86ed5c0142c3da48faa536ec4ab77b17cc1e4320
[platform/kernel/linux-starfive.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 #define NFSDBG_FACILITY         NFSDBG_PROC
75
76 #define NFS4_BITMASK_SZ         3
77
78 #define NFS4_POLL_RETRY_MIN     (HZ/10)
79 #define NFS4_POLL_RETRY_MAX     (15*HZ)
80
81 /* file attributes which can be mapped to nfs attributes */
82 #define NFS4_VALID_ATTRS (ATTR_MODE \
83         | ATTR_UID \
84         | ATTR_GID \
85         | ATTR_SIZE \
86         | ATTR_ATIME \
87         | ATTR_MTIME \
88         | ATTR_CTIME \
89         | ATTR_ATIME_SET \
90         | ATTR_MTIME_SET)
91
92 struct nfs4_opendata;
93 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
94 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
95 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
96 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
97                               struct nfs_fattr *fattr, struct inode *inode);
98 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
99                             struct nfs_fattr *fattr, struct iattr *sattr,
100                             struct nfs_open_context *ctx, struct nfs4_label *ilabel);
101 #ifdef CONFIG_NFS_V4_1
102 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
103                 const struct cred *cred,
104                 struct nfs4_slot *slot,
105                 bool is_privileged);
106 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
107                 const struct cred *);
108 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
109                 const struct cred *, bool);
110 #endif
111
112 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
113 static inline struct nfs4_label *
114 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
115         struct iattr *sattr, struct nfs4_label *label)
116 {
117         int err;
118
119         if (label == NULL)
120                 return NULL;
121
122         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
123                 return NULL;
124
125         err = security_dentry_init_security(dentry, sattr->ia_mode,
126                                 &dentry->d_name, NULL,
127                                 (void **)&label->label, &label->len);
128         if (err == 0)
129                 return label;
130
131         return NULL;
132 }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 {
136         if (label)
137                 security_release_secctx(label->label, label->len);
138 }
139 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
140 {
141         if (label)
142                 return server->attr_bitmask;
143
144         return server->attr_bitmask_nl;
145 }
146 #else
147 static inline struct nfs4_label *
148 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
149         struct iattr *sattr, struct nfs4_label *l)
150 { return NULL; }
151 static inline void
152 nfs4_label_release_security(struct nfs4_label *label)
153 { return; }
154 static inline u32 *
155 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
156 { return server->attr_bitmask; }
157 #endif
158
159 /* Prevent leaks of NFSv4 errors into userland */
160 static int nfs4_map_errors(int err)
161 {
162         if (err >= -1000)
163                 return err;
164         switch (err) {
165         case -NFS4ERR_RESOURCE:
166         case -NFS4ERR_LAYOUTTRYLATER:
167         case -NFS4ERR_RECALLCONFLICT:
168                 return -EREMOTEIO;
169         case -NFS4ERR_WRONGSEC:
170         case -NFS4ERR_WRONG_CRED:
171                 return -EPERM;
172         case -NFS4ERR_BADOWNER:
173         case -NFS4ERR_BADNAME:
174                 return -EINVAL;
175         case -NFS4ERR_SHARE_DENIED:
176                 return -EACCES;
177         case -NFS4ERR_MINOR_VERS_MISMATCH:
178                 return -EPROTONOSUPPORT;
179         case -NFS4ERR_FILE_OPEN:
180                 return -EBUSY;
181         case -NFS4ERR_NOT_SAME:
182                 return -ENOTSYNC;
183         default:
184                 dprintk("%s could not handle NFSv4 error %d\n",
185                                 __func__, -err);
186                 break;
187         }
188         return -EIO;
189 }
190
191 /*
192  * This is our standard bitmap for GETATTR requests.
193  */
194 const u32 nfs4_fattr_bitmap[3] = {
195         FATTR4_WORD0_TYPE
196         | FATTR4_WORD0_CHANGE
197         | FATTR4_WORD0_SIZE
198         | FATTR4_WORD0_FSID
199         | FATTR4_WORD0_FILEID,
200         FATTR4_WORD1_MODE
201         | FATTR4_WORD1_NUMLINKS
202         | FATTR4_WORD1_OWNER
203         | FATTR4_WORD1_OWNER_GROUP
204         | FATTR4_WORD1_RAWDEV
205         | FATTR4_WORD1_SPACE_USED
206         | FATTR4_WORD1_TIME_ACCESS
207         | FATTR4_WORD1_TIME_METADATA
208         | FATTR4_WORD1_TIME_MODIFY
209         | FATTR4_WORD1_MOUNTED_ON_FILEID,
210 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
211         FATTR4_WORD2_SECURITY_LABEL
212 #endif
213 };
214
215 static const u32 nfs4_pnfs_open_bitmap[3] = {
216         FATTR4_WORD0_TYPE
217         | FATTR4_WORD0_CHANGE
218         | FATTR4_WORD0_SIZE
219         | FATTR4_WORD0_FSID
220         | FATTR4_WORD0_FILEID,
221         FATTR4_WORD1_MODE
222         | FATTR4_WORD1_NUMLINKS
223         | FATTR4_WORD1_OWNER
224         | FATTR4_WORD1_OWNER_GROUP
225         | FATTR4_WORD1_RAWDEV
226         | FATTR4_WORD1_SPACE_USED
227         | FATTR4_WORD1_TIME_ACCESS
228         | FATTR4_WORD1_TIME_METADATA
229         | FATTR4_WORD1_TIME_MODIFY,
230         FATTR4_WORD2_MDSTHRESHOLD
231 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
232         | FATTR4_WORD2_SECURITY_LABEL
233 #endif
234 };
235
236 static const u32 nfs4_open_noattr_bitmap[3] = {
237         FATTR4_WORD0_TYPE
238         | FATTR4_WORD0_FILEID,
239 };
240
241 const u32 nfs4_statfs_bitmap[3] = {
242         FATTR4_WORD0_FILES_AVAIL
243         | FATTR4_WORD0_FILES_FREE
244         | FATTR4_WORD0_FILES_TOTAL,
245         FATTR4_WORD1_SPACE_AVAIL
246         | FATTR4_WORD1_SPACE_FREE
247         | FATTR4_WORD1_SPACE_TOTAL
248 };
249
250 const u32 nfs4_pathconf_bitmap[3] = {
251         FATTR4_WORD0_MAXLINK
252         | FATTR4_WORD0_MAXNAME,
253         0
254 };
255
256 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
257                         | FATTR4_WORD0_MAXREAD
258                         | FATTR4_WORD0_MAXWRITE
259                         | FATTR4_WORD0_LEASE_TIME,
260                         FATTR4_WORD1_TIME_DELTA
261                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
262                         FATTR4_WORD2_LAYOUT_BLKSIZE
263                         | FATTR4_WORD2_CLONE_BLKSIZE
264                         | FATTR4_WORD2_CHANGE_ATTR_TYPE
265                         | FATTR4_WORD2_XATTR_SUPPORT
266 };
267
268 const u32 nfs4_fs_locations_bitmap[3] = {
269         FATTR4_WORD0_CHANGE
270         | FATTR4_WORD0_SIZE
271         | FATTR4_WORD0_FSID
272         | FATTR4_WORD0_FILEID
273         | FATTR4_WORD0_FS_LOCATIONS,
274         FATTR4_WORD1_OWNER
275         | FATTR4_WORD1_OWNER_GROUP
276         | FATTR4_WORD1_RAWDEV
277         | FATTR4_WORD1_SPACE_USED
278         | FATTR4_WORD1_TIME_ACCESS
279         | FATTR4_WORD1_TIME_METADATA
280         | FATTR4_WORD1_TIME_MODIFY
281         | FATTR4_WORD1_MOUNTED_ON_FILEID,
282 };
283
284 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
285                                     struct inode *inode, unsigned long flags)
286 {
287         unsigned long cache_validity;
288
289         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
290         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
291                 return;
292
293         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
294
295         /* Remove the attributes over which we have full control */
296         dst[1] &= ~FATTR4_WORD1_RAWDEV;
297         if (!(cache_validity & NFS_INO_INVALID_SIZE))
298                 dst[0] &= ~FATTR4_WORD0_SIZE;
299
300         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
301                 dst[0] &= ~FATTR4_WORD0_CHANGE;
302
303         if (!(cache_validity & NFS_INO_INVALID_MODE))
304                 dst[1] &= ~FATTR4_WORD1_MODE;
305         if (!(cache_validity & NFS_INO_INVALID_OTHER))
306                 dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
307 }
308
309 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
310                 struct nfs4_readdir_arg *readdir)
311 {
312         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
313         __be32 *start, *p;
314
315         if (cookie > 2) {
316                 readdir->cookie = cookie;
317                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
318                 return;
319         }
320
321         readdir->cookie = 0;
322         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
323         if (cookie == 2)
324                 return;
325         
326         /*
327          * NFSv4 servers do not return entries for '.' and '..'
328          * Therefore, we fake these entries here.  We let '.'
329          * have cookie 0 and '..' have cookie 1.  Note that
330          * when talking to the server, we always send cookie 0
331          * instead of 1 or 2.
332          */
333         start = p = kmap_atomic(*readdir->pages);
334         
335         if (cookie == 0) {
336                 *p++ = xdr_one;                                  /* next */
337                 *p++ = xdr_zero;                   /* cookie, first word */
338                 *p++ = xdr_one;                   /* cookie, second word */
339                 *p++ = xdr_one;                             /* entry len */
340                 memcpy(p, ".\0\0\0", 4);                        /* entry */
341                 p++;
342                 *p++ = xdr_one;                         /* bitmap length */
343                 *p++ = htonl(attrs);                           /* bitmap */
344                 *p++ = htonl(12);             /* attribute buffer length */
345                 *p++ = htonl(NF4DIR);
346                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
347         }
348         
349         *p++ = xdr_one;                                  /* next */
350         *p++ = xdr_zero;                   /* cookie, first word */
351         *p++ = xdr_two;                   /* cookie, second word */
352         *p++ = xdr_two;                             /* entry len */
353         memcpy(p, "..\0\0", 4);                         /* entry */
354         p++;
355         *p++ = xdr_one;                         /* bitmap length */
356         *p++ = htonl(attrs);                           /* bitmap */
357         *p++ = htonl(12);             /* attribute buffer length */
358         *p++ = htonl(NF4DIR);
359         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
360
361         readdir->pgbase = (char *)p - (char *)start;
362         readdir->count -= readdir->pgbase;
363         kunmap_atomic(start);
364 }
365
366 static void nfs4_fattr_set_prechange(struct nfs_fattr *fattr, u64 version)
367 {
368         if (!(fattr->valid & NFS_ATTR_FATTR_PRECHANGE)) {
369                 fattr->pre_change_attr = version;
370                 fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
371         }
372 }
373
374 static void nfs4_test_and_free_stateid(struct nfs_server *server,
375                 nfs4_stateid *stateid,
376                 const struct cred *cred)
377 {
378         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
379
380         ops->test_and_free_expired(server, stateid, cred);
381 }
382
383 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
384                 nfs4_stateid *stateid,
385                 const struct cred *cred)
386 {
387         stateid->type = NFS4_REVOKED_STATEID_TYPE;
388         nfs4_test_and_free_stateid(server, stateid, cred);
389 }
390
391 static void nfs4_free_revoked_stateid(struct nfs_server *server,
392                 const nfs4_stateid *stateid,
393                 const struct cred *cred)
394 {
395         nfs4_stateid tmp;
396
397         nfs4_stateid_copy(&tmp, stateid);
398         __nfs4_free_revoked_stateid(server, &tmp, cred);
399 }
400
401 static long nfs4_update_delay(long *timeout)
402 {
403         long ret;
404         if (!timeout)
405                 return NFS4_POLL_RETRY_MAX;
406         if (*timeout <= 0)
407                 *timeout = NFS4_POLL_RETRY_MIN;
408         if (*timeout > NFS4_POLL_RETRY_MAX)
409                 *timeout = NFS4_POLL_RETRY_MAX;
410         ret = *timeout;
411         *timeout <<= 1;
412         return ret;
413 }
414
415 static int nfs4_delay_killable(long *timeout)
416 {
417         might_sleep();
418
419         __set_current_state(TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
420         schedule_timeout(nfs4_update_delay(timeout));
421         if (!__fatal_signal_pending(current))
422                 return 0;
423         return -EINTR;
424 }
425
426 static int nfs4_delay_interruptible(long *timeout)
427 {
428         might_sleep();
429
430         __set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE_UNSAFE);
431         schedule_timeout(nfs4_update_delay(timeout));
432         if (!signal_pending(current))
433                 return 0;
434         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
435 }
436
437 static int nfs4_delay(long *timeout, bool interruptible)
438 {
439         if (interruptible)
440                 return nfs4_delay_interruptible(timeout);
441         return nfs4_delay_killable(timeout);
442 }
443
444 static const nfs4_stateid *
445 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
446 {
447         if (!stateid)
448                 return NULL;
449         switch (stateid->type) {
450         case NFS4_OPEN_STATEID_TYPE:
451         case NFS4_LOCK_STATEID_TYPE:
452         case NFS4_DELEGATION_STATEID_TYPE:
453                 return stateid;
454         default:
455                 break;
456         }
457         return NULL;
458 }
459
460 /* This is the error handling routine for processes that are allowed
461  * to sleep.
462  */
463 static int nfs4_do_handle_exception(struct nfs_server *server,
464                 int errorcode, struct nfs4_exception *exception)
465 {
466         struct nfs_client *clp = server->nfs_client;
467         struct nfs4_state *state = exception->state;
468         const nfs4_stateid *stateid;
469         struct inode *inode = exception->inode;
470         int ret = errorcode;
471
472         exception->delay = 0;
473         exception->recovering = 0;
474         exception->retry = 0;
475
476         stateid = nfs4_recoverable_stateid(exception->stateid);
477         if (stateid == NULL && state != NULL)
478                 stateid = nfs4_recoverable_stateid(&state->stateid);
479
480         switch(errorcode) {
481                 case 0:
482                         return 0;
483                 case -NFS4ERR_BADHANDLE:
484                 case -ESTALE:
485                         if (inode != NULL && S_ISREG(inode->i_mode))
486                                 pnfs_destroy_layout(NFS_I(inode));
487                         break;
488                 case -NFS4ERR_DELEG_REVOKED:
489                 case -NFS4ERR_ADMIN_REVOKED:
490                 case -NFS4ERR_EXPIRED:
491                 case -NFS4ERR_BAD_STATEID:
492                 case -NFS4ERR_PARTNER_NO_AUTH:
493                         if (inode != NULL && stateid != NULL) {
494                                 nfs_inode_find_state_and_recover(inode,
495                                                 stateid);
496                                 goto wait_on_recovery;
497                         }
498                         fallthrough;
499                 case -NFS4ERR_OPENMODE:
500                         if (inode) {
501                                 int err;
502
503                                 err = nfs_async_inode_return_delegation(inode,
504                                                 stateid);
505                                 if (err == 0)
506                                         goto wait_on_recovery;
507                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
508                                         exception->retry = 1;
509                                         break;
510                                 }
511                         }
512                         if (state == NULL)
513                                 break;
514                         ret = nfs4_schedule_stateid_recovery(server, state);
515                         if (ret < 0)
516                                 break;
517                         goto wait_on_recovery;
518                 case -NFS4ERR_STALE_STATEID:
519                 case -NFS4ERR_STALE_CLIENTID:
520                         nfs4_schedule_lease_recovery(clp);
521                         goto wait_on_recovery;
522                 case -NFS4ERR_MOVED:
523                         ret = nfs4_schedule_migration_recovery(server);
524                         if (ret < 0)
525                                 break;
526                         goto wait_on_recovery;
527                 case -NFS4ERR_LEASE_MOVED:
528                         nfs4_schedule_lease_moved_recovery(clp);
529                         goto wait_on_recovery;
530 #if defined(CONFIG_NFS_V4_1)
531                 case -NFS4ERR_BADSESSION:
532                 case -NFS4ERR_BADSLOT:
533                 case -NFS4ERR_BAD_HIGH_SLOT:
534                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
535                 case -NFS4ERR_DEADSESSION:
536                 case -NFS4ERR_SEQ_FALSE_RETRY:
537                 case -NFS4ERR_SEQ_MISORDERED:
538                         /* Handled in nfs41_sequence_process() */
539                         goto wait_on_recovery;
540 #endif /* defined(CONFIG_NFS_V4_1) */
541                 case -NFS4ERR_FILE_OPEN:
542                         if (exception->timeout > HZ) {
543                                 /* We have retried a decent amount, time to
544                                  * fail
545                                  */
546                                 ret = -EBUSY;
547                                 break;
548                         }
549                         fallthrough;
550                 case -NFS4ERR_DELAY:
551                         nfs_inc_server_stats(server, NFSIOS_DELAY);
552                         fallthrough;
553                 case -NFS4ERR_GRACE:
554                 case -NFS4ERR_LAYOUTTRYLATER:
555                 case -NFS4ERR_RECALLCONFLICT:
556                         exception->delay = 1;
557                         return 0;
558
559                 case -NFS4ERR_RETRY_UNCACHED_REP:
560                 case -NFS4ERR_OLD_STATEID:
561                         exception->retry = 1;
562                         break;
563                 case -NFS4ERR_BADOWNER:
564                         /* The following works around a Linux server bug! */
565                 case -NFS4ERR_BADNAME:
566                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
567                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
568                                 exception->retry = 1;
569                                 printk(KERN_WARNING "NFS: v4 server %s "
570                                                 "does not accept raw "
571                                                 "uid/gids. "
572                                                 "Reenabling the idmapper.\n",
573                                                 server->nfs_client->cl_hostname);
574                         }
575         }
576         /* We failed to handle the error */
577         return nfs4_map_errors(ret);
578 wait_on_recovery:
579         exception->recovering = 1;
580         return 0;
581 }
582
583 /* This is the error handling routine for processes that are allowed
584  * to sleep.
585  */
586 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
587 {
588         struct nfs_client *clp = server->nfs_client;
589         int ret;
590
591         ret = nfs4_do_handle_exception(server, errorcode, exception);
592         if (exception->delay) {
593                 ret = nfs4_delay(&exception->timeout,
594                                 exception->interruptible);
595                 goto out_retry;
596         }
597         if (exception->recovering) {
598                 if (exception->task_is_privileged)
599                         return -EDEADLOCK;
600                 ret = nfs4_wait_clnt_recover(clp);
601                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
602                         return -EIO;
603                 goto out_retry;
604         }
605         return ret;
606 out_retry:
607         if (ret == 0)
608                 exception->retry = 1;
609         return ret;
610 }
611
612 static int
613 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
614                 int errorcode, struct nfs4_exception *exception)
615 {
616         struct nfs_client *clp = server->nfs_client;
617         int ret;
618
619         ret = nfs4_do_handle_exception(server, errorcode, exception);
620         if (exception->delay) {
621                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
622                 goto out_retry;
623         }
624         if (exception->recovering) {
625                 if (exception->task_is_privileged)
626                         return -EDEADLOCK;
627                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
628                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
629                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
630                 goto out_retry;
631         }
632         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
633                 ret = -EIO;
634         return ret;
635 out_retry:
636         if (ret == 0) {
637                 exception->retry = 1;
638                 /*
639                  * For NFS4ERR_MOVED, the client transport will need to
640                  * be recomputed after migration recovery has completed.
641                  */
642                 if (errorcode == -NFS4ERR_MOVED)
643                         rpc_task_release_transport(task);
644         }
645         return ret;
646 }
647
648 int
649 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
650                         struct nfs4_state *state, long *timeout)
651 {
652         struct nfs4_exception exception = {
653                 .state = state,
654         };
655
656         if (task->tk_status >= 0)
657                 return 0;
658         if (timeout)
659                 exception.timeout = *timeout;
660         task->tk_status = nfs4_async_handle_exception(task, server,
661                         task->tk_status,
662                         &exception);
663         if (exception.delay && timeout)
664                 *timeout = exception.timeout;
665         if (exception.retry)
666                 return -EAGAIN;
667         return 0;
668 }
669
670 /*
671  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
672  * or 'false' otherwise.
673  */
674 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
675 {
676         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
677         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
678 }
679
680 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
681 {
682         spin_lock(&clp->cl_lock);
683         if (time_before(clp->cl_last_renewal,timestamp))
684                 clp->cl_last_renewal = timestamp;
685         spin_unlock(&clp->cl_lock);
686 }
687
688 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
689 {
690         struct nfs_client *clp = server->nfs_client;
691
692         if (!nfs4_has_session(clp))
693                 do_renew_lease(clp, timestamp);
694 }
695
696 struct nfs4_call_sync_data {
697         const struct nfs_server *seq_server;
698         struct nfs4_sequence_args *seq_args;
699         struct nfs4_sequence_res *seq_res;
700 };
701
702 void nfs4_init_sequence(struct nfs4_sequence_args *args,
703                         struct nfs4_sequence_res *res, int cache_reply,
704                         int privileged)
705 {
706         args->sa_slot = NULL;
707         args->sa_cache_this = cache_reply;
708         args->sa_privileged = privileged;
709
710         res->sr_slot = NULL;
711 }
712
713 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
714 {
715         struct nfs4_slot *slot = res->sr_slot;
716         struct nfs4_slot_table *tbl;
717
718         tbl = slot->table;
719         spin_lock(&tbl->slot_tbl_lock);
720         if (!nfs41_wake_and_assign_slot(tbl, slot))
721                 nfs4_free_slot(tbl, slot);
722         spin_unlock(&tbl->slot_tbl_lock);
723
724         res->sr_slot = NULL;
725 }
726
727 static int nfs40_sequence_done(struct rpc_task *task,
728                                struct nfs4_sequence_res *res)
729 {
730         if (res->sr_slot != NULL)
731                 nfs40_sequence_free_slot(res);
732         return 1;
733 }
734
735 #if defined(CONFIG_NFS_V4_1)
736
737 static void nfs41_release_slot(struct nfs4_slot *slot)
738 {
739         struct nfs4_session *session;
740         struct nfs4_slot_table *tbl;
741         bool send_new_highest_used_slotid = false;
742
743         if (!slot)
744                 return;
745         tbl = slot->table;
746         session = tbl->session;
747
748         /* Bump the slot sequence number */
749         if (slot->seq_done)
750                 slot->seq_nr++;
751         slot->seq_done = 0;
752
753         spin_lock(&tbl->slot_tbl_lock);
754         /* Be nice to the server: try to ensure that the last transmitted
755          * value for highest_user_slotid <= target_highest_slotid
756          */
757         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
758                 send_new_highest_used_slotid = true;
759
760         if (nfs41_wake_and_assign_slot(tbl, slot)) {
761                 send_new_highest_used_slotid = false;
762                 goto out_unlock;
763         }
764         nfs4_free_slot(tbl, slot);
765
766         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
767                 send_new_highest_used_slotid = false;
768 out_unlock:
769         spin_unlock(&tbl->slot_tbl_lock);
770         if (send_new_highest_used_slotid)
771                 nfs41_notify_server(session->clp);
772         if (waitqueue_active(&tbl->slot_waitq))
773                 wake_up_all(&tbl->slot_waitq);
774 }
775
776 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
777 {
778         nfs41_release_slot(res->sr_slot);
779         res->sr_slot = NULL;
780 }
781
782 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
783                 u32 seqnr)
784 {
785         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
786                 slot->seq_nr_highest_sent = seqnr;
787 }
788 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot, u32 seqnr)
789 {
790         nfs4_slot_sequence_record_sent(slot, seqnr);
791         slot->seq_nr_last_acked = seqnr;
792 }
793
794 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
795                                 struct nfs4_slot *slot)
796 {
797         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
798         if (!IS_ERR(task))
799                 rpc_put_task_async(task);
800 }
801
802 static int nfs41_sequence_process(struct rpc_task *task,
803                 struct nfs4_sequence_res *res)
804 {
805         struct nfs4_session *session;
806         struct nfs4_slot *slot = res->sr_slot;
807         struct nfs_client *clp;
808         int status;
809         int ret = 1;
810
811         if (slot == NULL)
812                 goto out_noaction;
813         /* don't increment the sequence number if the task wasn't sent */
814         if (!RPC_WAS_SENT(task) || slot->seq_done)
815                 goto out;
816
817         session = slot->table->session;
818         clp = session->clp;
819
820         trace_nfs4_sequence_done(session, res);
821
822         status = res->sr_status;
823         if (task->tk_status == -NFS4ERR_DEADSESSION)
824                 status = -NFS4ERR_DEADSESSION;
825
826         /* Check the SEQUENCE operation status */
827         switch (status) {
828         case 0:
829                 /* Mark this sequence number as having been acked */
830                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
831                 /* Update the slot's sequence and clientid lease timer */
832                 slot->seq_done = 1;
833                 do_renew_lease(clp, res->sr_timestamp);
834                 /* Check sequence flags */
835                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
836                                 !!slot->privileged);
837                 nfs41_update_target_slotid(slot->table, slot, res);
838                 break;
839         case 1:
840                 /*
841                  * sr_status remains 1 if an RPC level error occurred.
