9e0ca9b2b210b01e07e40f535b28703b1eaf9ff4
[linux-2.6-microblaze.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57 #include <linux/iversion.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "sysfs.h"
67 #include "nfs4idmap.h"
68 #include "nfs4session.h"
69 #include "fscache.h"
70 #include "nfs42.h"
71
72 #include "nfs4trace.h"
73
74 #ifdef CONFIG_NFS_V4_2
75 #include "nfs42.h"
76 #endif /* CONFIG_NFS_V4_2 */
77
78 #define NFSDBG_FACILITY         NFSDBG_PROC
79
80 #define NFS4_BITMASK_SZ         3
81
82 #define NFS4_POLL_RETRY_MIN     (HZ/10)
83 #define NFS4_POLL_RETRY_MAX     (15*HZ)
84
85 /* file attributes which can be mapped to nfs attributes */
86 #define NFS4_VALID_ATTRS (ATTR_MODE \
87         | ATTR_UID \
88         | ATTR_GID \
89         | ATTR_SIZE \
90         | ATTR_ATIME \
91         | ATTR_MTIME \
92         | ATTR_CTIME \
93         | ATTR_ATIME_SET \
94         | ATTR_MTIME_SET)
95
96 struct nfs4_opendata;
97 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
98 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
99 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
100 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
101 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
102                             struct nfs_fattr *fattr, struct iattr *sattr,
103                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
104                             struct nfs4_label *olabel);
105 #ifdef CONFIG_NFS_V4_1
106 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
107                 const struct cred *cred,
108                 struct nfs4_slot *slot,
109                 bool is_privileged);
110 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
111                 const struct cred *);
112 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
113                 const struct cred *, bool);
114 #endif
115 static void nfs4_bitmask_adjust(__u32 *bitmask, struct inode *inode,
116                 struct nfs_server *server,
117                 struct nfs4_label *label);
118
119 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
120 static inline struct nfs4_label *
121 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
122         struct iattr *sattr, struct nfs4_label *label)
123 {
124         int err;
125
126         if (label == NULL)
127                 return NULL;
128
129         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
130                 return NULL;
131
132         err = security_dentry_init_security(dentry, sattr->ia_mode,
133                                 &dentry->d_name, (void **)&label->label, &label->len);
134         if (err == 0)
135                 return label;
136
137         return NULL;
138 }
139 static inline void
140 nfs4_label_release_security(struct nfs4_label *label)
141 {
142         if (label)
143                 security_release_secctx(label->label, label->len);
144 }
145 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
146 {
147         if (label)
148                 return server->attr_bitmask;
149
150         return server->attr_bitmask_nl;
151 }
152 #else
153 static inline struct nfs4_label *
154 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
155         struct iattr *sattr, struct nfs4_label *l)
156 { return NULL; }
157 static inline void
158 nfs4_label_release_security(struct nfs4_label *label)
159 { return; }
160 static inline u32 *
161 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
162 { return server->attr_bitmask; }
163 #endif
164
165 /* Prevent leaks of NFSv4 errors into userland */
166 static int nfs4_map_errors(int err)
167 {
168         if (err >= -1000)
169                 return err;
170         switch (err) {
171         case -NFS4ERR_RESOURCE:
172         case -NFS4ERR_LAYOUTTRYLATER:
173         case -NFS4ERR_RECALLCONFLICT:
174                 return -EREMOTEIO;
175         case -NFS4ERR_WRONGSEC:
176         case -NFS4ERR_WRONG_CRED:
177                 return -EPERM;
178         case -NFS4ERR_BADOWNER:
179         case -NFS4ERR_BADNAME:
180                 return -EINVAL;
181         case -NFS4ERR_SHARE_DENIED:
182                 return -EACCES;
183         case -NFS4ERR_MINOR_VERS_MISMATCH:
184                 return -EPROTONOSUPPORT;
185         case -NFS4ERR_FILE_OPEN:
186                 return -EBUSY;
187         default:
188                 dprintk("%s could not handle NFSv4 error %d\n",
189                                 __func__, -err);
190                 break;
191         }
192         return -EIO;
193 }
194
195 /*
196  * This is our standard bitmap for GETATTR requests.
197  */
198 const u32 nfs4_fattr_bitmap[3] = {
199         FATTR4_WORD0_TYPE
200         | FATTR4_WORD0_CHANGE
201         | FATTR4_WORD0_SIZE
202         | FATTR4_WORD0_FSID
203         | FATTR4_WORD0_FILEID,
204         FATTR4_WORD1_MODE
205         | FATTR4_WORD1_NUMLINKS
206         | FATTR4_WORD1_OWNER
207         | FATTR4_WORD1_OWNER_GROUP
208         | FATTR4_WORD1_RAWDEV
209         | FATTR4_WORD1_SPACE_USED
210         | FATTR4_WORD1_TIME_ACCESS
211         | FATTR4_WORD1_TIME_METADATA
212         | FATTR4_WORD1_TIME_MODIFY
213         | FATTR4_WORD1_MOUNTED_ON_FILEID,
214 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
215         FATTR4_WORD2_SECURITY_LABEL
216 #endif
217 };
218
219 static const u32 nfs4_pnfs_open_bitmap[3] = {
220         FATTR4_WORD0_TYPE
221         | FATTR4_WORD0_CHANGE
222         | FATTR4_WORD0_SIZE
223         | FATTR4_WORD0_FSID
224         | FATTR4_WORD0_FILEID,
225         FATTR4_WORD1_MODE
226         | FATTR4_WORD1_NUMLINKS
227         | FATTR4_WORD1_OWNER
228         | FATTR4_WORD1_OWNER_GROUP
229         | FATTR4_WORD1_RAWDEV
230         | FATTR4_WORD1_SPACE_USED
231         | FATTR4_WORD1_TIME_ACCESS
232         | FATTR4_WORD1_TIME_METADATA
233         | FATTR4_WORD1_TIME_MODIFY,
234         FATTR4_WORD2_MDSTHRESHOLD
235 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
236         | FATTR4_WORD2_SECURITY_LABEL
237 #endif
238 };
239
240 static const u32 nfs4_open_noattr_bitmap[3] = {
241         FATTR4_WORD0_TYPE
242         | FATTR4_WORD0_FILEID,
243 };
244
245 const u32 nfs4_statfs_bitmap[3] = {
246         FATTR4_WORD0_FILES_AVAIL
247         | FATTR4_WORD0_FILES_FREE
248         | FATTR4_WORD0_FILES_TOTAL,
249         FATTR4_WORD1_SPACE_AVAIL
250         | FATTR4_WORD1_SPACE_FREE
251         | FATTR4_WORD1_SPACE_TOTAL
252 };
253
254 const u32 nfs4_pathconf_bitmap[3] = {
255         FATTR4_WORD0_MAXLINK
256         | FATTR4_WORD0_MAXNAME,
257         0
258 };
259
260 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
261                         | FATTR4_WORD0_MAXREAD
262                         | FATTR4_WORD0_MAXWRITE
263                         | FATTR4_WORD0_LEASE_TIME,
264                         FATTR4_WORD1_TIME_DELTA
265                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
266                         FATTR4_WORD2_LAYOUT_BLKSIZE
267                         | FATTR4_WORD2_CLONE_BLKSIZE
268                         | FATTR4_WORD2_XATTR_SUPPORT
269 };
270
271 const u32 nfs4_fs_locations_bitmap[3] = {
272         FATTR4_WORD0_CHANGE
273         | FATTR4_WORD0_SIZE
274         | FATTR4_WORD0_FSID
275         | FATTR4_WORD0_FILEID
276         | FATTR4_WORD0_FS_LOCATIONS,
277         FATTR4_WORD1_OWNER
278         | FATTR4_WORD1_OWNER_GROUP
279         | FATTR4_WORD1_RAWDEV
280         | FATTR4_WORD1_SPACE_USED
281         | FATTR4_WORD1_TIME_ACCESS
282         | FATTR4_WORD1_TIME_METADATA
283         | FATTR4_WORD1_TIME_MODIFY
284         | FATTR4_WORD1_MOUNTED_ON_FILEID,
285 };
286
287 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
288                 struct inode *inode)
289 {
290         unsigned long cache_validity;
291
292         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
293         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
294                 return;
295
296         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
297         if (!(cache_validity & NFS_INO_REVAL_FORCED))
298                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
299                                 | NFS_INO_INVALID_SIZE);
300
301         if (!(cache_validity & NFS_INO_INVALID_SIZE))
302                 dst[0] &= ~FATTR4_WORD0_SIZE;
303
304         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
305                 dst[0] &= ~FATTR4_WORD0_CHANGE;
306 }
307
308 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
309                 const __u32 *src, struct inode *inode)
310 {
311         nfs4_bitmap_copy_adjust(dst, src, inode);
312 }
313
314 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
315                 struct nfs4_readdir_arg *readdir)
316 {
317         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
318         __be32 *start, *p;
319
320         if (cookie > 2) {
321                 readdir->cookie = cookie;
322                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
323                 return;
324         }
325
326         readdir->cookie = 0;
327         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
328         if (cookie == 2)
329                 return;
330         
331         /*
332          * NFSv4 servers do not return entries for '.' and '..'
333          * Therefore, we fake these entries here.  We let '.'
334          * have cookie 0 and '..' have cookie 1.  Note that
335          * when talking to the server, we always send cookie 0
336          * instead of 1 or 2.
337          */
338         start = p = kmap_atomic(*readdir->pages);
339         
340         if (cookie == 0) {
341                 *p++ = xdr_one;                                  /* next */
342                 *p++ = xdr_zero;                   /* cookie, first word */
343                 *p++ = xdr_one;                   /* cookie, second word */
344                 *p++ = xdr_one;                             /* entry len */
345                 memcpy(p, ".\0\0\0", 4);                        /* entry */
346                 p++;
347                 *p++ = xdr_one;                         /* bitmap length */
348                 *p++ = htonl(attrs);                           /* bitmap */
349                 *p++ = htonl(12);             /* attribute buffer length */
350                 *p++ = htonl(NF4DIR);
351                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
352         }
353         
354         *p++ = xdr_one;                                  /* next */
355         *p++ = xdr_zero;                   /* cookie, first word */
356         *p++ = xdr_two;                   /* cookie, second word */
357         *p++ = xdr_two;                             /* entry len */
358         memcpy(p, "..\0\0", 4);                         /* entry */
359         p++;
360         *p++ = xdr_one;                         /* bitmap length */
361         *p++ = htonl(attrs);                           /* bitmap */
362         *p++ = htonl(12);             /* attribute buffer length */
363         *p++ = htonl(NF4DIR);
364         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
365
366         readdir->pgbase = (char *)p - (char *)start;
367         readdir->count -= readdir->pgbase;
368         kunmap_atomic(start);
369 }
370
371 static void nfs4_test_and_free_stateid(struct nfs_server *server,
372                 nfs4_stateid *stateid,
373                 const struct cred *cred)
374 {
375         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
376
377         ops->test_and_free_expired(server, stateid, cred);
378 }
379
380 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
381                 nfs4_stateid *stateid,
382                 const struct cred *cred)
383 {
384         stateid->type = NFS4_REVOKED_STATEID_TYPE;
385         nfs4_test_and_free_stateid(server, stateid, cred);
386 }
387
388 static void nfs4_free_revoked_stateid(struct nfs_server *server,
389                 const nfs4_stateid *stateid,
390                 const struct cred *cred)
391 {
392         nfs4_stateid tmp;
393
394         nfs4_stateid_copy(&tmp, stateid);
395         __nfs4_free_revoked_stateid(server, &tmp, cred);
396 }
397
398 static long nfs4_update_delay(long *timeout)
399 {
400         long ret;
401         if (!timeout)
402                 return NFS4_POLL_RETRY_MAX;
403         if (*timeout <= 0)
404                 *timeout = NFS4_POLL_RETRY_MIN;
405         if (*timeout > NFS4_POLL_RETRY_MAX)
406                 *timeout = NFS4_POLL_RETRY_MAX;
407         ret = *timeout;
408         *timeout <<= 1;
409         return ret;
410 }
411
412 static int nfs4_delay_killable(long *timeout)
413 {
414         might_sleep();
415
416         freezable_schedule_timeout_killable_unsafe(
417                 nfs4_update_delay(timeout));
418         if (!__fatal_signal_pending(current))
419                 return 0;
420         return -EINTR;
421 }
422
423 static int nfs4_delay_interruptible(long *timeout)
424 {
425         might_sleep();
426
427         freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
428         if (!signal_pending(current))
429                 return 0;
430         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
431 }
432
433 static int nfs4_delay(long *timeout, bool interruptible)
434 {
435         if (interruptible)
436                 return nfs4_delay_interruptible(timeout);
437         return nfs4_delay_killable(timeout);
438 }
439
440 static const nfs4_stateid *
441 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
442 {
443         if (!stateid)
444                 return NULL;
445         switch (stateid->type) {
446         case NFS4_OPEN_STATEID_TYPE:
447         case NFS4_LOCK_STATEID_TYPE:
448         case NFS4_DELEGATION_STATEID_TYPE:
449                 return stateid;
450         default:
451                 break;
452         }
453         return NULL;
454 }
455
456 /* This is the error handling routine for processes that are allowed
457  * to sleep.
458  */
459 static int nfs4_do_handle_exception(struct nfs_server *server,
460                 int errorcode, struct nfs4_exception *exception)
461 {
462         struct nfs_client *clp = server->nfs_client;
463         struct nfs4_state *state = exception->state;
464         const nfs4_stateid *stateid;
465         struct inode *inode = exception->inode;
466         int ret = errorcode;
467
468         exception->delay = 0;
469         exception->recovering = 0;
470         exception->retry = 0;
471
472         stateid = nfs4_recoverable_stateid(exception->stateid);
473         if (stateid == NULL && state != NULL)
474                 stateid = nfs4_recoverable_stateid(&state->stateid);
475
476         switch(errorcode) {
477                 case 0:
478                         return 0;
479                 case -NFS4ERR_BADHANDLE:
480                 case -ESTALE:
481                         if (inode != NULL && S_ISREG(inode->i_mode))
482                                 pnfs_destroy_layout(NFS_I(inode));
483                         break;
484                 case -NFS4ERR_DELEG_REVOKED:
485                 case -NFS4ERR_ADMIN_REVOKED:
486                 case -NFS4ERR_EXPIRED:
487                 case -NFS4ERR_BAD_STATEID:
488                 case -NFS4ERR_PARTNER_NO_AUTH:
489                         if (inode != NULL && stateid != NULL) {
490                                 nfs_inode_find_state_and_recover(inode,
491                                                 stateid);
492                                 goto wait_on_recovery;
493                         }
494                         fallthrough;
495                 case -NFS4ERR_OPENMODE:
496                         if (inode) {
497                                 int err;
498
499                                 err = nfs_async_inode_return_delegation(inode,
500                                                 stateid);
501                                 if (err == 0)
502                                         goto wait_on_recovery;
503                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
504                                         exception->retry = 1;
505                                         break;
506                                 }
507                         }
508                         if (state == NULL)
509                                 break;
510                         ret = nfs4_schedule_stateid_recovery(server, state);
511                         if (ret < 0)
512                                 break;
513                         goto wait_on_recovery;
514                 case -NFS4ERR_STALE_STATEID:
515                 case -NFS4ERR_STALE_CLIENTID:
516                         nfs4_schedule_lease_recovery(clp);
517                         goto wait_on_recovery;
518                 case -NFS4ERR_MOVED:
519                         ret = nfs4_schedule_migration_recovery(server);
520                         if (ret < 0)
521                                 break;
522                         goto wait_on_recovery;
523                 case -NFS4ERR_LEASE_MOVED:
524                         nfs4_schedule_lease_moved_recovery(clp);
525                         goto wait_on_recovery;
526 #if defined(CONFIG_NFS_V4_1)
527                 case -NFS4ERR_BADSESSION:
528                 case -NFS4ERR_BADSLOT:
529                 case -NFS4ERR_BAD_HIGH_SLOT:
530                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
531                 case -NFS4ERR_DEADSESSION:
532                 case -NFS4ERR_SEQ_FALSE_RETRY:
533                 case -NFS4ERR_SEQ_MISORDERED:
534                         /* Handled in nfs41_sequence_process() */
535                         goto wait_on_recovery;
536 #endif /* defined(CONFIG_NFS_V4_1) */
537                 case -NFS4ERR_FILE_OPEN:
538                         if (exception->timeout > HZ) {
539                                 /* We have retried a decent amount, time to
540                                  * fail
541                                  */
542                                 ret = -EBUSY;
543                                 break;
544                         }
545                         fallthrough;
546                 case -NFS4ERR_DELAY:
547                         nfs_inc_server_stats(server, NFSIOS_DELAY);
548                         fallthrough;
549                 case -NFS4ERR_GRACE:
550                 case -NFS4ERR_LAYOUTTRYLATER:
551                 case -NFS4ERR_RECALLCONFLICT:
552                         exception->delay = 1;
553                         return 0;
554
555                 case -NFS4ERR_RETRY_UNCACHED_REP:
556                 case -NFS4ERR_OLD_STATEID:
557                         exception->retry = 1;
558                         break;
559                 case -NFS4ERR_BADOWNER:
560                         /* The following works around a Linux server bug! */
561                 case -NFS4ERR_BADNAME:
562                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
563                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
564                                 exception->retry = 1;
565                                 printk(KERN_WARNING "NFS: v4 server %s "
566                                                 "does not accept raw "
567                                                 "uid/gids. "
568                                                 "Reenabling the idmapper.\n",
569                                                 server->nfs_client->cl_hostname);
570                         }
571         }
572         /* We failed to handle the error */
573         return nfs4_map_errors(ret);
574 wait_on_recovery:
575         exception->recovering = 1;
576         return 0;
577 }
578
579 /* This is the error handling routine for processes that are allowed
580  * to sleep.
581  */
582 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
583 {
584         struct nfs_client *clp = server->nfs_client;
585         int ret;
586
587         ret = nfs4_do_handle_exception(server, errorcode, exception);
588         if (exception->delay) {
589                 ret = nfs4_delay(&exception->timeout,
590                                 exception->interruptible);
591                 goto out_retry;
592         }
593         if (exception->recovering) {
594                 ret = nfs4_wait_clnt_recover(clp);
595                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
596                         return -EIO;
597                 goto out_retry;
598         }
599         return ret;
600 out_retry:
601         if (ret == 0)
602                 exception->retry = 1;
603         return ret;
604 }
605
606 static int
607 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
608                 int errorcode, struct nfs4_exception *exception)
609 {
610         struct nfs_client *clp = server->nfs_client;
611         int ret;
612
613         ret = nfs4_do_handle_exception(server, errorcode, exception);
614         if (exception->delay) {
615                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
616                 goto out_retry;
617         }
618         if (exception->recovering) {
619                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
620                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
621                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
622                 goto out_retry;
623         }
624         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
625                 ret = -EIO;
626         return ret;
627 out_retry:
628         if (ret == 0) {
629                 exception->retry = 1;
630                 /*
631                  * For NFS4ERR_MOVED, the client transport will need to
632                  * be recomputed after migration recovery has completed.
633                  */
634                 if (errorcode == -NFS4ERR_MOVED)
635                         rpc_task_release_transport(task);
636         }
637         return ret;
638 }
639
640 int
641 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
642                         struct nfs4_state *state, long *timeout)
643 {
644         struct nfs4_exception exception = {
645                 .state = state,
646         };
647
648         if (task->tk_status >= 0)
649                 return 0;
650         if (timeout)
651                 exception.timeout = *timeout;
652         task->tk_status = nfs4_async_handle_exception(task, server,
653                         task->tk_status,
654                         &exception);
655         if (exception.delay && timeout)
656                 *timeout = exception.timeout;
657         if (exception.retry)
658                 return -EAGAIN;
659         return 0;
660 }
661
662 /*
663  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
664  * or 'false' otherwise.
665  */
666 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
667 {
668         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
669         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
670 }
671
672 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
673 {
674         spin_lock(&clp->cl_lock);
675         if (time_before(clp->cl_last_renewal,timestamp))
676                 clp->cl_last_renewal = timestamp;
677         spin_unlock(&clp->cl_lock);
678 }
679
680 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
681 {
682         struct nfs_client *clp = server->nfs_client;
683
684         if (!nfs4_has_session(clp))
685                 do_renew_lease(clp, timestamp);
686 }
687
688 struct nfs4_call_sync_data {
689         const struct nfs_server *seq_server;
690         struct nfs4_sequence_args *seq_args;
691         struct nfs4_sequence_res *seq_res;
692 };
693
694 void nfs4_init_sequence(struct nfs4_sequence_args *args,
695                         struct nfs4_sequence_res *res, int cache_reply,
696                         int privileged)
697 {
698         args->sa_slot = NULL;
699         args->sa_cache_this = cache_reply;
700         args->sa_privileged = privileged;
701
702         res->sr_slot = NULL;
703 }
704
705 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
706 {
707         struct nfs4_slot *slot = res->sr_slot;
708         struct nfs4_slot_table *tbl;
709
710         tbl = slot->table;
711         spin_lock(&tbl->slot_tbl_lock);
712         if (!nfs41_wake_and_assign_slot(tbl, slot))
713                 nfs4_free_slot(tbl, slot);
714         spin_unlock(&tbl->slot_tbl_lock);
715
716         res->sr_slot = NULL;
717 }
718
719 static int nfs40_sequence_done(struct rpc_task *task,
720                                struct nfs4_sequence_res *res)
721 {
722         if (res->sr_slot != NULL)
723                 nfs40_sequence_free_slot(res);
724         return 1;
725 }
726
727 #if defined(CONFIG_NFS_V4_1)
728
729 static void nfs41_release_slot(struct nfs4_slot *slot)
730 {
731         struct nfs4_session *session;
732         struct nfs4_slot_table *tbl;
733         bool send_new_highest_used_slotid = false;
734
735         if (!slot)
736                 return;
737         tbl = slot->table;
738         session = tbl->session;
739
740         /* Bump the slot sequence number */
741         if (slot->seq_done)
742                 slot->seq_nr++;
743         slot->seq_done = 0;
744
745         spin_lock(&tbl->slot_tbl_lock);
746         /* Be nice to the server: try to ensure that the last transmitted
747          * value for highest_user_slotid <= target_highest_slotid
748          */
749         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
750                 send_new_highest_used_slotid = true;
751
752         if (nfs41_wake_and_assign_slot(tbl, slot)) {
753                 send_new_highest_used_slotid = false;
754                 goto out_unlock;
755         }
756         nfs4_free_slot(tbl, slot);
757
758         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
759                 send_new_highest_used_slotid = false;
760 out_unlock:
761         spin_unlock(&tbl->slot_tbl_lock);
762         if (send_new_highest_used_slotid)
763                 nfs41_notify_server(session->clp);
764         if (waitqueue_active(&tbl->slot_waitq))
765                 wake_up_all(&tbl->slot_waitq);
766 }
767
768 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
769 {
770         nfs41_release_slot(res->sr_slot);
771         res->sr_slot = NULL;
772 }
773
774 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
775                 u32 seqnr)
776 {
777         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
778                 slot->seq_nr_highest_sent = seqnr;
779 }
780 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
781                 u32 seqnr)
782 {
783         slot->seq_nr_highest_sent = seqnr;
784         slot->seq_nr_last_acked = seqnr;
785 }
786
787 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
788                                 struct nfs4_slot *slot)
789 {
790         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
791         if (!IS_ERR(task))
792                 rpc_put_task_async(task);
793 }
794
795 static int nfs41_sequence_process(struct rpc_task *task,
796                 struct nfs4_sequence_res *res)
797 {
798         struct nfs4_session *session;
799         struct nfs4_slot *slot = res->sr_slot;
800         struct nfs_client *clp;
801         int status;
802         int ret = 1;
803
804         if (slot == NULL)
805                 goto out_noaction;
806         /* don't increment the sequence number if the task wasn't sent */
807         if (!RPC_WAS_SENT(task) || slot->seq_done)
808                 goto out;
809
810         session = slot->table->session;
811         clp = session->clp;
812
813         trace_nfs4_sequence_done(session, res);
814
815         status = res->sr_status;
816         if (task->tk_status == -NFS4ERR_DEADSESSION)
817                 status = -NFS4ERR_DEADSESSION;
818
819         /* Check the SEQUENCE operation status */
820         switch (status) {
821         case 0:
822                 /* Mark this sequence number as having been acked */
823                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
824                 /* Update the slot's sequence and clientid lease timer */
825                 slot->seq_done = 1;
826                 do_renew_lease(clp, res->sr_timestamp);
827                 /* Check sequence flags */
828                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
829                                 !!slot->privileged);
830                 nfs41_update_target_slotid(slot->table, slot, res);
831                 break;
832         case 1:
833                 /*
834                  * sr_status remains 1 if an RPC level error occurred.
835                  * The server may or may not have processed the sequence
836                  * operation..
837                  */
838                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
839                 slot->seq_done = 1;
840                 goto out;
841         case -NFS4ERR_DELAY:
842                 /* The server detected a resend of the RPC call and
843                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
844                  * of RFC5661.
845                  */
846                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
847                         __func__,
848                         slot->slot_nr,
849                         slot->seq_nr);
850                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
851                 goto out_retry;
852         case -NFS4ERR_RETRY_UNCACHED_REP:
853         case -NFS4ERR_SEQ_FALSE_RETRY:
854                 /*
855                  * The server thinks we tried to replay a request.
856                  * Retry the call after bumping the sequence ID.
857                  */
858                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
859                 goto retry_new_seq;
860         case -NFS4ERR_BADSLOT:
861                 /*
862                  * The slot id we used was probably retired. Try again
863                  * using a different slot id.
864                  */
865                 if (slot->slot_nr < slot->table->target_highest_slotid)
866                         goto session_recover;
867                 goto retry_nowait;
868         case -NFS4ERR_SEQ_MISORDERED:
869                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
870                 /*
871                  * Were one or more calls using this slot interrupted?
872                  * If the server never received the request, then our
873                  * transmitted slot sequence number may be too high. However,
874                  * if the server did receive the request then it might
875                  * accidentally give us a reply with a mismatched operation.
876                  * We can sort this out by sending a lone sequence operation
877                  * to the server on the same slot.
878                  */
879                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
880                         slot->seq_nr--;
881                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
882                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
883                                 res->sr_slot = NULL;
884                         }
885                         goto retry_nowait;
886                 }
887                 /*
888                  * RFC5661:
889                  * A retry might be sent while the original request is
890                  * still in progress on the replier. The replier SHOULD
891                  * deal with the issue by returning NFS4ERR_DELAY as the
892                  * reply to SEQUENCE or CB_SEQUENCE operation, but
893                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
894                  *
895                  * Restart the search after a delay.
896                  */
897                 slot->seq_nr = slot->seq_nr_highest_sent;
898                 goto out_retry;
899         case -NFS4ERR_BADSESSION:
900         case -NFS4ERR_DEADSESSION:
901         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
902                 goto session_recover;
903         default:
904                 /* Just update the slot sequence no. */
905                 slot->seq_done = 1;
906         }
907 out:
908         /* The session may be reset by one of the error handlers. */
909         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
910 out_noaction:
911         return ret;
912 session_recover:
913         nfs4_schedule_session_recovery(session, status);
914         dprintk("%s ERROR: %d Reset session\n", __func__, status);
915         nfs41_sequence_free_slot(res);
916         goto out;
917 retry_new_seq:
918         ++slot->seq_nr;
919 retry_nowait:
920         if (rpc_restart_call_prepare(task)) {
921                 nfs41_sequence_free_slot(res);
922                 task->tk_status = 0;
923                 ret = 0;
924         }
925         goto out;
926 out_retry:
927         if (!rpc_restart_call(task))
928                 goto out;
929         rpc_delay(task, NFS4_POLL_RETRY_MAX);
930         return 0;
931 }
932
933 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
934 {
935         if (!nfs41_sequence_process(task, res))
936                 return 0;
937         if (res->sr_slot != NULL)
938                 nfs41_sequence_free_slot(res);
939         return 1;
940
941 }
942 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
943
944 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
945 {
946         if (res->sr_slot == NULL)
947                 return 1;
948         if (res->sr_slot->table->session != NULL)
949                 return nfs41_sequence_process(task, res);
950         return nfs40_sequence_done(task, res);
951 }
952
953 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
954 {
955         if (res->sr_slot != NULL) {
956                 if (res->sr_slot->table->session != NULL)
957                         nfs41_sequence_free_slot(res);
958                 else
959                         nfs40_sequence_free_slot(res);
960         }
961 }
962
963 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
964 {
965         if (res->sr_slot == NULL)
966                 return 1;
967         if (!res->sr_slot->table->session)
968                 return nfs40_sequence_done(task, res);
969         return nfs41_sequence_done(task, res);
970 }
971 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
972
973 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
974 {
975         struct nfs4_call_sync_data *data = calldata;
976
977         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
978
979         nfs4_setup_sequence(data->seq_server->nfs_client,
980                             data->seq_args, data->seq_res, task);
981 }
982
983 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
984 {
985         struct nfs4_call_sync_data *data = calldata;
986
987         nfs41_sequence_done(task, data->seq_res);
988 }
989
990 static const struct rpc_call_ops nfs41_call_sync_ops = {
991         .rpc_call_prepare = nfs41_call_sync_prepare,
992         .rpc_call_done = nfs41_call_sync_done,
993 };
994
995 #else   /* !CONFIG_NFS_V4_1 */
996
997 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
998 {
999         return nfs40_sequence_done(task, res);
1000 }
1001
1002 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1003 {
1004         if (res->sr_slot != NULL)
1005                 nfs40_sequence_free_slot(res);
1006 }
1007
1008 int nfs4_sequence_done(struct rpc_task *task,
1009                        struct nfs4_sequence_res *res)
1010 {
1011         return nfs40_sequence_done(task, res);
1012 }
1013 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1014
1015 #endif  /* !CONFIG_NFS_V4_1 */
1016
1017 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1018 {
1019         res->sr_timestamp = jiffies;
1020         res->sr_status_flags = 0;
1021         res->sr_status = 1;
1022 }
1023
1024 static
1025 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1026                 struct nfs4_sequence_res *res,
1027                 struct nfs4_slot *slot)
1028 {
1029         if (!slot)
1030                 return;
1031         slot->privileged = args->sa_privileged ? 1 : 0;
1032         args->sa_slot = slot;
1033
1034         res->sr_slot = slot;
1035 }
1036
1037 int nfs4_setup_sequence(struct nfs_client *client,
1038                         struct nfs4_sequence_args *args,
1039                         struct nfs4_sequence_res *res,
1040                         struct rpc_task *task)
1041 {
1042         struct nfs4_session *session = nfs4_get_session(client);
1043         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1044         struct nfs4_slot *slot;
1045
1046         /* slot already allocated? */
1047         if (res->sr_slot != NULL)
1048                 goto out_start;
1049
1050         if (session)
1051                 tbl = &session->fc_slot_table;
1052
1053         spin_lock(&tbl->slot_tbl_lock);
1054         /* The state manager will wait until the slot table is empty */
1055         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1056                 goto out_sleep;
1057
1058         slot = nfs4_alloc_slot(tbl);
1059         if (IS_ERR(slot)) {
1060                 if (slot == ERR_PTR(-ENOMEM))
1061                         goto out_sleep_timeout;
1062                 goto out_sleep;
1063         }
1064         spin_unlock(&tbl->slot_tbl_lock);
1065
1066         nfs4_sequence_attach_slot(args, res, slot);
1067
1068         trace_nfs4_setup_sequence(session, args);
1069 out_start:
1070         nfs41_sequence_res_init(res);
1071         rpc_call_start(task);
1072         return 0;
1073 out_sleep_timeout:
1074         /* Try again in 1/4 second */
1075         if (args->sa_privileged)
1076                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1077                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1078         else
1079                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1080                                 NULL, jiffies + (HZ >> 2));
1081         spin_unlock(&tbl->slot_tbl_lock);
1082         return -EAGAIN;
1083 out_sleep:
1084         if (args->sa_privileged)
1085                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1086                                 RPC_PRIORITY_PRIVILEGED);
1087         else
1088                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1089         spin_unlock(&tbl->slot_tbl_lock);
1090         return -EAGAIN;
1091 }
1092 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1093
1094 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1095 {
1096         struct nfs4_call_sync_data *data = calldata;
1097         nfs4_setup_sequence(data->seq_server->nfs_client,
1098                                 data->seq_args, data->seq_res, task);
1099 }
1100
1101 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1102 {
1103         struct nfs4_call_sync_data *data = calldata;
1104         nfs4_sequence_done(task, data->seq_res);
1105 }
1106
1107 static const struct rpc_call_ops nfs40_call_sync_ops = {
1108         .rpc_call_prepare = nfs40_call_sync_prepare,
1109         .rpc_call_done = nfs40_call_sync_done,
1110 };
1111
1112 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1113 {
1114         int ret;
1115         struct rpc_task *task;
1116
1117         task = rpc_run_task(task_setup);
1118         if (IS_ERR(task))
1119                 return PTR_ERR(task);
1120
1121         ret = task->tk_status;
1122         rpc_put_task(task);
1123         return ret;
1124 }
1125
1126 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1127                              struct nfs_server *server,
1128                              struct rpc_message *msg,
1129                              struct nfs4_sequence_args *args,
1130                              struct nfs4_sequence_res *res,
1131                              unsigned short task_flags)
1132 {
1133         struct nfs_client *clp = server->nfs_client;
1134         struct nfs4_call_sync_data data = {
1135                 .seq_server = server,
1136                 .seq_args = args,
1137                 .seq_res = res,
1138         };
1139         struct rpc_task_setup task_setup = {
1140                 .rpc_client = clnt,
1141                 .rpc_message = msg,
1142                 .callback_ops = clp->cl_mvops->call_sync_ops,
1143                 .callback_data = &data,
1144                 .flags = task_flags,
1145         };
1146
1147         return nfs4_call_sync_custom(&task_setup);
1148 }
1149
1150 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1151                                    struct nfs_server *server,
1152                                    struct rpc_message *msg,
1153                                    struct nfs4_sequence_args *args,
1154                                    struct nfs4_sequence_res *res)
1155 {
1156         return nfs4_do_call_sync(clnt, server, msg, args, res, 0);
1157 }
1158
1159
1160 int nfs4_call_sync(struct rpc_clnt *clnt,
1161                    struct nfs_server *server,
1162                    struct rpc_message *msg,
1163                    struct nfs4_sequence_args *args,
1164                    struct nfs4_sequence_res *res,
1165                    int cache_reply)
1166 {
1167         nfs4_init_sequence(args, res, cache_reply, 0);
1168         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1169 }
1170
1171 static void
1172 nfs4_inc_nlink_locked(struct inode *inode)
1173 {
1174         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1175         inc_nlink(inode);
1176 }
1177
1178 static void
1179 nfs4_dec_nlink_locked(struct inode *inode)
1180 {
1181         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1182         drop_nlink(inode);
1183 }
1184
1185 static void
1186 nfs4_update_changeattr_locked(struct inode *inode,
1187                 struct nfs4_change_info *cinfo,
1188                 unsigned long timestamp, unsigned long cache_validity)
1189 {
1190         struct nfs_inode *nfsi = NFS_I(inode);
1191
1192         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1193                 | NFS_INO_INVALID_MTIME
1194                 | cache_validity;
1195
1196         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(inode)) {
1197                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1198                 nfsi->attrtimeo_timestamp = jiffies;
1199         } else {
1200                 if (S_ISDIR(inode->i_mode)) {
1201                         nfsi->cache_validity |= NFS_INO_INVALID_DATA;
1202                         nfs_force_lookup_revalidate(inode);
1203                 } else {
1204                         if (!NFS_PROTO(inode)->have_delegation(inode,
1205                                                                FMODE_READ))
1206                                 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
1207                 }
1208
1209                 if (cinfo->before != inode_peek_iversion_raw(inode))
1210                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1211                                                 NFS_INO_INVALID_ACL |
1212                                                 NFS_INO_INVALID_XATTR;
1213         }
1214         inode_set_iversion_raw(inode, cinfo->after);
1215         nfsi->read_cache_jiffies = timestamp;
1216         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1217         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1218
1219         if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
1220                 nfs_fscache_invalidate(inode);
1221 }
1222
1223 void
1224 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1225                 unsigned long timestamp, unsigned long cache_validity)
1226 {
1227         spin_lock(&dir->i_lock);
1228         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1229         spin_unlock(&dir->i_lock);
1230 }
1231
1232 struct nfs4_open_createattrs {
1233         struct nfs4_label *label;
1234         struct iattr *sattr;
1235         const __u32 verf[2];
1236 };
1237
1238 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1239                 int err, struct nfs4_exception *exception)
1240 {
1241         if (err != -EINVAL)
1242                 return false;
1243         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1244                 return false;
1245         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1246         exception->retry = 1;
1247         return true;
1248 }
1249
1250 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1251 {
1252          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1253 }
1254
1255 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1256 {
1257         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1258
1259         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1260 }
1261
1262 static u32
1263 nfs4_map_atomic_open_share(struct nfs_server *server,
1264                 fmode_t fmode, int openflags)
1265 {
1266         u32 res = 0;
1267
1268         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1269         case FMODE_READ:
1270                 res = NFS4_SHARE_ACCESS_READ;
1271                 break;
1272         case FMODE_WRITE:
1273                 res = NFS4_SHARE_ACCESS_WRITE;
1274                 break;
1275         case FMODE_READ|FMODE_WRITE:
1276                 res = NFS4_SHARE_ACCESS_BOTH;
1277         }
1278         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1279                 goto out;
1280         /* Want no delegation if we're using O_DIRECT */
1281         if (openflags & O_DIRECT)
1282                 res |= NFS4_SHARE_WANT_NO_DELEG;
1283 out:
1284         return res;
1285 }
1286
1287 static enum open_claim_type4
1288 nfs4_map_atomic_open_claim(struct nfs_server *server,
1289                 enum open_claim_type4 claim)
1290 {
1291         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1292                 return claim;
1293         switch (claim) {
1294         default:
1295                 return claim;
1296         case NFS4_OPEN_CLAIM_FH:
1297                 return NFS4_OPEN_CLAIM_NULL;
1298         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1299                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1300         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1301                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1302         }
1303 }
1304
1305 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1306 {
1307         p->o_res.f_attr = &p->f_attr;
1308         p->o_res.f_label = p->f_label;
1309         p->o_res.seqid = p->o_arg.seqid;
1310         p->c_res.seqid = p->c_arg.seqid;
1311         p->o_res.server = p->o_arg.server;
1312         p->o_res.access_request = p->o_arg.access;
1313         nfs_fattr_init(&p->f_attr);
1314         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1315 }
1316
1317 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1318                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1319                 const struct nfs4_open_createattrs *c,
1320                 enum open_claim_type4 claim,
1321                 gfp_t gfp_mask)
1322 {
1323         struct dentry *parent = dget_parent(dentry);
1324         struct inode *dir = d_inode(parent);
1325         struct nfs_server *server = NFS_SERVER(dir);
1326         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1327         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1328         struct nfs4_opendata *p;
1329
1330         p = kzalloc(sizeof(*p), gfp_mask);
1331         if (p == NULL)
1332                 goto err;
1333
1334         p->f_label = nfs4_label_alloc(server, gfp_mask);
1335         if (IS_ERR(p->f_label))
1336                 goto err_free_p;
1337
1338         p->a_label = nfs4_label_alloc(server, gfp_mask);
1339         if (IS_ERR(p->a_label))
1340                 goto err_free_f;
1341
1342         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1343         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1344         if (IS_ERR(p->o_arg.seqid))
1345                 goto err_free_label;
1346         nfs_sb_active(dentry->d_sb);
1347         p->dentry = dget(dentry);
1348         p->dir = parent;
1349         p->owner = sp;
1350         atomic_inc(&sp->so_count);
1351         p->o_arg.open_flags = flags;
1352         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1353         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1354         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1355                         fmode, flags);
1356         if (flags & O_CREAT) {
1357                 p->o_arg.umask = current_umask();
1358                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1359                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1360                         p->o_arg.u.attrs = &p->attrs;
1361                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1362
1363                         memcpy(p->o_arg.u.verifier.data, c->verf,
1364                                         sizeof(p->o_arg.u.verifier.data));
1365                 }
1366         }
1367         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1368          * will return permission denied for all bits until close */
1369         if (!(flags & O_EXCL)) {
1370                 /* ask server to check for all possible rights as results
1371                  * are cached */
1372                 switch (p->o_arg.claim) {
1373                 default:
1374                         break;
1375                 case NFS4_OPEN_CLAIM_NULL:
1376                 case NFS4_OPEN_CLAIM_FH:
1377                         p->o_arg.access = NFS4_ACCESS_READ |
1378                                 NFS4_ACCESS_MODIFY |
1379                                 NFS4_ACCESS_EXTEND |
1380                                 NFS4_ACCESS_EXECUTE;
1381 #ifdef CONFIG_NFS_V4_2
1382                         if (server->caps & NFS_CAP_XATTR)
1383                                 p->o_arg.access |= NFS4_ACCESS_XAREAD |
1384                                     NFS4_ACCESS_XAWRITE |
1385                                     NFS4_ACCESS_XALIST;
1386 #endif
1387                 }
1388         }
1389         p->o_arg.clientid = server->nfs_client->cl_clientid;
1390         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1391         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1392         p->o_arg.name = &dentry->d_name;
1393         p->o_arg.server = server;
1394         p->o_arg.bitmask = nfs4_bitmask(server, label);
1395         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1396         switch (p->o_arg.claim) {
1397         case NFS4_OPEN_CLAIM_NULL:
1398         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1399         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1400                 p->o_arg.fh = NFS_FH(dir);
1401                 break;
1402         case NFS4_OPEN_CLAIM_PREVIOUS:
1403         case NFS4_OPEN_CLAIM_FH:
1404         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1405         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1406                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1407         }
1408         p->c_arg.fh = &p->o_res.fh;
1409         p->c_arg.stateid = &p->o_res.stateid;
1410         p->c_arg.seqid = p->o_arg.seqid;
1411         nfs4_init_opendata_res(p);
1412         kref_init(&p->kref);
1413         return p;
1414
1415 err_free_label:
1416         nfs4_label_free(p->a_label);
1417 err_free_f:
1418         nfs4_label_free(p->f_label);
1419 err_free_p:
1420         kfree(p);
1421 err:
1422         dput(parent);
1423         return NULL;
1424 }
1425
1426 static void nfs4_opendata_free(struct kref *kref)
1427 {
1428         struct nfs4_opendata *p = container_of(kref,
1429                         struct nfs4_opendata, kref);
1430         struct super_block *sb = p->dentry->d_sb;
1431
1432         nfs4_lgopen_release(p->lgp);
1433         nfs_free_seqid(p->o_arg.seqid);
1434         nfs4_sequence_free_slot(&p->o_res.seq_res);
1435         if (p->state != NULL)
1436                 nfs4_put_open_state(p->state);
1437         nfs4_put_state_owner(p->owner);
1438
1439         nfs4_label_free(p->a_label);
1440         nfs4_label_free(p->f_label);
1441
1442         dput(p->dir);
1443         dput(p->dentry);
1444         nfs_sb_deactive(sb);
1445         nfs_fattr_free_names(&p->f_attr);
1446         kfree(p->f_attr.mdsthreshold);
1447         kfree(p);
1448 }
1449
1450 static void nfs4_opendata_put(struct nfs4_opendata *p)
1451 {
1452         if (p != NULL)
1453                 kref_put(&p->kref, nfs4_opendata_free);
1454 }
1455
1456 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1457                 fmode_t fmode)
1458 {
1459         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1460         case FMODE_READ|FMODE_WRITE:
1461                 return state->n_rdwr != 0;
1462         case FMODE_WRITE:
1463                 return state->n_wronly != 0;
1464         case FMODE_READ:
1465                 return state->n_rdonly != 0;
1466         }
1467         WARN_ON_ONCE(1);
1468         return false;
1469 }
1470
1471 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1472                 int open_mode, enum open_claim_type4 claim)
1473 {
1474         int ret = 0;
1475
1476         if (open_mode & (O_EXCL|O_TRUNC))
1477                 goto out;
1478         switch (claim) {
1479         case NFS4_OPEN_CLAIM_NULL:
1480         case NFS4_OPEN_CLAIM_FH:
1481                 goto out;
1482         default:
1483                 break;
1484         }
1485         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1486                 case FMODE_READ:
1487                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1488                                 && state->n_rdonly != 0;
1489                         break;
1490                 case FMODE_WRITE:
1491                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1492                                 && state->n_wronly != 0;
1493                         break;
1494                 case FMODE_READ|FMODE_WRITE:
1495                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1496                                 && state->n_rdwr != 0;
1497         }
1498 out:
1499         return ret;
1500 }
1501
1502 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1503                 enum open_claim_type4 claim)
1504 {
1505         if (delegation == NULL)
1506                 return 0;
1507         if ((delegation->type & fmode) != fmode)
1508                 return 0;
1509         switch (claim) {
1510         case NFS4_OPEN_CLAIM_NULL:
1511         case NFS4_OPEN_CLAIM_FH:
1512                 break;
1513         case NFS4_OPEN_CLAIM_PREVIOUS:
1514                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1515                         break;
1516                 fallthrough;
1517         default:
1518                 return 0;
1519         }
1520         nfs_mark_delegation_referenced(delegation);
1521         return 1;
1522 }
1523
1524 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1525 {
1526         switch (fmode) {
1527                 case FMODE_WRITE:
1528                         state->n_wronly++;
1529                         break;
1530                 case FMODE_READ:
1531                         state->n_rdonly++;
1532                         break;
1533                 case FMODE_READ|FMODE_WRITE:
1534                         state->n_rdwr++;
1535         }
1536         nfs4_state_set_mode_locked(state, state->state | fmode);
1537 }
1538
1539 #ifdef CONFIG_NFS_V4_1
1540 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1541 {
1542         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1543                 return true;
1544         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1545                 return true;
1546         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1547                 return true;
1548         return false;
1549 }
1550 #endif /* CONFIG_NFS_V4_1 */
1551
1552 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1553 {
1554         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1555                 wake_up_all(&state->waitq);
1556 }
1557
1558 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1559 {
1560         struct nfs_client *clp = state->owner->so_server->nfs_client;
1561         bool need_recover = false;
1562
1563         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1564                 need_recover = true;
1565         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1566                 need_recover = true;
1567         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1568                 need_recover = true;
1569         if (need_recover)
1570                 nfs4_state_mark_reclaim_nograce(clp, state);
1571 }
1572
1573 /*
1574  * Check for whether or not the caller may update the open stateid
1575  * to the value passed in by stateid.
