         struct nfs_delegation *deleg_cur;
1750         nfs4_stateid freeme = { };
1751         int ret = 0;
1752
1753         fmode &= (FMODE_READ|FMODE_WRITE);
1754
1755         rcu_read_lock();
1756         spin_lock(&state->owner->so_lock);
1757         if (open_stateid != NULL) {
1758                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1759                 ret = 1;
1760         }
1761
1762         deleg_cur = nfs4_get_valid_delegation(state->inode);
1763         if (deleg_cur == NULL)
1764                 goto no_delegation;
1765
1766         spin_lock(&deleg_cur->lock);
1767         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1768            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1769             (deleg_cur->type & fmode) != fmode)
1770                 goto no_delegation_unlock;
1771
1772         if (delegation == NULL)
1773                 delegation = &deleg_cur->stateid;
1774         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1775                 goto no_delegation_unlock;
1776
1777         nfs_mark_delegation_referenced(deleg_cur);
1778         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1779         ret = 1;
1780 no_delegation_unlock:
1781         spin_unlock(&deleg_cur->lock);
1782 no_delegation:
1783         if (ret)
1784                 update_open_stateflags(state, fmode);
1785         spin_unlock(&state->owner->so_lock);
1786         rcu_read_unlock();
1787
1788         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1789                 nfs4_schedule_state_manager(clp);
1790         if (freeme.type != 0)
1791                 nfs4_test_and_free_stateid(server, &freeme,
1792                                 state->owner->so_cred);
1793
1794         return ret;
1795 }
1796
1797 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1798                 const nfs4_stateid *stateid)
1799 {
1800         struct nfs4_state *state = lsp->ls_state;
1801         bool ret = false;
1802
1803         spin_lock(&state->state_lock);
1804         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1805                 goto out_noupdate;
1806         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1807                 goto out_noupdate;
1808         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1809         ret = true;
1810 out_noupdate:
1811         spin_unlock(&state->state_lock);
1812         return ret;
1813 }
1814
1815 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1816 {
1817         struct nfs_delegation *delegation;
1818
1819         fmode &= FMODE_READ|FMODE_WRITE;
1820         rcu_read_lock();
1821         delegation = nfs4_get_valid_delegation(inode);
1822         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1823                 rcu_read_unlock();
1824                 return;
1825         }
1826         rcu_read_unlock();
1827         nfs4_inode_return_delegation(inode);
1828 }
1829
1830 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1831 {
1832         struct nfs4_state *state = opendata->state;
1833         struct nfs_delegation *delegation;
1834         int open_mode = opendata->o_arg.open_flags;
1835         fmode_t fmode = opendata->o_arg.fmode;
1836         enum open_claim_type4 claim = opendata->o_arg.claim;
1837         nfs4_stateid stateid;
1838         int ret = -EAGAIN;
1839
1840         for (;;) {
1841                 spin_lock(&state->owner->so_lock);
1842                 if (can_open_cached(state, fmode, open_mode, claim)) {
1843                         update_open_stateflags(state, fmode);
1844                         spin_unlock(&state->owner->so_lock);
1845                         goto out_return_state;
1846                 }
1847                 spin_unlock(&state->owner->so_lock);
1848                 rcu_read_lock();
1849                 delegation = nfs4_get_valid_delegation(state->inode);
1850                 if (!can_open_delegated(delegation, fmode, claim)) {
1851                         rcu_read_unlock();
1852                         break;
1853                 }
1854                 /* Save the delegation */
1855                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1856                 rcu_read_unlock();
1857                 nfs_release_seqid(opendata->o_arg.seqid);
1858                 if (!opendata->is_recover) {
1859                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1860                         if (ret != 0)
1861                                 goto out;
1862                 }
1863                 ret = -EAGAIN;
1864
1865                 /* Try to update the stateid using the delegation */
1866                 if (update_open_stateid(state, NULL, &stateid, fmode))
1867                         goto out_return_state;
1868         }
1869 out:
1870         return ERR_PTR(ret);
1871 out_return_state:
1872         refcount_inc(&state->count);
1873         return state;
1874 }
1875
1876 static void
1877 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1878 {
1879         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1880         struct nfs_delegation *delegation;
1881         int delegation_flags = 0;
1882
1883         rcu_read_lock();
1884         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1885         if (delegation)
1886                 delegation_flags = delegation->flags;
1887         rcu_read_unlock();
1888         switch (data->o_arg.claim) {
1889         default:
1890                 break;
1891         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1892         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1893                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1894                                    "returning a delegation for "
1895                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1896                                    clp->cl_hostname);
1897                 return;
1898         }
1899         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1900                 nfs_inode_set_delegation(state->inode,
1901                                 data->owner->so_cred,
1902                                 data->o_res.delegation_type,
1903                                 &data->o_res.delegation,
1904                                 data->o_res.pagemod_limit);
1905         else
1906                 nfs_inode_reclaim_delegation(state->inode,
1907                                 data->owner->so_cred,
1908                                 data->o_res.delegation_type,
1909                                 &data->o_res.delegation,
1910                                 data->o_res.pagemod_limit);
1911
1912         if (data->o_res.do_recall)
1913                 nfs_async_inode_return_delegation(state->inode,
1914                                                   &data->o_res.delegation);
1915 }
1916
1917 /*
1918  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1919  * and update the nfs4_state.
1920  */
1921 static struct nfs4_state *
1922 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1923 {
1924         struct inode *inode = data->state->inode;
1925         struct nfs4_state *state = data->state;
1926         int ret;
1927
1928         if (!data->rpc_done) {
1929                 if (data->rpc_status)
1930                         return ERR_PTR(data->rpc_status);
1931                 /* cached opens have already been processed */
1932                 goto update;
1933         }
1934
1935         ret = nfs_refresh_inode(inode, &data->f_attr);
1936         if (ret)
1937                 return ERR_PTR(ret);
1938
1939         if (data->o_res.delegation_type != 0)
1940                 nfs4_opendata_check_deleg(data, state);
1941 update:
1942         if (!update_open_stateid(state, &data->o_res.stateid,
1943                                 NULL, data->o_arg.fmode))
1944                 return ERR_PTR(-EAGAIN);
1945         refcount_inc(&state->count);
1946
1947         return state;
1948 }
1949
1950 static struct inode *
1951 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1952 {
1953         struct inode *inode;
1954
1955         switch (data->o_arg.claim) {
1956         case NFS4_OPEN_CLAIM_NULL:
1957         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1958         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1959                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1960                         return ERR_PTR(-EAGAIN);
1961                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1962                                 &data->f_attr, data->f_label);
1963                 break;
1964         default:
1965                 inode = d_inode(data->dentry);
1966                 ihold(inode);
1967                 nfs_refresh_inode(inode, &data->f_attr);
1968         }
1969         return inode;
1970 }
1971
1972 static struct nfs4_state *
1973 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1974 {
1975         struct nfs4_state *state;
1976         struct inode *inode;
1977
1978         inode = nfs4_opendata_get_inode(data);
1979         if (IS_ERR(inode))
1980                 return ERR_CAST(inode);
1981         if (data->state != NULL && data->state->inode == inode) {
1982                 state = data->state;
1983                 refcount_inc(&state->count);
1984         } else
1985                 state = nfs4_get_open_state(inode, data->owner);
1986         iput(inode);
1987         if (state == NULL)
1988                 state = ERR_PTR(-ENOMEM);
1989         return state;
1990 }
1991
1992 static struct nfs4_state *
1993 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1994 {
1995         struct nfs4_state *state;
1996
1997         if (!data->rpc_done) {
1998                 state = nfs4_try_open_cached(data);
1999                 trace_nfs4_cached_open(data->state);
2000                 goto out;
2001         }
2002
2003         state = nfs4_opendata_find_nfs4_state(data);
2004         if (IS_ERR(state))
2005                 goto out;
2006
2007         if (data->o_res.delegation_type != 0)
2008                 nfs4_opendata_check_deleg(data, state);
2009         if (!update_open_stateid(state, &data->o_res.stateid,
2010                                 NULL, data->o_arg.fmode)) {
2011                 nfs4_put_open_state(state);
2012                 state = ERR_PTR(-EAGAIN);
2013         }
2014 out:
2015         nfs_release_seqid(data->o_arg.seqid);
2016         return state;
2017 }
2018
2019 static struct nfs4_state *
2020 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2021 {
2022         struct nfs4_state *ret;
2023
2024         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2025                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2026         else
2027                 ret = _nfs4_opendata_to_nfs4_state(data);
2028         nfs4_sequence_free_slot(&data->o_res.seq_res);
2029         return ret;
2030 }
2031
2032 static struct nfs_open_context *
2033 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2034 {
2035         struct nfs_inode *nfsi = NFS_I(state->inode);
2036         struct nfs_open_context *ctx;
2037
2038         rcu_read_lock();
2039         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2040                 if (ctx->state != state)
2041                         continue;
2042                 if ((ctx->mode & mode) != mode)
2043                         continue;
2044                 if (!get_nfs_open_context(ctx))
2045                         continue;
2046                 rcu_read_unlock();
2047                 return ctx;
2048         }
2049         rcu_read_unlock();
2050         return ERR_PTR(-ENOENT);
2051 }
2052
2053 static struct nfs_open_context *
2054 nfs4_state_find_open_context(struct nfs4_state *state)
2055 {
2056         struct nfs_open_context *ctx;
2057
2058         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2059         if (!IS_ERR(ctx))
2060                 return ctx;
2061         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2062         if (!IS_ERR(ctx))
2063                 return ctx;
2064         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2065 }
2066
2067 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2068                 struct nfs4_state *state, enum open_claim_type4 claim)
2069 {
2070         struct nfs4_opendata *opendata;
2071
2072         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2073                         NULL, claim, GFP_NOFS);
2074         if (opendata == NULL)
2075                 return ERR_PTR(-ENOMEM);
2076         opendata->state = state;
2077         refcount_inc(&state->count);
2078         return opendata;
2079 }
2080
2081 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2082                 fmode_t fmode)
2083 {
2084         struct nfs4_state *newstate;
2085         int ret;
2086
2087         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2088                 return 0;
2089         opendata->o_arg.open_flags = 0;
2090         opendata->o_arg.fmode = fmode;
2091         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2092                         NFS_SB(opendata->dentry->d_sb),
2093                         fmode, 0);
2094         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2095         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2096         nfs4_init_opendata_res(opendata);
2097         ret = _nfs4_recover_proc_open(opendata);
2098         if (ret != 0)
2099                 return ret; 
2100         newstate = nfs4_opendata_to_nfs4_state(opendata);
2101         if (IS_ERR(newstate))
2102                 return PTR_ERR(newstate);
2103         if (newstate != opendata->state)
2104                 ret = -ESTALE;
2105         nfs4_close_state(newstate, fmode);
2106         return ret;
2107 }
2108
2109 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2110 {
2111         int ret;
2112
2113         /* memory barrier prior to reading state->n_* */
2114         smp_rmb();
2115         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2116         if (ret != 0)
2117                 return ret;
2118         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2119         if (ret != 0)
2120                 return ret;
2121         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2122         if (ret != 0)
2123                 return ret;
2124         /*
2125          * We may have performed cached opens for all three recoveries.
2126          * Check if we need to update the current stateid.
2127          */
2128         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2129             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2130                 write_seqlock(&state->seqlock);
2131                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2132                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2133                 write_sequnlock(&state->seqlock);
2134         }
2135         return 0;
2136 }
2137
2138 /*
2139  * OPEN_RECLAIM:
2140  *      reclaim state on the server after a reboot.
2141  */
2142 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2143 {
2144         struct nfs_delegation *delegation;
2145         struct nfs4_opendata *opendata;
2146         fmode_t delegation_type = 0;
2147         int status;
2148
2149         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2150                         NFS4_OPEN_CLAIM_PREVIOUS);
2151         if (IS_ERR(opendata))
2152                 return PTR_ERR(opendata);
2153         rcu_read_lock();
2154         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2155         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2156                 delegation_type = delegation->type;
2157         rcu_read_unlock();
2158         opendata->o_arg.u.delegation_type = delegation_type;
2159         status = nfs4_open_recover(opendata, state);
2160         nfs4_opendata_put(opendata);
2161         return status;
2162 }
2163
2164 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2165 {
2166         struct nfs_server *server = NFS_SERVER(state->inode);
2167         struct nfs4_exception exception = { };
2168         int err;
2169         do {
2170                 err = _nfs4_do_open_reclaim(ctx, state);
2171                 trace_nfs4_open_reclaim(ctx, 0, err);
2172                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2173                         continue;
2174                 if (err != -NFS4ERR_DELAY)
2175                         break;
2176                 nfs4_handle_exception(server, err, &exception);
2177         } while (exception.retry);
2178         return err;
2179 }
2180
2181 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2182 {
2183         struct nfs_open_context *ctx;
2184         int ret;
2185
2186         ctx = nfs4_state_find_open_context(state);
2187         if (IS_ERR(ctx))
2188                 return -EAGAIN;
2189         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2190         nfs_state_clear_open_state_flags(state);
2191         ret = nfs4_do_open_reclaim(ctx, state);
2192         put_nfs_open_context(ctx);
2193         return ret;
2194 }
2195
2196 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)
2197 {
2198         switch (err) {
2199                 default:
2200                         printk(KERN_ERR "NFS: %s: unhandled error "
2201                                         "%d.\n", __func__, err);
2202                 case 0:
2203                 case -ENOENT:
2204                 case -EAGAIN:
2205                 case -ESTALE:
2206                 case -ETIMEDOUT:
2207                         break;
2208                 case -NFS4ERR_BADSESSION:
2209                 case -NFS4ERR_BADSLOT:
2210                 case -NFS4ERR_BAD_HIGH_SLOT:
2211                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2212                 case -NFS4ERR_DEADSESSION:
2213                         return -EAGAIN;
2214                 case -NFS4ERR_STALE_CLIENTID:
2215                 case -NFS4ERR_STALE_STATEID:
2216                         /* Don't recall a delegation if it was lost */
2217                         nfs4_schedule_lease_recovery(server->nfs_client);
2218                         return -EAGAIN;
2219                 case -NFS4ERR_MOVED:
2220                         nfs4_schedule_migration_recovery(server);
2221                         return -EAGAIN;
2222                 case -NFS4ERR_LEASE_MOVED:
2223                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2224                         return -EAGAIN;
2225                 case -NFS4ERR_DELEG_REVOKED:
2226                 case -NFS4ERR_ADMIN_REVOKED:
2227                 case -NFS4ERR_EXPIRED:
2228                 case -NFS4ERR_BAD_STATEID:
2229                 case -NFS4ERR_OPENMODE:
2230                         nfs_inode_find_state_and_recover(state->inode,
2231                                         stateid);
2232                         nfs4_schedule_stateid_recovery(server, state);
2233                         return -EAGAIN;
2234                 case -NFS4ERR_DELAY:
2235                 case -NFS4ERR_GRACE:
2236                         ssleep(1);
2237                         return -EAGAIN;
2238                 case -ENOMEM:
2239                 case -NFS4ERR_DENIED:
2240                         if (fl) {
2241                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2242                                 if (lsp)
2243                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2244                         }
2245                         return 0;
2246         }
2247         return err;
2248 }
2249
2250 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2251                 struct nfs4_state *state, const nfs4_stateid *stateid)
2252 {
2253         struct nfs_server *server = NFS_SERVER(state->inode);
2254         struct nfs4_opendata *opendata;
2255         int err = 0;
2256
2257         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2258                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2259         if (IS_ERR(opendata))
2260                 return PTR_ERR(opendata);
2261         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2262         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2263                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2264                 if (err)
2265                         goto out;
2266         }
2267         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2268                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2269                 if (err)
2270                         goto out;
2271         }
2272         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2273                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2274                 if (err)
2275                         goto out;
2276         }
2277         nfs_state_clear_delegation(state);
2278 out:
2279         nfs4_opendata_put(opendata);
2280         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2281 }
2282
2283 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2284 {
2285         struct nfs4_opendata *data = calldata;
2286
2287         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2288                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2289 }
2290
2291 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2292 {
2293         struct nfs4_opendata *data = calldata;
2294
2295         nfs40_sequence_done(task, &data->c_res.seq_res);
2296
2297         data->rpc_status = task->tk_status;
2298         if (data->rpc_status == 0) {
2299                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2300                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2301                 renew_lease(data->o_res.server, data->timestamp);
2302                 data->rpc_done = true;
2303         }
2304 }
2305
2306 static void nfs4_open_confirm_release(void *calldata)
2307 {
2308         struct nfs4_opendata *data = calldata;
2309         struct nfs4_state *state = NULL;
2310
2311         /* If this request hasn't been cancelled, do nothing */
2312         if (!data->cancelled)
2313                 goto out_free;
2314         /* In case of error, no cleanup! */
2315         if (!data->rpc_done)
2316                 goto out_free;
2317         state = nfs4_opendata_to_nfs4_state(data);
2318         if (!IS_ERR(state))
2319                 nfs4_close_state(state, data->o_arg.fmode);
2320 out_free:
2321         nfs4_opendata_put(data);
2322 }
2323
2324 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2325         .rpc_call_prepare = nfs4_open_confirm_prepare,
2326         .rpc_call_done = nfs4_open_confirm_done,
2327         .rpc_release = nfs4_open_confirm_release,
2328 };
2329
2330 /*
2331  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2332  */
2333 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2334 {
2335         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2336         struct rpc_task *task;
2337         struct  rpc_message msg = {
2338                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2339                 .rpc_argp = &data->c_arg,
2340                 .rpc_resp = &data->c_res,
2341                 .rpc_cred = data->owner->so_cred,
2342         };
2343         struct rpc_task_setup task_setup_data = {
2344                 .rpc_client = server->client,
2345                 .rpc_message = &msg,
2346                 .callback_ops = &nfs4_open_confirm_ops,
2347                 .callback_data = data,
2348                 .workqueue = nfsiod_workqueue,
2349                 .flags = RPC_TASK_ASYNC,
2350         };
2351         int status;
2352
2353         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2354                                 data->is_recover);
2355         kref_get(&data->kref);
2356         data->rpc_done = false;
2357         data->rpc_status = 0;
2358         data->timestamp = jiffies;
2359         task = rpc_run_task(&task_setup_data);
2360         if (IS_ERR(task))
2361                 return PTR_ERR(task);
2362         status = rpc_wait_for_completion_task(task);
2363         if (status != 0) {
2364                 data->cancelled = true;
2365                 smp_wmb();
2366         } else
2367                 status = data->rpc_status;
2368         rpc_put_task(task);
2369         return status;
2370 }
2371
2372 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2373 {
2374         struct nfs4_opendata *data = calldata;
2375         struct nfs4_state_owner *sp = data->owner;
2376         struct nfs_client *clp = sp->so_server->nfs_client;
2377         enum open_claim_type4 claim = data->o_arg.claim;
2378
2379         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2380                 goto out_wait;
2381         /*
2382          * Check if we still need to send an OPEN call, or if we can use
2383          * a delegation instead.
2384          */
2385         if (data->state != NULL) {
2386                 struct nfs_delegation *delegation;
2387
2388                 if (can_open_cached(data->state, data->o_arg.fmode,
2389                                         data->o_arg.open_flags, claim))
2390                         goto out_no_action;
2391                 rcu_read_lock();
2392                 delegation = nfs4_get_valid_delegation(data->state->inode);
2393                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2394                         goto unlock_no_action;
2395                 rcu_read_unlock();
2396         }
2397         /* Update client id. */
2398         data->o_arg.clientid = clp->cl_clientid;
2399         switch (claim) {
2400         default:
2401                 break;
2402         case NFS4_OPEN_CLAIM_PREVIOUS:
2403         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2404         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2405                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2406                 /* Fall through */
2407         case NFS4_OPEN_CLAIM_FH:
2408                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2409         }
2410         data->timestamp = jiffies;
2411         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2412                                 &data->o_arg.seq_args,
2413                                 &data->o_res.seq_res,
2414                                 task) != 0)
2415                 nfs_release_seqid(data->o_arg.seqid);
2416
2417         /* Set the create mode (note dependency on the session type) */
2418         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2419         if (data->o_arg.open_flags & O_EXCL) {
2420                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2421                 if (nfs4_has_persistent_session(clp))
2422                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2423                 else if (clp->cl_mvops->minor_version > 0)
2424                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2425         }
2426         return;
2427 unlock_no_action:
2428         trace_nfs4_cached_open(data->state);
2429         rcu_read_unlock();
2430 out_no_action:
2431         task->tk_action = NULL;
2432 out_wait:
2433         nfs4_sequence_done(task, &data->o_res.seq_res);
2434 }
2435
2436 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2437 {
2438         struct nfs4_opendata *data = calldata;
2439
2440         data->rpc_status = task->tk_status;
2441
2442         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2443                 return;
2444
2445         if (task->tk_status == 0) {
2446                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2447                         switch (data->o_res.f_attr->mode & S_IFMT) {
2448                         case S_IFREG:
2449                                 break;
2450                         case S_IFLNK:
2451                                 data->rpc_status = -ELOOP;
2452                                 break;
2453                         case S_IFDIR:
2454                                 data->rpc_status = -EISDIR;
2455                                 break;
2456                         default:
2457                                 data->rpc_status = -ENOTDIR;
2458                         }
2459                 }
2460                 renew_lease(data->o_res.server, data->timestamp);
2461                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2462                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2463         }
2464         data->rpc_done = true;
2465 }
2466
2467 static void nfs4_open_release(void *calldata)
2468 {
2469         struct nfs4_opendata *data = calldata;
2470         struct nfs4_state *state = NULL;
2471
2472         /* If this request hasn't been cancelled, do nothing */
2473         if (!data->cancelled)
2474                 goto out_free;
2475         /* In case of error, no cleanup! */
2476         if (data->rpc_status != 0 || !data->rpc_done)
2477                 goto out_free;
2478         /* In case we need an open_confirm, no cleanup! */
2479         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2480                 goto out_free;
2481         state = nfs4_opendata_to_nfs4_state(data);
2482         if (!IS_ERR(state))
2483                 nfs4_close_state(state, data->o_arg.fmode);
2484 out_free:
2485         nfs4_opendata_put(data);
2486 }
2487
2488 static const struct rpc_call_ops nfs4_open_ops = {
2489         .rpc_call_prepare = nfs4_open_prepare,
2490         .rpc_call_done = nfs4_open_done,
2491         .rpc_release = nfs4_open_release,
2492 };
2493
2494 static int nfs4_run_open_task(struct nfs4_opendata *data,
2495                               struct nfs_open_context *ctx)
2496 {
2497         struct inode *dir = d_inode(data->dir);
2498         struct nfs_server *server = NFS_SERVER(dir);
2499         struct nfs_openargs *o_arg = &data->o_arg;
2500         struct nfs_openres *o_res = &data->o_res;
2501         struct rpc_task *task;
2502         struct rpc_message msg = {
2503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2504                 .rpc_argp = o_arg,
2505                 .rpc_resp = o_res,
2506                 .rpc_cred = data->owner->so_cred,
2507         };
2508         struct rpc_task_setup task_setup_data = {
2509                 .rpc_client = server->client,
2510                 .rpc_message = &msg,
2511                 .callback_ops = &nfs4_open_ops,
2512                 .callback_data = data,
2513                 .workqueue = nfsiod_workqueue,
2514                 .flags = RPC_TASK_ASYNC,
2515         };
2516         int status;
2517
2518         kref_get(&data->kref);
2519         data->rpc_done = false;
2520         data->rpc_status = 0;
2521         data->cancelled = false;
2522         data->is_recover = false;
2523         if (!ctx) {
2524                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2525                 data->is_recover = true;
2526                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2527         } else {
2528                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2529                 pnfs_lgopen_prepare(data, ctx);
2530         }
2531         task = rpc_run_task(&task_setup_data);
2532         if (IS_ERR(task))
2533                 return PTR_ERR(task);
2534         status = rpc_wait_for_completion_task(task);
2535         if (status != 0) {
2536                 data->cancelled = true;
2537                 smp_wmb();
2538         } else
2539                 status = data->rpc_status;
2540         rpc_put_task(task);
2541
2542         return status;
2543 }
2544
2545 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2546 {
2547         struct inode *dir = d_inode(data->dir);
2548         struct nfs_openres *o_res = &data->o_res;
2549         int status;
2550
2551         status = nfs4_run_open_task(data, NULL);
2552         if (status != 0 || !data->rpc_done)
2553                 return status;
2554
2555         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2556
2557         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2558                 status = _nfs4_proc_open_confirm(data);
2559
2560         return status;
2561 }
2562
2563 /*
2564  * Additional permission checks in order to distinguish between an
2565  * open for read, and an open for execute. This works around the
2566  * fact that NFSv4 OPEN treats read and execute permissions as being
2567  * the same.
2568  * Note that in the non-execute case, we want to turn off permission
2569  * checking if we just created a new file (POSIX open() semantics).
2570  */
2571 static int nfs4_opendata_access(const struct cred *cred,
2572                                 struct nfs4_opendata *opendata,
2573                                 struct nfs4_state *state, fmode_t fmode,
2574                                 int openflags)
2575 {
2576         struct nfs_access_entry cache;
2577         u32 mask, flags;
2578
2579         /* access call failed or for some reason the server doesn't
2580          * support any access modes -- defer access call until later */
2581         if (opendata->o_res.access_supported == 0)
2582                 return 0;
2583
2584         mask = 0;
2585         /*
2586          * Use openflags to check for exec, because fmode won't
2587          * always have FMODE_EXEC set when file open for exec.
2588          */
2589         if (openflags & __FMODE_EXEC) {
2590                 /* ONLY check for exec rights */
2591                 if (S_ISDIR(state->inode->i_mode))
2592                         mask = NFS4_ACCESS_LOOKUP;
2593                 else
2594                         mask = NFS4_ACCESS_EXECUTE;
2595         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2596                 mask = NFS4_ACCESS_READ;
2597
2598         cache.cred = cred;
2599         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2600         nfs_access_add_cache(state->inode, &cache);
2601
2602         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2603         if ((mask & ~cache.mask & flags) == 0)
2604                 return 0;
2605
2606         return -EACCES;
2607 }
2608
2609 /*
2610  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2611  */
2612 static int _nfs4_proc_open(struct nfs4_opendata *data,
2613                            struct nfs_open_context *ctx)
2614 {
2615         struct inode *dir = d_inode(data->dir);
2616         struct nfs_server *server = NFS_SERVER(dir);
2617         struct nfs_openargs *o_arg = &data->o_arg;
2618         struct nfs_openres *o_res = &data->o_res;
2619         int status;
2620
2621         status = nfs4_run_open_task(data, ctx);
2622         if (!data->rpc_done)
2623                 return status;
2624         if (status != 0) {
2625                 if (status == -NFS4ERR_BADNAME &&
2626                                 !(o_arg->open_flags & O_CREAT))
2627                         return -ENOENT;
2628                 return status;
2629         }
2630
2631         nfs_fattr_map_and_free_names(server, &data->f_attr);
2632
2633         if (o_arg->open_flags & O_CREAT) {
2634                 if (o_arg->open_flags & O_EXCL)
2635                         data->file_created = true;
2636                 else if (o_res->cinfo.before != o_res->cinfo.after)
2637                         data->file_created = true;
2638                 if (data->file_created ||
2639                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2640                         update_changeattr(dir, &o_res->cinfo,
2641                                         o_res->f_attr->time_start, 0);
2642         }
2643         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2644                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2645         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2646                 status = _nfs4_proc_open_confirm(data);
2647                 if (status != 0)
2648                         return status;
2649         }
2650         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2651                 nfs4_sequence_free_slot(&o_res->seq_res);
2652                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2653                                 o_res->f_label, NULL);
2654         }
2655         return 0;
2656 }
2657
2658 /*
2659  * OPEN_EXPIRED:
2660  *      reclaim state on the server after a network partition.
2661  *      Assumes caller holds the appropriate lock
2662  */
2663 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2664 {
2665         struct nfs4_opendata *opendata;
2666         int ret;
2667
2668         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2669                         NFS4_OPEN_CLAIM_FH);
2670         if (IS_ERR(opendata))
2671                 return PTR_ERR(opendata);
2672         ret = nfs4_open_recover(opendata, state);
2673         if (ret == -ESTALE)
2674                 d_drop(ctx->dentry);
2675         nfs4_opendata_put(opendata);
2676         return ret;
2677 }
2678
2679 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2680 {
2681         struct nfs_server *server = NFS_SERVER(state->inode);
2682         struct nfs4_exception exception = { };
2683         int err;
2684
2685         do {
2686                 err = _nfs4_open_expired(ctx, state);
2687                 trace_nfs4_open_expired(ctx, 0, err);
2688                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2689                         continue;
2690                 switch (err) {
2691                 default:
2692                         goto out;
2693                 case -NFS4ERR_GRACE:
2694                 case -NFS4ERR_DELAY:
2695                         nfs4_handle_exception(server, err, &exception);
2696                         err = 0;
2697                 }
2698         } while (exception.retry);
2699 out:
2700         return err;
2701 }
2702
2703 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2704 {
2705         struct nfs_open_context *ctx;
2706         int ret;
2707
2708         ctx = nfs4_state_find_open_context(state);
2709         if (IS_ERR(ctx))
2710                 return -EAGAIN;
2711         ret = nfs4_do_open_expired(ctx, state);
2712         put_nfs_open_context(ctx);
2713         return ret;
2714 }
2715
2716 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2717                 const nfs4_stateid *stateid)
2718 {
2719         nfs_remove_bad_delegation(state->inode, stateid);
2720         nfs_state_clear_delegation(state);
2721 }
2722
2723 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2724 {
2725         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2726                 nfs_finish_clear_delegation_stateid(state, NULL);
2727 }
2728
2729 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2730 {
2731         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2732         nfs40_clear_delegation_stateid(state);
2733         nfs_state_clear_open_state_flags(state);
2734         return nfs4_open_expired(sp, state);
2735 }
2736
2737 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2738                 nfs4_stateid *stateid,
2739                 const struct cred *cred)
2740 {
2741         return -NFS4ERR_BAD_STATEID;
2742 }
2743
2744 #if defined(CONFIG_NFS_V4_1)
2745 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2746                 nfs4_stateid *stateid,
2747                 const struct cred *cred)
2748 {
2749         int status;
2750
2751         switch (stateid->type) {
2752         default:
2753                 break;
2754         case NFS4_INVALID_STATEID_TYPE:
2755         case NFS4_SPECIAL_STATEID_TYPE:
2756                 return -NFS4ERR_BAD_STATEID;
2757         case NFS4_REVOKED_STATEID_TYPE:
2758                 goto out_free;
2759         }
2760
2761         status = nfs41_test_stateid(server, stateid, cred);
2762         switch (status) {
2763         case -NFS4ERR_EXPIRED:
2764         case -NFS4ERR_ADMIN_REVOKED:
2765         case -NFS4ERR_DELEG_REVOKED:
2766                 break;
2767         default:
2768                 return status;
2769         }
2770 out_free:
2771         /* Ack the revoked state to the server */
2772         nfs41_free_stateid(server, stateid, cred, true);
2773         return -NFS4ERR_EXPIRED;
2774 }
2775
2776 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2777 {
2778         struct nfs_server *server = NFS_SERVER(state->inode);
2779         nfs4_stateid stateid;
2780         struct nfs_delegation *delegation;
2781         const struct cred *cred = NULL;
2782         int status, ret = NFS_OK;
2783
2784         /* Get the delegation credential for use by test/free_stateid */
2785         rcu_read_lock();
2786         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2787         if (delegation == NULL) {
2788                 rcu_read_unlock();
2789                 nfs_state_clear_delegation(state);
2790                 return NFS_OK;
2791         }
2792
2793         nfs4_stateid_copy(&stateid, &delegation->stateid);
2794
2795         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2796                                 &delegation->flags)) {
2797                 rcu_read_unlock();
2798                 return NFS_OK;
2799         }
2800
2801         if (delegation->cred)
2802                 cred = get_cred(delegation->cred);
2803         rcu_read_unlock();
2804         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2805         trace_nfs4_test_delegation_stateid(state, NULL, status);
2806         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2807                 nfs_finish_clear_delegation_stateid(state, &stateid);
2808         else
2809                 ret = status;
2810
2811         put_cred(cred);
2812         return ret;
2813 }
2814
2815 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2816 {
2817         nfs4_stateid tmp;
2818
2819         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2820             nfs4_copy_delegation_stateid(state->inode, state->state,
2821                                 &tmp, NULL) &&
2822             nfs4_stateid_match_other(&state->stateid, &tmp))
2823                 nfs_state_set_delegation(state, &tmp, state->state);
2824         else
2825                 nfs_state_clear_delegation(state);
2826 }
2827
2828 /**
2829  * nfs41_check_expired_locks - possibly free a lock stateid
2830  *
2831  * @state: NFSv4 state for an inode
2832  *
2833  * Returns NFS_OK if recovery for this stateid is now finished.