842                  * The server may or may not have processed the sequence
843                  * operation..
844                  */
845                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
846                 slot->seq_done = 1;
847                 goto out;
848         case -NFS4ERR_DELAY:
849                 /* The server detected a resend of the RPC call and
850                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
851                  * of RFC5661.
852                  */
853                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
854                         __func__,
855                         slot->slot_nr,
856                         slot->seq_nr);
857                 goto out_retry;
858         case -NFS4ERR_RETRY_UNCACHED_REP:
859         case -NFS4ERR_SEQ_FALSE_RETRY:
860                 /*
861                  * The server thinks we tried to replay a request.
862                  * Retry the call after bumping the sequence ID.
863                  */
864                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
865                 goto retry_new_seq;
866         case -NFS4ERR_BADSLOT:
867                 /*
868                  * The slot id we used was probably retired. Try again
869                  * using a different slot id.
870                  */
871                 if (slot->slot_nr < slot->table->target_highest_slotid)
872                         goto session_recover;
873                 goto retry_nowait;
874         case -NFS4ERR_SEQ_MISORDERED:
875                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
876                 /*
877                  * Were one or more calls using this slot interrupted?
878                  * If the server never received the request, then our
879                  * transmitted slot sequence number may be too high. However,
880                  * if the server did receive the request then it might
881                  * accidentally give us a reply with a mismatched operation.
882                  * We can sort this out by sending a lone sequence operation
883                  * to the server on the same slot.
884                  */
885                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
886                         slot->seq_nr--;
887                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
888                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
889                                 res->sr_slot = NULL;
890                         }
891                         goto retry_nowait;
892                 }
893                 /*
894                  * RFC5661:
895                  * A retry might be sent while the original request is
896                  * still in progress on the replier. The replier SHOULD
897                  * deal with the issue by returning NFS4ERR_DELAY as the
898                  * reply to SEQUENCE or CB_SEQUENCE operation, but
899                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
900                  *
901                  * Restart the search after a delay.
902                  */
903                 slot->seq_nr = slot->seq_nr_highest_sent;
904                 goto out_retry;
905         case -NFS4ERR_BADSESSION:
906         case -NFS4ERR_DEADSESSION:
907         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
908                 goto session_recover;
909         default:
910                 /* Just update the slot sequence no. */
911                 slot->seq_done = 1;
912         }
913 out:
914         /* The session may be reset by one of the error handlers. */
915         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
916 out_noaction:
917         return ret;
918 session_recover:
919         nfs4_schedule_session_recovery(session, status);
920         dprintk("%s ERROR: %d Reset session\n", __func__, status);
921         nfs41_sequence_free_slot(res);
922         goto out;
923 retry_new_seq:
924         ++slot->seq_nr;
925 retry_nowait:
926         if (rpc_restart_call_prepare(task)) {
927                 nfs41_sequence_free_slot(res);
928                 task->tk_status = 0;
929                 ret = 0;
930         }
931         goto out;
932 out_retry:
933         if (!rpc_restart_call(task))
934                 goto out;
935         rpc_delay(task, NFS4_POLL_RETRY_MAX);
936         return 0;
937 }
938
939 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
940 {
941         if (!nfs41_sequence_process(task, res))
942                 return 0;
943         if (res->sr_slot != NULL)
944                 nfs41_sequence_free_slot(res);
945         return 1;
946
947 }
948 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
949
950 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
951 {
952         if (res->sr_slot == NULL)
953                 return 1;
954         if (res->sr_slot->table->session != NULL)
955                 return nfs41_sequence_process(task, res);
956         return nfs40_sequence_done(task, res);
957 }
958
959 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
960 {
961         if (res->sr_slot != NULL) {
962                 if (res->sr_slot->table->session != NULL)
963                         nfs41_sequence_free_slot(res);
964                 else
965                         nfs40_sequence_free_slot(res);
966         }
967 }
968
969 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
970 {
971         if (res->sr_slot == NULL)
972                 return 1;
973         if (!res->sr_slot->table->session)
974                 return nfs40_sequence_done(task, res);
975         return nfs41_sequence_done(task, res);
976 }
977 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
978
979 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
980 {
981         struct nfs4_call_sync_data *data = calldata;
982
983         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
984
985         nfs4_setup_sequence(data->seq_server->nfs_client,
986                             data->seq_args, data->seq_res, task);
987 }
988
989 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
990 {
991         struct nfs4_call_sync_data *data = calldata;
992
993         nfs41_sequence_done(task, data->seq_res);
994 }
995
996 static const struct rpc_call_ops nfs41_call_sync_ops = {
997         .rpc_call_prepare = nfs41_call_sync_prepare,
998         .rpc_call_done = nfs41_call_sync_done,
999 };
1000
1001 #else   /* !CONFIG_NFS_V4_1 */
1002
1003 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1004 {
1005         return nfs40_sequence_done(task, res);
1006 }
1007
1008 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1009 {
1010         if (res->sr_slot != NULL)
1011                 nfs40_sequence_free_slot(res);
1012 }
1013
1014 int nfs4_sequence_done(struct rpc_task *task,
1015                        struct nfs4_sequence_res *res)
1016 {
1017         return nfs40_sequence_done(task, res);
1018 }
1019 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1020
1021 #endif  /* !CONFIG_NFS_V4_1 */
1022
1023 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1024 {
1025         res->sr_timestamp = jiffies;
1026         res->sr_status_flags = 0;
1027         res->sr_status = 1;
1028 }
1029
1030 static
1031 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1032                 struct nfs4_sequence_res *res,
1033                 struct nfs4_slot *slot)
1034 {
1035         if (!slot)
1036                 return;
1037         slot->privileged = args->sa_privileged ? 1 : 0;
1038         args->sa_slot = slot;
1039
1040         res->sr_slot = slot;
1041 }
1042
1043 int nfs4_setup_sequence(struct nfs_client *client,
1044                         struct nfs4_sequence_args *args,
1045                         struct nfs4_sequence_res *res,
1046                         struct rpc_task *task)
1047 {
1048         struct nfs4_session *session = nfs4_get_session(client);
1049         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1050         struct nfs4_slot *slot;
1051
1052         /* slot already allocated? */
1053         if (res->sr_slot != NULL)
1054                 goto out_start;
1055
1056         if (session)
1057                 tbl = &session->fc_slot_table;
1058
1059         spin_lock(&tbl->slot_tbl_lock);
1060         /* The state manager will wait until the slot table is empty */
1061         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1062                 goto out_sleep;
1063
1064         slot = nfs4_alloc_slot(tbl);
1065         if (IS_ERR(slot)) {
1066                 if (slot == ERR_PTR(-ENOMEM))
1067                         goto out_sleep_timeout;
1068                 goto out_sleep;
1069         }
1070         spin_unlock(&tbl->slot_tbl_lock);
1071
1072         nfs4_sequence_attach_slot(args, res, slot);
1073
1074         trace_nfs4_setup_sequence(session, args);
1075 out_start:
1076         nfs41_sequence_res_init(res);
1077         rpc_call_start(task);
1078         return 0;
1079 out_sleep_timeout:
1080         /* Try again in 1/4 second */
1081         if (args->sa_privileged)
1082                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1083                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1084         else
1085                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1086                                 NULL, jiffies + (HZ >> 2));
1087         spin_unlock(&tbl->slot_tbl_lock);
1088         return -EAGAIN;
1089 out_sleep:
1090         if (args->sa_privileged)
1091                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1092                                 RPC_PRIORITY_PRIVILEGED);
1093         else
1094                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1095         spin_unlock(&tbl->slot_tbl_lock);
1096         return -EAGAIN;
1097 }
1098 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1099
1100 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1101 {
1102         struct nfs4_call_sync_data *data = calldata;
1103         nfs4_setup_sequence(data->seq_server->nfs_client,
1104                                 data->seq_args, data->seq_res, task);
1105 }
1106
1107 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1108 {
1109         struct nfs4_call_sync_data *data = calldata;
1110         nfs4_sequence_done(task, data->seq_res);
1111 }
1112
1113 static const struct rpc_call_ops nfs40_call_sync_ops = {
1114         .rpc_call_prepare = nfs40_call_sync_prepare,
1115         .rpc_call_done = nfs40_call_sync_done,
1116 };
1117
1118 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1119 {
1120         int ret;
1121         struct rpc_task *task;
1122
1123         task = rpc_run_task(task_setup);
1124         if (IS_ERR(task))
1125                 return PTR_ERR(task);
1126
1127         ret = task->tk_status;
1128         rpc_put_task(task);
1129         return ret;
1130 }
1131
1132 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1133                              struct nfs_server *server,
1134                              struct rpc_message *msg,
1135                              struct nfs4_sequence_args *args,
1136                              struct nfs4_sequence_res *res,
1137                              unsigned short task_flags)
1138 {
1139         struct nfs_client *clp = server->nfs_client;
1140         struct nfs4_call_sync_data data = {
1141                 .seq_server = server,
1142                 .seq_args = args,
1143                 .seq_res = res,
1144         };
1145         struct rpc_task_setup task_setup = {
1146                 .rpc_client = clnt,
1147                 .rpc_message = msg,
1148                 .callback_ops = clp->cl_mvops->call_sync_ops,
1149                 .callback_data = &data,
1150                 .flags = task_flags,
1151         };
1152
1153         return nfs4_call_sync_custom(&task_setup);
1154 }
1155
1156 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1157                                    struct nfs_server *server,
1158                                    struct rpc_message *msg,
1159                                    struct nfs4_sequence_args *args,
1160                                    struct nfs4_sequence_res *res)
1161 {
1162         unsigned short task_flags = 0;
1163
1164         if (server->caps & NFS_CAP_MOVEABLE)
1165                 task_flags = RPC_TASK_MOVEABLE;
1166         return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1167 }
1168
1169
1170 int nfs4_call_sync(struct rpc_clnt *clnt,
1171                    struct nfs_server *server,
1172                    struct rpc_message *msg,
1173                    struct nfs4_sequence_args *args,
1174                    struct nfs4_sequence_res *res,
1175                    int cache_reply)
1176 {
1177         nfs4_init_sequence(args, res, cache_reply, 0);
1178         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1179 }
1180
1181 static void
1182 nfs4_inc_nlink_locked(struct inode *inode)
1183 {
1184         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1185                                              NFS_INO_INVALID_CTIME |
1186                                              NFS_INO_INVALID_NLINK);
1187         inc_nlink(inode);
1188 }
1189
1190 static void
1191 nfs4_inc_nlink(struct inode *inode)
1192 {
1193         spin_lock(&inode->i_lock);
1194         nfs4_inc_nlink_locked(inode);
1195         spin_unlock(&inode->i_lock);
1196 }
1197
1198 static void
1199 nfs4_dec_nlink_locked(struct inode *inode)
1200 {
1201         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1202                                              NFS_INO_INVALID_CTIME |
1203                                              NFS_INO_INVALID_NLINK);
1204         drop_nlink(inode);
1205 }
1206
1207 static void
1208 nfs4_update_changeattr_locked(struct inode *inode,
1209                 struct nfs4_change_info *cinfo,
1210                 unsigned long timestamp, unsigned long cache_validity)
1211 {
1212         struct nfs_inode *nfsi = NFS_I(inode);
1213         u64 change_attr = inode_peek_iversion_raw(inode);
1214
1215         cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1216         if (S_ISDIR(inode->i_mode))
1217                 cache_validity |= NFS_INO_INVALID_DATA;
1218
1219         switch (NFS_SERVER(inode)->change_attr_type) {
1220         case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1221                 if (cinfo->after == change_attr)
1222                         goto out;
1223                 break;
1224         default:
1225                 if ((s64)(change_attr - cinfo->after) >= 0)
1226                         goto out;
1227         }
1228
1229         inode_set_iversion_raw(inode, cinfo->after);
1230         if (!cinfo->atomic || cinfo->before != change_attr) {
1231                 if (S_ISDIR(inode->i_mode))
1232                         nfs_force_lookup_revalidate(inode);
1233
1234                 if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1235                         cache_validity |=
1236                                 NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1237                                 NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1238                                 NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1239                                 NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR;
1240                 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1241         }
1242         nfsi->attrtimeo_timestamp = jiffies;
1243         nfsi->read_cache_jiffies = timestamp;
1244         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1245         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1246 out:
1247         nfs_set_cache_invalid(inode, cache_validity);
1248 }
1249
1250 void
1251 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1252                 unsigned long timestamp, unsigned long cache_validity)
1253 {
1254         spin_lock(&dir->i_lock);
1255         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1256         spin_unlock(&dir->i_lock);
1257 }
1258
1259 struct nfs4_open_createattrs {
1260         struct nfs4_label *label;
1261         struct iattr *sattr;
1262         const __u32 verf[2];
1263 };
1264
1265 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1266                 int err, struct nfs4_exception *exception)
1267 {
1268         if (err != -EINVAL)
1269                 return false;
1270         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1271                 return false;
1272         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1273         exception->retry = 1;
1274         return true;
1275 }
1276
1277 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1278 {
1279          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1280 }
1281
1282 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1283 {
1284         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1285
1286         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1287 }
1288
1289 static u32
1290 nfs4_map_atomic_open_share(struct nfs_server *server,
1291                 fmode_t fmode, int openflags)
1292 {
1293         u32 res = 0;
1294
1295         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1296         case FMODE_READ:
1297                 res = NFS4_SHARE_ACCESS_READ;
1298                 break;
1299         case FMODE_WRITE:
1300                 res = NFS4_SHARE_ACCESS_WRITE;
1301                 break;
1302         case FMODE_READ|FMODE_WRITE:
1303                 res = NFS4_SHARE_ACCESS_BOTH;
1304         }
1305         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1306                 goto out;
1307         /* Want no delegation if we're using O_DIRECT */
1308         if (openflags & O_DIRECT)
1309                 res |= NFS4_SHARE_WANT_NO_DELEG;
1310 out:
1311         return res;
1312 }
1313
1314 static enum open_claim_type4
1315 nfs4_map_atomic_open_claim(struct nfs_server *server,
1316                 enum open_claim_type4 claim)
1317 {
1318         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1319                 return claim;
1320         switch (claim) {
1321         default:
1322                 return claim;
1323         case NFS4_OPEN_CLAIM_FH:
1324                 return NFS4_OPEN_CLAIM_NULL;
1325         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1326                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1327         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1328                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1329         }
1330 }
1331
1332 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1333 {
1334         p->o_res.f_attr = &p->f_attr;
1335         p->o_res.seqid = p->o_arg.seqid;
1336         p->c_res.seqid = p->c_arg.seqid;
1337         p->o_res.server = p->o_arg.server;
1338         p->o_res.access_request = p->o_arg.access;
1339         nfs_fattr_init(&p->f_attr);
1340         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1341 }
1342
1343 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1344                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1345                 const struct nfs4_open_createattrs *c,
1346                 enum open_claim_type4 claim,
1347                 gfp_t gfp_mask)
1348 {
1349         struct dentry *parent = dget_parent(dentry);
1350         struct inode *dir = d_inode(parent);
1351         struct nfs_server *server = NFS_SERVER(dir);
1352         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1353         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1354         struct nfs4_opendata *p;
1355
1356         p = kzalloc(sizeof(*p), gfp_mask);
1357         if (p == NULL)
1358                 goto err;
1359
1360         p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1361         if (IS_ERR(p->f_attr.label))
1362                 goto err_free_p;
1363
1364         p->a_label = nfs4_label_alloc(server, gfp_mask);
1365         if (IS_ERR(p->a_label))
1366                 goto err_free_f;
1367
1368         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1369         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1370         if (IS_ERR(p->o_arg.seqid))
1371                 goto err_free_label;
1372         nfs_sb_active(dentry->d_sb);
1373         p->dentry = dget(dentry);
1374         p->dir = parent;
1375         p->owner = sp;
1376         atomic_inc(&sp->so_count);
1377         p->o_arg.open_flags = flags;
1378         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1379         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1380         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1381                         fmode, flags);
1382         if (flags & O_CREAT) {
1383                 p->o_arg.umask = current_umask();
1384                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1385                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1386                         p->o_arg.u.attrs = &p->attrs;
1387                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1388
1389                         memcpy(p->o_arg.u.verifier.data, c->verf,
1390                                         sizeof(p->o_arg.u.verifier.data));
1391                 }
1392         }
1393         /* ask server to check for all possible rights as results
1394          * are cached */
1395         switch (p->o_arg.claim) {
1396         default:
1397                 break;
1398         case NFS4_OPEN_CLAIM_NULL:
1399         case NFS4_OPEN_CLAIM_FH:
1400                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1401                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1402                                   NFS4_ACCESS_EXECUTE |
1403                                   nfs_access_xattr_mask(server);
1404         }
1405         p->o_arg.clientid = server->nfs_client->cl_clientid;
1406         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1407         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1408         p->o_arg.name = &dentry->d_name;
1409         p->o_arg.server = server;
1410         p->o_arg.bitmask = nfs4_bitmask(server, label);
1411         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1412         switch (p->o_arg.claim) {
1413         case NFS4_OPEN_CLAIM_NULL:
1414         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1415         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1416                 p->o_arg.fh = NFS_FH(dir);
1417                 break;
1418         case NFS4_OPEN_CLAIM_PREVIOUS:
1419         case NFS4_OPEN_CLAIM_FH:
1420         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1421         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1422                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1423         }
1424         p->c_arg.fh = &p->o_res.fh;
1425         p->c_arg.stateid = &p->o_res.stateid;
1426         p->c_arg.seqid = p->o_arg.seqid;
1427         nfs4_init_opendata_res(p);
1428         kref_init(&p->kref);
1429         return p;
1430
1431 err_free_label:
1432         nfs4_label_free(p->a_label);
1433 err_free_f:
1434         nfs4_label_free(p->f_attr.label);
1435 err_free_p:
1436         kfree(p);
1437 err:
1438         dput(parent);
1439         return NULL;
1440 }
1441
1442 static void nfs4_opendata_free(struct kref *kref)
1443 {
1444         struct nfs4_opendata *p = container_of(kref,
1445                         struct nfs4_opendata, kref);
1446         struct super_block *sb = p->dentry->d_sb;
1447
1448         nfs4_lgopen_release(p->lgp);
1449         nfs_free_seqid(p->o_arg.seqid);
1450         nfs4_sequence_free_slot(&p->o_res.seq_res);
1451         if (p->state != NULL)
1452                 nfs4_put_open_state(p->state);
1453         nfs4_put_state_owner(p->owner);
1454
1455         nfs4_label_free(p->a_label);
1456         nfs4_label_free(p->f_attr.label);
1457
1458         dput(p->dir);
1459         dput(p->dentry);
1460         nfs_sb_deactive(sb);
1461         nfs_fattr_free_names(&p->f_attr);
1462         kfree(p->f_attr.mdsthreshold);
1463         kfree(p);
1464 }
1465
1466 static void nfs4_opendata_put(struct nfs4_opendata *p)
1467 {
1468         if (p != NULL)
1469                 kref_put(&p->kref, nfs4_opendata_free);
1470 }
1471
1472 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1473                 fmode_t fmode)
1474 {
1475         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1476         case FMODE_READ|FMODE_WRITE:
1477                 return state->n_rdwr != 0;
1478         case FMODE_WRITE:
1479                 return state->n_wronly != 0;
1480         case FMODE_READ:
1481                 return state->n_rdonly != 0;
1482         }
1483         WARN_ON_ONCE(1);
1484         return false;
1485 }
1486
1487 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1488                 int open_mode, enum open_claim_type4 claim)
1489 {
1490         int ret = 0;
1491
1492         if (open_mode & (O_EXCL|O_TRUNC))
1493                 goto out;
1494         switch (claim) {
1495         case NFS4_OPEN_CLAIM_NULL:
1496         case NFS4_OPEN_CLAIM_FH:
1497                 goto out;
1498         default:
1499                 break;
1500         }
1501         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1502                 case FMODE_READ:
1503                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1504                                 && state->n_rdonly != 0;
1505                         break;
1506                 case FMODE_WRITE:
1507                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1508                                 && state->n_wronly != 0;
1509                         break;
1510                 case FMODE_READ|FMODE_WRITE:
1511                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1512                                 && state->n_rdwr != 0;
1513         }
1514 out:
1515         return ret;
1516 }
1517
1518 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1519                 enum open_claim_type4 claim)
1520 {
1521         if (delegation == NULL)
1522                 return 0;
1523         if ((delegation->type & fmode) != fmode)
1524                 return 0;
1525         switch (claim) {
1526         case NFS4_OPEN_CLAIM_NULL:
1527         case NFS4_OPEN_CLAIM_FH:
1528                 break;
1529         case NFS4_OPEN_CLAIM_PREVIOUS:
1530                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1531                         break;
1532                 fallthrough;
1533         default:
1534                 return 0;
1535         }
1536         nfs_mark_delegation_referenced(delegation);
1537         return 1;
1538 }
1539
1540 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1541 {
1542         switch (fmode) {
1543                 case FMODE_WRITE:
1544                         state->n_wronly++;
1545                         break;
1546                 case FMODE_READ:
1547                         state->n_rdonly++;
1548                         break;
1549                 case FMODE_READ|FMODE_WRITE:
1550                         state->n_rdwr++;
1551         }
1552         nfs4_state_set_mode_locked(state, state->state | fmode);
1553 }
1554
1555 #ifdef CONFIG_NFS_V4_1
1556 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1557 {
1558         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1559                 return true;
1560         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1561                 return true;
1562         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1563                 return true;
1564         return false;
1565 }
1566 #endif /* CONFIG_NFS_V4_1 */
1567
1568 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1569 {
1570         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1571                 wake_up_all(&state->waitq);
1572 }
1573
1574 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1575 {
1576         struct nfs_client *clp = state->owner->so_server->nfs_client;
1577         bool need_recover = false;
1578
1579         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1580                 need_recover = true;
1581         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1582                 need_recover = true;
1583         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1584                 need_recover = true;
1585         if (need_recover)
1586                 nfs4_state_mark_reclaim_nograce(clp, state);
1587 }
1588
1589 /*
1590  * Check for whether or not the caller may update the open stateid
1591  * to the value passed in by stateid.
1592  *
1593  * Note: This function relies heavily on the server implementing
1594  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1595  * correctly.
1596  * i.e. The stateid seqids have to be initialised to 1, and
1597  * are then incremented on every state transition.
1598  */
1599 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1600                 const nfs4_stateid *stateid)
1601 {
1602         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1603                 /* The common case - we're updating to a new sequence number */
1604                 if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1605                         if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1606                                 return true;
1607                         return false;
1608                 }
1609                 /* The server returned a new stateid */
1610         }
1611         /* This is the first OPEN in this generation */
1612         if (stateid->seqid == cpu_to_be32(1))
1613                 return true;
1614         return false;
1615 }
1616
1617 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1618 {
1619         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1620                 return;
1621         if (state->n_wronly)
1622                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1623         if (state->n_rdonly)
1624                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1625         if (state->n_rdwr)
1626                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1627         set_bit(NFS_OPEN_STATE, &state->flags);
1628 }
1629
1630 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1631                 nfs4_stateid *stateid, fmode_t fmode)
1632 {
1633         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1634         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1635         case FMODE_WRITE:
1636                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1637                 break;
1638         case FMODE_READ:
1639                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1640                 break;
1641         case 0:
1642                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1643                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1644                 clear_bit(NFS_OPEN_STATE, &state->flags);
1645         }
1646         if (stateid == NULL)
1647                 return;
1648         /* Handle OPEN+OPEN_DOWNGRADE races */
1649         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1650             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1651                 nfs_resync_open_stateid_locked(state);
1652                 goto out;
1653         }
1654         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1655                 nfs4_stateid_copy(&state->stateid, stateid);
1656         nfs4_stateid_copy(&state->open_stateid, stateid);
1657         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1658 out:
1659         nfs_state_log_update_open_stateid(state);
1660 }
1661
1662 static void nfs_clear_open_stateid(struct nfs4_state *state,
1663         nfs4_stateid *arg_stateid,
1664         nfs4_stateid *stateid, fmode_t fmode)
1665 {
1666         write_seqlock(&state->seqlock);
1667         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1668         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1669                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1670         write_sequnlock(&state->seqlock);
1671         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1672                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1673 }
1674
1675 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1676                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1677         __must_hold(&state->owner->so_lock)
1678         __must_hold(&state->seqlock)
1679         __must_hold(RCU)
1680
1681 {
1682         DEFINE_WAIT(wait);
1683         int status = 0;
1684         for (;;) {
1685
1686                 if (nfs_stateid_is_sequential(state, stateid))
1687                         break;
1688
1689                 if (status)
1690                         break;
1691                 /* Rely on seqids for serialisation with NFSv4.0 */
1692                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1693                         break;
1694
1695                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1696                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1697                 /*
1698                  * Ensure we process the state changes in the same order
1699                  * in which the server processed them by delaying the
1700                  * update of the stateid until we are in sequence.