1576  *
1577  * Note: This function relies heavily on the server implementing
1578  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1579  * correctly.
1580  * i.e. The stateid seqids have to be initialised to 1, and
1581  * are then incremented on every state transition.
1582  */
1583 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1584                 const nfs4_stateid *stateid)
1585 {
1586         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1587                 /* The common case - we're updating to a new sequence number */
1588                 if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1589                         nfs4_stateid_is_next(&state->open_stateid, stateid)) {
1590                         return true;
1591                 }
1592         } else {
1593                 /* This is the first OPEN in this generation */
1594                 if (stateid->seqid == cpu_to_be32(1))
1595                         return true;
1596         }
1597         return false;
1598 }
1599
1600 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1601 {
1602         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1603                 return;
1604         if (state->n_wronly)
1605                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1606         if (state->n_rdonly)
1607                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1608         if (state->n_rdwr)
1609                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1610         set_bit(NFS_OPEN_STATE, &state->flags);
1611 }
1612
1613 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1614                 nfs4_stateid *stateid, fmode_t fmode)
1615 {
1616         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1617         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1618         case FMODE_WRITE:
1619                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1620                 break;
1621         case FMODE_READ:
1622                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1623                 break;
1624         case 0:
1625                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1626                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1627                 clear_bit(NFS_OPEN_STATE, &state->flags);
1628         }
1629         if (stateid == NULL)
1630                 return;
1631         /* Handle OPEN+OPEN_DOWNGRADE races */
1632         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1633             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1634                 nfs_resync_open_stateid_locked(state);
1635                 goto out;
1636         }
1637         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1638                 nfs4_stateid_copy(&state->stateid, stateid);
1639         nfs4_stateid_copy(&state->open_stateid, stateid);
1640         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1641 out:
1642         nfs_state_log_update_open_stateid(state);
1643 }
1644
1645 static void nfs_clear_open_stateid(struct nfs4_state *state,
1646         nfs4_stateid *arg_stateid,
1647         nfs4_stateid *stateid, fmode_t fmode)
1648 {
1649         write_seqlock(&state->seqlock);
1650         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1651         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1652                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1653         write_sequnlock(&state->seqlock);
1654         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1655                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1656 }
1657
1658 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1659                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1660         __must_hold(&state->owner->so_lock)
1661         __must_hold(&state->seqlock)
1662         __must_hold(RCU)
1663
1664 {
1665         DEFINE_WAIT(wait);
1666         int status = 0;
1667         for (;;) {
1668
1669                 if (nfs_stateid_is_sequential(state, stateid))
1670                         break;
1671
1672                 if (status)
1673                         break;
1674                 /* Rely on seqids for serialisation with NFSv4.0 */
1675                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1676                         break;
1677
1678                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1679                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1680                 /*
1681                  * Ensure we process the state changes in the same order
1682                  * in which the server processed them by delaying the
1683                  * update of the stateid until we are in sequence.
1684                  */
1685                 write_sequnlock(&state->seqlock);
1686                 spin_unlock(&state->owner->so_lock);
1687                 rcu_read_unlock();
1688                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1689
1690                 if (!signal_pending(current)) {
1691                         if (schedule_timeout(5*HZ) == 0)
1692                                 status = -EAGAIN;
1693                         else
1694                                 status = 0;
1695                 } else
1696                         status = -EINTR;
1697                 finish_wait(&state->waitq, &wait);
1698                 rcu_read_lock();
1699                 spin_lock(&state->owner->so_lock);
1700                 write_seqlock(&state->seqlock);
1701         }
1702
1703         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1704             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1705                 nfs4_stateid_copy(freeme, &state->open_stateid);
1706                 nfs_test_and_clear_all_open_stateid(state);
1707         }
1708
1709         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1710                 nfs4_stateid_copy(&state->stateid, stateid);
1711         nfs4_stateid_copy(&state->open_stateid, stateid);
1712         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1713         nfs_state_log_update_open_stateid(state);
1714 }
1715
1716 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1717                 const nfs4_stateid *open_stateid,
1718                 fmode_t fmode,
1719                 nfs4_stateid *freeme)
1720 {
1721         /*
1722          * Protect the call to nfs4_state_set_mode_locked and
1723          * serialise the stateid update
1724          */
1725         write_seqlock(&state->seqlock);
1726         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1727         switch (fmode) {
1728         case FMODE_READ:
1729                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1730                 break;
1731         case FMODE_WRITE:
1732                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1733                 break;
1734         case FMODE_READ|FMODE_WRITE:
1735                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1736         }
1737         set_bit(NFS_OPEN_STATE, &state->flags);
1738         write_sequnlock(&state->seqlock);
1739 }
1740
1741 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1742 {
1743         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1744         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1745         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1746         clear_bit(NFS_OPEN_STATE, &state->flags);
1747 }
1748
1749 static void nfs_state_set_delegation(struct nfs4_state *state,
1750                 const nfs4_stateid *deleg_stateid,
1751                 fmode_t fmode)
1752 {
1753         /*
1754          * Protect the call to nfs4_state_set_mode_locked and
1755          * serialise the stateid update
1756          */
1757         write_seqlock(&state->seqlock);
1758         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1759         set_bit(NFS_DELEGATED_STATE, &state->flags);
1760         write_sequnlock(&state->seqlock);
1761 }
1762
1763 static void nfs_state_clear_delegation(struct nfs4_state *state)
1764 {
1765         write_seqlock(&state->seqlock);
1766         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1767         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1768         write_sequnlock(&state->seqlock);
1769 }
1770
1771 int update_open_stateid(struct nfs4_state *state,
1772                 const nfs4_stateid *open_stateid,
1773                 const nfs4_stateid *delegation,
1774                 fmode_t fmode)
1775 {
1776         struct nfs_server *server = NFS_SERVER(state->inode);
1777         struct nfs_client *clp = server->nfs_client;
1778         struct nfs_inode *nfsi = NFS_I(state->inode);
1779         struct nfs_delegation *deleg_cur;
1780         nfs4_stateid freeme = { };
1781         int ret = 0;
1782
1783         fmode &= (FMODE_READ|FMODE_WRITE);
1784
1785         rcu_read_lock();
1786         spin_lock(&state->owner->so_lock);
1787         if (open_stateid != NULL) {
1788                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1789                 ret = 1;
1790         }
1791
1792         deleg_cur = nfs4_get_valid_delegation(state->inode);
1793         if (deleg_cur == NULL)
1794                 goto no_delegation;
1795
1796         spin_lock(&deleg_cur->lock);
1797         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1798            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1799             (deleg_cur->type & fmode) != fmode)
1800                 goto no_delegation_unlock;
1801
1802         if (delegation == NULL)
1803                 delegation = &deleg_cur->stateid;
1804         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1805                 goto no_delegation_unlock;
1806
1807         nfs_mark_delegation_referenced(deleg_cur);
1808         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1809         ret = 1;
1810 no_delegation_unlock:
1811         spin_unlock(&deleg_cur->lock);
1812 no_delegation:
1813         if (ret)
1814                 update_open_stateflags(state, fmode);
1815         spin_unlock(&state->owner->so_lock);
1816         rcu_read_unlock();
1817
1818         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1819                 nfs4_schedule_state_manager(clp);
1820         if (freeme.type != 0)
1821                 nfs4_test_and_free_stateid(server, &freeme,
1822                                 state->owner->so_cred);
1823
1824         return ret;
1825 }
1826
1827 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1828                 const nfs4_stateid *stateid)
1829 {
1830         struct nfs4_state *state = lsp->ls_state;
1831         bool ret = false;
1832
1833         spin_lock(&state->state_lock);
1834         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1835                 goto out_noupdate;
1836         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1837                 goto out_noupdate;
1838         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1839         ret = true;
1840 out_noupdate:
1841         spin_unlock(&state->state_lock);
1842         return ret;
1843 }
1844
1845 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1846 {
1847         struct nfs_delegation *delegation;
1848
1849         fmode &= FMODE_READ|FMODE_WRITE;
1850         rcu_read_lock();
1851         delegation = nfs4_get_valid_delegation(inode);
1852         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1853                 rcu_read_unlock();
1854                 return;
1855         }
1856         rcu_read_unlock();
1857         nfs4_inode_return_delegation(inode);
1858 }
1859
1860 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1861 {
1862         struct nfs4_state *state = opendata->state;
1863         struct nfs_delegation *delegation;
1864         int open_mode = opendata->o_arg.open_flags;
1865         fmode_t fmode = opendata->o_arg.fmode;
1866         enum open_claim_type4 claim = opendata->o_arg.claim;
1867         nfs4_stateid stateid;
1868         int ret = -EAGAIN;
1869
1870         for (;;) {
1871                 spin_lock(&state->owner->so_lock);
1872                 if (can_open_cached(state, fmode, open_mode, claim)) {
1873                         update_open_stateflags(state, fmode);
1874                         spin_unlock(&state->owner->so_lock);
1875                         goto out_return_state;
1876                 }
1877                 spin_unlock(&state->owner->so_lock);
1878                 rcu_read_lock();
1879                 delegation = nfs4_get_valid_delegation(state->inode);
1880                 if (!can_open_delegated(delegation, fmode, claim)) {
1881                         rcu_read_unlock();
1882                         break;
1883                 }
1884                 /* Save the delegation */
1885                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1886                 rcu_read_unlock();
1887                 nfs_release_seqid(opendata->o_arg.seqid);
1888                 if (!opendata->is_recover) {
1889                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1890                         if (ret != 0)
1891                                 goto out;
1892                 }
1893                 ret = -EAGAIN;
1894
1895                 /* Try to update the stateid using the delegation */
1896                 if (update_open_stateid(state, NULL, &stateid, fmode))
1897                         goto out_return_state;
1898         }
1899 out:
1900         return ERR_PTR(ret);
1901 out_return_state:
1902         refcount_inc(&state->count);
1903         return state;
1904 }
1905
1906 static void
1907 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1908 {
1909         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1910         struct nfs_delegation *delegation;
1911         int delegation_flags = 0;
1912
1913         rcu_read_lock();
1914         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1915         if (delegation)
1916                 delegation_flags = delegation->flags;
1917         rcu_read_unlock();
1918         switch (data->o_arg.claim) {
1919         default:
1920                 break;
1921         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1922         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1923                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1924                                    "returning a delegation for "
1925                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1926                                    clp->cl_hostname);
1927                 return;
1928         }
1929         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1930                 nfs_inode_set_delegation(state->inode,
1931                                 data->owner->so_cred,
1932                                 data->o_res.delegation_type,
1933                                 &data->o_res.delegation,
1934                                 data->o_res.pagemod_limit);
1935         else
1936                 nfs_inode_reclaim_delegation(state->inode,
1937                                 data->owner->so_cred,
1938                                 data->o_res.delegation_type,
1939                                 &data->o_res.delegation,
1940                                 data->o_res.pagemod_limit);
1941
1942         if (data->o_res.do_recall)
1943                 nfs_async_inode_return_delegation(state->inode,
1944                                                   &data->o_res.delegation);
1945 }
1946
1947 /*
1948  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1949  * and update the nfs4_state.
1950  */
1951 static struct nfs4_state *
1952 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1953 {
1954         struct inode *inode = data->state->inode;
1955         struct nfs4_state *state = data->state;
1956         int ret;
1957
1958         if (!data->rpc_done) {
1959                 if (data->rpc_status)
1960                         return ERR_PTR(data->rpc_status);
1961                 /* cached opens have already been processed */
1962                 goto update;
1963         }
1964
1965         ret = nfs_refresh_inode(inode, &data->f_attr);
1966         if (ret)
1967                 return ERR_PTR(ret);
1968
1969         if (data->o_res.delegation_type != 0)
1970                 nfs4_opendata_check_deleg(data, state);
1971 update:
1972         if (!update_open_stateid(state, &data->o_res.stateid,
1973                                 NULL, data->o_arg.fmode))
1974                 return ERR_PTR(-EAGAIN);
1975         refcount_inc(&state->count);
1976
1977         return state;
1978 }
1979
1980 static struct inode *
1981 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1982 {
1983         struct inode *inode;
1984
1985         switch (data->o_arg.claim) {
1986         case NFS4_OPEN_CLAIM_NULL:
1987         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1988         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1989                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1990                         return ERR_PTR(-EAGAIN);
1991                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1992                                 &data->f_attr, data->f_label);
1993                 break;
1994         default:
1995                 inode = d_inode(data->dentry);
1996                 ihold(inode);
1997                 nfs_refresh_inode(inode, &data->f_attr);
1998         }
1999         return inode;
2000 }
2001
2002 static struct nfs4_state *
2003 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2004 {
2005         struct nfs4_state *state;
2006         struct inode *inode;
2007
2008         inode = nfs4_opendata_get_inode(data);
2009         if (IS_ERR(inode))
2010                 return ERR_CAST(inode);
2011         if (data->state != NULL && data->state->inode == inode) {
2012                 state = data->state;
2013                 refcount_inc(&state->count);
2014         } else
2015                 state = nfs4_get_open_state(inode, data->owner);
2016         iput(inode);
2017         if (state == NULL)
2018                 state = ERR_PTR(-ENOMEM);
2019         return state;
2020 }
2021
2022 static struct nfs4_state *
2023 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2024 {
2025         struct nfs4_state *state;
2026
2027         if (!data->rpc_done) {
2028                 state = nfs4_try_open_cached(data);
2029                 trace_nfs4_cached_open(data->state);
2030                 goto out;
2031         }
2032
2033         state = nfs4_opendata_find_nfs4_state(data);
2034         if (IS_ERR(state))
2035                 goto out;
2036
2037         if (data->o_res.delegation_type != 0)
2038                 nfs4_opendata_check_deleg(data, state);
2039         if (!update_open_stateid(state, &data->o_res.stateid,
2040                                 NULL, data->o_arg.fmode)) {
2041                 nfs4_put_open_state(state);
2042                 state = ERR_PTR(-EAGAIN);
2043         }
2044 out:
2045         nfs_release_seqid(data->o_arg.seqid);
2046         return state;
2047 }
2048
2049 static struct nfs4_state *
2050 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2051 {
2052         struct nfs4_state *ret;
2053
2054         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2055                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2056         else
2057                 ret = _nfs4_opendata_to_nfs4_state(data);
2058         nfs4_sequence_free_slot(&data->o_res.seq_res);
2059         return ret;
2060 }
2061
2062 static struct nfs_open_context *
2063 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2064 {
2065         struct nfs_inode *nfsi = NFS_I(state->inode);
2066         struct nfs_open_context *ctx;
2067
2068         rcu_read_lock();
2069         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2070                 if (ctx->state != state)
2071                         continue;
2072                 if ((ctx->mode & mode) != mode)
2073                         continue;
2074                 if (!get_nfs_open_context(ctx))
2075                         continue;
2076                 rcu_read_unlock();
2077                 return ctx;
2078         }
2079         rcu_read_unlock();
2080         return ERR_PTR(-ENOENT);
2081 }
2082
2083 static struct nfs_open_context *
2084 nfs4_state_find_open_context(struct nfs4_state *state)
2085 {
2086         struct nfs_open_context *ctx;
2087
2088         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2089         if (!IS_ERR(ctx))
2090                 return ctx;
2091         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2092         if (!IS_ERR(ctx))
2093                 return ctx;
2094         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2095 }
2096
2097 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2098                 struct nfs4_state *state, enum open_claim_type4 claim)
2099 {
2100         struct nfs4_opendata *opendata;
2101
2102         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2103                         NULL, claim, GFP_NOFS);
2104         if (opendata == NULL)
2105                 return ERR_PTR(-ENOMEM);
2106         opendata->state = state;
2107         refcount_inc(&state->count);
2108         return opendata;
2109 }
2110
2111 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2112                 fmode_t fmode)
2113 {
2114         struct nfs4_state *newstate;
2115         int ret;
2116
2117         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2118                 return 0;
2119         opendata->o_arg.open_flags = 0;
2120         opendata->o_arg.fmode = fmode;
2121         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2122                         NFS_SB(opendata->dentry->d_sb),
2123                         fmode, 0);
2124         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2125         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2126         nfs4_init_opendata_res(opendata);
2127         ret = _nfs4_recover_proc_open(opendata);
2128         if (ret != 0)
2129                 return ret; 
2130         newstate = nfs4_opendata_to_nfs4_state(opendata);
2131         if (IS_ERR(newstate))
2132                 return PTR_ERR(newstate);
2133         if (newstate != opendata->state)
2134                 ret = -ESTALE;
2135         nfs4_close_state(newstate, fmode);
2136         return ret;
2137 }
2138
2139 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2140 {
2141         int ret;
2142
2143         /* memory barrier prior to reading state->n_* */
2144         smp_rmb();
2145         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2146         if (ret != 0)
2147                 return ret;
2148         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2149         if (ret != 0)
2150                 return ret;
2151         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2152         if (ret != 0)
2153                 return ret;
2154         /*
2155          * We may have performed cached opens for all three recoveries.
2156          * Check if we need to update the current stateid.
2157          */
2158         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2159             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2160                 write_seqlock(&state->seqlock);
2161                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2162                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2163                 write_sequnlock(&state->seqlock);
2164         }
2165         return 0;
2166 }
2167
2168 /*
2169  * OPEN_RECLAIM:
2170  *      reclaim state on the server after a reboot.
2171  */
2172 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2173 {
2174         struct nfs_delegation *delegation;
2175         struct nfs4_opendata *opendata;
2176         fmode_t delegation_type = 0;
2177         int status;
2178
2179         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2180                         NFS4_OPEN_CLAIM_PREVIOUS);
2181         if (IS_ERR(opendata))
2182                 return PTR_ERR(opendata);
2183         rcu_read_lock();
2184         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2185         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2186                 delegation_type = delegation->type;
2187         rcu_read_unlock();
2188         opendata->o_arg.u.delegation_type = delegation_type;
2189         status = nfs4_open_recover(opendata, state);
2190         nfs4_opendata_put(opendata);
2191         return status;
2192 }
2193
2194 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2195 {
2196         struct nfs_server *server = NFS_SERVER(state->inode);
2197         struct nfs4_exception exception = { };
2198         int err;
2199         do {
2200                 err = _nfs4_do_open_reclaim(ctx, state);
2201                 trace_nfs4_open_reclaim(ctx, 0, err);
2202                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2203                         continue;
2204                 if (err != -NFS4ERR_DELAY)
2205                         break;
2206                 nfs4_handle_exception(server, err, &exception);
2207         } while (exception.retry);
2208         return err;
2209 }
2210
2211 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2212 {
2213         struct nfs_open_context *ctx;
2214         int ret;
2215
2216         ctx = nfs4_state_find_open_context(state);
2217         if (IS_ERR(ctx))
2218                 return -EAGAIN;
2219         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2220         nfs_state_clear_open_state_flags(state);
2221         ret = nfs4_do_open_reclaim(ctx, state);
2222         put_nfs_open_context(ctx);
2223         return ret;
2224 }
2225
2226 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)
2227 {
2228         switch (err) {
2229                 default:
2230                         printk(KERN_ERR "NFS: %s: unhandled error "
2231                                         "%d.\n", __func__, err);
2232                 case 0:
2233                 case -ENOENT:
2234                 case -EAGAIN:
2235                 case -ESTALE:
2236                 case -ETIMEDOUT:
2237                         break;
2238                 case -NFS4ERR_BADSESSION:
2239                 case -NFS4ERR_BADSLOT:
2240                 case -NFS4ERR_BAD_HIGH_SLOT:
2241                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2242                 case -NFS4ERR_DEADSESSION:
2243                         return -EAGAIN;
2244                 case -NFS4ERR_STALE_CLIENTID:
2245                 case -NFS4ERR_STALE_STATEID:
2246                         /* Don't recall a delegation if it was lost */
2247                         nfs4_schedule_lease_recovery(server->nfs_client);
2248                         return -EAGAIN;
2249                 case -NFS4ERR_MOVED:
2250                         nfs4_schedule_migration_recovery(server);
2251                         return -EAGAIN;
2252                 case -NFS4ERR_LEASE_MOVED:
2253                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2254                         return -EAGAIN;
2255                 case -NFS4ERR_DELEG_REVOKED:
2256                 case -NFS4ERR_ADMIN_REVOKED:
2257                 case -NFS4ERR_EXPIRED:
2258                 case -NFS4ERR_BAD_STATEID:
2259                 case -NFS4ERR_OPENMODE:
2260                         nfs_inode_find_state_and_recover(state->inode,
2261                                         stateid);
2262                         nfs4_schedule_stateid_recovery(server, state);
2263                         return -EAGAIN;
2264                 case -NFS4ERR_DELAY:
2265                 case -NFS4ERR_GRACE:
2266                         ssleep(1);
2267                         return -EAGAIN;
2268                 case -ENOMEM:
2269                 case -NFS4ERR_DENIED:
2270                         if (fl) {
2271                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2272                                 if (lsp)
2273                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2274                         }
2275                         return 0;
2276         }
2277         return err;
2278 }
2279
2280 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2281                 struct nfs4_state *state, const nfs4_stateid *stateid)
2282 {
2283         struct nfs_server *server = NFS_SERVER(state->inode);
2284         struct nfs4_opendata *opendata;
2285         int err = 0;
2286
2287         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2288                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2289         if (IS_ERR(opendata))
2290                 return PTR_ERR(opendata);
2291         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2292         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2293                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2294                 if (err)
2295                         goto out;
2296         }
2297         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2298                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2299                 if (err)
2300                         goto out;
2301         }
2302         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2303                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2304                 if (err)
2305                         goto out;
2306         }
2307         nfs_state_clear_delegation(state);
2308 out:
2309         nfs4_opendata_put(opendata);
2310         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2311 }
2312
2313 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2314 {
2315         struct nfs4_opendata *data = calldata;
2316
2317         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2318                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2319 }
2320
2321 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2322 {
2323         struct nfs4_opendata *data = calldata;
2324
2325         nfs40_sequence_done(task, &data->c_res.seq_res);
2326
2327         data->rpc_status = task->tk_status;
2328         if (data->rpc_status == 0) {
2329                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2330                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2331                 renew_lease(data->o_res.server, data->timestamp);
2332                 data->rpc_done = true;
2333         }
2334 }
2335
2336 static void nfs4_open_confirm_release(void *calldata)
2337 {
2338         struct nfs4_opendata *data = calldata;
2339         struct nfs4_state *state = NULL;
2340
2341         /* If this request hasn't been cancelled, do nothing */
2342         if (!data->cancelled)
2343                 goto out_free;
2344         /* In case of error, no cleanup! */
2345         if (!data->rpc_done)
2346                 goto out_free;
2347         state = nfs4_opendata_to_nfs4_state(data);
2348         if (!IS_ERR(state))
2349                 nfs4_close_state(state, data->o_arg.fmode);
2350 out_free:
2351         nfs4_opendata_put(data);
2352 }
2353
2354 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2355         .rpc_call_prepare = nfs4_open_confirm_prepare,
2356         .rpc_call_done = nfs4_open_confirm_done,
2357         .rpc_release = nfs4_open_confirm_release,
2358 };
2359
2360 /*
2361  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2362  */
2363 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2364 {
2365         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2366         struct rpc_task *task;
2367         struct  rpc_message msg = {
2368                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2369                 .rpc_argp = &data->c_arg,
2370                 .rpc_resp = &data->c_res,
2371                 .rpc_cred = data->owner->so_cred,
2372         };
2373         struct rpc_task_setup task_setup_data = {
2374                 .rpc_client = server->client,
2375                 .rpc_message = &msg,
2376                 .callback_ops = &nfs4_open_confirm_ops,
2377                 .callback_data = data,
2378                 .workqueue = nfsiod_workqueue,
2379                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2380         };
2381         int status;
2382
2383         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2384                                 data->is_recover);
2385         kref_get(&data->kref);
2386         data->rpc_done = false;
2387         data->rpc_status = 0;
2388         data->timestamp = jiffies;
2389         task = rpc_run_task(&task_setup_data);
2390         if (IS_ERR(task))
2391                 return PTR_ERR(task);
2392         status = rpc_wait_for_completion_task(task);
2393         if (status != 0) {
2394                 data->cancelled = true;
2395                 smp_wmb();
2396         } else
2397                 status = data->rpc_status;
2398         rpc_put_task(task);
2399         return status;
2400 }
2401
2402 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2403 {
2404         struct nfs4_opendata *data = calldata;
2405         struct nfs4_state_owner *sp = data->owner;
2406         struct nfs_client *clp = sp->so_server->nfs_client;
2407         enum open_claim_type4 claim = data->o_arg.claim;
2408
2409         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2410                 goto out_wait;
2411         /*
2412          * Check if we still need to send an OPEN call, or if we can use
2413          * a delegation instead.
2414          */
2415         if (data->state != NULL) {
2416                 struct nfs_delegation *delegation;
2417
2418                 if (can_open_cached(data->state, data->o_arg.fmode,
2419                                         data->o_arg.open_flags, claim))
2420                         goto out_no_action;
2421                 rcu_read_lock();
2422                 delegation = nfs4_get_valid_delegation(data->state->inode);
2423                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2424                         goto unlock_no_action;
2425                 rcu_read_unlock();
2426         }
2427         /* Update client id. */
2428         data->o_arg.clientid = clp->cl_clientid;
2429         switch (claim) {
2430         default:
2431                 break;
2432         case NFS4_OPEN_CLAIM_PREVIOUS:
2433         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2434         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2435                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2436                 fallthrough;
2437         case NFS4_OPEN_CLAIM_FH:
2438                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2439         }
2440         data->timestamp = jiffies;
2441         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2442                                 &data->o_arg.seq_args,
2443                                 &data->o_res.seq_res,
2444                                 task) != 0)
2445                 nfs_release_seqid(data->o_arg.seqid);
2446
2447         /* Set the create mode (note dependency on the session type) */
2448         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2449         if (data->o_arg.open_flags & O_EXCL) {
2450                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2451                 if (nfs4_has_persistent_session(clp))
2452                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2453                 else if (clp->cl_mvops->minor_version > 0)
2454                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2455         }
2456         return;
2457 unlock_no_action:
2458         trace_nfs4_cached_open(data->state);
2459         rcu_read_unlock();
2460 out_no_action:
2461         task->tk_action = NULL;
2462 out_wait:
2463         nfs4_sequence_done(task, &data->o_res.seq_res);
2464 }
2465
2466 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2467 {
2468         struct nfs4_opendata *data = calldata;
2469
2470         data->rpc_status = task->tk_status;
2471
2472         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2473                 return;
2474
2475         if (task->tk_status == 0) {
2476                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2477                         switch (data->o_res.f_attr->mode & S_IFMT) {
2478                         case S_IFREG:
2479                                 break;
2480                         case S_IFLNK:
2481                                 data->rpc_status = -ELOOP;
2482                                 break;
2483                         case S_IFDIR:
2484                                 data->rpc_status = -EISDIR;
2485                                 break;
2486                         default:
2487                                 data->rpc_status = -ENOTDIR;
2488                         }
2489                 }
2490                 renew_lease(data->o_res.server, data->timestamp);
2491                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2492                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2493         }
2494         data->rpc_done = true;
2495 }
2496
2497 static void nfs4_open_release(void *calldata)
2498 {
2499         struct nfs4_opendata *data = calldata;
2500         struct nfs4_state *state = NULL;
2501
2502         /* If this request hasn't been cancelled, do nothing */
2503         if (!data->cancelled)
2504                 goto out_free;
2505         /* In case of error, no cleanup! */
2506         if (data->rpc_status != 0 || !data->rpc_done)
2507                 goto out_free;
2508         /* In case we need an open_confirm, no cleanup! */
2509         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2510                 goto out_free;
2511         state = nfs4_opendata_to_nfs4_state(data);
2512         if (!IS_ERR(state))
2513                 nfs4_close_state(state, data->o_arg.fmode);
2514 out_free:
2515         nfs4_opendata_put(data);
2516 }
2517
2518 static const struct rpc_call_ops nfs4_open_ops = {
2519         .rpc_call_prepare = nfs4_open_prepare,
2520         .rpc_call_done = nfs4_open_done,
2521         .rpc_release = nfs4_open_release,
2522 };
2523
2524 static int nfs4_run_open_task(struct nfs4_opendata *data,
2525                               struct nfs_open_context *ctx)
2526 {
2527         struct inode *dir = d_inode(data->dir);
2528         struct nfs_server *server = NFS_SERVER(dir);
2529         struct nfs_openargs *o_arg = &data->o_arg;
2530         struct nfs_openres *o_res = &data->o_res;
2531         struct rpc_task *task;
2532         struct rpc_message msg = {
2533                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2534                 .rpc_argp = o_arg,
2535                 .rpc_resp = o_res,
2536                 .rpc_cred = data->owner->so_cred,
2537         };
2538         struct rpc_task_setup task_setup_data = {
2539                 .rpc_client = server->client,
2540                 .rpc_message = &msg,
2541                 .callback_ops = &nfs4_open_ops,
2542                 .callback_data = data,
2543                 .workqueue = nfsiod_workqueue,
2544                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2545         };
2546         int status;
2547
2548         kref_get(&data->kref);
2549         data->rpc_done = false;
2550         data->rpc_status = 0;
2551         data->cancelled = false;
2552         data->is_recover = false;
2553         if (!ctx) {
2554                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2555                 data->is_recover = true;
2556                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2557         } else {
2558                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2559                 pnfs_lgopen_prepare(data, ctx);
2560         }
2561         task = rpc_run_task(&task_setup_data);
2562         if (IS_ERR(task))
2563                 return PTR_ERR(task);
2564         status = rpc_wait_for_completion_task(task);
2565         if (status != 0) {
2566                 data->cancelled = true;
2567                 smp_wmb();
2568         } else
2569                 status = data->rpc_status;
2570         rpc_put_task(task);
2571
2572         return status;
2573 }
2574
2575 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2576 {
2577         struct inode *dir = d_inode(data->dir);
2578         struct nfs_openres *o_res = &data->o_res;
2579         int status;
2580
2581         status = nfs4_run_open_task(data, NULL);
2582         if (status != 0 || !data->rpc_done)
2583                 return status;
2584
2585         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2586
2587         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2588                 status = _nfs4_proc_open_confirm(data);
2589
2590         return status;
2591 }
2592
2593 /*
2594  * Additional permission checks in order to distinguish between an
2595  * open for read, and an open for execute. This works around the
2596  * fact that NFSv4 OPEN treats read and execute permissions as being
2597  * the same.
2598  * Note that in the non-execute case, we want to turn off permission
2599  * checking if we just created a new file (POSIX open() semantics).
2600  */
2601 static int nfs4_opendata_access(const struct cred *cred,
2602                                 struct nfs4_opendata *opendata,
2603                                 struct nfs4_state *state, fmode_t fmode,
2604                                 int openflags)
2605 {
2606         struct nfs_access_entry cache;
2607         u32 mask, flags;
2608
2609         /* access call failed or for some reason the server doesn't
2610          * support any access modes -- defer access call until later */
2611         if (opendata->o_res.access_supported == 0)
2612                 return 0;
2613
2614         mask = 0;
2615         /*
2616          * Use openflags to check for exec, because fmode won't
2617          * always have FMODE_EXEC set when file open for exec.