2834  * Otherwise a negative NFS4ERR value is returned.
2835  */
2836 static int nfs41_check_expired_locks(struct nfs4_state *state)
2837 {
2838         int status, ret = NFS_OK;
2839         struct nfs4_lock_state *lsp, *prev = NULL;
2840         struct nfs_server *server = NFS_SERVER(state->inode);
2841
2842         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2843                 goto out;
2844
2845         spin_lock(&state->state_lock);
2846         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2847                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2848                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2849
2850                         refcount_inc(&lsp->ls_count);
2851                         spin_unlock(&state->state_lock);
2852
2853                         nfs4_put_lock_state(prev);
2854                         prev = lsp;
2855
2856                         status = nfs41_test_and_free_expired_stateid(server,
2857                                         &lsp->ls_stateid,
2858                                         cred);
2859                         trace_nfs4_test_lock_stateid(state, lsp, status);
2860                         if (status == -NFS4ERR_EXPIRED ||
2861                             status == -NFS4ERR_BAD_STATEID) {
2862                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2863                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2864                                 if (!recover_lost_locks)
2865                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2866                         } else if (status != NFS_OK) {
2867                                 ret = status;
2868                                 nfs4_put_lock_state(prev);
2869                                 goto out;
2870                         }
2871                         spin_lock(&state->state_lock);
2872                 }
2873         }
2874         spin_unlock(&state->state_lock);
2875         nfs4_put_lock_state(prev);
2876 out:
2877         return ret;
2878 }
2879
2880 /**
2881  * nfs41_check_open_stateid - possibly free an open stateid
2882  *
2883  * @state: NFSv4 state for an inode
2884  *
2885  * Returns NFS_OK if recovery for this stateid is now finished.
2886  * Otherwise a negative NFS4ERR value is returned.
2887  */
2888 static int nfs41_check_open_stateid(struct nfs4_state *state)
2889 {
2890         struct nfs_server *server = NFS_SERVER(state->inode);
2891         nfs4_stateid *stateid = &state->open_stateid;
2892         const struct cred *cred = state->owner->so_cred;
2893         int status;
2894
2895         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2896                 return -NFS4ERR_BAD_STATEID;
2897         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2898         trace_nfs4_test_open_stateid(state, NULL, status);
2899         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2900                 nfs_state_clear_open_state_flags(state);
2901                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2902                 return status;
2903         }
2904         if (nfs_open_stateid_recover_openmode(state))
2905                 return -NFS4ERR_OPENMODE;
2906         return NFS_OK;
2907 }
2908
2909 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2910 {
2911         int status;
2912
2913         status = nfs41_check_delegation_stateid(state);
2914         if (status != NFS_OK)
2915                 return status;
2916         nfs41_delegation_recover_stateid(state);
2917
2918         status = nfs41_check_expired_locks(state);
2919         if (status != NFS_OK)
2920                 return status;
2921         status = nfs41_check_open_stateid(state);
2922         if (status != NFS_OK)
2923                 status = nfs4_open_expired(sp, state);
2924         return status;
2925 }
2926 #endif
2927
2928 /*
2929  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2930  * fields corresponding to attributes that were used to store the verifier.
2931  * Make sure we clobber those fields in the later setattr call
2932  */
2933 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2934                                 struct iattr *sattr, struct nfs4_label **label)
2935 {
2936         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2937         __u32 attrset[3];
2938         unsigned ret;
2939         unsigned i;
2940
2941         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2942                 attrset[i] = opendata->o_res.attrset[i];
2943                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2944                         attrset[i] &= ~bitmask[i];
2945         }
2946
2947         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2948                 sattr->ia_valid : 0;
2949
2950         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2951                 if (sattr->ia_valid & ATTR_ATIME_SET)
2952                         ret |= ATTR_ATIME_SET;
2953                 else
2954                         ret |= ATTR_ATIME;
2955         }
2956
2957         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2958                 if (sattr->ia_valid & ATTR_MTIME_SET)
2959                         ret |= ATTR_MTIME_SET;
2960                 else
2961                         ret |= ATTR_MTIME;
2962         }
2963
2964         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2965                 *label = NULL;
2966         return ret;
2967 }
2968
2969 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2970                 int flags, struct nfs_open_context *ctx)
2971 {
2972         struct nfs4_state_owner *sp = opendata->owner;
2973         struct nfs_server *server = sp->so_server;
2974         struct dentry *dentry;
2975         struct nfs4_state *state;
2976         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
2977         unsigned int seq;
2978         int ret;
2979
2980         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2981
2982         ret = _nfs4_proc_open(opendata, ctx);
2983         if (ret != 0)
2984                 goto out;
2985
2986         state = _nfs4_opendata_to_nfs4_state(opendata);
2987         ret = PTR_ERR(state);
2988         if (IS_ERR(state))
2989                 goto out;
2990         ctx->state = state;
2991         if (server->caps & NFS_CAP_POSIX_LOCK)
2992                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2993         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2994                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2995
2996         dentry = opendata->dentry;
2997         if (d_really_is_negative(dentry)) {
2998                 struct dentry *alias;
2999                 d_drop(dentry);
3000                 alias = d_exact_alias(dentry, state->inode);
3001                 if (!alias)
3002                         alias = d_splice_alias(igrab(state->inode), dentry);
3003                 /* d_splice_alias() can't fail here - it's a non-directory */
3004                 if (alias) {
3005                         dput(ctx->dentry);
3006                         ctx->dentry = dentry = alias;
3007                 }
3008                 nfs_set_verifier(dentry,
3009                                 nfs_save_change_attribute(d_inode(opendata->dir)));
3010         }
3011
3012         /* Parse layoutget results before we check for access */
3013         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3014
3015         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3016                         acc_mode, flags);
3017         if (ret != 0)
3018                 goto out;
3019
3020         if (d_inode(dentry) == state->inode) {
3021                 nfs_inode_attach_open_context(ctx);
3022                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3023                         nfs4_schedule_stateid_recovery(server, state);
3024         }
3025
3026 out:
3027         if (!opendata->cancelled)
3028                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3029         return ret;
3030 }
3031
3032 /*
3033  * Returns a referenced nfs4_state
3034  */
3035 static int _nfs4_do_open(struct inode *dir,
3036                         struct nfs_open_context *ctx,
3037                         int flags,
3038                         const struct nfs4_open_createattrs *c,
3039                         int *opened)
3040 {
3041         struct nfs4_state_owner  *sp;
3042         struct nfs4_state     *state = NULL;
3043         struct nfs_server       *server = NFS_SERVER(dir);
3044         struct nfs4_opendata *opendata;
3045         struct dentry *dentry = ctx->dentry;
3046         const struct cred *cred = ctx->cred;
3047         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3048         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3049         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3050         struct iattr *sattr = c->sattr;
3051         struct nfs4_label *label = c->label;
3052         struct nfs4_label *olabel = NULL;
3053         int status;
3054
3055         /* Protect against reboot recovery conflicts */
3056         status = -ENOMEM;
3057         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3058         if (sp == NULL) {
3059                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3060                 goto out_err;
3061         }
3062         status = nfs4_client_recover_expired_lease(server->nfs_client);
3063         if (status != 0)
3064                 goto err_put_state_owner;
3065         if (d_really_is_positive(dentry))
3066                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3067         status = -ENOMEM;
3068         if (d_really_is_positive(dentry))
3069                 claim = NFS4_OPEN_CLAIM_FH;
3070         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3071                         c, claim, GFP_KERNEL);
3072         if (opendata == NULL)
3073                 goto err_put_state_owner;
3074
3075         if (label) {
3076                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3077                 if (IS_ERR(olabel)) {
3078                         status = PTR_ERR(olabel);
3079                         goto err_opendata_put;
3080                 }
3081         }
3082
3083         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3084                 if (!opendata->f_attr.mdsthreshold) {
3085                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3086                         if (!opendata->f_attr.mdsthreshold)
3087                                 goto err_free_label;
3088                 }
3089                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3090         }
3091         if (d_really_is_positive(dentry))
3092                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3093
3094         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3095         if (status != 0)
3096                 goto err_free_label;
3097         state = ctx->state;
3098
3099         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3100             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3101                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3102                 /*
3103                  * send create attributes which was not set by open
3104                  * with an extra setattr.
3105                  */
3106                 if (attrs || label) {
3107                         unsigned ia_old = sattr->ia_valid;
3108
3109                         sattr->ia_valid = attrs;
3110                         nfs_fattr_init(opendata->o_res.f_attr);
3111                         status = nfs4_do_setattr(state->inode, cred,
3112                                         opendata->o_res.f_attr, sattr,
3113                                         ctx, label, olabel);
3114                         if (status == 0) {
3115                                 nfs_setattr_update_inode(state->inode, sattr,
3116                                                 opendata->o_res.f_attr);
3117                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3118                         }
3119                         sattr->ia_valid = ia_old;
3120                 }
3121         }
3122         if (opened && opendata->file_created)
3123                 *opened = 1;
3124
3125         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3126                 *ctx_th = opendata->f_attr.mdsthreshold;
3127                 opendata->f_attr.mdsthreshold = NULL;
3128         }
3129
3130         nfs4_label_free(olabel);
3131
3132         nfs4_opendata_put(opendata);
3133         nfs4_put_state_owner(sp);
3134         return 0;
3135 err_free_label:
3136         nfs4_label_free(olabel);
3137 err_opendata_put:
3138         nfs4_opendata_put(opendata);
3139 err_put_state_owner:
3140         nfs4_put_state_owner(sp);
3141 out_err:
3142         return status;
3143 }
3144
3145
3146 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3147                                         struct nfs_open_context *ctx,
3148                                         int flags,
3149                                         struct iattr *sattr,
3150                                         struct nfs4_label *label,
3151                                         int *opened)
3152 {
3153         struct nfs_server *server = NFS_SERVER(dir);
3154         struct nfs4_exception exception = {
3155                 .interruptible = true,
3156         };
3157         struct nfs4_state *res;
3158         struct nfs4_open_createattrs c = {
3159                 .label = label,
3160                 .sattr = sattr,
3161                 .verf = {
3162                         [0] = (__u32)jiffies,
3163                         [1] = (__u32)current->pid,
3164                 },
3165         };
3166         int status;
3167
3168         do {
3169                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3170                 res = ctx->state;
3171                 trace_nfs4_open_file(ctx, flags, status);
3172                 if (status == 0)
3173                         break;
3174                 /* NOTE: BAD_SEQID means the server and client disagree about the
3175                  * book-keeping w.r.t. state-changing operations
3176                  * (OPEN/CLOSE/LOCK/LOCKU...)
3177                  * It is actually a sign of a bug on the client or on the server.
3178                  *
3179                  * If we receive a BAD_SEQID error in the particular case of
3180                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3181                  * have unhashed the old state_owner for us, and that we can
3182                  * therefore safely retry using a new one. We should still warn
3183                  * the user though...
3184                  */
3185                 if (status == -NFS4ERR_BAD_SEQID) {
3186                         pr_warn_ratelimited("NFS: v4 server %s "
3187                                         " returned a bad sequence-id error!\n",
3188                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3189                         exception.retry = 1;
3190                         continue;
3191                 }
3192                 /*
3193                  * BAD_STATEID on OPEN means that the server cancelled our
3194                  * state before it received the OPEN_CONFIRM.
3195                  * Recover by retrying the request as per the discussion
3196                  * on Page 181 of RFC3530.
3197                  */
3198                 if (status == -NFS4ERR_BAD_STATEID) {
3199                         exception.retry = 1;
3200                         continue;
3201                 }
3202                 if (status == -NFS4ERR_EXPIRED) {
3203                         nfs4_schedule_lease_recovery(server->nfs_client);
3204                         exception.retry = 1;
3205                         continue;
3206                 }
3207                 if (status == -EAGAIN) {
3208                         /* We must have found a delegation */
3209                         exception.retry = 1;
3210                         continue;
3211                 }
3212                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3213                         continue;
3214                 res = ERR_PTR(nfs4_handle_exception(server,
3215                                         status, &exception));
3216         } while (exception.retry);
3217         return res;
3218 }
3219
3220 static int _nfs4_do_setattr(struct inode *inode,
3221                             struct nfs_setattrargs *arg,
3222                             struct nfs_setattrres *res,
3223                             const struct cred *cred,
3224                             struct nfs_open_context *ctx)
3225 {
3226         struct nfs_server *server = NFS_SERVER(inode);
3227         struct rpc_message msg = {
3228                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3229                 .rpc_argp       = arg,
3230                 .rpc_resp       = res,
3231                 .rpc_cred       = cred,
3232         };
3233         const struct cred *delegation_cred = NULL;
3234         unsigned long timestamp = jiffies;
3235         bool truncate;
3236         int status;
3237
3238         nfs_fattr_init(res->fattr);
3239
3240         /* Servers should only apply open mode checks for file size changes */
3241         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3242         if (!truncate)
3243                 goto zero_stateid;
3244
3245         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3246                 /* Use that stateid */
3247         } else if (ctx != NULL && ctx->state) {
3248                 struct nfs_lock_context *l_ctx;
3249                 if (!nfs4_valid_open_stateid(ctx->state))
3250                         return -EBADF;
3251                 l_ctx = nfs_get_lock_context(ctx);
3252                 if (IS_ERR(l_ctx))
3253                         return PTR_ERR(l_ctx);
3254                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3255                                                 &arg->stateid, &delegation_cred);
3256                 nfs_put_lock_context(l_ctx);
3257                 if (status == -EIO)
3258                         return -EBADF;
3259                 else if (status == -EAGAIN)
3260                         goto zero_stateid;
3261         } else {
3262 zero_stateid:
3263                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3264         }
3265         if (delegation_cred)
3266                 msg.rpc_cred = delegation_cred;
3267
3268         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3269
3270         put_cred(delegation_cred);
3271         if (status == 0 && ctx != NULL)
3272                 renew_lease(server, timestamp);
3273         trace_nfs4_setattr(inode, &arg->stateid, status);
3274         return status;
3275 }
3276
3277 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3278                            struct nfs_fattr *fattr, struct iattr *sattr,
3279                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3280                            struct nfs4_label *olabel)
3281 {
3282         struct nfs_server *server = NFS_SERVER(inode);
3283         __u32 bitmask[NFS4_BITMASK_SZ];
3284         struct nfs4_state *state = ctx ? ctx->state : NULL;
3285         struct nfs_setattrargs  arg = {
3286                 .fh             = NFS_FH(inode),
3287                 .iap            = sattr,
3288                 .server         = server,
3289                 .bitmask = bitmask,
3290                 .label          = ilabel,
3291         };
3292         struct nfs_setattrres  res = {
3293                 .fattr          = fattr,
3294                 .label          = olabel,
3295                 .server         = server,
3296         };
3297         struct nfs4_exception exception = {
3298                 .state = state,
3299                 .inode = inode,
3300                 .stateid = &arg.stateid,
3301         };
3302         int err;
3303
3304         do {
3305                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3306                                 nfs4_bitmask(server, olabel),
3307                                 inode);
3308
3309                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3310                 switch (err) {
3311                 case -NFS4ERR_OPENMODE:
3312                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3313                                 pr_warn_once("NFSv4: server %s is incorrectly "
3314                                                 "applying open mode checks to "
3315                                                 "a SETATTR that is not "
3316                                                 "changing file size.\n",
3317                                                 server->nfs_client->cl_hostname);
3318                         }
3319                         if (state && !(state->state & FMODE_WRITE)) {
3320                                 err = -EBADF;
3321                                 if (sattr->ia_valid & ATTR_OPEN)
3322                                         err = -EACCES;
3323                                 goto out;
3324                         }
3325                 }
3326                 err = nfs4_handle_exception(server, err, &exception);
3327         } while (exception.retry);
3328 out:
3329         return err;
3330 }
3331
3332 static bool
3333 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3334 {
3335         if (inode == NULL || !nfs_have_layout(inode))
3336                 return false;
3337
3338         return pnfs_wait_on_layoutreturn(inode, task);
3339 }
3340
3341 /*
3342  * Update the seqid of an open stateid
3343  */
3344 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3345                 struct nfs4_state *state)
3346 {
3347         __be32 seqid_open;
3348         u32 dst_seqid;
3349         int seq;
3350
3351         for (;;) {
3352                 if (!nfs4_valid_open_stateid(state))
3353                         break;
3354                 seq = read_seqbegin(&state->seqlock);
3355                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3356                         nfs4_stateid_copy(dst, &state->open_stateid);
3357                         if (read_seqretry(&state->seqlock, seq))
3358                                 continue;
3359                         break;
3360                 }
3361                 seqid_open = state->open_stateid.seqid;
3362                 if (read_seqretry(&state->seqlock, seq))
3363                         continue;
3364
3365                 dst_seqid = be32_to_cpu(dst->seqid);
3366                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3367                         dst->seqid = seqid_open;
3368                 break;
3369         }
3370 }
3371
3372 /*
3373  * Update the seqid of an open stateid after receiving
3374  * NFS4ERR_OLD_STATEID
3375  */
3376 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3377                 struct nfs4_state *state)
3378 {
3379         __be32 seqid_open;
3380         u32 dst_seqid;
3381         bool ret;
3382         int seq;
3383
3384         for (;;) {
3385                 ret = false;
3386                 if (!nfs4_valid_open_stateid(state))
3387                         break;
3388                 seq = read_seqbegin(&state->seqlock);
3389                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
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 = cpu_to_be32(dst_seqid + 1);
3401                 else
3402                         dst->seqid = seqid_open;
3403                 ret = true;
3404                 break;
3405         }
3406
3407         return ret;
3408 }
3409
3410 struct nfs4_closedata {
3411         struct inode *inode;
3412         struct nfs4_state *state;
3413         struct nfs_closeargs arg;
3414         struct nfs_closeres res;
3415         struct {
3416                 struct nfs4_layoutreturn_args arg;
3417                 struct nfs4_layoutreturn_res res;
3418                 struct nfs4_xdr_opaque_data ld_private;
3419                 u32 roc_barrier;
3420                 bool roc;
3421         } lr;
3422         struct nfs_fattr fattr;
3423         unsigned long timestamp;
3424 };
3425
3426 static void nfs4_free_closedata(void *data)
3427 {
3428         struct nfs4_closedata *calldata = data;
3429         struct nfs4_state_owner *sp = calldata->state->owner;
3430         struct super_block *sb = calldata->state->inode->i_sb;
3431
3432         if (calldata->lr.roc)
3433                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3434                                 calldata->res.lr_ret);
3435         nfs4_put_open_state(calldata->state);
3436         nfs_free_seqid(calldata->arg.seqid);
3437         nfs4_put_state_owner(sp);
3438         nfs_sb_deactive(sb);
3439         kfree(calldata);
3440 }
3441
3442 static void nfs4_close_done(struct rpc_task *task, void *data)
3443 {
3444         struct nfs4_closedata *calldata = data;
3445         struct nfs4_state *state = calldata->state;
3446         struct nfs_server *server = NFS_SERVER(calldata->inode);
3447         nfs4_stateid *res_stateid = NULL;
3448         struct nfs4_exception exception = {
3449                 .state = state,
3450                 .inode = calldata->inode,
3451                 .stateid = &calldata->arg.stateid,
3452         };
3453
3454         dprintk("%s: begin!\n", __func__);
3455         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3456                 return;
3457         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3458
3459         /* Handle Layoutreturn errors */
3460         if (pnfs_roc_done(task, calldata->inode,
3461                                 &calldata->arg.lr_args,
3462                                 &calldata->res.lr_res,
3463                                 &calldata->res.lr_ret) == -EAGAIN)
3464                 goto out_restart;
3465
3466         /* hmm. we are done with the inode, and in the process of freeing
3467          * the state_owner. we keep this around to process errors
3468          */
3469         switch (task->tk_status) {
3470                 case 0:
3471                         res_stateid = &calldata->res.stateid;
3472                         renew_lease(server, calldata->timestamp);
3473                         break;
3474                 case -NFS4ERR_ACCESS:
3475                         if (calldata->arg.bitmask != NULL) {
3476                                 calldata->arg.bitmask = NULL;
3477                                 calldata->res.fattr = NULL;
3478                                 goto out_restart;
3479
3480                         }
3481                         break;
3482                 case -NFS4ERR_OLD_STATEID:
3483                         /* Did we race with OPEN? */
3484                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3485                                                 state))
3486                                 goto out_restart;
3487                         goto out_release;
3488                 case -NFS4ERR_ADMIN_REVOKED:
3489                 case -NFS4ERR_STALE_STATEID:
3490                 case -NFS4ERR_EXPIRED:
3491                         nfs4_free_revoked_stateid(server,
3492                                         &calldata->arg.stateid,
3493                                         task->tk_msg.rpc_cred);
3494                         /* Fallthrough */
3495                 case -NFS4ERR_BAD_STATEID:
3496                         if (calldata->arg.fmode == 0)
3497                                 break;
3498                         /* Fallthrough */
3499                 default:
3500                         task->tk_status = nfs4_async_handle_exception(task,
3501                                         server, task->tk_status, &exception);
3502                         if (exception.retry)
3503                                 goto out_restart;
3504         }
3505         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3506                         res_stateid, calldata->arg.fmode);
3507 out_release:
3508         task->tk_status = 0;
3509         nfs_release_seqid(calldata->arg.seqid);
3510         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3511         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3512         return;
3513 out_restart:
3514         task->tk_status = 0;
3515         rpc_restart_call_prepare(task);
3516         goto out_release;
3517 }
3518
3519 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3520 {
3521         struct nfs4_closedata *calldata = data;
3522         struct nfs4_state *state = calldata->state;
3523         struct inode *inode = calldata->inode;
3524         struct pnfs_layout_hdr *lo;
3525         bool is_rdonly, is_wronly, is_rdwr;
3526         int call_close = 0;
3527
3528         dprintk("%s: begin!\n", __func__);
3529         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3530                 goto out_wait;
3531
3532         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3533         spin_lock(&state->owner->so_lock);
3534         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3535         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3536         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3537         /* Calculate the change in open mode */
3538         calldata->arg.fmode = 0;
3539         if (state->n_rdwr == 0) {
3540                 if (state->n_rdonly == 0)
3541                         call_close |= is_rdonly;
3542                 else if (is_rdonly)
3543                         calldata->arg.fmode |= FMODE_READ;
3544                 if (state->n_wronly == 0)
3545                         call_close |= is_wronly;
3546                 else if (is_wronly)
3547                         calldata->arg.fmode |= FMODE_WRITE;
3548                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3549                         call_close |= is_rdwr;
3550         } else if (is_rdwr)
3551                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3552
3553         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3554         if (!nfs4_valid_open_stateid(state))
3555                 call_close = 0;
3556         spin_unlock(&state->owner->so_lock);
3557
3558         if (!call_close) {
3559                 /* Note: exit _without_ calling nfs4_close_done */
3560                 goto out_no_action;
3561         }
3562
3563         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3564                 nfs_release_seqid(calldata->arg.seqid);
3565                 goto out_wait;
3566         }
3567
3568         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3569         if (lo && !pnfs_layout_is_valid(lo)) {
3570                 calldata->arg.lr_args = NULL;
3571                 calldata->res.lr_res = NULL;
3572         }
3573
3574         if (calldata->arg.fmode == 0)
3575                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3576
3577         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3578                 /* Close-to-open cache consistency revalidation */
3579                 if (!nfs4_have_delegation(inode, FMODE_READ))
3580                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3581                 else
3582                         calldata->arg.bitmask = NULL;
3583         }
3584
3585         calldata->arg.share_access =
3586                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3587                                 calldata->arg.fmode, 0);
3588
3589         if (calldata->res.fattr == NULL)
3590                 calldata->arg.bitmask = NULL;
3591         else if (calldata->arg.bitmask == NULL)
3592                 calldata->res.fattr = NULL;
3593         calldata->timestamp = jiffies;
3594         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3595                                 &calldata->arg.seq_args,
3596                                 &calldata->res.seq_res,
3597                                 task) != 0)
3598                 nfs_release_seqid(calldata->arg.seqid);
3599         dprintk("%s: done!\n", __func__);
3600         return;
3601 out_no_action:
3602         task->tk_action = NULL;
3603 out_wait:
3604         nfs4_sequence_done(task, &calldata->res.seq_res);
3605 }
3606
3607 static const struct rpc_call_ops nfs4_close_ops = {
3608         .rpc_call_prepare = nfs4_close_prepare,
3609         .rpc_call_done = nfs4_close_done,
3610         .rpc_release = nfs4_free_closedata,
3611 };
3612
3613 /* 
3614  * It is possible for data to be read/written from a mem-mapped file 
3615  * after the sys_close call (which hits the vfs layer as a flush).
3616  * This means that we can't safely call nfsv4 close on a file until 
3617  * the inode is cleared. This in turn means that we are not good
3618  * NFSv4 citizens - we do not indicate to the server to update the file's 
3619  * share state even when we are done with one of the three share 
3620  * stateid's in the inode.
3621  *
3622  * NOTE: Caller must be holding the sp->so_owner semaphore!
3623  */
3624 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3625 {
3626         struct nfs_server *server = NFS_SERVER(state->inode);
3627         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3628         struct nfs4_closedata *calldata;
3629         struct nfs4_state_owner *sp = state->owner;
3630         struct rpc_task *task;
3631         struct rpc_message msg = {
3632                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3633                 .rpc_cred = state->owner->so_cred,
3634         };
3635         struct rpc_task_setup task_setup_data = {
3636                 .rpc_client = server->client,
3637                 .rpc_message = &msg,
3638                 .callback_ops = &nfs4_close_ops,
3639                 .workqueue = nfsiod_workqueue,
3640                 .flags = RPC_TASK_ASYNC,
3641         };
3642         int status = -ENOMEM;
3643
3644         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3645                 &task_setup_data.rpc_client, &msg);
3646
3647         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3648         if (calldata == NULL)
3649                 goto out;
3650         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3651         calldata->inode = state->inode;
3652         calldata->state = state;
3653         calldata->arg.fh = NFS_FH(state->inode);
3654         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3655                 goto out_free_calldata;
3656         /* Serialization for the sequence id */
3657         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3658         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3659         if (IS_ERR(calldata->arg.seqid))
3660                 goto out_free_calldata;
3661         nfs_fattr_init(&calldata->fattr);
3662         calldata->arg.fmode = 0;
3663         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3664         calldata->res.fattr = &calldata->fattr;
3665         calldata->res.seqid = calldata->arg.seqid;
3666         calldata->res.server = server;
3667         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3668         calldata->lr.roc = pnfs_roc(state->inode,
3669                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3670         if (calldata->lr.roc) {
3671                 calldata->arg.lr_args = &calldata->lr.arg;
3672                 calldata->res.lr_res = &calldata->lr.res;
3673         }
3674         nfs_sb_active(calldata->inode->i_sb);
3675
3676         msg.rpc_argp = &calldata->arg;
3677         msg.rpc_resp = &calldata->res;
3678         task_setup_data.callback_data = calldata;
3679         task = rpc_run_task(&task_setup_data);
3680         if (IS_ERR(task))
3681                 return PTR_ERR(task);
3682         status = 0;
3683         if (wait)
3684                 status = rpc_wait_for_completion_task(task);
3685         rpc_put_task(task);
3686         return status;
3687 out_free_calldata:
3688         kfree(calldata);
3689 out:
3690         nfs4_put_open_state(state);
3691         nfs4_put_state_owner(sp);
3692         return status;
3693 }
3694
3695 static struct inode *
3696 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3697                 int open_flags, struct iattr *attr, int *opened)
3698 {
3699         struct nfs4_state *state;
3700         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3701
3702         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3703
3704         /* Protect against concurrent sillydeletes */
3705         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3706
3707         nfs4_label_release_security(label);
3708
3709         if (IS_ERR(state))
3710                 return ERR_CAST(state);
3711         return state->inode;
3712 }
3713
3714 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3715 {
3716         if (ctx->state == NULL)
3717                 return;
3718         if (is_sync)
3719                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3720         else
3721                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3722 }
3723
3724 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3725 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3726 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3727
3728 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3729 {
3730         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3731         struct nfs4_server_caps_arg args = {
3732                 .fhandle = fhandle,
3733                 .bitmask = bitmask,
3734         };
3735         struct nfs4_server_caps_res res = {};
3736         struct rpc_message msg = {
3737                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3738                 .rpc_argp = &args,
3739                 .rpc_resp = &res,
3740         };
3741         int status;
3742         int i;
3743
3744         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3745                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3746                      FATTR4_WORD0_LINK_SUPPORT |
3747                      FATTR4_WORD0_SYMLINK_SUPPORT |
3748                      FATTR4_WORD0_ACLSUPPORT;
3749         if (minorversion)
3750                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3751
3752         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3753         if (status == 0) {
3754                 /* Sanity check the server answers */
3755                 switch (minorversion) {
3756                 case 0:
3757                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3758                         res.attr_bitmask[2] = 0;
3759                         break;
3760                 case 1:
3761                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3762                         break;
3763                 case 2:
3764                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3765                 }
3766                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3767                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3768                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3769                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3770                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3771                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3772                                 NFS_CAP_SECURITY_LABEL);
3773                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3774                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3775                         server->caps |= NFS_CAP_ACLS;
3776                 if (res.has_links != 0)
3777                         server->caps |= NFS_CAP_HARDLINKS;
3778                 if (res.has_symlinks != 0)
3779                         server->caps |= NFS_CAP_SYMLINKS;
3780                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3781                         server->caps |= NFS_CAP_FILEID;
3782                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3783                         server->caps |= NFS_CAP_MODE;
3784                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3785                         server->caps |= NFS_CAP_NLINK;
3786                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3787                         server->caps |= NFS_CAP_OWNER;
3788                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3789                         server->caps |= NFS_CAP_OWNER_GROUP;
3790                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3791                         server->caps |= NFS_CAP_ATIME;
3792                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3793                         server->caps |= NFS_CAP_CTIME;
3794                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3795                         server->caps |= NFS_CAP_MTIME;
3796 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3797                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3798                         server->caps |= NFS_CAP_SECURITY_LABEL;
3799 #endif
3800                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3801                                 sizeof(server->attr_bitmask));
3802                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3803
3804                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3805                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3806                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3807                 server->cache_consistency_bitmask[2] = 0;
3808
3809                 /* Avoid a regression due to buggy server */
3810                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3811                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3812                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3813                         sizeof(server->exclcreat_bitmask));
3814
3815                 server->acl_bitmask = res.acl_bitmask;
3816                 server->fh_expire_type = res.fh_expire_type;
3817         }
3818
3819         return status;
3820 }
3821
3822 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3823 {
3824         struct nfs4_exception exception = {
3825                 .interruptible = true,
3826         };
3827         int err;
3828         do {
3829                 err = nfs4_handle_exception(server,
3830                                 _nfs4_server_capabilities(server, fhandle),
3831                                 &exception);
3832         } while (exception.retry);
3833         return err;
3834 }
3835
3836 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3837                 struct nfs_fsinfo *info)
3838 {
3839         u32 bitmask[3];
3840         struct nfs4_lookup_root_arg args = {
3841                 .bitmask = bitmask,
3842         };
3843         struct nfs4_lookup_res res = {
3844                 .server = server,
3845                 .fattr = info->fattr,
3846                 .fh = fhandle,
3847         };
3848         struct rpc_message msg = {
3849                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3850                 .rpc_argp = &args,
3851                 .rpc_resp = &res,
3852         };
3853
3854         bitmask[0] = nfs4_fattr_bitmap[0];
3855         bitmask[1] = nfs4_fattr_bitmap[1];
3856         /*
3857          * Process the label in the upcoming getfattr
3858          */
3859         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3860
3861         nfs_fattr_init(info->fattr);
3862         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3863 }
3864
3865 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3866                 struct nfs_fsinfo *info)
3867 {
3868         struct nfs4_exception exception = {
3869                 .interruptible = true,
3870         };
3871         int err;
3872         do {
3873                 err = _nfs4_lookup_root(server, fhandle, info);
3874                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3875                 switch (err) {
3876                 case 0:
3877                 case -NFS4ERR_WRONGSEC:
3878                         goto out;
3879                 default:
3880                         err = nfs4_handle_exception(server, err, &exception);
3881                 }
3882         } while (exception.retry);
3883 out:
3884         return err;
3885 }
3886
3887 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3888                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3889 {
3890         struct rpc_auth_create_args auth_args = {
3891                 .pseudoflavor = flavor,
3892         };
3893         struct rpc_auth *auth;
3894
3895         auth = rpcauth_create(&auth_args, server->client);
3896         if (IS_ERR(auth))
3897                 return -EACCES;
3898         return nfs4_lookup_root(server, fhandle, info);
3899 }
3900
3901 /*
3902  * Retry pseudoroot lookup with various security flavors.  We do this when:
3903  *
3904  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3905  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3906  *
3907  * Returns zero on success, or a negative NFS4ERR value, or a
3908  * negative errno value.