1701                  */
1702                 write_sequnlock(&state->seqlock);
1703                 spin_unlock(&state->owner->so_lock);
1704                 rcu_read_unlock();
1705                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1706
1707                 if (!fatal_signal_pending(current)) {
1708                         if (schedule_timeout(5*HZ) == 0)
1709                                 status = -EAGAIN;
1710                         else
1711                                 status = 0;
1712                 } else
1713                         status = -EINTR;
1714                 finish_wait(&state->waitq, &wait);
1715                 rcu_read_lock();
1716                 spin_lock(&state->owner->so_lock);
1717                 write_seqlock(&state->seqlock);
1718         }
1719
1720         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1721             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1722                 nfs4_stateid_copy(freeme, &state->open_stateid);
1723                 nfs_test_and_clear_all_open_stateid(state);
1724         }
1725
1726         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1727                 nfs4_stateid_copy(&state->stateid, stateid);
1728         nfs4_stateid_copy(&state->open_stateid, stateid);
1729         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1730         nfs_state_log_update_open_stateid(state);
1731 }
1732
1733 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1734                 const nfs4_stateid *open_stateid,
1735                 fmode_t fmode,
1736                 nfs4_stateid *freeme)
1737 {
1738         /*
1739          * Protect the call to nfs4_state_set_mode_locked and
1740          * serialise the stateid update
1741          */
1742         write_seqlock(&state->seqlock);
1743         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1744         switch (fmode) {
1745         case FMODE_READ:
1746                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1747                 break;
1748         case FMODE_WRITE:
1749                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1750                 break;
1751         case FMODE_READ|FMODE_WRITE:
1752                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1753         }
1754         set_bit(NFS_OPEN_STATE, &state->flags);
1755         write_sequnlock(&state->seqlock);
1756 }
1757
1758 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1759 {
1760         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1761         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1762         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1763         clear_bit(NFS_OPEN_STATE, &state->flags);
1764 }
1765
1766 static void nfs_state_set_delegation(struct nfs4_state *state,
1767                 const nfs4_stateid *deleg_stateid,
1768                 fmode_t fmode)
1769 {
1770         /*
1771          * Protect the call to nfs4_state_set_mode_locked and
1772          * serialise the stateid update
1773          */
1774         write_seqlock(&state->seqlock);
1775         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1776         set_bit(NFS_DELEGATED_STATE, &state->flags);
1777         write_sequnlock(&state->seqlock);
1778 }
1779
1780 static void nfs_state_clear_delegation(struct nfs4_state *state)
1781 {
1782         write_seqlock(&state->seqlock);
1783         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1784         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1785         write_sequnlock(&state->seqlock);
1786 }
1787
1788 int update_open_stateid(struct nfs4_state *state,
1789                 const nfs4_stateid *open_stateid,
1790                 const nfs4_stateid *delegation,
1791                 fmode_t fmode)
1792 {
1793         struct nfs_server *server = NFS_SERVER(state->inode);
1794         struct nfs_client *clp = server->nfs_client;
1795         struct nfs_inode *nfsi = NFS_I(state->inode);
1796         struct nfs_delegation *deleg_cur;
1797         nfs4_stateid freeme = { };
1798         int ret = 0;
1799
1800         fmode &= (FMODE_READ|FMODE_WRITE);
1801
1802         rcu_read_lock();
1803         spin_lock(&state->owner->so_lock);
1804         if (open_stateid != NULL) {
1805                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1806                 ret = 1;
1807         }
1808
1809         deleg_cur = nfs4_get_valid_delegation(state->inode);
1810         if (deleg_cur == NULL)
1811                 goto no_delegation;
1812
1813         spin_lock(&deleg_cur->lock);
1814         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1815            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1816             (deleg_cur->type & fmode) != fmode)
1817                 goto no_delegation_unlock;
1818
1819         if (delegation == NULL)
1820                 delegation = &deleg_cur->stateid;
1821         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1822                 goto no_delegation_unlock;
1823
1824         nfs_mark_delegation_referenced(deleg_cur);
1825         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1826         ret = 1;
1827 no_delegation_unlock:
1828         spin_unlock(&deleg_cur->lock);
1829 no_delegation:
1830         if (ret)
1831                 update_open_stateflags(state, fmode);
1832         spin_unlock(&state->owner->so_lock);
1833         rcu_read_unlock();
1834
1835         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1836                 nfs4_schedule_state_manager(clp);
1837         if (freeme.type != 0)
1838                 nfs4_test_and_free_stateid(server, &freeme,
1839                                 state->owner->so_cred);
1840
1841         return ret;
1842 }
1843
1844 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1845                 const nfs4_stateid *stateid)
1846 {
1847         struct nfs4_state *state = lsp->ls_state;
1848         bool ret = false;
1849
1850         spin_lock(&state->state_lock);
1851         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1852                 goto out_noupdate;
1853         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1854                 goto out_noupdate;
1855         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1856         ret = true;
1857 out_noupdate:
1858         spin_unlock(&state->state_lock);
1859         return ret;
1860 }
1861
1862 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1863 {
1864         struct nfs_delegation *delegation;
1865
1866         fmode &= FMODE_READ|FMODE_WRITE;
1867         rcu_read_lock();
1868         delegation = nfs4_get_valid_delegation(inode);
1869         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1870                 rcu_read_unlock();
1871                 return;
1872         }
1873         rcu_read_unlock();
1874         nfs4_inode_return_delegation(inode);
1875 }
1876
1877 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1878 {
1879         struct nfs4_state *state = opendata->state;
1880         struct nfs_delegation *delegation;
1881         int open_mode = opendata->o_arg.open_flags;
1882         fmode_t fmode = opendata->o_arg.fmode;
1883         enum open_claim_type4 claim = opendata->o_arg.claim;
1884         nfs4_stateid stateid;
1885         int ret = -EAGAIN;
1886
1887         for (;;) {
1888                 spin_lock(&state->owner->so_lock);
1889                 if (can_open_cached(state, fmode, open_mode, claim)) {
1890                         update_open_stateflags(state, fmode);
1891                         spin_unlock(&state->owner->so_lock);
1892                         goto out_return_state;
1893                 }
1894                 spin_unlock(&state->owner->so_lock);
1895                 rcu_read_lock();
1896                 delegation = nfs4_get_valid_delegation(state->inode);
1897                 if (!can_open_delegated(delegation, fmode, claim)) {
1898                         rcu_read_unlock();
1899                         break;
1900                 }
1901                 /* Save the delegation */
1902                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1903                 rcu_read_unlock();
1904                 nfs_release_seqid(opendata->o_arg.seqid);
1905                 if (!opendata->is_recover) {
1906                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1907                         if (ret != 0)
1908                                 goto out;
1909                 }
1910                 ret = -EAGAIN;
1911
1912                 /* Try to update the stateid using the delegation */
1913                 if (update_open_stateid(state, NULL, &stateid, fmode))
1914                         goto out_return_state;
1915         }
1916 out:
1917         return ERR_PTR(ret);
1918 out_return_state:
1919         refcount_inc(&state->count);
1920         return state;
1921 }
1922
1923 static void
1924 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1925 {
1926         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1927         struct nfs_delegation *delegation;
1928         int delegation_flags = 0;
1929
1930         rcu_read_lock();
1931         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1932         if (delegation)
1933                 delegation_flags = delegation->flags;
1934         rcu_read_unlock();
1935         switch (data->o_arg.claim) {
1936         default:
1937                 break;
1938         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1939         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1940                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1941                                    "returning a delegation for "
1942                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1943                                    clp->cl_hostname);
1944                 return;
1945         }
1946         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1947                 nfs_inode_set_delegation(state->inode,
1948                                 data->owner->so_cred,
1949                                 data->o_res.delegation_type,
1950                                 &data->o_res.delegation,
1951                                 data->o_res.pagemod_limit);
1952         else
1953                 nfs_inode_reclaim_delegation(state->inode,
1954                                 data->owner->so_cred,
1955                                 data->o_res.delegation_type,
1956                                 &data->o_res.delegation,
1957                                 data->o_res.pagemod_limit);
1958
1959         if (data->o_res.do_recall)
1960                 nfs_async_inode_return_delegation(state->inode,
1961                                                   &data->o_res.delegation);
1962 }
1963
1964 /*
1965  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1966  * and update the nfs4_state.
1967  */
1968 static struct nfs4_state *
1969 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1970 {
1971         struct inode *inode = data->state->inode;
1972         struct nfs4_state *state = data->state;
1973         int ret;
1974
1975         if (!data->rpc_done) {
1976                 if (data->rpc_status)
1977                         return ERR_PTR(data->rpc_status);
1978                 /* cached opens have already been processed */
1979                 goto update;
1980         }
1981
1982         ret = nfs_refresh_inode(inode, &data->f_attr);
1983         if (ret)
1984                 return ERR_PTR(ret);
1985
1986         if (data->o_res.delegation_type != 0)
1987                 nfs4_opendata_check_deleg(data, state);
1988 update:
1989         if (!update_open_stateid(state, &data->o_res.stateid,
1990                                 NULL, data->o_arg.fmode))
1991                 return ERR_PTR(-EAGAIN);
1992         refcount_inc(&state->count);
1993
1994         return state;
1995 }
1996
1997 static struct inode *
1998 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1999 {
2000         struct inode *inode;
2001
2002         switch (data->o_arg.claim) {
2003         case NFS4_OPEN_CLAIM_NULL:
2004         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2005         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2006                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2007                         return ERR_PTR(-EAGAIN);
2008                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2009                                 &data->f_attr);
2010                 break;
2011         default:
2012                 inode = d_inode(data->dentry);
2013                 ihold(inode);
2014                 nfs_refresh_inode(inode, &data->f_attr);
2015         }
2016         return inode;
2017 }
2018
2019 static struct nfs4_state *
2020 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2021 {
2022         struct nfs4_state *state;
2023         struct inode *inode;
2024
2025         inode = nfs4_opendata_get_inode(data);
2026         if (IS_ERR(inode))
2027                 return ERR_CAST(inode);
2028         if (data->state != NULL && data->state->inode == inode) {
2029                 state = data->state;
2030                 refcount_inc(&state->count);
2031         } else
2032                 state = nfs4_get_open_state(inode, data->owner);
2033         iput(inode);
2034         if (state == NULL)
2035                 state = ERR_PTR(-ENOMEM);
2036         return state;
2037 }
2038
2039 static struct nfs4_state *
2040 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2041 {
2042         struct nfs4_state *state;
2043
2044         if (!data->rpc_done) {
2045                 state = nfs4_try_open_cached(data);
2046                 trace_nfs4_cached_open(data->state);
2047                 goto out;
2048         }
2049
2050         state = nfs4_opendata_find_nfs4_state(data);
2051         if (IS_ERR(state))
2052                 goto out;
2053
2054         if (data->o_res.delegation_type != 0)
2055                 nfs4_opendata_check_deleg(data, state);
2056         if (!update_open_stateid(state, &data->o_res.stateid,
2057                                 NULL, data->o_arg.fmode)) {
2058                 nfs4_put_open_state(state);
2059                 state = ERR_PTR(-EAGAIN);
2060         }
2061 out:
2062         nfs_release_seqid(data->o_arg.seqid);
2063         return state;
2064 }
2065
2066 static struct nfs4_state *
2067 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2068 {
2069         struct nfs4_state *ret;
2070
2071         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2072                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2073         else
2074                 ret = _nfs4_opendata_to_nfs4_state(data);
2075         nfs4_sequence_free_slot(&data->o_res.seq_res);
2076         return ret;
2077 }
2078
2079 static struct nfs_open_context *
2080 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2081 {
2082         struct nfs_inode *nfsi = NFS_I(state->inode);
2083         struct nfs_open_context *ctx;
2084
2085         rcu_read_lock();
2086         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2087                 if (ctx->state != state)
2088                         continue;
2089                 if ((ctx->mode & mode) != mode)
2090                         continue;
2091                 if (!get_nfs_open_context(ctx))
2092                         continue;
2093                 rcu_read_unlock();
2094                 return ctx;
2095         }
2096         rcu_read_unlock();
2097         return ERR_PTR(-ENOENT);
2098 }
2099
2100 static struct nfs_open_context *
2101 nfs4_state_find_open_context(struct nfs4_state *state)
2102 {
2103         struct nfs_open_context *ctx;
2104
2105         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2106         if (!IS_ERR(ctx))
2107                 return ctx;
2108         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2109         if (!IS_ERR(ctx))
2110                 return ctx;
2111         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2112 }
2113
2114 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2115                 struct nfs4_state *state, enum open_claim_type4 claim)
2116 {
2117         struct nfs4_opendata *opendata;
2118
2119         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2120                         NULL, claim, GFP_NOFS);
2121         if (opendata == NULL)
2122                 return ERR_PTR(-ENOMEM);
2123         opendata->state = state;
2124         refcount_inc(&state->count);
2125         return opendata;
2126 }
2127
2128 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2129                 fmode_t fmode)
2130 {
2131         struct nfs4_state *newstate;
2132         int ret;
2133
2134         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2135                 return 0;
2136         opendata->o_arg.open_flags = 0;
2137         opendata->o_arg.fmode = fmode;
2138         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2139                         NFS_SB(opendata->dentry->d_sb),
2140                         fmode, 0);
2141         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2142         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2143         nfs4_init_opendata_res(opendata);
2144         ret = _nfs4_recover_proc_open(opendata);
2145         if (ret != 0)
2146                 return ret; 
2147         newstate = nfs4_opendata_to_nfs4_state(opendata);
2148         if (IS_ERR(newstate))
2149                 return PTR_ERR(newstate);
2150         if (newstate != opendata->state)
2151                 ret = -ESTALE;
2152         nfs4_close_state(newstate, fmode);
2153         return ret;
2154 }
2155
2156 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2157 {
2158         int ret;
2159
2160         /* memory barrier prior to reading state->n_* */
2161         smp_rmb();
2162         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2163         if (ret != 0)
2164                 return ret;
2165         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2166         if (ret != 0)
2167                 return ret;
2168         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2169         if (ret != 0)
2170                 return ret;
2171         /*
2172          * We may have performed cached opens for all three recoveries.
2173          * Check if we need to update the current stateid.
2174          */
2175         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2176             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2177                 write_seqlock(&state->seqlock);
2178                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2179                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2180                 write_sequnlock(&state->seqlock);
2181         }
2182         return 0;
2183 }
2184
2185 /*
2186  * OPEN_RECLAIM:
2187  *      reclaim state on the server after a reboot.
2188  */
2189 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2190 {
2191         struct nfs_delegation *delegation;
2192         struct nfs4_opendata *opendata;
2193         fmode_t delegation_type = 0;
2194         int status;
2195
2196         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2197                         NFS4_OPEN_CLAIM_PREVIOUS);
2198         if (IS_ERR(opendata))
2199                 return PTR_ERR(opendata);
2200         rcu_read_lock();
2201         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2202         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2203                 delegation_type = delegation->type;
2204         rcu_read_unlock();
2205         opendata->o_arg.u.delegation_type = delegation_type;
2206         status = nfs4_open_recover(opendata, state);
2207         nfs4_opendata_put(opendata);
2208         return status;
2209 }
2210
2211 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2212 {
2213         struct nfs_server *server = NFS_SERVER(state->inode);
2214         struct nfs4_exception exception = { };
2215         int err;
2216         do {
2217                 err = _nfs4_do_open_reclaim(ctx, state);
2218                 trace_nfs4_open_reclaim(ctx, 0, err);
2219                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2220                         continue;
2221                 if (err != -NFS4ERR_DELAY)
2222                         break;
2223                 nfs4_handle_exception(server, err, &exception);
2224         } while (exception.retry);
2225         return err;
2226 }
2227
2228 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2229 {
2230         struct nfs_open_context *ctx;
2231         int ret;
2232
2233         ctx = nfs4_state_find_open_context(state);
2234         if (IS_ERR(ctx))
2235                 return -EAGAIN;
2236         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2237         nfs_state_clear_open_state_flags(state);
2238         ret = nfs4_do_open_reclaim(ctx, state);
2239         put_nfs_open_context(ctx);
2240         return ret;
2241 }
2242
2243 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)
2244 {
2245         switch (err) {
2246                 default:
2247                         printk(KERN_ERR "NFS: %s: unhandled error "
2248                                         "%d.\n", __func__, err);
2249                         fallthrough;
2250                 case 0:
2251                 case -ENOENT:
2252                 case -EAGAIN:
2253                 case -ESTALE:
2254                 case -ETIMEDOUT:
2255                         break;
2256                 case -NFS4ERR_BADSESSION:
2257                 case -NFS4ERR_BADSLOT:
2258                 case -NFS4ERR_BAD_HIGH_SLOT:
2259                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2260                 case -NFS4ERR_DEADSESSION:
2261                         return -EAGAIN;
2262                 case -NFS4ERR_STALE_CLIENTID:
2263                 case -NFS4ERR_STALE_STATEID:
2264                         /* Don't recall a delegation if it was lost */
2265                         nfs4_schedule_lease_recovery(server->nfs_client);
2266                         return -EAGAIN;
2267                 case -NFS4ERR_MOVED:
2268                         nfs4_schedule_migration_recovery(server);
2269                         return -EAGAIN;
2270                 case -NFS4ERR_LEASE_MOVED:
2271                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2272                         return -EAGAIN;
2273                 case -NFS4ERR_DELEG_REVOKED:
2274                 case -NFS4ERR_ADMIN_REVOKED:
2275                 case -NFS4ERR_EXPIRED:
2276                 case -NFS4ERR_BAD_STATEID:
2277                 case -NFS4ERR_OPENMODE:
2278                         nfs_inode_find_state_and_recover(state->inode,
2279                                         stateid);
2280                         nfs4_schedule_stateid_recovery(server, state);
2281                         return -EAGAIN;
2282                 case -NFS4ERR_DELAY:
2283                 case -NFS4ERR_GRACE:
2284                         ssleep(1);
2285                         return -EAGAIN;
2286                 case -ENOMEM:
2287                 case -NFS4ERR_DENIED:
2288                         if (fl) {
2289                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2290                                 if (lsp)
2291                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2292                         }
2293                         return 0;
2294         }
2295         return err;
2296 }
2297
2298 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2299                 struct nfs4_state *state, const nfs4_stateid *stateid)
2300 {
2301         struct nfs_server *server = NFS_SERVER(state->inode);
2302         struct nfs4_opendata *opendata;
2303         int err = 0;
2304
2305         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2306                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2307         if (IS_ERR(opendata))
2308                 return PTR_ERR(opendata);
2309         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2310         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2311                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2312                 if (err)
2313                         goto out;
2314         }
2315         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2316                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2317                 if (err)
2318                         goto out;
2319         }
2320         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2321                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2322                 if (err)
2323                         goto out;
2324         }
2325         nfs_state_clear_delegation(state);
2326 out:
2327         nfs4_opendata_put(opendata);
2328         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2329 }
2330
2331 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2332 {
2333         struct nfs4_opendata *data = calldata;
2334
2335         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2336                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2337 }
2338
2339 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2340 {
2341         struct nfs4_opendata *data = calldata;
2342
2343         nfs40_sequence_done(task, &data->c_res.seq_res);
2344
2345         data->rpc_status = task->tk_status;
2346         if (data->rpc_status == 0) {
2347                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2348                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2349                 renew_lease(data->o_res.server, data->timestamp);
2350                 data->rpc_done = true;
2351         }
2352 }
2353
2354 static void nfs4_open_confirm_release(void *calldata)
2355 {
2356         struct nfs4_opendata *data = calldata;
2357         struct nfs4_state *state = NULL;
2358
2359         /* If this request hasn't been cancelled, do nothing */
2360         if (!data->cancelled)
2361                 goto out_free;
2362         /* In case of error, no cleanup! */
2363         if (!data->rpc_done)
2364                 goto out_free;
2365         state = nfs4_opendata_to_nfs4_state(data);
2366         if (!IS_ERR(state))
2367                 nfs4_close_state(state, data->o_arg.fmode);
2368 out_free:
2369         nfs4_opendata_put(data);
2370 }
2371
2372 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2373         .rpc_call_prepare = nfs4_open_confirm_prepare,
2374         .rpc_call_done = nfs4_open_confirm_done,
2375         .rpc_release = nfs4_open_confirm_release,
2376 };
2377
2378 /*
2379  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2380  */
2381 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2382 {
2383         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2384         struct rpc_task *task;
2385         struct  rpc_message msg = {
2386                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2387                 .rpc_argp = &data->c_arg,
2388                 .rpc_resp = &data->c_res,
2389                 .rpc_cred = data->owner->so_cred,
2390         };
2391         struct rpc_task_setup task_setup_data = {
2392                 .rpc_client = server->client,
2393                 .rpc_message = &msg,
2394                 .callback_ops = &nfs4_open_confirm_ops,
2395                 .callback_data = data,
2396                 .workqueue = nfsiod_workqueue,
2397                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2398         };
2399         int status;
2400
2401         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2402                                 data->is_recover);
2403         kref_get(&data->kref);
2404         data->rpc_done = false;
2405         data->rpc_status = 0;
2406         data->timestamp = jiffies;
2407         task = rpc_run_task(&task_setup_data);
2408         if (IS_ERR(task))
2409                 return PTR_ERR(task);
2410         status = rpc_wait_for_completion_task(task);
2411         if (status != 0) {
2412                 data->cancelled = true;
2413                 smp_wmb();
2414         } else
2415                 status = data->rpc_status;
2416         rpc_put_task(task);
2417         return status;
2418 }
2419
2420 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2421 {
2422         struct nfs4_opendata *data = calldata;
2423         struct nfs4_state_owner *sp = data->owner;
2424         struct nfs_client *clp = sp->so_server->nfs_client;
2425         enum open_claim_type4 claim = data->o_arg.claim;
2426
2427         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2428                 goto out_wait;
2429         /*
2430          * Check if we still need to send an OPEN call, or if we can use
2431          * a delegation instead.