2618          */
2619         if (openflags & __FMODE_EXEC) {
2620                 /* ONLY check for exec rights */
2621                 if (S_ISDIR(state->inode->i_mode))
2622                         mask = NFS4_ACCESS_LOOKUP;
2623                 else
2624                         mask = NFS4_ACCESS_EXECUTE;
2625         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2626                 mask = NFS4_ACCESS_READ;
2627
2628         cache.cred = cred;
2629         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2630         nfs_access_add_cache(state->inode, &cache);
2631
2632         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2633         if ((mask & ~cache.mask & flags) == 0)
2634                 return 0;
2635
2636         return -EACCES;
2637 }
2638
2639 /*
2640  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2641  */
2642 static int _nfs4_proc_open(struct nfs4_opendata *data,
2643                            struct nfs_open_context *ctx)
2644 {
2645         struct inode *dir = d_inode(data->dir);
2646         struct nfs_server *server = NFS_SERVER(dir);
2647         struct nfs_openargs *o_arg = &data->o_arg;
2648         struct nfs_openres *o_res = &data->o_res;
2649         int status;
2650
2651         status = nfs4_run_open_task(data, ctx);
2652         if (!data->rpc_done)
2653                 return status;
2654         if (status != 0) {
2655                 if (status == -NFS4ERR_BADNAME &&
2656                                 !(o_arg->open_flags & O_CREAT))
2657                         return -ENOENT;
2658                 return status;
2659         }
2660
2661         nfs_fattr_map_and_free_names(server, &data->f_attr);
2662
2663         if (o_arg->open_flags & O_CREAT) {
2664                 if (o_arg->open_flags & O_EXCL)
2665                         data->file_created = true;
2666                 else if (o_res->cinfo.before != o_res->cinfo.after)
2667                         data->file_created = true;
2668                 if (data->file_created ||
2669                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2670                         nfs4_update_changeattr(dir, &o_res->cinfo,
2671                                         o_res->f_attr->time_start,
2672                                         NFS_INO_INVALID_DATA);
2673         }
2674         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2675                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2676         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2677                 status = _nfs4_proc_open_confirm(data);
2678                 if (status != 0)
2679                         return status;
2680         }
2681         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2682                 nfs4_sequence_free_slot(&o_res->seq_res);
2683                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2684                                 o_res->f_label, NULL);
2685         }
2686         return 0;
2687 }
2688
2689 /*
2690  * OPEN_EXPIRED:
2691  *      reclaim state on the server after a network partition.
2692  *      Assumes caller holds the appropriate lock
2693  */
2694 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2695 {
2696         struct nfs4_opendata *opendata;
2697         int ret;
2698
2699         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2700                         NFS4_OPEN_CLAIM_FH);
2701         if (IS_ERR(opendata))
2702                 return PTR_ERR(opendata);
2703         ret = nfs4_open_recover(opendata, state);
2704         if (ret == -ESTALE)
2705                 d_drop(ctx->dentry);
2706         nfs4_opendata_put(opendata);
2707         return ret;
2708 }
2709
2710 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2711 {
2712         struct nfs_server *server = NFS_SERVER(state->inode);
2713         struct nfs4_exception exception = { };
2714         int err;
2715
2716         do {
2717                 err = _nfs4_open_expired(ctx, state);
2718                 trace_nfs4_open_expired(ctx, 0, err);
2719                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2720                         continue;
2721                 switch (err) {
2722                 default:
2723                         goto out;
2724                 case -NFS4ERR_GRACE:
2725                 case -NFS4ERR_DELAY:
2726                         nfs4_handle_exception(server, err, &exception);
2727                         err = 0;
2728                 }
2729         } while (exception.retry);
2730 out:
2731         return err;
2732 }
2733
2734 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2735 {
2736         struct nfs_open_context *ctx;
2737         int ret;
2738
2739         ctx = nfs4_state_find_open_context(state);
2740         if (IS_ERR(ctx))
2741                 return -EAGAIN;
2742         ret = nfs4_do_open_expired(ctx, state);
2743         put_nfs_open_context(ctx);
2744         return ret;
2745 }
2746
2747 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2748                 const nfs4_stateid *stateid)
2749 {
2750         nfs_remove_bad_delegation(state->inode, stateid);
2751         nfs_state_clear_delegation(state);
2752 }
2753
2754 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2755 {
2756         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2757                 nfs_finish_clear_delegation_stateid(state, NULL);
2758 }
2759
2760 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2761 {
2762         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2763         nfs40_clear_delegation_stateid(state);
2764         nfs_state_clear_open_state_flags(state);
2765         return nfs4_open_expired(sp, state);
2766 }
2767
2768 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2769                 nfs4_stateid *stateid,
2770                 const struct cred *cred)
2771 {
2772         return -NFS4ERR_BAD_STATEID;
2773 }
2774
2775 #if defined(CONFIG_NFS_V4_1)
2776 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2777                 nfs4_stateid *stateid,
2778                 const struct cred *cred)
2779 {
2780         int status;
2781
2782         switch (stateid->type) {
2783         default:
2784                 break;
2785         case NFS4_INVALID_STATEID_TYPE:
2786         case NFS4_SPECIAL_STATEID_TYPE:
2787                 return -NFS4ERR_BAD_STATEID;
2788         case NFS4_REVOKED_STATEID_TYPE:
2789                 goto out_free;
2790         }
2791
2792         status = nfs41_test_stateid(server, stateid, cred);
2793         switch (status) {
2794         case -NFS4ERR_EXPIRED:
2795         case -NFS4ERR_ADMIN_REVOKED:
2796         case -NFS4ERR_DELEG_REVOKED:
2797                 break;
2798         default:
2799                 return status;
2800         }
2801 out_free:
2802         /* Ack the revoked state to the server */
2803         nfs41_free_stateid(server, stateid, cred, true);
2804         return -NFS4ERR_EXPIRED;
2805 }
2806
2807 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2808 {
2809         struct nfs_server *server = NFS_SERVER(state->inode);
2810         nfs4_stateid stateid;
2811         struct nfs_delegation *delegation;
2812         const struct cred *cred = NULL;
2813         int status, ret = NFS_OK;
2814
2815         /* Get the delegation credential for use by test/free_stateid */
2816         rcu_read_lock();
2817         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2818         if (delegation == NULL) {
2819                 rcu_read_unlock();
2820                 nfs_state_clear_delegation(state);
2821                 return NFS_OK;
2822         }
2823
2824         spin_lock(&delegation->lock);
2825         nfs4_stateid_copy(&stateid, &delegation->stateid);
2826
2827         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2828                                 &delegation->flags)) {
2829                 spin_unlock(&delegation->lock);
2830                 rcu_read_unlock();
2831                 return NFS_OK;
2832         }
2833
2834         if (delegation->cred)
2835                 cred = get_cred(delegation->cred);
2836         spin_unlock(&delegation->lock);
2837         rcu_read_unlock();
2838         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2839         trace_nfs4_test_delegation_stateid(state, NULL, status);
2840         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2841                 nfs_finish_clear_delegation_stateid(state, &stateid);
2842         else
2843                 ret = status;
2844
2845         put_cred(cred);
2846         return ret;
2847 }
2848
2849 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2850 {
2851         nfs4_stateid tmp;
2852
2853         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2854             nfs4_copy_delegation_stateid(state->inode, state->state,
2855                                 &tmp, NULL) &&
2856             nfs4_stateid_match_other(&state->stateid, &tmp))
2857                 nfs_state_set_delegation(state, &tmp, state->state);
2858         else
2859                 nfs_state_clear_delegation(state);
2860 }
2861
2862 /**
2863  * nfs41_check_expired_locks - possibly free a lock stateid
2864  *
2865  * @state: NFSv4 state for an inode
2866  *
2867  * Returns NFS_OK if recovery for this stateid is now finished.
2868  * Otherwise a negative NFS4ERR value is returned.
2869  */
2870 static int nfs41_check_expired_locks(struct nfs4_state *state)
2871 {
2872         int status, ret = NFS_OK;
2873         struct nfs4_lock_state *lsp, *prev = NULL;
2874         struct nfs_server *server = NFS_SERVER(state->inode);
2875
2876         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2877                 goto out;
2878
2879         spin_lock(&state->state_lock);
2880         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2881                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2882                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2883
2884                         refcount_inc(&lsp->ls_count);
2885                         spin_unlock(&state->state_lock);
2886
2887                         nfs4_put_lock_state(prev);
2888                         prev = lsp;
2889
2890                         status = nfs41_test_and_free_expired_stateid(server,
2891                                         &lsp->ls_stateid,
2892                                         cred);
2893                         trace_nfs4_test_lock_stateid(state, lsp, status);
2894                         if (status == -NFS4ERR_EXPIRED ||
2895                             status == -NFS4ERR_BAD_STATEID) {
2896                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2897                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2898                                 if (!recover_lost_locks)
2899                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2900                         } else if (status != NFS_OK) {
2901                                 ret = status;
2902                                 nfs4_put_lock_state(prev);
2903                                 goto out;
2904                         }
2905                         spin_lock(&state->state_lock);
2906                 }
2907         }
2908         spin_unlock(&state->state_lock);
2909         nfs4_put_lock_state(prev);
2910 out:
2911         return ret;
2912 }
2913
2914 /**
2915  * nfs41_check_open_stateid - possibly free an open stateid
2916  *
2917  * @state: NFSv4 state for an inode
2918  *
2919  * Returns NFS_OK if recovery for this stateid is now finished.
2920  * Otherwise a negative NFS4ERR value is returned.
2921  */
2922 static int nfs41_check_open_stateid(struct nfs4_state *state)
2923 {
2924         struct nfs_server *server = NFS_SERVER(state->inode);
2925         nfs4_stateid *stateid = &state->open_stateid;
2926         const struct cred *cred = state->owner->so_cred;
2927         int status;
2928
2929         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2930                 return -NFS4ERR_BAD_STATEID;
2931         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2932         trace_nfs4_test_open_stateid(state, NULL, status);
2933         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2934                 nfs_state_clear_open_state_flags(state);
2935                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2936                 return status;
2937         }
2938         if (nfs_open_stateid_recover_openmode(state))
2939                 return -NFS4ERR_OPENMODE;
2940         return NFS_OK;
2941 }
2942
2943 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2944 {
2945         int status;
2946
2947         status = nfs41_check_delegation_stateid(state);
2948         if (status != NFS_OK)
2949                 return status;
2950         nfs41_delegation_recover_stateid(state);
2951
2952         status = nfs41_check_expired_locks(state);
2953         if (status != NFS_OK)
2954                 return status;
2955         status = nfs41_check_open_stateid(state);
2956         if (status != NFS_OK)
2957                 status = nfs4_open_expired(sp, state);
2958         return status;
2959 }
2960 #endif
2961
2962 /*
2963  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2964  * fields corresponding to attributes that were used to store the verifier.
2965  * Make sure we clobber those fields in the later setattr call
2966  */
2967 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2968                                 struct iattr *sattr, struct nfs4_label **label)
2969 {
2970         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2971         __u32 attrset[3];
2972         unsigned ret;
2973         unsigned i;
2974
2975         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2976                 attrset[i] = opendata->o_res.attrset[i];
2977                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2978                         attrset[i] &= ~bitmask[i];
2979         }
2980
2981         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2982                 sattr->ia_valid : 0;
2983
2984         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2985                 if (sattr->ia_valid & ATTR_ATIME_SET)
2986                         ret |= ATTR_ATIME_SET;
2987                 else
2988                         ret |= ATTR_ATIME;
2989         }
2990
2991         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2992                 if (sattr->ia_valid & ATTR_MTIME_SET)
2993                         ret |= ATTR_MTIME_SET;
2994                 else
2995                         ret |= ATTR_MTIME;
2996         }
2997
2998         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2999                 *label = NULL;
3000         return ret;
3001 }
3002
3003 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3004                 int flags, struct nfs_open_context *ctx)
3005 {
3006         struct nfs4_state_owner *sp = opendata->owner;
3007         struct nfs_server *server = sp->so_server;
3008         struct dentry *dentry;
3009         struct nfs4_state *state;
3010         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3011         struct inode *dir = d_inode(opendata->dir);
3012         unsigned long dir_verifier;
3013         unsigned int seq;
3014         int ret;
3015
3016         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3017         dir_verifier = nfs_save_change_attribute(dir);
3018
3019         ret = _nfs4_proc_open(opendata, ctx);
3020         if (ret != 0)
3021                 goto out;
3022
3023         state = _nfs4_opendata_to_nfs4_state(opendata);
3024         ret = PTR_ERR(state);
3025         if (IS_ERR(state))
3026                 goto out;
3027         ctx->state = state;
3028         if (server->caps & NFS_CAP_POSIX_LOCK)
3029                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3030         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3031                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3032
3033         dentry = opendata->dentry;
3034         if (d_really_is_negative(dentry)) {
3035                 struct dentry *alias;
3036                 d_drop(dentry);
3037                 alias = d_exact_alias(dentry, state->inode);
3038                 if (!alias)
3039                         alias = d_splice_alias(igrab(state->inode), dentry);
3040                 /* d_splice_alias() can't fail here - it's a non-directory */
3041                 if (alias) {
3042                         dput(ctx->dentry);
3043                         ctx->dentry = dentry = alias;
3044                 }
3045         }
3046
3047         switch(opendata->o_arg.claim) {
3048         default:
3049                 break;
3050         case NFS4_OPEN_CLAIM_NULL:
3051         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3052         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3053                 if (!opendata->rpc_done)
3054                         break;
3055                 if (opendata->o_res.delegation_type != 0)
3056                         dir_verifier = nfs_save_change_attribute(dir);
3057                 nfs_set_verifier(dentry, dir_verifier);
3058         }
3059
3060         /* Parse layoutget results before we check for access */
3061         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3062
3063         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3064                         acc_mode, flags);
3065         if (ret != 0)
3066                 goto out;
3067
3068         if (d_inode(dentry) == state->inode) {
3069                 nfs_inode_attach_open_context(ctx);
3070                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3071                         nfs4_schedule_stateid_recovery(server, state);
3072         }
3073
3074 out:
3075         if (!opendata->cancelled)
3076                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3077         return ret;
3078 }
3079
3080 /*
3081  * Returns a referenced nfs4_state
3082  */
3083 static int _nfs4_do_open(struct inode *dir,
3084                         struct nfs_open_context *ctx,
3085                         int flags,
3086                         const struct nfs4_open_createattrs *c,
3087                         int *opened)
3088 {
3089         struct nfs4_state_owner  *sp;
3090         struct nfs4_state     *state = NULL;
3091         struct nfs_server       *server = NFS_SERVER(dir);
3092         struct nfs4_opendata *opendata;
3093         struct dentry *dentry = ctx->dentry;
3094         const struct cred *cred = ctx->cred;
3095         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3096         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3097         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3098         struct iattr *sattr = c->sattr;
3099         struct nfs4_label *label = c->label;
3100         struct nfs4_label *olabel = NULL;
3101         int status;
3102
3103         /* Protect against reboot recovery conflicts */
3104         status = -ENOMEM;
3105         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3106         if (sp == NULL) {
3107                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3108                 goto out_err;
3109         }
3110         status = nfs4_client_recover_expired_lease(server->nfs_client);
3111         if (status != 0)
3112                 goto err_put_state_owner;
3113         if (d_really_is_positive(dentry))
3114                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3115         status = -ENOMEM;
3116         if (d_really_is_positive(dentry))
3117                 claim = NFS4_OPEN_CLAIM_FH;
3118         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3119                         c, claim, GFP_KERNEL);
3120         if (opendata == NULL)
3121                 goto err_put_state_owner;
3122
3123         if (label) {
3124                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3125                 if (IS_ERR(olabel)) {
3126                         status = PTR_ERR(olabel);
3127                         goto err_opendata_put;
3128                 }
3129         }
3130
3131         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3132                 if (!opendata->f_attr.mdsthreshold) {
3133                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3134                         if (!opendata->f_attr.mdsthreshold)
3135                                 goto err_free_label;
3136                 }
3137                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3138         }
3139         if (d_really_is_positive(dentry))
3140                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3141
3142         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3143         if (status != 0)
3144                 goto err_free_label;
3145         state = ctx->state;
3146
3147         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3148             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3149                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3150                 /*
3151                  * send create attributes which was not set by open
3152                  * with an extra setattr.
3153                  */
3154                 if (attrs || label) {
3155                         unsigned ia_old = sattr->ia_valid;
3156
3157                         sattr->ia_valid = attrs;
3158                         nfs_fattr_init(opendata->o_res.f_attr);
3159                         status = nfs4_do_setattr(state->inode, cred,
3160                                         opendata->o_res.f_attr, sattr,
3161                                         ctx, label, olabel);
3162                         if (status == 0) {
3163                                 nfs_setattr_update_inode(state->inode, sattr,
3164                                                 opendata->o_res.f_attr);
3165                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3166                         }
3167                         sattr->ia_valid = ia_old;
3168                 }
3169         }
3170         if (opened && opendata->file_created)
3171                 *opened = 1;
3172
3173         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3174                 *ctx_th = opendata->f_attr.mdsthreshold;
3175                 opendata->f_attr.mdsthreshold = NULL;
3176         }
3177
3178         nfs4_label_free(olabel);
3179
3180         nfs4_opendata_put(opendata);
3181         nfs4_put_state_owner(sp);
3182         return 0;
3183 err_free_label:
3184         nfs4_label_free(olabel);
3185 err_opendata_put:
3186         nfs4_opendata_put(opendata);
3187 err_put_state_owner:
3188         nfs4_put_state_owner(sp);
3189 out_err:
3190         return status;
3191 }
3192
3193
3194 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3195                                         struct nfs_open_context *ctx,
3196                                         int flags,
3197                                         struct iattr *sattr,
3198                                         struct nfs4_label *label,
3199                                         int *opened)
3200 {
3201         struct nfs_server *server = NFS_SERVER(dir);
3202         struct nfs4_exception exception = {
3203                 .interruptible = true,
3204         };
3205         struct nfs4_state *res;
3206         struct nfs4_open_createattrs c = {
3207                 .label = label,
3208                 .sattr = sattr,
3209                 .verf = {
3210                         [0] = (__u32)jiffies,
3211                         [1] = (__u32)current->pid,
3212                 },
3213         };
3214         int status;
3215
3216         do {
3217                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3218                 res = ctx->state;
3219                 trace_nfs4_open_file(ctx, flags, status);
3220                 if (status == 0)
3221                         break;
3222                 /* NOTE: BAD_SEQID means the server and client disagree about the
3223                  * book-keeping w.r.t. state-changing operations
3224                  * (OPEN/CLOSE/LOCK/LOCKU...)
3225                  * It is actually a sign of a bug on the client or on the server.
3226                  *
3227                  * If we receive a BAD_SEQID error in the particular case of
3228                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3229                  * have unhashed the old state_owner for us, and that we can
3230                  * therefore safely retry using a new one. We should still warn
3231                  * the user though...
3232                  */
3233                 if (status == -NFS4ERR_BAD_SEQID) {
3234                         pr_warn_ratelimited("NFS: v4 server %s "
3235                                         " returned a bad sequence-id error!\n",
3236                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3237                         exception.retry = 1;
3238                         continue;
3239                 }
3240                 /*
3241                  * BAD_STATEID on OPEN means that the server cancelled our
3242                  * state before it received the OPEN_CONFIRM.
3243                  * Recover by retrying the request as per the discussion
3244                  * on Page 181 of RFC3530.
3245                  */
3246                 if (status == -NFS4ERR_BAD_STATEID) {
3247                         exception.retry = 1;
3248                         continue;
3249                 }
3250                 if (status == -NFS4ERR_EXPIRED) {
3251                         nfs4_schedule_lease_recovery(server->nfs_client);
3252                         exception.retry = 1;
3253                         continue;
3254                 }
3255                 if (status == -EAGAIN) {
3256                         /* We must have found a delegation */
3257                         exception.retry = 1;
3258                         continue;
3259                 }
3260                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3261                         continue;
3262                 res = ERR_PTR(nfs4_handle_exception(server,
3263                                         status, &exception));
3264         } while (exception.retry);
3265         return res;
3266 }
3267
3268 static int _nfs4_do_setattr(struct inode *inode,
3269                             struct nfs_setattrargs *arg,
3270                             struct nfs_setattrres *res,
3271                             const struct cred *cred,
3272                             struct nfs_open_context *ctx)
3273 {
3274         struct nfs_server *server = NFS_SERVER(inode);
3275         struct rpc_message msg = {
3276                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3277                 .rpc_argp       = arg,
3278                 .rpc_resp       = res,
3279                 .rpc_cred       = cred,
3280         };
3281         const struct cred *delegation_cred = NULL;
3282         unsigned long timestamp = jiffies;
3283         bool truncate;
3284         int status;
3285
3286         nfs_fattr_init(res->fattr);
3287
3288         /* Servers should only apply open mode checks for file size changes */
3289         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3290         if (!truncate) {
3291                 nfs4_inode_make_writeable(inode);
3292                 goto zero_stateid;
3293         }
3294
3295         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3296                 /* Use that stateid */
3297         } else if (ctx != NULL && ctx->state) {
3298                 struct nfs_lock_context *l_ctx;
3299                 if (!nfs4_valid_open_stateid(ctx->state))
3300                         return -EBADF;
3301                 l_ctx = nfs_get_lock_context(ctx);
3302                 if (IS_ERR(l_ctx))
3303                         return PTR_ERR(l_ctx);
3304                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3305                                                 &arg->stateid, &delegation_cred);
3306                 nfs_put_lock_context(l_ctx);
3307                 if (status == -EIO)
3308                         return -EBADF;
3309                 else if (status == -EAGAIN)
3310                         goto zero_stateid;
3311         } else {
3312 zero_stateid:
3313                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3314         }
3315         if (delegation_cred)
3316                 msg.rpc_cred = delegation_cred;
3317
3318         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3319
3320         put_cred(delegation_cred);
3321         if (status == 0 && ctx != NULL)
3322                 renew_lease(server, timestamp);
3323         trace_nfs4_setattr(inode, &arg->stateid, status);
3324         return status;
3325 }
3326
3327 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3328                            struct nfs_fattr *fattr, struct iattr *sattr,
3329                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3330                            struct nfs4_label *olabel)
3331 {
3332         struct nfs_server *server = NFS_SERVER(inode);
3333         __u32 bitmask[NFS4_BITMASK_SZ];
3334         struct nfs4_state *state = ctx ? ctx->state : NULL;
3335         struct nfs_setattrargs  arg = {
3336                 .fh             = NFS_FH(inode),
3337                 .iap            = sattr,
3338                 .server         = server,
3339                 .bitmask = bitmask,
3340                 .label          = ilabel,
3341         };
3342         struct nfs_setattrres  res = {
3343                 .fattr          = fattr,
3344                 .label          = olabel,
3345                 .server         = server,
3346         };
3347         struct nfs4_exception exception = {
3348                 .state = state,
3349                 .inode = inode,
3350                 .stateid = &arg.stateid,
3351         };
3352         int err;
3353
3354         do {
3355                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3356                                 nfs4_bitmask(server, olabel),
3357                                 inode);
3358
3359                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3360                 switch (err) {
3361                 case -NFS4ERR_OPENMODE:
3362                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3363                                 pr_warn_once("NFSv4: server %s is incorrectly "
3364                                                 "applying open mode checks to "
3365                                                 "a SETATTR that is not "
3366                                                 "changing file size.\n",
3367                                                 server->nfs_client->cl_hostname);
3368                         }
3369                         if (state && !(state->state & FMODE_WRITE)) {
3370                                 err = -EBADF;
3371                                 if (sattr->ia_valid & ATTR_OPEN)
3372                                         err = -EACCES;
3373                                 goto out;
3374                         }
3375                 }
3376                 err = nfs4_handle_exception(server, err, &exception);
3377         } while (exception.retry);
3378 out:
3379         return err;
3380 }
3381
3382 static bool
3383 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3384 {
3385         if (inode == NULL || !nfs_have_layout(inode))
3386                 return false;
3387
3388         return pnfs_wait_on_layoutreturn(inode, task);
3389 }
3390
3391 /*
3392  * Update the seqid of an open stateid
3393  */
3394 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3395                 struct nfs4_state *state)
3396 {
3397         __be32 seqid_open;
3398         u32 dst_seqid;
3399         int seq;
3400
3401         for (;;) {
3402                 if (!nfs4_valid_open_stateid(state))
3403                         break;
3404                 seq = read_seqbegin(&state->seqlock);
3405                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3406                         nfs4_stateid_copy(dst, &state->open_stateid);
3407                         if (read_seqretry(&state->seqlock, seq))
3408                                 continue;
3409                         break;
3410                 }
3411                 seqid_open = state->open_stateid.seqid;
3412                 if (read_seqretry(&state->seqlock, seq))
3413                         continue;
3414
3415                 dst_seqid = be32_to_cpu(dst->seqid);
3416                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3417                         dst->seqid = seqid_open;
3418                 break;
3419         }
3420 }
3421
3422 /*
3423  * Update the seqid of an open stateid after receiving
3424  * NFS4ERR_OLD_STATEID
3425  */
3426 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3427                 struct nfs4_state *state)
3428 {
3429         __be32 seqid_open;
3430         u32 dst_seqid;
3431         bool ret;
3432         int seq, status = -EAGAIN;
3433         DEFINE_WAIT(wait);
3434
3435         for (;;) {
3436                 ret = false;
3437                 if (!nfs4_valid_open_stateid(state))
3438                         break;
3439                 seq = read_seqbegin(&state->seqlock);
3440                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3441                         if (read_seqretry(&state->seqlock, seq))
3442                                 continue;
3443                         break;
3444                 }
3445
3446                 write_seqlock(&state->seqlock);
3447                 seqid_open = state->open_stateid.seqid;
3448
3449                 dst_seqid = be32_to_cpu(dst->seqid);
3450
3451                 /* Did another OPEN bump the state's seqid?  try again: */
3452                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3453                         dst->seqid = seqid_open;
3454                         write_sequnlock(&state->seqlock);
3455                         ret = true;
3456                         break;
3457                 }
3458
3459                 /* server says we're behind but we haven't seen the update yet */
3460                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3461                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3462                 write_sequnlock(&state->seqlock);
3463                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3464
3465                 if (signal_pending(current))
3466                         status = -EINTR;
3467                 else
3468                         if (schedule_timeout(5*HZ) != 0)
3469                                 status = 0;
3470
3471                 finish_wait(&state->waitq, &wait);
3472
3473                 if (!status)
3474                         continue;
3475                 if (status == -EINTR)
3476                         break;
3477
3478                 /* we slept the whole 5 seconds, we must have lost a seqid */
3479                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3480                 ret = true;
3481                 break;
3482         }
3483
3484         return ret;
3485 }
3486
3487 struct nfs4_closedata {
3488         struct inode *inode;
3489         struct nfs4_state *state;
3490         struct nfs_closeargs arg;
3491         struct nfs_closeres res;
3492         struct {
3493                 struct nfs4_layoutreturn_args arg;
3494                 struct nfs4_layoutreturn_res res;
3495                 struct nfs4_xdr_opaque_data ld_private;
3496                 u32 roc_barrier;
3497                 bool roc;
3498         } lr;
3499         struct nfs_fattr fattr;
3500         unsigned long timestamp;
3501 };
3502
3503 static void nfs4_free_closedata(void *data)
3504 {
3505         struct nfs4_closedata *calldata = data;
3506         struct nfs4_state_owner *sp = calldata->state->owner;
3507         struct super_block *sb = calldata->state->inode->i_sb;
3508
3509         if (calldata->lr.roc)
3510                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3511                                 calldata->res.lr_ret);
3512         nfs4_put_open_state(calldata->state);
3513         nfs_free_seqid(calldata->arg.seqid);
3514         nfs4_put_state_owner(sp);
3515         nfs_sb_deactive(sb);
3516         kfree(calldata);
3517 }
3518
3519 static void nfs4_close_done(struct rpc_task *task, void *data)
3520 {
3521         struct nfs4_closedata *calldata = data;
3522         struct nfs4_state *state = calldata->state;
3523         struct nfs_server *server = NFS_SERVER(calldata->inode);
3524         nfs4_stateid *res_stateid = NULL;
3525         struct nfs4_exception exception = {
3526                 .state = state,
3527                 .inode = calldata->inode,
3528                 .stateid = &calldata->arg.stateid,
3529         };
3530
3531         dprintk("%s: begin!\n", __func__);
3532         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3533                 return;
3534         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3535
3536         /* Handle Layoutreturn errors */
3537         if (pnfs_roc_done(task, calldata->inode,
3538                                 &calldata->arg.lr_args,
3539                                 &calldata->res.lr_res,
3540                                 &calldata->res.lr_ret) == -EAGAIN)
3541                 goto out_restart;
3542
3543         /* hmm. we are done with the inode, and in the process of freeing
3544          * the state_owner. we keep this around to process errors
3545          */
3546         switch (task->tk_status) {
3547                 case 0:
3548                         res_stateid = &calldata->res.stateid;
3549                         renew_lease(server, calldata->timestamp);
3550                         break;
3551                 case -NFS4ERR_ACCESS:
3552                         if (calldata->arg.bitmask != NULL) {
3553                                 calldata->arg.bitmask = NULL;
3554                                 calldata->res.fattr = NULL;
3555                                 goto out_restart;
3556
3557                         }
3558                         break;
3559                 case -NFS4ERR_OLD_STATEID:
3560                         /* Did we race with OPEN? */
3561                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3562                                                 state))
3563                                 goto out_restart;
3564                         goto out_release;
3565                 case -NFS4ERR_ADMIN_REVOKED:
3566                 case -NFS4ERR_STALE_STATEID:
3567                 case -NFS4ERR_EXPIRED:
3568                         nfs4_free_revoked_stateid(server,
3569                                         &calldata->arg.stateid,
3570                                         task->tk_msg.rpc_cred);
3571                         fallthrough;
3572                 case -NFS4ERR_BAD_STATEID:
3573                         if (calldata->arg.fmode == 0)
3574                                 break;
3575                         fallthrough;
3576                 default:
3577                         task->tk_status = nfs4_async_handle_exception(task,
3578                                         server, task->tk_status, &exception);
3579                         if (exception.retry)
3580                                 goto out_restart;
3581         }
3582         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3583                         res_stateid, calldata->arg.fmode);
3584 out_release:
3585         task->tk_status = 0;
3586         nfs_release_seqid(calldata->arg.seqid);
3587         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3588         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3589         return;
3590 out_restart:
3591         task->tk_status = 0;
3592         rpc_restart_call_prepare(task);
3593         goto out_release;
3594 }
3595
3596 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3597 {
3598         struct nfs4_closedata *calldata = data;
3599         struct nfs4_state *state = calldata->state;
3600         struct inode *inode = calldata->inode;
3601         struct pnfs_layout_hdr *lo;
3602         bool is_rdonly, is_wronly, is_rdwr;
3603         int call_close = 0;
3604
3605         dprintk("%s: begin!\n", __func__);
3606         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3607                 goto out_wait;
3608
3609         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3610         spin_lock(&state->owner->so_lock);
3611         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3612         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3613         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3614         /* Calculate the change in open mode */
3615         calldata->arg.fmode = 0;
3616         if (state->n_rdwr == 0) {
3617                 if (state->n_rdonly == 0)
3618                         call_close |= is_rdonly;
3619                 else if (is_rdonly)
3620                         calldata->arg.fmode |= FMODE_READ;
3621                 if (state->n_wronly == 0)
3622                         call_close |= is_wronly;
3623                 else if (is_wronly)
3624                         calldata->arg.fmode |= FMODE_WRITE;
3625                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3626                         call_close |= is_rdwr;
3627         } else if (is_rdwr)
3628                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3629
3630         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3631         if (!nfs4_valid_open_stateid(state))
3632                 call_close = 0;
3633         spin_unlock(&state->owner->so_lock);
3634
3635         if (!call_close) {
3636                 /* Note: exit _without_ calling nfs4_close_done */
3637                 goto out_no_action;
3638         }
3639
3640         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3641                 nfs_release_seqid(calldata->arg.seqid);
3642                 goto out_wait;
3643         }
3644
3645         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3646         if (lo && !pnfs_layout_is_valid(lo)) {
3647                 calldata->arg.lr_args = NULL;
3648                 calldata->res.lr_res = NULL;
3649         }
3650
3651         if (calldata->arg.fmode == 0)
3652                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3653
3654         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3655                 /* Close-to-open cache consistency revalidation */
3656                 if (!nfs4_have_delegation(inode, FMODE_READ)) {
3657                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3658                         nfs4_bitmask_adjust(calldata->arg.bitmask, inode, NFS_SERVER(inode), NULL);
3659                 } else
3660                         calldata->arg.bitmask = NULL;
3661         }
3662
3663         calldata->arg.share_access =
3664                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3665                                 calldata->arg.fmode, 0);
3666
3667         if (calldata->res.fattr == NULL)
3668                 calldata->arg.bitmask = NULL;
3669         else if (calldata->arg.bitmask == NULL)
3670                 calldata->res.fattr = NULL;
3671         calldata->timestamp = jiffies;
3672         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3673                                 &calldata->arg.seq_args,
3674                                 &calldata->res.seq_res,
3675                                 task) != 0)
3676                 nfs_release_seqid(calldata->arg.seqid);
3677         dprintk("%s: done!\n", __func__);
3678         return;
3679 out_no_action:
3680         task->tk_action = NULL;
3681 out_wait:
3682         nfs4_sequence_done(task, &calldata->res.seq_res);
3683 }
3684
3685 static const struct rpc_call_ops nfs4_close_ops = {
3686         .rpc_call_prepare = nfs4_close_prepare,
3687         .rpc_call_done = nfs4_close_done,
3688         .rpc_release = nfs4_free_closedata,
3689 };
3690
3691 /* 
3692  * It is possible for data to be read/written from a mem-mapped file 
3693  * after the sys_close call (which hits the vfs layer as a flush).
3694  * This means that we can't safely call nfsv4 close on a file until 
3695  * the inode is cleared. This in turn means that we are not good
3696  * NFSv4 citizens - we do not indicate to the server to update the file's 
3697  * share state even when we are done with one of the three share 
3698  * stateid's in the inode.
3699  *
3700  * NOTE: Caller must be holding the sp->so_owner semaphore!
3701  */
3702 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3703 {
3704         struct nfs_server *server = NFS_SERVER(state->inode);
3705         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3706         struct nfs4_closedata *calldata;
3707         struct nfs4_state_owner *sp = state->owner;
3708         struct rpc_task *task;
3709         struct rpc_message msg = {
3710                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3711                 .rpc_cred = state->owner->so_cred,
3712         };
3713         struct rpc_task_setup task_setup_data = {
3714                 .rpc_client = server->client,
3715                 .rpc_message = &msg,
3716                 .callback_ops = &nfs4_close_ops,
3717                 .workqueue = nfsiod_workqueue,
3718                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3719         };
3720         int status = -ENOMEM;
3721
3722         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3723                 &task_setup_data.rpc_client, &msg);
3724
3725         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3726         if (calldata == NULL)
3727                 goto out;
3728         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3729         calldata->inode = state->inode;
3730         calldata->state = state;
3731         calldata->arg.fh = NFS_FH(state->inode);
3732         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3733                 goto out_free_calldata;
3734         /* Serialization for the sequence id */
3735         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3736         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3737         if (IS_ERR(calldata->arg.seqid))
3738                 goto out_free_calldata;
3739         nfs_fattr_init(&calldata->fattr);
3740         calldata->arg.fmode = 0;
3741         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3742         calldata->res.fattr = &calldata->fattr;
3743         calldata->res.seqid = calldata->arg.seqid;
3744         calldata->res.server = server;
3745         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3746         calldata->lr.roc = pnfs_roc(state->inode,
3747                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3748         if (calldata->lr.roc) {
3749                 calldata->arg.lr_args = &calldata->lr.arg;
3750                 calldata->res.lr_res = &calldata->lr.res;
3751         }
3752         nfs_sb_active(calldata->inode->i_sb);
3753
3754         msg.rpc_argp = &calldata->arg;
3755         msg.rpc_resp = &calldata->res;
3756         task_setup_data.callback_data = calldata;
3757         task = rpc_run_task(&task_setup_data);
3758         if (IS_ERR(task))
3759                 return PTR_ERR(task);
3760         status = 0;
3761         if (wait)
3762                 status = rpc_wait_for_completion_task(task);
3763         rpc_put_task(task);
3764         return status;
3765 out_free_calldata:
3766         kfree(calldata);
3767 out:
3768         nfs4_put_open_state(state);
3769         nfs4_put_state_owner(sp);
3770         return status;
3771 }
3772
3773 static struct inode *
3774 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3775                 int open_flags, struct iattr *attr, int *opened)
3776 {
3777         struct nfs4_state *state;
3778         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3779
3780         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3781
3782         /* Protect against concurrent sillydeletes */
3783         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3784
3785         nfs4_label_release_security(label);
3786
3787         if (IS_ERR(state))
3788                 return ERR_CAST(state);
3789         return state->inode;
3790 }
3791
3792 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3793 {
3794         if (ctx->state == NULL)
3795                 return;
3796         if (is_sync)
3797                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3798         else
3799                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3800 }
3801
3802 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3803 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3804 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3805
3806 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3807 {
3808         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3809         struct nfs4_server_caps_arg args = {
3810                 .fhandle = fhandle,
3811                 .bitmask = bitmask,
3812         };
3813         struct nfs4_server_caps_res res = {};
3814         struct rpc_message msg = {
3815                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3816                 .rpc_argp = &args,
3817                 .rpc_resp = &res,
3818         };
3819         int status;
3820         int i;
3821
3822         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3823                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3824                      FATTR4_WORD0_LINK_SUPPORT |
3825                      FATTR4_WORD0_SYMLINK_SUPPORT |
3826                      FATTR4_WORD0_ACLSUPPORT;
3827         if (minorversion)
3828                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3829
3830         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3831         if (status == 0) {
3832                 /* Sanity check the server answers */
3833                 switch (minorversion) {
3834                 case 0:
3835                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3836                         res.attr_bitmask[2] = 0;
3837                         break;
3838                 case 1:
3839                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3840                         break;
3841                 case 2:
3842                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3843                 }
3844                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3845                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3846                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3847                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3848                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3849                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3850                                 NFS_CAP_SECURITY_LABEL);
3851                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3852                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3853                         server->caps |= NFS_CAP_ACLS;
3854                 if (res.has_links != 0)
3855                         server->caps |= NFS_CAP_HARDLINKS;
3856                 if (res.has_symlinks != 0)
3857                         server->caps |= NFS_CAP_SYMLINKS;
3858                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3859                         server->caps |= NFS_CAP_FILEID;
3860                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3861                         server->caps |= NFS_CAP_MODE;
3862                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3863                         server->caps |= NFS_CAP_NLINK;
3864                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3865                         server->caps |= NFS_CAP_OWNER;
3866                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3867                         server->caps |= NFS_CAP_OWNER_GROUP;
3868                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3869                         server->caps |= NFS_CAP_ATIME;
3870                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3871                         server->caps |= NFS_CAP_CTIME;
3872                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3873                         server->caps |= NFS_CAP_MTIME;
3874 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3875                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3876                         server->caps |= NFS_CAP_SECURITY_LABEL;
3877 #endif
3878                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3879                                 sizeof(server->attr_bitmask));
3880                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3881
3882                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3883                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3884                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3885                 server->cache_consistency_bitmask[2] = 0;
3886
3887                 /* Avoid a regression due to buggy server */
3888                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3889                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3890                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3891                         sizeof(server->exclcreat_bitmask));
3892
3893                 server->acl_bitmask = res.acl_bitmask;
3894                 server->fh_expire_type = res.fh_expire_type;
3895         }
3896
3897         return status;
3898 }
3899
3900 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3901 {
3902         struct nfs4_exception exception = {
3903                 .interruptible = true,
3904         };
3905         int err;
3906         do {
3907                 err = nfs4_handle_exception(server,
3908                                 _nfs4_server_capabilities(server, fhandle),
3909                                 &exception);
3910         } while (exception.retry);
3911         return err;
3912 }
3913
3914 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3915                 struct nfs_fsinfo *info)
3916 {
3917         u32 bitmask[3];
3918         struct nfs4_lookup_root_arg args = {
3919                 .bitmask = bitmask,
3920         };
3921         struct nfs4_lookup_res res = {
3922                 .server = server,
3923                 .fattr = info->fattr,
3924                 .fh = fhandle,
3925         };
3926         struct rpc_message msg = {
3927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3928                 .rpc_argp = &args,
3929                 .rpc_resp = &res,
3930         };
3931
3932         bitmask[0] = nfs4_fattr_bitmap[0];
3933         bitmask[1] = nfs4_fattr_bitmap[1];
3934         /*
3935          * Process the label in the upcoming getfattr
3936          */
3937         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3938
3939         nfs_fattr_init(info->fattr);
3940         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3941 }
3942
3943 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3944                 struct nfs_fsinfo *info)
3945 {
3946         struct nfs4_exception exception = {
3947                 .interruptible = true,
3948         };
3949         int err;
3950         do {
3951                 err = _nfs4_lookup_root(server, fhandle, info);
3952                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3953                 switch (err) {
3954                 case 0:
3955                 case -NFS4ERR_WRONGSEC:
3956                         goto out;
3957                 default:
3958                         err = nfs4_handle_exception(server, err, &exception);
3959                 }
3960         } while (exception.retry);
3961 out:
3962         return err;
3963 }
3964
3965 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3966                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3967 {
3968         struct rpc_auth_create_args auth_args = {
3969                 .pseudoflavor = flavor,
3970         };
3971         struct rpc_auth *auth;
3972
3973         auth = rpcauth_create(&auth_args, server->client);
3974         if (IS_ERR(auth))
3975                 return -EACCES;
3976         return nfs4_lookup_root(server, fhandle, info);
3977 }
3978
3979 /*
3980  * Retry pseudoroot lookup with various security flavors.  We do this when:
3981  *
3982  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3983  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3984  *
3985  * Returns zero on success, or a negative NFS4ERR value, or a
3986  * negative errno value.