3909  */
3910 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3911                               struct nfs_fsinfo *info)
3912 {
3913         /* Per 3530bis 15.33.5 */
3914         static const rpc_authflavor_t flav_array[] = {
3915                 RPC_AUTH_GSS_KRB5P,
3916                 RPC_AUTH_GSS_KRB5I,
3917                 RPC_AUTH_GSS_KRB5,
3918                 RPC_AUTH_UNIX,                  /* courtesy */
3919                 RPC_AUTH_NULL,
3920         };
3921         int status = -EPERM;
3922         size_t i;
3923
3924         if (server->auth_info.flavor_len > 0) {
3925                 /* try each flavor specified by user */
3926                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3927                         status = nfs4_lookup_root_sec(server, fhandle, info,
3928                                                 server->auth_info.flavors[i]);
3929                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3930                                 continue;
3931                         break;
3932                 }
3933         } else {
3934                 /* no flavors specified by user, try default list */
3935                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3936                         status = nfs4_lookup_root_sec(server, fhandle, info,
3937                                                       flav_array[i]);
3938                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3939                                 continue;
3940                         break;
3941                 }
3942         }
3943
3944         /*
3945          * -EACCES could mean that the user doesn't have correct permissions
3946          * to access the mount.  It could also mean that we tried to mount
3947          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3948          * existing mount programs don't handle -EACCES very well so it should
3949          * be mapped to -EPERM instead.
3950          */
3951         if (status == -EACCES)
3952                 status = -EPERM;
3953         return status;
3954 }
3955
3956 /**
3957  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3958  * @server: initialized nfs_server handle
3959  * @fhandle: we fill in the pseudo-fs root file handle
3960  * @info: we fill in an FSINFO struct
3961  * @auth_probe: probe the auth flavours
3962  *
3963  * Returns zero on success, or a negative errno.
3964  */
3965 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3966                          struct nfs_fsinfo *info,
3967                          bool auth_probe)
3968 {
3969         int status = 0;
3970
3971         if (!auth_probe)
3972                 status = nfs4_lookup_root(server, fhandle, info);
3973
3974         if (auth_probe || status == NFS4ERR_WRONGSEC)
3975                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3976                                 fhandle, info);
3977
3978         if (status == 0)
3979                 status = nfs4_server_capabilities(server, fhandle);
3980         if (status == 0)
3981                 status = nfs4_do_fsinfo(server, fhandle, info);
3982
3983         return nfs4_map_errors(status);
3984 }
3985
3986 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3987                               struct nfs_fsinfo *info)
3988 {
3989         int error;
3990         struct nfs_fattr *fattr = info->fattr;
3991         struct nfs4_label *label = NULL;
3992
3993         error = nfs4_server_capabilities(server, mntfh);
3994         if (error < 0) {
3995                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3996                 return error;
3997         }
3998
3999         label = nfs4_label_alloc(server, GFP_KERNEL);
4000         if (IS_ERR(label))
4001                 return PTR_ERR(label);
4002
4003         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4004         if (error < 0) {
4005                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4006                 goto err_free_label;
4007         }
4008
4009         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4010             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4011                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4012
4013 err_free_label:
4014         nfs4_label_free(label);
4015
4016         return error;
4017 }
4018
4019 /*
4020  * Get locations and (maybe) other attributes of a referral.
4021  * Note that we'll actually follow the referral later when
4022  * we detect fsid mismatch in inode revalidation
4023  */
4024 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4025                              const struct qstr *name, struct nfs_fattr *fattr,
4026                              struct nfs_fh *fhandle)
4027 {
4028         int status = -ENOMEM;
4029         struct page *page = NULL;
4030         struct nfs4_fs_locations *locations = NULL;
4031
4032         page = alloc_page(GFP_KERNEL);
4033         if (page == NULL)
4034                 goto out;
4035         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4036         if (locations == NULL)
4037                 goto out;
4038
4039         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4040         if (status != 0)
4041                 goto out;
4042
4043         /*
4044          * If the fsid didn't change, this is a migration event, not a
4045          * referral.  Cause us to drop into the exception handler, which
4046          * will kick off migration recovery.
4047          */
4048         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4049                 dprintk("%s: server did not return a different fsid for"
4050                         " a referral at %s\n", __func__, name->name);
4051                 status = -NFS4ERR_MOVED;
4052                 goto out;
4053         }
4054         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4055         nfs_fixup_referral_attributes(&locations->fattr);
4056
4057         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4058         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4059         memset(fhandle, 0, sizeof(struct nfs_fh));
4060 out:
4061         if (page)
4062                 __free_page(page);
4063         kfree(locations);
4064         return status;
4065 }
4066
4067 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4068                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4069                                 struct inode *inode)
4070 {
4071         __u32 bitmask[NFS4_BITMASK_SZ];
4072         struct nfs4_getattr_arg args = {
4073                 .fh = fhandle,
4074                 .bitmask = bitmask,
4075         };
4076         struct nfs4_getattr_res res = {
4077                 .fattr = fattr,
4078                 .label = label,
4079                 .server = server,
4080         };
4081         struct rpc_message msg = {
4082                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4083                 .rpc_argp = &args,
4084                 .rpc_resp = &res,
4085         };
4086         unsigned short task_flags = 0;
4087
4088         /* Is this is an attribute revalidation, subject to softreval? */
4089         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4090                 task_flags |= RPC_TASK_TIMEOUT;
4091
4092         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4093
4094         nfs_fattr_init(fattr);
4095         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4096         return nfs4_do_call_sync(server->client, server, &msg,
4097                         &args.seq_args, &res.seq_res, task_flags);
4098 }
4099
4100 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4101                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4102                                 struct inode *inode)
4103 {
4104         struct nfs4_exception exception = {
4105                 .interruptible = true,
4106         };
4107         int err;
4108         do {
4109                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4110                 trace_nfs4_getattr(server, fhandle, fattr, err);
4111                 err = nfs4_handle_exception(server, err,
4112                                 &exception);
4113         } while (exception.retry);
4114         return err;
4115 }
4116
4117 /* 
4118  * The file is not closed if it is opened due to the a request to change
4119  * the size of the file. The open call will not be needed once the
4120  * VFS layer lookup-intents are implemented.
4121  *
4122  * Close is called when the inode is destroyed.
4123  * If we haven't opened the file for O_WRONLY, we
4124  * need to in the size_change case to obtain a stateid.
4125  *
4126  * Got race?
4127  * Because OPEN is always done by name in nfsv4, it is
4128  * possible that we opened a different file by the same
4129  * name.  We can recognize this race condition, but we
4130  * can't do anything about it besides returning an error.
4131  *
4132  * This will be fixed with VFS changes (lookup-intent).
4133  */
4134 static int
4135 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4136                   struct iattr *sattr)
4137 {
4138         struct inode *inode = d_inode(dentry);
4139         const struct cred *cred = NULL;
4140         struct nfs_open_context *ctx = NULL;
4141         struct nfs4_label *label = NULL;
4142         int status;
4143
4144         if (pnfs_ld_layoutret_on_setattr(inode) &&
4145             sattr->ia_valid & ATTR_SIZE &&
4146             sattr->ia_size < i_size_read(inode))
4147                 pnfs_commit_and_return_layout(inode);
4148
4149         nfs_fattr_init(fattr);
4150         
4151         /* Deal with open(O_TRUNC) */
4152         if (sattr->ia_valid & ATTR_OPEN)
4153                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4154
4155         /* Optimization: if the end result is no change, don't RPC */
4156         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4157                 return 0;
4158
4159         /* Search for an existing open(O_WRITE) file */
4160         if (sattr->ia_valid & ATTR_FILE) {
4161
4162                 ctx = nfs_file_open_context(sattr->ia_file);
4163                 if (ctx)
4164                         cred = ctx->cred;
4165         }
4166
4167         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4168         if (IS_ERR(label))
4169                 return PTR_ERR(label);
4170
4171         /* Return any delegations if we're going to change ACLs */
4172         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4173                 nfs4_inode_make_writeable(inode);
4174
4175         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4176         if (status == 0) {
4177                 nfs_setattr_update_inode(inode, sattr, fattr);
4178                 nfs_setsecurity(inode, fattr, label);
4179         }
4180         nfs4_label_free(label);
4181         return status;
4182 }
4183
4184 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4185                 struct dentry *dentry, struct nfs_fh *fhandle,
4186                 struct nfs_fattr *fattr, struct nfs4_label *label)
4187 {
4188         struct nfs_server *server = NFS_SERVER(dir);
4189         int                    status;
4190         struct nfs4_lookup_arg args = {
4191                 .bitmask = server->attr_bitmask,
4192                 .dir_fh = NFS_FH(dir),
4193                 .name = &dentry->d_name,
4194         };
4195         struct nfs4_lookup_res res = {
4196                 .server = server,
4197                 .fattr = fattr,
4198                 .label = label,
4199                 .fh = fhandle,
4200         };
4201         struct rpc_message msg = {
4202                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4203                 .rpc_argp = &args,
4204                 .rpc_resp = &res,
4205         };
4206         unsigned short task_flags = 0;
4207
4208         /* Is this is an attribute revalidation, subject to softreval? */
4209         if (nfs_lookup_is_soft_revalidate(dentry))
4210                 task_flags |= RPC_TASK_TIMEOUT;
4211
4212         args.bitmask = nfs4_bitmask(server, label);
4213
4214         nfs_fattr_init(fattr);
4215
4216         dprintk("NFS call  lookup %pd2\n", dentry);
4217         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4218         status = nfs4_do_call_sync(clnt, server, &msg,
4219                         &args.seq_args, &res.seq_res, task_flags);
4220         dprintk("NFS reply lookup: %d\n", status);
4221         return status;
4222 }
4223
4224 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4225 {
4226         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4227                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4228         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4229         fattr->nlink = 2;
4230 }
4231
4232 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4233                                    struct dentry *dentry, struct nfs_fh *fhandle,
4234                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4235 {
4236         struct nfs4_exception exception = {
4237                 .interruptible = true,
4238         };
4239         struct rpc_clnt *client = *clnt;
4240         const struct qstr *name = &dentry->d_name;
4241         int err;
4242         do {
4243                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4244                 trace_nfs4_lookup(dir, name, err);
4245                 switch (err) {
4246                 case -NFS4ERR_BADNAME:
4247                         err = -ENOENT;
4248                         goto out;
4249                 case -NFS4ERR_MOVED:
4250                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4251                         if (err == -NFS4ERR_MOVED)
4252                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4253                         goto out;
4254                 case -NFS4ERR_WRONGSEC:
4255                         err = -EPERM;
4256                         if (client != *clnt)
4257                                 goto out;
4258                         client = nfs4_negotiate_security(client, dir, name);
4259                         if (IS_ERR(client))
4260                                 return PTR_ERR(client);
4261
4262                         exception.retry = 1;
4263                         break;
4264                 default:
4265                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4266                 }
4267         } while (exception.retry);
4268
4269 out:
4270         if (err == 0)
4271                 *clnt = client;
4272         else if (client != *clnt)
4273                 rpc_shutdown_client(client);
4274
4275         return err;
4276 }
4277
4278 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4279                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4280                             struct nfs4_label *label)
4281 {
4282         int status;
4283         struct rpc_clnt *client = NFS_CLIENT(dir);
4284
4285         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4286         if (client != NFS_CLIENT(dir)) {
4287                 rpc_shutdown_client(client);
4288                 nfs_fixup_secinfo_attributes(fattr);
4289         }
4290         return status;
4291 }
4292
4293 struct rpc_clnt *
4294 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4295                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4296 {
4297         struct rpc_clnt *client = NFS_CLIENT(dir);
4298         int status;
4299
4300         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4301         if (status < 0)
4302                 return ERR_PTR(status);
4303         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4304 }
4305
4306 static int _nfs4_proc_lookupp(struct inode *inode,
4307                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4308                 struct nfs4_label *label)
4309 {
4310         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4311         struct nfs_server *server = NFS_SERVER(inode);
4312         int                    status;
4313         struct nfs4_lookupp_arg args = {
4314                 .bitmask = server->attr_bitmask,
4315                 .fh = NFS_FH(inode),
4316         };
4317         struct nfs4_lookupp_res res = {
4318                 .server = server,
4319                 .fattr = fattr,
4320                 .label = label,
4321                 .fh = fhandle,
4322         };
4323         struct rpc_message msg = {
4324                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4325                 .rpc_argp = &args,
4326                 .rpc_resp = &res,
4327         };
4328
4329         args.bitmask = nfs4_bitmask(server, label);
4330
4331         nfs_fattr_init(fattr);
4332
4333         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4334         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4335                                 &res.seq_res, 0);
4336         dprintk("NFS reply lookupp: %d\n", status);
4337         return status;
4338 }
4339
4340 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4341                              struct nfs_fattr *fattr, struct nfs4_label *label)
4342 {
4343         struct nfs4_exception exception = {
4344                 .interruptible = true,
4345         };
4346         int err;
4347         do {
4348                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4349                 trace_nfs4_lookupp(inode, err);
4350                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4351                                 &exception);
4352         } while (exception.retry);
4353         return err;
4354 }
4355
4356 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4357 {
4358         struct nfs_server *server = NFS_SERVER(inode);
4359         struct nfs4_accessargs args = {
4360                 .fh = NFS_FH(inode),
4361                 .access = entry->mask,
4362         };
4363         struct nfs4_accessres res = {
4364                 .server = server,
4365         };
4366         struct rpc_message msg = {
4367                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4368                 .rpc_argp = &args,
4369                 .rpc_resp = &res,
4370                 .rpc_cred = entry->cred,
4371         };
4372         int status = 0;
4373
4374         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4375                 res.fattr = nfs_alloc_fattr();
4376                 if (res.fattr == NULL)
4377                         return -ENOMEM;
4378                 args.bitmask = server->cache_consistency_bitmask;
4379         }
4380         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4381         if (!status) {
4382                 nfs_access_set_mask(entry, res.access);
4383                 if (res.fattr)
4384                         nfs_refresh_inode(inode, res.fattr);
4385         }
4386         nfs_free_fattr(res.fattr);
4387         return status;
4388 }
4389
4390 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4391 {
4392         struct nfs4_exception exception = {
4393                 .interruptible = true,
4394         };
4395         int err;
4396         do {
4397                 err = _nfs4_proc_access(inode, entry);
4398                 trace_nfs4_access(inode, err);
4399                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4400                                 &exception);
4401         } while (exception.retry);
4402         return err;
4403 }
4404
4405 /*
4406  * TODO: For the time being, we don't try to get any attributes
4407  * along with any of the zero-copy operations READ, READDIR,
4408  * READLINK, WRITE.
4409  *
4410  * In the case of the first three, we want to put the GETATTR
4411  * after the read-type operation -- this is because it is hard
4412  * to predict the length of a GETATTR response in v4, and thus
4413  * align the READ data correctly.  This means that the GETATTR
4414  * may end up partially falling into the page cache, and we should
4415  * shift it into the 'tail' of the xdr_buf before processing.
4416  * To do this efficiently, we need to know the total length
4417  * of data received, which doesn't seem to be available outside
4418  * of the RPC layer.
4419  *
4420  * In the case of WRITE, we also want to put the GETATTR after
4421  * the operation -- in this case because we want to make sure
4422  * we get the post-operation mtime and size.
4423  *
4424  * Both of these changes to the XDR layer would in fact be quite
4425  * minor, but I decided to leave them for a subsequent patch.
4426  */
4427 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4428                 unsigned int pgbase, unsigned int pglen)
4429 {
4430         struct nfs4_readlink args = {
4431                 .fh       = NFS_FH(inode),
4432                 .pgbase   = pgbase,
4433                 .pglen    = pglen,
4434                 .pages    = &page,
4435         };
4436         struct nfs4_readlink_res res;
4437         struct rpc_message msg = {
4438                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4439                 .rpc_argp = &args,
4440                 .rpc_resp = &res,
4441         };
4442
4443         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4444 }
4445
4446 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4447                 unsigned int pgbase, unsigned int pglen)
4448 {
4449         struct nfs4_exception exception = {
4450                 .interruptible = true,
4451         };
4452         int err;
4453         do {
4454                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4455                 trace_nfs4_readlink(inode, err);
4456                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4457                                 &exception);
4458         } while (exception.retry);
4459         return err;
4460 }
4461
4462 /*
4463  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4464  */
4465 static int
4466 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4467                  int flags)
4468 {
4469         struct nfs_server *server = NFS_SERVER(dir);
4470         struct nfs4_label l, *ilabel = NULL;
4471         struct nfs_open_context *ctx;
4472         struct nfs4_state *state;
4473         int status = 0;
4474
4475         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4476         if (IS_ERR(ctx))
4477                 return PTR_ERR(ctx);
4478
4479         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4480
4481         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4482                 sattr->ia_mode &= ~current_umask();
4483         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4484         if (IS_ERR(state)) {
4485                 status = PTR_ERR(state);
4486                 goto out;
4487         }
4488 out:
4489         nfs4_label_release_security(ilabel);
4490         put_nfs_open_context(ctx);
4491         return status;
4492 }
4493
4494 static int
4495 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4496 {
4497         struct nfs_server *server = NFS_SERVER(dir);
4498         struct nfs_removeargs args = {
4499                 .fh = NFS_FH(dir),
4500                 .name = *name,
4501         };
4502         struct nfs_removeres res = {
4503                 .server = server,
4504         };
4505         struct rpc_message msg = {
4506                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4507                 .rpc_argp = &args,
4508                 .rpc_resp = &res,
4509         };
4510         unsigned long timestamp = jiffies;
4511         int status;
4512
4513         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4514         if (status == 0) {
4515                 spin_lock(&dir->i_lock);
4516                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4517                 /* Removing a directory decrements nlink in the parent */
4518                 if (ftype == NF4DIR && dir->i_nlink > 2)
4519                         nfs4_dec_nlink_locked(dir);
4520                 spin_unlock(&dir->i_lock);
4521         }
4522         return status;
4523 }
4524
4525 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4526 {
4527         struct nfs4_exception exception = {
4528                 .interruptible = true,
4529         };
4530         struct inode *inode = d_inode(dentry);
4531         int err;
4532
4533         if (inode) {
4534                 if (inode->i_nlink == 1)
4535                         nfs4_inode_return_delegation(inode);
4536                 else
4537                         nfs4_inode_make_writeable(inode);
4538         }
4539         do {
4540                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4541                 trace_nfs4_remove(dir, &dentry->d_name, err);
4542                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4543                                 &exception);
4544         } while (exception.retry);
4545         return err;
4546 }
4547
4548 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4549 {
4550         struct nfs4_exception exception = {
4551                 .interruptible = true,
4552         };
4553         int err;
4554
4555         do {
4556                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4557                 trace_nfs4_remove(dir, name, err);
4558                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4559                                 &exception);
4560         } while (exception.retry);
4561         return err;
4562 }
4563
4564 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4565                 struct dentry *dentry,
4566                 struct inode *inode)
4567 {
4568         struct nfs_removeargs *args = msg->rpc_argp;
4569         struct nfs_removeres *res = msg->rpc_resp;
4570
4571         res->server = NFS_SB(dentry->d_sb);
4572         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4573         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4574
4575         nfs_fattr_init(res->dir_attr);
4576
4577         if (inode)
4578                 nfs4_inode_return_delegation(inode);
4579 }
4580
4581 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4582 {
4583         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4584                         &data->args.seq_args,
4585                         &data->res.seq_res,
4586                         task);
4587 }
4588
4589 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4590 {
4591         struct nfs_unlinkdata *data = task->tk_calldata;
4592         struct nfs_removeres *res = &data->res;
4593
4594         if (!nfs4_sequence_done(task, &res->seq_res))
4595                 return 0;
4596         if (nfs4_async_handle_error(task, res->server, NULL,
4597                                     &data->timeout) == -EAGAIN)
4598                 return 0;
4599         if (task->tk_status == 0)
4600                 update_changeattr(dir, &res->cinfo,
4601                                 res->dir_attr->time_start, 0);
4602         return 1;
4603 }
4604
4605 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4606                 struct dentry *old_dentry,
4607                 struct dentry *new_dentry)
4608 {
4609         struct nfs_renameargs *arg = msg->rpc_argp;
4610         struct nfs_renameres *res = msg->rpc_resp;
4611         struct inode *old_inode = d_inode(old_dentry);
4612         struct inode *new_inode = d_inode(new_dentry);
4613
4614         if (old_inode)
4615                 nfs4_inode_make_writeable(old_inode);
4616         if (new_inode)
4617                 nfs4_inode_return_delegation(new_inode);
4618         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4619         res->server = NFS_SB(old_dentry->d_sb);
4620         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4621 }
4622
4623 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4624 {
4625         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4626                         &data->args.seq_args,
4627                         &data->res.seq_res,
4628                         task);
4629 }
4630
4631 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4632                                  struct inode *new_dir)
4633 {
4634         struct nfs_renamedata *data = task->tk_calldata;
4635         struct nfs_renameres *res = &data->res;
4636
4637         if (!nfs4_sequence_done(task, &res->seq_res))
4638                 return 0;
4639         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4640                 return 0;
4641
4642         if (task->tk_status == 0) {
4643                 if (new_dir != old_dir) {
4644                         /* Note: If we moved a directory, nlink will change */
4645                         update_changeattr(old_dir, &res->old_cinfo,
4646                                         res->old_fattr->time_start,
4647                                         NFS_INO_INVALID_OTHER);
4648                         update_changeattr(new_dir, &res->new_cinfo,
4649                                         res->new_fattr->time_start,
4650                                         NFS_INO_INVALID_OTHER);
4651                 } else
4652                         update_changeattr(old_dir, &res->old_cinfo,
4653                                         res->old_fattr->time_start,
4654                                         0);
4655         }
4656         return 1;
4657 }
4658
4659 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4660 {
4661         struct nfs_server *server = NFS_SERVER(inode);
4662         __u32 bitmask[NFS4_BITMASK_SZ];
4663         struct nfs4_link_arg arg = {
4664                 .fh     = NFS_FH(inode),
4665                 .dir_fh = NFS_FH(dir),
4666                 .name   = name,
4667                 .bitmask = bitmask,
4668         };
4669         struct nfs4_link_res res = {
4670                 .server = server,
4671                 .label = NULL,
4672         };
4673         struct rpc_message msg = {
4674                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4675                 .rpc_argp = &arg,
4676                 .rpc_resp = &res,
4677         };
4678         int status = -ENOMEM;
4679
4680         res.fattr = nfs_alloc_fattr();
4681         if (res.fattr == NULL)
4682                 goto out;
4683
4684         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4685         if (IS_ERR(res.label)) {
4686                 status = PTR_ERR(res.label);
4687                 goto out;
4688         }
4689
4690         nfs4_inode_make_writeable(inode);
4691         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4692
4693         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4694         if (!status) {
4695                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4696                 status = nfs_post_op_update_inode(inode, res.fattr);
4697                 if (!status)
4698                         nfs_setsecurity(inode, res.fattr, res.label);
4699         }
4700
4701
4702         nfs4_label_free(res.label);
4703
4704 out:
4705         nfs_free_fattr(res.fattr);
4706         return status;
4707 }
4708
4709 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4710 {
4711         struct nfs4_exception exception = {
4712                 .interruptible = true,
4713         };
4714         int err;
4715         do {
4716                 err = nfs4_handle_exception(NFS_SERVER(inode),
4717                                 _nfs4_proc_link(inode, dir, name),
4718                                 &exception);
4719         } while (exception.retry);
4720         return err;
4721 }
4722
4723 struct nfs4_createdata {
4724         struct rpc_message msg;
4725         struct nfs4_create_arg arg;
4726         struct nfs4_create_res res;
4727         struct nfs_fh fh;
4728         struct nfs_fattr fattr;
4729         struct nfs4_label *label;
4730 };
4731
4732 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4733                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4734 {
4735         struct nfs4_createdata *data;
4736
4737         data = kzalloc(sizeof(*data), GFP_KERNEL);
4738         if (data != NULL) {
4739                 struct nfs_server *server = NFS_SERVER(dir);
4740
4741                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4742                 if (IS_ERR(data->label))
4743                         goto out_free;
4744
4745                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4746                 data->msg.rpc_argp = &data->arg;
4747                 data->msg.rpc_resp = &data->res;
4748                 data->arg.dir_fh = NFS_FH(dir);
4749                 data->arg.server = server;
4750                 data->arg.name = name;
4751                 data->arg.attrs = sattr;
4752                 data->arg.ftype = ftype;
4753                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4754                 data->arg.umask = current_umask();
4755                 data->res.server = server;
4756                 data->res.fh = &data->fh;
4757                 data->res.fattr = &data->fattr;
4758                 data->res.label = data->label;
4759                 nfs_fattr_init(data->res.fattr);
4760         }
4761         return data;
4762 out_free:
4763         kfree(data);
4764         return NULL;
4765 }
4766
4767 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4768 {
4769         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4770                                     &data->arg.seq_args, &data->res.seq_res, 1);
4771         if (status == 0) {
4772                 spin_lock(&dir->i_lock);
4773                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4774                                 data->res.fattr->time_start, 0);
4775                 /* Creating a directory bumps nlink in the parent */
4776                 if (data->arg.ftype == NF4DIR)
4777                         nfs4_inc_nlink_locked(dir);
4778                 spin_unlock(&dir->i_lock);
4779                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4780         }
4781         return status;
4782 }
4783
4784 static void nfs4_free_createdata(struct nfs4_createdata *data)
4785 {
4786         nfs4_label_free(data->label);
4787         kfree(data);
4788 }
4789
4790 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4791                 struct page *page, unsigned int len, struct iattr *sattr,
4792                 struct nfs4_label *label)
4793 {
4794         struct nfs4_createdata *data;
4795         int status = -ENAMETOOLONG;
4796
4797         if (len > NFS4_MAXPATHLEN)
4798                 goto out;
4799
4800         status = -ENOMEM;
4801         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4802         if (data == NULL)
4803                 goto out;
4804
4805         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4806         data->arg.u.symlink.pages = &page;
4807         data->arg.u.symlink.len = len;
4808         data->arg.label = label;
4809         
4810         status = nfs4_do_create(dir, dentry, data);
4811
4812         nfs4_free_createdata(data);
4813 out:
4814         return status;
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 {
4820         struct nfs4_exception exception = {
4821                 .interruptible = true,
4822         };
4823         struct nfs4_label l, *label = NULL;
4824         int err;
4825
4826         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4827
4828         do {
4829                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4830                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4831                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4832                                 &exception);
4833         } while (exception.retry);
4834
4835         nfs4_label_release_security(label);
4836         return err;
4837 }
4838
4839 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4840                 struct iattr *sattr, struct nfs4_label *label)
4841 {
4842         struct nfs4_createdata *data;
4843         int status = -ENOMEM;
4844
4845         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4846         if (data == NULL)
4847                 goto out;
4848
4849         data->arg.label = label;
4850         status = nfs4_do_create(dir, dentry, data);
4851
4852         nfs4_free_createdata(data);
4853 out:
4854         return status;
4855 }
4856
4857 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4858                 struct iattr *sattr)
4859 {
4860         struct nfs_server *server = NFS_SERVER(dir);
4861         struct nfs4_exception exception = {
4862                 .interruptible = true,
4863         };
4864         struct nfs4_label l, *label = NULL;
4865         int err;
4866
4867         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4868
4869         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4870                 sattr->ia_mode &= ~current_umask();
4871         do {
4872                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4873                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4874                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4875                                 &exception);
4876         } while (exception.retry);
4877         nfs4_label_release_security(label);
4878
4879         return err;
4880 }
4881
4882 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4883                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4884 {
4885         struct inode            *dir = d_inode(dentry);
4886         struct nfs4_readdir_arg args = {
4887                 .fh = NFS_FH(dir),
4888                 .pages = pages,
4889                 .pgbase = 0,
4890                 .count = count,
4891                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4892                 .plus = plus,
4893         };
4894         struct nfs4_readdir_res res;
4895         struct rpc_message msg = {
4896                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4897                 .rpc_argp = &args,
4898                 .rpc_resp = &res,
4899                 .rpc_cred = cred,
4900         };
4901         int                     status;
4902
4903         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4904                         dentry,
4905                         (unsigned long long)cookie);
4906         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4907         res.pgbase = args.pgbase;
4908         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4909         if (status >= 0) {
4910                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4911                 status += args.pgbase;
4912         }
4913
4914         nfs_invalidate_atime(dir);
4915
4916         dprintk("%s: returns %d\n", __func__, status);
4917         return status;
4918 }
4919
4920 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4921                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4922 {
4923         struct nfs4_exception exception = {
4924                 .interruptible = true,
4925         };
4926         int err;
4927         do {
4928                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4929                                 pages, count, plus);
4930                 trace_nfs4_readdir(d_inode(dentry), err);
4931                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4932                                 &exception);
4933         } while (exception.retry);
4934         return err;
4935 }
4936