2432          */
2433         if (data->state != NULL) {
2434                 struct nfs_delegation *delegation;
2435
2436                 if (can_open_cached(data->state, data->o_arg.fmode,
2437                                         data->o_arg.open_flags, claim))
2438                         goto out_no_action;
2439                 rcu_read_lock();
2440                 delegation = nfs4_get_valid_delegation(data->state->inode);
2441                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2442                         goto unlock_no_action;
2443                 rcu_read_unlock();
2444         }
2445         /* Update client id. */
2446         data->o_arg.clientid = clp->cl_clientid;
2447         switch (claim) {
2448         default:
2449                 break;
2450         case NFS4_OPEN_CLAIM_PREVIOUS:
2451         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2452         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2453                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2454                 fallthrough;
2455         case NFS4_OPEN_CLAIM_FH:
2456                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2457         }
2458         data->timestamp = jiffies;
2459         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2460                                 &data->o_arg.seq_args,
2461                                 &data->o_res.seq_res,
2462                                 task) != 0)
2463                 nfs_release_seqid(data->o_arg.seqid);
2464
2465         /* Set the create mode (note dependency on the session type) */
2466         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2467         if (data->o_arg.open_flags & O_EXCL) {
2468                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2469                 if (clp->cl_mvops->minor_version == 0) {
2470                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2471                         /* don't put an ACCESS op in OPEN compound if O_EXCL,
2472                          * because ACCESS will return permission denied for
2473                          * all bits until close */
2474                         data->o_res.access_request = data->o_arg.access = 0;
2475                 } else if (nfs4_has_persistent_session(clp))
2476                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2477         }
2478         return;
2479 unlock_no_action:
2480         trace_nfs4_cached_open(data->state);
2481         rcu_read_unlock();
2482 out_no_action:
2483         task->tk_action = NULL;
2484 out_wait:
2485         nfs4_sequence_done(task, &data->o_res.seq_res);
2486 }
2487
2488 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2489 {
2490         struct nfs4_opendata *data = calldata;
2491
2492         data->rpc_status = task->tk_status;
2493
2494         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2495                 return;
2496
2497         if (task->tk_status == 0) {
2498                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2499                         switch (data->o_res.f_attr->mode & S_IFMT) {
2500                         case S_IFREG:
2501                                 break;
2502                         case S_IFLNK:
2503                                 data->rpc_status = -ELOOP;
2504                                 break;
2505                         case S_IFDIR:
2506                                 data->rpc_status = -EISDIR;
2507                                 break;
2508                         default:
2509                                 data->rpc_status = -ENOTDIR;
2510                         }
2511                 }
2512                 renew_lease(data->o_res.server, data->timestamp);
2513                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2514                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2515         }
2516         data->rpc_done = true;
2517 }
2518
2519 static void nfs4_open_release(void *calldata)
2520 {
2521         struct nfs4_opendata *data = calldata;
2522         struct nfs4_state *state = NULL;
2523
2524         /* If this request hasn't been cancelled, do nothing */
2525         if (!data->cancelled)
2526                 goto out_free;
2527         /* In case of error, no cleanup! */
2528         if (data->rpc_status != 0 || !data->rpc_done)
2529                 goto out_free;
2530         /* In case we need an open_confirm, no cleanup! */
2531         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2532                 goto out_free;
2533         state = nfs4_opendata_to_nfs4_state(data);
2534         if (!IS_ERR(state))
2535                 nfs4_close_state(state, data->o_arg.fmode);
2536 out_free:
2537         nfs4_opendata_put(data);
2538 }
2539
2540 static const struct rpc_call_ops nfs4_open_ops = {
2541         .rpc_call_prepare = nfs4_open_prepare,
2542         .rpc_call_done = nfs4_open_done,
2543         .rpc_release = nfs4_open_release,
2544 };
2545
2546 static int nfs4_run_open_task(struct nfs4_opendata *data,
2547                               struct nfs_open_context *ctx)
2548 {
2549         struct inode *dir = d_inode(data->dir);
2550         struct nfs_server *server = NFS_SERVER(dir);
2551         struct nfs_openargs *o_arg = &data->o_arg;
2552         struct nfs_openres *o_res = &data->o_res;
2553         struct rpc_task *task;
2554         struct rpc_message msg = {
2555                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2556                 .rpc_argp = o_arg,
2557                 .rpc_resp = o_res,
2558                 .rpc_cred = data->owner->so_cred,
2559         };
2560         struct rpc_task_setup task_setup_data = {
2561                 .rpc_client = server->client,
2562                 .rpc_message = &msg,
2563                 .callback_ops = &nfs4_open_ops,
2564                 .callback_data = data,
2565                 .workqueue = nfsiod_workqueue,
2566                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2567         };
2568         int status;
2569
2570         if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
2571                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
2572
2573         kref_get(&data->kref);
2574         data->rpc_done = false;
2575         data->rpc_status = 0;
2576         data->cancelled = false;
2577         data->is_recover = false;
2578         if (!ctx) {
2579                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2580                 data->is_recover = true;
2581                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2582         } else {
2583                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2584                 pnfs_lgopen_prepare(data, ctx);
2585         }
2586         task = rpc_run_task(&task_setup_data);
2587         if (IS_ERR(task))
2588                 return PTR_ERR(task);
2589         status = rpc_wait_for_completion_task(task);
2590         if (status != 0) {
2591                 data->cancelled = true;
2592                 smp_wmb();
2593         } else
2594                 status = data->rpc_status;
2595         rpc_put_task(task);
2596
2597         return status;
2598 }
2599
2600 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2601 {
2602         struct inode *dir = d_inode(data->dir);
2603         struct nfs_openres *o_res = &data->o_res;
2604         int status;
2605
2606         status = nfs4_run_open_task(data, NULL);
2607         if (status != 0 || !data->rpc_done)
2608                 return status;
2609
2610         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2611
2612         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2613                 status = _nfs4_proc_open_confirm(data);
2614
2615         return status;
2616 }
2617
2618 /*
2619  * Additional permission checks in order to distinguish between an
2620  * open for read, and an open for execute. This works around the
2621  * fact that NFSv4 OPEN treats read and execute permissions as being
2622  * the same.
2623  * Note that in the non-execute case, we want to turn off permission
2624  * checking if we just created a new file (POSIX open() semantics).
2625  */
2626 static int nfs4_opendata_access(const struct cred *cred,
2627                                 struct nfs4_opendata *opendata,
2628                                 struct nfs4_state *state, fmode_t fmode,
2629                                 int openflags)
2630 {
2631         struct nfs_access_entry cache;
2632         u32 mask, flags;
2633
2634         /* access call failed or for some reason the server doesn't
2635          * support any access modes -- defer access call until later */
2636         if (opendata->o_res.access_supported == 0)
2637                 return 0;
2638
2639         mask = 0;
2640         /*
2641          * Use openflags to check for exec, because fmode won't
2642          * always have FMODE_EXEC set when file open for exec.
2643          */
2644         if (openflags & __FMODE_EXEC) {
2645                 /* ONLY check for exec rights */
2646                 if (S_ISDIR(state->inode->i_mode))
2647                         mask = NFS4_ACCESS_LOOKUP;
2648                 else
2649                         mask = NFS4_ACCESS_EXECUTE;
2650         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2651                 mask = NFS4_ACCESS_READ;
2652
2653         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2654         nfs_access_add_cache(state->inode, &cache, cred);
2655
2656         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2657         if ((mask & ~cache.mask & flags) == 0)
2658                 return 0;
2659
2660         return -EACCES;
2661 }
2662
2663 /*
2664  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2665  */
2666 static int _nfs4_proc_open(struct nfs4_opendata *data,
2667                            struct nfs_open_context *ctx)
2668 {
2669         struct inode *dir = d_inode(data->dir);
2670         struct nfs_server *server = NFS_SERVER(dir);
2671         struct nfs_openargs *o_arg = &data->o_arg;
2672         struct nfs_openres *o_res = &data->o_res;
2673         int status;
2674
2675         status = nfs4_run_open_task(data, ctx);
2676         if (!data->rpc_done)
2677                 return status;
2678         if (status != 0) {
2679                 if (status == -NFS4ERR_BADNAME &&
2680                                 !(o_arg->open_flags & O_CREAT))
2681                         return -ENOENT;
2682                 return status;
2683         }
2684
2685         nfs_fattr_map_and_free_names(server, &data->f_attr);
2686
2687         if (o_arg->open_flags & O_CREAT) {
2688                 if (o_arg->open_flags & O_EXCL)
2689                         data->file_created = true;
2690                 else if (o_res->cinfo.before != o_res->cinfo.after)
2691                         data->file_created = true;
2692                 if (data->file_created ||
2693                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2694                         nfs4_update_changeattr(dir, &o_res->cinfo,
2695                                         o_res->f_attr->time_start,
2696                                         NFS_INO_INVALID_DATA);
2697         }
2698         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2699                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2700         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2701                 status = _nfs4_proc_open_confirm(data);
2702                 if (status != 0)
2703                         return status;
2704         }
2705         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2706                 nfs4_sequence_free_slot(&o_res->seq_res);
2707                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, NULL);
2708         }
2709         return 0;
2710 }
2711
2712 /*
2713  * OPEN_EXPIRED:
2714  *      reclaim state on the server after a network partition.
2715  *      Assumes caller holds the appropriate lock
2716  */
2717 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2718 {
2719         struct nfs4_opendata *opendata;
2720         int ret;
2721
2722         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2723                         NFS4_OPEN_CLAIM_FH);
2724         if (IS_ERR(opendata))
2725                 return PTR_ERR(opendata);
2726         ret = nfs4_open_recover(opendata, state);
2727         if (ret == -ESTALE)
2728                 d_drop(ctx->dentry);
2729         nfs4_opendata_put(opendata);
2730         return ret;
2731 }
2732
2733 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2734 {
2735         struct nfs_server *server = NFS_SERVER(state->inode);
2736         struct nfs4_exception exception = { };
2737         int err;
2738
2739         do {
2740                 err = _nfs4_open_expired(ctx, state);
2741                 trace_nfs4_open_expired(ctx, 0, err);
2742                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2743                         continue;
2744                 switch (err) {
2745                 default:
2746                         goto out;
2747                 case -NFS4ERR_GRACE:
2748                 case -NFS4ERR_DELAY:
2749                         nfs4_handle_exception(server, err, &exception);
2750                         err = 0;
2751                 }
2752         } while (exception.retry);
2753 out:
2754         return err;
2755 }
2756
2757 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2758 {
2759         struct nfs_open_context *ctx;
2760         int ret;
2761
2762         ctx = nfs4_state_find_open_context(state);
2763         if (IS_ERR(ctx))
2764                 return -EAGAIN;
2765         ret = nfs4_do_open_expired(ctx, state);
2766         put_nfs_open_context(ctx);
2767         return ret;
2768 }
2769
2770 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2771                 const nfs4_stateid *stateid)
2772 {
2773         nfs_remove_bad_delegation(state->inode, stateid);
2774         nfs_state_clear_delegation(state);
2775 }
2776
2777 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2778 {
2779         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2780                 nfs_finish_clear_delegation_stateid(state, NULL);
2781 }
2782
2783 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2784 {
2785         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2786         nfs40_clear_delegation_stateid(state);
2787         nfs_state_clear_open_state_flags(state);
2788         return nfs4_open_expired(sp, state);
2789 }
2790
2791 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2792                 nfs4_stateid *stateid,
2793                 const struct cred *cred)
2794 {
2795         return -NFS4ERR_BAD_STATEID;
2796 }
2797
2798 #if defined(CONFIG_NFS_V4_1)
2799 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2800                 nfs4_stateid *stateid,
2801                 const struct cred *cred)
2802 {
2803         int status;
2804
2805         switch (stateid->type) {
2806         default:
2807                 break;
2808         case NFS4_INVALID_STATEID_TYPE:
2809         case NFS4_SPECIAL_STATEID_TYPE:
2810                 return -NFS4ERR_BAD_STATEID;
2811         case NFS4_REVOKED_STATEID_TYPE:
2812                 goto out_free;
2813         }
2814
2815         status = nfs41_test_stateid(server, stateid, cred);
2816         switch (status) {
2817         case -NFS4ERR_EXPIRED:
2818         case -NFS4ERR_ADMIN_REVOKED:
2819         case -NFS4ERR_DELEG_REVOKED:
2820                 break;
2821         default:
2822                 return status;
2823         }
2824 out_free:
2825         /* Ack the revoked state to the server */
2826         nfs41_free_stateid(server, stateid, cred, true);
2827         return -NFS4ERR_EXPIRED;
2828 }
2829
2830 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2831 {
2832         struct nfs_server *server = NFS_SERVER(state->inode);
2833         nfs4_stateid stateid;
2834         struct nfs_delegation *delegation;
2835         const struct cred *cred = NULL;
2836         int status, ret = NFS_OK;
2837
2838         /* Get the delegation credential for use by test/free_stateid */
2839         rcu_read_lock();
2840         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2841         if (delegation == NULL) {
2842                 rcu_read_unlock();
2843                 nfs_state_clear_delegation(state);
2844                 return NFS_OK;
2845         }
2846
2847         spin_lock(&delegation->lock);
2848         nfs4_stateid_copy(&stateid, &delegation->stateid);
2849
2850         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2851                                 &delegation->flags)) {
2852                 spin_unlock(&delegation->lock);
2853                 rcu_read_unlock();
2854                 return NFS_OK;
2855         }
2856
2857         if (delegation->cred)
2858                 cred = get_cred(delegation->cred);
2859         spin_unlock(&delegation->lock);
2860         rcu_read_unlock();
2861         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2862         trace_nfs4_test_delegation_stateid(state, NULL, status);
2863         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2864                 nfs_finish_clear_delegation_stateid(state, &stateid);
2865         else
2866                 ret = status;
2867
2868         put_cred(cred);
2869         return ret;
2870 }
2871
2872 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2873 {
2874         nfs4_stateid tmp;
2875
2876         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2877             nfs4_copy_delegation_stateid(state->inode, state->state,
2878                                 &tmp, NULL) &&
2879             nfs4_stateid_match_other(&state->stateid, &tmp))
2880                 nfs_state_set_delegation(state, &tmp, state->state);
2881         else
2882                 nfs_state_clear_delegation(state);
2883 }
2884
2885 /**
2886  * nfs41_check_expired_locks - possibly free a lock stateid
2887  *
2888  * @state: NFSv4 state for an inode
2889  *
2890  * Returns NFS_OK if recovery for this stateid is now finished.
2891  * Otherwise a negative NFS4ERR value is returned.
2892  */
2893 static int nfs41_check_expired_locks(struct nfs4_state *state)
2894 {
2895         int status, ret = NFS_OK;
2896         struct nfs4_lock_state *lsp, *prev = NULL;
2897         struct nfs_server *server = NFS_SERVER(state->inode);
2898
2899         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2900                 goto out;
2901
2902         spin_lock(&state->state_lock);
2903         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2904                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2905                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2906
2907                         refcount_inc(&lsp->ls_count);
2908                         spin_unlock(&state->state_lock);
2909
2910                         nfs4_put_lock_state(prev);
2911                         prev = lsp;
2912
2913                         status = nfs41_test_and_free_expired_stateid(server,
2914                                         &lsp->ls_stateid,
2915                                         cred);
2916                         trace_nfs4_test_lock_stateid(state, lsp, status);
2917                         if (status == -NFS4ERR_EXPIRED ||
2918                             status == -NFS4ERR_BAD_STATEID) {
2919                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2920                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2921                                 if (!recover_lost_locks)
2922                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2923                         } else if (status != NFS_OK) {
2924                                 ret = status;
2925                                 nfs4_put_lock_state(prev);
2926                                 goto out;
2927                         }
2928                         spin_lock(&state->state_lock);
2929                 }
2930         }
2931         spin_unlock(&state->state_lock);
2932         nfs4_put_lock_state(prev);
2933 out:
2934         return ret;
2935 }
2936
2937 /**
2938  * nfs41_check_open_stateid - possibly free an open stateid
2939  *
2940  * @state: NFSv4 state for an inode
2941  *
2942  * Returns NFS_OK if recovery for this stateid is now finished.
2943  * Otherwise a negative NFS4ERR value is returned.
2944  */
2945 static int nfs41_check_open_stateid(struct nfs4_state *state)
2946 {
2947         struct nfs_server *server = NFS_SERVER(state->inode);
2948         nfs4_stateid *stateid = &state->open_stateid;
2949         const struct cred *cred = state->owner->so_cred;
2950         int status;
2951
2952         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2953                 return -NFS4ERR_BAD_STATEID;
2954         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2955         trace_nfs4_test_open_stateid(state, NULL, status);
2956         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2957                 nfs_state_clear_open_state_flags(state);
2958                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2959                 return status;
2960         }
2961         if (nfs_open_stateid_recover_openmode(state))
2962                 return -NFS4ERR_OPENMODE;
2963         return NFS_OK;
2964 }
2965
2966 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2967 {
2968         int status;
2969
2970         status = nfs41_check_delegation_stateid(state);
2971         if (status != NFS_OK)
2972                 return status;
2973         nfs41_delegation_recover_stateid(state);
2974
2975         status = nfs41_check_expired_locks(state);
2976         if (status != NFS_OK)
2977                 return status;
2978         status = nfs41_check_open_stateid(state);
2979         if (status != NFS_OK)
2980                 status = nfs4_open_expired(sp, state);
2981         return status;
2982 }
2983 #endif
2984
2985 /*
2986  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2987  * fields corresponding to attributes that were used to store the verifier.
2988  * Make sure we clobber those fields in the later setattr call
2989  */
2990 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2991                                 struct iattr *sattr, struct nfs4_label **label)
2992 {
2993         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2994         __u32 attrset[3];
2995         unsigned ret;
2996         unsigned i;
2997
2998         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2999                 attrset[i] = opendata->o_res.attrset[i];
3000                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3001                         attrset[i] &= ~bitmask[i];
3002         }
3003
3004         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3005                 sattr->ia_valid : 0;
3006
3007         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3008                 if (sattr->ia_valid & ATTR_ATIME_SET)
3009                         ret |= ATTR_ATIME_SET;
3010                 else
3011                         ret |= ATTR_ATIME;
3012         }
3013
3014         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3015                 if (sattr->ia_valid & ATTR_MTIME_SET)
3016                         ret |= ATTR_MTIME_SET;
3017                 else
3018                         ret |= ATTR_MTIME;
3019         }
3020
3021         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3022                 *label = NULL;
3023         return ret;
3024 }
3025
3026 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3027                 int flags, struct nfs_open_context *ctx)
3028 {
3029         struct nfs4_state_owner *sp = opendata->owner;
3030         struct nfs_server *server = sp->so_server;
3031         struct dentry *dentry;
3032         struct nfs4_state *state;
3033         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3034         struct inode *dir = d_inode(opendata->dir);
3035         unsigned long dir_verifier;
3036         unsigned int seq;
3037         int ret;
3038
3039         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3040         dir_verifier = nfs_save_change_attribute(dir);
3041
3042         ret = _nfs4_proc_open(opendata, ctx);
3043         if (ret != 0)
3044                 goto out;
3045
3046         state = _nfs4_opendata_to_nfs4_state(opendata);
3047         ret = PTR_ERR(state);
3048         if (IS_ERR(state))
3049                 goto out;
3050         ctx->state = state;
3051         if (server->caps & NFS_CAP_POSIX_LOCK)
3052                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3053         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3054                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3055         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_PRESERVE_UNLINKED)
3056                 set_bit(NFS_INO_PRESERVE_UNLINKED, &NFS_I(state->inode)->flags);
3057
3058         dentry = opendata->dentry;
3059         if (d_really_is_negative(dentry)) {
3060                 struct dentry *alias;
3061                 d_drop(dentry);
3062                 alias = d_exact_alias(dentry, state->inode);
3063                 if (!alias)
3064                         alias = d_splice_alias(igrab(state->inode), dentry);
3065                 /* d_splice_alias() can't fail here - it's a non-directory */
3066                 if (alias) {
3067                         dput(ctx->dentry);
3068                         ctx->dentry = dentry = alias;
3069                 }
3070         }
3071
3072         switch(opendata->o_arg.claim) {
3073         default:
3074                 break;
3075         case NFS4_OPEN_CLAIM_NULL:
3076         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3077         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3078                 if (!opendata->rpc_done)
3079                         break;
3080                 if (opendata->o_res.delegation_type != 0)
3081                         dir_verifier = nfs_save_change_attribute(dir);
3082                 nfs_set_verifier(dentry, dir_verifier);
3083         }
3084
3085         /* Parse layoutget results before we check for access */
3086         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3087
3088         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3089                         acc_mode, flags);
3090         if (ret != 0)
3091                 goto out;
3092
3093         if (d_inode(dentry) == state->inode) {
3094                 nfs_inode_attach_open_context(ctx);
3095                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3096                         nfs4_schedule_stateid_recovery(server, state);
3097         }
3098
3099 out:
3100         if (!opendata->cancelled) {
3101                 if (opendata->lgp) {
3102                         nfs4_lgopen_release(opendata->lgp);
3103                         opendata->lgp = NULL;
3104                 }
3105                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3106         }
3107         return ret;
3108 }
3109
3110 /*
3111  * Returns a referenced nfs4_state
3112  */
3113 static int _nfs4_do_open(struct inode *dir,
3114                         struct nfs_open_context *ctx,
3115                         int flags,
3116                         const struct nfs4_open_createattrs *c,
3117                         int *opened)
3118 {
3119         struct nfs4_state_owner  *sp;
3120         struct nfs4_state     *state = NULL;
3121         struct nfs_server       *server = NFS_SERVER(dir);
3122         struct nfs4_opendata *opendata;
3123         struct dentry *dentry = ctx->dentry;
3124         const struct cred *cred = ctx->cred;
3125         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3126         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3127         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3128         struct iattr *sattr = c->sattr;
3129         struct nfs4_label *label = c->label;
3130         int status;
3131
3132         /* Protect against reboot recovery conflicts */
3133         status = -ENOMEM;
3134         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3135         if (sp == NULL) {
3136                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3137                 goto out_err;
3138         }
3139         status = nfs4_client_recover_expired_lease(server->nfs_client);
3140         if (status != 0)
3141                 goto err_put_state_owner;
3142         if (d_really_is_positive(dentry))
3143                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3144         status = -ENOMEM;
3145         if (d_really_is_positive(dentry))
3146                 claim = NFS4_OPEN_CLAIM_FH;
3147         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3148                         c, claim, GFP_KERNEL);
3149         if (opendata == NULL)
3150                 goto err_put_state_owner;
3151
3152         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3153                 if (!opendata->f_attr.mdsthreshold) {
3154                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3155                         if (!opendata->f_attr.mdsthreshold)
3156                                 goto err_opendata_put;
3157                 }
3158                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3159         }
3160         if (d_really_is_positive(dentry))
3161                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3162
3163         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3164         if (status != 0)
3165                 goto err_opendata_put;
3166         state = ctx->state;
3167
3168         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3169             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3170                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3171                 /*
3172                  * send create attributes which was not set by open
3173                  * with an extra setattr.
3174                  */
3175                 if (attrs || label) {
3176                         unsigned ia_old = sattr->ia_valid;
3177
3178                         sattr->ia_valid = attrs;
3179                         nfs_fattr_init(opendata->o_res.f_attr);
3180                         status = nfs4_do_setattr(state->inode, cred,
3181                                         opendata->o_res.f_attr, sattr,
3182                                         ctx, label);
3183                         if (status == 0) {
3184                                 nfs_setattr_update_inode(state->inode, sattr,
3185                                                 opendata->o_res.f_attr);
3186                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3187                         }
3188                         sattr->ia_valid = ia_old;
3189                 }
3190         }
3191         if (opened && opendata->file_created)
3192                 *opened = 1;
3193
3194         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3195                 *ctx_th = opendata->f_attr.mdsthreshold;
3196                 opendata->f_attr.mdsthreshold = NULL;
3197         }
3198
3199         nfs4_opendata_put(opendata);
3200         nfs4_put_state_owner(sp);
3201         return 0;
3202 err_opendata_put:
3203         nfs4_opendata_put(opendata);
3204 err_put_state_owner:
3205         nfs4_put_state_owner(sp);
3206 out_err:
3207         return status;
3208 }
3209
3210
3211 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3212                                         struct nfs_open_context *ctx,
3213                                         int flags,
3214                                         struct iattr *sattr,
3215                                         struct nfs4_label *label,
3216                                         int *opened)
3217 {
3218         struct nfs_server *server = NFS_SERVER(dir);
3219         struct nfs4_exception exception = {
3220                 .interruptible = true,
3221         };
3222         struct nfs4_state *res;
3223         struct nfs4_open_createattrs c = {
3224                 .label = label,
3225                 .sattr = sattr,
3226                 .verf = {
3227                         [0] = (__u32)jiffies,
3228                         [1] = (__u32)current->pid,
3229                 },
3230         };
3231         int status;
3232
3233         do {
3234                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3235                 res = ctx->state;
3236                 trace_nfs4_open_file(ctx, flags, status);
3237                 if (status == 0)
3238                         break;
3239                 /* NOTE: BAD_SEQID means the server and client disagree about the
3240                  * book-keeping w.r.t. state-changing operations
3241                  * (OPEN/CLOSE/LOCK/LOCKU...)
3242                  * It is actually a sign of a bug on the client or on the server.
3243                  *
3244                  * If we receive a BAD_SEQID error in the particular case of
3245                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3246                  * have unhashed the old state_owner for us, and that we can
3247                  * therefore safely retry using a new one. We should still warn
3248                  * the user though...