3987  */
3988 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3989                               struct nfs_fsinfo *info)
3990 {
3991         /* Per 3530bis 15.33.5 */
3992         static const rpc_authflavor_t flav_array[] = {
3993                 RPC_AUTH_GSS_KRB5P,
3994                 RPC_AUTH_GSS_KRB5I,
3995                 RPC_AUTH_GSS_KRB5,
3996                 RPC_AUTH_UNIX,                  /* courtesy */
3997                 RPC_AUTH_NULL,
3998         };
3999         int status = -EPERM;
4000         size_t i;
4001
4002         if (server->auth_info.flavor_len > 0) {
4003                 /* try each flavor specified by user */
4004                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4005                         status = nfs4_lookup_root_sec(server, fhandle, info,
4006                                                 server->auth_info.flavors[i]);
4007                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4008                                 continue;
4009                         break;
4010                 }
4011         } else {
4012                 /* no flavors specified by user, try default list */
4013                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4014                         status = nfs4_lookup_root_sec(server, fhandle, info,
4015                                                       flav_array[i]);
4016                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4017                                 continue;
4018                         break;
4019                 }
4020         }
4021
4022         /*
4023          * -EACCES could mean that the user doesn't have correct permissions
4024          * to access the mount.  It could also mean that we tried to mount
4025          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4026          * existing mount programs don't handle -EACCES very well so it should
4027          * be mapped to -EPERM instead.
4028          */
4029         if (status == -EACCES)
4030                 status = -EPERM;
4031         return status;
4032 }
4033
4034 /**
4035  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4036  * @server: initialized nfs_server handle
4037  * @fhandle: we fill in the pseudo-fs root file handle
4038  * @info: we fill in an FSINFO struct
4039  * @auth_probe: probe the auth flavours
4040  *
4041  * Returns zero on success, or a negative errno.
4042  */
4043 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4044                          struct nfs_fsinfo *info,
4045                          bool auth_probe)
4046 {
4047         int status = 0;
4048
4049         if (!auth_probe)
4050                 status = nfs4_lookup_root(server, fhandle, info);
4051
4052         if (auth_probe || status == NFS4ERR_WRONGSEC)
4053                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4054                                 fhandle, info);
4055
4056         if (status == 0)
4057                 status = nfs4_server_capabilities(server, fhandle);
4058         if (status == 0)
4059                 status = nfs4_do_fsinfo(server, fhandle, info);
4060
4061         return nfs4_map_errors(status);
4062 }
4063
4064 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4065                               struct nfs_fsinfo *info)
4066 {
4067         int error;
4068         struct nfs_fattr *fattr = info->fattr;
4069         struct nfs4_label *label = fattr->label;
4070
4071         error = nfs4_server_capabilities(server, mntfh);
4072         if (error < 0) {
4073                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4074                 return error;
4075         }
4076
4077         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4078         if (error < 0) {
4079                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4080                 goto out;
4081         }
4082
4083         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4084             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4085                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4086
4087 out:
4088         return error;
4089 }
4090
4091 /*
4092  * Get locations and (maybe) other attributes of a referral.
4093  * Note that we'll actually follow the referral later when
4094  * we detect fsid mismatch in inode revalidation
4095  */
4096 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4097                              const struct qstr *name, struct nfs_fattr *fattr,
4098                              struct nfs_fh *fhandle)
4099 {
4100         int status = -ENOMEM;
4101         struct page *page = NULL;
4102         struct nfs4_fs_locations *locations = NULL;
4103
4104         page = alloc_page(GFP_KERNEL);
4105         if (page == NULL)
4106                 goto out;
4107         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4108         if (locations == NULL)
4109                 goto out;
4110
4111         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4112         if (status != 0)
4113                 goto out;
4114
4115         /*
4116          * If the fsid didn't change, this is a migration event, not a
4117          * referral.  Cause us to drop into the exception handler, which
4118          * will kick off migration recovery.
4119          */
4120         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4121                 dprintk("%s: server did not return a different fsid for"
4122                         " a referral at %s\n", __func__, name->name);
4123                 status = -NFS4ERR_MOVED;
4124                 goto out;
4125         }
4126         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4127         nfs_fixup_referral_attributes(&locations->fattr);
4128
4129         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4130         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4131         memset(fhandle, 0, sizeof(struct nfs_fh));
4132 out:
4133         if (page)
4134                 __free_page(page);
4135         kfree(locations);
4136         return status;
4137 }
4138
4139 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4140                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4141                                 struct inode *inode)
4142 {
4143         __u32 bitmask[NFS4_BITMASK_SZ];
4144         struct nfs4_getattr_arg args = {
4145                 .fh = fhandle,
4146                 .bitmask = bitmask,
4147         };
4148         struct nfs4_getattr_res res = {
4149                 .fattr = fattr,
4150                 .label = label,
4151                 .server = server,
4152         };
4153         struct rpc_message msg = {
4154                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4155                 .rpc_argp = &args,
4156                 .rpc_resp = &res,
4157         };
4158         unsigned short task_flags = 0;
4159
4160         /* Is this is an attribute revalidation, subject to softreval? */
4161         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4162                 task_flags |= RPC_TASK_TIMEOUT;
4163
4164         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4165
4166         nfs_fattr_init(fattr);
4167         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4168         return nfs4_do_call_sync(server->client, server, &msg,
4169                         &args.seq_args, &res.seq_res, task_flags);
4170 }
4171
4172 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4173                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4174                                 struct inode *inode)
4175 {
4176         struct nfs4_exception exception = {
4177                 .interruptible = true,
4178         };
4179         int err;
4180         do {
4181                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4182                 trace_nfs4_getattr(server, fhandle, fattr, err);
4183                 err = nfs4_handle_exception(server, err,
4184                                 &exception);
4185         } while (exception.retry);
4186         return err;
4187 }
4188
4189 /* 
4190  * The file is not closed if it is opened due to the a request to change
4191  * the size of the file. The open call will not be needed once the
4192  * VFS layer lookup-intents are implemented.
4193  *
4194  * Close is called when the inode is destroyed.
4195  * If we haven't opened the file for O_WRONLY, we
4196  * need to in the size_change case to obtain a stateid.
4197  *
4198  * Got race?
4199  * Because OPEN is always done by name in nfsv4, it is
4200  * possible that we opened a different file by the same
4201  * name.  We can recognize this race condition, but we
4202  * can't do anything about it besides returning an error.
4203  *
4204  * This will be fixed with VFS changes (lookup-intent).
4205  */
4206 static int
4207 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4208                   struct iattr *sattr)
4209 {
4210         struct inode *inode = d_inode(dentry);
4211         const struct cred *cred = NULL;
4212         struct nfs_open_context *ctx = NULL;
4213         struct nfs4_label *label = NULL;
4214         int status;
4215
4216         if (pnfs_ld_layoutret_on_setattr(inode) &&
4217             sattr->ia_valid & ATTR_SIZE &&
4218             sattr->ia_size < i_size_read(inode))
4219                 pnfs_commit_and_return_layout(inode);
4220
4221         nfs_fattr_init(fattr);
4222         
4223         /* Deal with open(O_TRUNC) */
4224         if (sattr->ia_valid & ATTR_OPEN)
4225                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4226
4227         /* Optimization: if the end result is no change, don't RPC */
4228         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4229                 return 0;
4230
4231         /* Search for an existing open(O_WRITE) file */
4232         if (sattr->ia_valid & ATTR_FILE) {
4233
4234                 ctx = nfs_file_open_context(sattr->ia_file);
4235                 if (ctx)
4236                         cred = ctx->cred;
4237         }
4238
4239         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4240         if (IS_ERR(label))
4241                 return PTR_ERR(label);
4242
4243         /* Return any delegations if we're going to change ACLs */
4244         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4245                 nfs4_inode_make_writeable(inode);
4246
4247         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4248         if (status == 0) {
4249                 nfs_setattr_update_inode(inode, sattr, fattr);
4250                 nfs_setsecurity(inode, fattr, label);
4251         }
4252         nfs4_label_free(label);
4253         return status;
4254 }
4255
4256 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4257                 struct dentry *dentry, struct nfs_fh *fhandle,
4258                 struct nfs_fattr *fattr, struct nfs4_label *label)
4259 {
4260         struct nfs_server *server = NFS_SERVER(dir);
4261         int                    status;
4262         struct nfs4_lookup_arg args = {
4263                 .bitmask = server->attr_bitmask,
4264                 .dir_fh = NFS_FH(dir),
4265                 .name = &dentry->d_name,
4266         };
4267         struct nfs4_lookup_res res = {
4268                 .server = server,
4269                 .fattr = fattr,
4270                 .label = label,
4271                 .fh = fhandle,
4272         };
4273         struct rpc_message msg = {
4274                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4275                 .rpc_argp = &args,
4276                 .rpc_resp = &res,
4277         };
4278         unsigned short task_flags = 0;
4279
4280         /* Is this is an attribute revalidation, subject to softreval? */
4281         if (nfs_lookup_is_soft_revalidate(dentry))
4282                 task_flags |= RPC_TASK_TIMEOUT;
4283
4284         args.bitmask = nfs4_bitmask(server, label);
4285
4286         nfs_fattr_init(fattr);
4287
4288         dprintk("NFS call  lookup %pd2\n", dentry);
4289         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4290         status = nfs4_do_call_sync(clnt, server, &msg,
4291                         &args.seq_args, &res.seq_res, task_flags);
4292         dprintk("NFS reply lookup: %d\n", status);
4293         return status;
4294 }
4295
4296 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4297 {
4298         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4299                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4300         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4301         fattr->nlink = 2;
4302 }
4303
4304 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4305                                    struct dentry *dentry, struct nfs_fh *fhandle,
4306                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4307 {
4308         struct nfs4_exception exception = {
4309                 .interruptible = true,
4310         };
4311         struct rpc_clnt *client = *clnt;
4312         const struct qstr *name = &dentry->d_name;
4313         int err;
4314         do {
4315                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4316                 trace_nfs4_lookup(dir, name, err);
4317                 switch (err) {
4318                 case -NFS4ERR_BADNAME:
4319                         err = -ENOENT;
4320                         goto out;
4321                 case -NFS4ERR_MOVED:
4322                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4323                         if (err == -NFS4ERR_MOVED)
4324                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4325                         goto out;
4326                 case -NFS4ERR_WRONGSEC:
4327                         err = -EPERM;
4328                         if (client != *clnt)
4329                                 goto out;
4330                         client = nfs4_negotiate_security(client, dir, name);
4331                         if (IS_ERR(client))
4332                                 return PTR_ERR(client);
4333
4334                         exception.retry = 1;
4335                         break;
4336                 default:
4337                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4338                 }
4339         } while (exception.retry);
4340
4341 out:
4342         if (err == 0)
4343                 *clnt = client;
4344         else if (client != *clnt)
4345                 rpc_shutdown_client(client);
4346
4347         return err;
4348 }
4349
4350 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4351                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4352                             struct nfs4_label *label)
4353 {
4354         int status;
4355         struct rpc_clnt *client = NFS_CLIENT(dir);
4356
4357         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4358         if (client != NFS_CLIENT(dir)) {
4359                 rpc_shutdown_client(client);
4360                 nfs_fixup_secinfo_attributes(fattr);
4361         }
4362         return status;
4363 }
4364
4365 struct rpc_clnt *
4366 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4367                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4368 {
4369         struct rpc_clnt *client = NFS_CLIENT(dir);
4370         int status;
4371
4372         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4373         if (status < 0)
4374                 return ERR_PTR(status);
4375         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4376 }
4377
4378 static int _nfs4_proc_lookupp(struct inode *inode,
4379                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4380                 struct nfs4_label *label)
4381 {
4382         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4383         struct nfs_server *server = NFS_SERVER(inode);
4384         int                    status;
4385         struct nfs4_lookupp_arg args = {
4386                 .bitmask = server->attr_bitmask,
4387                 .fh = NFS_FH(inode),
4388         };
4389         struct nfs4_lookupp_res res = {
4390                 .server = server,
4391                 .fattr = fattr,
4392                 .label = label,
4393                 .fh = fhandle,
4394         };
4395         struct rpc_message msg = {
4396                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4397                 .rpc_argp = &args,
4398                 .rpc_resp = &res,
4399         };
4400
4401         args.bitmask = nfs4_bitmask(server, label);
4402
4403         nfs_fattr_init(fattr);
4404
4405         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4406         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4407                                 &res.seq_res, 0);
4408         dprintk("NFS reply lookupp: %d\n", status);
4409         return status;
4410 }
4411
4412 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4413                              struct nfs_fattr *fattr, struct nfs4_label *label)
4414 {
4415         struct nfs4_exception exception = {
4416                 .interruptible = true,
4417         };
4418         int err;
4419         do {
4420                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4421                 trace_nfs4_lookupp(inode, err);
4422                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4423                                 &exception);
4424         } while (exception.retry);
4425         return err;
4426 }
4427
4428 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4429 {
4430         struct nfs_server *server = NFS_SERVER(inode);
4431         struct nfs4_accessargs args = {
4432                 .fh = NFS_FH(inode),
4433                 .access = entry->mask,
4434         };
4435         struct nfs4_accessres res = {
4436                 .server = server,
4437         };
4438         struct rpc_message msg = {
4439                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4440                 .rpc_argp = &args,
4441                 .rpc_resp = &res,
4442                 .rpc_cred = entry->cred,
4443         };
4444         int status = 0;
4445
4446         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4447                 res.fattr = nfs_alloc_fattr();
4448                 if (res.fattr == NULL)
4449                         return -ENOMEM;
4450                 args.bitmask = server->cache_consistency_bitmask;
4451         }
4452         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4453         if (!status) {
4454                 nfs_access_set_mask(entry, res.access);
4455                 if (res.fattr)
4456                         nfs_refresh_inode(inode, res.fattr);
4457         }
4458         nfs_free_fattr(res.fattr);
4459         return status;
4460 }
4461
4462 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4463 {
4464         struct nfs4_exception exception = {
4465                 .interruptible = true,
4466         };
4467         int err;
4468         do {
4469                 err = _nfs4_proc_access(inode, entry);
4470                 trace_nfs4_access(inode, err);
4471                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4472                                 &exception);
4473         } while (exception.retry);
4474         return err;
4475 }
4476
4477 /*
4478  * TODO: For the time being, we don't try to get any attributes
4479  * along with any of the zero-copy operations READ, READDIR,
4480  * READLINK, WRITE.
4481  *
4482  * In the case of the first three, we want to put the GETATTR
4483  * after the read-type operation -- this is because it is hard
4484  * to predict the length of a GETATTR response in v4, and thus
4485  * align the READ data correctly.  This means that the GETATTR
4486  * may end up partially falling into the page cache, and we should
4487  * shift it into the 'tail' of the xdr_buf before processing.
4488  * To do this efficiently, we need to know the total length
4489  * of data received, which doesn't seem to be available outside
4490  * of the RPC layer.
4491  *
4492  * In the case of WRITE, we also want to put the GETATTR after
4493  * the operation -- in this case because we want to make sure
4494  * we get the post-operation mtime and size.
4495  *
4496  * Both of these changes to the XDR layer would in fact be quite
4497  * minor, but I decided to leave them for a subsequent patch.
4498  */
4499 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4500                 unsigned int pgbase, unsigned int pglen)
4501 {
4502         struct nfs4_readlink args = {
4503                 .fh       = NFS_FH(inode),
4504                 .pgbase   = pgbase,
4505                 .pglen    = pglen,
4506                 .pages    = &page,
4507         };
4508         struct nfs4_readlink_res res;
4509         struct rpc_message msg = {
4510                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4511                 .rpc_argp = &args,
4512                 .rpc_resp = &res,
4513         };
4514
4515         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4516 }
4517
4518 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4519                 unsigned int pgbase, unsigned int pglen)
4520 {
4521         struct nfs4_exception exception = {
4522                 .interruptible = true,
4523         };
4524         int err;
4525         do {
4526                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4527                 trace_nfs4_readlink(inode, err);
4528                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4529                                 &exception);
4530         } while (exception.retry);
4531         return err;
4532 }
4533
4534 /*
4535  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4536  */
4537 static int
4538 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4539                  int flags)
4540 {
4541         struct nfs_server *server = NFS_SERVER(dir);
4542         struct nfs4_label l, *ilabel = NULL;
4543         struct nfs_open_context *ctx;
4544         struct nfs4_state *state;
4545         int status = 0;
4546
4547         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4548         if (IS_ERR(ctx))
4549                 return PTR_ERR(ctx);
4550
4551         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4552
4553         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4554                 sattr->ia_mode &= ~current_umask();
4555         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4556         if (IS_ERR(state)) {
4557                 status = PTR_ERR(state);
4558                 goto out;
4559         }
4560 out:
4561         nfs4_label_release_security(ilabel);
4562         put_nfs_open_context(ctx);
4563         return status;
4564 }
4565
4566 static int
4567 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4568 {
4569         struct nfs_server *server = NFS_SERVER(dir);
4570         struct nfs_removeargs args = {
4571                 .fh = NFS_FH(dir),
4572                 .name = *name,
4573         };
4574         struct nfs_removeres res = {
4575                 .server = server,
4576         };
4577         struct rpc_message msg = {
4578                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4579                 .rpc_argp = &args,
4580                 .rpc_resp = &res,
4581         };
4582         unsigned long timestamp = jiffies;
4583         int status;
4584
4585         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4586         if (status == 0) {
4587                 spin_lock(&dir->i_lock);
4588                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4589                                               NFS_INO_INVALID_DATA);
4590                 /* Removing a directory decrements nlink in the parent */
4591                 if (ftype == NF4DIR && dir->i_nlink > 2)
4592                         nfs4_dec_nlink_locked(dir);
4593                 spin_unlock(&dir->i_lock);
4594         }
4595         return status;
4596 }
4597
4598 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4599 {
4600         struct nfs4_exception exception = {
4601                 .interruptible = true,
4602         };
4603         struct inode *inode = d_inode(dentry);
4604         int err;
4605
4606         if (inode) {
4607                 if (inode->i_nlink == 1)
4608                         nfs4_inode_return_delegation(inode);
4609                 else
4610                         nfs4_inode_make_writeable(inode);
4611         }
4612         do {
4613                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4614                 trace_nfs4_remove(dir, &dentry->d_name, err);
4615                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4616                                 &exception);
4617         } while (exception.retry);
4618         return err;
4619 }
4620
4621 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4622 {
4623         struct nfs4_exception exception = {
4624                 .interruptible = true,
4625         };
4626         int err;
4627
4628         do {
4629                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4630                 trace_nfs4_remove(dir, name, err);
4631                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4632                                 &exception);
4633         } while (exception.retry);
4634         return err;
4635 }
4636
4637 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4638                 struct dentry *dentry,
4639                 struct inode *inode)
4640 {
4641         struct nfs_removeargs *args = msg->rpc_argp;
4642         struct nfs_removeres *res = msg->rpc_resp;
4643
4644         res->server = NFS_SB(dentry->d_sb);
4645         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4646         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4647
4648         nfs_fattr_init(res->dir_attr);
4649
4650         if (inode)
4651                 nfs4_inode_return_delegation(inode);
4652 }
4653
4654 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4655 {
4656         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4657                         &data->args.seq_args,
4658                         &data->res.seq_res,
4659                         task);
4660 }
4661
4662 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4663 {
4664         struct nfs_unlinkdata *data = task->tk_calldata;
4665         struct nfs_removeres *res = &data->res;
4666
4667         if (!nfs4_sequence_done(task, &res->seq_res))
4668                 return 0;
4669         if (nfs4_async_handle_error(task, res->server, NULL,
4670                                     &data->timeout) == -EAGAIN)
4671                 return 0;
4672         if (task->tk_status == 0)
4673                 nfs4_update_changeattr(dir, &res->cinfo,
4674                                 res->dir_attr->time_start,
4675                                 NFS_INO_INVALID_DATA);
4676         return 1;
4677 }
4678
4679 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4680                 struct dentry *old_dentry,
4681                 struct dentry *new_dentry)
4682 {
4683         struct nfs_renameargs *arg = msg->rpc_argp;
4684         struct nfs_renameres *res = msg->rpc_resp;
4685         struct inode *old_inode = d_inode(old_dentry);
4686         struct inode *new_inode = d_inode(new_dentry);
4687
4688         if (old_inode)
4689                 nfs4_inode_make_writeable(old_inode);
4690         if (new_inode)
4691                 nfs4_inode_return_delegation(new_inode);
4692         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4693         res->server = NFS_SB(old_dentry->d_sb);
4694         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4695 }
4696
4697 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4698 {
4699         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4700                         &data->args.seq_args,
4701                         &data->res.seq_res,
4702                         task);
4703 }
4704
4705 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4706                                  struct inode *new_dir)
4707 {
4708         struct nfs_renamedata *data = task->tk_calldata;
4709         struct nfs_renameres *res = &data->res;
4710
4711         if (!nfs4_sequence_done(task, &res->seq_res))
4712                 return 0;
4713         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4714                 return 0;
4715
4716         if (task->tk_status == 0) {
4717                 if (new_dir != old_dir) {
4718                         /* Note: If we moved a directory, nlink will change */
4719                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4720                                         res->old_fattr->time_start,
4721                                         NFS_INO_INVALID_OTHER |
4722                                             NFS_INO_INVALID_DATA);
4723                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4724                                         res->new_fattr->time_start,
4725                                         NFS_INO_INVALID_OTHER |
4726                                             NFS_INO_INVALID_DATA);
4727                 } else
4728                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4729                                         res->old_fattr->time_start,
4730                                         NFS_INO_INVALID_DATA);
4731         }
4732         return 1;
4733 }
4734
4735 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4736 {
4737         struct nfs_server *server = NFS_SERVER(inode);
4738         __u32 bitmask[NFS4_BITMASK_SZ];
4739         struct nfs4_link_arg arg = {
4740                 .fh     = NFS_FH(inode),
4741                 .dir_fh = NFS_FH(dir),
4742                 .name   = name,
4743                 .bitmask = bitmask,
4744         };
4745         struct nfs4_link_res res = {
4746                 .server = server,
4747                 .label = NULL,
4748         };
4749         struct rpc_message msg = {
4750                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4751                 .rpc_argp = &arg,
4752                 .rpc_resp = &res,
4753         };
4754         int status = -ENOMEM;
4755
4756         res.fattr = nfs_alloc_fattr();
4757         if (res.fattr == NULL)
4758                 goto out;
4759
4760         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4761         if (IS_ERR(res.label)) {
4762                 status = PTR_ERR(res.label);
4763                 goto out;
4764         }
4765
4766         nfs4_inode_make_writeable(inode);
4767         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4768
4769         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4770         if (!status) {
4771                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4772                                        NFS_INO_INVALID_DATA);
4773                 status = nfs_post_op_update_inode(inode, res.fattr);
4774                 if (!status)
4775                         nfs_setsecurity(inode, res.fattr, res.label);
4776         }
4777
4778
4779         nfs4_label_free(res.label);
4780
4781 out:
4782         nfs_free_fattr(res.fattr);
4783         return status;
4784 }
4785
4786 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4787 {
4788         struct nfs4_exception exception = {
4789                 .interruptible = true,
4790         };
4791         int err;
4792         do {
4793                 err = nfs4_handle_exception(NFS_SERVER(inode),
4794                                 _nfs4_proc_link(inode, dir, name),
4795                                 &exception);
4796         } while (exception.retry);
4797         return err;
4798 }
4799
4800 struct nfs4_createdata {
4801         struct rpc_message msg;
4802         struct nfs4_create_arg arg;
4803         struct nfs4_create_res res;
4804         struct nfs_fh fh;
4805         struct nfs_fattr fattr;
4806         struct nfs4_label *label;
4807 };
4808
4809 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4810                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4811 {
4812         struct nfs4_createdata *data;
4813
4814         data = kzalloc(sizeof(*data), GFP_KERNEL);
4815         if (data != NULL) {
4816                 struct nfs_server *server = NFS_SERVER(dir);
4817
4818                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4819                 if (IS_ERR(data->label))
4820                         goto out_free;
4821
4822                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4823                 data->msg.rpc_argp = &data->arg;
4824                 data->msg.rpc_resp = &data->res;
4825                 data->arg.dir_fh = NFS_FH(dir);
4826                 data->arg.server = server;
4827                 data->arg.name = name;
4828                 data->arg.attrs = sattr;
4829                 data->arg.ftype = ftype;
4830                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4831                 data->arg.umask = current_umask();
4832                 data->res.server = server;
4833                 data->res.fh = &data->fh;
4834                 data->res.fattr = &data->fattr;
4835                 data->res.label = data->label;
4836                 nfs_fattr_init(data->res.fattr);
4837         }
4838         return data;
4839 out_free:
4840         kfree(data);
4841         return NULL;
4842 }
4843
4844 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4845 {
4846         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4847                                     &data->arg.seq_args, &data->res.seq_res, 1);
4848         if (status == 0) {
4849                 spin_lock(&dir->i_lock);
4850                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4851                                 data->res.fattr->time_start,
4852                                 NFS_INO_INVALID_DATA);
4853                 /* Creating a directory bumps nlink in the parent */
4854                 if (data->arg.ftype == NF4DIR)
4855                         nfs4_inc_nlink_locked(dir);
4856                 spin_unlock(&dir->i_lock);
4857                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4858         }
4859         return status;
4860 }
4861
4862 static void nfs4_free_createdata(struct nfs4_createdata *data)
4863 {
4864         nfs4_label_free(data->label);
4865         kfree(data);
4866 }
4867
4868 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4869                 struct page *page, unsigned int len, struct iattr *sattr,
4870                 struct nfs4_label *label)
4871 {
4872         struct nfs4_createdata *data;
4873         int status = -ENAMETOOLONG;
4874
4875         if (len > NFS4_MAXPATHLEN)
4876                 goto out;
4877
4878         status = -ENOMEM;
4879         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4880         if (data == NULL)
4881                 goto out;
4882
4883         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4884         data->arg.u.symlink.pages = &page;
4885         data->arg.u.symlink.len = len;
4886         data->arg.label = label;
4887         
4888         status = nfs4_do_create(dir, dentry, data);
4889
4890         nfs4_free_createdata(data);
4891 out:
4892         return status;
4893 }
4894
4895 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4896                 struct page *page, unsigned int len, struct iattr *sattr)
4897 {
4898         struct nfs4_exception exception = {
4899                 .interruptible = true,
4900         };
4901         struct nfs4_label l, *label = NULL;
4902         int err;
4903
4904         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4905
4906         do {
4907                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4908                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4909                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4910                                 &exception);
4911         } while (exception.retry);
4912
4913         nfs4_label_release_security(label);
4914         return err;
4915 }
4916
4917 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4918                 struct iattr *sattr, struct nfs4_label *label)
4919 {
4920         struct nfs4_createdata *data;
4921         int status = -ENOMEM;
4922
4923         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4924         if (data == NULL)
4925                 goto out;
4926
4927         data->arg.label = label;
4928         status = nfs4_do_create(dir, dentry, data);
4929
4930         nfs4_free_createdata(data);
4931 out:
4932         return status;
4933 }
4934
4935 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4936                 struct iattr *sattr)
4937 {
4938         struct nfs_server *server = NFS_SERVER(dir);
4939         struct nfs4_exception exception = {
4940                 .interruptible = true,
4941         };
4942         struct nfs4_label l, *label = NULL;
4943         int err;
4944
4945         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4946
4947         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4948                 sattr->ia_mode &= ~current_umask();
4949         do {
4950                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4951                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4952                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4953                                 &exception);
4954         } while (exception.retry);
4955         nfs4_label_release_security(label);
4956
4957         return err;
4958 }
4959
4960 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4961                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4962 {
4963         struct inode            *dir = d_inode(dentry);
4964         struct nfs4_readdir_arg args = {
4965                 .fh = NFS_FH(dir),
4966                 .pages = pages,
4967                 .pgbase = 0,
4968                 .count = count,
4969                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4970                 .plus = plus,
4971         };
4972         struct nfs4_readdir_res res;
4973         struct rpc_message msg = {
4974                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4975                 .rpc_argp = &args,
4976                 .rpc_resp = &res,
4977                 .rpc_cred = cred,
4978         };
4979         int                     status;
4980
4981         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4982                         dentry,
4983                         (unsigned long long)cookie);
4984         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4985         res.pgbase = args.pgbase;
4986         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4987         if (status >= 0) {
4988                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4989                 status += args.pgbase;
4990         }
4991
4992         nfs_invalidate_atime(dir);
4993
4994         dprintk("%s: returns %d\n", __func__, status);
4995         return status;
4996 }
4997
4998 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4999                 u64 cookie, struct page **pages, unsigned int count, bool plus)
5000 {
5001         struct nfs4_exception exception = {
5002                 .interruptible = true,
5003         };
5004         int err;
5005         do {
5006                 err = _nfs4_proc_readdir(dentry, cred, cookie,
5007                                 pages, count, plus);
5008                 trace_nfs4_readdir(d_inode(dentry), err);
5009                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
5010                                 &exception);
5011         } while (exception.retry);
5012         return err;
5013 }
5014
5015 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5016                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5017 {
5018         struct nfs4_createdata *data;
5019         int mode = sattr->ia_mode;
5020         int status = -ENOMEM;
5021
5022         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5023         if (data == NULL)
5024                 goto out;
5025
5026         if (S_ISFIFO(mode))
5027                 data->arg.ftype = NF4FIFO;
5028         else if (S_ISBLK(mode)) {
5029                 data->arg.ftype = NF4BLK;
5030                 data->arg.u.device.specdata1 = MAJOR(rdev);
5031                 data->arg.u.device.specdata2 = MINOR(rdev);
5032         }
5033         else if (S_ISCHR(mode)) {
5034                 data->arg.ftype = NF4CHR;
5035                 data->arg.u.device.specdata1 = MAJOR(rdev);
5036                 data->arg.u.device.specdata2 = MINOR(rdev);
5037         } else if (!S_ISSOCK(mode)) {
5038                 status = -EINVAL;
5039                 goto out_free;
5040         }
5041
5042         data->arg.label = label;
5043         status = nfs4_do_create(dir, dentry, data);
5044 out_free:
5045         nfs4_free_createdata(data);
5046 out:
5047         return status;
5048 }
5049
5050 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5051                 struct iattr *sattr, dev_t rdev)
5052 {
5053         struct nfs_server *server = NFS_SERVER(dir);
5054         struct nfs4_exception exception = {
5055                 .interruptible = true,
5056         };
5057         struct nfs4_label l, *label = NULL;
5058         int err;
5059
5060         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5061
5062         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5063                 sattr->ia_mode &= ~current_umask();
5064         do {
5065                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5066                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5067                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5068                                 &exception);
5069         } while (exception.retry);
5070
5071         nfs4_label_release_security(label);
5072
5073         return err;
5074 }
5075
5076 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5077                  struct nfs_fsstat *fsstat)
5078 {
5079         struct nfs4_statfs_arg args = {
5080                 .fh = fhandle,
5081                 .bitmask = server->attr_bitmask,
5082         };
5083         struct nfs4_statfs_res res = {
5084                 .fsstat = fsstat,
5085         };
5086         struct rpc_message msg = {
5087                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5088                 .rpc_argp = &args,
5089                 .rpc_resp = &res,
5090         };
5091
5092         nfs_fattr_init(fsstat->fattr);
5093         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5094 }
5095
5096 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5097 {
5098         struct nfs4_exception exception = {
5099                 .interruptible = true,
5100         };
5101         int err;
5102         do {
5103                 err = nfs4_handle_exception(server,
5104                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5105                                 &exception);
5106         } while (exception.retry);
5107         return err;
5108 }
5109
5110 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5111                 struct nfs_fsinfo *fsinfo)
5112 {
5113         struct nfs4_fsinfo_arg args = {
5114                 .fh = fhandle,
5115                 .bitmask = server->attr_bitmask,
5116         };
5117         struct nfs4_fsinfo_res res = {
5118                 .fsinfo = fsinfo,
5119         };
5120         struct rpc_message msg = {
5121                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5122                 .rpc_argp = &args,
5123                 .rpc_resp = &res,
5124         };
5125
5126         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5127 }
5128
5129 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5130 {
5131         struct nfs4_exception exception = {
5132                 .interruptible = true,
5133         };
5134         int err;
5135
5136         do {
5137                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5138                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5139                 if (err == 0) {
5140                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5141                         break;
5142                 }
5143                 err = nfs4_handle_exception(server, err, &exception);
5144         } while (exception.retry);
5145         return err;
5146 }
5147
5148 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5149 {
5150         int error;
5151
5152         nfs_fattr_init(fsinfo->fattr);
5153         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5154         if (error == 0) {
5155                 /* block layout checks this! */
5156                 server->pnfs_blksize = fsinfo->blksize;
5157                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5158         }
5159
5160         return error;
5161 }
5162
5163 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5164                 struct nfs_pathconf *pathconf)
5165 {
5166         struct nfs4_pathconf_arg args = {
5167                 .fh = fhandle,
5168                 .bitmask = server->attr_bitmask,
5169         };
5170         struct nfs4_pathconf_res res = {
5171                 .pathconf = pathconf,
5172         };
5173         struct rpc_message msg = {
5174                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5175                 .rpc_argp = &args,
5176                 .rpc_resp = &res,
5177         };
5178
5179         /* None of the pathconf attributes are mandatory to implement */
5180         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5181                 memset(pathconf, 0, sizeof(*pathconf));
5182                 return 0;
5183         }
5184
5185         nfs_fattr_init(pathconf->fattr);
5186         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5187 }
5188
5189 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5190                 struct nfs_pathconf *pathconf)
5191 {
5192         struct nfs4_exception exception = {
5193                 .interruptible = true,
5194         };
5195         int err;
5196
5197         do {
5198                 err = nfs4_handle_exception(server,
5199                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5200                                 &exception);
5201         } while (exception.retry);
5202         return err;
5203 }
5204
5205 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5206                 const struct nfs_open_context *ctx,
5207                 const struct nfs_lock_context *l_ctx,
5208                 fmode_t fmode)
5209 {
5210         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5211 }
5212 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5213
5214 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5215                 const struct nfs_open_context *ctx,
5216                 const struct nfs_lock_context *l_ctx,
5217                 fmode_t fmode)
5218 {
5219         nfs4_stateid _current_stateid;
5220
5221         /* If the current stateid represents a lost lock, then exit */
5222         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5223                 return true;
5224         return nfs4_stateid_match(stateid, &_current_stateid);
5225 }
5226
5227 static bool nfs4_error_stateid_expired(int err)
5228 {
5229         switch (err) {
5230         case -NFS4ERR_DELEG_REVOKED:
5231         case -NFS4ERR_ADMIN_REVOKED:
5232         case -NFS4ERR_BAD_STATEID:
5233         case -NFS4ERR_STALE_STATEID:
5234         case -NFS4ERR_OLD_STATEID:
5235         case -NFS4ERR_OPENMODE:
5236         case -NFS4ERR_EXPIRED:
5237                 return true;
5238         }
5239         return false;
5240 }
5241
5242 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5243 {
5244         struct nfs_server *server = NFS_SERVER(hdr->inode);
5245
5246         trace_nfs4_read(hdr, task->tk_status);
5247         if (task->tk_status < 0) {
5248                 struct nfs4_exception exception = {
5249                         .inode = hdr->inode,
5250                         .state = hdr->args.context->state,
5251                         .stateid = &hdr->args.stateid,
5252                 };
5253                 task->tk_status = nfs4_async_handle_exception(task,
5254                                 server, task->tk_status, &exception);
5255                 if (exception.retry) {
5256                         rpc_restart_call_prepare(task);
5257                         return -EAGAIN;
5258                 }
5259         }
5260
5261         if (task->tk_status > 0)
5262                 renew_lease(server, hdr->timestamp);
5263         return 0;
5264 }
5265
5266 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5267                 struct nfs_pgio_args *args)
5268 {
5269
5270         if (!nfs4_error_stateid_expired(task->tk_status) ||
5271                 nfs4_stateid_is_current(&args->stateid,
5272                                 args->context,
5273                                 args->lock_context,
5274                                 FMODE_READ))
5275                 return false;
5276         rpc_restart_call_prepare(task);
5277         return true;
5278 }
5279
5280 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5281                                          struct nfs_pgio_header *hdr)
5282 {
5283         struct nfs_server *server = NFS_SERVER(hdr->inode);
5284         struct rpc_message *msg = &task->tk_msg;
5285
5286         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5287             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5288                 server->caps &= ~NFS_CAP_READ_PLUS;
5289                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5290                 rpc_restart_call_prepare(task);
5291                 return true;
5292         }
5293         return false;
5294 }
5295
5296 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5297 {
5298         dprintk("--> %s\n", __func__);
5299
5300         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5301                 return -EAGAIN;
5302         if (nfs4_read_stateid_changed(task, &hdr->args))
5303                 return -EAGAIN;
5304         if (nfs4_read_plus_not_supported(task, hdr))
5305                 return -EAGAIN;
5306         if (task->tk_status > 0)
5307                 nfs_invalidate_atime(hdr->inode);
5308         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5309                                     nfs4_read_done_cb(task, hdr);
5310 }
5311
5312 #ifdef CONFIG_NFS_V4_2
5313 static void nfs42_read_plus_support(struct nfs_server *server, struct rpc_message *msg)
5314 {
5315         if (server->caps & NFS_CAP_READ_PLUS)
5316                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5317         else
5318                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5319 }
5320 #else
5321 static void nfs42_read_plus_support(struct nfs_server *server, struct rpc_message *msg)
5322 {
5323         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5324 }
5325 #endif /* CONFIG_NFS_V4_2 */
5326
5327 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5328                                  struct rpc_message *msg)
5329 {
5330         hdr->timestamp   = jiffies;
5331         if (!hdr->pgio_done_cb)
5332                 hdr->pgio_done_cb = nfs4_read_done_cb;
5333         nfs42_read_plus_support(NFS_SERVER(hdr->inode), msg);
5334         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5335 }
5336
5337 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5338                                       struct nfs_pgio_header *hdr)
5339 {
5340         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5341                         &hdr->args.seq_args,
5342                         &hdr->res.seq_res,
5343                         task))
5344                 return 0;
5345         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5346                                 hdr->args.lock_context,
5347                                 hdr->rw_mode) == -EIO)
5348                 return -EIO;
5349         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5350                 return -EIO;
5351         return 0;
5352 }
5353
5354 static int nfs4_write_done_cb(struct rpc_task *task,
5355                               struct nfs_pgio_header *hdr)
5356 {
5357         struct inode *inode = hdr->inode;
5358
5359         trace_nfs4_write(hdr, task->tk_status);
5360         if (task->tk_status < 0) {
5361                 struct nfs4_exception exception = {
5362                         .inode = hdr->inode,
5363                         .state = hdr->args.context->state,
5364                         .stateid = &hdr->args.stateid,
5365                 };
5366                 task->tk_status = nfs4_async_handle_exception(task,
5367                                 NFS_SERVER(inode), task->tk_status,
5368                                 &exception);
5369                 if (exception.retry) {
5370                         rpc_restart_call_prepare(task);
5371                         return -EAGAIN;
5372                 }
5373         }
5374         if (task->tk_status >= 0) {
5375                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5376                 nfs_writeback_update_inode(hdr);
5377         }
5378         return 0;
5379 }
5380
5381 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5382                 struct nfs_pgio_args *args)
5383 {
5384
5385         if (!nfs4_error_stateid_expired(task->tk_status) ||
5386                 nfs4_stateid_is_current(&args->stateid,
5387                                 args->context,
5388                                 args->lock_context,
5389                                 FMODE_WRITE))
5390                 return false;
5391         rpc_restart_call_prepare(task);
5392         return true;
5393 }
5394
5395 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5396 {
5397         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5398                 return -EAGAIN;
5399         if (nfs4_write_stateid_changed(task, &hdr->args))
5400                 return -EAGAIN;
5401         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5402                 nfs4_write_done_cb(task, hdr);
5403 }
5404
5405 static
5406 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5407 {
5408         /* Don't request attributes for pNFS or O_DIRECT writes */
5409         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5410                 return false;
5411         /* Otherwise, request attributes if and only if we don't hold
5412          * a delegation
5413          */
5414         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5415 }
5416
5417 static void nfs4_bitmask_adjust(__u32 *bitmask, struct inode *inode,
5418                                 struct nfs_server *server,
5419                                 struct nfs4_label *label)
5420 {
5421
5422         unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
5423
5424         if ((cache_validity & NFS_INO_INVALID_DATA) ||
5425                 (cache_validity & NFS_INO_REVAL_PAGECACHE) ||
5426                 (cache_validity & NFS_INO_REVAL_FORCED) ||
5427                 (cache_validity & NFS_INO_INVALID_OTHER))
5428                 nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
5429
5430         if (cache_validity & NFS_INO_INVALID_ATIME)
5431                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5432         if (cache_validity & NFS_INO_INVALID_ACCESS)
5433                 bitmask[0] |= FATTR4_WORD1_MODE | FATTR4_WORD1_OWNER |
5434                                 FATTR4_WORD1_OWNER_GROUP;
5435         if (cache_validity & NFS_INO_INVALID_ACL)
5436                 bitmask[0] |= FATTR4_WORD0_ACL;
5437         if (cache_validity & NFS_INO_INVALID_LABEL)
5438                 bitmask[2] |= FATTR4_WORD2_SECURITY_LABEL;
5439         if (cache_validity & NFS_INO_INVALID_CTIME)
5440                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5441         if (cache_validity & NFS_INO_INVALID_MTIME)
5442                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5443         if (cache_validity & NFS_INO_INVALID_SIZE)
5444                 bitmask[0] |= FATTR4_WORD0_SIZE;
5445         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5446                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5447 }
5448
5449 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5450                                   struct rpc_message *msg,
5451                                   struct rpc_clnt **clnt)
5452 {
5453         struct nfs_server *server = NFS_SERVER(hdr->inode);
5454
5455         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5456                 hdr->args.bitmask = NULL;
5457                 hdr->res.fattr = NULL;
5458         } else {
5459                 hdr->args.bitmask = server->cache_consistency_bitmask;
5460                 nfs4_bitmask_adjust(hdr->args.bitmask, hdr->inode, server, NULL);
5461         }
5462
5463         if (!hdr->pgio_done_cb)
5464                 hdr->pgio_done_cb = nfs4_write_done_cb;
5465         hdr->res.server = server;
5466         hdr->timestamp   = jiffies;
5467
5468         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5469         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5470         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5471 }
5472
5473 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5474 {
5475         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5476                         &data->args.seq_args,
5477                         &data->res.seq_res,
5478                         task);
5479 }
5480
5481 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5482 {
5483         struct inode *inode = data->inode;
5484
5485         trace_nfs4_commit(data, task->tk_status);
5486         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5487                                     NULL, NULL) == -EAGAIN) {
5488                 rpc_restart_call_prepare(task);
5489                 return -EAGAIN;
5490         }
5491         return 0;
5492 }
5493
5494 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5495 {
5496         if (!nfs4_sequence_done(task, &data->res.seq_res))
5497                 return -EAGAIN;
5498         return data->commit_done_cb(task, data);
5499 }
5500
5501 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5502                                    struct rpc_clnt **clnt)
5503 {
5504         struct nfs_server *server = NFS_SERVER(data->inode);
5505
5506         if (data->commit_done_cb == NULL)
5507                 data->commit_done_cb = nfs4_commit_done_cb;
5508         data->res.server = server;
5509         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5510         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5511         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5512 }
5513
5514 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5515                                 struct nfs_commitres *res)
5516 {
5517         struct inode *dst_inode = file_inode(dst);
5518         struct nfs_server *server = NFS_SERVER(dst_inode);
5519         struct rpc_message msg = {
5520                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5521                 .rpc_argp = args,
5522                 .rpc_resp = res,
5523         };
5524
5525         args->fh = NFS_FH(dst_inode);
5526         return nfs4_call_sync(server->client, server, &msg,
5527                         &args->seq_args, &res->seq_res, 1);
5528 }
5529
5530 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5531 {
5532         struct nfs_commitargs args = {
5533                 .offset = offset,
5534                 .count = count,
5535         };
5536         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5537         struct nfs4_exception exception = { };
5538         int status;
5539
5540         do {
5541                 status = _nfs4_proc_commit(dst, &args, res);
5542                 status = nfs4_handle_exception(dst_server, status, &exception);
5543         } while (exception.retry);
5544
5545         return status;
5546 }
5547
5548 struct nfs4_renewdata {
5549         struct nfs_client       *client;
5550         unsigned long           timestamp;
5551 };
5552
5553 /*
5554  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5555  * standalone procedure for queueing an asynchronous RENEW.