4937 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4938                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4939 {
4940         struct nfs4_createdata *data;
4941         int mode = sattr->ia_mode;
4942         int status = -ENOMEM;
4943
4944         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4945         if (data == NULL)
4946                 goto out;
4947
4948         if (S_ISFIFO(mode))
4949                 data->arg.ftype = NF4FIFO;
4950         else if (S_ISBLK(mode)) {
4951                 data->arg.ftype = NF4BLK;
4952                 data->arg.u.device.specdata1 = MAJOR(rdev);
4953                 data->arg.u.device.specdata2 = MINOR(rdev);
4954         }
4955         else if (S_ISCHR(mode)) {
4956                 data->arg.ftype = NF4CHR;
4957                 data->arg.u.device.specdata1 = MAJOR(rdev);
4958                 data->arg.u.device.specdata2 = MINOR(rdev);
4959         } else if (!S_ISSOCK(mode)) {
4960                 status = -EINVAL;
4961                 goto out_free;
4962         }
4963
4964         data->arg.label = label;
4965         status = nfs4_do_create(dir, dentry, data);
4966 out_free:
4967         nfs4_free_createdata(data);
4968 out:
4969         return status;
4970 }
4971
4972 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4973                 struct iattr *sattr, dev_t rdev)
4974 {
4975         struct nfs_server *server = NFS_SERVER(dir);
4976         struct nfs4_exception exception = {
4977                 .interruptible = true,
4978         };
4979         struct nfs4_label l, *label = NULL;
4980         int err;
4981
4982         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4983
4984         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4985                 sattr->ia_mode &= ~current_umask();
4986         do {
4987                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4988                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4989                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4990                                 &exception);
4991         } while (exception.retry);
4992
4993         nfs4_label_release_security(label);
4994
4995         return err;
4996 }
4997
4998 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4999                  struct nfs_fsstat *fsstat)
5000 {
5001         struct nfs4_statfs_arg args = {
5002                 .fh = fhandle,
5003                 .bitmask = server->attr_bitmask,
5004         };
5005         struct nfs4_statfs_res res = {
5006                 .fsstat = fsstat,
5007         };
5008         struct rpc_message msg = {
5009                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5010                 .rpc_argp = &args,
5011                 .rpc_resp = &res,
5012         };
5013
5014         nfs_fattr_init(fsstat->fattr);
5015         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5016 }
5017
5018 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5019 {
5020         struct nfs4_exception exception = {
5021                 .interruptible = true,
5022         };
5023         int err;
5024         do {
5025                 err = nfs4_handle_exception(server,
5026                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5027                                 &exception);
5028         } while (exception.retry);
5029         return err;
5030 }
5031
5032 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5033                 struct nfs_fsinfo *fsinfo)
5034 {
5035         struct nfs4_fsinfo_arg args = {
5036                 .fh = fhandle,
5037                 .bitmask = server->attr_bitmask,
5038         };
5039         struct nfs4_fsinfo_res res = {
5040                 .fsinfo = fsinfo,
5041         };
5042         struct rpc_message msg = {
5043                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5044                 .rpc_argp = &args,
5045                 .rpc_resp = &res,
5046         };
5047
5048         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5049 }
5050
5051 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5052 {
5053         struct nfs4_exception exception = {
5054                 .interruptible = true,
5055         };
5056         int err;
5057
5058         do {
5059                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5060                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5061                 if (err == 0) {
5062                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5063                         break;
5064                 }
5065                 err = nfs4_handle_exception(server, err, &exception);
5066         } while (exception.retry);
5067         return err;
5068 }
5069
5070 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5071 {
5072         int error;
5073
5074         nfs_fattr_init(fsinfo->fattr);
5075         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5076         if (error == 0) {
5077                 /* block layout checks this! */
5078                 server->pnfs_blksize = fsinfo->blksize;
5079                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5080         }
5081
5082         return error;
5083 }
5084
5085 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5086                 struct nfs_pathconf *pathconf)
5087 {
5088         struct nfs4_pathconf_arg args = {
5089                 .fh = fhandle,
5090                 .bitmask = server->attr_bitmask,
5091         };
5092         struct nfs4_pathconf_res res = {
5093                 .pathconf = pathconf,
5094         };
5095         struct rpc_message msg = {
5096                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5097                 .rpc_argp = &args,
5098                 .rpc_resp = &res,
5099         };
5100
5101         /* None of the pathconf attributes are mandatory to implement */
5102         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5103                 memset(pathconf, 0, sizeof(*pathconf));
5104                 return 0;
5105         }
5106
5107         nfs_fattr_init(pathconf->fattr);
5108         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5109 }
5110
5111 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5112                 struct nfs_pathconf *pathconf)
5113 {
5114         struct nfs4_exception exception = {
5115                 .interruptible = true,
5116         };
5117         int err;
5118
5119         do {
5120                 err = nfs4_handle_exception(server,
5121                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5122                                 &exception);
5123         } while (exception.retry);
5124         return err;
5125 }
5126
5127 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5128                 const struct nfs_open_context *ctx,
5129                 const struct nfs_lock_context *l_ctx,
5130                 fmode_t fmode)
5131 {
5132         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5133 }
5134 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5135
5136 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5137                 const struct nfs_open_context *ctx,
5138                 const struct nfs_lock_context *l_ctx,
5139                 fmode_t fmode)
5140 {
5141         nfs4_stateid _current_stateid;
5142
5143         /* If the current stateid represents a lost lock, then exit */
5144         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5145                 return true;
5146         return nfs4_stateid_match(stateid, &_current_stateid);
5147 }
5148
5149 static bool nfs4_error_stateid_expired(int err)
5150 {
5151         switch (err) {
5152         case -NFS4ERR_DELEG_REVOKED:
5153         case -NFS4ERR_ADMIN_REVOKED:
5154         case -NFS4ERR_BAD_STATEID:
5155         case -NFS4ERR_STALE_STATEID:
5156         case -NFS4ERR_OLD_STATEID:
5157         case -NFS4ERR_OPENMODE:
5158         case -NFS4ERR_EXPIRED:
5159                 return true;
5160         }
5161         return false;
5162 }
5163
5164 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5165 {
5166         struct nfs_server *server = NFS_SERVER(hdr->inode);
5167
5168         trace_nfs4_read(hdr, task->tk_status);
5169         if (task->tk_status < 0) {
5170                 struct nfs4_exception exception = {
5171                         .inode = hdr->inode,
5172                         .state = hdr->args.context->state,
5173                         .stateid = &hdr->args.stateid,
5174                 };
5175                 task->tk_status = nfs4_async_handle_exception(task,
5176                                 server, task->tk_status, &exception);
5177                 if (exception.retry) {
5178                         rpc_restart_call_prepare(task);
5179                         return -EAGAIN;
5180                 }
5181         }
5182
5183         if (task->tk_status > 0)
5184                 renew_lease(server, hdr->timestamp);
5185         return 0;
5186 }
5187
5188 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5189                 struct nfs_pgio_args *args)
5190 {
5191
5192         if (!nfs4_error_stateid_expired(task->tk_status) ||
5193                 nfs4_stateid_is_current(&args->stateid,
5194                                 args->context,
5195                                 args->lock_context,
5196                                 FMODE_READ))
5197                 return false;
5198         rpc_restart_call_prepare(task);
5199         return true;
5200 }
5201
5202 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5203 {
5204
5205         dprintk("--> %s\n", __func__);
5206
5207         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5208                 return -EAGAIN;
5209         if (nfs4_read_stateid_changed(task, &hdr->args))
5210                 return -EAGAIN;
5211         if (task->tk_status > 0)
5212                 nfs_invalidate_atime(hdr->inode);
5213         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5214                                     nfs4_read_done_cb(task, hdr);
5215 }
5216
5217 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5218                                  struct rpc_message *msg)
5219 {
5220         hdr->timestamp   = jiffies;
5221         if (!hdr->pgio_done_cb)
5222                 hdr->pgio_done_cb = nfs4_read_done_cb;
5223         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5224         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5225 }
5226
5227 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5228                                       struct nfs_pgio_header *hdr)
5229 {
5230         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5231                         &hdr->args.seq_args,
5232                         &hdr->res.seq_res,
5233                         task))
5234                 return 0;
5235         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5236                                 hdr->args.lock_context,
5237                                 hdr->rw_mode) == -EIO)
5238                 return -EIO;
5239         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5240                 return -EIO;
5241         return 0;
5242 }
5243
5244 static int nfs4_write_done_cb(struct rpc_task *task,
5245                               struct nfs_pgio_header *hdr)
5246 {
5247         struct inode *inode = hdr->inode;
5248
5249         trace_nfs4_write(hdr, task->tk_status);
5250         if (task->tk_status < 0) {
5251                 struct nfs4_exception exception = {
5252                         .inode = hdr->inode,
5253                         .state = hdr->args.context->state,
5254                         .stateid = &hdr->args.stateid,
5255                 };
5256                 task->tk_status = nfs4_async_handle_exception(task,
5257                                 NFS_SERVER(inode), task->tk_status,
5258                                 &exception);
5259                 if (exception.retry) {
5260                         rpc_restart_call_prepare(task);
5261                         return -EAGAIN;
5262                 }
5263         }
5264         if (task->tk_status >= 0) {
5265                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5266                 nfs_writeback_update_inode(hdr);
5267         }
5268         return 0;
5269 }
5270
5271 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5272                 struct nfs_pgio_args *args)
5273 {
5274
5275         if (!nfs4_error_stateid_expired(task->tk_status) ||
5276                 nfs4_stateid_is_current(&args->stateid,
5277                                 args->context,
5278                                 args->lock_context,
5279                                 FMODE_WRITE))
5280                 return false;
5281         rpc_restart_call_prepare(task);
5282         return true;
5283 }
5284
5285 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5286 {
5287         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5288                 return -EAGAIN;
5289         if (nfs4_write_stateid_changed(task, &hdr->args))
5290                 return -EAGAIN;
5291         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5292                 nfs4_write_done_cb(task, hdr);
5293 }
5294
5295 static
5296 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5297 {
5298         /* Don't request attributes for pNFS or O_DIRECT writes */
5299         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5300                 return false;
5301         /* Otherwise, request attributes if and only if we don't hold
5302          * a delegation
5303          */
5304         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5305 }
5306
5307 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5308                                   struct rpc_message *msg,
5309                                   struct rpc_clnt **clnt)
5310 {
5311         struct nfs_server *server = NFS_SERVER(hdr->inode);
5312
5313         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5314                 hdr->args.bitmask = NULL;
5315                 hdr->res.fattr = NULL;
5316         } else
5317                 hdr->args.bitmask = server->cache_consistency_bitmask;
5318
5319         if (!hdr->pgio_done_cb)
5320                 hdr->pgio_done_cb = nfs4_write_done_cb;
5321         hdr->res.server = server;
5322         hdr->timestamp   = jiffies;
5323
5324         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5325         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5326         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5327 }
5328
5329 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5330 {
5331         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5332                         &data->args.seq_args,
5333                         &data->res.seq_res,
5334                         task);
5335 }
5336
5337 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5338 {
5339         struct inode *inode = data->inode;
5340
5341         trace_nfs4_commit(data, task->tk_status);
5342         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5343                                     NULL, NULL) == -EAGAIN) {
5344                 rpc_restart_call_prepare(task);
5345                 return -EAGAIN;
5346         }
5347         return 0;
5348 }
5349
5350 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5351 {
5352         if (!nfs4_sequence_done(task, &data->res.seq_res))
5353                 return -EAGAIN;
5354         return data->commit_done_cb(task, data);
5355 }
5356
5357 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5358                                    struct rpc_clnt **clnt)
5359 {
5360         struct nfs_server *server = NFS_SERVER(data->inode);
5361
5362         if (data->commit_done_cb == NULL)
5363                 data->commit_done_cb = nfs4_commit_done_cb;
5364         data->res.server = server;
5365         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5366         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5367         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5368 }
5369
5370 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5371                                 struct nfs_commitres *res)
5372 {
5373         struct inode *dst_inode = file_inode(dst);
5374         struct nfs_server *server = NFS_SERVER(dst_inode);
5375         struct rpc_message msg = {
5376                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5377                 .rpc_argp = args,
5378                 .rpc_resp = res,
5379         };
5380
5381         args->fh = NFS_FH(dst_inode);
5382         return nfs4_call_sync(server->client, server, &msg,
5383                         &args->seq_args, &res->seq_res, 1);
5384 }
5385
5386 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5387 {
5388         struct nfs_commitargs args = {
5389                 .offset = offset,
5390                 .count = count,
5391         };
5392         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5393         struct nfs4_exception exception = { };
5394         int status;
5395
5396         do {
5397                 status = _nfs4_proc_commit(dst, &args, res);
5398                 status = nfs4_handle_exception(dst_server, status, &exception);
5399         } while (exception.retry);
5400
5401         return status;
5402 }
5403
5404 struct nfs4_renewdata {
5405         struct nfs_client       *client;
5406         unsigned long           timestamp;
5407 };
5408
5409 /*
5410  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5411  * standalone procedure for queueing an asynchronous RENEW.
5412  */
5413 static void nfs4_renew_release(void *calldata)
5414 {
5415         struct nfs4_renewdata *data = calldata;
5416         struct nfs_client *clp = data->client;
5417
5418         if (refcount_read(&clp->cl_count) > 1)
5419                 nfs4_schedule_state_renewal(clp);
5420         nfs_put_client(clp);
5421         kfree(data);
5422 }
5423
5424 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5425 {
5426         struct nfs4_renewdata *data = calldata;
5427         struct nfs_client *clp = data->client;
5428         unsigned long timestamp = data->timestamp;
5429
5430         trace_nfs4_renew_async(clp, task->tk_status);
5431         switch (task->tk_status) {
5432         case 0:
5433                 break;
5434         case -NFS4ERR_LEASE_MOVED:
5435                 nfs4_schedule_lease_moved_recovery(clp);
5436                 break;
5437         default:
5438                 /* Unless we're shutting down, schedule state recovery! */
5439                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5440                         return;
5441                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5442                         nfs4_schedule_lease_recovery(clp);
5443                         return;
5444                 }
5445                 nfs4_schedule_path_down_recovery(clp);
5446         }
5447         do_renew_lease(clp, timestamp);
5448 }
5449
5450 static const struct rpc_call_ops nfs4_renew_ops = {
5451         .rpc_call_done = nfs4_renew_done,
5452         .rpc_release = nfs4_renew_release,
5453 };
5454
5455 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5456 {
5457         struct rpc_message msg = {
5458                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5459                 .rpc_argp       = clp,
5460                 .rpc_cred       = cred,
5461         };
5462         struct nfs4_renewdata *data;
5463
5464         if (renew_flags == 0)
5465                 return 0;
5466         if (!refcount_inc_not_zero(&clp->cl_count))
5467                 return -EIO;
5468         data = kmalloc(sizeof(*data), GFP_NOFS);
5469         if (data == NULL) {
5470                 nfs_put_client(clp);
5471                 return -ENOMEM;
5472         }
5473         data->client = clp;
5474         data->timestamp = jiffies;
5475         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5476                         &nfs4_renew_ops, data);
5477 }
5478
5479 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5480 {
5481         struct rpc_message msg = {
5482                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5483                 .rpc_argp       = clp,
5484                 .rpc_cred       = cred,
5485         };
5486         unsigned long now = jiffies;
5487         int status;
5488
5489         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5490         if (status < 0)
5491                 return status;
5492         do_renew_lease(clp, now);
5493         return 0;
5494 }
5495
5496 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5497 {
5498         return server->caps & NFS_CAP_ACLS;
5499 }
5500
5501 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5502  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5503  * the stack.
5504  */
5505 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5506
5507 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5508                 struct page **pages)
5509 {
5510         struct page *newpage, **spages;
5511         int rc = 0;
5512         size_t len;
5513         spages = pages;
5514
5515         do {
5516                 len = min_t(size_t, PAGE_SIZE, buflen);
5517                 newpage = alloc_page(GFP_KERNEL);
5518
5519                 if (newpage == NULL)
5520                         goto unwind;
5521                 memcpy(page_address(newpage), buf, len);
5522                 buf += len;
5523                 buflen -= len;
5524                 *pages++ = newpage;
5525                 rc++;
5526         } while (buflen != 0);
5527
5528         return rc;
5529
5530 unwind:
5531         for(; rc > 0; rc--)
5532                 __free_page(spages[rc-1]);
5533         return -ENOMEM;
5534 }
5535
5536 struct nfs4_cached_acl {
5537         int cached;
5538         size_t len;
5539         char data[0];
5540 };
5541
5542 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5543 {
5544         struct nfs_inode *nfsi = NFS_I(inode);
5545
5546         spin_lock(&inode->i_lock);
5547         kfree(nfsi->nfs4_acl);
5548         nfsi->nfs4_acl = acl;
5549         spin_unlock(&inode->i_lock);
5550 }
5551
5552 static void nfs4_zap_acl_attr(struct inode *inode)
5553 {
5554         nfs4_set_cached_acl(inode, NULL);
5555 }
5556
5557 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5558 {
5559         struct nfs_inode *nfsi = NFS_I(inode);
5560         struct nfs4_cached_acl *acl;
5561         int ret = -ENOENT;
5562
5563         spin_lock(&inode->i_lock);
5564         acl = nfsi->nfs4_acl;
5565         if (acl == NULL)
5566                 goto out;
5567         if (buf == NULL) /* user is just asking for length */
5568                 goto out_len;
5569         if (acl->cached == 0)
5570                 goto out;
5571         ret = -ERANGE; /* see getxattr(2) man page */
5572         if (acl->len > buflen)
5573                 goto out;
5574         memcpy(buf, acl->data, acl->len);
5575 out_len:
5576         ret = acl->len;
5577 out:
5578         spin_unlock(&inode->i_lock);
5579         return ret;
5580 }
5581
5582 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5583 {
5584         struct nfs4_cached_acl *acl;
5585         size_t buflen = sizeof(*acl) + acl_len;
5586
5587         if (buflen <= PAGE_SIZE) {
5588                 acl = kmalloc(buflen, GFP_KERNEL);
5589                 if (acl == NULL)
5590                         goto out;
5591                 acl->cached = 1;
5592                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5593         } else {
5594                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5595                 if (acl == NULL)
5596                         goto out;
5597                 acl->cached = 0;
5598         }
5599         acl->len = acl_len;
5600 out:
5601         nfs4_set_cached_acl(inode, acl);
5602 }
5603
5604 /*
5605  * The getxattr API returns the required buffer length when called with a
5606  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5607  * the required buf.  On a NULL buf, we send a page of data to the server
5608  * guessing that the ACL request can be serviced by a page. If so, we cache
5609  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5610  * the cache. If not so, we throw away the page, and cache the required
5611  * length. The next getxattr call will then produce another round trip to
5612  * the server, this time with the input buf of the required size.
5613  */
5614 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5615 {
5616         struct page **pages;
5617         struct nfs_getaclargs args = {
5618                 .fh = NFS_FH(inode),
5619                 .acl_len = buflen,
5620         };
5621         struct nfs_getaclres res = {
5622                 .acl_len = buflen,
5623         };
5624         struct rpc_message msg = {
5625                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5626                 .rpc_argp = &args,
5627                 .rpc_resp = &res,
5628         };
5629         unsigned int npages;
5630         int ret = -ENOMEM, i;
5631         struct nfs_server *server = NFS_SERVER(inode);
5632
5633         if (buflen == 0)
5634                 buflen = server->rsize;
5635
5636         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5637         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5638         if (!pages)
5639                 return -ENOMEM;
5640
5641         args.acl_pages = pages;
5642
5643         for (i = 0; i < npages; i++) {
5644                 pages[i] = alloc_page(GFP_KERNEL);
5645                 if (!pages[i])
5646                         goto out_free;
5647         }
5648
5649         /* for decoding across pages */
5650         res.acl_scratch = alloc_page(GFP_KERNEL);
5651         if (!res.acl_scratch)
5652                 goto out_free;
5653
5654         args.acl_len = npages * PAGE_SIZE;
5655
5656         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5657                 __func__, buf, buflen, npages, args.acl_len);
5658         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5659                              &msg, &args.seq_args, &res.seq_res, 0);
5660         if (ret)
5661                 goto out_free;
5662
5663         /* Handle the case where the passed-in buffer is too short */
5664         if (res.acl_flags & NFS4_ACL_TRUNC) {
5665                 /* Did the user only issue a request for the acl length? */
5666                 if (buf == NULL)
5667                         goto out_ok;
5668                 ret = -ERANGE;
5669                 goto out_free;
5670         }
5671         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5672         if (buf) {
5673                 if (res.acl_len > buflen) {
5674                         ret = -ERANGE;
5675                         goto out_free;
5676                 }
5677                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5678         }
5679 out_ok:
5680         ret = res.acl_len;
5681 out_free:
5682         for (i = 0; i < npages; i++)
5683                 if (pages[i])
5684                         __free_page(pages[i]);
5685         if (res.acl_scratch)
5686                 __free_page(res.acl_scratch);
5687         kfree(pages);
5688         return ret;
5689 }
5690
5691 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5692 {
5693         struct nfs4_exception exception = {
5694                 .interruptible = true,
5695         };
5696         ssize_t ret;
5697         do {
5698                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5699                 trace_nfs4_get_acl(inode, ret);
5700                 if (ret >= 0)
5701                         break;
5702                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5703         } while (exception.retry);
5704         return ret;
5705 }
5706
5707 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5708 {
5709         struct nfs_server *server = NFS_SERVER(inode);
5710         int ret;
5711
5712         if (!nfs4_server_supports_acls(server))
5713                 return -EOPNOTSUPP;
5714         ret = nfs_revalidate_inode(server, inode);
5715         if (ret < 0)
5716                 return ret;
5717         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5718                 nfs_zap_acl_cache(inode);
5719         ret = nfs4_read_cached_acl(inode, buf, buflen);
5720         if (ret != -ENOENT)
5721                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5722                  * but no cached acl data, just the acl length */
5723                 return ret;
5724         return nfs4_get_acl_uncached(inode, buf, buflen);
5725 }
5726
5727 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5728 {
5729         struct nfs_server *server = NFS_SERVER(inode);
5730         struct page *pages[NFS4ACL_MAXPAGES];
5731         struct nfs_setaclargs arg = {
5732                 .fh             = NFS_FH(inode),
5733                 .acl_pages      = pages,
5734                 .acl_len        = buflen,
5735         };
5736         struct nfs_setaclres res;
5737         struct rpc_message msg = {
5738                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5739                 .rpc_argp       = &arg,
5740                 .rpc_resp       = &res,
5741         };
5742         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5743         int ret, i;
5744
5745         if (!nfs4_server_supports_acls(server))
5746                 return -EOPNOTSUPP;
5747         if (npages > ARRAY_SIZE(pages))
5748                 return -ERANGE;
5749         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5750         if (i < 0)
5751                 return i;
5752         nfs4_inode_make_writeable(inode);
5753         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5754
5755         /*
5756          * Free each page after tx, so the only ref left is
5757          * held by the network stack
5758          */
5759         for (; i > 0; i--)
5760                 put_page(pages[i-1]);
5761
5762         /*
5763          * Acl update can result in inode attribute update.
5764          * so mark the attribute cache invalid.
5765          */
5766         spin_lock(&inode->i_lock);
5767         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5768                 | NFS_INO_INVALID_CTIME
5769                 | NFS_INO_REVAL_FORCED;
5770         spin_unlock(&inode->i_lock);
5771         nfs_access_zap_cache(inode);
5772         nfs_zap_acl_cache(inode);
5773         return ret;
5774 }
5775
5776 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5777 {
5778         struct nfs4_exception exception = { };
5779         int err;
5780         do {
5781                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5782                 trace_nfs4_set_acl(inode, err);
5783                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5784                                 &exception);
5785         } while (exception.retry);
5786         return err;
5787 }
5788
5789 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5790 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5791                                         size_t buflen)
5792 {
5793         struct nfs_server *server = NFS_SERVER(inode);
5794         struct nfs_fattr fattr;
5795         struct nfs4_label label = {0, 0, buflen, buf};
5796
5797         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5798         struct nfs4_getattr_arg arg = {
5799                 .fh             = NFS_FH(inode),
5800                 .bitmask        = bitmask,
5801         };
5802         struct nfs4_getattr_res res = {
5803                 .fattr          = &fattr,
5804                 .label          = &label,
5805                 .server         = server,
5806         };
5807         struct rpc_message msg = {
5808                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5809                 .rpc_argp       = &arg,
5810                 .rpc_resp       = &res,
5811         };
5812         int ret;
5813
5814         nfs_fattr_init(&fattr);
5815
5816         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5817         if (ret)
5818                 return ret;
5819         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5820                 return -ENOENT;
5821         if (buflen < label.len)
5822                 return -ERANGE;
5823         return 0;
5824 }
5825
5826 static int nfs4_get_security_label(struct inode *inode, void *buf,
5827                                         size_t buflen)
5828 {
5829         struct nfs4_exception exception = {
5830                 .interruptible = true,
5831         };
5832         int err;
5833
5834         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5835                 return -EOPNOTSUPP;
5836
5837         do {
5838                 err = _nfs4_get_security_label(inode, buf, buflen);
5839                 trace_nfs4_get_security_label(inode, err);
5840                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5841                                 &exception);
5842         } while (exception.retry);
5843         return err;
5844 }
5845
5846 static int _nfs4_do_set_security_label(struct inode *inode,
5847                 struct nfs4_label *ilabel,
5848                 struct nfs_fattr *fattr,
5849                 struct nfs4_label *olabel)
5850 {
5851
5852         struct iattr sattr = {0};
5853         struct nfs_server *server = NFS_SERVER(inode);
5854         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5855         struct nfs_setattrargs arg = {
5856                 .fh             = NFS_FH(inode),
5857                 .iap            = &sattr,
5858                 .server         = server,
5859                 .bitmask        = bitmask,
5860                 .label          = ilabel,
5861         };
5862         struct nfs_setattrres res = {
5863                 .fattr          = fattr,
5864                 .label          = olabel,
5865                 .server         = server,
5866         };
5867         struct rpc_message msg = {
5868                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5869                 .rpc_argp       = &arg,
5870                 .rpc_resp       = &res,
5871         };
5872         int status;
5873
5874         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5875
5876         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5877         if (status)
5878                 dprintk("%s failed: %d\n", __func__, status);
5879
5880         return status;
5881 }
5882
5883 static int nfs4_do_set_security_label(struct inode *inode,
5884                 struct nfs4_label *ilabel,
5885                 struct nfs_fattr *fattr,
5886                 struct nfs4_label *olabel)
5887 {
5888         struct nfs4_exception exception = { };
5889         int err;
5890
5891         do {
5892                 err = _nfs4_do_set_security_label(inode, ilabel,
5893                                 fattr, olabel);
5894                 trace_nfs4_set_security_label(inode, err);
5895                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5896                                 &exception);
5897         } while (exception.retry);
5898         return err;
5899 }
5900
5901 static int
5902 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5903 {
5904         struct nfs4_label ilabel, *olabel = NULL;
5905         struct nfs_fattr fattr;
5906         int status;
5907
5908         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5909                 return -EOPNOTSUPP;
5910
5911         nfs_fattr_init(&fattr);
5912
5913         ilabel.pi = 0;
5914         ilabel.lfs = 0;
5915         ilabel.label = (char *)buf;
5916         ilabel.len = buflen;
5917
5918         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5919         if (IS_ERR(olabel)) {
5920                 status = -PTR_ERR(olabel);
5921                 goto out;
5922         }
5923
5924         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5925         if (status == 0)
5926                 nfs_setsecurity(inode, &fattr, olabel);
5927
5928         nfs4_label_free(olabel);
5929 out:
5930         return status;
5931 }
5932 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5933
5934
5935 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5936                                     nfs4_verifier *bootverf)
5937 {
5938         __be32 verf[2];
5939
5940         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5941                 /* An impossible timestamp guarantees this value
5942                  * will never match a generated boot time. */
5943                 verf[0] = cpu_to_be32(U32_MAX);
5944                 verf[1] = cpu_to_be32(U32_MAX);
5945         } else {
5946                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5947                 u64 ns = ktime_to_ns(nn->boot_time);
5948
5949                 verf[0] = cpu_to_be32(ns >> 32);
5950                 verf[1] = cpu_to_be32(ns);
5951         }
5952         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5953 }
5954
5955 static int
5956 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5957 {
5958         size_t len;
5959         char *str;
5960
5961         if (clp->cl_owner_id != NULL)
5962                 return 0;
5963
5964         rcu_read_lock();
5965         len = 14 +
5966                 strlen(clp->cl_rpcclient->cl_nodename) +
5967                 1 +
5968                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5969                 1;
5970         rcu_read_unlock();
5971         if (nfs4_client_id_uniquifier[0] != '\0')
5972                 len += strlen(nfs4_client_id_uniquifier) + 1;
5973         if (len > NFS4_OPAQUE_LIMIT + 1)
5974                 return -EINVAL;
5975
5976         /*
5977          * Since this string is allocated at mount time, and held until the
5978          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5979          * about a memory-reclaim deadlock.
5980          */
5981         str = kmalloc(len, GFP_KERNEL);
5982         if (!str)
5983                 return -ENOMEM;
5984
5985         rcu_read_lock();
5986         if (nfs4_client_id_uniquifier[0] != '\0')
5987                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5988                           clp->cl_rpcclient->cl_nodename,
5989                           nfs4_client_id_uniquifier,
5990                           rpc_peeraddr2str(clp->cl_rpcclient,
5991                                            RPC_DISPLAY_ADDR));
5992         else
5993                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5994                           clp->cl_rpcclient->cl_nodename,
5995                           rpc_peeraddr2str(clp->cl_rpcclient,
5996                                            RPC_DISPLAY_ADDR));
5997         rcu_read_unlock();
5998
5999         clp->cl_owner_id = str;
6000         return 0;
6001 }
6002
6003 static int
6004 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
6005 {
6006         size_t len;
6007         char *str;
6008
6009         len = 10 + 10 + 1 + 10 + 1 +
6010                 strlen(nfs4_client_id_uniquifier) + 1 +
6011                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6012
6013         if (len > NFS4_OPAQUE_LIMIT + 1)
6014                 return -EINVAL;
6015
6016         /*
6017          * Since this string is allocated at mount time, and held until the
6018          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6019          * about a memory-reclaim deadlock.