3249                  */
3250                 if (status == -NFS4ERR_BAD_SEQID) {
3251                         pr_warn_ratelimited("NFS: v4 server %s "
3252                                         " returned a bad sequence-id error!\n",
3253                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3254                         exception.retry = 1;
3255                         continue;
3256                 }
3257                 /*
3258                  * BAD_STATEID on OPEN means that the server cancelled our
3259                  * state before it received the OPEN_CONFIRM.
3260                  * Recover by retrying the request as per the discussion
3261                  * on Page 181 of RFC3530.
3262                  */
3263                 if (status == -NFS4ERR_BAD_STATEID) {
3264                         exception.retry = 1;
3265                         continue;
3266                 }
3267                 if (status == -NFS4ERR_EXPIRED) {
3268                         nfs4_schedule_lease_recovery(server->nfs_client);
3269                         exception.retry = 1;
3270                         continue;
3271                 }
3272                 if (status == -EAGAIN) {
3273                         /* We must have found a delegation */
3274                         exception.retry = 1;
3275                         continue;
3276                 }
3277                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3278                         continue;
3279                 res = ERR_PTR(nfs4_handle_exception(server,
3280                                         status, &exception));
3281         } while (exception.retry);
3282         return res;
3283 }
3284
3285 static int _nfs4_do_setattr(struct inode *inode,
3286                             struct nfs_setattrargs *arg,
3287                             struct nfs_setattrres *res,
3288                             const struct cred *cred,
3289                             struct nfs_open_context *ctx)
3290 {
3291         struct nfs_server *server = NFS_SERVER(inode);
3292         struct rpc_message msg = {
3293                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3294                 .rpc_argp       = arg,
3295                 .rpc_resp       = res,
3296                 .rpc_cred       = cred,
3297         };
3298         const struct cred *delegation_cred = NULL;
3299         unsigned long timestamp = jiffies;
3300         bool truncate;
3301         int status;
3302
3303         nfs_fattr_init(res->fattr);
3304
3305         /* Servers should only apply open mode checks for file size changes */
3306         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3307         if (!truncate) {
3308                 nfs4_inode_make_writeable(inode);
3309                 goto zero_stateid;
3310         }
3311
3312         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3313                 /* Use that stateid */
3314         } else if (ctx != NULL && ctx->state) {
3315                 struct nfs_lock_context *l_ctx;
3316                 if (!nfs4_valid_open_stateid(ctx->state))
3317                         return -EBADF;
3318                 l_ctx = nfs_get_lock_context(ctx);
3319                 if (IS_ERR(l_ctx))
3320                         return PTR_ERR(l_ctx);
3321                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3322                                                 &arg->stateid, &delegation_cred);
3323                 nfs_put_lock_context(l_ctx);
3324                 if (status == -EIO)
3325                         return -EBADF;
3326                 else if (status == -EAGAIN)
3327                         goto zero_stateid;
3328         } else {
3329 zero_stateid:
3330                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3331         }
3332         if (delegation_cred)
3333                 msg.rpc_cred = delegation_cred;
3334
3335         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3336
3337         put_cred(delegation_cred);
3338         if (status == 0 && ctx != NULL)
3339                 renew_lease(server, timestamp);
3340         trace_nfs4_setattr(inode, &arg->stateid, status);
3341         return status;
3342 }
3343
3344 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3345                            struct nfs_fattr *fattr, struct iattr *sattr,
3346                            struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3347 {
3348         struct nfs_server *server = NFS_SERVER(inode);
3349         __u32 bitmask[NFS4_BITMASK_SZ];
3350         struct nfs4_state *state = ctx ? ctx->state : NULL;
3351         struct nfs_setattrargs  arg = {
3352                 .fh             = NFS_FH(inode),
3353                 .iap            = sattr,
3354                 .server         = server,
3355                 .bitmask = bitmask,
3356                 .label          = ilabel,
3357         };
3358         struct nfs_setattrres  res = {
3359                 .fattr          = fattr,
3360                 .server         = server,
3361         };
3362         struct nfs4_exception exception = {
3363                 .state = state,
3364                 .inode = inode,
3365                 .stateid = &arg.stateid,
3366         };
3367         unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3368         int err;
3369
3370         if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3371                 adjust_flags |= NFS_INO_INVALID_MODE;
3372         if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3373                 adjust_flags |= NFS_INO_INVALID_OTHER;
3374
3375         do {
3376                 nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3377                                         inode, adjust_flags);
3378
3379                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3380                 switch (err) {
3381                 case -NFS4ERR_OPENMODE:
3382                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3383                                 pr_warn_once("NFSv4: server %s is incorrectly "
3384                                                 "applying open mode checks to "
3385                                                 "a SETATTR that is not "
3386                                                 "changing file size.\n",
3387                                                 server->nfs_client->cl_hostname);
3388                         }
3389                         if (state && !(state->state & FMODE_WRITE)) {
3390                                 err = -EBADF;
3391                                 if (sattr->ia_valid & ATTR_OPEN)
3392                                         err = -EACCES;
3393                                 goto out;
3394                         }
3395                 }
3396                 err = nfs4_handle_exception(server, err, &exception);
3397         } while (exception.retry);
3398 out:
3399         return err;
3400 }
3401
3402 static bool
3403 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3404 {
3405         if (inode == NULL || !nfs_have_layout(inode))
3406                 return false;
3407
3408         return pnfs_wait_on_layoutreturn(inode, task);
3409 }
3410
3411 /*
3412  * Update the seqid of an open stateid
3413  */
3414 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3415                 struct nfs4_state *state)
3416 {
3417         __be32 seqid_open;
3418         u32 dst_seqid;
3419         int seq;
3420
3421         for (;;) {
3422                 if (!nfs4_valid_open_stateid(state))
3423                         break;
3424                 seq = read_seqbegin(&state->seqlock);
3425                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3426                         nfs4_stateid_copy(dst, &state->open_stateid);
3427                         if (read_seqretry(&state->seqlock, seq))
3428                                 continue;
3429                         break;
3430                 }
3431                 seqid_open = state->open_stateid.seqid;
3432                 if (read_seqretry(&state->seqlock, seq))
3433                         continue;
3434
3435                 dst_seqid = be32_to_cpu(dst->seqid);
3436                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3437                         dst->seqid = seqid_open;
3438                 break;
3439         }
3440 }
3441
3442 /*
3443  * Update the seqid of an open stateid after receiving
3444  * NFS4ERR_OLD_STATEID
3445  */
3446 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3447                 struct nfs4_state *state)
3448 {
3449         __be32 seqid_open;
3450         u32 dst_seqid;
3451         bool ret;
3452         int seq, status = -EAGAIN;
3453         DEFINE_WAIT(wait);
3454
3455         for (;;) {
3456                 ret = false;
3457                 if (!nfs4_valid_open_stateid(state))
3458                         break;
3459                 seq = read_seqbegin(&state->seqlock);
3460                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3461                         if (read_seqretry(&state->seqlock, seq))
3462                                 continue;
3463                         break;
3464                 }
3465
3466                 write_seqlock(&state->seqlock);
3467                 seqid_open = state->open_stateid.seqid;
3468
3469                 dst_seqid = be32_to_cpu(dst->seqid);
3470
3471                 /* Did another OPEN bump the state's seqid?  try again: */
3472                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3473                         dst->seqid = seqid_open;
3474                         write_sequnlock(&state->seqlock);
3475                         ret = true;
3476                         break;
3477                 }
3478
3479                 /* server says we're behind but we haven't seen the update yet */
3480                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3481                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3482                 write_sequnlock(&state->seqlock);
3483                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3484
3485                 if (fatal_signal_pending(current))
3486                         status = -EINTR;
3487                 else
3488                         if (schedule_timeout(5*HZ) != 0)
3489                                 status = 0;
3490
3491                 finish_wait(&state->waitq, &wait);
3492
3493                 if (!status)
3494                         continue;
3495                 if (status == -EINTR)
3496                         break;
3497
3498                 /* we slept the whole 5 seconds, we must have lost a seqid */
3499                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3500                 ret = true;
3501                 break;
3502         }
3503
3504         return ret;
3505 }
3506
3507 struct nfs4_closedata {
3508         struct inode *inode;
3509         struct nfs4_state *state;
3510         struct nfs_closeargs arg;
3511         struct nfs_closeres res;
3512         struct {
3513                 struct nfs4_layoutreturn_args arg;
3514                 struct nfs4_layoutreturn_res res;
3515                 struct nfs4_xdr_opaque_data ld_private;
3516                 u32 roc_barrier;
3517                 bool roc;
3518         } lr;
3519         struct nfs_fattr fattr;
3520         unsigned long timestamp;
3521 };
3522
3523 static void nfs4_free_closedata(void *data)
3524 {
3525         struct nfs4_closedata *calldata = data;
3526         struct nfs4_state_owner *sp = calldata->state->owner;
3527         struct super_block *sb = calldata->state->inode->i_sb;
3528
3529         if (calldata->lr.roc)
3530                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3531                                 calldata->res.lr_ret);
3532         nfs4_put_open_state(calldata->state);
3533         nfs_free_seqid(calldata->arg.seqid);
3534         nfs4_put_state_owner(sp);
3535         nfs_sb_deactive(sb);
3536         kfree(calldata);
3537 }
3538
3539 static void nfs4_close_done(struct rpc_task *task, void *data)
3540 {
3541         struct nfs4_closedata *calldata = data;
3542         struct nfs4_state *state = calldata->state;
3543         struct nfs_server *server = NFS_SERVER(calldata->inode);
3544         nfs4_stateid *res_stateid = NULL;
3545         struct nfs4_exception exception = {
3546                 .state = state,
3547                 .inode = calldata->inode,
3548                 .stateid = &calldata->arg.stateid,
3549         };
3550
3551         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3552                 return;
3553         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3554
3555         /* Handle Layoutreturn errors */
3556         if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3557                           &calldata->res.lr_ret) == -EAGAIN)
3558                 goto out_restart;
3559
3560         /* hmm. we are done with the inode, and in the process of freeing
3561          * the state_owner. we keep this around to process errors
3562          */
3563         switch (task->tk_status) {
3564                 case 0:
3565                         res_stateid = &calldata->res.stateid;
3566                         renew_lease(server, calldata->timestamp);
3567                         break;
3568                 case -NFS4ERR_ACCESS:
3569                         if (calldata->arg.bitmask != NULL) {
3570                                 calldata->arg.bitmask = NULL;
3571                                 calldata->res.fattr = NULL;
3572                                 goto out_restart;
3573
3574                         }
3575                         break;
3576                 case -NFS4ERR_OLD_STATEID:
3577                         /* Did we race with OPEN? */
3578                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3579                                                 state))
3580                                 goto out_restart;
3581                         goto out_release;
3582                 case -NFS4ERR_ADMIN_REVOKED:
3583                 case -NFS4ERR_STALE_STATEID:
3584                 case -NFS4ERR_EXPIRED:
3585                         nfs4_free_revoked_stateid(server,
3586                                         &calldata->arg.stateid,
3587                                         task->tk_msg.rpc_cred);
3588                         fallthrough;
3589                 case -NFS4ERR_BAD_STATEID:
3590                         if (calldata->arg.fmode == 0)
3591                                 break;
3592                         fallthrough;
3593                 default:
3594                         task->tk_status = nfs4_async_handle_exception(task,
3595                                         server, task->tk_status, &exception);
3596                         if (exception.retry)
3597                                 goto out_restart;
3598         }
3599         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3600                         res_stateid, calldata->arg.fmode);
3601 out_release:
3602         task->tk_status = 0;
3603         nfs_release_seqid(calldata->arg.seqid);
3604         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3605         dprintk("%s: ret = %d\n", __func__, task->tk_status);
3606         return;
3607 out_restart:
3608         task->tk_status = 0;
3609         rpc_restart_call_prepare(task);
3610         goto out_release;
3611 }
3612
3613 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3614 {
3615         struct nfs4_closedata *calldata = data;
3616         struct nfs4_state *state = calldata->state;
3617         struct inode *inode = calldata->inode;
3618         struct nfs_server *server = NFS_SERVER(inode);
3619         struct pnfs_layout_hdr *lo;
3620         bool is_rdonly, is_wronly, is_rdwr;
3621         int call_close = 0;
3622
3623         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3624                 goto out_wait;
3625
3626         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3627         spin_lock(&state->owner->so_lock);
3628         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3629         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3630         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3631         /* Calculate the change in open mode */
3632         calldata->arg.fmode = 0;
3633         if (state->n_rdwr == 0) {
3634                 if (state->n_rdonly == 0)
3635                         call_close |= is_rdonly;
3636                 else if (is_rdonly)
3637                         calldata->arg.fmode |= FMODE_READ;
3638                 if (state->n_wronly == 0)
3639                         call_close |= is_wronly;
3640                 else if (is_wronly)
3641                         calldata->arg.fmode |= FMODE_WRITE;
3642                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3643                         call_close |= is_rdwr;
3644         } else if (is_rdwr)
3645                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3646
3647         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3648         if (!nfs4_valid_open_stateid(state))
3649                 call_close = 0;
3650         spin_unlock(&state->owner->so_lock);
3651
3652         if (!call_close) {
3653                 /* Note: exit _without_ calling nfs4_close_done */
3654                 goto out_no_action;
3655         }
3656
3657         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3658                 nfs_release_seqid(calldata->arg.seqid);
3659                 goto out_wait;
3660         }
3661
3662         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3663         if (lo && !pnfs_layout_is_valid(lo)) {
3664                 calldata->arg.lr_args = NULL;
3665                 calldata->res.lr_res = NULL;
3666         }
3667
3668         if (calldata->arg.fmode == 0)
3669                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3670
3671         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3672                 /* Close-to-open cache consistency revalidation */
3673                 if (!nfs4_have_delegation(inode, FMODE_READ)) {
3674                         nfs4_bitmask_set(calldata->arg.bitmask_store,
3675                                          server->cache_consistency_bitmask,
3676                                          inode, 0);
3677                         calldata->arg.bitmask = calldata->arg.bitmask_store;
3678                 } else
3679                         calldata->arg.bitmask = NULL;
3680         }
3681
3682         calldata->arg.share_access =
3683                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3684                                 calldata->arg.fmode, 0);
3685
3686         if (calldata->res.fattr == NULL)
3687                 calldata->arg.bitmask = NULL;
3688         else if (calldata->arg.bitmask == NULL)
3689                 calldata->res.fattr = NULL;
3690         calldata->timestamp = jiffies;
3691         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3692                                 &calldata->arg.seq_args,
3693                                 &calldata->res.seq_res,
3694                                 task) != 0)
3695                 nfs_release_seqid(calldata->arg.seqid);
3696         return;
3697 out_no_action:
3698         task->tk_action = NULL;
3699 out_wait:
3700         nfs4_sequence_done(task, &calldata->res.seq_res);
3701 }
3702
3703 static const struct rpc_call_ops nfs4_close_ops = {
3704         .rpc_call_prepare = nfs4_close_prepare,
3705         .rpc_call_done = nfs4_close_done,
3706         .rpc_release = nfs4_free_closedata,
3707 };
3708
3709 /* 
3710  * It is possible for data to be read/written from a mem-mapped file 
3711  * after the sys_close call (which hits the vfs layer as a flush).
3712  * This means that we can't safely call nfsv4 close on a file until 
3713  * the inode is cleared. This in turn means that we are not good
3714  * NFSv4 citizens - we do not indicate to the server to update the file's 
3715  * share state even when we are done with one of the three share 
3716  * stateid's in the inode.
3717  *
3718  * NOTE: Caller must be holding the sp->so_owner semaphore!
3719  */
3720 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3721 {
3722         struct nfs_server *server = NFS_SERVER(state->inode);
3723         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3724         struct nfs4_closedata *calldata;
3725         struct nfs4_state_owner *sp = state->owner;
3726         struct rpc_task *task;
3727         struct rpc_message msg = {
3728                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3729                 .rpc_cred = state->owner->so_cred,
3730         };
3731         struct rpc_task_setup task_setup_data = {
3732                 .rpc_client = server->client,
3733                 .rpc_message = &msg,
3734                 .callback_ops = &nfs4_close_ops,
3735                 .workqueue = nfsiod_workqueue,
3736                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3737         };
3738         int status = -ENOMEM;
3739
3740         if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
3741                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
3742
3743         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3744                 &task_setup_data.rpc_client, &msg);
3745
3746         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3747         if (calldata == NULL)
3748                 goto out;
3749         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3750         calldata->inode = state->inode;
3751         calldata->state = state;
3752         calldata->arg.fh = NFS_FH(state->inode);
3753         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3754                 goto out_free_calldata;
3755         /* Serialization for the sequence id */
3756         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3757         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3758         if (IS_ERR(calldata->arg.seqid))
3759                 goto out_free_calldata;
3760         nfs_fattr_init(&calldata->fattr);
3761         calldata->arg.fmode = 0;
3762         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3763         calldata->res.fattr = &calldata->fattr;
3764         calldata->res.seqid = calldata->arg.seqid;
3765         calldata->res.server = server;
3766         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3767         calldata->lr.roc = pnfs_roc(state->inode,
3768                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3769         if (calldata->lr.roc) {
3770                 calldata->arg.lr_args = &calldata->lr.arg;
3771                 calldata->res.lr_res = &calldata->lr.res;
3772         }
3773         nfs_sb_active(calldata->inode->i_sb);
3774
3775         msg.rpc_argp = &calldata->arg;
3776         msg.rpc_resp = &calldata->res;
3777         task_setup_data.callback_data = calldata;
3778         task = rpc_run_task(&task_setup_data);
3779         if (IS_ERR(task))
3780                 return PTR_ERR(task);
3781         status = 0;
3782         if (wait)
3783                 status = rpc_wait_for_completion_task(task);
3784         rpc_put_task(task);
3785         return status;
3786 out_free_calldata:
3787         kfree(calldata);
3788 out:
3789         nfs4_put_open_state(state);
3790         nfs4_put_state_owner(sp);
3791         return status;
3792 }
3793
3794 static struct inode *
3795 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3796                 int open_flags, struct iattr *attr, int *opened)
3797 {
3798         struct nfs4_state *state;
3799         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3800
3801         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3802
3803         /* Protect against concurrent sillydeletes */
3804         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3805
3806         nfs4_label_release_security(label);
3807
3808         if (IS_ERR(state))
3809                 return ERR_CAST(state);
3810         return state->inode;
3811 }
3812
3813 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3814 {
3815         if (ctx->state == NULL)
3816                 return;
3817         if (is_sync)
3818                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3819         else
3820                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3821 }
3822
3823 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3824 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3825 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3826
3827 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3828 {
3829         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3830         struct nfs4_server_caps_arg args = {
3831                 .fhandle = fhandle,
3832                 .bitmask = bitmask,
3833         };
3834         struct nfs4_server_caps_res res = {};
3835         struct rpc_message msg = {
3836                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3837                 .rpc_argp = &args,
3838                 .rpc_resp = &res,
3839         };
3840         int status;
3841         int i;
3842
3843         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3844                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3845                      FATTR4_WORD0_LINK_SUPPORT |
3846                      FATTR4_WORD0_SYMLINK_SUPPORT |
3847                      FATTR4_WORD0_ACLSUPPORT |
3848                      FATTR4_WORD0_CASE_INSENSITIVE |
3849                      FATTR4_WORD0_CASE_PRESERVING;
3850         if (minorversion)
3851                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3852
3853         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3854         if (status == 0) {
3855                 /* Sanity check the server answers */
3856                 switch (minorversion) {
3857                 case 0:
3858                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3859                         res.attr_bitmask[2] = 0;
3860                         break;
3861                 case 1:
3862                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3863                         break;
3864                 case 2:
3865                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3866                 }
3867                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3868                 server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3869                                   NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3870                 server->fattr_valid = NFS_ATTR_FATTR_V4;
3871                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3872                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3873                         server->caps |= NFS_CAP_ACLS;
3874                 if (res.has_links != 0)
3875                         server->caps |= NFS_CAP_HARDLINKS;
3876                 if (res.has_symlinks != 0)
3877                         server->caps |= NFS_CAP_SYMLINKS;
3878                 if (res.case_insensitive)
3879                         server->caps |= NFS_CAP_CASE_INSENSITIVE;
3880                 if (res.case_preserving)
3881                         server->caps |= NFS_CAP_CASE_PRESERVING;
3882 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3883                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3884                         server->caps |= NFS_CAP_SECURITY_LABEL;
3885 #endif
3886                 if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3887                         server->caps |= NFS_CAP_FS_LOCATIONS;
3888                 if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3889                         server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3890                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3891                         server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3892                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3893                         server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3894                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3895                         server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3896                                 NFS_ATTR_FATTR_OWNER_NAME);
3897                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3898                         server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3899                                 NFS_ATTR_FATTR_GROUP_NAME);
3900                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3901                         server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3902                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3903                         server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3904                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3905                         server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3906                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3907                         server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3908                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3909                                 sizeof(server->attr_bitmask));
3910                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3911
3912                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3913                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3914                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3915                 server->cache_consistency_bitmask[2] = 0;
3916
3917                 /* Avoid a regression due to buggy server */
3918                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3919                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3920                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3921                         sizeof(server->exclcreat_bitmask));
3922
3923                 server->acl_bitmask = res.acl_bitmask;
3924                 server->fh_expire_type = res.fh_expire_type;
3925         }
3926
3927         return status;
3928 }
3929
3930 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3931 {
3932         struct nfs4_exception exception = {
3933                 .interruptible = true,
3934         };
3935         int err;
3936
3937         nfs4_server_set_init_caps(server);
3938         do {
3939                 err = nfs4_handle_exception(server,
3940                                 _nfs4_server_capabilities(server, fhandle),
3941                                 &exception);
3942         } while (exception.retry);
3943         return err;
3944 }
3945
3946 static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
3947                                           struct nfs_client *clp,
3948                                           struct nfs_server *server)
3949 {
3950         int i;
3951
3952         for (i = 0; i < location->nservers; i++) {
3953                 struct nfs4_string *srv_loc = &location->servers[i];
3954                 struct sockaddr_storage addr;
3955                 size_t addrlen;
3956                 struct xprt_create xprt_args = {
3957                         .ident = 0,
3958                         .net = clp->cl_net,
3959                 };
3960                 struct nfs4_add_xprt_data xprtdata = {
3961                         .clp = clp,
3962                 };
3963                 struct rpc_add_xprt_test rpcdata = {
3964                         .add_xprt_test = clp->cl_mvops->session_trunk,
3965                         .data = &xprtdata,
3966                 };
3967                 char *servername = NULL;
3968
3969                 if (!srv_loc->len)
3970                         continue;
3971
3972                 addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
3973                                                 &addr, sizeof(addr),
3974                                                 clp->cl_net, server->port);
3975                 if (!addrlen)
3976                         return;
3977                 xprt_args.dstaddr = (struct sockaddr *)&addr;
3978                 xprt_args.addrlen = addrlen;
3979                 servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
3980                 if (!servername)
3981                         return;
3982                 memcpy(servername, srv_loc->data, srv_loc->len);
3983                 servername[srv_loc->len] = '\0';
3984                 xprt_args.servername = servername;
3985
3986                 xprtdata.cred = nfs4_get_clid_cred(clp);
3987                 rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
3988                                   rpc_clnt_setup_test_and_add_xprt,
3989                                   &rpcdata);
3990                 if (xprtdata.cred)
3991                         put_cred(xprtdata.cred);
3992                 kfree(servername);
3993         }
3994 }
3995
3996 static int _nfs4_discover_trunking(struct nfs_server *server,
3997                                    struct nfs_fh *fhandle)
3998 {
3999         struct nfs4_fs_locations *locations = NULL;
4000         struct page *page;
4001         const struct cred *cred;
4002         struct nfs_client *clp = server->nfs_client;
4003         const struct nfs4_state_maintenance_ops *ops =
4004                 clp->cl_mvops->state_renewal_ops;
4005         int status = -ENOMEM, i;
4006
4007         cred = ops->get_state_renewal_cred(clp);
4008         if (cred == NULL) {
4009                 cred = nfs4_get_clid_cred(clp);
4010                 if (cred == NULL)
4011                         return -ENOKEY;
4012         }
4013
4014         page = alloc_page(GFP_KERNEL);
4015         if (!page)
4016                 return -ENOMEM;
4017         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4018         if (!locations)
4019                 goto out_free;
4020         locations->fattr = nfs_alloc_fattr();
4021         if (!locations->fattr)
4022                 goto out_free_2;
4023
4024         status = nfs4_proc_get_locations(server, fhandle, locations, page,
4025                                          cred);
4026         if (status)
4027                 goto out_free_3;
4028
4029         for (i = 0; i < locations->nlocations; i++)
4030                 test_fs_location_for_trunking(&locations->locations[i], clp,
4031                                               server);
4032 out_free_3:
4033         kfree(locations->fattr);
4034 out_free_2:
4035         kfree(locations);
4036 out_free:
4037         __free_page(page);
4038         return status;
4039 }
4040
4041 static int nfs4_discover_trunking(struct nfs_server *server,
4042                                   struct nfs_fh *fhandle)
4043 {
4044         struct nfs4_exception exception = {
4045                 .interruptible = true,
4046         };
4047         struct nfs_client *clp = server->nfs_client;
4048         int err = 0;
4049
4050         if (!nfs4_has_session(clp))
4051                 goto out;
4052         do {
4053                 err = nfs4_handle_exception(server,
4054                                 _nfs4_discover_trunking(server, fhandle),
4055                                 &exception);
4056         } while (exception.retry);
4057 out:
4058         return err;
4059 }
4060
4061 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4062                 struct nfs_fsinfo *info)
4063 {
4064         u32 bitmask[3];
4065         struct nfs4_lookup_root_arg args = {
4066                 .bitmask = bitmask,
4067         };
4068         struct nfs4_lookup_res res = {
4069                 .server = server,
4070                 .fattr = info->fattr,
4071                 .fh = fhandle,
4072         };
4073         struct rpc_message msg = {
4074                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4075                 .rpc_argp = &args,
4076                 .rpc_resp = &res,
4077         };
4078
4079         bitmask[0] = nfs4_fattr_bitmap[0];
4080         bitmask[1] = nfs4_fattr_bitmap[1];
4081         /*
4082          * Process the label in the upcoming getfattr
4083          */
4084         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4085
4086         nfs_fattr_init(info->fattr);
4087         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4088 }
4089
4090 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4091                 struct nfs_fsinfo *info)
4092 {
4093         struct nfs4_exception exception = {
4094                 .interruptible = true,
4095         };
4096         int err;
4097         do {
4098                 err = _nfs4_lookup_root(server, fhandle, info);
4099                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4100                 switch (err) {
4101                 case 0:
4102                 case -NFS4ERR_WRONGSEC:
4103                         goto out;
4104                 default:
4105                         err = nfs4_handle_exception(server, err, &exception);
4106                 }
4107         } while (exception.retry);
4108 out:
4109         return err;
4110 }
4111
4112 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4113                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4114 {
4115         struct rpc_auth_create_args auth_args = {
4116                 .pseudoflavor = flavor,
4117         };
4118         struct rpc_auth *auth;
4119
4120         auth = rpcauth_create(&auth_args, server->client);
4121         if (IS_ERR(auth))
4122                 return -EACCES;
4123         return nfs4_lookup_root(server, fhandle, info);
4124 }
4125
4126 /*
4127  * Retry pseudoroot lookup with various security flavors.  We do this when:
4128  *
4129  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4130  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4131  *
4132  * Returns zero on success, or a negative NFS4ERR value, or a
4133  * negative errno value.