5556  */
5557 static void nfs4_renew_release(void *calldata)
5558 {
5559         struct nfs4_renewdata *data = calldata;
5560         struct nfs_client *clp = data->client;
5561
5562         if (refcount_read(&clp->cl_count) > 1)
5563                 nfs4_schedule_state_renewal(clp);
5564         nfs_put_client(clp);
5565         kfree(data);
5566 }
5567
5568 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5569 {
5570         struct nfs4_renewdata *data = calldata;
5571         struct nfs_client *clp = data->client;
5572         unsigned long timestamp = data->timestamp;
5573
5574         trace_nfs4_renew_async(clp, task->tk_status);
5575         switch (task->tk_status) {
5576         case 0:
5577                 break;
5578         case -NFS4ERR_LEASE_MOVED:
5579                 nfs4_schedule_lease_moved_recovery(clp);
5580                 break;
5581         default:
5582                 /* Unless we're shutting down, schedule state recovery! */
5583                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5584                         return;
5585                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5586                         nfs4_schedule_lease_recovery(clp);
5587                         return;
5588                 }
5589                 nfs4_schedule_path_down_recovery(clp);
5590         }
5591         do_renew_lease(clp, timestamp);
5592 }
5593
5594 static const struct rpc_call_ops nfs4_renew_ops = {
5595         .rpc_call_done = nfs4_renew_done,
5596         .rpc_release = nfs4_renew_release,
5597 };
5598
5599 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5600 {
5601         struct rpc_message msg = {
5602                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5603                 .rpc_argp       = clp,
5604                 .rpc_cred       = cred,
5605         };
5606         struct nfs4_renewdata *data;
5607
5608         if (renew_flags == 0)
5609                 return 0;
5610         if (!refcount_inc_not_zero(&clp->cl_count))
5611                 return -EIO;
5612         data = kmalloc(sizeof(*data), GFP_NOFS);
5613         if (data == NULL) {
5614                 nfs_put_client(clp);
5615                 return -ENOMEM;
5616         }
5617         data->client = clp;
5618         data->timestamp = jiffies;
5619         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5620                         &nfs4_renew_ops, data);
5621 }
5622
5623 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5624 {
5625         struct rpc_message msg = {
5626                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5627                 .rpc_argp       = clp,
5628                 .rpc_cred       = cred,
5629         };
5630         unsigned long now = jiffies;
5631         int status;
5632
5633         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5634         if (status < 0)
5635                 return status;
5636         do_renew_lease(clp, now);
5637         return 0;
5638 }
5639
5640 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5641 {
5642         return server->caps & NFS_CAP_ACLS;
5643 }
5644
5645 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5646  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5647  * the stack.
5648  */
5649 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5650
5651 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5652                 struct page **pages)
5653 {
5654         struct page *newpage, **spages;
5655         int rc = 0;
5656         size_t len;
5657         spages = pages;
5658
5659         do {
5660                 len = min_t(size_t, PAGE_SIZE, buflen);
5661                 newpage = alloc_page(GFP_KERNEL);
5662
5663                 if (newpage == NULL)
5664                         goto unwind;
5665                 memcpy(page_address(newpage), buf, len);
5666                 buf += len;
5667                 buflen -= len;
5668                 *pages++ = newpage;
5669                 rc++;
5670         } while (buflen != 0);
5671
5672         return rc;
5673
5674 unwind:
5675         for(; rc > 0; rc--)
5676                 __free_page(spages[rc-1]);
5677         return -ENOMEM;
5678 }
5679
5680 struct nfs4_cached_acl {
5681         int cached;
5682         size_t len;
5683         char data[];
5684 };
5685
5686 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5687 {
5688         struct nfs_inode *nfsi = NFS_I(inode);
5689
5690         spin_lock(&inode->i_lock);
5691         kfree(nfsi->nfs4_acl);
5692         nfsi->nfs4_acl = acl;
5693         spin_unlock(&inode->i_lock);
5694 }
5695
5696 static void nfs4_zap_acl_attr(struct inode *inode)
5697 {
5698         nfs4_set_cached_acl(inode, NULL);
5699 }
5700
5701 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5702 {
5703         struct nfs_inode *nfsi = NFS_I(inode);
5704         struct nfs4_cached_acl *acl;
5705         int ret = -ENOENT;
5706
5707         spin_lock(&inode->i_lock);
5708         acl = nfsi->nfs4_acl;
5709         if (acl == NULL)
5710                 goto out;
5711         if (buf == NULL) /* user is just asking for length */
5712                 goto out_len;
5713         if (acl->cached == 0)
5714                 goto out;
5715         ret = -ERANGE; /* see getxattr(2) man page */
5716         if (acl->len > buflen)
5717                 goto out;
5718         memcpy(buf, acl->data, acl->len);
5719 out_len:
5720         ret = acl->len;
5721 out:
5722         spin_unlock(&inode->i_lock);
5723         return ret;
5724 }
5725
5726 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5727 {
5728         struct nfs4_cached_acl *acl;
5729         size_t buflen = sizeof(*acl) + acl_len;
5730
5731         if (buflen <= PAGE_SIZE) {
5732                 acl = kmalloc(buflen, GFP_KERNEL);
5733                 if (acl == NULL)
5734                         goto out;
5735                 acl->cached = 1;
5736                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5737         } else {
5738                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5739                 if (acl == NULL)
5740                         goto out;
5741                 acl->cached = 0;
5742         }
5743         acl->len = acl_len;
5744 out:
5745         nfs4_set_cached_acl(inode, acl);
5746 }
5747
5748 /*
5749  * The getxattr API returns the required buffer length when called with a
5750  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5751  * the required buf.  On a NULL buf, we send a page of data to the server
5752  * guessing that the ACL request can be serviced by a page. If so, we cache
5753  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5754  * the cache. If not so, we throw away the page, and cache the required
5755  * length. The next getxattr call will then produce another round trip to
5756  * the server, this time with the input buf of the required size.
5757  */
5758 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5759 {
5760         struct page **pages;
5761         struct nfs_getaclargs args = {
5762                 .fh = NFS_FH(inode),
5763                 .acl_len = buflen,
5764         };
5765         struct nfs_getaclres res = {
5766                 .acl_len = buflen,
5767         };
5768         struct rpc_message msg = {
5769                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5770                 .rpc_argp = &args,
5771                 .rpc_resp = &res,
5772         };
5773         unsigned int npages;
5774         int ret = -ENOMEM, i;
5775         struct nfs_server *server = NFS_SERVER(inode);
5776
5777         if (buflen == 0)
5778                 buflen = server->rsize;
5779
5780         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5781         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5782         if (!pages)
5783                 return -ENOMEM;
5784
5785         args.acl_pages = pages;
5786
5787         for (i = 0; i < npages; i++) {
5788                 pages[i] = alloc_page(GFP_KERNEL);
5789                 if (!pages[i])
5790                         goto out_free;
5791         }
5792
5793         /* for decoding across pages */
5794         res.acl_scratch = alloc_page(GFP_KERNEL);
5795         if (!res.acl_scratch)
5796                 goto out_free;
5797
5798         args.acl_len = npages * PAGE_SIZE;
5799
5800         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5801                 __func__, buf, buflen, npages, args.acl_len);
5802         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5803                              &msg, &args.seq_args, &res.seq_res, 0);
5804         if (ret)
5805                 goto out_free;
5806
5807         /* Handle the case where the passed-in buffer is too short */
5808         if (res.acl_flags & NFS4_ACL_TRUNC) {
5809                 /* Did the user only issue a request for the acl length? */
5810                 if (buf == NULL)
5811                         goto out_ok;
5812                 ret = -ERANGE;
5813                 goto out_free;
5814         }
5815         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5816         if (buf) {
5817                 if (res.acl_len > buflen) {
5818                         ret = -ERANGE;
5819                         goto out_free;
5820                 }
5821                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5822         }
5823 out_ok:
5824         ret = res.acl_len;
5825 out_free:
5826         for (i = 0; i < npages; i++)
5827                 if (pages[i])
5828                         __free_page(pages[i]);
5829         if (res.acl_scratch)
5830                 __free_page(res.acl_scratch);
5831         kfree(pages);
5832         return ret;
5833 }
5834
5835 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5836 {
5837         struct nfs4_exception exception = {
5838                 .interruptible = true,
5839         };
5840         ssize_t ret;
5841         do {
5842                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5843                 trace_nfs4_get_acl(inode, ret);
5844                 if (ret >= 0)
5845                         break;
5846                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5847         } while (exception.retry);
5848         return ret;
5849 }
5850
5851 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5852 {
5853         struct nfs_server *server = NFS_SERVER(inode);
5854         int ret;
5855
5856         if (!nfs4_server_supports_acls(server))
5857                 return -EOPNOTSUPP;
5858         ret = nfs_revalidate_inode(server, inode);
5859         if (ret < 0)
5860                 return ret;
5861         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5862                 nfs_zap_acl_cache(inode);
5863         ret = nfs4_read_cached_acl(inode, buf, buflen);
5864         if (ret != -ENOENT)
5865                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5866                  * but no cached acl data, just the acl length */
5867                 return ret;
5868         return nfs4_get_acl_uncached(inode, buf, buflen);
5869 }
5870
5871 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5872 {
5873         struct nfs_server *server = NFS_SERVER(inode);
5874         struct page *pages[NFS4ACL_MAXPAGES];
5875         struct nfs_setaclargs arg = {
5876                 .fh             = NFS_FH(inode),
5877                 .acl_pages      = pages,
5878                 .acl_len        = buflen,
5879         };
5880         struct nfs_setaclres res;
5881         struct rpc_message msg = {
5882                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5883                 .rpc_argp       = &arg,
5884                 .rpc_resp       = &res,
5885         };
5886         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5887         int ret, i;
5888
5889         if (!nfs4_server_supports_acls(server))
5890                 return -EOPNOTSUPP;
5891         if (npages > ARRAY_SIZE(pages))
5892                 return -ERANGE;
5893         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5894         if (i < 0)
5895                 return i;
5896         nfs4_inode_make_writeable(inode);
5897         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5898
5899         /*
5900          * Free each page after tx, so the only ref left is
5901          * held by the network stack
5902          */
5903         for (; i > 0; i--)
5904                 put_page(pages[i-1]);
5905
5906         /*
5907          * Acl update can result in inode attribute update.
5908          * so mark the attribute cache invalid.
5909          */
5910         spin_lock(&inode->i_lock);
5911         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5912                 | NFS_INO_INVALID_CTIME
5913                 | NFS_INO_REVAL_FORCED;
5914         spin_unlock(&inode->i_lock);
5915         nfs_access_zap_cache(inode);
5916         nfs_zap_acl_cache(inode);
5917         return ret;
5918 }
5919
5920 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5921 {
5922         struct nfs4_exception exception = { };
5923         int err;
5924         do {
5925                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5926                 trace_nfs4_set_acl(inode, err);
5927                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5928                                 &exception);
5929         } while (exception.retry);
5930         return err;
5931 }
5932
5933 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5934 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5935                                         size_t buflen)
5936 {
5937         struct nfs_server *server = NFS_SERVER(inode);
5938         struct nfs_fattr fattr;
5939         struct nfs4_label label = {0, 0, buflen, buf};
5940
5941         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5942         struct nfs4_getattr_arg arg = {
5943                 .fh             = NFS_FH(inode),
5944                 .bitmask        = bitmask,
5945         };
5946         struct nfs4_getattr_res res = {
5947                 .fattr          = &fattr,
5948                 .label          = &label,
5949                 .server         = server,
5950         };
5951         struct rpc_message msg = {
5952                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5953                 .rpc_argp       = &arg,
5954                 .rpc_resp       = &res,
5955         };
5956         int ret;
5957
5958         nfs_fattr_init(&fattr);
5959
5960         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5961         if (ret)
5962                 return ret;
5963         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5964                 return -ENOENT;
5965         return 0;
5966 }
5967
5968 static int nfs4_get_security_label(struct inode *inode, void *buf,
5969                                         size_t buflen)
5970 {
5971         struct nfs4_exception exception = {
5972                 .interruptible = true,
5973         };
5974         int err;
5975
5976         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5977                 return -EOPNOTSUPP;
5978
5979         do {
5980                 err = _nfs4_get_security_label(inode, buf, buflen);
5981                 trace_nfs4_get_security_label(inode, err);
5982                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5983                                 &exception);
5984         } while (exception.retry);
5985         return err;
5986 }
5987
5988 static int _nfs4_do_set_security_label(struct inode *inode,
5989                 struct nfs4_label *ilabel,
5990                 struct nfs_fattr *fattr,
5991                 struct nfs4_label *olabel)
5992 {
5993
5994         struct iattr sattr = {0};
5995         struct nfs_server *server = NFS_SERVER(inode);
5996         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5997         struct nfs_setattrargs arg = {
5998                 .fh             = NFS_FH(inode),
5999                 .iap            = &sattr,
6000                 .server         = server,
6001                 .bitmask        = bitmask,
6002                 .label          = ilabel,
6003         };
6004         struct nfs_setattrres res = {
6005                 .fattr          = fattr,
6006                 .label          = olabel,
6007                 .server         = server,
6008         };
6009         struct rpc_message msg = {
6010                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6011                 .rpc_argp       = &arg,
6012                 .rpc_resp       = &res,
6013         };
6014         int status;
6015
6016         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6017
6018         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6019         if (status)
6020                 dprintk("%s failed: %d\n", __func__, status);
6021
6022         return status;
6023 }
6024
6025 static int nfs4_do_set_security_label(struct inode *inode,
6026                 struct nfs4_label *ilabel,
6027                 struct nfs_fattr *fattr,
6028                 struct nfs4_label *olabel)
6029 {
6030         struct nfs4_exception exception = { };
6031         int err;
6032
6033         do {
6034                 err = _nfs4_do_set_security_label(inode, ilabel,
6035                                 fattr, olabel);
6036                 trace_nfs4_set_security_label(inode, err);
6037                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6038                                 &exception);
6039         } while (exception.retry);
6040         return err;
6041 }
6042
6043 static int
6044 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6045 {
6046         struct nfs4_label ilabel, *olabel = NULL;
6047         struct nfs_fattr fattr;
6048         int status;
6049
6050         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6051                 return -EOPNOTSUPP;
6052
6053         nfs_fattr_init(&fattr);
6054
6055         ilabel.pi = 0;
6056         ilabel.lfs = 0;
6057         ilabel.label = (char *)buf;
6058         ilabel.len = buflen;
6059
6060         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
6061         if (IS_ERR(olabel)) {
6062                 status = -PTR_ERR(olabel);
6063                 goto out;
6064         }
6065
6066         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
6067         if (status == 0)
6068                 nfs_setsecurity(inode, &fattr, olabel);
6069
6070         nfs4_label_free(olabel);
6071 out:
6072         return status;
6073 }
6074 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6075
6076
6077 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6078                                     nfs4_verifier *bootverf)
6079 {
6080         __be32 verf[2];
6081
6082         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6083                 /* An impossible timestamp guarantees this value
6084                  * will never match a generated boot time. */
6085                 verf[0] = cpu_to_be32(U32_MAX);
6086                 verf[1] = cpu_to_be32(U32_MAX);
6087         } else {
6088                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6089                 u64 ns = ktime_to_ns(nn->boot_time);
6090
6091                 verf[0] = cpu_to_be32(ns >> 32);
6092                 verf[1] = cpu_to_be32(ns);
6093         }
6094         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6095 }
6096
6097 static size_t
6098 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6099 {
6100         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6101         struct nfs_netns_client *nn_clp = nn->nfs_client;
6102         const char *id;
6103
6104         buf[0] = '\0';
6105
6106         if (nn_clp) {
6107                 rcu_read_lock();
6108                 id = rcu_dereference(nn_clp->identifier);
6109                 if (id)
6110                         strscpy(buf, id, buflen);
6111                 rcu_read_unlock();
6112         }
6113
6114         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6115                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6116
6117         return strlen(buf);
6118 }
6119
6120 static int
6121 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6122 {
6123         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6124         size_t buflen;
6125         size_t len;
6126         char *str;
6127
6128         if (clp->cl_owner_id != NULL)
6129                 return 0;
6130
6131         rcu_read_lock();
6132         len = 14 +
6133                 strlen(clp->cl_rpcclient->cl_nodename) +
6134                 1 +
6135                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6136                 1;
6137         rcu_read_unlock();
6138
6139         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6140         if (buflen)
6141                 len += buflen + 1;
6142
6143         if (len > NFS4_OPAQUE_LIMIT + 1)
6144                 return -EINVAL;
6145
6146         /*
6147          * Since this string is allocated at mount time, and held until the
6148          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6149          * about a memory-reclaim deadlock.
6150          */
6151         str = kmalloc(len, GFP_KERNEL);
6152         if (!str)
6153                 return -ENOMEM;
6154
6155         rcu_read_lock();
6156         if (buflen)
6157                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6158                           clp->cl_rpcclient->cl_nodename, buf,
6159                           rpc_peeraddr2str(clp->cl_rpcclient,
6160                                            RPC_DISPLAY_ADDR));
6161         else
6162                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6163                           clp->cl_rpcclient->cl_nodename,
6164                           rpc_peeraddr2str(clp->cl_rpcclient,
6165                                            RPC_DISPLAY_ADDR));
6166         rcu_read_unlock();
6167
6168         clp->cl_owner_id = str;
6169         return 0;
6170 }
6171
6172 static int
6173 nfs4_init_uniform_client_string(struct nfs_client *clp)
6174 {
6175         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6176         size_t buflen;
6177         size_t len;
6178         char *str;
6179
6180         if (clp->cl_owner_id != NULL)
6181                 return 0;
6182
6183         len = 10 + 10 + 1 + 10 + 1 +
6184                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6185
6186         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6187         if (buflen)
6188                 len += buflen + 1;
6189
6190         if (len > NFS4_OPAQUE_LIMIT + 1)
6191                 return -EINVAL;
6192
6193         /*
6194          * Since this string is allocated at mount time, and held until the
6195          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6196          * about a memory-reclaim deadlock.
6197          */
6198         str = kmalloc(len, GFP_KERNEL);
6199         if (!str)
6200                 return -ENOMEM;
6201
6202         if (buflen)
6203                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6204                           clp->rpc_ops->version, clp->cl_minorversion,
6205                           buf, clp->cl_rpcclient->cl_nodename);
6206         else
6207                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6208                           clp->rpc_ops->version, clp->cl_minorversion,
6209                           clp->cl_rpcclient->cl_nodename);
6210         clp->cl_owner_id = str;
6211         return 0;
6212 }
6213
6214 /*
6215  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6216  * services.  Advertise one based on the address family of the
6217  * clientaddr.
6218  */
6219 static unsigned int
6220 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6221 {
6222         if (strchr(clp->cl_ipaddr, ':') != NULL)
6223                 return scnprintf(buf, len, "tcp6");
6224         else
6225                 return scnprintf(buf, len, "tcp");
6226 }
6227
6228 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6229 {
6230         struct nfs4_setclientid *sc = calldata;
6231
6232         if (task->tk_status == 0)
6233                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6234 }
6235
6236 static const struct rpc_call_ops nfs4_setclientid_ops = {
6237         .rpc_call_done = nfs4_setclientid_done,
6238 };
6239
6240 /**
6241  * nfs4_proc_setclientid - Negotiate client ID
6242  * @clp: state data structure
6243  * @program: RPC program for NFSv4 callback service
6244  * @port: IP port number for NFS4 callback service
6245  * @cred: credential to use for this call
6246  * @res: where to place the result
6247  *
6248  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6249  */
6250 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6251                 unsigned short port, const struct cred *cred,
6252                 struct nfs4_setclientid_res *res)
6253 {
6254         nfs4_verifier sc_verifier;
6255         struct nfs4_setclientid setclientid = {
6256                 .sc_verifier = &sc_verifier,
6257                 .sc_prog = program,
6258                 .sc_clnt = clp,
6259         };
6260         struct rpc_message msg = {
6261                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6262                 .rpc_argp = &setclientid,
6263                 .rpc_resp = res,
6264                 .rpc_cred = cred,
6265         };
6266         struct rpc_task_setup task_setup_data = {
6267                 .rpc_client = clp->cl_rpcclient,
6268                 .rpc_message = &msg,
6269                 .callback_ops = &nfs4_setclientid_ops,
6270                 .callback_data = &setclientid,
6271                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6272         };
6273         unsigned long now = jiffies;
6274         int status;
6275
6276         /* nfs_client_id4 */
6277         nfs4_init_boot_verifier(clp, &sc_verifier);
6278
6279         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6280                 status = nfs4_init_uniform_client_string(clp);
6281         else
6282                 status = nfs4_init_nonuniform_client_string(clp);
6283
6284         if (status)
6285                 goto out;
6286
6287         /* cb_client4 */
6288         setclientid.sc_netid_len =
6289                                 nfs4_init_callback_netid(clp,
6290                                                 setclientid.sc_netid,
6291                                                 sizeof(setclientid.sc_netid));
6292         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6293                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6294                                 clp->cl_ipaddr, port >> 8, port & 255);
6295
6296         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6297                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6298                 clp->cl_owner_id);
6299
6300         status = nfs4_call_sync_custom(&task_setup_data);
6301         if (setclientid.sc_cred) {
6302                 kfree(clp->cl_acceptor);
6303                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6304                 put_rpccred(setclientid.sc_cred);
6305         }
6306
6307         if (status == 0)
6308                 do_renew_lease(clp, now);
6309 out:
6310         trace_nfs4_setclientid(clp, status);
6311         dprintk("NFS reply setclientid: %d\n", status);
6312         return status;
6313 }
6314
6315 /**
6316  * nfs4_proc_setclientid_confirm - Confirm client ID
6317  * @clp: state data structure
6318  * @arg: result of a previous SETCLIENTID
6319  * @cred: credential to use for this call
6320  *
6321  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6322  */
6323 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6324                 struct nfs4_setclientid_res *arg,
6325                 const struct cred *cred)
6326 {
6327         struct rpc_message msg = {
6328                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6329                 .rpc_argp = arg,
6330                 .rpc_cred = cred,
6331         };
6332         int status;
6333
6334         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6335                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6336                 clp->cl_clientid);
6337         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6338                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6339         trace_nfs4_setclientid_confirm(clp, status);
6340         dprintk("NFS reply setclientid_confirm: %d\n", status);
6341         return status;
6342 }
6343
6344 struct nfs4_delegreturndata {
6345         struct nfs4_delegreturnargs args;
6346         struct nfs4_delegreturnres res;
6347         struct nfs_fh fh;
6348         nfs4_stateid stateid;
6349         unsigned long timestamp;
6350         struct {
6351                 struct nfs4_layoutreturn_args arg;
6352                 struct nfs4_layoutreturn_res res;
6353                 struct nfs4_xdr_opaque_data ld_private;
6354                 u32 roc_barrier;
6355                 bool roc;
6356         } lr;
6357         struct nfs_fattr fattr;
6358         int rpc_status;
6359         struct inode *inode;
6360 };
6361
6362 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6363 {
6364         struct nfs4_delegreturndata *data = calldata;
6365         struct nfs4_exception exception = {
6366                 .inode = data->inode,
6367                 .stateid = &data->stateid,
6368         };
6369
6370         if (!nfs4_sequence_done(task, &data->res.seq_res))
6371                 return;
6372
6373         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6374
6375         /* Handle Layoutreturn errors */
6376         if (pnfs_roc_done(task, data->inode,
6377                                 &data->args.lr_args,
6378                                 &data->res.lr_res,
6379                                 &data->res.lr_ret) == -EAGAIN)
6380                 goto out_restart;
6381
6382         switch (task->tk_status) {
6383         case 0:
6384                 renew_lease(data->res.server, data->timestamp);
6385                 break;
6386         case -NFS4ERR_ADMIN_REVOKED:
6387         case -NFS4ERR_DELEG_REVOKED:
6388         case -NFS4ERR_EXPIRED:
6389                 nfs4_free_revoked_stateid(data->res.server,
6390                                 data->args.stateid,
6391                                 task->tk_msg.rpc_cred);
6392                 fallthrough;
6393         case -NFS4ERR_BAD_STATEID:
6394         case -NFS4ERR_STALE_STATEID:
6395         case -ETIMEDOUT:
6396                 task->tk_status = 0;
6397                 break;
6398         case -NFS4ERR_OLD_STATEID:
6399                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6400                         nfs4_stateid_seqid_inc(&data->stateid);
6401                 if (data->args.bitmask) {
6402                         data->args.bitmask = NULL;
6403                         data->res.fattr = NULL;
6404                 }
6405                 goto out_restart;
6406         case -NFS4ERR_ACCESS:
6407                 if (data->args.bitmask) {
6408                         data->args.bitmask = NULL;
6409                         data->res.fattr = NULL;
6410                         goto out_restart;
6411                 }
6412                 fallthrough;
6413         default:
6414                 task->tk_status = nfs4_async_handle_exception(task,
6415                                 data->res.server, task->tk_status,
6416                                 &exception);
6417                 if (exception.retry)
6418                         goto out_restart;
6419         }
6420         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6421         data->rpc_status = task->tk_status;
6422         return;
6423 out_restart:
6424         task->tk_status = 0;
6425         rpc_restart_call_prepare(task);
6426 }
6427
6428 static void nfs4_delegreturn_release(void *calldata)
6429 {
6430         struct nfs4_delegreturndata *data = calldata;
6431         struct inode *inode = data->inode;
6432
6433         if (inode) {
6434                 if (data->lr.roc)
6435                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6436                                         data->res.lr_ret);
6437                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6438                 nfs_iput_and_deactive(inode);
6439         }
6440         kfree(calldata);
6441 }
6442
6443 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6444 {
6445         struct nfs4_delegreturndata *d_data;
6446         struct pnfs_layout_hdr *lo;
6447
6448         d_data = (struct nfs4_delegreturndata *)data;
6449
6450         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6451                 nfs4_sequence_done(task, &d_data->res.seq_res);
6452                 return;
6453         }
6454
6455         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6456         if (lo && !pnfs_layout_is_valid(lo)) {
6457                 d_data->args.lr_args = NULL;
6458                 d_data->res.lr_res = NULL;
6459         }
6460
6461         nfs4_setup_sequence(d_data->res.server->nfs_client,
6462                         &d_data->args.seq_args,
6463                         &d_data->res.seq_res,
6464                         task);
6465 }
6466
6467 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6468         .rpc_call_prepare = nfs4_delegreturn_prepare,
6469         .rpc_call_done = nfs4_delegreturn_done,
6470         .rpc_release = nfs4_delegreturn_release,
6471 };
6472
6473 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6474 {
6475         struct nfs4_delegreturndata *data;
6476         struct nfs_server *server = NFS_SERVER(inode);
6477         struct rpc_task *task;
6478         struct rpc_message msg = {
6479                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6480                 .rpc_cred = cred,
6481         };
6482         struct rpc_task_setup task_setup_data = {
6483                 .rpc_client = server->client,
6484                 .rpc_message = &msg,
6485                 .callback_ops = &nfs4_delegreturn_ops,
6486                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6487         };
6488         int status = 0;
6489
6490         data = kzalloc(sizeof(*data), GFP_NOFS);
6491         if (data == NULL)
6492                 return -ENOMEM;
6493         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6494
6495         nfs4_state_protect(server->nfs_client,
6496                         NFS_SP4_MACH_CRED_CLEANUP,
6497                         &task_setup_data.rpc_client, &msg);
6498
6499         data->args.fhandle = &data->fh;
6500         data->args.stateid = &data->stateid;
6501         data->args.bitmask = server->cache_consistency_bitmask;
6502         nfs4_bitmask_adjust(data->args.bitmask, inode, server, NULL);
6503         nfs_copy_fh(&data->fh, NFS_FH(inode));
6504         nfs4_stateid_copy(&data->stateid, stateid);
6505         data->res.fattr = &data->fattr;
6506         data->res.server = server;
6507         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6508         data->lr.arg.ld_private = &data->lr.ld_private;
6509         nfs_fattr_init(data->res.fattr);
6510         data->timestamp = jiffies;
6511         data->rpc_status = 0;
6512         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6513         data->inode = nfs_igrab_and_active(inode);
6514         if (data->inode) {
6515                 if (data->lr.roc) {
6516                         data->args.lr_args = &data->lr.arg;
6517                         data->res.lr_res = &data->lr.res;
6518                 }
6519         } else if (data->lr.roc) {
6520                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6521                 data->lr.roc = false;
6522         }
6523
6524         task_setup_data.callback_data = data;
6525         msg.rpc_argp = &data->args;
6526         msg.rpc_resp = &data->res;
6527         task = rpc_run_task(&task_setup_data);
6528         if (IS_ERR(task))
6529                 return PTR_ERR(task);
6530         if (!issync)
6531                 goto out;
6532         status = rpc_wait_for_completion_task(task);
6533         if (status != 0)
6534                 goto out;
6535         status = data->rpc_status;
6536 out:
6537         rpc_put_task(task);
6538         return status;
6539 }
6540
6541 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6542 {
6543         struct nfs_server *server = NFS_SERVER(inode);
6544         struct nfs4_exception exception = { };
6545         int err;
6546         do {
6547                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6548                 trace_nfs4_delegreturn(inode, stateid, err);
6549                 switch (err) {
6550                         case -NFS4ERR_STALE_STATEID:
6551                         case -NFS4ERR_EXPIRED:
6552                         case 0:
6553                                 return 0;
6554                 }
6555                 err = nfs4_handle_exception(server, err, &exception);
6556         } while (exception.retry);
6557         return err;
6558 }
6559
6560 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6561 {
6562         struct inode *inode = state->inode;
6563         struct nfs_server *server = NFS_SERVER(inode);
6564         struct nfs_client *clp = server->nfs_client;
6565         struct nfs_lockt_args arg = {
6566                 .fh = NFS_FH(inode),
6567                 .fl = request,
6568         };
6569         struct nfs_lockt_res res = {
6570                 .denied = request,
6571         };
6572         struct rpc_message msg = {
6573                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6574                 .rpc_argp       = &arg,
6575                 .rpc_resp       = &res,
6576                 .rpc_cred       = state->owner->so_cred,
6577         };
6578         struct nfs4_lock_state *lsp;
6579         int status;
6580
6581         arg.lock_owner.clientid = clp->cl_clientid;
6582         status = nfs4_set_lock_state(state, request);
6583         if (status != 0)
6584                 goto out;
6585         lsp = request->fl_u.nfs4_fl.owner;
6586         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6587         arg.lock_owner.s_dev = server->s_dev;
6588         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6589         switch (status) {
6590                 case 0:
6591                         request->fl_type = F_UNLCK;
6592                         break;
6593                 case -NFS4ERR_DENIED:
6594                         status = 0;
6595         }
6596         request->fl_ops->fl_release_private(request);
6597         request->fl_ops = NULL;
6598 out:
6599         return status;
6600 }
6601
6602 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6603 {
6604         struct nfs4_exception exception = {
6605                 .interruptible = true,
6606         };
6607         int err;
6608
6609         do {
6610                 err = _nfs4_proc_getlk(state, cmd, request);
6611                 trace_nfs4_get_lock(request, state, cmd, err);
6612                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6613                                 &exception);
6614         } while (exception.retry);
6615         return err;
6616 }
6617
6618 /*
6619  * Update the seqid of a lock stateid after receiving
6620  * NFS4ERR_OLD_STATEID
6621  */
6622 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6623                 struct nfs4_lock_state *lsp)
6624 {
6625         struct nfs4_state *state = lsp->ls_state;
6626         bool ret = false;
6627
6628         spin_lock(&state->state_lock);
6629         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6630                 goto out;
6631         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6632                 nfs4_stateid_seqid_inc(dst);
6633         else
6634                 dst->seqid = lsp->ls_stateid.seqid;
6635         ret = true;
6636 out:
6637         spin_unlock(&state->state_lock);
6638         return ret;
6639 }
6640
6641 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6642                 struct nfs4_lock_state *lsp)
6643 {
6644         struct nfs4_state *state = lsp->ls_state;
6645         bool ret;
6646
6647         spin_lock(&state->state_lock);
6648         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6649         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6650         spin_unlock(&state->state_lock);
6651         return ret;
6652 }
6653
6654 struct nfs4_unlockdata {
6655         struct nfs_locku_args arg;
6656         struct nfs_locku_res res;
6657         struct nfs4_lock_state *lsp;
6658         struct nfs_open_context *ctx;
6659         struct nfs_lock_context *l_ctx;
6660         struct file_lock fl;
6661         struct nfs_server *server;
6662         unsigned long timestamp;
6663 };
6664
6665 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6666                 struct nfs_open_context *ctx,
6667                 struct nfs4_lock_state *lsp,
6668                 struct nfs_seqid *seqid)
6669 {
6670         struct nfs4_unlockdata *p;
6671         struct nfs4_state *state = lsp->ls_state;
6672         struct inode *inode = state->inode;
6673
6674         p = kzalloc(sizeof(*p), GFP_NOFS);
6675         if (p == NULL)
6676                 return NULL;
6677         p->arg.fh = NFS_FH(inode);
6678         p->arg.fl = &p->fl;
6679         p->arg.seqid = seqid;
6680         p->res.seqid = seqid;
6681         p->lsp = lsp;
6682         /* Ensure we don't close file until we're done freeing locks! */
6683         p->ctx = get_nfs_open_context(ctx);
6684         p->l_ctx = nfs_get_lock_context(ctx);
6685         locks_init_lock(&p->fl);
6686         locks_copy_lock(&p->fl, fl);
6687         p->server = NFS_SERVER(inode);
6688         spin_lock(&state->state_lock);
6689         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6690         spin_unlock(&state->state_lock);
6691         return p;
6692 }
6693
6694 static void nfs4_locku_release_calldata(void *data)
6695 {
6696         struct nfs4_unlockdata *calldata = data;
6697         nfs_free_seqid(calldata->arg.seqid);
6698         nfs4_put_lock_state(calldata->lsp);
6699         nfs_put_lock_context(calldata->l_ctx);
6700         put_nfs_open_context(calldata->ctx);
6701         kfree(calldata);
6702 }
6703
6704 static void nfs4_locku_done(struct rpc_task *task, void *data)
6705 {
6706         struct nfs4_unlockdata *calldata = data;
6707         struct nfs4_exception exception = {
6708                 .inode = calldata->lsp->ls_state->inode,
6709                 .stateid = &calldata->arg.stateid,
6710         };
6711
6712         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6713                 return;
6714         switch (task->tk_status) {
6715                 case 0:
6716                         renew_lease(calldata->server, calldata->timestamp);
6717                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6718                         if (nfs4_update_lock_stateid(calldata->lsp,
6719                                         &calldata->res.stateid))
6720                                 break;
6721                         fallthrough;
6722                 case -NFS4ERR_ADMIN_REVOKED:
6723                 case -NFS4ERR_EXPIRED:
6724                         nfs4_free_revoked_stateid(calldata->server,
6725                                         &calldata->arg.stateid,
6726                                         task->tk_msg.rpc_cred);
6727                         fallthrough;
6728                 case -NFS4ERR_BAD_STATEID:
6729                 case -NFS4ERR_STALE_STATEID:
6730                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6731                                                 calldata->lsp))
6732                                 rpc_restart_call_prepare(task);
6733                         break;
6734                 case -NFS4ERR_OLD_STATEID:
6735                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6736                                                 calldata->lsp))
6737                                 rpc_restart_call_prepare(task);
6738                         break;
6739                 default:
6740                         task->tk_status = nfs4_async_handle_exception(task,
6741                                         calldata->server, task->tk_status,
6742                                         &exception);
6743                         if (exception.retry)
6744                                 rpc_restart_call_prepare(task);
6745         }
6746         nfs_release_seqid(calldata->arg.seqid);
6747 }
6748
6749 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6750 {
6751         struct nfs4_unlockdata *calldata = data;
6752
6753         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6754                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6755                 return;
6756
6757         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6758                 goto out_wait;
6759         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6760                 /* Note: exit _without_ running nfs4_locku_done */
6761                 goto out_no_action;
6762         }
6763         calldata->timestamp = jiffies;
6764         if (nfs4_setup_sequence(calldata->server->nfs_client,
6765                                 &calldata->arg.seq_args,
6766                                 &calldata->res.seq_res,
6767                                 task) != 0)
6768                 nfs_release_seqid(calldata->arg.seqid);
6769         return;
6770 out_no_action:
6771         task->tk_action = NULL;
6772 out_wait:
6773         nfs4_sequence_done(task, &calldata->res.seq_res);
6774 }
6775
6776 static const struct rpc_call_ops nfs4_locku_ops = {
6777         .rpc_call_prepare = nfs4_locku_prepare,
6778         .rpc_call_done = nfs4_locku_done,
6779         .rpc_release = nfs4_locku_release_calldata,
6780 };
6781
6782 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6783                 struct nfs_open_context *ctx,
6784                 struct nfs4_lock_state *lsp,
6785                 struct nfs_seqid *seqid)
6786 {
6787         struct nfs4_unlockdata *data;
6788         struct rpc_message msg = {
6789                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6790                 .rpc_cred = ctx->cred,
6791         };
6792         struct rpc_task_setup task_setup_data = {
6793                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6794                 .rpc_message = &msg,
6795                 .callback_ops = &nfs4_locku_ops,
6796                 .workqueue = nfsiod_workqueue,
6797                 .flags = RPC_TASK_ASYNC,
6798         };
6799
6800         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6801                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6802
6803         /* Ensure this is an unlock - when canceling a lock, the
6804          * canceled lock is passed in, and it won't be an unlock.