6020          */
6021         str = kmalloc(len, GFP_KERNEL);
6022         if (!str)
6023                 return -ENOMEM;
6024
6025         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6026                         clp->rpc_ops->version, clp->cl_minorversion,
6027                         nfs4_client_id_uniquifier,
6028                         clp->cl_rpcclient->cl_nodename);
6029         clp->cl_owner_id = str;
6030         return 0;
6031 }
6032
6033 static int
6034 nfs4_init_uniform_client_string(struct nfs_client *clp)
6035 {
6036         size_t len;
6037         char *str;
6038
6039         if (clp->cl_owner_id != NULL)
6040                 return 0;
6041
6042         if (nfs4_client_id_uniquifier[0] != '\0')
6043                 return nfs4_init_uniquifier_client_string(clp);
6044
6045         len = 10 + 10 + 1 + 10 + 1 +
6046                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6047
6048         if (len > NFS4_OPAQUE_LIMIT + 1)
6049                 return -EINVAL;
6050
6051         /*
6052          * Since this string is allocated at mount time, and held until the
6053          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6054          * about a memory-reclaim deadlock.
6055          */
6056         str = kmalloc(len, GFP_KERNEL);
6057         if (!str)
6058                 return -ENOMEM;
6059
6060         scnprintf(str, len, "Linux NFSv%u.%u %s",
6061                         clp->rpc_ops->version, clp->cl_minorversion,
6062                         clp->cl_rpcclient->cl_nodename);
6063         clp->cl_owner_id = str;
6064         return 0;
6065 }
6066
6067 /*
6068  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6069  * services.  Advertise one based on the address family of the
6070  * clientaddr.
6071  */
6072 static unsigned int
6073 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6074 {
6075         if (strchr(clp->cl_ipaddr, ':') != NULL)
6076                 return scnprintf(buf, len, "tcp6");
6077         else
6078                 return scnprintf(buf, len, "tcp");
6079 }
6080
6081 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6082 {
6083         struct nfs4_setclientid *sc = calldata;
6084
6085         if (task->tk_status == 0)
6086                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6087 }
6088
6089 static const struct rpc_call_ops nfs4_setclientid_ops = {
6090         .rpc_call_done = nfs4_setclientid_done,
6091 };
6092
6093 /**
6094  * nfs4_proc_setclientid - Negotiate client ID
6095  * @clp: state data structure
6096  * @program: RPC program for NFSv4 callback service
6097  * @port: IP port number for NFS4 callback service
6098  * @cred: credential to use for this call
6099  * @res: where to place the result
6100  *
6101  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6102  */
6103 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6104                 unsigned short port, const struct cred *cred,
6105                 struct nfs4_setclientid_res *res)
6106 {
6107         nfs4_verifier sc_verifier;
6108         struct nfs4_setclientid setclientid = {
6109                 .sc_verifier = &sc_verifier,
6110                 .sc_prog = program,
6111                 .sc_clnt = clp,
6112         };
6113         struct rpc_message msg = {
6114                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6115                 .rpc_argp = &setclientid,
6116                 .rpc_resp = res,
6117                 .rpc_cred = cred,
6118         };
6119         struct rpc_task_setup task_setup_data = {
6120                 .rpc_client = clp->cl_rpcclient,
6121                 .rpc_message = &msg,
6122                 .callback_ops = &nfs4_setclientid_ops,
6123                 .callback_data = &setclientid,
6124                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6125         };
6126         unsigned long now = jiffies;
6127         int status;
6128
6129         /* nfs_client_id4 */
6130         nfs4_init_boot_verifier(clp, &sc_verifier);
6131
6132         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6133                 status = nfs4_init_uniform_client_string(clp);
6134         else
6135                 status = nfs4_init_nonuniform_client_string(clp);
6136
6137         if (status)
6138                 goto out;
6139
6140         /* cb_client4 */
6141         setclientid.sc_netid_len =
6142                                 nfs4_init_callback_netid(clp,
6143                                                 setclientid.sc_netid,
6144                                                 sizeof(setclientid.sc_netid));
6145         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6146                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6147                                 clp->cl_ipaddr, port >> 8, port & 255);
6148
6149         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6150                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6151                 clp->cl_owner_id);
6152
6153         status = nfs4_call_sync_custom(&task_setup_data);
6154         if (setclientid.sc_cred) {
6155                 kfree(clp->cl_acceptor);
6156                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6157                 put_rpccred(setclientid.sc_cred);
6158         }
6159
6160         if (status == 0)
6161                 do_renew_lease(clp, now);
6162 out:
6163         trace_nfs4_setclientid(clp, status);
6164         dprintk("NFS reply setclientid: %d\n", status);
6165         return status;
6166 }
6167
6168 /**
6169  * nfs4_proc_setclientid_confirm - Confirm client ID
6170  * @clp: state data structure
6171  * @arg: result of a previous SETCLIENTID
6172  * @cred: credential to use for this call
6173  *
6174  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6175  */
6176 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6177                 struct nfs4_setclientid_res *arg,
6178                 const struct cred *cred)
6179 {
6180         struct rpc_message msg = {
6181                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6182                 .rpc_argp = arg,
6183                 .rpc_cred = cred,
6184         };
6185         int status;
6186
6187         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6188                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6189                 clp->cl_clientid);
6190         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6191                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6192         trace_nfs4_setclientid_confirm(clp, status);
6193         dprintk("NFS reply setclientid_confirm: %d\n", status);
6194         return status;
6195 }
6196
6197 struct nfs4_delegreturndata {
6198         struct nfs4_delegreturnargs args;
6199         struct nfs4_delegreturnres res;
6200         struct nfs_fh fh;
6201         nfs4_stateid stateid;
6202         unsigned long timestamp;
6203         struct {
6204                 struct nfs4_layoutreturn_args arg;
6205                 struct nfs4_layoutreturn_res res;
6206                 struct nfs4_xdr_opaque_data ld_private;
6207                 u32 roc_barrier;
6208                 bool roc;
6209         } lr;
6210         struct nfs_fattr fattr;
6211         int rpc_status;
6212         struct inode *inode;
6213 };
6214
6215 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6216 {
6217         struct nfs4_delegreturndata *data = calldata;
6218         struct nfs4_exception exception = {
6219                 .inode = data->inode,
6220                 .stateid = &data->stateid,
6221         };
6222
6223         if (!nfs4_sequence_done(task, &data->res.seq_res))
6224                 return;
6225
6226         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6227
6228         /* Handle Layoutreturn errors */
6229         if (pnfs_roc_done(task, data->inode,
6230                                 &data->args.lr_args,
6231                                 &data->res.lr_res,
6232                                 &data->res.lr_ret) == -EAGAIN)
6233                 goto out_restart;
6234
6235         switch (task->tk_status) {
6236         case 0:
6237                 renew_lease(data->res.server, data->timestamp);
6238                 break;
6239         case -NFS4ERR_ADMIN_REVOKED:
6240         case -NFS4ERR_DELEG_REVOKED:
6241         case -NFS4ERR_EXPIRED:
6242                 nfs4_free_revoked_stateid(data->res.server,
6243                                 data->args.stateid,
6244                                 task->tk_msg.rpc_cred);
6245                 /* Fallthrough */
6246         case -NFS4ERR_BAD_STATEID:
6247         case -NFS4ERR_STALE_STATEID:
6248                 task->tk_status = 0;
6249                 break;
6250         case -NFS4ERR_OLD_STATEID:
6251                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6252                         nfs4_stateid_seqid_inc(&data->stateid);
6253                 if (data->args.bitmask) {
6254                         data->args.bitmask = NULL;
6255                         data->res.fattr = NULL;
6256                 }
6257                 goto out_restart;
6258         case -NFS4ERR_ACCESS:
6259                 if (data->args.bitmask) {
6260                         data->args.bitmask = NULL;
6261                         data->res.fattr = NULL;
6262                         goto out_restart;
6263                 }
6264                 /* Fallthrough */
6265         default:
6266                 task->tk_status = nfs4_async_handle_exception(task,
6267                                 data->res.server, task->tk_status,
6268                                 &exception);
6269                 if (exception.retry)
6270                         goto out_restart;
6271         }
6272         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6273         data->rpc_status = task->tk_status;
6274         return;
6275 out_restart:
6276         task->tk_status = 0;
6277         rpc_restart_call_prepare(task);
6278 }
6279
6280 static void nfs4_delegreturn_release(void *calldata)
6281 {
6282         struct nfs4_delegreturndata *data = calldata;
6283         struct inode *inode = data->inode;
6284
6285         if (inode) {
6286                 if (data->lr.roc)
6287                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6288                                         data->res.lr_ret);
6289                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6290                 nfs_iput_and_deactive(inode);
6291         }
6292         kfree(calldata);
6293 }
6294
6295 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6296 {
6297         struct nfs4_delegreturndata *d_data;
6298         struct pnfs_layout_hdr *lo;
6299
6300         d_data = (struct nfs4_delegreturndata *)data;
6301
6302         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6303                 nfs4_sequence_done(task, &d_data->res.seq_res);
6304                 return;
6305         }
6306
6307         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6308         if (lo && !pnfs_layout_is_valid(lo)) {
6309                 d_data->args.lr_args = NULL;
6310                 d_data->res.lr_res = NULL;
6311         }
6312
6313         nfs4_setup_sequence(d_data->res.server->nfs_client,
6314                         &d_data->args.seq_args,
6315                         &d_data->res.seq_res,
6316                         task);
6317 }
6318
6319 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6320         .rpc_call_prepare = nfs4_delegreturn_prepare,
6321         .rpc_call_done = nfs4_delegreturn_done,
6322         .rpc_release = nfs4_delegreturn_release,
6323 };
6324
6325 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6326 {
6327         struct nfs4_delegreturndata *data;
6328         struct nfs_server *server = NFS_SERVER(inode);
6329         struct rpc_task *task;
6330         struct rpc_message msg = {
6331                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6332                 .rpc_cred = cred,
6333         };
6334         struct rpc_task_setup task_setup_data = {
6335                 .rpc_client = server->client,
6336                 .rpc_message = &msg,
6337                 .callback_ops = &nfs4_delegreturn_ops,
6338                 .flags = RPC_TASK_ASYNC,
6339         };
6340         int status = 0;
6341
6342         data = kzalloc(sizeof(*data), GFP_NOFS);
6343         if (data == NULL)
6344                 return -ENOMEM;
6345         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6346
6347         nfs4_state_protect(server->nfs_client,
6348                         NFS_SP4_MACH_CRED_CLEANUP,
6349                         &task_setup_data.rpc_client, &msg);
6350
6351         data->args.fhandle = &data->fh;
6352         data->args.stateid = &data->stateid;
6353         data->args.bitmask = server->cache_consistency_bitmask;
6354         nfs_copy_fh(&data->fh, NFS_FH(inode));
6355         nfs4_stateid_copy(&data->stateid, stateid);
6356         data->res.fattr = &data->fattr;
6357         data->res.server = server;
6358         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6359         data->lr.arg.ld_private = &data->lr.ld_private;
6360         nfs_fattr_init(data->res.fattr);
6361         data->timestamp = jiffies;
6362         data->rpc_status = 0;
6363         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6364         data->inode = nfs_igrab_and_active(inode);
6365         if (data->inode) {
6366                 if (data->lr.roc) {
6367                         data->args.lr_args = &data->lr.arg;
6368                         data->res.lr_res = &data->lr.res;
6369                 }
6370         } else if (data->lr.roc) {
6371                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6372                 data->lr.roc = false;
6373         }
6374
6375         task_setup_data.callback_data = data;
6376         msg.rpc_argp = &data->args;
6377         msg.rpc_resp = &data->res;
6378         task = rpc_run_task(&task_setup_data);
6379         if (IS_ERR(task))
6380                 return PTR_ERR(task);
6381         if (!issync)
6382                 goto out;
6383         status = rpc_wait_for_completion_task(task);
6384         if (status != 0)
6385                 goto out;
6386         status = data->rpc_status;
6387 out:
6388         rpc_put_task(task);
6389         return status;
6390 }
6391
6392 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6393 {
6394         struct nfs_server *server = NFS_SERVER(inode);
6395         struct nfs4_exception exception = { };
6396         int err;
6397         do {
6398                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6399                 trace_nfs4_delegreturn(inode, stateid, err);
6400                 switch (err) {
6401                         case -NFS4ERR_STALE_STATEID:
6402                         case -NFS4ERR_EXPIRED:
6403                         case 0:
6404                                 return 0;
6405                 }
6406                 err = nfs4_handle_exception(server, err, &exception);
6407         } while (exception.retry);
6408         return err;
6409 }
6410
6411 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6412 {
6413         struct inode *inode = state->inode;
6414         struct nfs_server *server = NFS_SERVER(inode);
6415         struct nfs_client *clp = server->nfs_client;
6416         struct nfs_lockt_args arg = {
6417                 .fh = NFS_FH(inode),
6418                 .fl = request,
6419         };
6420         struct nfs_lockt_res res = {
6421                 .denied = request,
6422         };
6423         struct rpc_message msg = {
6424                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6425                 .rpc_argp       = &arg,
6426                 .rpc_resp       = &res,
6427                 .rpc_cred       = state->owner->so_cred,
6428         };
6429         struct nfs4_lock_state *lsp;
6430         int status;
6431
6432         arg.lock_owner.clientid = clp->cl_clientid;
6433         status = nfs4_set_lock_state(state, request);
6434         if (status != 0)
6435                 goto out;
6436         lsp = request->fl_u.nfs4_fl.owner;
6437         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6438         arg.lock_owner.s_dev = server->s_dev;
6439         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6440         switch (status) {
6441                 case 0:
6442                         request->fl_type = F_UNLCK;
6443                         break;
6444                 case -NFS4ERR_DENIED:
6445                         status = 0;
6446         }
6447         request->fl_ops->fl_release_private(request);
6448         request->fl_ops = NULL;
6449 out:
6450         return status;
6451 }
6452
6453 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6454 {
6455         struct nfs4_exception exception = {
6456                 .interruptible = true,
6457         };
6458         int err;
6459
6460         do {
6461                 err = _nfs4_proc_getlk(state, cmd, request);
6462                 trace_nfs4_get_lock(request, state, cmd, err);
6463                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6464                                 &exception);
6465         } while (exception.retry);
6466         return err;
6467 }
6468
6469 /*
6470  * Update the seqid of a lock stateid after receiving
6471  * NFS4ERR_OLD_STATEID
6472  */
6473 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6474                 struct nfs4_lock_state *lsp)
6475 {
6476         struct nfs4_state *state = lsp->ls_state;
6477         bool ret = false;
6478
6479         spin_lock(&state->state_lock);
6480         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6481                 goto out;
6482         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6483                 nfs4_stateid_seqid_inc(dst);
6484         else
6485                 dst->seqid = lsp->ls_stateid.seqid;
6486         ret = true;
6487 out:
6488         spin_unlock(&state->state_lock);
6489         return ret;
6490 }
6491
6492 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6493                 struct nfs4_lock_state *lsp)
6494 {
6495         struct nfs4_state *state = lsp->ls_state;
6496         bool ret;
6497
6498         spin_lock(&state->state_lock);
6499         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6500         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6501         spin_unlock(&state->state_lock);
6502         return ret;
6503 }
6504
6505 struct nfs4_unlockdata {
6506         struct nfs_locku_args arg;
6507         struct nfs_locku_res res;
6508         struct nfs4_lock_state *lsp;
6509         struct nfs_open_context *ctx;
6510         struct nfs_lock_context *l_ctx;
6511         struct file_lock fl;
6512         struct nfs_server *server;
6513         unsigned long timestamp;
6514 };
6515
6516 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6517                 struct nfs_open_context *ctx,
6518                 struct nfs4_lock_state *lsp,
6519                 struct nfs_seqid *seqid)
6520 {
6521         struct nfs4_unlockdata *p;
6522         struct nfs4_state *state = lsp->ls_state;
6523         struct inode *inode = state->inode;
6524
6525         p = kzalloc(sizeof(*p), GFP_NOFS);
6526         if (p == NULL)
6527                 return NULL;
6528         p->arg.fh = NFS_FH(inode);
6529         p->arg.fl = &p->fl;
6530         p->arg.seqid = seqid;
6531         p->res.seqid = seqid;
6532         p->lsp = lsp;
6533         /* Ensure we don't close file until we're done freeing locks! */
6534         p->ctx = get_nfs_open_context(ctx);
6535         p->l_ctx = nfs_get_lock_context(ctx);
6536         locks_init_lock(&p->fl);
6537         locks_copy_lock(&p->fl, fl);
6538         p->server = NFS_SERVER(inode);
6539         spin_lock(&state->state_lock);
6540         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6541         spin_unlock(&state->state_lock);
6542         return p;
6543 }
6544
6545 static void nfs4_locku_release_calldata(void *data)
6546 {
6547         struct nfs4_unlockdata *calldata = data;
6548         nfs_free_seqid(calldata->arg.seqid);
6549         nfs4_put_lock_state(calldata->lsp);
6550         nfs_put_lock_context(calldata->l_ctx);
6551         put_nfs_open_context(calldata->ctx);
6552         kfree(calldata);
6553 }
6554
6555 static void nfs4_locku_done(struct rpc_task *task, void *data)
6556 {
6557         struct nfs4_unlockdata *calldata = data;
6558         struct nfs4_exception exception = {
6559                 .inode = calldata->lsp->ls_state->inode,
6560                 .stateid = &calldata->arg.stateid,
6561         };
6562
6563         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6564                 return;
6565         switch (task->tk_status) {
6566                 case 0:
6567                         renew_lease(calldata->server, calldata->timestamp);
6568                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6569                         if (nfs4_update_lock_stateid(calldata->lsp,
6570                                         &calldata->res.stateid))
6571                                 break;
6572                         /* Fall through */
6573                 case -NFS4ERR_ADMIN_REVOKED:
6574                 case -NFS4ERR_EXPIRED:
6575                         nfs4_free_revoked_stateid(calldata->server,
6576                                         &calldata->arg.stateid,
6577                                         task->tk_msg.rpc_cred);
6578                         /* Fall through */
6579                 case -NFS4ERR_BAD_STATEID:
6580                 case -NFS4ERR_STALE_STATEID:
6581                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6582                                                 calldata->lsp))
6583                                 rpc_restart_call_prepare(task);
6584                         break;
6585                 case -NFS4ERR_OLD_STATEID:
6586                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6587                                                 calldata->lsp))
6588                                 rpc_restart_call_prepare(task);
6589                         break;
6590                 default:
6591                         task->tk_status = nfs4_async_handle_exception(task,
6592                                         calldata->server, task->tk_status,
6593                                         &exception);
6594                         if (exception.retry)
6595                                 rpc_restart_call_prepare(task);
6596         }
6597         nfs_release_seqid(calldata->arg.seqid);
6598 }
6599
6600 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6601 {
6602         struct nfs4_unlockdata *calldata = data;
6603
6604         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6605                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6606                 return;
6607
6608         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6609                 goto out_wait;
6610         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6611                 /* Note: exit _without_ running nfs4_locku_done */
6612                 goto out_no_action;
6613         }
6614         calldata->timestamp = jiffies;
6615         if (nfs4_setup_sequence(calldata->server->nfs_client,
6616                                 &calldata->arg.seq_args,
6617                                 &calldata->res.seq_res,
6618                                 task) != 0)
6619                 nfs_release_seqid(calldata->arg.seqid);
6620         return;
6621 out_no_action:
6622         task->tk_action = NULL;
6623 out_wait:
6624         nfs4_sequence_done(task, &calldata->res.seq_res);
6625 }
6626
6627 static const struct rpc_call_ops nfs4_locku_ops = {
6628         .rpc_call_prepare = nfs4_locku_prepare,
6629         .rpc_call_done = nfs4_locku_done,
6630         .rpc_release = nfs4_locku_release_calldata,
6631 };
6632
6633 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6634                 struct nfs_open_context *ctx,
6635                 struct nfs4_lock_state *lsp,
6636                 struct nfs_seqid *seqid)
6637 {
6638         struct nfs4_unlockdata *data;
6639         struct rpc_message msg = {
6640                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6641                 .rpc_cred = ctx->cred,
6642         };
6643         struct rpc_task_setup task_setup_data = {
6644                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6645                 .rpc_message = &msg,
6646                 .callback_ops = &nfs4_locku_ops,
6647                 .workqueue = nfsiod_workqueue,
6648                 .flags = RPC_TASK_ASYNC,
6649         };
6650
6651         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6652                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6653
6654         /* Ensure this is an unlock - when canceling a lock, the
6655          * canceled lock is passed in, and it won't be an unlock.
6656          */
6657         fl->fl_type = F_UNLCK;
6658         if (fl->fl_flags & FL_CLOSE)
6659                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6660
6661         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6662         if (data == NULL) {
6663                 nfs_free_seqid(seqid);
6664                 return ERR_PTR(-ENOMEM);
6665         }
6666
6667         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6668         msg.rpc_argp = &data->arg;
6669         msg.rpc_resp = &data->res;
6670         task_setup_data.callback_data = data;
6671         return rpc_run_task(&task_setup_data);
6672 }
6673
6674 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6675 {
6676         struct inode *inode = state->inode;
6677         struct nfs4_state_owner *sp = state->owner;
6678         struct nfs_inode *nfsi = NFS_I(inode);
6679         struct nfs_seqid *seqid;
6680         struct nfs4_lock_state *lsp;
6681         struct rpc_task *task;
6682         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6683         int status = 0;
6684         unsigned char fl_flags = request->fl_flags;
6685
6686         status = nfs4_set_lock_state(state, request);
6687         /* Unlock _before_ we do the RPC call */
6688         request->fl_flags |= FL_EXISTS;
6689         /* Exclude nfs_delegation_claim_locks() */
6690         mutex_lock(&sp->so_delegreturn_mutex);
6691         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6692         down_read(&nfsi->rwsem);
6693         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6694                 up_read(&nfsi->rwsem);
6695                 mutex_unlock(&sp->so_delegreturn_mutex);
6696                 goto out;
6697         }
6698         up_read(&nfsi->rwsem);
6699         mutex_unlock(&sp->so_delegreturn_mutex);
6700         if (status != 0)
6701                 goto out;
6702         /* Is this a delegated lock? */
6703         lsp = request->fl_u.nfs4_fl.owner;
6704         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6705                 goto out;
6706         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6707         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6708         status = -ENOMEM;
6709         if (IS_ERR(seqid))
6710                 goto out;
6711         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6712         status = PTR_ERR(task);
6713         if (IS_ERR(task))
6714                 goto out;
6715         status = rpc_wait_for_completion_task(task);
6716         rpc_put_task(task);
6717 out:
6718         request->fl_flags = fl_flags;
6719         trace_nfs4_unlock(request, state, F_SETLK, status);
6720         return status;
6721 }
6722
6723 struct nfs4_lockdata {
6724         struct nfs_lock_args arg;
6725         struct nfs_lock_res res;
6726         struct nfs4_lock_state *lsp;
6727         struct nfs_open_context *ctx;
6728         struct file_lock fl;
6729         unsigned long timestamp;
6730         int rpc_status;
6731         int cancelled;
6732         struct nfs_server *server;
6733 };
6734
6735 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6736                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6737                 gfp_t gfp_mask)
6738 {
6739         struct nfs4_lockdata *p;
6740         struct inode *inode = lsp->ls_state->inode;
6741         struct nfs_server *server = NFS_SERVER(inode);
6742         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6743
6744         p = kzalloc(sizeof(*p), gfp_mask);
6745         if (p == NULL)
6746                 return NULL;
6747
6748         p->arg.fh = NFS_FH(inode);
6749         p->arg.fl = &p->fl;
6750         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6751         if (IS_ERR(p->arg.open_seqid))
6752                 goto out_free;
6753         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6754         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6755         if (IS_ERR(p->arg.lock_seqid))
6756                 goto out_free_seqid;
6757         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6758         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6759         p->arg.lock_owner.s_dev = server->s_dev;
6760         p->res.lock_seqid = p->arg.lock_seqid;
6761         p->lsp = lsp;
6762         p->server = server;
6763         p->ctx = get_nfs_open_context(ctx);
6764         locks_init_lock(&p->fl);
6765         locks_copy_lock(&p->fl, fl);
6766         return p;
6767 out_free_seqid:
6768         nfs_free_seqid(p->arg.open_seqid);
6769 out_free:
6770         kfree(p);
6771         return NULL;
6772 }
6773
6774 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6775 {
6776         struct nfs4_lockdata *data = calldata;
6777         struct nfs4_state *state = data->lsp->ls_state;
6778
6779         dprintk("%s: begin!\n", __func__);
6780         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6781                 goto out_wait;
6782         /* Do we need to do an open_to_lock_owner? */
6783         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6784                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6785                         goto out_release_lock_seqid;
6786                 }
6787                 nfs4_stateid_copy(&data->arg.open_stateid,
6788                                 &state->open_stateid);
6789                 data->arg.new_lock_owner = 1;
6790                 data->res.open_seqid = data->arg.open_seqid;
6791         } else {
6792                 data->arg.new_lock_owner = 0;
6793                 nfs4_stateid_copy(&data->arg.lock_stateid,
6794                                 &data->lsp->ls_stateid);
6795         }
6796         if (!nfs4_valid_open_stateid(state)) {
6797                 data->rpc_status = -EBADF;
6798                 task->tk_action = NULL;
6799                 goto out_release_open_seqid;
6800         }
6801         data->timestamp = jiffies;
6802         if (nfs4_setup_sequence(data->server->nfs_client,
6803                                 &data->arg.seq_args,
6804                                 &data->res.seq_res,
6805                                 task) == 0)
6806                 return;
6807 out_release_open_seqid:
6808         nfs_release_seqid(data->arg.open_seqid);
6809 out_release_lock_seqid:
6810         nfs_release_seqid(data->arg.lock_seqid);
6811 out_wait:
6812         nfs4_sequence_done(task, &data->res.seq_res);
6813         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6814 }
6815
6816 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6817 {
6818         struct nfs4_lockdata *data = calldata;
6819         struct nfs4_lock_state *lsp = data->lsp;
6820
6821         dprintk("%s: begin!\n", __func__);
6822
6823         if (!nfs4_sequence_done(task, &data->res.seq_res))
6824                 return;
6825
6826         data->rpc_status = task->tk_status;
6827         switch (task->tk_status) {
6828         case 0:
6829                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6830                                 data->timestamp);
6831                 if (data->arg.new_lock && !data->cancelled) {
6832                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6833                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6834                                 goto out_restart;
6835                 }
6836                 if (data->arg.new_lock_owner != 0) {
6837                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6838                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6839                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6840                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6841                         goto out_restart;
6842                 break;
6843         case -NFS4ERR_BAD_STATEID:
6844         case -NFS4ERR_OLD_STATEID:
6845         case -NFS4ERR_STALE_STATEID:
6846         case -NFS4ERR_EXPIRED:
6847                 if (data->arg.new_lock_owner != 0) {
6848                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6849                                                 &lsp->ls_state->open_stateid))
6850                                 goto out_restart;
6851                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6852                                                 &lsp->ls_stateid))
6853                                 goto out_restart;
6854         }
6855 out_done:
6856         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6857         return;
6858 out_restart:
6859         if (!data->cancelled)
6860                 rpc_restart_call_prepare(task);
6861         goto out_done;
6862 }
6863
6864 static void nfs4_lock_release(void *calldata)
6865 {
6866         struct nfs4_lockdata *data = calldata;
6867
6868         dprintk("%s: begin!\n", __func__);
6869         nfs_free_seqid(data->arg.open_seqid);
6870         if (data->cancelled && data->rpc_status == 0) {
6871                 struct rpc_task *task;
6872                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6873                                 data->arg.lock_seqid);
6874                 if (!IS_ERR(task))
6875                         rpc_put_task_async(task);
6876                 dprintk("%s: cancelling lock!\n", __func__);
6877         } else
6878                 nfs_free_seqid(data->arg.lock_seqid);
6879         nfs4_put_lock_state(data->lsp);
6880         put_nfs_open_context(data->ctx);
6881         kfree(data);
6882         dprintk("%s: done!\n", __func__);
6883 }
6884
6885 static const struct rpc_call_ops nfs4_lock_ops = {
6886         .rpc_call_prepare = nfs4_lock_prepare,
6887         .rpc_call_done = nfs4_lock_done,
6888         .rpc_release = nfs4_lock_release,
6889 };
6890
6891 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6892 {
6893         switch (error) {
6894         case -NFS4ERR_ADMIN_REVOKED:
6895         case -NFS4ERR_EXPIRED:
6896         case -NFS4ERR_BAD_STATEID:
6897                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6898                 if (new_lock_owner != 0 ||
6899                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6900                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6901                 break;
6902         case -NFS4ERR_STALE_STATEID:
6903                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6904                 nfs4_schedule_lease_recovery(server->nfs_client);
6905         }
6906 }
6907
6908 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6909 {
6910         struct nfs4_lockdata *data;
6911         struct rpc_task *task;
6912         struct rpc_message msg = {
6913                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6914                 .rpc_cred = state->owner->so_cred,
6915         };
6916         struct rpc_task_setup task_setup_data = {
6917                 .rpc_client = NFS_CLIENT(state->inode),
6918                 .rpc_message = &msg,
6919                 .callback_ops = &nfs4_lock_ops,
6920                 .workqueue = nfsiod_workqueue,
6921                 .flags = RPC_TASK_ASYNC,
6922         };
6923         int ret;
6924
6925         dprintk("%s: begin!\n", __func__);
6926         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6927                         fl->fl_u.nfs4_fl.owner,
6928                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6929         if (data == NULL)
6930                 return -ENOMEM;
6931         if (IS_SETLKW(cmd))
6932                 data->arg.block = 1;
6933         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6934                                 recovery_type > NFS_LOCK_NEW);
6935         msg.rpc_argp = &data->arg;
6936         msg.rpc_resp = &data->res;
6937         task_setup_data.callback_data = data;
6938         if (recovery_type > NFS_LOCK_NEW) {
6939                 if (recovery_type == NFS_LOCK_RECLAIM)
6940                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6941         } else
6942                 data->arg.new_lock = 1;
6943         task = rpc_run_task(&task_setup_data);
6944         if (IS_ERR(task))
6945                 return PTR_ERR(task);
6946         ret = rpc_wait_for_completion_task(task);
6947         if (ret == 0) {
6948                 ret = data->rpc_status;
6949                 if (ret)
6950                         nfs4_handle_setlk_error(data->server, data->lsp,
6951                                         data->arg.new_lock_owner, ret);
6952         } else
6953                 data->cancelled = true;
6954         rpc_put_task(task);
6955         dprintk("%s: done, ret = %d!\n", __func__, ret);
6956         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6957         return ret;
6958 }
6959
6960 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6961 {
6962         struct nfs_server *server = NFS_SERVER(state->inode);
6963         struct nfs4_exception exception = {
6964                 .inode = state->inode,
6965         };
6966         int err;
6967
6968         do {
6969                 /* Cache the lock if possible... */
6970                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6971                         return 0;
6972                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6973                 if (err != -NFS4ERR_DELAY)
6974                         break;
6975                 nfs4_handle_exception(server, err, &exception);
6976         } while (exception.retry);
6977         return err;
6978 }
6979
6980 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6981 {
6982         struct nfs_server *server = NFS_SERVER(state->inode);
6983         struct nfs4_exception exception = {
6984                 .inode = state->inode,
6985         };
6986         int err;
6987
6988         err = nfs4_set_lock_state(state, request);
6989         if (err != 0)
6990                 return err;
6991         if (!recover_lost_locks) {
6992                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6993                 return 0;
6994         }
6995         do {
6996                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6997                         return 0;
6998                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6999                 switch (err) {
7000                 default:
7001                         goto out;
7002                 case -NFS4ERR_GRACE:
7003                 case -NFS4ERR_DELAY:
7004                         nfs4_handle_exception(server, err, &exception);
7005                         err = 0;
7006                 }
7007         } while (exception.retry);
7008 out:
7009         return err;
7010 }
7011
7012 #if defined(CONFIG_NFS_V4_1)
7013 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7014 {
7015         struct nfs4_lock_state *lsp;
7016         int status;
7017
7018         status = nfs4_set_lock_state(state, request);
7019         if (status != 0)
7020                 return status;
7021         lsp = request->fl_u.nfs4_fl.owner;
7022         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7023             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7024                 return 0;
7025         return nfs4_lock_expired(state, request);
7026 }
7027 #endif
7028
7029 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7030 {
7031         struct nfs_inode *nfsi = NFS_I(state->inode);
7032         struct nfs4_state_owner *sp = state->owner;
7033         unsigned char fl_flags = request->fl_flags;
7034         int status;
7035
7036         request->fl_flags |= FL_ACCESS;
7037         status = locks_lock_inode_wait(state->inode, request);
7038         if (status < 0)
7039                 goto out;
7040         mutex_lock(&sp->so_delegreturn_mutex);
7041         down_read(&nfsi->rwsem);
7042         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7043                 /* Yes: cache locks! */
7044                 /* ...but avoid races with delegation recall... */
7045                 request->fl_flags = fl_flags & ~FL_SLEEP;
7046                 status = locks_lock_inode_wait(state->inode, request);
7047                 up_read(&nfsi->rwsem);
7048                 mutex_unlock(&sp->so_delegreturn_mutex);
7049                 goto out;
7050         }
7051         up_read(&nfsi->rwsem);
7052         mutex_unlock(&sp->so_delegreturn_mutex);
7053         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7054 out:
7055         request->fl_flags = fl_flags;
7056         return status;
7057 }
7058
7059 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7060 {
7061         struct nfs4_exception exception = {
7062                 .state = state,
7063                 .inode = state->inode,
7064                 .interruptible = true,
7065         };
7066         int err;
7067
7068         do {
7069                 err = _nfs4_proc_setlk(state, cmd, request);
7070                 if (err == -NFS4ERR_DENIED)
7071                         err = -EAGAIN;
7072                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7073                                 err, &exception);
7074         } while (exception.retry);
7075         return err;
7076 }
7077
7078 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7079 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7080
7081 static int
7082 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7083                         struct file_lock *request)
7084 {
7085         int             status = -ERESTARTSYS;
7086         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7087
7088         while(!signalled()) {
7089                 status = nfs4_proc_setlk(state, cmd, request);
7090                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7091                         break;
7092                 freezable_schedule_timeout_interruptible(timeout);
7093                 timeout *= 2;
7094                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7095                 status = -ERESTARTSYS;
7096         }
7097         return status;
7098 }
7099
7100 #ifdef CONFIG_NFS_V4_1
7101 struct nfs4_lock_waiter {
7102         struct task_struct      *task;
7103         struct inode            *inode;
7104         struct nfs_lowner       *owner;
7105 };
7106
7107 static int
7108 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7109 {
7110         int ret;
7111         struct nfs4_lock_waiter *waiter = wait->private;
7112
7113         /* NULL key means to wake up everyone */
7114         if (key) {
7115                 struct cb_notify_lock_args      *cbnl = key;
7116                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7117                                                 *wowner = waiter->owner;
7118
7119                 /* Only wake if the callback was for the same owner. */
7120                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7121                         return 0;
7122
7123                 /* Make sure it's for the right inode */
7124                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7125                         return 0;
7126         }
7127
7128         /* override "private" so we can use default_wake_function */
7129         wait->private = waiter->task;
7130         ret = woken_wake_function(wait, mode, flags, key);
7131         if (ret)
7132                 list_del_init(&wait->entry);
7133         wait->private = waiter;
7134         return ret;
7135 }
7136
7137 static int
7138 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7139 {
7140         int status = -ERESTARTSYS;
7141         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7142         struct nfs_server *server = NFS_SERVER(state->inode);
7143         struct nfs_client *clp = server->nfs_client;
7144         wait_queue_head_t *q = &clp->cl_lock_waitq;
7145         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7146                                     .id = lsp->ls_seqid.owner_id,
7147                                     .s_dev = server->s_dev };
7148         struct nfs4_lock_waiter waiter = { .task  = current,
7149                                            .inode = state->inode,
7150                                            .owner = &owner};
7151         wait_queue_entry_t wait;
7152
7153         /* Don't bother with waitqueue if we don't expect a callback */
7154         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7155                 return nfs4_retry_setlk_simple(state, cmd, request);
7156
7157         init_wait(&wait);
7158         wait.private = &waiter;
7159         wait.func = nfs4_wake_lock_waiter;
7160
7161         while(!signalled()) {
7162                 add_wait_queue(q, &wait);
7163                 status = nfs4_proc_setlk(state, cmd, request);
7164                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7165                         finish_wait(q, &wait);
7166                         break;
7167                 }
7168
7169                 status = -ERESTARTSYS;
7170                 freezer_do_not_count();
7171                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7172                 freezer_count();
7173                 finish_wait(q, &wait);
7174         }
7175
7176         return status;
7177 }
7178 #else /* !CONFIG_NFS_V4_1 */
7179 static inline int
7180 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7181 {
7182         return nfs4_retry_setlk_simple(state, cmd, request);
7183 }
7184 #endif
7185
7186 static int
7187 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7188 {
7189         struct nfs_open_context *ctx;
7190         struct nfs4_state *state;
7191         int status;
7192
7193         /* verify open state */
7194         ctx = nfs_file_open_context(filp);
7195         state = ctx->state;
7196
7197         if (IS_GETLK(cmd)) {
7198                 if (state != NULL)
7199                         return nfs4_proc_getlk(state, F_GETLK, request);
7200                 return 0;
7201         }
7202
7203         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7204                 return -EINVAL;
7205
7206         if (request->fl_type == F_UNLCK) {
7207                 if (state != NULL)
7208                         return nfs4_proc_unlck(state, cmd, request);
7209                 return 0;
7210         }
7211
7212         if (state == NULL)
7213                 return -ENOLCK;
7214
7215         if ((request->fl_flags & FL_POSIX) &&
7216             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7217                 return -ENOLCK;
7218
7219         /*
7220          * Don't rely on the VFS having checked the file open mode,
7221          * since it won't do this for flock() locks.