4134  */
4135 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4136                               struct nfs_fsinfo *info)
4137 {
4138         /* Per 3530bis 15.33.5 */
4139         static const rpc_authflavor_t flav_array[] = {
4140                 RPC_AUTH_GSS_KRB5P,
4141                 RPC_AUTH_GSS_KRB5I,
4142                 RPC_AUTH_GSS_KRB5,
4143                 RPC_AUTH_UNIX,                  /* courtesy */
4144                 RPC_AUTH_NULL,
4145         };
4146         int status = -EPERM;
4147         size_t i;
4148
4149         if (server->auth_info.flavor_len > 0) {
4150                 /* try each flavor specified by user */
4151                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4152                         status = nfs4_lookup_root_sec(server, fhandle, info,
4153                                                 server->auth_info.flavors[i]);
4154                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4155                                 continue;
4156                         break;
4157                 }
4158         } else {
4159                 /* no flavors specified by user, try default list */
4160                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4161                         status = nfs4_lookup_root_sec(server, fhandle, info,
4162                                                       flav_array[i]);
4163                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4164                                 continue;
4165                         break;
4166                 }
4167         }
4168
4169         /*
4170          * -EACCES could mean that the user doesn't have correct permissions
4171          * to access the mount.  It could also mean that we tried to mount
4172          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4173          * existing mount programs don't handle -EACCES very well so it should
4174          * be mapped to -EPERM instead.
4175          */
4176         if (status == -EACCES)
4177                 status = -EPERM;
4178         return status;
4179 }
4180
4181 /**
4182  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4183  * @server: initialized nfs_server handle
4184  * @fhandle: we fill in the pseudo-fs root file handle
4185  * @info: we fill in an FSINFO struct
4186  * @auth_probe: probe the auth flavours
4187  *
4188  * Returns zero on success, or a negative errno.
4189  */
4190 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4191                          struct nfs_fsinfo *info,
4192                          bool auth_probe)
4193 {
4194         int status = 0;
4195
4196         if (!auth_probe)
4197                 status = nfs4_lookup_root(server, fhandle, info);
4198
4199         if (auth_probe || status == NFS4ERR_WRONGSEC)
4200                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4201                                 fhandle, info);
4202
4203         if (status == 0)
4204                 status = nfs4_server_capabilities(server, fhandle);
4205         if (status == 0)
4206                 status = nfs4_do_fsinfo(server, fhandle, info);
4207
4208         return nfs4_map_errors(status);
4209 }
4210
4211 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4212                               struct nfs_fsinfo *info)
4213 {
4214         int error;
4215         struct nfs_fattr *fattr = info->fattr;
4216
4217         error = nfs4_server_capabilities(server, mntfh);
4218         if (error < 0) {
4219                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4220                 return error;
4221         }
4222
4223         error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4224         if (error < 0) {
4225                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4226                 goto out;
4227         }
4228
4229         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4230             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4231                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4232
4233 out:
4234         return error;
4235 }
4236
4237 /*
4238  * Get locations and (maybe) other attributes of a referral.
4239  * Note that we'll actually follow the referral later when
4240  * we detect fsid mismatch in inode revalidation
4241  */
4242 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4243                              const struct qstr *name, struct nfs_fattr *fattr,
4244                              struct nfs_fh *fhandle)
4245 {
4246         int status = -ENOMEM;
4247         struct page *page = NULL;
4248         struct nfs4_fs_locations *locations = NULL;
4249
4250         page = alloc_page(GFP_KERNEL);
4251         if (page == NULL)
4252                 goto out;
4253         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4254         if (locations == NULL)
4255                 goto out;
4256
4257         locations->fattr = fattr;
4258
4259         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4260         if (status != 0)
4261                 goto out;
4262
4263         /*
4264          * If the fsid didn't change, this is a migration event, not a
4265          * referral.  Cause us to drop into the exception handler, which
4266          * will kick off migration recovery.
4267          */
4268         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4269                 dprintk("%s: server did not return a different fsid for"
4270                         " a referral at %s\n", __func__, name->name);
4271                 status = -NFS4ERR_MOVED;
4272                 goto out;
4273         }
4274         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4275         nfs_fixup_referral_attributes(fattr);
4276         memset(fhandle, 0, sizeof(struct nfs_fh));
4277 out:
4278         if (page)
4279                 __free_page(page);
4280         kfree(locations);
4281         return status;
4282 }
4283
4284 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4285                                 struct nfs_fattr *fattr, struct inode *inode)
4286 {
4287         __u32 bitmask[NFS4_BITMASK_SZ];
4288         struct nfs4_getattr_arg args = {
4289                 .fh = fhandle,
4290                 .bitmask = bitmask,
4291         };
4292         struct nfs4_getattr_res res = {
4293                 .fattr = fattr,
4294                 .server = server,
4295         };
4296         struct rpc_message msg = {
4297                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4298                 .rpc_argp = &args,
4299                 .rpc_resp = &res,
4300         };
4301         unsigned short task_flags = 0;
4302
4303         if (nfs4_has_session(server->nfs_client))
4304                 task_flags = RPC_TASK_MOVEABLE;
4305
4306         /* Is this is an attribute revalidation, subject to softreval? */
4307         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4308                 task_flags |= RPC_TASK_TIMEOUT;
4309
4310         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4311         nfs_fattr_init(fattr);
4312         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4313         return nfs4_do_call_sync(server->client, server, &msg,
4314                         &args.seq_args, &res.seq_res, task_flags);
4315 }
4316
4317 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4318                                 struct nfs_fattr *fattr, struct inode *inode)
4319 {
4320         struct nfs4_exception exception = {
4321                 .interruptible = true,
4322         };
4323         int err;
4324         do {
4325                 err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4326                 trace_nfs4_getattr(server, fhandle, fattr, err);
4327                 err = nfs4_handle_exception(server, err,
4328                                 &exception);
4329         } while (exception.retry);
4330         return err;
4331 }
4332
4333 /* 
4334  * The file is not closed if it is opened due to the a request to change
4335  * the size of the file. The open call will not be needed once the
4336  * VFS layer lookup-intents are implemented.
4337  *
4338  * Close is called when the inode is destroyed.
4339  * If we haven't opened the file for O_WRONLY, we
4340  * need to in the size_change case to obtain a stateid.
4341  *
4342  * Got race?
4343  * Because OPEN is always done by name in nfsv4, it is
4344  * possible that we opened a different file by the same
4345  * name.  We can recognize this race condition, but we
4346  * can't do anything about it besides returning an error.
4347  *
4348  * This will be fixed with VFS changes (lookup-intent).
4349  */
4350 static int
4351 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4352                   struct iattr *sattr)
4353 {
4354         struct inode *inode = d_inode(dentry);
4355         const struct cred *cred = NULL;
4356         struct nfs_open_context *ctx = NULL;
4357         int status;
4358
4359         if (pnfs_ld_layoutret_on_setattr(inode) &&
4360             sattr->ia_valid & ATTR_SIZE &&
4361             sattr->ia_size < i_size_read(inode))
4362                 pnfs_commit_and_return_layout(inode);
4363
4364         nfs_fattr_init(fattr);
4365         
4366         /* Deal with open(O_TRUNC) */
4367         if (sattr->ia_valid & ATTR_OPEN)
4368                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4369
4370         /* Optimization: if the end result is no change, don't RPC */
4371         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4372                 return 0;
4373
4374         /* Search for an existing open(O_WRITE) file */
4375         if (sattr->ia_valid & ATTR_FILE) {
4376
4377                 ctx = nfs_file_open_context(sattr->ia_file);
4378                 if (ctx)
4379                         cred = ctx->cred;
4380         }
4381
4382         /* Return any delegations if we're going to change ACLs */
4383         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4384                 nfs4_inode_make_writeable(inode);
4385
4386         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4387         if (status == 0) {
4388                 nfs_setattr_update_inode(inode, sattr, fattr);
4389                 nfs_setsecurity(inode, fattr);
4390         }
4391         return status;
4392 }
4393
4394 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4395                 struct dentry *dentry, struct nfs_fh *fhandle,
4396                 struct nfs_fattr *fattr)
4397 {
4398         struct nfs_server *server = NFS_SERVER(dir);
4399         int                    status;
4400         struct nfs4_lookup_arg args = {
4401                 .bitmask = server->attr_bitmask,
4402                 .dir_fh = NFS_FH(dir),
4403                 .name = &dentry->d_name,
4404         };
4405         struct nfs4_lookup_res res = {
4406                 .server = server,
4407                 .fattr = fattr,
4408                 .fh = fhandle,
4409         };
4410         struct rpc_message msg = {
4411                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4412                 .rpc_argp = &args,
4413                 .rpc_resp = &res,
4414         };
4415         unsigned short task_flags = 0;
4416
4417         if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4418                 task_flags = RPC_TASK_MOVEABLE;
4419
4420         /* Is this is an attribute revalidation, subject to softreval? */
4421         if (nfs_lookup_is_soft_revalidate(dentry))
4422                 task_flags |= RPC_TASK_TIMEOUT;
4423
4424         args.bitmask = nfs4_bitmask(server, fattr->label);
4425
4426         nfs_fattr_init(fattr);
4427
4428         dprintk("NFS call  lookup %pd2\n", dentry);
4429         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4430         status = nfs4_do_call_sync(clnt, server, &msg,
4431                         &args.seq_args, &res.seq_res, task_flags);
4432         dprintk("NFS reply lookup: %d\n", status);
4433         return status;
4434 }
4435
4436 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4437 {
4438         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4439                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4440         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4441         fattr->nlink = 2;
4442 }
4443
4444 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4445                                    struct dentry *dentry, struct nfs_fh *fhandle,
4446                                    struct nfs_fattr *fattr)
4447 {
4448         struct nfs4_exception exception = {
4449                 .interruptible = true,
4450         };
4451         struct rpc_clnt *client = *clnt;
4452         const struct qstr *name = &dentry->d_name;
4453         int err;
4454         do {
4455                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4456                 trace_nfs4_lookup(dir, name, err);
4457                 switch (err) {
4458                 case -NFS4ERR_BADNAME:
4459                         err = -ENOENT;
4460                         goto out;
4461                 case -NFS4ERR_MOVED:
4462                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4463                         if (err == -NFS4ERR_MOVED)
4464                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4465                         goto out;
4466                 case -NFS4ERR_WRONGSEC:
4467                         err = -EPERM;
4468                         if (client != *clnt)
4469                                 goto out;
4470                         client = nfs4_negotiate_security(client, dir, name);
4471                         if (IS_ERR(client))
4472                                 return PTR_ERR(client);
4473
4474                         exception.retry = 1;
4475                         break;
4476                 default:
4477                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4478                 }
4479         } while (exception.retry);
4480
4481 out:
4482         if (err == 0)
4483                 *clnt = client;
4484         else if (client != *clnt)
4485                 rpc_shutdown_client(client);
4486
4487         return err;
4488 }
4489
4490 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4491                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4492 {
4493         int status;
4494         struct rpc_clnt *client = NFS_CLIENT(dir);
4495
4496         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4497         if (client != NFS_CLIENT(dir)) {
4498                 rpc_shutdown_client(client);
4499                 nfs_fixup_secinfo_attributes(fattr);
4500         }
4501         return status;
4502 }
4503
4504 struct rpc_clnt *
4505 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4506                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4507 {
4508         struct rpc_clnt *client = NFS_CLIENT(dir);
4509         int status;
4510
4511         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4512         if (status < 0)
4513                 return ERR_PTR(status);
4514         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4515 }
4516
4517 static int _nfs4_proc_lookupp(struct inode *inode,
4518                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4519 {
4520         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4521         struct nfs_server *server = NFS_SERVER(inode);
4522         int                    status;
4523         struct nfs4_lookupp_arg args = {
4524                 .bitmask = server->attr_bitmask,
4525                 .fh = NFS_FH(inode),
4526         };
4527         struct nfs4_lookupp_res res = {
4528                 .server = server,
4529                 .fattr = fattr,
4530                 .fh = fhandle,
4531         };
4532         struct rpc_message msg = {
4533                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4534                 .rpc_argp = &args,
4535                 .rpc_resp = &res,
4536         };
4537         unsigned short task_flags = 0;
4538
4539         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4540                 task_flags |= RPC_TASK_TIMEOUT;
4541
4542         args.bitmask = nfs4_bitmask(server, fattr->label);
4543
4544         nfs_fattr_init(fattr);
4545
4546         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4547         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4548                                 &res.seq_res, task_flags);
4549         dprintk("NFS reply lookupp: %d\n", status);
4550         return status;
4551 }
4552
4553 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4554                              struct nfs_fattr *fattr)
4555 {
4556         struct nfs4_exception exception = {
4557                 .interruptible = true,
4558         };
4559         int err;
4560         do {
4561                 err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4562                 trace_nfs4_lookupp(inode, err);
4563                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4564                                 &exception);
4565         } while (exception.retry);
4566         return err;
4567 }
4568
4569 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4570                              const struct cred *cred)
4571 {
4572         struct nfs_server *server = NFS_SERVER(inode);
4573         struct nfs4_accessargs args = {
4574                 .fh = NFS_FH(inode),
4575                 .access = entry->mask,
4576         };
4577         struct nfs4_accessres res = {
4578                 .server = server,
4579         };
4580         struct rpc_message msg = {
4581                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4582                 .rpc_argp = &args,
4583                 .rpc_resp = &res,
4584                 .rpc_cred = cred,
4585         };
4586         int status = 0;
4587
4588         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4589                 res.fattr = nfs_alloc_fattr();
4590                 if (res.fattr == NULL)
4591                         return -ENOMEM;
4592                 args.bitmask = server->cache_consistency_bitmask;
4593         }
4594         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4595         if (!status) {
4596                 nfs_access_set_mask(entry, res.access);
4597                 if (res.fattr)
4598                         nfs_refresh_inode(inode, res.fattr);
4599         }
4600         nfs_free_fattr(res.fattr);
4601         return status;
4602 }
4603
4604 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4605                             const struct cred *cred)
4606 {
4607         struct nfs4_exception exception = {
4608                 .interruptible = true,
4609         };
4610         int err;
4611         do {
4612                 err = _nfs4_proc_access(inode, entry, cred);
4613                 trace_nfs4_access(inode, err);
4614                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4615                                 &exception);
4616         } while (exception.retry);
4617         return err;
4618 }
4619
4620 /*
4621  * TODO: For the time being, we don't try to get any attributes
4622  * along with any of the zero-copy operations READ, READDIR,
4623  * READLINK, WRITE.
4624  *
4625  * In the case of the first three, we want to put the GETATTR
4626  * after the read-type operation -- this is because it is hard
4627  * to predict the length of a GETATTR response in v4, and thus
4628  * align the READ data correctly.  This means that the GETATTR
4629  * may end up partially falling into the page cache, and we should
4630  * shift it into the 'tail' of the xdr_buf before processing.
4631  * To do this efficiently, we need to know the total length
4632  * of data received, which doesn't seem to be available outside
4633  * of the RPC layer.
4634  *
4635  * In the case of WRITE, we also want to put the GETATTR after
4636  * the operation -- in this case because we want to make sure
4637  * we get the post-operation mtime and size.
4638  *
4639  * Both of these changes to the XDR layer would in fact be quite
4640  * minor, but I decided to leave them for a subsequent patch.