6805          */
6806         fl->fl_type = F_UNLCK;
6807         if (fl->fl_flags & FL_CLOSE)
6808                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6809
6810         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6811         if (data == NULL) {
6812                 nfs_free_seqid(seqid);
6813                 return ERR_PTR(-ENOMEM);
6814         }
6815
6816         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6817         msg.rpc_argp = &data->arg;
6818         msg.rpc_resp = &data->res;
6819         task_setup_data.callback_data = data;
6820         return rpc_run_task(&task_setup_data);
6821 }
6822
6823 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6824 {
6825         struct inode *inode = state->inode;
6826         struct nfs4_state_owner *sp = state->owner;
6827         struct nfs_inode *nfsi = NFS_I(inode);
6828         struct nfs_seqid *seqid;
6829         struct nfs4_lock_state *lsp;
6830         struct rpc_task *task;
6831         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6832         int status = 0;
6833         unsigned char fl_flags = request->fl_flags;
6834
6835         status = nfs4_set_lock_state(state, request);
6836         /* Unlock _before_ we do the RPC call */
6837         request->fl_flags |= FL_EXISTS;
6838         /* Exclude nfs_delegation_claim_locks() */
6839         mutex_lock(&sp->so_delegreturn_mutex);
6840         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6841         down_read(&nfsi->rwsem);
6842         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6843                 up_read(&nfsi->rwsem);
6844                 mutex_unlock(&sp->so_delegreturn_mutex);
6845                 goto out;
6846         }
6847         up_read(&nfsi->rwsem);
6848         mutex_unlock(&sp->so_delegreturn_mutex);
6849         if (status != 0)
6850                 goto out;
6851         /* Is this a delegated lock? */
6852         lsp = request->fl_u.nfs4_fl.owner;
6853         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6854                 goto out;
6855         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6856         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6857         status = -ENOMEM;
6858         if (IS_ERR(seqid))
6859                 goto out;
6860         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6861         status = PTR_ERR(task);
6862         if (IS_ERR(task))
6863                 goto out;
6864         status = rpc_wait_for_completion_task(task);
6865         rpc_put_task(task);
6866 out:
6867         request->fl_flags = fl_flags;
6868         trace_nfs4_unlock(request, state, F_SETLK, status);
6869         return status;
6870 }
6871
6872 struct nfs4_lockdata {
6873         struct nfs_lock_args arg;
6874         struct nfs_lock_res res;
6875         struct nfs4_lock_state *lsp;
6876         struct nfs_open_context *ctx;
6877         struct file_lock fl;
6878         unsigned long timestamp;
6879         int rpc_status;
6880         int cancelled;
6881         struct nfs_server *server;
6882 };
6883
6884 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6885                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6886                 gfp_t gfp_mask)
6887 {
6888         struct nfs4_lockdata *p;
6889         struct inode *inode = lsp->ls_state->inode;
6890         struct nfs_server *server = NFS_SERVER(inode);
6891         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6892
6893         p = kzalloc(sizeof(*p), gfp_mask);
6894         if (p == NULL)
6895                 return NULL;
6896
6897         p->arg.fh = NFS_FH(inode);
6898         p->arg.fl = &p->fl;
6899         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6900         if (IS_ERR(p->arg.open_seqid))
6901                 goto out_free;
6902         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6903         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6904         if (IS_ERR(p->arg.lock_seqid))
6905                 goto out_free_seqid;
6906         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6907         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6908         p->arg.lock_owner.s_dev = server->s_dev;
6909         p->res.lock_seqid = p->arg.lock_seqid;
6910         p->lsp = lsp;
6911         p->server = server;
6912         p->ctx = get_nfs_open_context(ctx);
6913         locks_init_lock(&p->fl);
6914         locks_copy_lock(&p->fl, fl);
6915         return p;
6916 out_free_seqid:
6917         nfs_free_seqid(p->arg.open_seqid);
6918 out_free:
6919         kfree(p);
6920         return NULL;
6921 }
6922
6923 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6924 {
6925         struct nfs4_lockdata *data = calldata;
6926         struct nfs4_state *state = data->lsp->ls_state;
6927
6928         dprintk("%s: begin!\n", __func__);
6929         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6930                 goto out_wait;
6931         /* Do we need to do an open_to_lock_owner? */
6932         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6933                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6934                         goto out_release_lock_seqid;
6935                 }
6936                 nfs4_stateid_copy(&data->arg.open_stateid,
6937                                 &state->open_stateid);
6938                 data->arg.new_lock_owner = 1;
6939                 data->res.open_seqid = data->arg.open_seqid;
6940         } else {
6941                 data->arg.new_lock_owner = 0;
6942                 nfs4_stateid_copy(&data->arg.lock_stateid,
6943                                 &data->lsp->ls_stateid);
6944         }
6945         if (!nfs4_valid_open_stateid(state)) {
6946                 data->rpc_status = -EBADF;
6947                 task->tk_action = NULL;
6948                 goto out_release_open_seqid;
6949         }
6950         data->timestamp = jiffies;
6951         if (nfs4_setup_sequence(data->server->nfs_client,
6952                                 &data->arg.seq_args,
6953                                 &data->res.seq_res,
6954                                 task) == 0)
6955                 return;
6956 out_release_open_seqid:
6957         nfs_release_seqid(data->arg.open_seqid);
6958 out_release_lock_seqid:
6959         nfs_release_seqid(data->arg.lock_seqid);
6960 out_wait:
6961         nfs4_sequence_done(task, &data->res.seq_res);
6962         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6963 }
6964
6965 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6966 {
6967         struct nfs4_lockdata *data = calldata;
6968         struct nfs4_lock_state *lsp = data->lsp;
6969
6970         dprintk("%s: begin!\n", __func__);
6971
6972         if (!nfs4_sequence_done(task, &data->res.seq_res))
6973                 return;
6974
6975         data->rpc_status = task->tk_status;
6976         switch (task->tk_status) {
6977         case 0:
6978                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6979                                 data->timestamp);
6980                 if (data->arg.new_lock && !data->cancelled) {
6981                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6982                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6983                                 goto out_restart;
6984                 }
6985                 if (data->arg.new_lock_owner != 0) {
6986                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6987                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6988                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6989                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6990                         goto out_restart;
6991                 break;
6992         case -NFS4ERR_BAD_STATEID:
6993         case -NFS4ERR_OLD_STATEID:
6994         case -NFS4ERR_STALE_STATEID:
6995         case -NFS4ERR_EXPIRED:
6996                 if (data->arg.new_lock_owner != 0) {
6997                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6998                                                 &lsp->ls_state->open_stateid))
6999                                 goto out_restart;
7000                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7001                                                 &lsp->ls_stateid))
7002                                 goto out_restart;
7003         }
7004 out_done:
7005         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
7006         return;
7007 out_restart:
7008         if (!data->cancelled)
7009                 rpc_restart_call_prepare(task);
7010         goto out_done;
7011 }
7012
7013 static void nfs4_lock_release(void *calldata)
7014 {
7015         struct nfs4_lockdata *data = calldata;
7016
7017         dprintk("%s: begin!\n", __func__);
7018         nfs_free_seqid(data->arg.open_seqid);
7019         if (data->cancelled && data->rpc_status == 0) {
7020                 struct rpc_task *task;
7021                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7022                                 data->arg.lock_seqid);
7023                 if (!IS_ERR(task))
7024                         rpc_put_task_async(task);
7025                 dprintk("%s: cancelling lock!\n", __func__);
7026         } else
7027                 nfs_free_seqid(data->arg.lock_seqid);
7028         nfs4_put_lock_state(data->lsp);
7029         put_nfs_open_context(data->ctx);
7030         kfree(data);
7031         dprintk("%s: done!\n", __func__);
7032 }
7033
7034 static const struct rpc_call_ops nfs4_lock_ops = {
7035         .rpc_call_prepare = nfs4_lock_prepare,
7036         .rpc_call_done = nfs4_lock_done,
7037         .rpc_release = nfs4_lock_release,
7038 };
7039
7040 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7041 {
7042         switch (error) {
7043         case -NFS4ERR_ADMIN_REVOKED:
7044         case -NFS4ERR_EXPIRED:
7045         case -NFS4ERR_BAD_STATEID:
7046                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7047                 if (new_lock_owner != 0 ||
7048                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7049                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7050                 break;
7051         case -NFS4ERR_STALE_STATEID:
7052                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7053                 nfs4_schedule_lease_recovery(server->nfs_client);
7054         }
7055 }
7056
7057 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7058 {
7059         struct nfs4_lockdata *data;
7060         struct rpc_task *task;
7061         struct rpc_message msg = {
7062                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7063                 .rpc_cred = state->owner->so_cred,
7064         };
7065         struct rpc_task_setup task_setup_data = {
7066                 .rpc_client = NFS_CLIENT(state->inode),
7067                 .rpc_message = &msg,
7068                 .callback_ops = &nfs4_lock_ops,
7069                 .workqueue = nfsiod_workqueue,
7070                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7071         };
7072         int ret;
7073
7074         dprintk("%s: begin!\n", __func__);
7075         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7076                         fl->fl_u.nfs4_fl.owner,
7077                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
7078         if (data == NULL)
7079                 return -ENOMEM;
7080         if (IS_SETLKW(cmd))
7081                 data->arg.block = 1;
7082         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7083                                 recovery_type > NFS_LOCK_NEW);
7084         msg.rpc_argp = &data->arg;
7085         msg.rpc_resp = &data->res;
7086         task_setup_data.callback_data = data;
7087         if (recovery_type > NFS_LOCK_NEW) {
7088                 if (recovery_type == NFS_LOCK_RECLAIM)
7089                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7090         } else
7091                 data->arg.new_lock = 1;
7092         task = rpc_run_task(&task_setup_data);
7093         if (IS_ERR(task))
7094                 return PTR_ERR(task);
7095         ret = rpc_wait_for_completion_task(task);
7096         if (ret == 0) {
7097                 ret = data->rpc_status;
7098                 if (ret)
7099                         nfs4_handle_setlk_error(data->server, data->lsp,
7100                                         data->arg.new_lock_owner, ret);
7101         } else
7102                 data->cancelled = true;
7103         rpc_put_task(task);
7104         dprintk("%s: done, ret = %d!\n", __func__, ret);
7105         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7106         return ret;
7107 }
7108
7109 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7110 {
7111         struct nfs_server *server = NFS_SERVER(state->inode);
7112         struct nfs4_exception exception = {
7113                 .inode = state->inode,
7114         };
7115         int err;
7116
7117         do {
7118                 /* Cache the lock if possible... */
7119                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7120                         return 0;
7121                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7122                 if (err != -NFS4ERR_DELAY)
7123                         break;
7124                 nfs4_handle_exception(server, err, &exception);
7125         } while (exception.retry);
7126         return err;
7127 }
7128
7129 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7130 {
7131         struct nfs_server *server = NFS_SERVER(state->inode);
7132         struct nfs4_exception exception = {
7133                 .inode = state->inode,
7134         };
7135         int err;
7136
7137         err = nfs4_set_lock_state(state, request);
7138         if (err != 0)
7139                 return err;
7140         if (!recover_lost_locks) {
7141                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7142                 return 0;
7143         }
7144         do {
7145                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7146                         return 0;
7147                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7148                 switch (err) {
7149                 default:
7150                         goto out;
7151                 case -NFS4ERR_GRACE:
7152                 case -NFS4ERR_DELAY:
7153                         nfs4_handle_exception(server, err, &exception);
7154                         err = 0;
7155                 }
7156         } while (exception.retry);
7157 out:
7158         return err;
7159 }
7160
7161 #if defined(CONFIG_NFS_V4_1)
7162 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7163 {
7164         struct nfs4_lock_state *lsp;
7165         int status;
7166
7167         status = nfs4_set_lock_state(state, request);
7168         if (status != 0)
7169                 return status;
7170         lsp = request->fl_u.nfs4_fl.owner;
7171         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7172             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7173                 return 0;
7174         return nfs4_lock_expired(state, request);
7175 }
7176 #endif
7177
7178 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7179 {
7180         struct nfs_inode *nfsi = NFS_I(state->inode);
7181         struct nfs4_state_owner *sp = state->owner;
7182         unsigned char fl_flags = request->fl_flags;
7183         int status;
7184
7185         request->fl_flags |= FL_ACCESS;
7186         status = locks_lock_inode_wait(state->inode, request);
7187         if (status < 0)
7188                 goto out;
7189         mutex_lock(&sp->so_delegreturn_mutex);
7190         down_read(&nfsi->rwsem);
7191         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7192                 /* Yes: cache locks! */
7193                 /* ...but avoid races with delegation recall... */
7194                 request->fl_flags = fl_flags & ~FL_SLEEP;
7195                 status = locks_lock_inode_wait(state->inode, request);
7196                 up_read(&nfsi->rwsem);
7197                 mutex_unlock(&sp->so_delegreturn_mutex);
7198                 goto out;
7199         }
7200         up_read(&nfsi->rwsem);
7201         mutex_unlock(&sp->so_delegreturn_mutex);
7202         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7203 out:
7204         request->fl_flags = fl_flags;
7205         return status;
7206 }
7207
7208 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7209 {
7210         struct nfs4_exception exception = {
7211                 .state = state,
7212                 .inode = state->inode,
7213                 .interruptible = true,
7214         };
7215         int err;
7216
7217         do {
7218                 err = _nfs4_proc_setlk(state, cmd, request);
7219                 if (err == -NFS4ERR_DENIED)
7220                         err = -EAGAIN;
7221                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7222                                 err, &exception);
7223         } while (exception.retry);
7224         return err;
7225 }
7226
7227 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7228 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7229
7230 static int
7231 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7232                         struct file_lock *request)
7233 {
7234         int             status = -ERESTARTSYS;
7235         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7236
7237         while(!signalled()) {
7238                 status = nfs4_proc_setlk(state, cmd, request);
7239                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7240                         break;
7241                 freezable_schedule_timeout_interruptible(timeout);
7242                 timeout *= 2;
7243                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7244                 status = -ERESTARTSYS;
7245         }
7246         return status;
7247 }
7248
7249 #ifdef CONFIG_NFS_V4_1
7250 struct nfs4_lock_waiter {
7251         struct task_struct      *task;
7252         struct inode            *inode;
7253         struct nfs_lowner       *owner;
7254 };
7255
7256 static int
7257 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7258 {
7259         int ret;
7260         struct nfs4_lock_waiter *waiter = wait->private;
7261
7262         /* NULL key means to wake up everyone */
7263         if (key) {
7264                 struct cb_notify_lock_args      *cbnl = key;
7265                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7266                                                 *wowner = waiter->owner;
7267
7268                 /* Only wake if the callback was for the same owner. */
7269                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7270                         return 0;
7271
7272                 /* Make sure it's for the right inode */
7273                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7274                         return 0;
7275         }
7276
7277         /* override "private" so we can use default_wake_function */
7278         wait->private = waiter->task;
7279         ret = woken_wake_function(wait, mode, flags, key);
7280         if (ret)
7281                 list_del_init(&wait->entry);
7282         wait->private = waiter;
7283         return ret;
7284 }
7285
7286 static int
7287 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7288 {
7289         int status = -ERESTARTSYS;
7290         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7291         struct nfs_server *server = NFS_SERVER(state->inode);
7292         struct nfs_client *clp = server->nfs_client;
7293         wait_queue_head_t *q = &clp->cl_lock_waitq;
7294         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7295                                     .id = lsp->ls_seqid.owner_id,
7296                                     .s_dev = server->s_dev };
7297         struct nfs4_lock_waiter waiter = { .task  = current,
7298                                            .inode = state->inode,
7299                                            .owner = &owner};
7300         wait_queue_entry_t wait;
7301
7302         /* Don't bother with waitqueue if we don't expect a callback */
7303         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7304                 return nfs4_retry_setlk_simple(state, cmd, request);
7305
7306         init_wait(&wait);
7307         wait.private = &waiter;
7308         wait.func = nfs4_wake_lock_waiter;
7309
7310         while(!signalled()) {
7311                 add_wait_queue(q, &wait);
7312                 status = nfs4_proc_setlk(state, cmd, request);
7313                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7314                         finish_wait(q, &wait);
7315                         break;
7316                 }
7317
7318                 status = -ERESTARTSYS;
7319                 freezer_do_not_count();
7320                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7321                 freezer_count();
7322                 finish_wait(q, &wait);
7323         }
7324
7325         return status;
7326 }
7327 #else /* !CONFIG_NFS_V4_1 */
7328 static inline int
7329 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7330 {
7331         return nfs4_retry_setlk_simple(state, cmd, request);
7332 }
7333 #endif
7334
7335 static int
7336 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7337 {
7338         struct nfs_open_context *ctx;
7339         struct nfs4_state *state;
7340         int status;
7341
7342         /* verify open state */
7343         ctx = nfs_file_open_context(filp);
7344         state = ctx->state;
7345
7346         if (IS_GETLK(cmd)) {
7347                 if (state != NULL)
7348                         return nfs4_proc_getlk(state, F_GETLK, request);
7349                 return 0;
7350         }
7351
7352         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7353                 return -EINVAL;
7354
7355         if (request->fl_type == F_UNLCK) {
7356                 if (state != NULL)
7357                         return nfs4_proc_unlck(state, cmd, request);
7358                 return 0;
7359         }
7360
7361         if (state == NULL)
7362                 return -ENOLCK;
7363
7364         if ((request->fl_flags & FL_POSIX) &&
7365             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7366                 return -ENOLCK;
7367
7368         /*
7369          * Don't rely on the VFS having checked the file open mode,
7370          * since it won't do this for flock() locks.
7371          */
7372         switch (request->fl_type) {
7373         case F_RDLCK:
7374                 if (!(filp->f_mode & FMODE_READ))
7375                         return -EBADF;
7376                 break;
7377         case F_WRLCK:
7378                 if (!(filp->f_mode & FMODE_WRITE))
7379                         return -EBADF;
7380         }
7381
7382         status = nfs4_set_lock_state(state, request);
7383         if (status != 0)
7384                 return status;
7385
7386         return nfs4_retry_setlk(state, cmd, request);
7387 }
7388
7389 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7390 {
7391         struct nfs_server *server = NFS_SERVER(state->inode);
7392         int err;
7393
7394         err = nfs4_set_lock_state(state, fl);
7395         if (err != 0)
7396                 return err;
7397         do {
7398                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7399                 if (err != -NFS4ERR_DELAY)
7400                         break;
7401                 ssleep(1);
7402         } while (err == -NFS4ERR_DELAY);
7403         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7404 }
7405
7406 struct nfs_release_lockowner_data {
7407         struct nfs4_lock_state *lsp;
7408         struct nfs_server *server;
7409         struct nfs_release_lockowner_args args;
7410         struct nfs_release_lockowner_res res;
7411         unsigned long timestamp;
7412 };
7413
7414 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7415 {
7416         struct nfs_release_lockowner_data *data = calldata;
7417         struct nfs_server *server = data->server;
7418         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7419                            &data->res.seq_res, task);
7420         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7421         data->timestamp = jiffies;
7422 }
7423
7424 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7425 {
7426         struct nfs_release_lockowner_data *data = calldata;
7427         struct nfs_server *server = data->server;
7428
7429         nfs40_sequence_done(task, &data->res.seq_res);
7430
7431         switch (task->tk_status) {
7432         case 0:
7433                 renew_lease(server, data->timestamp);
7434                 break;
7435         case -NFS4ERR_STALE_CLIENTID:
7436         case -NFS4ERR_EXPIRED:
7437                 nfs4_schedule_lease_recovery(server->nfs_client);
7438                 break;
7439         case -NFS4ERR_LEASE_MOVED:
7440         case -NFS4ERR_DELAY:
7441                 if (nfs4_async_handle_error(task, server,
7442                                             NULL, NULL) == -EAGAIN)
7443                         rpc_restart_call_prepare(task);
7444         }
7445 }
7446
7447 static void nfs4_release_lockowner_release(void *calldata)
7448 {
7449         struct nfs_release_lockowner_data *data = calldata;
7450         nfs4_free_lock_state(data->server, data->lsp);
7451         kfree(calldata);
7452 }
7453
7454 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7455         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7456         .rpc_call_done = nfs4_release_lockowner_done,
7457         .rpc_release = nfs4_release_lockowner_release,
7458 };
7459
7460 static void
7461 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7462 {
7463         struct nfs_release_lockowner_data *data;
7464         struct rpc_message msg = {
7465                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7466         };
7467
7468         if (server->nfs_client->cl_mvops->minor_version != 0)
7469                 return;
7470
7471         data = kmalloc(sizeof(*data), GFP_NOFS);
7472         if (!data)
7473                 return;
7474         data->lsp = lsp;
7475         data->server = server;
7476         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7477         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7478         data->args.lock_owner.s_dev = server->s_dev;
7479
7480         msg.rpc_argp = &data->args;
7481         msg.rpc_resp = &data->res;
7482         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7483         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7484 }
7485
7486 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7487
7488 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7489                                    struct dentry *unused, struct inode *inode,
7490                                    const char *key, const void *buf,
7491                                    size_t buflen, int flags)
7492 {
7493         return nfs4_proc_set_acl(inode, buf, buflen);
7494 }
7495
7496 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7497                                    struct dentry *unused, struct inode *inode,
7498                                    const char *key, void *buf, size_t buflen)
7499 {
7500         return nfs4_proc_get_acl(inode, buf, buflen);
7501 }
7502
7503 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7504 {
7505         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7506 }
7507
7508 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7509
7510 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7511                                      struct dentry *unused, struct inode *inode,
7512                                      const char *key, const void *buf,
7513                                      size_t buflen, int flags)
7514 {
7515         if (security_ismaclabel(key))
7516                 return nfs4_set_security_label(inode, buf, buflen);
7517
7518         return -EOPNOTSUPP;
7519 }
7520
7521 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7522                                      struct dentry *unused, struct inode *inode,
7523                                      const char *key, void *buf, size_t buflen)
7524 {
7525         if (security_ismaclabel(key))
7526                 return nfs4_get_security_label(inode, buf, buflen);
7527         return -EOPNOTSUPP;
7528 }
7529
7530 static ssize_t
7531 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7532 {
7533         int len = 0;
7534
7535         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7536                 len = security_inode_listsecurity(inode, list, list_len);
7537                 if (len >= 0 && list_len && len > list_len)
7538                         return -ERANGE;
7539         }
7540         return len;
7541 }
7542
7543 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7544         .prefix = XATTR_SECURITY_PREFIX,
7545         .get    = nfs4_xattr_get_nfs4_label,
7546         .set    = nfs4_xattr_set_nfs4_label,
7547 };
7548
7549 #else
7550
7551 static ssize_t
7552 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7553 {
7554         return 0;
7555 }
7556
7557 #endif
7558
7559 #ifdef CONFIG_NFS_V4_2
7560 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7561                                     struct dentry *unused, struct inode *inode,
7562                                     const char *key, const void *buf,
7563                                     size_t buflen, int flags)
7564 {
7565         struct nfs_access_entry cache;
7566         int ret;
7567
7568         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7569                 return -EOPNOTSUPP;
7570
7571         /*
7572          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7573          * flags right now. Handling of xattr operations use the normal
7574          * file read/write permissions.
7575          *
7576          * Just in case the server has other ideas (which RFC 8276 allows),
7577          * do a cached access check for the XA* flags to possibly avoid
7578          * doing an RPC and getting EACCES back.
7579          */
7580         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7581                 if (!(cache.mask & NFS_ACCESS_XAWRITE))
7582                         return -EACCES;
7583         }
7584
7585         if (buf == NULL) {
7586                 ret = nfs42_proc_removexattr(inode, key);
7587                 if (!ret)
7588                         nfs4_xattr_cache_remove(inode, key);
7589         } else {
7590                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7591                 if (!ret)
7592                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7593         }
7594
7595         return ret;
7596 }
7597
7598 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7599                                     struct dentry *unused, struct inode *inode,
7600                                     const char *key, void *buf, size_t buflen)
7601 {
7602         struct nfs_access_entry cache;
7603         ssize_t ret;
7604
7605         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7606                 return -EOPNOTSUPP;
7607
7608         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7609                 if (!(cache.mask & NFS_ACCESS_XAREAD))
7610                         return -EACCES;
7611         }
7612
7613         ret = nfs_revalidate_inode(NFS_SERVER(inode), inode);
7614         if (ret)
7615                 return ret;
7616
7617         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7618         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7619                 return ret;
7620
7621         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7622
7623         return ret;
7624 }
7625
7626 static ssize_t
7627 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7628 {
7629         u64 cookie;
7630         bool eof;
7631         ssize_t ret, size;
7632         char *buf;
7633         size_t buflen;
7634         struct nfs_access_entry cache;
7635
7636         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7637                 return 0;
7638
7639         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7640                 if (!(cache.mask & NFS_ACCESS_XALIST))
7641                         return 0;
7642         }
7643
7644         ret = nfs_revalidate_inode(NFS_SERVER(inode), inode);
7645         if (ret)
7646                 return ret;
7647
7648         ret = nfs4_xattr_cache_list(inode, list, list_len);
7649         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7650                 return ret;
7651
7652         cookie = 0;
7653         eof = false;
7654         buflen = list_len ? list_len : XATTR_LIST_MAX;
7655         buf = list_len ? list : NULL;
7656         size = 0;
7657
7658         while (!eof) {
7659                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7660                     &cookie, &eof);
7661                 if (ret < 0)
7662                         return ret;
7663
7664                 if (list_len) {
7665                         buf += ret;
7666                         buflen -= ret;
7667                 }
7668                 size += ret;
7669         }
7670
7671         if (list_len)
7672                 nfs4_xattr_cache_set_list(inode, list, size);
7673
7674         return size;
7675 }
7676
7677 #else
7678
7679 static ssize_t
7680 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7681 {
7682         return 0;
7683 }
7684 #endif /* CONFIG_NFS_V4_2 */
7685
7686 /*
7687  * nfs_fhget will use either the mounted_on_fileid or the fileid
7688  */
7689 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7690 {
7691         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7692                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7693               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7694               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7695                 return;
7696
7697         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7698                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7699         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7700         fattr->nlink = 2;
7701 }
7702
7703 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7704                                    const struct qstr *name,
7705                                    struct nfs4_fs_locations *fs_locations,
7706                                    struct page *page)
7707 {
7708         struct nfs_server *server = NFS_SERVER(dir);
7709         u32 bitmask[3];
7710         struct nfs4_fs_locations_arg args = {
7711                 .dir_fh = NFS_FH(dir),
7712                 .name = name,
7713                 .page = page,
7714                 .bitmask = bitmask,
7715         };
7716         struct nfs4_fs_locations_res res = {
7717                 .fs_locations = fs_locations,
7718         };
7719         struct rpc_message msg = {
7720                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7721                 .rpc_argp = &args,
7722                 .rpc_resp = &res,
7723         };
7724         int status;
7725
7726         dprintk("%s: start\n", __func__);
7727
7728         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7729         bitmask[1] = nfs4_fattr_bitmap[1];
7730
7731         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7732          * is not supported */
7733         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7734                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7735         else
7736                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7737
7738         nfs_fattr_init(&fs_locations->fattr);
7739         fs_locations->server = server;
7740         fs_locations->nlocations = 0;
7741         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7742         dprintk("%s: returned status = %d\n", __func__, status);
7743         return status;
7744 }
7745
7746 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7747                            const struct qstr *name,
7748                            struct nfs4_fs_locations *fs_locations,
7749                            struct page *page)
7750 {
7751         struct nfs4_exception exception = {
7752                 .interruptible = true,
7753         };
7754         int err;
7755         do {
7756                 err = _nfs4_proc_fs_locations(client, dir, name,
7757                                 fs_locations, page);
7758                 trace_nfs4_get_fs_locations(dir, name, err);
7759                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7760                                 &exception);
7761         } while (exception.retry);
7762         return err;
7763 }
7764
7765 /*
7766  * This operation also signals the server that this client is
7767  * performing migration recovery.  The server can stop returning
7768  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7769  * appended to this compound to identify the client ID which is
7770  * performing recovery.
7771  */
7772 static int _nfs40_proc_get_locations(struct inode *inode,
7773                                      struct nfs4_fs_locations *locations,
7774                                      struct page *page, const struct cred *cred)
7775 {
7776         struct nfs_server *server = NFS_SERVER(inode);
7777         struct rpc_clnt *clnt = server->client;
7778         u32 bitmask[2] = {
7779                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7780         };
7781         struct nfs4_fs_locations_arg args = {
7782                 .clientid       = server->nfs_client->cl_clientid,
7783                 .fh             = NFS_FH(inode),
7784                 .page           = page,
7785                 .bitmask        = bitmask,
7786                 .migration      = 1,            /* skip LOOKUP */
7787                 .renew          = 1,            /* append RENEW */
7788         };
7789         struct nfs4_fs_locations_res res = {
7790                 .fs_locations   = locations,
7791                 .migration      = 1,
7792                 .renew          = 1,
7793         };
7794         struct rpc_message msg = {
7795                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7796                 .rpc_argp       = &args,
7797                 .rpc_resp       = &res,
7798                 .rpc_cred       = cred,
7799         };
7800         unsigned long now = jiffies;
7801         int status;
7802
7803         nfs_fattr_init(&locations->fattr);
7804         locations->server = server;
7805         locations->nlocations = 0;
7806
7807         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7808         status = nfs4_call_sync_sequence(clnt, server, &msg,
7809                                         &args.seq_args, &res.seq_res);
7810         if (status)
7811                 return status;
7812
7813         renew_lease(server, now);
7814         return 0;
7815 }
7816
7817 #ifdef CONFIG_NFS_V4_1
7818
7819 /*
7820  * This operation also signals the server that this client is
7821  * performing migration recovery.  The server can stop asserting
7822  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7823  * performing this operation is identified in the SEQUENCE
7824  * operation in this compound.
7825  *
7826  * When the client supports GETATTR(fs_locations_info), it can
7827  * be plumbed in here.
7828  */
7829 static int _nfs41_proc_get_locations(struct inode *inode,
7830                                      struct nfs4_fs_locations *locations,
7831                                      struct page *page, const struct cred *cred)
7832 {
7833         struct nfs_server *server = NFS_SERVER(inode);
7834         struct rpc_clnt *clnt = server->client;
7835         u32 bitmask[2] = {
7836                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7837         };
7838         struct nfs4_fs_locations_arg args = {
7839                 .fh             = NFS_FH(inode),
7840                 .page           = page,
7841                 .bitmask        = bitmask,
7842                 .migration      = 1,            /* skip LOOKUP */
7843         };
7844         struct nfs4_fs_locations_res res = {
7845                 .fs_locations   = locations,
7846                 .migration      = 1,
7847         };
7848         struct rpc_message msg = {
7849                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7850                 .rpc_argp       = &args,
7851                 .rpc_resp       = &res,
7852                 .rpc_cred       = cred,
7853         };
7854         int status;
7855
7856         nfs_fattr_init(&locations->fattr);
7857         locations->server = server;
7858         locations->nlocations = 0;
7859
7860         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7861         status = nfs4_call_sync_sequence(clnt, server, &msg,
7862                                         &args.seq_args, &res.seq_res);
7863         if (status == NFS4_OK &&
7864             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7865                 status = -NFS4ERR_LEASE_MOVED;
7866         return status;
7867 }
7868
7869 #endif  /* CONFIG_NFS_V4_1 */
7870
7871 /**
7872  * nfs4_proc_get_locations - discover locations for a migrated FSID
7873  * @inode: inode on FSID that is migrating
7874  * @locations: result of query
7875  * @page: buffer
7876  * @cred: credential to use for this operation
7877  *
7878  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7879  * operation failed, or a negative errno if a local error occurred.