7222          */
7223         switch (request->fl_type) {
7224         case F_RDLCK:
7225                 if (!(filp->f_mode & FMODE_READ))
7226                         return -EBADF;
7227                 break;
7228         case F_WRLCK:
7229                 if (!(filp->f_mode & FMODE_WRITE))
7230                         return -EBADF;
7231         }
7232
7233         status = nfs4_set_lock_state(state, request);
7234         if (status != 0)
7235                 return status;
7236
7237         return nfs4_retry_setlk(state, cmd, request);
7238 }
7239
7240 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7241 {
7242         struct nfs_server *server = NFS_SERVER(state->inode);
7243         int err;
7244
7245         err = nfs4_set_lock_state(state, fl);
7246         if (err != 0)
7247                 return err;
7248         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7249         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7250 }
7251
7252 struct nfs_release_lockowner_data {
7253         struct nfs4_lock_state *lsp;
7254         struct nfs_server *server;
7255         struct nfs_release_lockowner_args args;
7256         struct nfs_release_lockowner_res res;
7257         unsigned long timestamp;
7258 };
7259
7260 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7261 {
7262         struct nfs_release_lockowner_data *data = calldata;
7263         struct nfs_server *server = data->server;
7264         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7265                            &data->res.seq_res, task);
7266         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7267         data->timestamp = jiffies;
7268 }
7269
7270 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7271 {
7272         struct nfs_release_lockowner_data *data = calldata;
7273         struct nfs_server *server = data->server;
7274
7275         nfs40_sequence_done(task, &data->res.seq_res);
7276
7277         switch (task->tk_status) {
7278         case 0:
7279                 renew_lease(server, data->timestamp);
7280                 break;
7281         case -NFS4ERR_STALE_CLIENTID:
7282         case -NFS4ERR_EXPIRED:
7283                 nfs4_schedule_lease_recovery(server->nfs_client);
7284                 break;
7285         case -NFS4ERR_LEASE_MOVED:
7286         case -NFS4ERR_DELAY:
7287                 if (nfs4_async_handle_error(task, server,
7288                                             NULL, NULL) == -EAGAIN)
7289                         rpc_restart_call_prepare(task);
7290         }
7291 }
7292
7293 static void nfs4_release_lockowner_release(void *calldata)
7294 {
7295         struct nfs_release_lockowner_data *data = calldata;
7296         nfs4_free_lock_state(data->server, data->lsp);
7297         kfree(calldata);
7298 }
7299
7300 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7301         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7302         .rpc_call_done = nfs4_release_lockowner_done,
7303         .rpc_release = nfs4_release_lockowner_release,
7304 };
7305
7306 static void
7307 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7308 {
7309         struct nfs_release_lockowner_data *data;
7310         struct rpc_message msg = {
7311                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7312         };
7313
7314         if (server->nfs_client->cl_mvops->minor_version != 0)
7315                 return;
7316
7317         data = kmalloc(sizeof(*data), GFP_NOFS);
7318         if (!data)
7319                 return;
7320         data->lsp = lsp;
7321         data->server = server;
7322         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7323         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7324         data->args.lock_owner.s_dev = server->s_dev;
7325
7326         msg.rpc_argp = &data->args;
7327         msg.rpc_resp = &data->res;
7328         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7329         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7330 }
7331
7332 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7333
7334 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7335                                    struct dentry *unused, struct inode *inode,
7336                                    const char *key, const void *buf,
7337                                    size_t buflen, int flags)
7338 {
7339         return nfs4_proc_set_acl(inode, buf, buflen);
7340 }
7341
7342 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7343                                    struct dentry *unused, struct inode *inode,
7344                                    const char *key, void *buf, size_t buflen)
7345 {
7346         return nfs4_proc_get_acl(inode, buf, buflen);
7347 }
7348
7349 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7350 {
7351         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7352 }
7353
7354 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7355
7356 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7357                                      struct dentry *unused, struct inode *inode,
7358                                      const char *key, const void *buf,
7359                                      size_t buflen, int flags)
7360 {
7361         if (security_ismaclabel(key))
7362                 return nfs4_set_security_label(inode, buf, buflen);
7363
7364         return -EOPNOTSUPP;
7365 }
7366
7367 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7368                                      struct dentry *unused, struct inode *inode,
7369                                      const char *key, void *buf, size_t buflen)
7370 {
7371         if (security_ismaclabel(key))
7372                 return nfs4_get_security_label(inode, buf, buflen);
7373         return -EOPNOTSUPP;
7374 }
7375
7376 static ssize_t
7377 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7378 {
7379         int len = 0;
7380
7381         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7382                 len = security_inode_listsecurity(inode, list, list_len);
7383                 if (list_len && len > list_len)
7384                         return -ERANGE;
7385         }
7386         return len;
7387 }
7388
7389 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7390         .prefix = XATTR_SECURITY_PREFIX,
7391         .get    = nfs4_xattr_get_nfs4_label,
7392         .set    = nfs4_xattr_set_nfs4_label,
7393 };
7394
7395 #else
7396
7397 static ssize_t
7398 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7399 {
7400         return 0;
7401 }
7402
7403 #endif
7404
7405 /*
7406  * nfs_fhget will use either the mounted_on_fileid or the fileid
7407  */
7408 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7409 {
7410         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7411                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7412               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7413               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7414                 return;
7415
7416         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7417                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7418         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7419         fattr->nlink = 2;
7420 }
7421
7422 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7423                                    const struct qstr *name,
7424                                    struct nfs4_fs_locations *fs_locations,
7425                                    struct page *page)
7426 {
7427         struct nfs_server *server = NFS_SERVER(dir);
7428         u32 bitmask[3];
7429         struct nfs4_fs_locations_arg args = {
7430                 .dir_fh = NFS_FH(dir),
7431                 .name = name,
7432                 .page = page,
7433                 .bitmask = bitmask,
7434         };
7435         struct nfs4_fs_locations_res res = {
7436                 .fs_locations = fs_locations,
7437         };
7438         struct rpc_message msg = {
7439                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7440                 .rpc_argp = &args,
7441                 .rpc_resp = &res,
7442         };
7443         int status;
7444
7445         dprintk("%s: start\n", __func__);
7446
7447         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7448         bitmask[1] = nfs4_fattr_bitmap[1];
7449
7450         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7451          * is not supported */
7452         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7453                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7454         else
7455                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7456
7457         nfs_fattr_init(&fs_locations->fattr);
7458         fs_locations->server = server;
7459         fs_locations->nlocations = 0;
7460         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7461         dprintk("%s: returned status = %d\n", __func__, status);
7462         return status;
7463 }
7464
7465 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7466                            const struct qstr *name,
7467                            struct nfs4_fs_locations *fs_locations,
7468                            struct page *page)
7469 {
7470         struct nfs4_exception exception = {
7471                 .interruptible = true,
7472         };
7473         int err;
7474         do {
7475                 err = _nfs4_proc_fs_locations(client, dir, name,
7476                                 fs_locations, page);
7477                 trace_nfs4_get_fs_locations(dir, name, err);
7478                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7479                                 &exception);
7480         } while (exception.retry);
7481         return err;
7482 }
7483
7484 /*
7485  * This operation also signals the server that this client is
7486  * performing migration recovery.  The server can stop returning
7487  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7488  * appended to this compound to identify the client ID which is
7489  * performing recovery.
7490  */
7491 static int _nfs40_proc_get_locations(struct inode *inode,
7492                                      struct nfs4_fs_locations *locations,
7493                                      struct page *page, const struct cred *cred)
7494 {
7495         struct nfs_server *server = NFS_SERVER(inode);
7496         struct rpc_clnt *clnt = server->client;
7497         u32 bitmask[2] = {
7498                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7499         };
7500         struct nfs4_fs_locations_arg args = {
7501                 .clientid       = server->nfs_client->cl_clientid,
7502                 .fh             = NFS_FH(inode),
7503                 .page           = page,
7504                 .bitmask        = bitmask,
7505                 .migration      = 1,            /* skip LOOKUP */
7506                 .renew          = 1,            /* append RENEW */
7507         };
7508         struct nfs4_fs_locations_res res = {
7509                 .fs_locations   = locations,
7510                 .migration      = 1,
7511                 .renew          = 1,
7512         };
7513         struct rpc_message msg = {
7514                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7515                 .rpc_argp       = &args,
7516                 .rpc_resp       = &res,
7517                 .rpc_cred       = cred,
7518         };
7519         unsigned long now = jiffies;
7520         int status;
7521
7522         nfs_fattr_init(&locations->fattr);
7523         locations->server = server;
7524         locations->nlocations = 0;
7525
7526         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7527         status = nfs4_call_sync_sequence(clnt, server, &msg,
7528                                         &args.seq_args, &res.seq_res);
7529         if (status)
7530                 return status;
7531
7532         renew_lease(server, now);
7533         return 0;
7534 }
7535
7536 #ifdef CONFIG_NFS_V4_1
7537
7538 /*
7539  * This operation also signals the server that this client is
7540  * performing migration recovery.  The server can stop asserting
7541  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7542  * performing this operation is identified in the SEQUENCE
7543  * operation in this compound.
7544  *
7545  * When the client supports GETATTR(fs_locations_info), it can
7546  * be plumbed in here.
7547  */
7548 static int _nfs41_proc_get_locations(struct inode *inode,
7549                                      struct nfs4_fs_locations *locations,
7550                                      struct page *page, const struct cred *cred)
7551 {
7552         struct nfs_server *server = NFS_SERVER(inode);
7553         struct rpc_clnt *clnt = server->client;
7554         u32 bitmask[2] = {
7555                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7556         };
7557         struct nfs4_fs_locations_arg args = {
7558                 .fh             = NFS_FH(inode),
7559                 .page           = page,
7560                 .bitmask        = bitmask,
7561                 .migration      = 1,            /* skip LOOKUP */
7562         };
7563         struct nfs4_fs_locations_res res = {
7564                 .fs_locations   = locations,
7565                 .migration      = 1,
7566         };
7567         struct rpc_message msg = {
7568                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7569                 .rpc_argp       = &args,
7570                 .rpc_resp       = &res,
7571                 .rpc_cred       = cred,
7572         };
7573         int status;
7574
7575         nfs_fattr_init(&locations->fattr);
7576         locations->server = server;
7577         locations->nlocations = 0;
7578
7579         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7580         status = nfs4_call_sync_sequence(clnt, server, &msg,
7581                                         &args.seq_args, &res.seq_res);
7582         if (status == NFS4_OK &&
7583             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7584                 status = -NFS4ERR_LEASE_MOVED;
7585         return status;
7586 }
7587
7588 #endif  /* CONFIG_NFS_V4_1 */
7589
7590 /**
7591  * nfs4_proc_get_locations - discover locations for a migrated FSID
7592  * @inode: inode on FSID that is migrating
7593  * @locations: result of query
7594  * @page: buffer
7595  * @cred: credential to use for this operation
7596  *
7597  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7598  * operation failed, or a negative errno if a local error occurred.
7599  *
7600  * On success, "locations" is filled in, but if the server has
7601  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7602  * asserted.
7603  *
7604  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7605  * from this client that require migration recovery.
7606  */
7607 int nfs4_proc_get_locations(struct inode *inode,
7608                             struct nfs4_fs_locations *locations,
7609                             struct page *page, const struct cred *cred)
7610 {
7611         struct nfs_server *server = NFS_SERVER(inode);
7612         struct nfs_client *clp = server->nfs_client;
7613         const struct nfs4_mig_recovery_ops *ops =
7614                                         clp->cl_mvops->mig_recovery_ops;
7615         struct nfs4_exception exception = {
7616                 .interruptible = true,
7617         };
7618         int status;
7619
7620         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7621                 (unsigned long long)server->fsid.major,
7622                 (unsigned long long)server->fsid.minor,
7623                 clp->cl_hostname);
7624         nfs_display_fhandle(NFS_FH(inode), __func__);
7625
7626         do {
7627                 status = ops->get_locations(inode, locations, page, cred);
7628                 if (status != -NFS4ERR_DELAY)
7629                         break;
7630                 nfs4_handle_exception(server, status, &exception);
7631         } while (exception.retry);
7632         return status;
7633 }
7634
7635 /*
7636  * This operation also signals the server that this client is
7637  * performing "lease moved" recovery.  The server can stop
7638  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7639  * is appended to this compound to identify the client ID which is
7640  * performing recovery.
7641  */
7642 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7643 {
7644         struct nfs_server *server = NFS_SERVER(inode);
7645         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7646         struct rpc_clnt *clnt = server->client;
7647         struct nfs4_fsid_present_arg args = {
7648                 .fh             = NFS_FH(inode),
7649                 .clientid       = clp->cl_clientid,
7650                 .renew          = 1,            /* append RENEW */
7651         };
7652         struct nfs4_fsid_present_res res = {
7653                 .renew          = 1,
7654         };
7655         struct rpc_message msg = {
7656                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7657                 .rpc_argp       = &args,
7658                 .rpc_resp       = &res,
7659                 .rpc_cred       = cred,
7660         };
7661         unsigned long now = jiffies;
7662         int status;
7663
7664         res.fh = nfs_alloc_fhandle();
7665         if (res.fh == NULL)
7666                 return -ENOMEM;
7667
7668         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7669         status = nfs4_call_sync_sequence(clnt, server, &msg,
7670                                                 &args.seq_args, &res.seq_res);
7671         nfs_free_fhandle(res.fh);
7672         if (status)
7673                 return status;
7674
7675         do_renew_lease(clp, now);
7676         return 0;
7677 }
7678
7679 #ifdef CONFIG_NFS_V4_1
7680
7681 /*
7682  * This operation also signals the server that this client is
7683  * performing "lease moved" recovery.  The server can stop asserting
7684  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7685  * this operation is identified in the SEQUENCE operation in this
7686  * compound.
7687  */
7688 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7689 {
7690         struct nfs_server *server = NFS_SERVER(inode);
7691         struct rpc_clnt *clnt = server->client;
7692         struct nfs4_fsid_present_arg args = {
7693                 .fh             = NFS_FH(inode),
7694         };
7695         struct nfs4_fsid_present_res res = {
7696         };
7697         struct rpc_message msg = {
7698                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7699                 .rpc_argp       = &args,
7700                 .rpc_resp       = &res,
7701                 .rpc_cred       = cred,
7702         };
7703         int status;
7704
7705         res.fh = nfs_alloc_fhandle();
7706         if (res.fh == NULL)
7707                 return -ENOMEM;
7708
7709         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7710         status = nfs4_call_sync_sequence(clnt, server, &msg,
7711                                                 &args.seq_args, &res.seq_res);
7712         nfs_free_fhandle(res.fh);
7713         if (status == NFS4_OK &&
7714             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7715                 status = -NFS4ERR_LEASE_MOVED;
7716         return status;
7717 }
7718
7719 #endif  /* CONFIG_NFS_V4_1 */
7720
7721 /**
7722  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7723  * @inode: inode on FSID to check
7724  * @cred: credential to use for this operation
7725  *
7726  * Server indicates whether the FSID is present, moved, or not
7727  * recognized.  This operation is necessary to clear a LEASE_MOVED
7728  * condition for this client ID.
7729  *
7730  * Returns NFS4_OK if the FSID is present on this server,
7731  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7732  *  NFS4ERR code if some error occurred on the server, or a
7733  *  negative errno if a local failure occurred.
7734  */
7735 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7736 {
7737         struct nfs_server *server = NFS_SERVER(inode);
7738         struct nfs_client *clp = server->nfs_client;
7739         const struct nfs4_mig_recovery_ops *ops =
7740                                         clp->cl_mvops->mig_recovery_ops;
7741         struct nfs4_exception exception = {
7742                 .interruptible = true,
7743         };
7744         int status;
7745
7746         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7747                 (unsigned long long)server->fsid.major,
7748                 (unsigned long long)server->fsid.minor,
7749                 clp->cl_hostname);
7750         nfs_display_fhandle(NFS_FH(inode), __func__);
7751
7752         do {
7753                 status = ops->fsid_present(inode, cred);
7754                 if (status != -NFS4ERR_DELAY)
7755                         break;
7756                 nfs4_handle_exception(server, status, &exception);
7757         } while (exception.retry);
7758         return status;
7759 }
7760
7761 /*
7762  * If 'use_integrity' is true and the state managment nfs_client
7763  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7764  * and the machine credential as per RFC3530bis and RFC5661 Security
7765  * Considerations sections. Otherwise, just use the user cred with the
7766  * filesystem's rpc_client.
7767  */
7768 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7769 {
7770         int status;
7771         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7772         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
7773         struct nfs4_secinfo_arg args = {
7774                 .dir_fh = NFS_FH(dir),
7775                 .name   = name,
7776         };
7777         struct nfs4_secinfo_res res = {
7778                 .flavors     = flavors,
7779         };
7780         struct rpc_message msg = {
7781                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7782                 .rpc_argp = &args,
7783                 .rpc_resp = &res,
7784         };
7785         struct nfs4_call_sync_data data = {
7786                 .seq_server = NFS_SERVER(dir),
7787                 .seq_args = &args.seq_args,
7788                 .seq_res = &res.seq_res,
7789         };
7790         struct rpc_task_setup task_setup = {
7791                 .rpc_client = clnt,
7792                 .rpc_message = &msg,
7793                 .callback_ops = clp->cl_mvops->call_sync_ops,
7794                 .callback_data = &data,
7795                 .flags = RPC_TASK_NO_ROUND_ROBIN,
7796         };
7797         const struct cred *cred = NULL;
7798
7799         if (use_integrity) {
7800                 clnt = clp->cl_rpcclient;
7801                 task_setup.rpc_client = clnt;
7802
7803                 cred = nfs4_get_clid_cred(clp);
7804                 msg.rpc_cred = cred;
7805         }
7806
7807         dprintk("NFS call  secinfo %s\n", name->name);
7808
7809         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7810         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
7811         status = nfs4_call_sync_custom(&task_setup);
7812
7813         dprintk("NFS reply  secinfo: %d\n", status);
7814
7815         put_cred(cred);
7816         return status;
7817 }
7818
7819 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7820                       struct nfs4_secinfo_flavors *flavors)
7821 {
7822         struct nfs4_exception exception = {
7823                 .interruptible = true,
7824         };
7825         int err;
7826         do {
7827                 err = -NFS4ERR_WRONGSEC;
7828
7829                 /* try to use integrity protection with machine cred */
7830                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7831                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7832
7833                 /*
7834                  * if unable to use integrity protection, or SECINFO with
7835                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7836                  * disallowed by spec, but exists in deployed servers) use
7837                  * the current filesystem's rpc_client and the user cred.
7838                  */
7839                 if (err == -NFS4ERR_WRONGSEC)
7840                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7841
7842                 trace_nfs4_secinfo(dir, name, err);
7843                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7844                                 &exception);
7845         } while (exception.retry);
7846         return err;
7847 }
7848
7849 #ifdef CONFIG_NFS_V4_1
7850 /*
7851  * Check the exchange flags returned by the server for invalid flags, having
7852  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7853  * DS flags set.
7854  */
7855 static int nfs4_check_cl_exchange_flags(u32 flags)
7856 {
7857         if (flags & ~EXCHGID4_FLAG_MASK_R)
7858                 goto out_inval;
7859         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7860             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7861                 goto out_inval;
7862         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7863                 goto out_inval;
7864         return NFS_OK;
7865 out_inval:
7866         return -NFS4ERR_INVAL;
7867 }
7868
7869 static bool
7870 nfs41_same_server_scope(struct nfs41_server_scope *a,
7871                         struct nfs41_server_scope *b)
7872 {
7873         if (a->server_scope_sz != b->server_scope_sz)
7874                 return false;
7875         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7876 }
7877
7878 static void
7879 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7880 {
7881         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
7882         struct nfs_client *clp = args->client;
7883
7884         switch (task->tk_status) {
7885         case -NFS4ERR_BADSESSION:
7886         case -NFS4ERR_DEADSESSION:
7887                 nfs4_schedule_session_recovery(clp->cl_session,
7888                                 task->tk_status);
7889         }
7890 }
7891
7892 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7893         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7894 };
7895
7896 /*
7897  * nfs4_proc_bind_one_conn_to_session()
7898  *
7899  * The 4.1 client currently uses the same TCP connection for the
7900  * fore and backchannel.
7901  */
7902 static
7903 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7904                 struct rpc_xprt *xprt,
7905                 struct nfs_client *clp,
7906                 const struct cred *cred)
7907 {
7908         int status;
7909         struct nfs41_bind_conn_to_session_args args = {
7910                 .client = clp,
7911                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7912         };
7913         struct nfs41_bind_conn_to_session_res res;
7914         struct rpc_message msg = {
7915                 .rpc_proc =
7916                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7917                 .rpc_argp = &args,
7918                 .rpc_resp = &res,
7919                 .rpc_cred = cred,
7920         };
7921         struct rpc_task_setup task_setup_data = {
7922                 .rpc_client = clnt,
7923                 .rpc_xprt = xprt,
7924                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7925                 .rpc_message = &msg,
7926                 .flags = RPC_TASK_TIMEOUT,
7927         };
7928         struct rpc_task *task;
7929
7930         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7931         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7932                 args.dir = NFS4_CDFC4_FORE;
7933
7934         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7935         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7936                 args.dir = NFS4_CDFC4_FORE;
7937
7938         task = rpc_run_task(&task_setup_data);
7939         if (!IS_ERR(task)) {
7940                 status = task->tk_status;
7941                 rpc_put_task(task);
7942         } else
7943                 status = PTR_ERR(task);
7944         trace_nfs4_bind_conn_to_session(clp, status);
7945         if (status == 0) {
7946                 if (memcmp(res.sessionid.data,
7947                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7948                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7949                         return -EIO;
7950                 }
7951                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7952                         dprintk("NFS: %s: Unexpected direction from server\n",
7953                                 __func__);
7954                         return -EIO;
7955                 }
7956                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7957                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7958                                 __func__);
7959                         return -EIO;
7960                 }
7961         }
7962
7963         return status;
7964 }
7965
7966 struct rpc_bind_conn_calldata {
7967         struct nfs_client *clp;
7968         const struct cred *cred;
7969 };
7970
7971 static int
7972 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7973                 struct rpc_xprt *xprt,
7974                 void *calldata)
7975 {
7976         struct rpc_bind_conn_calldata *p = calldata;
7977
7978         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7979 }
7980
7981 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
7982 {
7983         struct rpc_bind_conn_calldata data = {
7984                 .clp = clp,
7985                 .cred = cred,
7986         };
7987         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7988                         nfs4_proc_bind_conn_to_session_callback, &data);
7989 }
7990
7991 /*
7992  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7993  * and operations we'd like to see to enable certain features in the allow map
7994  */
7995 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7996         .how = SP4_MACH_CRED,
7997         .enforce.u.words = {
7998                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7999                       1 << (OP_EXCHANGE_ID - 32) |
8000                       1 << (OP_CREATE_SESSION - 32) |
8001                       1 << (OP_DESTROY_SESSION - 32) |
8002                       1 << (OP_DESTROY_CLIENTID - 32)
8003         },
8004         .allow.u.words = {
8005                 [0] = 1 << (OP_CLOSE) |
8006                       1 << (OP_OPEN_DOWNGRADE) |
8007                       1 << (OP_LOCKU) |
8008                       1 << (OP_DELEGRETURN) |
8009                       1 << (OP_COMMIT),
8010                 [1] = 1 << (OP_SECINFO - 32) |
8011                       1 << (OP_SECINFO_NO_NAME - 32) |
8012                       1 << (OP_LAYOUTRETURN - 32) |
8013                       1 << (OP_TEST_STATEID - 32) |
8014                       1 << (OP_FREE_STATEID - 32) |
8015                       1 << (OP_WRITE - 32)
8016         }
8017 };
8018
8019 /*
8020  * Select the state protection mode for client `clp' given the server results
8021  * from exchange_id in `sp'.
8022  *
8023  * Returns 0 on success, negative errno otherwise.
8024  */
8025 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8026                                  struct nfs41_state_protection *sp)
8027 {
8028         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8029                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8030                       1 << (OP_EXCHANGE_ID - 32) |
8031                       1 << (OP_CREATE_SESSION - 32) |
8032                       1 << (OP_DESTROY_SESSION - 32) |
8033                       1 << (OP_DESTROY_CLIENTID - 32)
8034         };
8035         unsigned long flags = 0;
8036         unsigned int i;
8037         int ret = 0;
8038
8039         if (sp->how == SP4_MACH_CRED) {
8040                 /* Print state protect result */
8041                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8042                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8043                         if (test_bit(i, sp->enforce.u.longs))
8044                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8045                         if (test_bit(i, sp->allow.u.longs))
8046                                 dfprintk(MOUNT, "  allow op %d\n", i);
8047                 }
8048
8049                 /* make sure nothing is on enforce list that isn't supported */
8050                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8051                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8052                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8053                                 ret = -EINVAL;
8054                                 goto out;
8055                         }
8056                 }
8057
8058                 /*
8059                  * Minimal mode - state operations are allowed to use machine
8060                  * credential.  Note this already happens by default, so the
8061                  * client doesn't have to do anything more than the negotiation.