4641  */
4642 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4643                 unsigned int pgbase, unsigned int pglen)
4644 {
4645         struct nfs4_readlink args = {
4646                 .fh       = NFS_FH(inode),
4647                 .pgbase   = pgbase,
4648                 .pglen    = pglen,
4649                 .pages    = &page,
4650         };
4651         struct nfs4_readlink_res res;
4652         struct rpc_message msg = {
4653                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4654                 .rpc_argp = &args,
4655                 .rpc_resp = &res,
4656         };
4657
4658         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4659 }
4660
4661 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4662                 unsigned int pgbase, unsigned int pglen)
4663 {
4664         struct nfs4_exception exception = {
4665                 .interruptible = true,
4666         };
4667         int err;
4668         do {
4669                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4670                 trace_nfs4_readlink(inode, err);
4671                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4672                                 &exception);
4673         } while (exception.retry);
4674         return err;
4675 }
4676
4677 /*
4678  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4679  */
4680 static int
4681 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4682                  int flags)
4683 {
4684         struct nfs_server *server = NFS_SERVER(dir);
4685         struct nfs4_label l, *ilabel = NULL;
4686         struct nfs_open_context *ctx;
4687         struct nfs4_state *state;
4688         int status = 0;
4689
4690         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4691         if (IS_ERR(ctx))
4692                 return PTR_ERR(ctx);
4693
4694         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4695
4696         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4697                 sattr->ia_mode &= ~current_umask();
4698         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4699         if (IS_ERR(state)) {
4700                 status = PTR_ERR(state);
4701                 goto out;
4702         }
4703 out:
4704         nfs4_label_release_security(ilabel);
4705         put_nfs_open_context(ctx);
4706         return status;
4707 }
4708
4709 static int
4710 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4711 {
4712         struct nfs_server *server = NFS_SERVER(dir);
4713         struct nfs_removeargs args = {
4714                 .fh = NFS_FH(dir),
4715                 .name = *name,
4716         };
4717         struct nfs_removeres res = {
4718                 .server = server,
4719         };
4720         struct rpc_message msg = {
4721                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4722                 .rpc_argp = &args,
4723                 .rpc_resp = &res,
4724         };
4725         unsigned long timestamp = jiffies;
4726         int status;
4727
4728         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4729         if (status == 0) {
4730                 spin_lock(&dir->i_lock);
4731                 /* Removing a directory decrements nlink in the parent */
4732                 if (ftype == NF4DIR && dir->i_nlink > 2)
4733                         nfs4_dec_nlink_locked(dir);
4734                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4735                                               NFS_INO_INVALID_DATA);
4736                 spin_unlock(&dir->i_lock);
4737         }
4738         return status;
4739 }
4740
4741 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4742 {
4743         struct nfs4_exception exception = {
4744                 .interruptible = true,
4745         };
4746         struct inode *inode = d_inode(dentry);
4747         int err;
4748
4749         if (inode) {
4750                 if (inode->i_nlink == 1)
4751                         nfs4_inode_return_delegation(inode);
4752                 else
4753                         nfs4_inode_make_writeable(inode);
4754         }
4755         do {
4756                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4757                 trace_nfs4_remove(dir, &dentry->d_name, err);
4758                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4759                                 &exception);
4760         } while (exception.retry);
4761         return err;
4762 }
4763
4764 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4765 {
4766         struct nfs4_exception exception = {
4767                 .interruptible = true,
4768         };
4769         int err;
4770
4771         do {
4772                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4773                 trace_nfs4_remove(dir, name, err);
4774                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4775                                 &exception);
4776         } while (exception.retry);
4777         return err;
4778 }
4779
4780 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4781                 struct dentry *dentry,
4782                 struct inode *inode)
4783 {
4784         struct nfs_removeargs *args = msg->rpc_argp;
4785         struct nfs_removeres *res = msg->rpc_resp;
4786
4787         res->server = NFS_SB(dentry->d_sb);
4788         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4789         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4790
4791         nfs_fattr_init(res->dir_attr);
4792
4793         if (inode) {
4794                 nfs4_inode_return_delegation(inode);
4795                 nfs_d_prune_case_insensitive_aliases(inode);
4796         }
4797 }
4798
4799 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4800 {
4801         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4802                         &data->args.seq_args,
4803                         &data->res.seq_res,
4804                         task);
4805 }
4806
4807 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4808 {
4809         struct nfs_unlinkdata *data = task->tk_calldata;
4810         struct nfs_removeres *res = &data->res;
4811
4812         if (!nfs4_sequence_done(task, &res->seq_res))
4813                 return 0;
4814         if (nfs4_async_handle_error(task, res->server, NULL,
4815                                     &data->timeout) == -EAGAIN)
4816                 return 0;
4817         if (task->tk_status == 0)
4818                 nfs4_update_changeattr(dir, &res->cinfo,
4819                                 res->dir_attr->time_start,
4820                                 NFS_INO_INVALID_DATA);
4821         return 1;
4822 }
4823
4824 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4825                 struct dentry *old_dentry,
4826                 struct dentry *new_dentry)
4827 {
4828         struct nfs_renameargs *arg = msg->rpc_argp;
4829         struct nfs_renameres *res = msg->rpc_resp;
4830         struct inode *old_inode = d_inode(old_dentry);
4831         struct inode *new_inode = d_inode(new_dentry);
4832
4833         if (old_inode)
4834                 nfs4_inode_make_writeable(old_inode);
4835         if (new_inode)
4836                 nfs4_inode_return_delegation(new_inode);
4837         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4838         res->server = NFS_SB(old_dentry->d_sb);
4839         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4840 }
4841
4842 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4843 {
4844         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4845                         &data->args.seq_args,
4846                         &data->res.seq_res,
4847                         task);
4848 }
4849
4850 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4851                                  struct inode *new_dir)
4852 {
4853         struct nfs_renamedata *data = task->tk_calldata;
4854         struct nfs_renameres *res = &data->res;
4855
4856         if (!nfs4_sequence_done(task, &res->seq_res))
4857                 return 0;
4858         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4859                 return 0;
4860
4861         if (task->tk_status == 0) {
4862                 nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4863                 if (new_dir != old_dir) {
4864                         /* Note: If we moved a directory, nlink will change */
4865                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4866                                         res->old_fattr->time_start,
4867                                         NFS_INO_INVALID_NLINK |
4868                                             NFS_INO_INVALID_DATA);
4869                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4870                                         res->new_fattr->time_start,
4871                                         NFS_INO_INVALID_NLINK |
4872                                             NFS_INO_INVALID_DATA);
4873                 } else
4874                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4875                                         res->old_fattr->time_start,
4876                                         NFS_INO_INVALID_DATA);
4877         }
4878         return 1;
4879 }
4880
4881 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4882 {
4883         struct nfs_server *server = NFS_SERVER(inode);
4884         __u32 bitmask[NFS4_BITMASK_SZ];
4885         struct nfs4_link_arg arg = {
4886                 .fh     = NFS_FH(inode),
4887                 .dir_fh = NFS_FH(dir),
4888                 .name   = name,
4889                 .bitmask = bitmask,
4890         };
4891         struct nfs4_link_res res = {
4892                 .server = server,
4893         };
4894         struct rpc_message msg = {
4895                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4896                 .rpc_argp = &arg,
4897                 .rpc_resp = &res,
4898         };
4899         int status = -ENOMEM;
4900
4901         res.fattr = nfs_alloc_fattr_with_label(server);
4902         if (res.fattr == NULL)
4903                 goto out;
4904
4905         nfs4_inode_make_writeable(inode);
4906         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4907                                 NFS_INO_INVALID_CHANGE);
4908         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4909         if (!status) {
4910                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4911                                        NFS_INO_INVALID_DATA);
4912                 nfs4_inc_nlink(inode);
4913                 status = nfs_post_op_update_inode(inode, res.fattr);
4914                 if (!status)
4915                         nfs_setsecurity(inode, res.fattr);
4916         }
4917
4918 out:
4919         nfs_free_fattr(res.fattr);
4920         return status;
4921 }
4922
4923 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4924 {
4925         struct nfs4_exception exception = {
4926                 .interruptible = true,
4927         };
4928         int err;
4929         do {
4930                 err = nfs4_handle_exception(NFS_SERVER(inode),
4931                                 _nfs4_proc_link(inode, dir, name),
4932                                 &exception);
4933         } while (exception.retry);
4934         return err;
4935 }
4936
4937 struct nfs4_createdata {
4938         struct rpc_message msg;
4939         struct nfs4_create_arg arg;
4940         struct nfs4_create_res res;
4941         struct nfs_fh fh;
4942         struct nfs_fattr fattr;
4943 };
4944
4945 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4946                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4947 {
4948         struct nfs4_createdata *data;
4949
4950         data = kzalloc(sizeof(*data), GFP_KERNEL);
4951         if (data != NULL) {
4952                 struct nfs_server *server = NFS_SERVER(dir);
4953
4954                 data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4955                 if (IS_ERR(data->fattr.label))
4956                         goto out_free;
4957
4958                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4959                 data->msg.rpc_argp = &data->arg;
4960                 data->msg.rpc_resp = &data->res;
4961                 data->arg.dir_fh = NFS_FH(dir);
4962                 data->arg.server = server;
4963                 data->arg.name = name;
4964                 data->arg.attrs = sattr;
4965                 data->arg.ftype = ftype;
4966                 data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4967                 data->arg.umask = current_umask();
4968                 data->res.server = server;
4969                 data->res.fh = &data->fh;
4970                 data->res.fattr = &data->fattr;
4971                 nfs_fattr_init(data->res.fattr);
4972         }
4973         return data;
4974 out_free:
4975         kfree(data);
4976         return NULL;
4977 }
4978
4979 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4980 {
4981         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4982                                     &data->arg.seq_args, &data->res.seq_res, 1);
4983         if (status == 0) {
4984                 spin_lock(&dir->i_lock);
4985                 /* Creating a directory bumps nlink in the parent */
4986                 if (data->arg.ftype == NF4DIR)
4987                         nfs4_inc_nlink_locked(dir);
4988                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4989                                               data->res.fattr->time_start,
4990                                               NFS_INO_INVALID_DATA);
4991                 spin_unlock(&dir->i_lock);
4992                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
4993         }
4994         return status;
4995 }
4996
4997 static void nfs4_free_createdata(struct nfs4_createdata *data)
4998 {
4999         nfs4_label_free(data->fattr.label);
5000         kfree(data);
5001 }
5002
5003 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5004                 struct page *page, unsigned int len, struct iattr *sattr,
5005                 struct nfs4_label *label)
5006 {
5007         struct nfs4_createdata *data;
5008         int status = -ENAMETOOLONG;
5009
5010         if (len > NFS4_MAXPATHLEN)
5011                 goto out;
5012
5013         status = -ENOMEM;
5014         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5015         if (data == NULL)
5016                 goto out;
5017
5018         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5019         data->arg.u.symlink.pages = &page;
5020         data->arg.u.symlink.len = len;
5021         data->arg.label = label;
5022         
5023         status = nfs4_do_create(dir, dentry, data);
5024
5025         nfs4_free_createdata(data);
5026 out:
5027         return status;
5028 }
5029
5030 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5031                 struct page *page, unsigned int len, struct iattr *sattr)
5032 {
5033         struct nfs4_exception exception = {
5034                 .interruptible = true,
5035         };
5036         struct nfs4_label l, *label = NULL;
5037         int err;
5038
5039         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5040
5041         do {
5042                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
5043                 trace_nfs4_symlink(dir, &dentry->d_name, err);
5044                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5045                                 &exception);
5046         } while (exception.retry);
5047
5048         nfs4_label_release_security(label);
5049         return err;
5050 }
5051
5052 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5053                 struct iattr *sattr, struct nfs4_label *label)
5054 {
5055         struct nfs4_createdata *data;
5056         int status = -ENOMEM;
5057
5058         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5059         if (data == NULL)
5060                 goto out;
5061
5062         data->arg.label = label;
5063         status = nfs4_do_create(dir, dentry, data);
5064
5065         nfs4_free_createdata(data);
5066 out:
5067         return status;
5068 }
5069
5070 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5071                 struct iattr *sattr)
5072 {
5073         struct nfs_server *server = NFS_SERVER(dir);
5074         struct nfs4_exception exception = {
5075                 .interruptible = true,
5076         };
5077         struct nfs4_label l, *label = NULL;
5078         int err;
5079
5080         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5081
5082         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5083                 sattr->ia_mode &= ~current_umask();
5084         do {
5085                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5086                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
5087                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5088                                 &exception);
5089         } while (exception.retry);
5090         nfs4_label_release_security(label);
5091
5092         return err;
5093 }
5094
5095 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5096                               struct nfs_readdir_res *nr_res)
5097 {
5098         struct inode            *dir = d_inode(nr_arg->dentry);
5099         struct nfs_server       *server = NFS_SERVER(dir);
5100         struct nfs4_readdir_arg args = {
5101                 .fh = NFS_FH(dir),
5102                 .pages = nr_arg->pages,
5103                 .pgbase = 0,
5104                 .count = nr_arg->page_len,
5105                 .plus = nr_arg->plus,
5106         };
5107         struct nfs4_readdir_res res;
5108         struct rpc_message msg = {
5109                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5110                 .rpc_argp = &args,
5111                 .rpc_resp = &res,
5112                 .rpc_cred = nr_arg->cred,
5113         };
5114         int                     status;
5115
5116         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5117                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5118         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5119                 args.bitmask = server->attr_bitmask_nl;
5120         else
5121                 args.bitmask = server->attr_bitmask;
5122
5123         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5124         res.pgbase = args.pgbase;
5125         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5126                         &res.seq_res, 0);
5127         if (status >= 0) {
5128                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5129                 status += args.pgbase;
5130         }
5131
5132         nfs_invalidate_atime(dir);
5133
5134         dprintk("%s: returns %d\n", __func__, status);
5135         return status;
5136 }
5137
5138 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5139                              struct nfs_readdir_res *res)
5140 {
5141         struct nfs4_exception exception = {
5142                 .interruptible = true,
5143         };
5144         int err;
5145         do {
5146                 err = _nfs4_proc_readdir(arg, res);
5147                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5148                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5149                                             err, &exception);
5150         } while (exception.retry);
5151         return err;
5152 }
5153
5154 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5155                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5156 {
5157         struct nfs4_createdata *data;
5158         int mode = sattr->ia_mode;
5159         int status = -ENOMEM;
5160
5161         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5162         if (data == NULL)
5163                 goto out;
5164
5165         if (S_ISFIFO(mode))
5166                 data->arg.ftype = NF4FIFO;
5167         else if (S_ISBLK(mode)) {
5168                 data->arg.ftype = NF4BLK;
5169                 data->arg.u.device.specdata1 = MAJOR(rdev);
5170                 data->arg.u.device.specdata2 = MINOR(rdev);
5171         }
5172         else if (S_ISCHR(mode)) {
5173                 data->arg.ftype = NF4CHR;
5174                 data->arg.u.device.specdata1 = MAJOR(rdev);
5175                 data->arg.u.device.specdata2 = MINOR(rdev);
5176         } else if (!S_ISSOCK(mode)) {
5177                 status = -EINVAL;
5178                 goto out_free;
5179         }
5180
5181         data->arg.label = label;
5182         status = nfs4_do_create(dir, dentry, data);
5183 out_free:
5184         nfs4_free_createdata(data);
5185 out:
5186         return status;
5187 }
5188
5189 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5190                 struct iattr *sattr, dev_t rdev)
5191 {
5192         struct nfs_server *server = NFS_SERVER(dir);
5193         struct nfs4_exception exception = {
5194                 .interruptible = true,
5195         };
5196         struct nfs4_label l, *label = NULL;
5197         int err;
5198
5199         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5200
5201         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5202                 sattr->ia_mode &= ~current_umask();
5203         do {
5204                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5205                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5206                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5207                                 &exception);
5208         } while (exception.retry);
5209
5210         nfs4_label_release_security(label);
5211
5212         return err;
5213 }
5214
5215 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5216                  struct nfs_fsstat *fsstat)
5217 {
5218         struct nfs4_statfs_arg args = {
5219                 .fh = fhandle,
5220                 .bitmask = server->attr_bitmask,
5221         };
5222         struct nfs4_statfs_res res = {
5223                 .fsstat = fsstat,
5224         };
5225         struct rpc_message msg = {
5226                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5227                 .rpc_argp = &args,
5228                 .rpc_resp = &res,
5229         };
5230
5231         nfs_fattr_init(fsstat->fattr);
5232         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5233 }
5234
5235 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5236 {
5237         struct nfs4_exception exception = {
5238                 .interruptible = true,
5239         };
5240         int err;
5241         do {
5242                 err = nfs4_handle_exception(server,
5243                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5244                                 &exception);
5245         } while (exception.retry);
5246         return err;
5247 }
5248
5249 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5250                 struct nfs_fsinfo *fsinfo)
5251 {
5252         struct nfs4_fsinfo_arg args = {
5253                 .fh = fhandle,
5254                 .bitmask = server->attr_bitmask,
5255         };
5256         struct nfs4_fsinfo_res res = {
5257                 .fsinfo = fsinfo,
5258         };
5259         struct rpc_message msg = {
5260                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5261                 .rpc_argp = &args,
5262                 .rpc_resp = &res,
5263         };
5264
5265         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5266 }
5267
5268 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5269 {
5270         struct nfs4_exception exception = {
5271                 .interruptible = true,
5272         };
5273         int err;
5274
5275         do {
5276                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5277                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5278                 if (err == 0) {
5279                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5280                         break;
5281                 }
5282                 err = nfs4_handle_exception(server, err, &exception);
5283         } while (exception.retry);
5284         return err;
5285 }
5286
5287 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5288 {
5289         int error;
5290
5291         nfs_fattr_init(fsinfo->fattr);
5292         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5293         if (error == 0) {
5294                 /* block layout checks this! */
5295                 server->pnfs_blksize = fsinfo->blksize;
5296                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5297         }
5298
5299         return error;
5300 }
5301
5302 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5303                 struct nfs_pathconf *pathconf)
5304 {
5305         struct nfs4_pathconf_arg args = {
5306                 .fh = fhandle,
5307                 .bitmask = server->attr_bitmask,
5308         };
5309         struct nfs4_pathconf_res res = {
5310                 .pathconf = pathconf,
5311         };
5312         struct rpc_message msg = {
5313                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5314                 .rpc_argp = &args,
5315                 .rpc_resp = &res,
5316         };
5317
5318         /* None of the pathconf attributes are mandatory to implement */
5319         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5320                 memset(pathconf, 0, sizeof(*pathconf));
5321                 return 0;
5322         }
5323
5324         nfs_fattr_init(pathconf->fattr);
5325         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5326 }
5327
5328 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5329                 struct nfs_pathconf *pathconf)
5330 {
5331         struct nfs4_exception exception = {
5332                 .interruptible = true,
5333         };
5334         int err;
5335
5336         do {
5337                 err = nfs4_handle_exception(server,
5338                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5339                                 &exception);
5340         } while (exception.retry);
5341         return err;
5342 }
5343
5344 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5345                 const struct nfs_open_context *ctx,
5346                 const struct nfs_lock_context *l_ctx,
5347                 fmode_t fmode)
5348 {
5349         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5350 }
5351 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5352
5353 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5354                 const struct nfs_open_context *ctx,
5355                 const struct nfs_lock_context *l_ctx,
5356                 fmode_t fmode)
5357 {
5358         nfs4_stateid _current_stateid;
5359
5360         /* If the current stateid represents a lost lock, then exit */
5361         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5362                 return true;
5363         return nfs4_stateid_match(stateid, &_current_stateid);
5364 }
5365
5366 static bool nfs4_error_stateid_expired(int err)
5367 {
5368         switch (err) {
5369         case -NFS4ERR_DELEG_REVOKED:
5370         case -NFS4ERR_ADMIN_REVOKED:
5371         case -NFS4ERR_BAD_STATEID:
5372         case -NFS4ERR_STALE_STATEID:
5373         case -NFS4ERR_OLD_STATEID:
5374         case -NFS4ERR_OPENMODE:
5375         case -NFS4ERR_EXPIRED:
5376                 return true;
5377         }
5378         return false;
5379 }
5380
5381 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5382 {
5383         struct nfs_server *server = NFS_SERVER(hdr->inode);
5384
5385         trace_nfs4_read(hdr, task->tk_status);
5386         if (task->tk_status < 0) {
5387                 struct nfs4_exception exception = {
5388                         .inode = hdr->inode,
5389                         .state = hdr->args.context->state,
5390                         .stateid = &hdr->args.stateid,
5391                 };
5392                 task->tk_status = nfs4_async_handle_exception(task,
5393                                 server, task->tk_status, &exception);
5394                 if (exception.retry) {
5395                         rpc_restart_call_prepare(task);
5396                         return -EAGAIN;
5397                 }
5398         }
5399
5400         if (task->tk_status > 0)
5401                 renew_lease(server, hdr->timestamp);
5402         return 0;
5403 }
5404
5405 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5406                 struct nfs_pgio_args *args)
5407 {
5408
5409         if (!nfs4_error_stateid_expired(task->tk_status) ||
5410                 nfs4_stateid_is_current(&args->stateid,
5411                                 args->context,
5412                                 args->lock_context,
5413                                 FMODE_READ))
5414                 return false;
5415         rpc_restart_call_prepare(task);
5416         return true;
5417 }
5418
5419 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5420                                          struct nfs_pgio_header *hdr)
5421 {
5422         struct nfs_server *server = NFS_SERVER(hdr->inode);
5423         struct rpc_message *msg = &task->tk_msg;
5424
5425         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5426             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5427                 server->caps &= ~NFS_CAP_READ_PLUS;
5428                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5429                 rpc_restart_call_prepare(task);
5430                 return true;
5431         }
5432         return false;
5433 }
5434
5435 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5436 {
5437         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5438                 return -EAGAIN;
5439         if (nfs4_read_stateid_changed(task, &hdr->args))
5440                 return -EAGAIN;
5441         if (nfs4_read_plus_not_supported(task, hdr))
5442                 return -EAGAIN;
5443         if (task->tk_status > 0)
5444                 nfs_invalidate_atime(hdr->inode);
5445         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5446                                     nfs4_read_done_cb(task, hdr);
5447 }
5448
5449 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5450 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5451                                     struct rpc_message *msg)
5452 {
5453         /* Note: We don't use READ_PLUS with pNFS yet */
5454         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5455                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5456 }
5457 #else
5458 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5459                                     struct rpc_message *msg)
5460 {
5461 }
5462 #endif /* CONFIG_NFS_V4_2 */
5463
5464 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5465                                  struct rpc_message *msg)
5466 {
5467         hdr->timestamp   = jiffies;
5468         if (!hdr->pgio_done_cb)
5469                 hdr->pgio_done_cb = nfs4_read_done_cb;
5470         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5471         nfs42_read_plus_support(hdr, msg);
5472         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5473 }
5474
5475 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5476                                       struct nfs_pgio_header *hdr)
5477 {
5478         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5479                         &hdr->args.seq_args,
5480                         &hdr->res.seq_res,
5481                         task))
5482                 return 0;
5483         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5484                                 hdr->args.lock_context,
5485                                 hdr->rw_mode) == -EIO)
5486                 return -EIO;
5487         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5488                 return -EIO;
5489         return 0;
5490 }
5491
5492 static int nfs4_write_done_cb(struct rpc_task *task,
5493                               struct nfs_pgio_header *hdr)
5494 {
5495         struct inode *inode = hdr->inode;
5496
5497         trace_nfs4_write(hdr, task->tk_status);
5498         if (task->tk_status < 0) {
5499                 struct nfs4_exception exception = {
5500                         .inode = hdr->inode,
5501                         .state = hdr->args.context->state,
5502                         .stateid = &hdr->args.stateid,
5503                 };
5504                 task->tk_status = nfs4_async_handle_exception(task,
5505                                 NFS_SERVER(inode), task->tk_status,
5506                                 &exception);
5507                 if (exception.retry) {
5508                         rpc_restart_call_prepare(task);
5509                         return -EAGAIN;
5510                 }
5511         }
5512         if (task->tk_status >= 0) {
5513                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5514                 nfs_writeback_update_inode(hdr);
5515         }
5516         return 0;
5517 }
5518
5519 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5520                 struct nfs_pgio_args *args)
5521 {
5522
5523         if (!nfs4_error_stateid_expired(task->tk_status) ||
5524                 nfs4_stateid_is_current(&args->stateid,
5525                                 args->context,
5526                                 args->lock_context,
5527                                 FMODE_WRITE))
5528                 return false;
5529         rpc_restart_call_prepare(task);
5530         return true;
5531 }
5532
5533 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5534 {
5535         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5536                 return -EAGAIN;
5537         if (nfs4_write_stateid_changed(task, &hdr->args))
5538                 return -EAGAIN;
5539         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5540                 nfs4_write_done_cb(task, hdr);
5541 }
5542
5543 static
5544 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5545 {
5546         /* Don't request attributes for pNFS or O_DIRECT writes */
5547         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5548                 return false;
5549         /* Otherwise, request attributes if and only if we don't hold
5550          * a delegation
5551          */
5552         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5553 }
5554
5555 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5556                       struct inode *inode, unsigned long cache_validity)
5557 {
5558         struct nfs_server *server = NFS_SERVER(inode);
5559         unsigned int i;
5560
5561         memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5562         cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5563
5564         if (cache_validity & NFS_INO_INVALID_CHANGE)
5565                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5566         if (cache_validity & NFS_INO_INVALID_ATIME)
5567                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5568         if (cache_validity & NFS_INO_INVALID_MODE)
5569                 bitmask[1] |= FATTR4_WORD1_MODE;
5570         if (cache_validity & NFS_INO_INVALID_OTHER)
5571                 bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5572         if (cache_validity & NFS_INO_INVALID_NLINK)
5573                 bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5574         if (cache_validity & NFS_INO_INVALID_CTIME)
5575                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5576         if (cache_validity & NFS_INO_INVALID_MTIME)
5577                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5578         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5579                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5580
5581         if (cache_validity & NFS_INO_INVALID_SIZE)
5582                 bitmask[0] |= FATTR4_WORD0_SIZE;
5583
5584         for (i = 0; i < NFS4_BITMASK_SZ; i++)
5585                 bitmask[i] &= server->attr_bitmask[i];
5586 }
5587
5588 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5589                                   struct rpc_message *msg,
5590                                   struct rpc_clnt **clnt)
5591 {
5592         struct nfs_server *server = NFS_SERVER(hdr->inode);
5593
5594         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5595                 hdr->args.bitmask = NULL;
5596                 hdr->res.fattr = NULL;
5597         } else {
5598                 nfs4_bitmask_set(hdr->args.bitmask_store,
5599                                  server->cache_consistency_bitmask,
5600                                  hdr->inode, NFS_INO_INVALID_BLOCKS);
5601                 hdr->args.bitmask = hdr->args.bitmask_store;
5602         }
5603
5604         if (!hdr->pgio_done_cb)
5605                 hdr->pgio_done_cb = nfs4_write_done_cb;
5606         hdr->res.server = server;
5607         hdr->timestamp   = jiffies;
5608
5609         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5610         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5611         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5612 }
5613
5614 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5615 {
5616         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5617                         &data->args.seq_args,
5618                         &data->res.seq_res,
5619                         task);
5620 }
5621
5622 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5623 {
5624         struct inode *inode = data->inode;
5625
5626         trace_nfs4_commit(data, task->tk_status);
5627         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5628                                     NULL, NULL) == -EAGAIN) {
5629                 rpc_restart_call_prepare(task);
5630                 return -EAGAIN;
5631         }
5632         return 0;
5633 }
5634
5635 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5636 {
5637         if (!nfs4_sequence_done(task, &data->res.seq_res))
5638                 return -EAGAIN;
5639         return data->commit_done_cb(task, data);
5640 }
5641
5642 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5643                                    struct rpc_clnt **clnt)
5644 {
5645         struct nfs_server *server = NFS_SERVER(data->inode);
5646
5647         if (data->commit_done_cb == NULL)
5648                 data->commit_done_cb = nfs4_commit_done_cb;
5649         data->res.server = server;
5650         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5651         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5652         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5653 }
5654
5655 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5656                                 struct nfs_commitres *res)
5657 {
5658         struct inode *dst_inode = file_inode(dst);
5659         struct nfs_server *server = NFS_SERVER(dst_inode);
5660         struct rpc_message msg = {
5661                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5662                 .rpc_argp = args,
5663                 .rpc_resp = res,
5664         };
5665
5666         args->fh = NFS_FH(dst_inode);
5667         return nfs4_call_sync(server->client, server, &msg,
5668                         &args->seq_args, &res->seq_res, 1);
5669 }
5670
5671 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5672 {
5673         struct nfs_commitargs args = {
5674                 .offset = offset,
5675                 .count = count,
5676         };
5677         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5678         struct nfs4_exception exception = { };
5679         int status;
5680
5681         do {
5682                 status = _nfs4_proc_commit(dst, &args, res);
5683                 status = nfs4_handle_exception(dst_server, status, &exception);
5684         } while (exception.retry);
5685
5686         return status;
5687 }
5688
5689 struct nfs4_renewdata {
5690         struct nfs_client       *client;
5691         unsigned long           timestamp;
5692 };
5693
5694 /*
5695  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5696  * standalone procedure for queueing an asynchronous RENEW.