7880  *
7881  * On success, "locations" is filled in, but if the server has
7882  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7883  * asserted.
7884  *
7885  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7886  * from this client that require migration recovery.
7887  */
7888 int nfs4_proc_get_locations(struct inode *inode,
7889                             struct nfs4_fs_locations *locations,
7890                             struct page *page, const struct cred *cred)
7891 {
7892         struct nfs_server *server = NFS_SERVER(inode);
7893         struct nfs_client *clp = server->nfs_client;
7894         const struct nfs4_mig_recovery_ops *ops =
7895                                         clp->cl_mvops->mig_recovery_ops;
7896         struct nfs4_exception exception = {
7897                 .interruptible = true,
7898         };
7899         int status;
7900
7901         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7902                 (unsigned long long)server->fsid.major,
7903                 (unsigned long long)server->fsid.minor,
7904                 clp->cl_hostname);
7905         nfs_display_fhandle(NFS_FH(inode), __func__);
7906
7907         do {
7908                 status = ops->get_locations(inode, locations, page, cred);
7909                 if (status != -NFS4ERR_DELAY)
7910                         break;
7911                 nfs4_handle_exception(server, status, &exception);
7912         } while (exception.retry);
7913         return status;
7914 }
7915
7916 /*
7917  * This operation also signals the server that this client is
7918  * performing "lease moved" recovery.  The server can stop
7919  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7920  * is appended to this compound to identify the client ID which is
7921  * performing recovery.
7922  */
7923 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7924 {
7925         struct nfs_server *server = NFS_SERVER(inode);
7926         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7927         struct rpc_clnt *clnt = server->client;
7928         struct nfs4_fsid_present_arg args = {
7929                 .fh             = NFS_FH(inode),
7930                 .clientid       = clp->cl_clientid,
7931                 .renew          = 1,            /* append RENEW */
7932         };
7933         struct nfs4_fsid_present_res res = {
7934                 .renew          = 1,
7935         };
7936         struct rpc_message msg = {
7937                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7938                 .rpc_argp       = &args,
7939                 .rpc_resp       = &res,
7940                 .rpc_cred       = cred,
7941         };
7942         unsigned long now = jiffies;
7943         int status;
7944
7945         res.fh = nfs_alloc_fhandle();
7946         if (res.fh == NULL)
7947                 return -ENOMEM;
7948
7949         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7950         status = nfs4_call_sync_sequence(clnt, server, &msg,
7951                                                 &args.seq_args, &res.seq_res);
7952         nfs_free_fhandle(res.fh);
7953         if (status)
7954                 return status;
7955
7956         do_renew_lease(clp, now);
7957         return 0;
7958 }
7959
7960 #ifdef CONFIG_NFS_V4_1
7961
7962 /*
7963  * This operation also signals the server that this client is
7964  * performing "lease moved" recovery.  The server can stop asserting
7965  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7966  * this operation is identified in the SEQUENCE operation in this
7967  * compound.
7968  */
7969 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7970 {
7971         struct nfs_server *server = NFS_SERVER(inode);
7972         struct rpc_clnt *clnt = server->client;
7973         struct nfs4_fsid_present_arg args = {
7974                 .fh             = NFS_FH(inode),
7975         };
7976         struct nfs4_fsid_present_res res = {
7977         };
7978         struct rpc_message msg = {
7979                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7980                 .rpc_argp       = &args,
7981                 .rpc_resp       = &res,
7982                 .rpc_cred       = cred,
7983         };
7984         int status;
7985
7986         res.fh = nfs_alloc_fhandle();
7987         if (res.fh == NULL)
7988                 return -ENOMEM;
7989
7990         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7991         status = nfs4_call_sync_sequence(clnt, server, &msg,
7992                                                 &args.seq_args, &res.seq_res);
7993         nfs_free_fhandle(res.fh);
7994         if (status == NFS4_OK &&
7995             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7996                 status = -NFS4ERR_LEASE_MOVED;
7997         return status;
7998 }
7999
8000 #endif  /* CONFIG_NFS_V4_1 */
8001
8002 /**
8003  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8004  * @inode: inode on FSID to check
8005  * @cred: credential to use for this operation
8006  *
8007  * Server indicates whether the FSID is present, moved, or not
8008  * recognized.  This operation is necessary to clear a LEASE_MOVED
8009  * condition for this client ID.
8010  *
8011  * Returns NFS4_OK if the FSID is present on this server,
8012  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8013  *  NFS4ERR code if some error occurred on the server, or a
8014  *  negative errno if a local failure occurred.
8015  */
8016 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8017 {
8018         struct nfs_server *server = NFS_SERVER(inode);
8019         struct nfs_client *clp = server->nfs_client;
8020         const struct nfs4_mig_recovery_ops *ops =
8021                                         clp->cl_mvops->mig_recovery_ops;
8022         struct nfs4_exception exception = {
8023                 .interruptible = true,
8024         };
8025         int status;
8026
8027         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8028                 (unsigned long long)server->fsid.major,
8029                 (unsigned long long)server->fsid.minor,
8030                 clp->cl_hostname);
8031         nfs_display_fhandle(NFS_FH(inode), __func__);
8032
8033         do {
8034                 status = ops->fsid_present(inode, cred);
8035                 if (status != -NFS4ERR_DELAY)
8036                         break;
8037                 nfs4_handle_exception(server, status, &exception);
8038         } while (exception.retry);
8039         return status;
8040 }
8041
8042 /*
8043  * If 'use_integrity' is true and the state managment nfs_client
8044  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8045  * and the machine credential as per RFC3530bis and RFC5661 Security
8046  * Considerations sections. Otherwise, just use the user cred with the
8047  * filesystem's rpc_client.
8048  */
8049 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8050 {
8051         int status;
8052         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8053         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8054         struct nfs4_secinfo_arg args = {
8055                 .dir_fh = NFS_FH(dir),
8056                 .name   = name,
8057         };
8058         struct nfs4_secinfo_res res = {
8059                 .flavors     = flavors,
8060         };
8061         struct rpc_message msg = {
8062                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8063                 .rpc_argp = &args,
8064                 .rpc_resp = &res,
8065         };
8066         struct nfs4_call_sync_data data = {
8067                 .seq_server = NFS_SERVER(dir),
8068                 .seq_args = &args.seq_args,
8069                 .seq_res = &res.seq_res,
8070         };
8071         struct rpc_task_setup task_setup = {
8072                 .rpc_client = clnt,
8073                 .rpc_message = &msg,
8074                 .callback_ops = clp->cl_mvops->call_sync_ops,
8075                 .callback_data = &data,
8076                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8077         };
8078         const struct cred *cred = NULL;
8079
8080         if (use_integrity) {
8081                 clnt = clp->cl_rpcclient;
8082                 task_setup.rpc_client = clnt;
8083
8084                 cred = nfs4_get_clid_cred(clp);
8085                 msg.rpc_cred = cred;
8086         }
8087
8088         dprintk("NFS call  secinfo %s\n", name->name);
8089
8090         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8091         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8092         status = nfs4_call_sync_custom(&task_setup);
8093
8094         dprintk("NFS reply  secinfo: %d\n", status);
8095
8096         put_cred(cred);
8097         return status;
8098 }
8099
8100 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8101                       struct nfs4_secinfo_flavors *flavors)
8102 {
8103         struct nfs4_exception exception = {
8104                 .interruptible = true,
8105         };
8106         int err;
8107         do {
8108                 err = -NFS4ERR_WRONGSEC;
8109
8110                 /* try to use integrity protection with machine cred */
8111                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8112                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8113
8114                 /*
8115                  * if unable to use integrity protection, or SECINFO with
8116                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8117                  * disallowed by spec, but exists in deployed servers) use
8118                  * the current filesystem's rpc_client and the user cred.
8119                  */
8120                 if (err == -NFS4ERR_WRONGSEC)
8121                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8122
8123                 trace_nfs4_secinfo(dir, name, err);
8124                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8125                                 &exception);
8126         } while (exception.retry);
8127         return err;
8128 }
8129
8130 #ifdef CONFIG_NFS_V4_1
8131 /*
8132  * Check the exchange flags returned by the server for invalid flags, having
8133  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8134  * DS flags set.
8135  */
8136 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8137 {
8138         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8139                 goto out_inval;
8140         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8141                 goto out_inval;
8142         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8143             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8144                 goto out_inval;
8145         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8146                 goto out_inval;
8147         return NFS_OK;
8148 out_inval:
8149         return -NFS4ERR_INVAL;
8150 }
8151
8152 static bool
8153 nfs41_same_server_scope(struct nfs41_server_scope *a,
8154                         struct nfs41_server_scope *b)
8155 {
8156         if (a->server_scope_sz != b->server_scope_sz)
8157                 return false;
8158         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8159 }
8160
8161 static void
8162 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8163 {
8164         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8165         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8166         struct nfs_client *clp = args->client;
8167
8168         switch (task->tk_status) {
8169         case -NFS4ERR_BADSESSION:
8170         case -NFS4ERR_DEADSESSION:
8171                 nfs4_schedule_session_recovery(clp->cl_session,
8172                                 task->tk_status);
8173         }
8174         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8175                         res->dir != NFS4_CDFS4_BOTH) {
8176                 rpc_task_close_connection(task);
8177                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8178                         rpc_restart_call(task);
8179         }
8180 }
8181
8182 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8183         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8184 };
8185
8186 /*
8187  * nfs4_proc_bind_one_conn_to_session()
8188  *
8189  * The 4.1 client currently uses the same TCP connection for the
8190  * fore and backchannel.
8191  */
8192 static
8193 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8194                 struct rpc_xprt *xprt,
8195                 struct nfs_client *clp,
8196                 const struct cred *cred)
8197 {
8198         int status;
8199         struct nfs41_bind_conn_to_session_args args = {
8200                 .client = clp,
8201                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8202                 .retries = 0,
8203         };
8204         struct nfs41_bind_conn_to_session_res res;
8205         struct rpc_message msg = {
8206                 .rpc_proc =
8207                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8208                 .rpc_argp = &args,
8209                 .rpc_resp = &res,
8210                 .rpc_cred = cred,
8211         };
8212         struct rpc_task_setup task_setup_data = {
8213                 .rpc_client = clnt,
8214                 .rpc_xprt = xprt,
8215                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8216                 .rpc_message = &msg,
8217                 .flags = RPC_TASK_TIMEOUT,
8218         };
8219         struct rpc_task *task;
8220
8221         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8222         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8223                 args.dir = NFS4_CDFC4_FORE;
8224
8225         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8226         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8227                 args.dir = NFS4_CDFC4_FORE;
8228
8229         task = rpc_run_task(&task_setup_data);
8230         if (!IS_ERR(task)) {
8231                 status = task->tk_status;
8232                 rpc_put_task(task);
8233         } else
8234                 status = PTR_ERR(task);
8235         trace_nfs4_bind_conn_to_session(clp, status);
8236         if (status == 0) {
8237                 if (memcmp(res.sessionid.data,
8238                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8239                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8240                         return -EIO;
8241                 }
8242                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8243                         dprintk("NFS: %s: Unexpected direction from server\n",
8244                                 __func__);
8245                         return -EIO;
8246                 }
8247                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8248                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8249                                 __func__);
8250                         return -EIO;
8251                 }
8252         }
8253
8254         return status;
8255 }
8256
8257 struct rpc_bind_conn_calldata {
8258         struct nfs_client *clp;
8259         const struct cred *cred;
8260 };
8261
8262 static int
8263 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8264                 struct rpc_xprt *xprt,
8265                 void *calldata)
8266 {
8267         struct rpc_bind_conn_calldata *p = calldata;
8268
8269         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8270 }
8271
8272 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8273 {
8274         struct rpc_bind_conn_calldata data = {
8275                 .clp = clp,
8276                 .cred = cred,
8277         };
8278         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8279                         nfs4_proc_bind_conn_to_session_callback, &data);
8280 }
8281
8282 /*
8283  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8284  * and operations we'd like to see to enable certain features in the allow map
8285  */
8286 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8287         .how = SP4_MACH_CRED,
8288         .enforce.u.words = {
8289                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8290                       1 << (OP_EXCHANGE_ID - 32) |
8291                       1 << (OP_CREATE_SESSION - 32) |
8292                       1 << (OP_DESTROY_SESSION - 32) |
8293                       1 << (OP_DESTROY_CLIENTID - 32)
8294         },
8295         .allow.u.words = {
8296                 [0] = 1 << (OP_CLOSE) |
8297                       1 << (OP_OPEN_DOWNGRADE) |
8298                       1 << (OP_LOCKU) |
8299                       1 << (OP_DELEGRETURN) |
8300                       1 << (OP_COMMIT),
8301                 [1] = 1 << (OP_SECINFO - 32) |
8302                       1 << (OP_SECINFO_NO_NAME - 32) |
8303                       1 << (OP_LAYOUTRETURN - 32) |
8304                       1 << (OP_TEST_STATEID - 32) |
8305                       1 << (OP_FREE_STATEID - 32) |
8306                       1 << (OP_WRITE - 32)
8307         }
8308 };
8309
8310 /*
8311  * Select the state protection mode for client `clp' given the server results
8312  * from exchange_id in `sp'.
8313  *
8314  * Returns 0 on success, negative errno otherwise.
8315  */
8316 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8317                                  struct nfs41_state_protection *sp)
8318 {
8319         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8320                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8321                       1 << (OP_EXCHANGE_ID - 32) |
8322                       1 << (OP_CREATE_SESSION - 32) |
8323                       1 << (OP_DESTROY_SESSION - 32) |
8324                       1 << (OP_DESTROY_CLIENTID - 32)
8325         };
8326         unsigned long flags = 0;
8327         unsigned int i;
8328         int ret = 0;
8329
8330         if (sp->how == SP4_MACH_CRED) {
8331                 /* Print state protect result */
8332                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8333                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8334                         if (test_bit(i, sp->enforce.u.longs))
8335                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8336                         if (test_bit(i, sp->allow.u.longs))
8337                                 dfprintk(MOUNT, "  allow op %d\n", i);
8338                 }
8339
8340                 /* make sure nothing is on enforce list that isn't supported */
8341                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8342                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8343                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8344                                 ret = -EINVAL;
8345                                 goto out;
8346                         }
8347                 }
8348
8349                 /*
8350                  * Minimal mode - state operations are allowed to use machine
8351                  * credential.  Note this already happens by default, so the
8352                  * client doesn't have to do anything more than the negotiation.
8353                  *
8354                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8355                  *       we're already using the machine cred for exchange_id
8356                  *       and will never use a different cred.
8357                  */
8358                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8359                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8360                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8361                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8362                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8363                         dfprintk(MOUNT, "  minimal mode enabled\n");
8364                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8365                 } else {
8366                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8367                         ret = -EINVAL;
8368                         goto out;
8369                 }
8370
8371                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8372                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8373                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8374                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8375                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8376                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8377                 }
8378
8379                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8380                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8381                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8382                 }
8383
8384                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8385                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8386                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8387                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8388                 }
8389
8390                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8391                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8392                         dfprintk(MOUNT, "  stateid mode enabled\n");
8393                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8394                 }
8395
8396                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8397                         dfprintk(MOUNT, "  write mode enabled\n");
8398                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8399                 }
8400
8401                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8402                         dfprintk(MOUNT, "  commit mode enabled\n");
8403                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8404                 }
8405         }
8406 out:
8407         clp->cl_sp4_flags = flags;
8408         return ret;
8409 }
8410
8411 struct nfs41_exchange_id_data {
8412         struct nfs41_exchange_id_res res;
8413         struct nfs41_exchange_id_args args;
8414 };
8415
8416 static void nfs4_exchange_id_release(void *data)
8417 {
8418         struct nfs41_exchange_id_data *cdata =
8419                                         (struct nfs41_exchange_id_data *)data;
8420
8421         nfs_put_client(cdata->args.client);
8422         kfree(cdata->res.impl_id);
8423         kfree(cdata->res.server_scope);
8424         kfree(cdata->res.server_owner);
8425         kfree(cdata);
8426 }
8427
8428 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8429         .rpc_release = nfs4_exchange_id_release,
8430 };
8431
8432 /*
8433  * _nfs4_proc_exchange_id()
8434  *
8435  * Wrapper for EXCHANGE_ID operation.
8436  */
8437 static struct rpc_task *
8438 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8439                         u32 sp4_how, struct rpc_xprt *xprt)
8440 {
8441         struct rpc_message msg = {
8442                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8443                 .rpc_cred = cred,
8444         };
8445         struct rpc_task_setup task_setup_data = {
8446                 .rpc_client = clp->cl_rpcclient,
8447                 .callback_ops = &nfs4_exchange_id_call_ops,
8448                 .rpc_message = &msg,
8449                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8450         };
8451         struct nfs41_exchange_id_data *calldata;
8452         int status;
8453
8454         if (!refcount_inc_not_zero(&clp->cl_count))
8455                 return ERR_PTR(-EIO);
8456
8457         status = -ENOMEM;
8458         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8459         if (!calldata)
8460                 goto out;
8461
8462         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8463
8464         status = nfs4_init_uniform_client_string(clp);
8465         if (status)
8466                 goto out_calldata;
8467
8468         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8469                                                 GFP_NOFS);
8470         status = -ENOMEM;
8471         if (unlikely(calldata->res.server_owner == NULL))
8472                 goto out_calldata;
8473
8474         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8475                                         GFP_NOFS);
8476         if (unlikely(calldata->res.server_scope == NULL))
8477                 goto out_server_owner;
8478
8479         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8480         if (unlikely(calldata->res.impl_id == NULL))
8481                 goto out_server_scope;
8482
8483         switch (sp4_how) {
8484         case SP4_NONE:
8485                 calldata->args.state_protect.how = SP4_NONE;
8486                 break;
8487
8488         case SP4_MACH_CRED:
8489                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8490                 break;
8491
8492         default:
8493                 /* unsupported! */
8494                 WARN_ON_ONCE(1);
8495                 status = -EINVAL;
8496                 goto out_impl_id;
8497         }
8498         if (xprt) {
8499                 task_setup_data.rpc_xprt = xprt;
8500                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8501                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8502                                 sizeof(calldata->args.verifier.data));
8503         }
8504         calldata->args.client = clp;
8505         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8506         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8507 #ifdef CONFIG_NFS_V4_1_MIGRATION
8508         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8509 #endif
8510         msg.rpc_argp = &calldata->args;
8511         msg.rpc_resp = &calldata->res;
8512         task_setup_data.callback_data = calldata;
8513
8514         return rpc_run_task(&task_setup_data);
8515
8516 out_impl_id:
8517         kfree(calldata->res.impl_id);
8518 out_server_scope:
8519         kfree(calldata->res.server_scope);
8520 out_server_owner:
8521         kfree(calldata->res.server_owner);
8522 out_calldata:
8523         kfree(calldata);
8524 out:
8525         nfs_put_client(clp);
8526         return ERR_PTR(status);
8527 }
8528
8529 /*
8530  * _nfs4_proc_exchange_id()
8531  *
8532  * Wrapper for EXCHANGE_ID operation.
8533  */
8534 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8535                         u32 sp4_how)
8536 {
8537         struct rpc_task *task;
8538         struct nfs41_exchange_id_args *argp;
8539         struct nfs41_exchange_id_res *resp;
8540         unsigned long now = jiffies;
8541         int status;
8542
8543         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8544         if (IS_ERR(task))
8545                 return PTR_ERR(task);
8546
8547         argp = task->tk_msg.rpc_argp;
8548         resp = task->tk_msg.rpc_resp;
8549         status = task->tk_status;
8550         if (status  != 0)
8551                 goto out;
8552
8553         status = nfs4_check_cl_exchange_flags(resp->flags,
8554                         clp->cl_mvops->minor_version);
8555         if (status  != 0)
8556                 goto out;
8557
8558         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8559         if (status != 0)
8560                 goto out;
8561
8562         do_renew_lease(clp, now);
8563
8564         clp->cl_clientid = resp->clientid;
8565         clp->cl_exchange_flags = resp->flags;
8566         clp->cl_seqid = resp->seqid;
8567         /* Client ID is not confirmed */
8568         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8569                 clear_bit(NFS4_SESSION_ESTABLISHED,
8570                           &clp->cl_session->session_state);
8571
8572         if (clp->cl_serverscope != NULL &&
8573             !nfs41_same_server_scope(clp->cl_serverscope,
8574                                 resp->server_scope)) {
8575                 dprintk("%s: server_scope mismatch detected\n",
8576                         __func__);
8577                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8578         }
8579
8580         swap(clp->cl_serverowner, resp->server_owner);
8581         swap(clp->cl_serverscope, resp->server_scope);
8582         swap(clp->cl_implid, resp->impl_id);
8583
8584         /* Save the EXCHANGE_ID verifier session trunk tests */
8585         memcpy(clp->cl_confirm.data, argp->verifier.data,
8586                sizeof(clp->cl_confirm.data));
8587 out:
8588         trace_nfs4_exchange_id(clp, status);
8589         rpc_put_task(task);
8590         return status;
8591 }
8592
8593 /*
8594  * nfs4_proc_exchange_id()
8595  *
8596  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8597  *
8598  * Since the clientid has expired, all compounds using sessions
8599  * associated with the stale clientid will be returning
8600  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8601  * be in some phase of session reset.
8602  *
8603  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8604  */
8605 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8606 {
8607         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8608         int status;
8609
8610         /* try SP4_MACH_CRED if krb5i/p */
8611         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8612             authflavor == RPC_AUTH_GSS_KRB5P) {
8613                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8614                 if (!status)
8615                         return 0;
8616         }
8617
8618         /* try SP4_NONE */
8619         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8620 }
8621
8622 /**
8623  * nfs4_test_session_trunk
8624  *
8625  * This is an add_xprt_test() test function called from
8626  * rpc_clnt_setup_test_and_add_xprt.
8627  *
8628  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8629  * and is dereferrenced in nfs4_exchange_id_release
8630  *
8631  * Upon success, add the new transport to the rpc_clnt
8632  *
8633  * @clnt: struct rpc_clnt to get new transport
8634  * @xprt: the rpc_xprt to test
8635  * @data: call data for _nfs4_proc_exchange_id.
8636  */
8637 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8638                             void *data)
8639 {
8640         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8641         struct rpc_task *task;
8642         int status;
8643
8644         u32 sp4_how;
8645
8646         dprintk("--> %s try %s\n", __func__,
8647                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8648
8649         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8650
8651         /* Test connection for session trunking. Async exchange_id call */
8652         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8653         if (IS_ERR(task))
8654                 return;
8655
8656         status = task->tk_status;
8657         if (status == 0)
8658                 status = nfs4_detect_session_trunking(adata->clp,
8659                                 task->tk_msg.rpc_resp, xprt);
8660
8661         if (status == 0)
8662                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8663
8664         rpc_put_task(task);
8665 }
8666 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8667
8668 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8669                 const struct cred *cred)
8670 {
8671         struct rpc_message msg = {
8672                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8673                 .rpc_argp = clp,
8674                 .rpc_cred = cred,
8675         };
8676         int status;
8677
8678         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8679                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8680         trace_nfs4_destroy_clientid(clp, status);
8681         if (status)
8682                 dprintk("NFS: Got error %d from the server %s on "
8683                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8684         return status;
8685 }
8686
8687 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8688                 const struct cred *cred)
8689 {
8690         unsigned int loop;
8691         int ret;
8692
8693         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8694                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8695                 switch (ret) {
8696                 case -NFS4ERR_DELAY:
8697                 case -NFS4ERR_CLIENTID_BUSY:
8698                         ssleep(1);
8699                         break;
8700                 default:
8701                         return ret;
8702                 }
8703         }
8704         return 0;
8705 }
8706
8707 int nfs4_destroy_clientid(struct nfs_client *clp)
8708 {
8709         const struct cred *cred;
8710         int ret = 0;
8711
8712         if (clp->cl_mvops->minor_version < 1)
8713                 goto out;
8714         if (clp->cl_exchange_flags == 0)
8715                 goto out;
8716         if (clp->cl_preserve_clid)
8717                 goto out;
8718         cred = nfs4_get_clid_cred(clp);
8719         ret = nfs4_proc_destroy_clientid(clp, cred);
8720         put_cred(cred);
8721         switch (ret) {
8722         case 0:
8723         case -NFS4ERR_STALE_CLIENTID:
8724                 clp->cl_exchange_flags = 0;
8725         }
8726 out:
8727         return ret;
8728 }
8729
8730 #endif /* CONFIG_NFS_V4_1 */
8731
8732 struct nfs4_get_lease_time_data {
8733         struct nfs4_get_lease_time_args *args;
8734         struct nfs4_get_lease_time_res *res;
8735         struct nfs_client *clp;
8736 };
8737
8738 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8739                                         void *calldata)
8740 {
8741         struct nfs4_get_lease_time_data *data =
8742                         (struct nfs4_get_lease_time_data *)calldata;
8743
8744         dprintk("--> %s\n", __func__);
8745         /* just setup sequence, do not trigger session recovery
8746            since we're invoked within one */
8747         nfs4_setup_sequence(data->clp,
8748                         &data->args->la_seq_args,
8749                         &data->res->lr_seq_res,
8750                         task);
8751         dprintk("<-- %s\n", __func__);
8752 }
8753
8754 /*
8755  * Called from nfs4_state_manager thread for session setup, so don't recover
8756  * from sequence operation or clientid errors.
8757  */
8758 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8759 {
8760         struct nfs4_get_lease_time_data *data =
8761                         (struct nfs4_get_lease_time_data *)calldata;
8762
8763         dprintk("--> %s\n", __func__);
8764         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8765                 return;
8766         switch (task->tk_status) {
8767         case -NFS4ERR_DELAY:
8768         case -NFS4ERR_GRACE:
8769                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8770                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8771                 task->tk_status = 0;
8772                 fallthrough;
8773         case -NFS4ERR_RETRY_UNCACHED_REP:
8774                 rpc_restart_call_prepare(task);
8775                 return;
8776         }
8777         dprintk("<-- %s\n", __func__);
8778 }
8779
8780 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8781         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8782         .rpc_call_done = nfs4_get_lease_time_done,
8783 };
8784
8785 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8786 {
8787         struct nfs4_get_lease_time_args args;
8788         struct nfs4_get_lease_time_res res = {
8789                 .lr_fsinfo = fsinfo,
8790         };
8791         struct nfs4_get_lease_time_data data = {
8792                 .args = &args,
8793                 .res = &res,
8794                 .clp = clp,
8795         };
8796         struct rpc_message msg = {
8797                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8798                 .rpc_argp = &args,
8799                 .rpc_resp = &res,
8800         };
8801         struct rpc_task_setup task_setup = {
8802                 .rpc_client = clp->cl_rpcclient,
8803                 .rpc_message = &msg,
8804                 .callback_ops = &nfs4_get_lease_time_ops,
8805                 .callback_data = &data,
8806                 .flags = RPC_TASK_TIMEOUT,
8807         };
8808
8809         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8810         return nfs4_call_sync_custom(&task_setup);
8811 }
8812
8813 #ifdef CONFIG_NFS_V4_1
8814
8815 /*
8816  * Initialize the values to be used by the client in CREATE_SESSION
8817  * If nfs4_init_session set the fore channel request and response sizes,
8818  * use them.
8819  *
8820  * Set the back channel max_resp_sz_cached to zero to force the client to
8821  * always set csa_cachethis to FALSE because the current implementation
8822  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8823  */
8824 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8825                                     struct rpc_clnt *clnt)
8826 {
8827         unsigned int max_rqst_sz, max_resp_sz;
8828         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8829         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8830
8831         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8832         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8833
8834         /* Fore channel attributes */
8835         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8836         args->fc_attrs.max_resp_sz = max_resp_sz;
8837         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8838         args->fc_attrs.max_reqs = max_session_slots;
8839
8840         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8841                 "max_ops=%u max_reqs=%u\n",
8842                 __func__,
8843                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8844                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8845
8846         /* Back channel attributes */
8847         args->bc_attrs.max_rqst_sz = max_bc_payload;
8848         args->bc_attrs.max_resp_sz = max_bc_payload;
8849         args->bc_attrs.max_resp_sz_cached = 0;
8850         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8851         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8852         if (args->bc_attrs.max_reqs > max_bc_slots)
8853                 args->bc_attrs.max_reqs = max_bc_slots;
8854
8855         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8856                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8857                 __func__,
8858                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8859                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8860                 args->bc_attrs.max_reqs);
8861 }
8862
8863 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8864                 struct nfs41_create_session_res *res)
8865 {
8866         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8867         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8868
8869         if (rcvd->max_resp_sz > sent->max_resp_sz)
8870                 return -EINVAL;
8871         /*
8872          * Our requested max_ops is the minimum we need; we're not
8873          * prepared to break up compounds into smaller pieces than that.
8874          * So, no point even trying to continue if the server won't
8875          * cooperate:
8876          */
8877         if (rcvd->max_ops < sent->max_ops)
8878                 return -EINVAL;
8879         if (rcvd->max_reqs == 0)
8880                 return -EINVAL;
8881         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8882                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8883         return 0;
8884 }
8885
8886 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8887                 struct nfs41_create_session_res *res)
8888 {
8889         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8890         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8891
8892         if (!(res->flags & SESSION4_BACK_CHAN))
8893                 goto out;
8894         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8895                 return -EINVAL;
8896         if (rcvd->max_resp_sz < sent->max_resp_sz)
8897                 return -EINVAL;
8898         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8899                 return -EINVAL;
8900         if (rcvd->max_ops > sent->max_ops)
8901                 return -EINVAL;
8902         if (rcvd->max_reqs > sent->max_reqs)
8903                 return -EINVAL;
8904 out:
8905         return 0;
8906 }
8907
8908 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8909                                      struct nfs41_create_session_res *res)
8910 {
8911         int ret;
8912
8913         ret = nfs4_verify_fore_channel_attrs(args, res);
8914         if (ret)
8915                 return ret;
8916         return nfs4_verify_back_channel_attrs(args, res);
8917 }
8918
8919 static void nfs4_update_session(struct nfs4_session *session,
8920                 struct nfs41_create_session_res *res)
8921 {
8922         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8923         /* Mark client id and session as being confirmed */
8924         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8925         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8926         session->flags = res->flags;
8927         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8928         if (res->flags & SESSION4_BACK_CHAN)
8929                 memcpy(&session->bc_attrs, &res->bc_attrs,
8930                                 sizeof(session->bc_attrs));
8931 }
8932
8933 static int _nfs4_proc_create_session(struct nfs_client *clp,
8934                 const struct cred *cred)
8935 {
8936         struct nfs4_session *session = clp->cl_session;
8937         struct nfs41_create_session_args args = {
8938                 .client = clp,
8939                 .clientid = clp->cl_clientid,
8940                 .seqid = clp->cl_seqid,
8941                 .cb_program = NFS4_CALLBACK,
8942         };
8943         struct nfs41_create_session_res res;
8944
8945         struct rpc_message msg = {
8946                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8947                 .rpc_argp = &args,
8948                 .rpc_resp = &res,
8949                 .rpc_cred = cred,
8950         };
8951         int status;
8952
8953         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8954         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8955
8956         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8957                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8958         trace_nfs4_create_session(clp, status);
8959
8960         switch (status) {
8961         case -NFS4ERR_STALE_CLIENTID:
8962         case -NFS4ERR_DELAY:
8963         case -ETIMEDOUT:
8964         case -EACCES:
8965         case -EAGAIN:
8966                 goto out;
8967         }
8968
8969         clp->cl_seqid++;
8970         if (!status) {
8971                 /* Verify the session's negotiated channel_attrs values */
8972                 status = nfs4_verify_channel_attrs(&args, &res);
8973                 /* Increment the clientid slot sequence id */
8974                 if (status)
8975                         goto out;
8976                 nfs4_update_session(session, &res);
8977         }
8978 out:
8979         return status;
8980 }
8981
8982 /*
8983  * Issues a CREATE_SESSION operation to the server.
8984  * It is the responsibility of the caller to verify the session is
8985  * expired before calling this routine.
8986  */
8987 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8988 {
8989         int status;
8990         unsigned *ptr;
8991         struct nfs4_session *session = clp->cl_session;
8992
8993         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8994
8995         status = _nfs4_proc_create_session(clp, cred);
8996         if (status)
8997                 goto out;
8998
8999         /* Init or reset the session slot tables */
9000         status = nfs4_setup_session_slot_tables(session);
9001         dprintk("slot table setup returned %d\n", status);
9002         if (status)
9003                 goto out;
9004
9005         ptr = (unsigned *)&session->sess_id.data[0];
9006         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9007                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9008 out:
9009         dprintk("<-- %s\n", __func__);
9010         return status;
9011 }
9012
9013 /*
9014  * Issue the over-the-wire RPC DESTROY_SESSION.
9015  * The caller must serialize access to this routine.
9016  */
9017 int nfs4_proc_destroy_session(struct nfs4_session *session,
9018                 const struct cred *cred)
9019 {
9020         struct rpc_message msg = {
9021                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9022                 .rpc_argp = session,
9023                 .rpc_cred = cred,
9024         };
9025         int status = 0;
9026
9027         dprintk("--> nfs4_proc_destroy_session\n");
9028
9029         /* session is still being setup */
9030         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9031                 return 0;
9032
9033         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9034                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9035         trace_nfs4_destroy_session(session->clp, status);
9036
9037         if (status)
9038                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9039                         "Session has been destroyed regardless...\n", status);
9040
9041         dprintk("<-- nfs4_proc_destroy_session\n");
9042         return status;
9043 }
9044
9045 /*
9046  * Renew the cl_session lease.
9047  */
9048 struct nfs4_sequence_data {
9049         struct nfs_client *clp;
9050         struct nfs4_sequence_args args;
9051         struct nfs4_sequence_res res;
9052 };
9053
9054 static void nfs41_sequence_release(void *data)
9055 {
9056         struct nfs4_sequence_data *calldata = data;
9057         struct nfs_client *clp = calldata->clp;
9058
9059         if (refcount_read(&clp->cl_count) > 1)
9060                 nfs4_schedule_state_renewal(clp);
9061         nfs_put_client(clp);
9062         kfree(calldata);
9063 }
9064
9065 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9066 {
9067         switch(task->tk_status) {
9068         case -NFS4ERR_DELAY:
9069                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9070                 return -EAGAIN;
9071         default:
9072                 nfs4_schedule_lease_recovery(clp);
9073         }
9074         return 0;
9075 }
9076
9077 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9078 {
9079         struct nfs4_sequence_data *calldata = data;
9080         struct nfs_client *clp = calldata->clp;
9081
9082         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9083                 return;
9084
9085         trace_nfs4_sequence(clp, task->tk_status);
9086         if (task->tk_status < 0) {
9087                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9088                 if (refcount_read(&clp->cl_count) == 1)
9089                         goto out;
9090
9091                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9092                         rpc_restart_call_prepare(task);
9093                         return;
9094                 }
9095         }
9096         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9097 out:
9098         dprintk("<-- %s\n", __func__);
9099 }
9100
9101 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9102 {
9103         struct nfs4_sequence_data *calldata = data;
9104         struct nfs_client *clp = calldata->clp;
9105         struct nfs4_sequence_args *args;
9106         struct nfs4_sequence_res *res;
9107
9108         args = task->tk_msg.rpc_argp;
9109         res = task->tk_msg.rpc_resp;
9110
9111         nfs4_setup_sequence(clp, args, res, task);
9112 }
9113
9114 static const struct rpc_call_ops nfs41_sequence_ops = {
9115         .rpc_call_done = nfs41_sequence_call_done,
9116         .rpc_call_prepare = nfs41_sequence_prepare,
9117         .rpc_release = nfs41_sequence_release,
9118 };
9119
9120 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9121                 const struct cred *cred,
9122                 struct nfs4_slot *slot,
9123                 bool is_privileged)
9124 {
9125         struct nfs4_sequence_data *calldata;
9126         struct rpc_message msg = {
9127                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9128                 .rpc_cred = cred,
9129         };
9130         struct rpc_task_setup task_setup_data = {
9131                 .rpc_client = clp->cl_rpcclient,
9132                 .rpc_message = &msg,
9133                 .callback_ops = &nfs41_sequence_ops,
9134                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
9135         };
9136         struct rpc_task *ret;
9137
9138         ret = ERR_PTR(-EIO);
9139         if (!refcount_inc_not_zero(&clp->cl_count))
9140                 goto out_err;
9141
9142         ret = ERR_PTR(-ENOMEM);
9143         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9144         if (calldata == NULL)
9145                 goto out_put_clp;
9146         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9147         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9148         msg.rpc_argp = &calldata->args;
9149         msg.rpc_resp = &calldata->res;
9150         calldata->clp = clp;
9151         task_setup_data.callback_data = calldata;
9152
9153         ret = rpc_run_task(&task_setup_data);
9154         if (IS_ERR(ret))
9155                 goto out_err;
9156         return ret;
9157 out_put_clp:
9158         nfs_put_client(clp);
9159 out_err:
9160         nfs41_release_slot(slot);
9161         return ret;
9162 }
9163
9164 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9165 {
9166         struct rpc_task *task;
9167         int ret = 0;
9168
9169         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9170                 return -EAGAIN;
9171         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9172         if (IS_ERR(task))
9173                 ret = PTR_ERR(task);
9174         else
9175                 rpc_put_task_async(task);
9176         dprintk("<-- %s status=%d\n", __func__, ret);
9177         return ret;
9178 }
9179
9180 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9181 {
9182         struct rpc_task *task;
9183         int ret;
9184
9185         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9186         if (IS_ERR(task)) {
9187                 ret = PTR_ERR(task);
9188                 goto out;
9189         }
9190         ret = rpc_wait_for_completion_task(task);
9191         if (!ret)
9192                 ret = task->tk_status;
9193         rpc_put_task(task);
9194 out:
9195         dprintk("<-- %s status=%d\n", __func__, ret);
9196         return ret;
9197 }
9198
9199 struct nfs4_reclaim_complete_data {
9200         struct nfs_client *clp;
9201         struct nfs41_reclaim_complete_args arg;
9202         struct nfs41_reclaim_complete_res res;
9203 };
9204
9205 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9206 {
9207         struct nfs4_reclaim_complete_data *calldata = data;
9208
9209         nfs4_setup_sequence(calldata->clp,
9210                         &calldata->arg.seq_args,
9211                         &calldata->res.seq_res,
9212                         task);
9213 }
9214
9215 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9216 {
9217         switch(task->tk_status) {
9218         case 0:
9219                 wake_up_all(&clp->cl_lock_waitq);
9220                 fallthrough;
9221         case -NFS4ERR_COMPLETE_ALREADY:
9222         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9223                 break;
9224         case -NFS4ERR_DELAY:
9225                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9226                 fallthrough;
9227         case -NFS4ERR_RETRY_UNCACHED_REP:
9228                 return -EAGAIN;
9229         case -NFS4ERR_BADSESSION:
9230         case -NFS4ERR_DEADSESSION:
9231         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9232                 break;
9233         default:
9234                 nfs4_schedule_lease_recovery(clp);
9235         }
9236         return 0;
9237 }
9238
9239 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9240 {
9241         struct nfs4_reclaim_complete_data *calldata = data;
9242         struct nfs_client *clp = calldata->clp;
9243         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9244
9245         dprintk("--> %s\n", __func__);
9246         if (!nfs41_sequence_done(task, res))
9247                 return;
9248
9249         trace_nfs4_reclaim_complete(clp, task->tk_status);
9250         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9251                 rpc_restart_call_prepare(task);
9252                 return;
9253         }
9254         dprintk("<-- %s\n", __func__);
9255 }
9256
9257 static void nfs4_free_reclaim_complete_data(void *data)
9258 {
9259         struct nfs4_reclaim_complete_data *calldata = data;
9260
9261         kfree(calldata);
9262 }
9263
9264 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9265         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9266         .rpc_call_done = nfs4_reclaim_complete_done,
9267         .rpc_release = nfs4_free_reclaim_complete_data,
9268 };
9269
9270 /*
9271  * Issue a global reclaim complete.