8062                  *
8063                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8064                  *       we're already using the machine cred for exchange_id
8065                  *       and will never use a different cred.
8066                  */
8067                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8068                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8069                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8070                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8071                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8072                         dfprintk(MOUNT, "  minimal mode enabled\n");
8073                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8074                 } else {
8075                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8076                         ret = -EINVAL;
8077                         goto out;
8078                 }
8079
8080                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8081                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8082                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8083                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8084                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8085                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8086                 }
8087
8088                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8089                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8090                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8091                 }
8092
8093                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8094                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8095                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8096                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8097                 }
8098
8099                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8100                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8101                         dfprintk(MOUNT, "  stateid mode enabled\n");
8102                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8103                 }
8104
8105                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8106                         dfprintk(MOUNT, "  write mode enabled\n");
8107                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8108                 }
8109
8110                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8111                         dfprintk(MOUNT, "  commit mode enabled\n");
8112                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8113                 }
8114         }
8115 out:
8116         clp->cl_sp4_flags = flags;
8117         return ret;
8118 }
8119
8120 struct nfs41_exchange_id_data {
8121         struct nfs41_exchange_id_res res;
8122         struct nfs41_exchange_id_args args;
8123 };
8124
8125 static void nfs4_exchange_id_release(void *data)
8126 {
8127         struct nfs41_exchange_id_data *cdata =
8128                                         (struct nfs41_exchange_id_data *)data;
8129
8130         nfs_put_client(cdata->args.client);
8131         kfree(cdata->res.impl_id);
8132         kfree(cdata->res.server_scope);
8133         kfree(cdata->res.server_owner);
8134         kfree(cdata);
8135 }
8136
8137 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8138         .rpc_release = nfs4_exchange_id_release,
8139 };
8140
8141 /*
8142  * _nfs4_proc_exchange_id()
8143  *
8144  * Wrapper for EXCHANGE_ID operation.
8145  */
8146 static struct rpc_task *
8147 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8148                         u32 sp4_how, struct rpc_xprt *xprt)
8149 {
8150         struct rpc_message msg = {
8151                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8152                 .rpc_cred = cred,
8153         };
8154         struct rpc_task_setup task_setup_data = {
8155                 .rpc_client = clp->cl_rpcclient,
8156                 .callback_ops = &nfs4_exchange_id_call_ops,
8157                 .rpc_message = &msg,
8158                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8159         };
8160         struct nfs41_exchange_id_data *calldata;
8161         int status;
8162
8163         if (!refcount_inc_not_zero(&clp->cl_count))
8164                 return ERR_PTR(-EIO);
8165
8166         status = -ENOMEM;
8167         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8168         if (!calldata)
8169                 goto out;
8170
8171         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8172
8173         status = nfs4_init_uniform_client_string(clp);
8174         if (status)
8175                 goto out_calldata;
8176
8177         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8178                                                 GFP_NOFS);
8179         status = -ENOMEM;
8180         if (unlikely(calldata->res.server_owner == NULL))
8181                 goto out_calldata;
8182
8183         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8184                                         GFP_NOFS);
8185         if (unlikely(calldata->res.server_scope == NULL))
8186                 goto out_server_owner;
8187
8188         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8189         if (unlikely(calldata->res.impl_id == NULL))
8190                 goto out_server_scope;
8191
8192         switch (sp4_how) {
8193         case SP4_NONE:
8194                 calldata->args.state_protect.how = SP4_NONE;
8195                 break;
8196
8197         case SP4_MACH_CRED:
8198                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8199                 break;
8200
8201         default:
8202                 /* unsupported! */
8203                 WARN_ON_ONCE(1);
8204                 status = -EINVAL;
8205                 goto out_impl_id;
8206         }
8207         if (xprt) {
8208                 task_setup_data.rpc_xprt = xprt;
8209                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8210                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8211                                 sizeof(calldata->args.verifier.data));
8212         }
8213         calldata->args.client = clp;
8214         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8215         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8216 #ifdef CONFIG_NFS_V4_1_MIGRATION
8217         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8218 #endif
8219         msg.rpc_argp = &calldata->args;
8220         msg.rpc_resp = &calldata->res;
8221         task_setup_data.callback_data = calldata;
8222
8223         return rpc_run_task(&task_setup_data);
8224
8225 out_impl_id:
8226         kfree(calldata->res.impl_id);
8227 out_server_scope:
8228         kfree(calldata->res.server_scope);
8229 out_server_owner:
8230         kfree(calldata->res.server_owner);
8231 out_calldata:
8232         kfree(calldata);
8233 out:
8234         nfs_put_client(clp);
8235         return ERR_PTR(status);
8236 }
8237
8238 /*
8239  * _nfs4_proc_exchange_id()
8240  *
8241  * Wrapper for EXCHANGE_ID operation.
8242  */
8243 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8244                         u32 sp4_how)
8245 {
8246         struct rpc_task *task;
8247         struct nfs41_exchange_id_args *argp;
8248         struct nfs41_exchange_id_res *resp;
8249         unsigned long now = jiffies;
8250         int status;
8251
8252         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8253         if (IS_ERR(task))
8254                 return PTR_ERR(task);
8255
8256         argp = task->tk_msg.rpc_argp;
8257         resp = task->tk_msg.rpc_resp;
8258         status = task->tk_status;
8259         if (status  != 0)
8260                 goto out;
8261
8262         status = nfs4_check_cl_exchange_flags(resp->flags);
8263         if (status  != 0)
8264                 goto out;
8265
8266         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8267         if (status != 0)
8268                 goto out;
8269
8270         do_renew_lease(clp, now);
8271
8272         clp->cl_clientid = resp->clientid;
8273         clp->cl_exchange_flags = resp->flags;
8274         clp->cl_seqid = resp->seqid;
8275         /* Client ID is not confirmed */
8276         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8277                 clear_bit(NFS4_SESSION_ESTABLISHED,
8278                           &clp->cl_session->session_state);
8279
8280         if (clp->cl_serverscope != NULL &&
8281             !nfs41_same_server_scope(clp->cl_serverscope,
8282                                 resp->server_scope)) {
8283                 dprintk("%s: server_scope mismatch detected\n",
8284                         __func__);
8285                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8286         }
8287
8288         swap(clp->cl_serverowner, resp->server_owner);
8289         swap(clp->cl_serverscope, resp->server_scope);
8290         swap(clp->cl_implid, resp->impl_id);
8291
8292         /* Save the EXCHANGE_ID verifier session trunk tests */
8293         memcpy(clp->cl_confirm.data, argp->verifier.data,
8294                sizeof(clp->cl_confirm.data));
8295 out:
8296         trace_nfs4_exchange_id(clp, status);
8297         rpc_put_task(task);
8298         return status;
8299 }
8300
8301 /*
8302  * nfs4_proc_exchange_id()
8303  *
8304  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8305  *
8306  * Since the clientid has expired, all compounds using sessions
8307  * associated with the stale clientid will be returning
8308  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8309  * be in some phase of session reset.
8310  *
8311  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8312  */
8313 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8314 {
8315         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8316         int status;
8317
8318         /* try SP4_MACH_CRED if krb5i/p */
8319         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8320             authflavor == RPC_AUTH_GSS_KRB5P) {
8321                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8322                 if (!status)
8323                         return 0;
8324         }
8325
8326         /* try SP4_NONE */
8327         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8328 }
8329
8330 /**
8331  * nfs4_test_session_trunk
8332  *
8333  * This is an add_xprt_test() test function called from
8334  * rpc_clnt_setup_test_and_add_xprt.
8335  *
8336  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8337  * and is dereferrenced in nfs4_exchange_id_release
8338  *
8339  * Upon success, add the new transport to the rpc_clnt
8340  *
8341  * @clnt: struct rpc_clnt to get new transport
8342  * @xprt: the rpc_xprt to test
8343  * @data: call data for _nfs4_proc_exchange_id.
8344  */
8345 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8346                             void *data)
8347 {
8348         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8349         struct rpc_task *task;
8350         int status;
8351
8352         u32 sp4_how;
8353
8354         dprintk("--> %s try %s\n", __func__,
8355                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8356
8357         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8358
8359         /* Test connection for session trunking. Async exchange_id call */
8360         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8361         if (IS_ERR(task))
8362                 return;
8363
8364         status = task->tk_status;
8365         if (status == 0)
8366                 status = nfs4_detect_session_trunking(adata->clp,
8367                                 task->tk_msg.rpc_resp, xprt);
8368
8369         if (status == 0)
8370                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8371
8372         rpc_put_task(task);
8373 }
8374 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8375
8376 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8377                 const struct cred *cred)
8378 {
8379         struct rpc_message msg = {
8380                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8381                 .rpc_argp = clp,
8382                 .rpc_cred = cred,
8383         };
8384         int status;
8385
8386         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8387                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8388         trace_nfs4_destroy_clientid(clp, status);
8389         if (status)
8390                 dprintk("NFS: Got error %d from the server %s on "
8391                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8392         return status;
8393 }
8394
8395 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8396                 const struct cred *cred)
8397 {
8398         unsigned int loop;
8399         int ret;
8400
8401         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8402                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8403                 switch (ret) {
8404                 case -NFS4ERR_DELAY:
8405                 case -NFS4ERR_CLIENTID_BUSY:
8406                         ssleep(1);
8407                         break;
8408                 default:
8409                         return ret;
8410                 }
8411         }
8412         return 0;
8413 }
8414
8415 int nfs4_destroy_clientid(struct nfs_client *clp)
8416 {
8417         const struct cred *cred;
8418         int ret = 0;
8419
8420         if (clp->cl_mvops->minor_version < 1)
8421                 goto out;
8422         if (clp->cl_exchange_flags == 0)
8423                 goto out;
8424         if (clp->cl_preserve_clid)
8425                 goto out;
8426         cred = nfs4_get_clid_cred(clp);
8427         ret = nfs4_proc_destroy_clientid(clp, cred);
8428         put_cred(cred);
8429         switch (ret) {
8430         case 0:
8431         case -NFS4ERR_STALE_CLIENTID:
8432                 clp->cl_exchange_flags = 0;
8433         }
8434 out:
8435         return ret;
8436 }
8437
8438 #endif /* CONFIG_NFS_V4_1 */
8439
8440 struct nfs4_get_lease_time_data {
8441         struct nfs4_get_lease_time_args *args;
8442         struct nfs4_get_lease_time_res *res;
8443         struct nfs_client *clp;
8444 };
8445
8446 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8447                                         void *calldata)
8448 {
8449         struct nfs4_get_lease_time_data *data =
8450                         (struct nfs4_get_lease_time_data *)calldata;
8451
8452         dprintk("--> %s\n", __func__);
8453         /* just setup sequence, do not trigger session recovery
8454            since we're invoked within one */
8455         nfs4_setup_sequence(data->clp,
8456                         &data->args->la_seq_args,
8457                         &data->res->lr_seq_res,
8458                         task);
8459         dprintk("<-- %s\n", __func__);
8460 }
8461
8462 /*
8463  * Called from nfs4_state_manager thread for session setup, so don't recover
8464  * from sequence operation or clientid errors.
8465  */
8466 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8467 {
8468         struct nfs4_get_lease_time_data *data =
8469                         (struct nfs4_get_lease_time_data *)calldata;
8470
8471         dprintk("--> %s\n", __func__);
8472         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8473                 return;
8474         switch (task->tk_status) {
8475         case -NFS4ERR_DELAY:
8476         case -NFS4ERR_GRACE:
8477                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8478                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8479                 task->tk_status = 0;
8480                 /* fall through */
8481         case -NFS4ERR_RETRY_UNCACHED_REP:
8482                 rpc_restart_call_prepare(task);
8483                 return;
8484         }
8485         dprintk("<-- %s\n", __func__);
8486 }
8487
8488 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8489         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8490         .rpc_call_done = nfs4_get_lease_time_done,
8491 };
8492
8493 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8494 {
8495         struct nfs4_get_lease_time_args args;
8496         struct nfs4_get_lease_time_res res = {
8497                 .lr_fsinfo = fsinfo,
8498         };
8499         struct nfs4_get_lease_time_data data = {
8500                 .args = &args,
8501                 .res = &res,
8502                 .clp = clp,
8503         };
8504         struct rpc_message msg = {
8505                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8506                 .rpc_argp = &args,
8507                 .rpc_resp = &res,
8508         };
8509         struct rpc_task_setup task_setup = {
8510                 .rpc_client = clp->cl_rpcclient,
8511                 .rpc_message = &msg,
8512                 .callback_ops = &nfs4_get_lease_time_ops,
8513                 .callback_data = &data,
8514                 .flags = RPC_TASK_TIMEOUT,
8515         };
8516
8517         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8518         return nfs4_call_sync_custom(&task_setup);
8519 }
8520
8521 #ifdef CONFIG_NFS_V4_1
8522
8523 /*
8524  * Initialize the values to be used by the client in CREATE_SESSION
8525  * If nfs4_init_session set the fore channel request and response sizes,
8526  * use them.
8527  *
8528  * Set the back channel max_resp_sz_cached to zero to force the client to
8529  * always set csa_cachethis to FALSE because the current implementation
8530  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8531  */
8532 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8533                                     struct rpc_clnt *clnt)
8534 {
8535         unsigned int max_rqst_sz, max_resp_sz;
8536         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8537         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8538
8539         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8540         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8541
8542         /* Fore channel attributes */
8543         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8544         args->fc_attrs.max_resp_sz = max_resp_sz;
8545         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8546         args->fc_attrs.max_reqs = max_session_slots;
8547
8548         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8549                 "max_ops=%u max_reqs=%u\n",
8550                 __func__,
8551                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8552                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8553
8554         /* Back channel attributes */
8555         args->bc_attrs.max_rqst_sz = max_bc_payload;
8556         args->bc_attrs.max_resp_sz = max_bc_payload;
8557         args->bc_attrs.max_resp_sz_cached = 0;
8558         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8559         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8560         if (args->bc_attrs.max_reqs > max_bc_slots)
8561                 args->bc_attrs.max_reqs = max_bc_slots;
8562
8563         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8564                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8565                 __func__,
8566                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8567                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8568                 args->bc_attrs.max_reqs);
8569 }
8570
8571 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8572                 struct nfs41_create_session_res *res)
8573 {
8574         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8575         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8576
8577         if (rcvd->max_resp_sz > sent->max_resp_sz)
8578                 return -EINVAL;
8579         /*
8580          * Our requested max_ops is the minimum we need; we're not
8581          * prepared to break up compounds into smaller pieces than that.
8582          * So, no point even trying to continue if the server won't
8583          * cooperate:
8584          */
8585         if (rcvd->max_ops < sent->max_ops)
8586                 return -EINVAL;
8587         if (rcvd->max_reqs == 0)
8588                 return -EINVAL;
8589         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8590                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8591         return 0;
8592 }
8593
8594 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8595                 struct nfs41_create_session_res *res)
8596 {
8597         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8598         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8599
8600         if (!(res->flags & SESSION4_BACK_CHAN))
8601                 goto out;
8602         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8603                 return -EINVAL;
8604         if (rcvd->max_resp_sz < sent->max_resp_sz)
8605                 return -EINVAL;
8606         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8607                 return -EINVAL;
8608         if (rcvd->max_ops > sent->max_ops)
8609                 return -EINVAL;
8610         if (rcvd->max_reqs > sent->max_reqs)
8611                 return -EINVAL;
8612 out:
8613         return 0;
8614 }
8615
8616 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8617                                      struct nfs41_create_session_res *res)
8618 {
8619         int ret;
8620
8621         ret = nfs4_verify_fore_channel_attrs(args, res);
8622         if (ret)
8623                 return ret;
8624         return nfs4_verify_back_channel_attrs(args, res);
8625 }
8626
8627 static void nfs4_update_session(struct nfs4_session *session,
8628                 struct nfs41_create_session_res *res)
8629 {
8630         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8631         /* Mark client id and session as being confirmed */
8632         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8633         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8634         session->flags = res->flags;
8635         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8636         if (res->flags & SESSION4_BACK_CHAN)
8637                 memcpy(&session->bc_attrs, &res->bc_attrs,
8638                                 sizeof(session->bc_attrs));
8639 }
8640
8641 static int _nfs4_proc_create_session(struct nfs_client *clp,
8642                 const struct cred *cred)
8643 {
8644         struct nfs4_session *session = clp->cl_session;
8645         struct nfs41_create_session_args args = {
8646                 .client = clp,
8647                 .clientid = clp->cl_clientid,
8648                 .seqid = clp->cl_seqid,
8649                 .cb_program = NFS4_CALLBACK,
8650         };
8651         struct nfs41_create_session_res res;
8652
8653         struct rpc_message msg = {
8654                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8655                 .rpc_argp = &args,
8656                 .rpc_resp = &res,
8657                 .rpc_cred = cred,
8658         };
8659         int status;
8660
8661         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8662         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8663
8664         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8665                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8666         trace_nfs4_create_session(clp, status);
8667
8668         switch (status) {
8669         case -NFS4ERR_STALE_CLIENTID:
8670         case -NFS4ERR_DELAY:
8671         case -ETIMEDOUT:
8672         case -EACCES:
8673         case -EAGAIN:
8674                 goto out;
8675         }
8676
8677         clp->cl_seqid++;
8678         if (!status) {
8679                 /* Verify the session's negotiated channel_attrs values */
8680                 status = nfs4_verify_channel_attrs(&args, &res);
8681                 /* Increment the clientid slot sequence id */
8682                 if (status)
8683                         goto out;
8684                 nfs4_update_session(session, &res);
8685         }
8686 out:
8687         return status;
8688 }
8689
8690 /*
8691  * Issues a CREATE_SESSION operation to the server.
8692  * It is the responsibility of the caller to verify the session is
8693  * expired before calling this routine.
8694  */
8695 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8696 {
8697         int status;
8698         unsigned *ptr;
8699         struct nfs4_session *session = clp->cl_session;
8700
8701         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8702
8703         status = _nfs4_proc_create_session(clp, cred);
8704         if (status)
8705                 goto out;
8706
8707         /* Init or reset the session slot tables */
8708         status = nfs4_setup_session_slot_tables(session);
8709         dprintk("slot table setup returned %d\n", status);
8710         if (status)
8711                 goto out;
8712
8713         ptr = (unsigned *)&session->sess_id.data[0];
8714         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8715                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8716 out:
8717         dprintk("<-- %s\n", __func__);
8718         return status;
8719 }
8720
8721 /*
8722  * Issue the over-the-wire RPC DESTROY_SESSION.
8723  * The caller must serialize access to this routine.
8724  */
8725 int nfs4_proc_destroy_session(struct nfs4_session *session,
8726                 const struct cred *cred)
8727 {
8728         struct rpc_message msg = {
8729                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8730                 .rpc_argp = session,
8731                 .rpc_cred = cred,
8732         };
8733         int status = 0;
8734
8735         dprintk("--> nfs4_proc_destroy_session\n");
8736
8737         /* session is still being setup */
8738         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8739                 return 0;
8740
8741         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8742                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8743         trace_nfs4_destroy_session(session->clp, status);
8744
8745         if (status)
8746                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8747                         "Session has been destroyed regardless...\n", status);
8748
8749         dprintk("<-- nfs4_proc_destroy_session\n");
8750         return status;
8751 }
8752
8753 /*
8754  * Renew the cl_session lease.
8755  */
8756 struct nfs4_sequence_data {
8757         struct nfs_client *clp;
8758         struct nfs4_sequence_args args;
8759         struct nfs4_sequence_res res;
8760 };
8761
8762 static void nfs41_sequence_release(void *data)
8763 {
8764         struct nfs4_sequence_data *calldata = data;
8765         struct nfs_client *clp = calldata->clp;
8766
8767         if (refcount_read(&clp->cl_count) > 1)
8768                 nfs4_schedule_state_renewal(clp);
8769         nfs_put_client(clp);
8770         kfree(calldata);
8771 }
8772
8773 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8774 {
8775         switch(task->tk_status) {
8776         case -NFS4ERR_DELAY:
8777                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8778                 return -EAGAIN;
8779         default:
8780                 nfs4_schedule_lease_recovery(clp);
8781         }
8782         return 0;
8783 }
8784
8785 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8786 {
8787         struct nfs4_sequence_data *calldata = data;
8788         struct nfs_client *clp = calldata->clp;
8789
8790         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8791                 return;
8792
8793         trace_nfs4_sequence(clp, task->tk_status);
8794         if (task->tk_status < 0) {
8795                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8796                 if (refcount_read(&clp->cl_count) == 1)
8797                         goto out;
8798
8799                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8800                         rpc_restart_call_prepare(task);
8801                         return;
8802                 }
8803         }
8804         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8805 out:
8806         dprintk("<-- %s\n", __func__);
8807 }
8808
8809 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8810 {
8811         struct nfs4_sequence_data *calldata = data;
8812         struct nfs_client *clp = calldata->clp;
8813         struct nfs4_sequence_args *args;
8814         struct nfs4_sequence_res *res;
8815
8816         args = task->tk_msg.rpc_argp;
8817         res = task->tk_msg.rpc_resp;
8818
8819         nfs4_setup_sequence(clp, args, res, task);
8820 }
8821
8822 static const struct rpc_call_ops nfs41_sequence_ops = {
8823         .rpc_call_done = nfs41_sequence_call_done,
8824         .rpc_call_prepare = nfs41_sequence_prepare,
8825         .rpc_release = nfs41_sequence_release,
8826 };
8827
8828 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8829                 const struct cred *cred,
8830                 struct nfs4_slot *slot,
8831                 bool is_privileged)
8832 {
8833         struct nfs4_sequence_data *calldata;
8834         struct rpc_message msg = {
8835                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8836                 .rpc_cred = cred,
8837         };
8838         struct rpc_task_setup task_setup_data = {
8839                 .rpc_client = clp->cl_rpcclient,
8840                 .rpc_message = &msg,
8841                 .callback_ops = &nfs41_sequence_ops,
8842                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8843         };
8844         struct rpc_task *ret;
8845
8846         ret = ERR_PTR(-EIO);
8847         if (!refcount_inc_not_zero(&clp->cl_count))
8848                 goto out_err;
8849
8850         ret = ERR_PTR(-ENOMEM);
8851         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8852         if (calldata == NULL)
8853                 goto out_put_clp;
8854         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8855         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8856         msg.rpc_argp = &calldata->args;
8857         msg.rpc_resp = &calldata->res;
8858         calldata->clp = clp;
8859         task_setup_data.callback_data = calldata;
8860
8861         ret = rpc_run_task(&task_setup_data);
8862         if (IS_ERR(ret))
8863                 goto out_err;
8864         return ret;
8865 out_put_clp:
8866         nfs_put_client(clp);
8867 out_err:
8868         nfs41_release_slot(slot);
8869         return ret;
8870 }
8871
8872 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8873 {
8874         struct rpc_task *task;
8875         int ret = 0;
8876
8877         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8878                 return -EAGAIN;
8879         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8880         if (IS_ERR(task))
8881                 ret = PTR_ERR(task);
8882         else
8883                 rpc_put_task_async(task);
8884         dprintk("<-- %s status=%d\n", __func__, ret);
8885         return ret;
8886 }
8887
8888 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8889 {
8890         struct rpc_task *task;
8891         int ret;
8892
8893         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8894         if (IS_ERR(task)) {
8895                 ret = PTR_ERR(task);
8896                 goto out;
8897         }
8898         ret = rpc_wait_for_completion_task(task);
8899         if (!ret)
8900                 ret = task->tk_status;
8901         rpc_put_task(task);
8902 out:
8903         dprintk("<-- %s status=%d\n", __func__, ret);
8904         return ret;
8905 }
8906
8907 struct nfs4_reclaim_complete_data {
8908         struct nfs_client *clp;
8909         struct nfs41_reclaim_complete_args arg;
8910         struct nfs41_reclaim_complete_res res;
8911 };
8912
8913 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8914 {
8915         struct nfs4_reclaim_complete_data *calldata = data;
8916
8917         nfs4_setup_sequence(calldata->clp,
8918                         &calldata->arg.seq_args,
8919                         &calldata->res.seq_res,
8920                         task);
8921 }
8922
8923 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8924 {
8925         switch(task->tk_status) {
8926         case 0:
8927                 wake_up_all(&clp->cl_lock_waitq);
8928                 /* Fallthrough */
8929         case -NFS4ERR_COMPLETE_ALREADY:
8930         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8931                 break;
8932         case -NFS4ERR_DELAY:
8933                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8934                 /* fall through */
8935         case -NFS4ERR_RETRY_UNCACHED_REP:
8936                 return -EAGAIN;
8937         case -NFS4ERR_BADSESSION:
8938         case -NFS4ERR_DEADSESSION:
8939         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8940                 break;
8941         default:
8942                 nfs4_schedule_lease_recovery(clp);
8943         }
8944         return 0;
8945 }
8946
8947 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8948 {
8949         struct nfs4_reclaim_complete_data *calldata = data;
8950         struct nfs_client *clp = calldata->clp;
8951         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8952
8953         dprintk("--> %s\n", __func__);
8954         if (!nfs41_sequence_done(task, res))
8955                 return;
8956
8957         trace_nfs4_reclaim_complete(clp, task->tk_status);
8958         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8959                 rpc_restart_call_prepare(task);
8960                 return;
8961         }
8962         dprintk("<-- %s\n", __func__);
8963 }
8964
8965 static void nfs4_free_reclaim_complete_data(void *data)
8966 {
8967         struct nfs4_reclaim_complete_data *calldata = data;
8968
8969         kfree(calldata);
8970 }
8971
8972 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8973         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8974         .rpc_call_done = nfs4_reclaim_complete_done,
8975         .rpc_release = nfs4_free_reclaim_complete_data,
8976 };
8977
8978 /*
8979  * Issue a global reclaim complete.
8980  */
8981 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8982                 const struct cred *cred)
8983 {
8984         struct nfs4_reclaim_complete_data *calldata;
8985         struct rpc_message msg = {
8986                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8987                 .rpc_cred = cred,
8988         };
8989         struct rpc_task_setup task_setup_data = {
8990                 .rpc_client = clp->cl_rpcclient,
8991                 .rpc_message = &msg,
8992                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8993                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8994         };
8995         int status = -ENOMEM;
8996
8997         dprintk("--> %s\n", __func__);
8998         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8999         if (calldata == NULL)
9000                 goto out;
9001         calldata->clp = clp;
9002         calldata->arg.one_fs = 0;
9003
9004         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9005         msg.rpc_argp = &calldata->arg;
9006         msg.rpc_resp = &calldata->res;
9007         task_setup_data.callback_data = calldata;
9008         status = nfs4_call_sync_custom(&task_setup_data);
9009 out:
9010         dprintk("<-- %s status=%d\n", __func__, status);
9011         return status;
9012 }
9013
9014 static void
9015 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9016 {
9017         struct nfs4_layoutget *lgp = calldata;
9018         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9019
9020         dprintk("--> %s\n", __func__);
9021         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9022                                 &lgp->res.seq_res, task);
9023         dprintk("<-- %s\n", __func__);
9024 }
9025
9026 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9027 {
9028         struct nfs4_layoutget *lgp = calldata;
9029
9030         dprintk("--> %s\n", __func__);
9031         nfs41_sequence_process(task, &lgp->res.seq_res);
9032         dprintk("<-- %s\n", __func__);
9033 }
9034
9035 static int
9036 nfs4_layoutget_handle_exception(struct rpc_task *task,
9037                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9038 {
9039         struct inode *inode = lgp->args.inode;
9040         struct nfs_server *server = NFS_SERVER(inode);
9041         struct pnfs_layout_hdr *lo;
9042         int nfs4err = task->tk_status;
9043         int err, status = 0;
9044         LIST_HEAD(head);
9045
9046         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9047
9048         nfs4_sequence_free_slot(&lgp->res.seq_res);
9049
9050         switch (nfs4err) {
9051         case 0:
9052                 goto out;
9053
9054         /*
9055          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9056          * on the file. set tk_status to -ENODATA to tell upper layer to
9057          * retry go inband.
9058          */
9059         case -NFS4ERR_LAYOUTUNAVAILABLE:
9060                 status = -ENODATA;
9061                 goto out;
9062         /*
9063          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9064          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9065          */
9066         case -NFS4ERR_BADLAYOUT:
9067                 status = -EOVERFLOW;
9068                 goto out;
9069         /*
9070          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9071          * (or clients) writing to the same RAID stripe except when
9072          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9073          *
9074          * Treat it like we would RECALLCONFLICT -- we retry for a little
9075          * while, and then eventually give up.
9076          */
9077         case -NFS4ERR_LAYOUTTRYLATER:
9078                 if (lgp->args.minlength == 0) {
9079                         status = -EOVERFLOW;
9080                         goto out;
9081                 }
9082                 status = -EBUSY;
9083                 break;
9084         case -NFS4ERR_RECALLCONFLICT:
9085                 status = -ERECALLCONFLICT;
9086                 break;
9087         case -NFS4ERR_DELEG_REVOKED:
9088         case -NFS4ERR_ADMIN_REVOKED:
9089         case -NFS4ERR_EXPIRED:
9090         case -NFS4ERR_BAD_STATEID:
9091                 exception->timeout = 0;
9092                 spin_lock(&inode->i_lock);
9093                 lo = NFS_I(inode)->layout;
9094                 /* If the open stateid was bad, then recover it. */
9095                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9096                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9097                         spin_unlock(&inode->i_lock);
9098                         exception->state = lgp->args.ctx->state;
9099                         exception->stateid = &lgp->args.stateid;
9100                         break;
9101                 }
9102
9103                 /*
9104                  * Mark the bad layout state as invalid, then retry
9105                  */
9106                 pnfs_mark_layout_stateid_invalid(lo, &head);
9107                 spin_unlock(&inode->i_lock);
9108                 nfs_commit_inode(inode, 0);
9109                 pnfs_free_lseg_list(&head);
9110                 status = -EAGAIN;
9111                 goto out;
9112         }
9113
9114         err = nfs4_handle_exception(server, nfs4err, exception);
9115         if (!status) {
9116                 if (exception->retry)
9117                         status = -EAGAIN;
9118                 else
9119                         status = err;
9120         }
9121 out:
9122         dprintk("<-- %s\n", __func__);
9123         return status;
9124 }
9125
9126 size_t max_response_pages(struct nfs_server *server)
9127 {
9128         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9129         return nfs_page_array_len(0, max_resp_sz);
9130 }
9131
9132 static void nfs4_layoutget_release(void *calldata)
9133 {
9134         struct nfs4_layoutget *lgp = calldata;
9135
9136         dprintk("--> %s\n", __func__);
9137         nfs4_sequence_free_slot(&lgp->res.seq_res);
9138         pnfs_layoutget_free(lgp);
9139         dprintk("<-- %s\n", __func__);
9140 }
9141
9142 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9143         .rpc_call_prepare = nfs4_layoutget_prepare,
9144         .rpc_call_done = nfs4_layoutget_done,
9145         .rpc_release = nfs4_layoutget_release,
9146 };
9147
9148 struct pnfs_layout_segment *
9149 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9150 {
9151         struct inode *inode = lgp->args.inode;
9152         struct nfs_server *server = NFS_SERVER(inode);
9153         struct rpc_task *task;
9154         struct rpc_message msg = {
9155                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9156                 .rpc_argp = &lgp->args,
9157                 .rpc_resp = &lgp->res,
9158                 .rpc_cred = lgp->cred,
9159         };
9160         struct rpc_task_setup task_setup_data = {
9161                 .rpc_client = server->client,
9162                 .rpc_message = &msg,
9163                 .callback_ops = &nfs4_layoutget_call_ops,
9164                 .callback_data = lgp,
9165                 .flags = RPC_TASK_ASYNC,
9166         };
9167         struct pnfs_layout_segment *lseg = NULL;
9168         struct nfs4_exception exception = {
9169                 .inode = inode,
9170                 .timeout = *timeout,
9171         };
9172         int status = 0;
9173
9174         dprintk("--> %s\n", __func__);
9175
9176         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9177         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9178
9179         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9180
9181         task = rpc_run_task(&task_setup_data);
9182         if (IS_ERR(task))
9183                 return ERR_CAST(task);
9184         status = rpc_wait_for_completion_task(task);
9185         if (status != 0)
9186                 goto out;
9187
9188         if (task->tk_status < 0) {
9189                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9190                 *timeout = exception.timeout;
9191         } else if (lgp->res.layoutp->len == 0) {
9192                 status = -EAGAIN;
9193                 *timeout = nfs4_update_delay(&exception.timeout);
9194         } else
9195                 lseg = pnfs_layout_process(lgp);
9196 out:
9197         trace_nfs4_layoutget(lgp->args.ctx,
9198                         &lgp->args.range,
9199                         &lgp->res.range,
9200                         &lgp->res.stateid,
9201                         status);
9202
9203         rpc_put_task(task);
9204         dprintk("<-- %s status=%d\n", __func__, status);
9205         if (status)
9206                 return ERR_PTR(status);
9207         return lseg;
9208 }
9209
9210 static void
9211 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9212 {
9213         struct nfs4_layoutreturn *lrp = calldata;
9214
9215         dprintk("--> %s\n", __func__);
9216         nfs4_setup_sequence(lrp->clp,
9217                         &lrp->args.seq_args,
9218                         &lrp->res.seq_res,
9219                         task);
9220         if (!pnfs_layout_is_valid(lrp->args.layout))
9221                 rpc_exit(task, 0);
9222 }
9223
9224 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9225 {
9226         struct nfs4_layoutreturn *lrp = calldata;
9227         struct nfs_server *server;
9228
9229         dprintk("--> %s\n", __func__);
9230
9231         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9232                 return;
9233
9234         /*
9235          * Was there an RPC level error? Assume the call succeeded,
9236          * and that we need to release the layout
9237          */
9238         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9239                 lrp->res.lrs_present = 0;
9240                 return;
9241         }
9242
9243         server = NFS_SERVER(lrp->args.inode);
9244         switch (task->tk_status) {
9245         case -NFS4ERR_OLD_STATEID:
9246                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9247                                         &lrp->args.range,
9248                                         lrp->args.inode))
9249                         goto out_restart;
9250                 /* Fallthrough */
9251         default:
9252                 task->tk_status = 0;
9253                 /* Fallthrough */
9254         case 0:
9255                 break;
9256         case -NFS4ERR_DELAY:
9257                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9258                         break;
9259                 goto out_restart;
9260         }
9261         dprintk("<-- %s\n", __func__);
9262         return;
9263 out_restart:
9264         task->tk_status = 0;
9265         nfs4_sequence_free_slot(&lrp->res.seq_res);
9266         rpc_restart_call_prepare(task);
9267 }
9268
9269 static void nfs4_layoutreturn_release(void *calldata)
9270 {
9271         struct nfs4_layoutreturn *lrp = calldata;
9272         struct pnfs_layout_hdr *lo = lrp->args.layout;
9273
9274         dprintk("--> %s\n", __func__);
9275         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9276                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9277         nfs4_sequence_free_slot(&lrp->res.seq_res);
9278         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9279                 lrp->ld_private.ops->free(&lrp->ld_private);
9280         pnfs_put_layout_hdr(lrp->args.layout);
9281         nfs_iput_and_deactive(lrp->inode);
9282         kfree(calldata);
9283         dprintk("<-- %s\n", __func__);
9284 }
9285
9286 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9287         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9288         .rpc_call_done = nfs4_layoutreturn_done,
9289         .rpc_release = nfs4_layoutreturn_release,
9290 };
9291
9292 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9293 {
9294         struct rpc_task *task;
9295         struct rpc_message msg = {
9296                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9297                 .rpc_argp = &lrp->args,
9298                 .rpc_resp = &lrp->res,
9299                 .rpc_cred = lrp->cred,
9300         };
9301         struct rpc_task_setup task_setup_data = {
9302                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9303                 .rpc_message = &msg,
9304                 .callback_ops = &nfs4_layoutreturn_call_ops,
9305                 .callback_data = lrp,
9306         };
9307         int status = 0;
9308
9309         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9310                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9311                         &task_setup_data.rpc_client, &msg);
9312
9313         dprintk("--> %s\n", __func__);
9314         if (!sync) {
9315                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9316                 if (!lrp->inode) {
9317                         nfs4_layoutreturn_release(lrp);
9318                         return -EAGAIN;
9319                 }
9320                 task_setup_data.flags |= RPC_TASK_ASYNC;
9321         }
9322         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9323         task = rpc_run_task(&task_setup_data);
9324         if (IS_ERR(task))
9325                 return PTR_ERR(task);
9326         if (sync)
9327                 status = task->tk_status;
9328         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9329         dprintk("<-- %s status=%d\n", __func__, status);
9330         rpc_put_task(task);
9331         return status;
9332 }
9333
9334 static int
9335 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9336                 struct pnfs_device *pdev,
9337                 const struct cred *cred)
9338 {
9339         struct nfs4_getdeviceinfo_args args = {
9340                 .pdev = pdev,
9341                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9342                         NOTIFY_DEVICEID4_DELETE,
9343         };
9344         struct nfs4_getdeviceinfo_res res = {
9345                 .pdev = pdev,
9346         };
9347         struct rpc_message msg = {
9348                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9349                 .rpc_argp = &args,
9350                 .rpc_resp = &res,
9351                 .rpc_cred = cred,
9352         };
9353         int status;
9354
9355         dprintk("--> %s\n", __func__);
9356         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9357         if (res.notification & ~args.notify_types)
9358                 dprintk("%s: unsupported notification\n", __func__);
9359         if (res.notification != args.notify_types)
9360                 pdev->nocache = 1;
9361
9362         dprintk("<-- %s status=%d\n", __func__, status);
9363
9364         return status;
9365 }
9366
9367 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9368                 struct pnfs_device *pdev,
9369                 const struct cred *cred)
9370 {
9371         struct nfs4_exception exception = { };
9372         int err;
9373
9374         do {
9375                 err = nfs4_handle_exception(server,
9376                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9377                                         &exception);
9378         } while (exception.retry);
9379         return err;
9380 }
9381 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9382
9383 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9384 {
9385         struct nfs4_layoutcommit_data *data = calldata;
9386         struct nfs_server *server = NFS_SERVER(data->args.inode);
9387
9388         nfs4_setup_sequence(server->nfs_client,
9389                         &data->args.seq_args,
9390                         &data->res.seq_res,
9391                         task);
9392 }
9393
9394 static void
9395 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9396 {
9397         struct nfs4_layoutcommit_data *data = calldata;
9398         struct nfs_server *server = NFS_SERVER(data->args.inode);
9399
9400         if (!nfs41_sequence_done(task, &data->res.seq_res))
9401                 return;
9402
9403         switch (task->tk_status) { /* Just ignore these failures */
9404         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9405         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9406         case -NFS4ERR_BADLAYOUT:     /* no layout */
9407         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9408                 task->tk_status = 0;
9409         case 0:
9410                 break;
9411         default:
9412                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9413                         rpc_restart_call_prepare(task);
9414                         return;
9415                 }
9416         }
9417 }
9418
9419 static void nfs4_layoutcommit_release(void *calldata)
9420 {
9421         struct nfs4_layoutcommit_data *data = calldata;
9422
9423         pnfs_cleanup_layoutcommit(data);
9424         nfs_post_op_update_inode_force_wcc(data->args.inode,
9425                                            data->res.fattr);
9426         put_cred(data->cred);
9427         nfs_iput_and_deactive(data->inode);
9428         kfree(data);
9429 }
9430
9431 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9432         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9433         .rpc_call_done = nfs4_layoutcommit_done,
9434         .rpc_release = nfs4_layoutcommit_release,
9435 };
9436
9437 int
9438 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9439 {
9440         struct rpc_message msg = {
9441                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9442                 .rpc_argp = &data->args,
9443                 .rpc_resp = &data->res,
9444                 .rpc_cred = data->cred,
9445         };
9446         struct rpc_task_setup task_setup_data = {
9447                 .task = &data->task,
9448                 .rpc_client = NFS_CLIENT(data->args.inode),
9449                 .rpc_message = &msg,
9450                 .callback_ops = &nfs4_layoutcommit_ops,
9451                 .callback_data = data,
9452         };
9453         struct rpc_task *task;
9454         int status = 0;
9455
9456         dprintk("NFS: initiating layoutcommit call. sync %d "
9457                 "lbw: %llu inode %lu\n", sync,
9458                 data->args.lastbytewritten,
9459                 data->args.inode->i_ino);
9460
9461         if (!sync) {
9462                 data->inode = nfs_igrab_and_active(data->args.inode);
9463                 if (data->inode == NULL) {
9464                         nfs4_layoutcommit_release(data);
9465                         return -EAGAIN;
9466                 }
9467                 task_setup_data.flags = RPC_TASK_ASYNC;
9468         }
9469         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9470         task = rpc_run_task(&task_setup_data);
9471         if (IS_ERR(task))
9472                 return PTR_ERR(task);
9473         if (sync)
9474                 status = task->tk_status;
9475         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9476         dprintk("%s: status %d\n", __func__, status);
9477         rpc_put_task(task);
9478         return status;
9479 }
9480
9481 /*
9482  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9483  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9484  */
9485 static int
9486 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9487                     struct nfs_fsinfo *info,
9488                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9489 {
9490         struct nfs41_secinfo_no_name_args args = {
9491                 .style = SECINFO_STYLE_CURRENT_FH,
9492         };
9493         struct nfs4_secinfo_res res = {
9494                 .flavors = flavors,
9495         };
9496         struct rpc_message msg = {
9497                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9498                 .rpc_argp = &args,
9499                 .rpc_resp = &res,
9500         };
9501         struct rpc_clnt *clnt = server->client;
9502         struct nfs4_call_sync_data data = {
9503                 .seq_server = server,
9504                 .seq_args = &args.seq_args,
9505                 .seq_res = &res.seq_res,
9506         };
9507         struct rpc_task_setup task_setup = {
9508                 .rpc_client = server->client,
9509                 .rpc_message = &msg,
9510                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9511                 .callback_data = &data,
9512                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9513         };
9514         const struct cred *cred = NULL;
9515         int status;
9516
9517         if (use_integrity) {
9518                 clnt = server->nfs_client->cl_rpcclient;
9519                 task_setup.rpc_client = clnt;
9520
9521                 cred = nfs4_get_clid_cred(server->nfs_client);
9522                 msg.rpc_cred = cred;
9523         }
9524
9525         dprintk("--> %s\n", __func__);
9526         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9527         status = nfs4_call_sync_custom(&task_setup);
9528         dprintk("<-- %s status=%d\n", __func__, status);
9529
9530         put_cred(cred);
9531
9532         return status;
9533 }
9534
9535 static int
9536 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9537                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9538 {
9539         struct nfs4_exception exception = {
9540                 .interruptible = true,
9541         };
9542         int err;
9543         do {
9544                 /* first try using integrity protection */
9545                 err = -NFS4ERR_WRONGSEC;
9546
9547                 /* try to use integrity protection with machine cred */
9548                 if (_nfs4_is_integrity_protected(server->nfs_client))
9549                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9550                                                           flavors, true);
9551
9552                 /*
9553                  * if unable to use integrity protection, or SECINFO with
9554                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9555                  * disallowed by spec, but exists in deployed servers) use
9556                  * the current filesystem's rpc_client and the user cred.
9557                  */
9558                 if (err == -NFS4ERR_WRONGSEC)
9559                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9560                                                           flavors, false);
9561
9562                 switch (err) {
9563                 case 0:
9564                 case -NFS4ERR_WRONGSEC:
9565                 case -ENOTSUPP:
9566                         goto out;
9567                 default:
9568                         err = nfs4_handle_exception(server, err, &exception);
9569                 }
9570         } while (exception.retry);
9571 out:
9572         return err;
9573 }
9574
9575 static int
9576 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9577                     struct nfs_fsinfo *info)
9578 {
9579         int err;
9580         struct page *page;
9581         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9582         struct nfs4_secinfo_flavors *flavors;
9583         struct nfs4_secinfo4 *secinfo;
9584         int i;
9585
9586         page = alloc_page(GFP_KERNEL);
9587         if (!page) {
9588                 err = -ENOMEM;
9589                 goto out;
9590         }
9591
9592         flavors = page_address(page);
9593         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9594
9595         /*
9596          * Fall back on "guess and check" method if
9597          * the server doesn't support SECINFO_NO_NAME
9598          */
9599         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9600                 err = nfs4_find_root_sec(server, fhandle, info);
9601                 goto out_freepage;
9602         }
9603         if (err)
9604                 goto out_freepage;
9605
9606         for (i = 0; i < flavors->num_flavors; i++) {
9607                 secinfo = &flavors->flavors[i];
9608
9609                 switch (secinfo->flavor) {
9610                 case RPC_AUTH_NULL:
9611                 case RPC_AUTH_UNIX:
9612                 case RPC_AUTH_GSS:
9613                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9614                                         &secinfo->flavor_info);
9615                         break;
9616                 default:
9617                         flavor = RPC_AUTH_MAXFLAVOR;
9618                         break;
9619                 }
9620
9621                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9622                         flavor = RPC_AUTH_MAXFLAVOR;
9623
9624                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9625                         err = nfs4_lookup_root_sec(server, fhandle,
9626                                                    info, flavor);
9627                         if (!err)
9628                                 break;
9629                 }
9630         }
9631
9632         if (flavor == RPC_AUTH_MAXFLAVOR)
9633                 err = -EPERM;
9634
9635 out_freepage:
9636         put_page(page);
9637         if (err == -EACCES)
9638                 return -EPERM;
9639 out:
9640         return err;
9641 }
9642
9643 static int _nfs41_test_stateid(struct nfs_server *server,
9644                 nfs4_stateid *stateid,
9645                 const struct cred *cred)
9646 {
9647         int status;
9648         struct nfs41_test_stateid_args args = {
9649                 .stateid = stateid,
9650         };
9651         struct nfs41_test_stateid_res res;
9652         struct rpc_message msg = {
9653                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9654                 .rpc_argp = &args,
9655                 .rpc_resp = &res,
9656                 .rpc_cred = cred,
9657         };
9658         struct rpc_clnt *rpc_client = server->client;
9659
9660         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9661                 &rpc_client, &msg);
9662
9663         dprintk("NFS call  test_stateid %p\n", stateid);
9664         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9665         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9666                         &args.seq_args, &res.seq_res);
9667         if (status != NFS_OK) {
9668                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9669                 return status;
9670         }
9671         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9672         return -res.status;
9673 }
9674
9675 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9676                 int err, struct nfs4_exception *exception)
9677 {
9678         exception->retry = 0;
9679         switch(err) {
9680         case -NFS4ERR_DELAY:
9681         case -NFS4ERR_RETRY_UNCACHED_REP:
9682                 nfs4_handle_exception(server, err, exception);
9683                 break;
9684         case -NFS4ERR_BADSESSION:
9685         case -NFS4ERR_BADSLOT:
9686         case -NFS4ERR_BAD_HIGH_SLOT:
9687         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9688         case -NFS4ERR_DEADSESSION:
9689                 nfs4_do_handle_exception(server, err, exception);
9690         }
9691 }
9692
9693 /**
9694  * nfs41_test_stateid - perform a TEST_STATEID operation
9695  *
9696  * @server: server / transport on which to perform the operation
9697  * @stateid: state ID to test
9698  * @cred: credential
9699  *
9700  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9701  * Otherwise a negative NFS4ERR value is returned if the operation
9702  * failed or the state ID is not currently valid.
9703  */
9704 static int nfs41_test_stateid(struct nfs_server *server,
9705                 nfs4_stateid *stateid,
9706                 const struct cred *cred)
9707 {
9708         struct nfs4_exception exception = {
9709                 .interruptible = true,
9710         };
9711         int err;
9712         do {
9713                 err = _nfs41_test_stateid(server, stateid, cred);
9714                 nfs4_handle_delay_or_session_error(server, err, &exception);
9715         } while (exception.retry);
9716         return err;
9717 }
9718
9719 struct nfs_free_stateid_data {
9720         struct nfs_server *server;
9721         struct nfs41_free_stateid_args args;
9722         struct nfs41_free_stateid_res res;
9723 };
9724
9725 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9726 {
9727         struct nfs_free_stateid_data *data = calldata;
9728         nfs4_setup_sequence(data->server->nfs_client,
9729                         &data->args.seq_args,
9730                         &data->res.seq_res,
9731                         task);
9732 }
9733
9734 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9735 {
9736         struct nfs_free_stateid_data *data = calldata;
9737
9738         nfs41_sequence_done(task, &data->res.seq_res);
9739
9740         switch (task->tk_status) {
9741         case -NFS4ERR_DELAY:
9742                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9743                         rpc_restart_call_prepare(task);
9744         }
9745 }
9746
9747 static void nfs41_free_stateid_release(void *calldata)
9748 {
9749         kfree(calldata);
9750 }
9751
9752 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9753         .rpc_call_prepare = nfs41_free_stateid_prepare,
9754         .rpc_call_done = nfs41_free_stateid_done,
9755         .rpc_release = nfs41_free_stateid_release,
9756 };
9757
9758 /**
9759  * nfs41_free_stateid - perform a FREE_STATEID operation
9760  *
9761  * @server: server / transport on which to perform the operation
9762  * @stateid: state ID to release
9763  * @cred: credential
9764  * @privileged: set to true if this call needs to be privileged
9765  *
9766  * Note: this function is always asynchronous.
9767  */
9768 static int nfs41_free_stateid(struct nfs_server *server,
9769                 const nfs4_stateid *stateid,
9770                 const struct cred *cred,
9771                 bool privileged)
9772 {
9773         struct rpc_message msg = {
9774                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9775                 .rpc_cred = cred,
9776         };
9777         struct rpc_task_setup task_setup = {
9778                 .rpc_client = server->client,
9779                 .rpc_message = &msg,
9780                 .callback_ops = &nfs41_free_stateid_ops,
9781                 .flags = RPC_TASK_ASYNC,
9782         };
9783         struct nfs_free_stateid_data *data;
9784         struct rpc_task *task;
9785
9786         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9787                 &task_setup.rpc_client, &msg);
9788
9789         dprintk("NFS call  free_stateid %p\n", stateid);
9790         data = kmalloc(sizeof(*data), GFP_NOFS);
9791         if (!data)
9792                 return -ENOMEM;
9793         data->server = server;
9794         nfs4_stateid_copy(&data->args.stateid, stateid);
9795
9796         task_setup.callback_data = data;
9797
9798         msg.rpc_argp = &data->args;
9799         msg.rpc_resp = &data->res;
9800         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9801         task = rpc_run_task(&task_setup);
9802         if (IS_ERR(task))
9803                 return PTR_ERR(task);
9804         rpc_put_task(task);
9805         return 0;
9806 }
9807
9808 static void
9809 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9810 {
9811         const struct cred *cred = lsp->ls_state->owner->so_cred;
9812
9813         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9814         nfs4_free_lock_state(server, lsp);
9815 }
9816
9817 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9818                 const nfs4_stateid *s2)
9819 {
9820         if (s1->type != s2->type)
9821                 return false;
9822
9823         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9824                 return false;
9825
9826         if (s1->seqid == s2->seqid)
9827                 return true;
9828
9829         return s1->seqid == 0 || s2->seqid == 0;
9830 }
9831
9832 #endif /* CONFIG_NFS_V4_1 */
9833
9834 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9835                 const nfs4_stateid *s2)
9836 {
9837         return nfs4_stateid_match(s1, s2);
9838 }
9839
9840
9841 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9842         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9843         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9844         .recover_open   = nfs4_open_reclaim,
9845         .recover_lock   = nfs4_lock_reclaim,
9846         .establish_clid = nfs4_init_clientid,
9847         .detect_trunking = nfs40_discover_server_trunking,
9848 };
9849
9850 #if defined(CONFIG_NFS_V4_1)
9851 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9852         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9853         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9854         .recover_open   = nfs4_open_reclaim,
9855         .recover_lock   = nfs4_lock_reclaim,
9856         .establish_clid = nfs41_init_clientid,
9857         .reclaim_complete = nfs41_proc_reclaim_complete,
9858         .detect_trunking = nfs41_discover_server_trunking,
9859 };
9860 #endif /* CONFIG_NFS_V4_1 */
9861
9862 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9863         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9864         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9865         .recover_open   = nfs40_open_expired,
9866         .recover_lock   = nfs4_lock_expired,
9867         .establish_clid = nfs4_init_clientid,
9868 };
9869
9870 #if defined(CONFIG_NFS_V4_1)
9871 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9872         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9873         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9874         .recover_open   = nfs41_open_expired,
9875         .recover_lock   = nfs41_lock_expired,
9876         .establish_clid = nfs41_init_clientid,
9877 };
9878 #endif /* CONFIG_NFS_V4_1 */
9879
9880 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9881         .sched_state_renewal = nfs4_proc_async_renew,
9882         .get_state_renewal_cred = nfs4_get_renew_cred,
9883         .renew_lease = nfs4_proc_renew,
9884 };
9885
9886 #if defined(CONFIG_NFS_V4_1)
9887 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9888         .sched_state_renewal = nfs41_proc_async_sequence,
9889         .get_state_renewal_cred = nfs4_get_machine_cred,
9890         .renew_lease = nfs4_proc_sequence,
9891 };
9892 #endif
9893
9894 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9895         .get_locations = _nfs40_proc_get_locations,
9896         .fsid_present = _nfs40_proc_fsid_present,
9897 };
9898
9899 #if defined(CONFIG_NFS_V4_1)
9900 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9901         .get_locations = _nfs41_proc_get_locations,
9902         .fsid_present = _nfs41_proc_fsid_present,
9903 };
9904 #endif  /* CONFIG_NFS_V4_1 */
9905
9906 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9907         .minor_version = 0,
9908         .init_caps = NFS_CAP_READDIRPLUS
9909                 | NFS_CAP_ATOMIC_OPEN
9910                 | NFS_CAP_POSIX_LOCK,
9911         .init_client = nfs40_init_client,
9912         .shutdown_client = nfs40_shutdown_client,
9913         .match_stateid = nfs4_match_stateid,
9914         .find_root_sec = nfs4_find_root_sec,
9915         .free_lock_state = nfs4_release_lockowner,
9916         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9917         .alloc_seqid = nfs_alloc_seqid,
9918         .call_sync_ops = &nfs40_call_sync_ops,
9919         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9920         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9921         .state_renewal_ops = &nfs40_state_renewal_ops,
9922         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9923 };
9924
9925 #if defined(CONFIG_NFS_V4_1)
9926 static struct nfs_seqid *
9927 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9928 {
9929         return NULL;
9930 }
9931
9932 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9933         .minor_version = 1,
9934         .init_caps = NFS_CAP_READDIRPLUS
9935                 | NFS_CAP_ATOMIC_OPEN
9936                 | NFS_CAP_POSIX_LOCK
9937                 | NFS_CAP_STATEID_NFSV41
9938                 | NFS_CAP_ATOMIC_OPEN_V1
9939                 | NFS_CAP_LGOPEN,
9940         .init_client = nfs41_init_client,
9941         .shutdown_client = nfs41_shutdown_client,
9942         .match_stateid = nfs41_match_stateid,
9943         .find_root_sec = nfs41_find_root_sec,
9944         .free_lock_state = nfs41_free_lock_state,
9945         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9946         .alloc_seqid = nfs_alloc_no_seqid,
9947         .session_trunk = nfs4_test_session_trunk,
9948         .call_sync_ops = &nfs41_call_sync_ops,
9949         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9950         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9951         .state_renewal_ops = &nfs41_state_renewal_ops,
9952         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9953 };
9954 #endif
9955
9956 #if defined(CONFIG_NFS_V4_2)
9957 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9958         .minor_version = 2,
9959         .init_caps = NFS_CAP_READDIRPLUS
9960                 | NFS_CAP_ATOMIC_OPEN
9961                 | NFS_CAP_POSIX_LOCK
9962                 | NFS_CAP_STATEID_NFSV41
9963                 | NFS_CAP_ATOMIC_OPEN_V1
9964                 | NFS_CAP_LGOPEN
9965                 | NFS_CAP_ALLOCATE
9966                 | NFS_CAP_COPY
9967                 | NFS_CAP_OFFLOAD_CANCEL
9968                 | NFS_CAP_COPY_NOTIFY
9969                 | NFS_CAP_DEALLOCATE
9970                 | NFS_CAP_SEEK
9971                 | NFS_CAP_LAYOUTSTATS
9972                 | NFS_CAP_CLONE
9973                 | NFS_CAP_LAYOUTERROR,
9974         .init_client = nfs41_init_client,
9975         .shutdown_client = nfs41_shutdown_client,
9976         .match_stateid = nfs41_match_stateid,
9977         .find_root_sec = nfs41_find_root_sec,
9978         .free_lock_state = nfs41_free_lock_state,
9979         .call_sync_ops = &nfs41_call_sync_ops,
9980         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9981         .alloc_seqid = nfs_alloc_no_seqid,
9982         .session_trunk = nfs4_test_session_trunk,
9983         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9984         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9985         .state_renewal_ops = &nfs41_state_renewal_ops,
9986         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9987 };
9988 #endif
9989
9990 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9991         [0] = &nfs_v4_0_minor_ops,
9992 #if defined(CONFIG_NFS_V4_1)
9993         [1] = &nfs_v4_1_minor_ops,
9994 #endif
9995 #if defined(CONFIG_NFS_V4_2)
9996         [2] = &nfs_v4_2_minor_ops,
9997 #endif
9998 };
9999
10000 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10001 {
10002         ssize_t error, error2;
10003
10004         error = generic_listxattr(dentry, list, size);
10005         if (error < 0)
10006                 return error;
10007         if (list) {
10008                 list += error;
10009                 size -= error;
10010         }
10011
10012         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10013         if (error2 < 0)
10014                 return error2;
10015         return error + error2;
10016 }
10017
10018 static const struct inode_operations nfs4_dir_inode_operations = {
10019         .create         = nfs_create,
10020         .lookup         = nfs_lookup,
10021         .atomic_open    = nfs_atomic_open,
10022         .link           = nfs_link,
10023         .unlink         = nfs_unlink,
10024         .symlink        = nfs_symlink,
10025         .mkdir          = nfs_mkdir,
10026         .rmdir          = nfs_rmdir,
10027         .mknod          = nfs_mknod,
10028         .rename         = nfs_rename,
10029         .permission     = nfs_permission,
10030         .getattr        = nfs_getattr,
10031         .setattr        = nfs_setattr,
10032         .listxattr      = nfs4_listxattr,
10033 };
10034
10035 static const struct inode_operations nfs4_file_inode_operations = {
10036         .permission     = nfs_permission,
10037         .getattr        = nfs_getattr,
10038         .setattr        = nfs_setattr,
10039         .listxattr      = nfs4_listxattr,
10040 };
10041
10042 const struct nfs_rpc_ops nfs_v4_clientops = {
10043         .version        = 4,                    /* protocol version */
10044         .dentry_ops     = &nfs4_dentry_operations,
10045         .dir_inode_ops  = &nfs4_dir_inode_operations,
10046         .file_inode_ops = &nfs4_file_inode_operations,
10047         .file_ops       = &nfs4_file_operations,
10048         .getroot        = nfs4_proc_get_root,
10049         .submount       = nfs4_submount,
10050         .try_get_tree   = nfs4_try_get_tree,
10051         .getattr        = nfs4_proc_getattr,
10052         .setattr        = nfs4_proc_setattr,
10053         .lookup         = nfs4_proc_lookup,
10054         .lookupp        = nfs4_proc_lookupp,
10055         .access         = nfs4_proc_access,
10056         .readlink       = nfs4_proc_readlink,
10057         .create         = nfs4_proc_create,
10058         .remove         = nfs4_proc_remove,
10059         .unlink_setup   = nfs4_proc_unlink_setup,
10060         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10061         .unlink_done    = nfs4_proc_unlink_done,
10062         .rename_setup   = nfs4_proc_rename_setup,
10063         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10064         .rename_done    = nfs4_proc_rename_done,
10065         .link           = nfs4_proc_link,
10066         .symlink        = nfs4_proc_symlink,
10067         .mkdir          = nfs4_proc_mkdir,
10068         .rmdir          = nfs4_proc_rmdir,
10069         .readdir        = nfs4_proc_readdir,
10070         .mknod          = nfs4_proc_mknod,
10071         .statfs         = nfs4_proc_statfs,
10072         .fsinfo         = nfs4_proc_fsinfo,
10073         .pathconf       = nfs4_proc_pathconf,
10074         .set_capabilities = nfs4_server_capabilities,
10075         .decode_dirent  = nfs4_decode_dirent,
10076         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10077         .read_setup     = nfs4_proc_read_setup,
10078         .read_done      = nfs4_read_done,
10079         .write_setup    = nfs4_proc_write_setup,
10080         .write_done     = nfs4_write_done,
10081         .commit_setup   = nfs4_proc_commit_setup,
10082         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10083         .commit_done    = nfs4_commit_done,
10084         .lock           = nfs4_proc_lock,
10085         .clear_acl_cache = nfs4_zap_acl_attr,
10086         .close_context  = nfs4_close_context,
10087         .open_context   = nfs4_atomic_open,
10088         .have_delegation = nfs4_have_delegation,
10089         .alloc_client   = nfs4_alloc_client,
10090         .init_client    = nfs4_init_client,
10091         .free_client    = nfs4_free_client,
10092         .create_server  = nfs4_create_server,
10093         .clone_server   = nfs_clone_server,
10094 };
10095
10096 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10097         .name   = XATTR_NAME_NFSV4_ACL,
10098         .list   = nfs4_xattr_list_nfs4_acl,
10099         .get    = nfs4_xattr_get_nfs4_acl,
10100         .set    = nfs4_xattr_set_nfs4_acl,
10101 };
10102
10103 const struct xattr_handler *nfs4_xattr_handlers[] = {
10104         &nfs4_xattr_nfs4_acl_handler,
10105 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10106         &nfs4_xattr_nfs4_label_handler,
10107 #endif
10108         NULL
10109 };
10110
10111 /*
10112  * Local variables:
10113  *  c-basic-offset: 8
10114  * End:
10115  */