9272  */
9273 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9274                 const struct cred *cred)
9275 {
9276         struct nfs4_reclaim_complete_data *calldata;
9277         struct rpc_message msg = {
9278                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9279                 .rpc_cred = cred,
9280         };
9281         struct rpc_task_setup task_setup_data = {
9282                 .rpc_client = clp->cl_rpcclient,
9283                 .rpc_message = &msg,
9284                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9285                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9286         };
9287         int status = -ENOMEM;
9288
9289         dprintk("--> %s\n", __func__);
9290         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9291         if (calldata == NULL)
9292                 goto out;
9293         calldata->clp = clp;
9294         calldata->arg.one_fs = 0;
9295
9296         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9297         msg.rpc_argp = &calldata->arg;
9298         msg.rpc_resp = &calldata->res;
9299         task_setup_data.callback_data = calldata;
9300         status = nfs4_call_sync_custom(&task_setup_data);
9301 out:
9302         dprintk("<-- %s status=%d\n", __func__, status);
9303         return status;
9304 }
9305
9306 static void
9307 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9308 {
9309         struct nfs4_layoutget *lgp = calldata;
9310         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9311
9312         dprintk("--> %s\n", __func__);
9313         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9314                                 &lgp->res.seq_res, task);
9315         dprintk("<-- %s\n", __func__);
9316 }
9317
9318 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9319 {
9320         struct nfs4_layoutget *lgp = calldata;
9321
9322         dprintk("--> %s\n", __func__);
9323         nfs41_sequence_process(task, &lgp->res.seq_res);
9324         dprintk("<-- %s\n", __func__);
9325 }
9326
9327 static int
9328 nfs4_layoutget_handle_exception(struct rpc_task *task,
9329                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9330 {
9331         struct inode *inode = lgp->args.inode;
9332         struct nfs_server *server = NFS_SERVER(inode);
9333         struct pnfs_layout_hdr *lo;
9334         int nfs4err = task->tk_status;
9335         int err, status = 0;
9336         LIST_HEAD(head);
9337
9338         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9339
9340         nfs4_sequence_free_slot(&lgp->res.seq_res);
9341
9342         switch (nfs4err) {
9343         case 0:
9344                 goto out;
9345
9346         /*
9347          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9348          * on the file. set tk_status to -ENODATA to tell upper layer to
9349          * retry go inband.
9350          */
9351         case -NFS4ERR_LAYOUTUNAVAILABLE:
9352                 status = -ENODATA;
9353                 goto out;
9354         /*
9355          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9356          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9357          */
9358         case -NFS4ERR_BADLAYOUT:
9359                 status = -EOVERFLOW;
9360                 goto out;
9361         /*
9362          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9363          * (or clients) writing to the same RAID stripe except when
9364          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9365          *
9366          * Treat it like we would RECALLCONFLICT -- we retry for a little
9367          * while, and then eventually give up.
9368          */
9369         case -NFS4ERR_LAYOUTTRYLATER:
9370                 if (lgp->args.minlength == 0) {
9371                         status = -EOVERFLOW;
9372                         goto out;
9373                 }
9374                 status = -EBUSY;
9375                 break;
9376         case -NFS4ERR_RECALLCONFLICT:
9377                 status = -ERECALLCONFLICT;
9378                 break;
9379         case -NFS4ERR_DELEG_REVOKED:
9380         case -NFS4ERR_ADMIN_REVOKED:
9381         case -NFS4ERR_EXPIRED:
9382         case -NFS4ERR_BAD_STATEID:
9383                 exception->timeout = 0;
9384                 spin_lock(&inode->i_lock);
9385                 lo = NFS_I(inode)->layout;
9386                 /* If the open stateid was bad, then recover it. */
9387                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9388                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9389                         spin_unlock(&inode->i_lock);
9390                         exception->state = lgp->args.ctx->state;
9391                         exception->stateid = &lgp->args.stateid;
9392                         break;
9393                 }
9394
9395                 /*
9396                  * Mark the bad layout state as invalid, then retry
9397                  */
9398                 pnfs_mark_layout_stateid_invalid(lo, &head);
9399                 spin_unlock(&inode->i_lock);
9400                 nfs_commit_inode(inode, 0);
9401                 pnfs_free_lseg_list(&head);
9402                 status = -EAGAIN;
9403                 goto out;
9404         }
9405
9406         err = nfs4_handle_exception(server, nfs4err, exception);
9407         if (!status) {
9408                 if (exception->retry)
9409                         status = -EAGAIN;
9410                 else
9411                         status = err;
9412         }
9413 out:
9414         dprintk("<-- %s\n", __func__);
9415         return status;
9416 }
9417
9418 size_t max_response_pages(struct nfs_server *server)
9419 {
9420         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9421         return nfs_page_array_len(0, max_resp_sz);
9422 }
9423
9424 static void nfs4_layoutget_release(void *calldata)
9425 {
9426         struct nfs4_layoutget *lgp = calldata;
9427
9428         dprintk("--> %s\n", __func__);
9429         nfs4_sequence_free_slot(&lgp->res.seq_res);
9430         pnfs_layoutget_free(lgp);
9431         dprintk("<-- %s\n", __func__);
9432 }
9433
9434 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9435         .rpc_call_prepare = nfs4_layoutget_prepare,
9436         .rpc_call_done = nfs4_layoutget_done,
9437         .rpc_release = nfs4_layoutget_release,
9438 };
9439
9440 struct pnfs_layout_segment *
9441 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9442 {
9443         struct inode *inode = lgp->args.inode;
9444         struct nfs_server *server = NFS_SERVER(inode);
9445         struct rpc_task *task;
9446         struct rpc_message msg = {
9447                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9448                 .rpc_argp = &lgp->args,
9449                 .rpc_resp = &lgp->res,
9450                 .rpc_cred = lgp->cred,
9451         };
9452         struct rpc_task_setup task_setup_data = {
9453                 .rpc_client = server->client,
9454                 .rpc_message = &msg,
9455                 .callback_ops = &nfs4_layoutget_call_ops,
9456                 .callback_data = lgp,
9457                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
9458         };
9459         struct pnfs_layout_segment *lseg = NULL;
9460         struct nfs4_exception exception = {
9461                 .inode = inode,
9462                 .timeout = *timeout,
9463         };
9464         int status = 0;
9465
9466         dprintk("--> %s\n", __func__);
9467
9468         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9469         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9470
9471         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9472
9473         task = rpc_run_task(&task_setup_data);
9474
9475         status = rpc_wait_for_completion_task(task);
9476         if (status != 0)
9477                 goto out;
9478
9479         if (task->tk_status < 0) {
9480                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9481                 *timeout = exception.timeout;
9482         } else if (lgp->res.layoutp->len == 0) {
9483                 status = -EAGAIN;
9484                 *timeout = nfs4_update_delay(&exception.timeout);
9485         } else
9486                 lseg = pnfs_layout_process(lgp);
9487 out:
9488         trace_nfs4_layoutget(lgp->args.ctx,
9489                         &lgp->args.range,
9490                         &lgp->res.range,
9491                         &lgp->res.stateid,
9492                         status);
9493
9494         rpc_put_task(task);
9495         dprintk("<-- %s status=%d\n", __func__, status);
9496         if (status)
9497                 return ERR_PTR(status);
9498         return lseg;
9499 }
9500
9501 static void
9502 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9503 {
9504         struct nfs4_layoutreturn *lrp = calldata;
9505
9506         dprintk("--> %s\n", __func__);
9507         nfs4_setup_sequence(lrp->clp,
9508                         &lrp->args.seq_args,
9509                         &lrp->res.seq_res,
9510                         task);
9511         if (!pnfs_layout_is_valid(lrp->args.layout))
9512                 rpc_exit(task, 0);
9513 }
9514
9515 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9516 {
9517         struct nfs4_layoutreturn *lrp = calldata;
9518         struct nfs_server *server;
9519
9520         dprintk("--> %s\n", __func__);
9521
9522         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9523                 return;
9524
9525         /*
9526          * Was there an RPC level error? Assume the call succeeded,
9527          * and that we need to release the layout
9528          */
9529         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9530                 lrp->res.lrs_present = 0;
9531                 return;
9532         }
9533
9534         server = NFS_SERVER(lrp->args.inode);
9535         switch (task->tk_status) {
9536         case -NFS4ERR_OLD_STATEID:
9537                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9538                                         &lrp->args.range,
9539                                         lrp->args.inode))
9540                         goto out_restart;
9541                 fallthrough;
9542         default:
9543                 task->tk_status = 0;
9544                 fallthrough;
9545         case 0:
9546                 break;
9547         case -NFS4ERR_DELAY:
9548                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9549                         break;
9550                 goto out_restart;
9551         }
9552         dprintk("<-- %s\n", __func__);
9553         return;
9554 out_restart:
9555         task->tk_status = 0;
9556         nfs4_sequence_free_slot(&lrp->res.seq_res);
9557         rpc_restart_call_prepare(task);
9558 }
9559
9560 static void nfs4_layoutreturn_release(void *calldata)
9561 {
9562         struct nfs4_layoutreturn *lrp = calldata;
9563         struct pnfs_layout_hdr *lo = lrp->args.layout;
9564
9565         dprintk("--> %s\n", __func__);
9566         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9567                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9568         nfs4_sequence_free_slot(&lrp->res.seq_res);
9569         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9570                 lrp->ld_private.ops->free(&lrp->ld_private);
9571         pnfs_put_layout_hdr(lrp->args.layout);
9572         nfs_iput_and_deactive(lrp->inode);
9573         put_cred(lrp->cred);
9574         kfree(calldata);
9575         dprintk("<-- %s\n", __func__);
9576 }
9577
9578 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9579         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9580         .rpc_call_done = nfs4_layoutreturn_done,
9581         .rpc_release = nfs4_layoutreturn_release,
9582 };
9583
9584 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9585 {
9586         struct rpc_task *task;
9587         struct rpc_message msg = {
9588                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9589                 .rpc_argp = &lrp->args,
9590                 .rpc_resp = &lrp->res,
9591                 .rpc_cred = lrp->cred,
9592         };
9593         struct rpc_task_setup task_setup_data = {
9594                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9595                 .rpc_message = &msg,
9596                 .callback_ops = &nfs4_layoutreturn_call_ops,
9597                 .callback_data = lrp,
9598         };
9599         int status = 0;
9600
9601         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9602                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9603                         &task_setup_data.rpc_client, &msg);
9604
9605         dprintk("--> %s\n", __func__);
9606         if (!sync) {
9607                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9608                 if (!lrp->inode) {
9609                         nfs4_layoutreturn_release(lrp);
9610                         return -EAGAIN;
9611                 }
9612                 task_setup_data.flags |= RPC_TASK_ASYNC;
9613         }
9614         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9615         task = rpc_run_task(&task_setup_data);
9616         if (IS_ERR(task))
9617                 return PTR_ERR(task);
9618         if (sync)
9619                 status = task->tk_status;
9620         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9621         dprintk("<-- %s status=%d\n", __func__, status);
9622         rpc_put_task(task);
9623         return status;
9624 }
9625
9626 static int
9627 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9628                 struct pnfs_device *pdev,
9629                 const struct cred *cred)
9630 {
9631         struct nfs4_getdeviceinfo_args args = {
9632                 .pdev = pdev,
9633                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9634                         NOTIFY_DEVICEID4_DELETE,
9635         };
9636         struct nfs4_getdeviceinfo_res res = {
9637                 .pdev = pdev,
9638         };
9639         struct rpc_message msg = {
9640                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9641                 .rpc_argp = &args,
9642                 .rpc_resp = &res,
9643                 .rpc_cred = cred,
9644         };
9645         int status;
9646
9647         dprintk("--> %s\n", __func__);
9648         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9649         if (res.notification & ~args.notify_types)
9650                 dprintk("%s: unsupported notification\n", __func__);
9651         if (res.notification != args.notify_types)
9652                 pdev->nocache = 1;
9653
9654         dprintk("<-- %s status=%d\n", __func__, status);
9655
9656         return status;
9657 }
9658
9659 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9660                 struct pnfs_device *pdev,
9661                 const struct cred *cred)
9662 {
9663         struct nfs4_exception exception = { };
9664         int err;
9665
9666         do {
9667                 err = nfs4_handle_exception(server,
9668                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9669                                         &exception);
9670         } while (exception.retry);
9671         return err;
9672 }
9673 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9674
9675 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9676 {
9677         struct nfs4_layoutcommit_data *data = calldata;
9678         struct nfs_server *server = NFS_SERVER(data->args.inode);
9679
9680         nfs4_setup_sequence(server->nfs_client,
9681                         &data->args.seq_args,
9682                         &data->res.seq_res,
9683                         task);
9684 }
9685
9686 static void
9687 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9688 {
9689         struct nfs4_layoutcommit_data *data = calldata;
9690         struct nfs_server *server = NFS_SERVER(data->args.inode);
9691
9692         if (!nfs41_sequence_done(task, &data->res.seq_res))
9693                 return;
9694
9695         switch (task->tk_status) { /* Just ignore these failures */
9696         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9697         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9698         case -NFS4ERR_BADLAYOUT:     /* no layout */
9699         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9700                 task->tk_status = 0;
9701         case 0:
9702                 break;
9703         default:
9704                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9705                         rpc_restart_call_prepare(task);
9706                         return;
9707                 }
9708         }
9709 }
9710
9711 static void nfs4_layoutcommit_release(void *calldata)
9712 {
9713         struct nfs4_layoutcommit_data *data = calldata;
9714
9715         pnfs_cleanup_layoutcommit(data);
9716         nfs_post_op_update_inode_force_wcc(data->args.inode,
9717                                            data->res.fattr);
9718         put_cred(data->cred);
9719         nfs_iput_and_deactive(data->inode);
9720         kfree(data);
9721 }
9722
9723 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9724         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9725         .rpc_call_done = nfs4_layoutcommit_done,
9726         .rpc_release = nfs4_layoutcommit_release,
9727 };
9728
9729 int
9730 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9731 {
9732         struct rpc_message msg = {
9733                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9734                 .rpc_argp = &data->args,
9735                 .rpc_resp = &data->res,
9736                 .rpc_cred = data->cred,
9737         };
9738         struct rpc_task_setup task_setup_data = {
9739                 .task = &data->task,
9740                 .rpc_client = NFS_CLIENT(data->args.inode),
9741                 .rpc_message = &msg,
9742                 .callback_ops = &nfs4_layoutcommit_ops,
9743                 .callback_data = data,
9744         };
9745         struct rpc_task *task;
9746         int status = 0;
9747
9748         dprintk("NFS: initiating layoutcommit call. sync %d "
9749                 "lbw: %llu inode %lu\n", sync,
9750                 data->args.lastbytewritten,
9751                 data->args.inode->i_ino);
9752
9753         if (!sync) {
9754                 data->inode = nfs_igrab_and_active(data->args.inode);
9755                 if (data->inode == NULL) {
9756                         nfs4_layoutcommit_release(data);
9757                         return -EAGAIN;
9758                 }
9759                 task_setup_data.flags = RPC_TASK_ASYNC;
9760         }
9761         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9762         task = rpc_run_task(&task_setup_data);
9763         if (IS_ERR(task))
9764                 return PTR_ERR(task);
9765         if (sync)
9766                 status = task->tk_status;
9767         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9768         dprintk("%s: status %d\n", __func__, status);
9769         rpc_put_task(task);
9770         return status;
9771 }
9772
9773 /*
9774  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9775  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9776  */
9777 static int
9778 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9779                     struct nfs_fsinfo *info,
9780                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9781 {
9782         struct nfs41_secinfo_no_name_args args = {
9783                 .style = SECINFO_STYLE_CURRENT_FH,
9784         };
9785         struct nfs4_secinfo_res res = {
9786                 .flavors = flavors,
9787         };
9788         struct rpc_message msg = {
9789                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9790                 .rpc_argp = &args,
9791                 .rpc_resp = &res,
9792         };
9793         struct nfs4_call_sync_data data = {
9794                 .seq_server = server,
9795                 .seq_args = &args.seq_args,
9796                 .seq_res = &res.seq_res,
9797         };
9798         struct rpc_task_setup task_setup = {
9799                 .rpc_client = server->client,
9800                 .rpc_message = &msg,
9801                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9802                 .callback_data = &data,
9803                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9804         };
9805         const struct cred *cred = NULL;
9806         int status;
9807
9808         if (use_integrity) {
9809                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
9810
9811                 cred = nfs4_get_clid_cred(server->nfs_client);
9812                 msg.rpc_cred = cred;
9813         }
9814
9815         dprintk("--> %s\n", __func__);
9816         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9817         status = nfs4_call_sync_custom(&task_setup);
9818         dprintk("<-- %s status=%d\n", __func__, status);
9819
9820         put_cred(cred);
9821
9822         return status;
9823 }
9824
9825 static int
9826 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9827                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9828 {
9829         struct nfs4_exception exception = {
9830                 .interruptible = true,
9831         };
9832         int err;
9833         do {
9834                 /* first try using integrity protection */
9835                 err = -NFS4ERR_WRONGSEC;
9836
9837                 /* try to use integrity protection with machine cred */
9838                 if (_nfs4_is_integrity_protected(server->nfs_client))
9839                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9840                                                           flavors, true);
9841
9842                 /*
9843                  * if unable to use integrity protection, or SECINFO with
9844                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9845                  * disallowed by spec, but exists in deployed servers) use
9846                  * the current filesystem's rpc_client and the user cred.
9847                  */
9848                 if (err == -NFS4ERR_WRONGSEC)
9849                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9850                                                           flavors, false);
9851
9852                 switch (err) {
9853                 case 0:
9854                 case -NFS4ERR_WRONGSEC:
9855                 case -ENOTSUPP:
9856                         goto out;
9857                 default:
9858                         err = nfs4_handle_exception(server, err, &exception);
9859                 }
9860         } while (exception.retry);
9861 out:
9862         return err;
9863 }
9864
9865 static int
9866 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9867                     struct nfs_fsinfo *info)
9868 {
9869         int err;
9870         struct page *page;
9871         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9872         struct nfs4_secinfo_flavors *flavors;
9873         struct nfs4_secinfo4 *secinfo;
9874         int i;
9875
9876         page = alloc_page(GFP_KERNEL);
9877         if (!page) {
9878                 err = -ENOMEM;
9879                 goto out;
9880         }
9881
9882         flavors = page_address(page);
9883         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9884
9885         /*
9886          * Fall back on "guess and check" method if
9887          * the server doesn't support SECINFO_NO_NAME
9888          */
9889         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9890                 err = nfs4_find_root_sec(server, fhandle, info);
9891                 goto out_freepage;
9892         }
9893         if (err)
9894                 goto out_freepage;
9895
9896         for (i = 0; i < flavors->num_flavors; i++) {
9897                 secinfo = &flavors->flavors[i];
9898
9899                 switch (secinfo->flavor) {
9900                 case RPC_AUTH_NULL:
9901                 case RPC_AUTH_UNIX:
9902                 case RPC_AUTH_GSS:
9903                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9904                                         &secinfo->flavor_info);
9905                         break;
9906                 default:
9907                         flavor = RPC_AUTH_MAXFLAVOR;
9908                         break;
9909                 }
9910
9911                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9912                         flavor = RPC_AUTH_MAXFLAVOR;
9913
9914                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9915                         err = nfs4_lookup_root_sec(server, fhandle,
9916                                                    info, flavor);
9917                         if (!err)
9918                                 break;
9919                 }
9920         }
9921
9922         if (flavor == RPC_AUTH_MAXFLAVOR)
9923                 err = -EPERM;
9924
9925 out_freepage:
9926         put_page(page);
9927         if (err == -EACCES)
9928                 return -EPERM;
9929 out:
9930         return err;
9931 }
9932
9933 static int _nfs41_test_stateid(struct nfs_server *server,
9934                 nfs4_stateid *stateid,
9935                 const struct cred *cred)
9936 {
9937         int status;
9938         struct nfs41_test_stateid_args args = {
9939                 .stateid = stateid,
9940         };
9941         struct nfs41_test_stateid_res res;
9942         struct rpc_message msg = {
9943                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9944                 .rpc_argp = &args,
9945                 .rpc_resp = &res,
9946                 .rpc_cred = cred,
9947         };
9948         struct rpc_clnt *rpc_client = server->client;
9949
9950         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9951                 &rpc_client, &msg);
9952
9953         dprintk("NFS call  test_stateid %p\n", stateid);
9954         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9955         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9956                         &args.seq_args, &res.seq_res);
9957         if (status != NFS_OK) {
9958                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9959                 return status;
9960         }
9961         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9962         return -res.status;
9963 }
9964
9965 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9966                 int err, struct nfs4_exception *exception)
9967 {
9968         exception->retry = 0;
9969         switch(err) {
9970         case -NFS4ERR_DELAY:
9971         case -NFS4ERR_RETRY_UNCACHED_REP:
9972                 nfs4_handle_exception(server, err, exception);
9973                 break;
9974         case -NFS4ERR_BADSESSION:
9975         case -NFS4ERR_BADSLOT:
9976         case -NFS4ERR_BAD_HIGH_SLOT:
9977         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9978         case -NFS4ERR_DEADSESSION:
9979                 nfs4_do_handle_exception(server, err, exception);
9980         }
9981 }
9982
9983 /**
9984  * nfs41_test_stateid - perform a TEST_STATEID operation
9985  *
9986  * @server: server / transport on which to perform the operation
9987  * @stateid: state ID to test
9988  * @cred: credential
9989  *
9990  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9991  * Otherwise a negative NFS4ERR value is returned if the operation
9992  * failed or the state ID is not currently valid.
9993  */
9994 static int nfs41_test_stateid(struct nfs_server *server,
9995                 nfs4_stateid *stateid,
9996                 const struct cred *cred)
9997 {
9998         struct nfs4_exception exception = {
9999                 .interruptible = true,
10000         };
10001         int err;
10002         do {
10003                 err = _nfs41_test_stateid(server, stateid, cred);
10004                 nfs4_handle_delay_or_session_error(server, err, &exception);
10005         } while (exception.retry);
10006         return err;
10007 }
10008
10009 struct nfs_free_stateid_data {
10010         struct nfs_server *server;
10011         struct nfs41_free_stateid_args args;
10012         struct nfs41_free_stateid_res res;
10013 };
10014
10015 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10016 {
10017         struct nfs_free_stateid_data *data = calldata;
10018         nfs4_setup_sequence(data->server->nfs_client,
10019                         &data->args.seq_args,
10020                         &data->res.seq_res,
10021                         task);
10022 }
10023
10024 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10025 {
10026         struct nfs_free_stateid_data *data = calldata;
10027
10028         nfs41_sequence_done(task, &data->res.seq_res);
10029
10030         switch (task->tk_status) {
10031         case -NFS4ERR_DELAY:
10032                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10033                         rpc_restart_call_prepare(task);
10034         }
10035 }
10036
10037 static void nfs41_free_stateid_release(void *calldata)
10038 {
10039         kfree(calldata);
10040 }
10041
10042 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10043         .rpc_call_prepare = nfs41_free_stateid_prepare,
10044         .rpc_call_done = nfs41_free_stateid_done,
10045         .rpc_release = nfs41_free_stateid_release,
10046 };
10047
10048 /**
10049  * nfs41_free_stateid - perform a FREE_STATEID operation
10050  *
10051  * @server: server / transport on which to perform the operation
10052  * @stateid: state ID to release
10053  * @cred: credential
10054  * @privileged: set to true if this call needs to be privileged
10055  *
10056  * Note: this function is always asynchronous.
10057  */
10058 static int nfs41_free_stateid(struct nfs_server *server,
10059                 const nfs4_stateid *stateid,
10060                 const struct cred *cred,
10061                 bool privileged)
10062 {
10063         struct rpc_message msg = {
10064                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10065                 .rpc_cred = cred,
10066         };
10067         struct rpc_task_setup task_setup = {
10068                 .rpc_client = server->client,
10069                 .rpc_message = &msg,
10070                 .callback_ops = &nfs41_free_stateid_ops,
10071                 .flags = RPC_TASK_ASYNC,
10072         };
10073         struct nfs_free_stateid_data *data;
10074         struct rpc_task *task;
10075
10076         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10077                 &task_setup.rpc_client, &msg);
10078
10079         dprintk("NFS call  free_stateid %p\n", stateid);
10080         data = kmalloc(sizeof(*data), GFP_NOFS);
10081         if (!data)
10082                 return -ENOMEM;
10083         data->server = server;
10084         nfs4_stateid_copy(&data->args.stateid, stateid);
10085
10086         task_setup.callback_data = data;
10087
10088         msg.rpc_argp = &data->args;
10089         msg.rpc_resp = &data->res;
10090         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10091         task = rpc_run_task(&task_setup);
10092         if (IS_ERR(task))
10093                 return PTR_ERR(task);
10094         rpc_put_task(task);
10095         return 0;
10096 }
10097
10098 static void
10099 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10100 {
10101         const struct cred *cred = lsp->ls_state->owner->so_cred;
10102
10103         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10104         nfs4_free_lock_state(server, lsp);
10105 }
10106
10107 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10108                 const nfs4_stateid *s2)
10109 {
10110         if (s1->type != s2->type)
10111                 return false;
10112
10113         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10114                 return false;
10115
10116         if (s1->seqid == s2->seqid)
10117                 return true;
10118
10119         return s1->seqid == 0 || s2->seqid == 0;
10120 }
10121
10122 #endif /* CONFIG_NFS_V4_1 */
10123
10124 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10125                 const nfs4_stateid *s2)
10126 {
10127         return nfs4_stateid_match(s1, s2);
10128 }
10129
10130
10131 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10132         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10133         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10134         .recover_open   = nfs4_open_reclaim,
10135         .recover_lock   = nfs4_lock_reclaim,
10136         .establish_clid = nfs4_init_clientid,
10137         .detect_trunking = nfs40_discover_server_trunking,
10138 };
10139
10140 #if defined(CONFIG_NFS_V4_1)
10141 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10142         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10143         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10144         .recover_open   = nfs4_open_reclaim,
10145         .recover_lock   = nfs4_lock_reclaim,
10146         .establish_clid = nfs41_init_clientid,
10147         .reclaim_complete = nfs41_proc_reclaim_complete,
10148         .detect_trunking = nfs41_discover_server_trunking,
10149 };
10150 #endif /* CONFIG_NFS_V4_1 */
10151
10152 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10153         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10154         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10155         .recover_open   = nfs40_open_expired,
10156         .recover_lock   = nfs4_lock_expired,
10157         .establish_clid = nfs4_init_clientid,
10158 };
10159
10160 #if defined(CONFIG_NFS_V4_1)
10161 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10162         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10163         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10164         .recover_open   = nfs41_open_expired,
10165         .recover_lock   = nfs41_lock_expired,
10166         .establish_clid = nfs41_init_clientid,
10167 };
10168 #endif /* CONFIG_NFS_V4_1 */
10169
10170 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10171         .sched_state_renewal = nfs4_proc_async_renew,
10172         .get_state_renewal_cred = nfs4_get_renew_cred,
10173         .renew_lease = nfs4_proc_renew,
10174 };
10175
10176 #if defined(CONFIG_NFS_V4_1)
10177 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10178         .sched_state_renewal = nfs41_proc_async_sequence,
10179         .get_state_renewal_cred = nfs4_get_machine_cred,
10180         .renew_lease = nfs4_proc_sequence,
10181 };
10182 #endif
10183
10184 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10185         .get_locations = _nfs40_proc_get_locations,
10186         .fsid_present = _nfs40_proc_fsid_present,
10187 };
10188
10189 #if defined(CONFIG_NFS_V4_1)
10190 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10191         .get_locations = _nfs41_proc_get_locations,
10192         .fsid_present = _nfs41_proc_fsid_present,
10193 };
10194 #endif  /* CONFIG_NFS_V4_1 */
10195
10196 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10197         .minor_version = 0,
10198         .init_caps = NFS_CAP_READDIRPLUS
10199                 | NFS_CAP_ATOMIC_OPEN
10200                 | NFS_CAP_POSIX_LOCK,
10201         .init_client = nfs40_init_client,
10202         .shutdown_client = nfs40_shutdown_client,
10203         .match_stateid = nfs4_match_stateid,
10204         .find_root_sec = nfs4_find_root_sec,
10205         .free_lock_state = nfs4_release_lockowner,
10206         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10207         .alloc_seqid = nfs_alloc_seqid,
10208         .call_sync_ops = &nfs40_call_sync_ops,
10209         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10210         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10211         .state_renewal_ops = &nfs40_state_renewal_ops,
10212         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10213 };
10214
10215 #if defined(CONFIG_NFS_V4_1)
10216 static struct nfs_seqid *
10217 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10218 {
10219         return NULL;
10220 }
10221
10222 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10223         .minor_version = 1,
10224         .init_caps = NFS_CAP_READDIRPLUS
10225                 | NFS_CAP_ATOMIC_OPEN
10226                 | NFS_CAP_POSIX_LOCK
10227                 | NFS_CAP_STATEID_NFSV41
10228                 | NFS_CAP_ATOMIC_OPEN_V1
10229                 | NFS_CAP_LGOPEN,
10230         .init_client = nfs41_init_client,
10231         .shutdown_client = nfs41_shutdown_client,
10232         .match_stateid = nfs41_match_stateid,
10233         .find_root_sec = nfs41_find_root_sec,
10234         .free_lock_state = nfs41_free_lock_state,
10235         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10236         .alloc_seqid = nfs_alloc_no_seqid,
10237         .session_trunk = nfs4_test_session_trunk,
10238         .call_sync_ops = &nfs41_call_sync_ops,
10239         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10240         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10241         .state_renewal_ops = &nfs41_state_renewal_ops,
10242         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10243 };
10244 #endif
10245
10246 #if defined(CONFIG_NFS_V4_2)
10247 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10248         .minor_version = 2,
10249         .init_caps = NFS_CAP_READDIRPLUS
10250                 | NFS_CAP_ATOMIC_OPEN
10251                 | NFS_CAP_POSIX_LOCK
10252                 | NFS_CAP_STATEID_NFSV41
10253                 | NFS_CAP_ATOMIC_OPEN_V1
10254                 | NFS_CAP_LGOPEN
10255                 | NFS_CAP_ALLOCATE
10256                 | NFS_CAP_COPY
10257                 | NFS_CAP_OFFLOAD_CANCEL
10258                 | NFS_CAP_COPY_NOTIFY
10259                 | NFS_CAP_DEALLOCATE
10260                 | NFS_CAP_SEEK
10261                 | NFS_CAP_LAYOUTSTATS
10262                 | NFS_CAP_CLONE
10263                 | NFS_CAP_LAYOUTERROR
10264                 | NFS_CAP_READ_PLUS,
10265         .init_client = nfs41_init_client,
10266         .shutdown_client = nfs41_shutdown_client,
10267         .match_stateid = nfs41_match_stateid,
10268         .find_root_sec = nfs41_find_root_sec,
10269         .free_lock_state = nfs41_free_lock_state,
10270         .call_sync_ops = &nfs41_call_sync_ops,
10271         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10272         .alloc_seqid = nfs_alloc_no_seqid,
10273         .session_trunk = nfs4_test_session_trunk,
10274         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10275         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10276         .state_renewal_ops = &nfs41_state_renewal_ops,
10277         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10278 };
10279 #endif
10280
10281 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10282         [0] = &nfs_v4_0_minor_ops,
10283 #if defined(CONFIG_NFS_V4_1)
10284         [1] = &nfs_v4_1_minor_ops,
10285 #endif
10286 #if defined(CONFIG_NFS_V4_2)
10287         [2] = &nfs_v4_2_minor_ops,
10288 #endif
10289 };
10290
10291 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10292 {
10293         ssize_t error, error2, error3;
10294
10295         error = generic_listxattr(dentry, list, size);
10296         if (error < 0)
10297                 return error;
10298         if (list) {
10299                 list += error;
10300                 size -= error;
10301         }
10302
10303         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10304         if (error2 < 0)
10305                 return error2;
10306
10307         if (list) {
10308                 list += error2;
10309                 size -= error2;
10310         }
10311
10312         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10313         if (error3 < 0)
10314                 return error3;
10315
10316         return error + error2 + error3;
10317 }
10318
10319 static const struct inode_operations nfs4_dir_inode_operations = {
10320         .create         = nfs_create,
10321         .lookup         = nfs_lookup,
10322         .atomic_open    = nfs_atomic_open,
10323         .link           = nfs_link,
10324         .unlink         = nfs_unlink,
10325         .symlink        = nfs_symlink,
10326         .mkdir          = nfs_mkdir,
10327         .rmdir          = nfs_rmdir,
10328         .mknod          = nfs_mknod,
10329         .rename         = nfs_rename,
10330         .permission     = nfs_permission,
10331         .getattr        = nfs_getattr,
10332         .setattr        = nfs_setattr,
10333         .listxattr      = nfs4_listxattr,
10334 };
10335
10336 static const struct inode_operations nfs4_file_inode_operations = {
10337         .permission     = nfs_permission,
10338         .getattr        = nfs_getattr,
10339         .setattr        = nfs_setattr,
10340         .listxattr      = nfs4_listxattr,
10341 };
10342
10343 const struct nfs_rpc_ops nfs_v4_clientops = {
10344         .version        = 4,                    /* protocol version */
10345         .dentry_ops     = &nfs4_dentry_operations,
10346         .dir_inode_ops  = &nfs4_dir_inode_operations,
10347         .file_inode_ops = &nfs4_file_inode_operations,
10348         .file_ops       = &nfs4_file_operations,
10349         .getroot        = nfs4_proc_get_root,
10350         .submount       = nfs4_submount,
10351         .try_get_tree   = nfs4_try_get_tree,
10352         .getattr        = nfs4_proc_getattr,
10353         .setattr        = nfs4_proc_setattr,
10354         .lookup         = nfs4_proc_lookup,
10355         .lookupp        = nfs4_proc_lookupp,
10356         .access         = nfs4_proc_access,
10357         .readlink       = nfs4_proc_readlink,
10358         .create         = nfs4_proc_create,
10359         .remove         = nfs4_proc_remove,
10360         .unlink_setup   = nfs4_proc_unlink_setup,
10361         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10362         .unlink_done    = nfs4_proc_unlink_done,
10363         .rename_setup   = nfs4_proc_rename_setup,
10364         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10365         .rename_done    = nfs4_proc_rename_done,
10366         .link           = nfs4_proc_link,
10367         .symlink        = nfs4_proc_symlink,
10368         .mkdir          = nfs4_proc_mkdir,
10369         .rmdir          = nfs4_proc_rmdir,
10370         .readdir        = nfs4_proc_readdir,
10371         .mknod          = nfs4_proc_mknod,
10372         .statfs         = nfs4_proc_statfs,
10373         .fsinfo         = nfs4_proc_fsinfo,
10374         .pathconf       = nfs4_proc_pathconf,
10375         .set_capabilities = nfs4_server_capabilities,
10376         .decode_dirent  = nfs4_decode_dirent,
10377         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10378         .read_setup     = nfs4_proc_read_setup,
10379         .read_done      = nfs4_read_done,
10380         .write_setup    = nfs4_proc_write_setup,
10381         .write_done     = nfs4_write_done,
10382         .commit_setup   = nfs4_proc_commit_setup,
10383         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10384         .commit_done    = nfs4_commit_done,
10385         .lock           = nfs4_proc_lock,
10386         .clear_acl_cache = nfs4_zap_acl_attr,
10387         .close_context  = nfs4_close_context,
10388         .open_context   = nfs4_atomic_open,
10389         .have_delegation = nfs4_have_delegation,
10390         .alloc_client   = nfs4_alloc_client,
10391         .init_client    = nfs4_init_client,
10392         .free_client    = nfs4_free_client,
10393         .create_server  = nfs4_create_server,
10394         .clone_server   = nfs_clone_server,
10395 };
10396
10397 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10398         .name   = XATTR_NAME_NFSV4_ACL,
10399         .list   = nfs4_xattr_list_nfs4_acl,
10400         .get    = nfs4_xattr_get_nfs4_acl,
10401         .set    = nfs4_xattr_set_nfs4_acl,
10402 };
10403
10404 #ifdef CONFIG_NFS_V4_2
10405 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10406         .prefix = XATTR_USER_PREFIX,
10407         .get    = nfs4_xattr_get_nfs4_user,
10408         .set    = nfs4_xattr_set_nfs4_user,
10409 };
10410 #endif
10411
10412 const struct xattr_handler *nfs4_xattr_handlers[] = {
10413         &nfs4_xattr_nfs4_acl_handler,
10414 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10415         &nfs4_xattr_nfs4_label_handler,
10416 #endif
10417 #ifdef CONFIG_NFS_V4_2
10418         &nfs4_xattr_nfs4_user_handler,
10419 #endif
10420         NULL
10421 };
10422
10423 /*
10424  * Local variables:
10425  *  c-basic-offset: 8
10426  * End:
10427  */