5697  */
5698 static void nfs4_renew_release(void *calldata)
5699 {
5700         struct nfs4_renewdata *data = calldata;
5701         struct nfs_client *clp = data->client;
5702
5703         if (refcount_read(&clp->cl_count) > 1)
5704                 nfs4_schedule_state_renewal(clp);
5705         nfs_put_client(clp);
5706         kfree(data);
5707 }
5708
5709 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5710 {
5711         struct nfs4_renewdata *data = calldata;
5712         struct nfs_client *clp = data->client;
5713         unsigned long timestamp = data->timestamp;
5714
5715         trace_nfs4_renew_async(clp, task->tk_status);
5716         switch (task->tk_status) {
5717         case 0:
5718                 break;
5719         case -NFS4ERR_LEASE_MOVED:
5720                 nfs4_schedule_lease_moved_recovery(clp);
5721                 break;
5722         default:
5723                 /* Unless we're shutting down, schedule state recovery! */
5724                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5725                         return;
5726                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5727                         nfs4_schedule_lease_recovery(clp);
5728                         return;
5729                 }
5730                 nfs4_schedule_path_down_recovery(clp);
5731         }
5732         do_renew_lease(clp, timestamp);
5733 }
5734
5735 static const struct rpc_call_ops nfs4_renew_ops = {
5736         .rpc_call_done = nfs4_renew_done,
5737         .rpc_release = nfs4_renew_release,
5738 };
5739
5740 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5741 {
5742         struct rpc_message msg = {
5743                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5744                 .rpc_argp       = clp,
5745                 .rpc_cred       = cred,
5746         };
5747         struct nfs4_renewdata *data;
5748
5749         if (renew_flags == 0)
5750                 return 0;
5751         if (!refcount_inc_not_zero(&clp->cl_count))
5752                 return -EIO;
5753         data = kmalloc(sizeof(*data), GFP_NOFS);
5754         if (data == NULL) {
5755                 nfs_put_client(clp);
5756                 return -ENOMEM;
5757         }
5758         data->client = clp;
5759         data->timestamp = jiffies;
5760         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5761                         &nfs4_renew_ops, data);
5762 }
5763
5764 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5765 {
5766         struct rpc_message msg = {
5767                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5768                 .rpc_argp       = clp,
5769                 .rpc_cred       = cred,
5770         };
5771         unsigned long now = jiffies;
5772         int status;
5773
5774         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5775         if (status < 0)
5776                 return status;
5777         do_renew_lease(clp, now);
5778         return 0;
5779 }
5780
5781 static bool nfs4_server_supports_acls(const struct nfs_server *server,
5782                                       enum nfs4_acl_type type)
5783 {
5784         switch (type) {
5785         default:
5786                 return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5787         case NFS4ACL_DACL:
5788                 return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5789         case NFS4ACL_SACL:
5790                 return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5791         }
5792 }
5793
5794 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5795  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5796  * the stack.
5797  */
5798 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5799
5800 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5801                 struct page **pages)
5802 {
5803         struct page *newpage, **spages;
5804         int rc = 0;
5805         size_t len;
5806         spages = pages;
5807
5808         do {
5809                 len = min_t(size_t, PAGE_SIZE, buflen);
5810                 newpage = alloc_page(GFP_KERNEL);
5811
5812                 if (newpage == NULL)
5813                         goto unwind;
5814                 memcpy(page_address(newpage), buf, len);
5815                 buf += len;
5816                 buflen -= len;
5817                 *pages++ = newpage;
5818                 rc++;
5819         } while (buflen != 0);
5820
5821         return rc;
5822
5823 unwind:
5824         for(; rc > 0; rc--)
5825                 __free_page(spages[rc-1]);
5826         return -ENOMEM;
5827 }
5828
5829 struct nfs4_cached_acl {
5830         enum nfs4_acl_type type;
5831         int cached;
5832         size_t len;
5833         char data[];
5834 };
5835
5836 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5837 {
5838         struct nfs_inode *nfsi = NFS_I(inode);
5839
5840         spin_lock(&inode->i_lock);
5841         kfree(nfsi->nfs4_acl);
5842         nfsi->nfs4_acl = acl;
5843         spin_unlock(&inode->i_lock);
5844 }
5845
5846 static void nfs4_zap_acl_attr(struct inode *inode)
5847 {
5848         nfs4_set_cached_acl(inode, NULL);
5849 }
5850
5851 static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5852                                     size_t buflen, enum nfs4_acl_type type)
5853 {
5854         struct nfs_inode *nfsi = NFS_I(inode);
5855         struct nfs4_cached_acl *acl;
5856         int ret = -ENOENT;
5857
5858         spin_lock(&inode->i_lock);
5859         acl = nfsi->nfs4_acl;
5860         if (acl == NULL)
5861                 goto out;
5862         if (acl->type != type)
5863                 goto out;
5864         if (buf == NULL) /* user is just asking for length */
5865                 goto out_len;
5866         if (acl->cached == 0)
5867                 goto out;
5868         ret = -ERANGE; /* see getxattr(2) man page */
5869         if (acl->len > buflen)
5870                 goto out;
5871         memcpy(buf, acl->data, acl->len);
5872 out_len:
5873         ret = acl->len;
5874 out:
5875         spin_unlock(&inode->i_lock);
5876         return ret;
5877 }
5878
5879 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
5880                                   size_t pgbase, size_t acl_len,
5881                                   enum nfs4_acl_type type)
5882 {
5883         struct nfs4_cached_acl *acl;
5884         size_t buflen = sizeof(*acl) + acl_len;
5885
5886         if (buflen <= PAGE_SIZE) {
5887                 acl = kmalloc(buflen, GFP_KERNEL);
5888                 if (acl == NULL)
5889                         goto out;
5890                 acl->cached = 1;
5891                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5892         } else {
5893                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5894                 if (acl == NULL)
5895                         goto out;
5896                 acl->cached = 0;
5897         }
5898         acl->type = type;
5899         acl->len = acl_len;
5900 out:
5901         nfs4_set_cached_acl(inode, acl);
5902 }
5903
5904 /*
5905  * The getxattr API returns the required buffer length when called with a
5906  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5907  * the required buf.  On a NULL buf, we send a page of data to the server
5908  * guessing that the ACL request can be serviced by a page. If so, we cache
5909  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5910  * the cache. If not so, we throw away the page, and cache the required
5911  * length. The next getxattr call will then produce another round trip to
5912  * the server, this time with the input buf of the required size.
5913  */
5914 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
5915                                        size_t buflen, enum nfs4_acl_type type)
5916 {
5917         struct page **pages;
5918         struct nfs_getaclargs args = {
5919                 .fh = NFS_FH(inode),
5920                 .acl_type = type,
5921                 .acl_len = buflen,
5922         };
5923         struct nfs_getaclres res = {
5924                 .acl_type = type,
5925                 .acl_len = buflen,
5926         };
5927         struct rpc_message msg = {
5928                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5929                 .rpc_argp = &args,
5930                 .rpc_resp = &res,
5931         };
5932         unsigned int npages;
5933         int ret = -ENOMEM, i;
5934         struct nfs_server *server = NFS_SERVER(inode);
5935
5936         if (buflen == 0)
5937                 buflen = server->rsize;
5938
5939         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5940         pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
5941         if (!pages)
5942                 return -ENOMEM;
5943
5944         args.acl_pages = pages;
5945
5946         for (i = 0; i < npages; i++) {
5947                 pages[i] = alloc_page(GFP_KERNEL);
5948                 if (!pages[i])
5949                         goto out_free;
5950         }
5951
5952         /* for decoding across pages */
5953         res.acl_scratch = alloc_page(GFP_KERNEL);
5954         if (!res.acl_scratch)
5955                 goto out_free;
5956
5957         args.acl_len = npages * PAGE_SIZE;
5958
5959         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5960                 __func__, buf, buflen, npages, args.acl_len);
5961         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5962                              &msg, &args.seq_args, &res.seq_res, 0);
5963         if (ret)
5964                 goto out_free;
5965
5966         /* Handle the case where the passed-in buffer is too short */
5967         if (res.acl_flags & NFS4_ACL_TRUNC) {
5968                 /* Did the user only issue a request for the acl length? */
5969                 if (buf == NULL)
5970                         goto out_ok;
5971                 ret = -ERANGE;
5972                 goto out_free;
5973         }
5974         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
5975                               type);
5976         if (buf) {
5977                 if (res.acl_len > buflen) {
5978                         ret = -ERANGE;
5979                         goto out_free;
5980                 }
5981                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5982         }
5983 out_ok:
5984         ret = res.acl_len;
5985 out_free:
5986         for (i = 0; i < npages; i++)
5987                 if (pages[i])
5988                         __free_page(pages[i]);
5989         if (res.acl_scratch)
5990                 __free_page(res.acl_scratch);
5991         kfree(pages);
5992         return ret;
5993 }
5994
5995 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
5996                                      size_t buflen, enum nfs4_acl_type type)
5997 {
5998         struct nfs4_exception exception = {
5999                 .interruptible = true,
6000         };
6001         ssize_t ret;
6002         do {
6003                 ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
6004                 trace_nfs4_get_acl(inode, ret);
6005                 if (ret >= 0)
6006                         break;
6007                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6008         } while (exception.retry);
6009         return ret;
6010 }
6011
6012 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6013                                  enum nfs4_acl_type type)
6014 {
6015         struct nfs_server *server = NFS_SERVER(inode);
6016         int ret;
6017
6018         if (!nfs4_server_supports_acls(server, type))
6019                 return -EOPNOTSUPP;
6020         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6021         if (ret < 0)
6022                 return ret;
6023         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6024                 nfs_zap_acl_cache(inode);
6025         ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6026         if (ret != -ENOENT)
6027                 /* -ENOENT is returned if there is no ACL or if there is an ACL
6028                  * but no cached acl data, just the acl length */
6029                 return ret;
6030         return nfs4_get_acl_uncached(inode, buf, buflen, type);
6031 }
6032
6033 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6034                                size_t buflen, enum nfs4_acl_type type)
6035 {
6036         struct nfs_server *server = NFS_SERVER(inode);
6037         struct page *pages[NFS4ACL_MAXPAGES];
6038         struct nfs_setaclargs arg = {
6039                 .fh = NFS_FH(inode),
6040                 .acl_type = type,
6041                 .acl_len = buflen,
6042                 .acl_pages = pages,
6043         };
6044         struct nfs_setaclres res;
6045         struct rpc_message msg = {
6046                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6047                 .rpc_argp       = &arg,
6048                 .rpc_resp       = &res,
6049         };
6050         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6051         int ret, i;
6052
6053         /* You can't remove system.nfs4_acl: */
6054         if (buflen == 0)
6055                 return -EINVAL;
6056         if (!nfs4_server_supports_acls(server, type))
6057                 return -EOPNOTSUPP;
6058         if (npages > ARRAY_SIZE(pages))
6059                 return -ERANGE;
6060         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6061         if (i < 0)
6062                 return i;
6063         nfs4_inode_make_writeable(inode);
6064         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6065
6066         /*
6067          * Free each page after tx, so the only ref left is
6068          * held by the network stack
6069          */
6070         for (; i > 0; i--)
6071                 put_page(pages[i-1]);
6072
6073         /*
6074          * Acl update can result in inode attribute update.
6075          * so mark the attribute cache invalid.
6076          */
6077         spin_lock(&inode->i_lock);
6078         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6079                                              NFS_INO_INVALID_CTIME |
6080                                              NFS_INO_REVAL_FORCED);
6081         spin_unlock(&inode->i_lock);
6082         nfs_access_zap_cache(inode);
6083         nfs_zap_acl_cache(inode);
6084         return ret;
6085 }
6086
6087 static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6088                              size_t buflen, enum nfs4_acl_type type)
6089 {
6090         struct nfs4_exception exception = { };
6091         int err;
6092         do {
6093                 err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6094                 trace_nfs4_set_acl(inode, err);
6095                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6096                         /*
6097                          * no need to retry since the kernel
6098                          * isn't involved in encoding the ACEs.
6099                          */
6100                         err = -EINVAL;
6101                         break;
6102                 }
6103                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6104                                 &exception);
6105         } while (exception.retry);
6106         return err;
6107 }
6108
6109 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6110 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6111                                         size_t buflen)
6112 {
6113         struct nfs_server *server = NFS_SERVER(inode);
6114         struct nfs4_label label = {0, 0, buflen, buf};
6115
6116         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6117         struct nfs_fattr fattr = {
6118                 .label = &label,
6119         };
6120         struct nfs4_getattr_arg arg = {
6121                 .fh             = NFS_FH(inode),
6122                 .bitmask        = bitmask,
6123         };
6124         struct nfs4_getattr_res res = {
6125                 .fattr          = &fattr,
6126                 .server         = server,
6127         };
6128         struct rpc_message msg = {
6129                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6130                 .rpc_argp       = &arg,
6131                 .rpc_resp       = &res,
6132         };
6133         int ret;
6134
6135         nfs_fattr_init(&fattr);
6136
6137         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6138         if (ret)
6139                 return ret;
6140         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6141                 return -ENOENT;
6142         return label.len;
6143 }
6144
6145 static int nfs4_get_security_label(struct inode *inode, void *buf,
6146                                         size_t buflen)
6147 {
6148         struct nfs4_exception exception = {
6149                 .interruptible = true,
6150         };
6151         int err;
6152
6153         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6154                 return -EOPNOTSUPP;
6155
6156         do {
6157                 err = _nfs4_get_security_label(inode, buf, buflen);
6158                 trace_nfs4_get_security_label(inode, err);
6159                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6160                                 &exception);
6161         } while (exception.retry);
6162         return err;
6163 }
6164
6165 static int _nfs4_do_set_security_label(struct inode *inode,
6166                 struct nfs4_label *ilabel,
6167                 struct nfs_fattr *fattr)
6168 {
6169
6170         struct iattr sattr = {0};
6171         struct nfs_server *server = NFS_SERVER(inode);
6172         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6173         struct nfs_setattrargs arg = {
6174                 .fh             = NFS_FH(inode),
6175                 .iap            = &sattr,
6176                 .server         = server,
6177                 .bitmask        = bitmask,
6178                 .label          = ilabel,
6179         };
6180         struct nfs_setattrres res = {
6181                 .fattr          = fattr,
6182                 .server         = server,
6183         };
6184         struct rpc_message msg = {
6185                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6186                 .rpc_argp       = &arg,
6187                 .rpc_resp       = &res,
6188         };
6189         int status;
6190
6191         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6192
6193         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6194         if (status)
6195                 dprintk("%s failed: %d\n", __func__, status);
6196
6197         return status;
6198 }
6199
6200 static int nfs4_do_set_security_label(struct inode *inode,
6201                 struct nfs4_label *ilabel,
6202                 struct nfs_fattr *fattr)
6203 {
6204         struct nfs4_exception exception = { };
6205         int err;
6206
6207         do {
6208                 err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6209                 trace_nfs4_set_security_label(inode, err);
6210                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6211                                 &exception);
6212         } while (exception.retry);
6213         return err;
6214 }
6215
6216 static int
6217 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6218 {
6219         struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6220         struct nfs_fattr *fattr;
6221         int status;
6222
6223         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6224                 return -EOPNOTSUPP;
6225
6226         fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6227         if (fattr == NULL)
6228                 return -ENOMEM;
6229
6230         status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6231         if (status == 0)
6232                 nfs_setsecurity(inode, fattr);
6233
6234         return status;
6235 }
6236 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6237
6238
6239 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6240                                     nfs4_verifier *bootverf)
6241 {
6242         __be32 verf[2];
6243
6244         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6245                 /* An impossible timestamp guarantees this value
6246                  * will never match a generated boot time. */
6247                 verf[0] = cpu_to_be32(U32_MAX);
6248                 verf[1] = cpu_to_be32(U32_MAX);
6249         } else {
6250                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6251                 u64 ns = ktime_to_ns(nn->boot_time);
6252
6253                 verf[0] = cpu_to_be32(ns >> 32);
6254                 verf[1] = cpu_to_be32(ns);
6255         }
6256         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6257 }
6258
6259 static size_t
6260 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6261 {
6262         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6263         struct nfs_netns_client *nn_clp = nn->nfs_client;
6264         const char *id;
6265
6266         buf[0] = '\0';
6267
6268         if (nn_clp) {
6269                 rcu_read_lock();
6270                 id = rcu_dereference(nn_clp->identifier);
6271                 if (id)
6272                         strscpy(buf, id, buflen);
6273                 rcu_read_unlock();
6274         }
6275
6276         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6277                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6278
6279         return strlen(buf);
6280 }
6281
6282 static int
6283 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6284 {
6285         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6286         size_t buflen;
6287         size_t len;
6288         char *str;
6289
6290         if (clp->cl_owner_id != NULL)
6291                 return 0;
6292
6293         rcu_read_lock();
6294         len = 14 +
6295                 strlen(clp->cl_rpcclient->cl_nodename) +
6296                 1 +
6297                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6298                 1;
6299         rcu_read_unlock();
6300
6301         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6302         if (buflen)
6303                 len += buflen + 1;
6304
6305         if (len > NFS4_OPAQUE_LIMIT + 1)
6306                 return -EINVAL;
6307
6308         /*
6309          * Since this string is allocated at mount time, and held until the
6310          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6311          * about a memory-reclaim deadlock.
6312          */
6313         str = kmalloc(len, GFP_KERNEL);
6314         if (!str)
6315                 return -ENOMEM;
6316
6317         rcu_read_lock();
6318         if (buflen)
6319                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6320                           clp->cl_rpcclient->cl_nodename, buf,
6321                           rpc_peeraddr2str(clp->cl_rpcclient,
6322                                            RPC_DISPLAY_ADDR));
6323         else
6324                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6325                           clp->cl_rpcclient->cl_nodename,
6326                           rpc_peeraddr2str(clp->cl_rpcclient,
6327                                            RPC_DISPLAY_ADDR));
6328         rcu_read_unlock();
6329
6330         clp->cl_owner_id = str;
6331         return 0;
6332 }
6333
6334 static int
6335 nfs4_init_uniform_client_string(struct nfs_client *clp)
6336 {
6337         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6338         size_t buflen;
6339         size_t len;
6340         char *str;
6341
6342         if (clp->cl_owner_id != NULL)
6343                 return 0;
6344
6345         len = 10 + 10 + 1 + 10 + 1 +
6346                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6347
6348         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6349         if (buflen)
6350                 len += buflen + 1;
6351
6352         if (len > NFS4_OPAQUE_LIMIT + 1)
6353                 return -EINVAL;
6354
6355         /*
6356          * Since this string is allocated at mount time, and held until the
6357          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6358          * about a memory-reclaim deadlock.
6359          */
6360         str = kmalloc(len, GFP_KERNEL);
6361         if (!str)
6362                 return -ENOMEM;
6363
6364         if (buflen)
6365                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6366                           clp->rpc_ops->version, clp->cl_minorversion,
6367                           buf, clp->cl_rpcclient->cl_nodename);
6368         else
6369                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6370                           clp->rpc_ops->version, clp->cl_minorversion,
6371                           clp->cl_rpcclient->cl_nodename);
6372         clp->cl_owner_id = str;
6373         return 0;
6374 }
6375
6376 /*
6377  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6378  * services.  Advertise one based on the address family of the
6379  * clientaddr.
6380  */
6381 static unsigned int
6382 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6383 {
6384         if (strchr(clp->cl_ipaddr, ':') != NULL)
6385                 return scnprintf(buf, len, "tcp6");
6386         else
6387                 return scnprintf(buf, len, "tcp");
6388 }
6389
6390 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6391 {
6392         struct nfs4_setclientid *sc = calldata;
6393
6394         if (task->tk_status == 0)
6395                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6396 }
6397
6398 static const struct rpc_call_ops nfs4_setclientid_ops = {
6399         .rpc_call_done = nfs4_setclientid_done,
6400 };
6401
6402 /**
6403  * nfs4_proc_setclientid - Negotiate client ID
6404  * @clp: state data structure
6405  * @program: RPC program for NFSv4 callback service
6406  * @port: IP port number for NFS4 callback service
6407  * @cred: credential to use for this call
6408  * @res: where to place the result
6409  *
6410  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6411  */
6412 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6413                 unsigned short port, const struct cred *cred,
6414                 struct nfs4_setclientid_res *res)
6415 {
6416         nfs4_verifier sc_verifier;
6417         struct nfs4_setclientid setclientid = {
6418                 .sc_verifier = &sc_verifier,
6419                 .sc_prog = program,
6420                 .sc_clnt = clp,
6421         };
6422         struct rpc_message msg = {
6423                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6424                 .rpc_argp = &setclientid,
6425                 .rpc_resp = res,
6426                 .rpc_cred = cred,
6427         };
6428         struct rpc_task_setup task_setup_data = {
6429                 .rpc_client = clp->cl_rpcclient,
6430                 .rpc_message = &msg,
6431                 .callback_ops = &nfs4_setclientid_ops,
6432                 .callback_data = &setclientid,
6433                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6434         };
6435         unsigned long now = jiffies;
6436         int status;
6437
6438         /* nfs_client_id4 */
6439         nfs4_init_boot_verifier(clp, &sc_verifier);
6440
6441         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6442                 status = nfs4_init_uniform_client_string(clp);
6443         else
6444                 status = nfs4_init_nonuniform_client_string(clp);
6445
6446         if (status)
6447                 goto out;
6448
6449         /* cb_client4 */
6450         setclientid.sc_netid_len =
6451                                 nfs4_init_callback_netid(clp,
6452                                                 setclientid.sc_netid,
6453                                                 sizeof(setclientid.sc_netid));
6454         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6455                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6456                                 clp->cl_ipaddr, port >> 8, port & 255);
6457
6458         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6459                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6460                 clp->cl_owner_id);
6461
6462         status = nfs4_call_sync_custom(&task_setup_data);
6463         if (setclientid.sc_cred) {
6464                 kfree(clp->cl_acceptor);
6465                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6466                 put_rpccred(setclientid.sc_cred);
6467         }
6468
6469         if (status == 0)
6470                 do_renew_lease(clp, now);
6471 out:
6472         trace_nfs4_setclientid(clp, status);
6473         dprintk("NFS reply setclientid: %d\n", status);
6474         return status;
6475 }
6476
6477 /**
6478  * nfs4_proc_setclientid_confirm - Confirm client ID
6479  * @clp: state data structure
6480  * @arg: result of a previous SETCLIENTID
6481  * @cred: credential to use for this call
6482  *
6483  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6484  */
6485 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6486                 struct nfs4_setclientid_res *arg,
6487                 const struct cred *cred)
6488 {
6489         struct rpc_message msg = {
6490                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6491                 .rpc_argp = arg,
6492                 .rpc_cred = cred,
6493         };
6494         int status;
6495
6496         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6497                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6498                 clp->cl_clientid);
6499         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6500                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6501         trace_nfs4_setclientid_confirm(clp, status);
6502         dprintk("NFS reply setclientid_confirm: %d\n", status);
6503         return status;
6504 }
6505
6506 struct nfs4_delegreturndata {
6507         struct nfs4_delegreturnargs args;
6508         struct nfs4_delegreturnres res;
6509         struct nfs_fh fh;
6510         nfs4_stateid stateid;
6511         unsigned long timestamp;
6512         struct {
6513                 struct nfs4_layoutreturn_args arg;
6514                 struct nfs4_layoutreturn_res res;
6515                 struct nfs4_xdr_opaque_data ld_private;
6516                 u32 roc_barrier;
6517                 bool roc;
6518         } lr;
6519         struct nfs_fattr fattr;
6520         int rpc_status;
6521         struct inode *inode;
6522 };
6523
6524 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6525 {
6526         struct nfs4_delegreturndata *data = calldata;
6527         struct nfs4_exception exception = {
6528                 .inode = data->inode,
6529                 .stateid = &data->stateid,
6530                 .task_is_privileged = data->args.seq_args.sa_privileged,
6531         };
6532
6533         if (!nfs4_sequence_done(task, &data->res.seq_res))
6534                 return;
6535
6536         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6537
6538         /* Handle Layoutreturn errors */
6539         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6540                           &data->res.lr_ret) == -EAGAIN)
6541                 goto out_restart;
6542
6543         switch (task->tk_status) {
6544         case 0:
6545                 renew_lease(data->res.server, data->timestamp);
6546                 break;
6547         case -NFS4ERR_ADMIN_REVOKED:
6548         case -NFS4ERR_DELEG_REVOKED:
6549         case -NFS4ERR_EXPIRED:
6550                 nfs4_free_revoked_stateid(data->res.server,
6551                                 data->args.stateid,
6552                                 task->tk_msg.rpc_cred);
6553                 fallthrough;
6554         case -NFS4ERR_BAD_STATEID:
6555         case -NFS4ERR_STALE_STATEID:
6556         case -ETIMEDOUT:
6557                 task->tk_status = 0;
6558                 break;
6559         case -NFS4ERR_OLD_STATEID:
6560                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6561                         nfs4_stateid_seqid_inc(&data->stateid);
6562                 if (data->args.bitmask) {
6563                         data->args.bitmask = NULL;
6564                         data->res.fattr = NULL;
6565                 }
6566                 goto out_restart;
6567         case -NFS4ERR_ACCESS:
6568                 if (data->args.bitmask) {
6569                         data->args.bitmask = NULL;
6570                         data->res.fattr = NULL;
6571                         goto out_restart;
6572                 }
6573                 fallthrough;
6574         default:
6575                 task->tk_status = nfs4_async_handle_exception(task,
6576                                 data->res.server, task->tk_status,
6577                                 &exception);
6578                 if (exception.retry)
6579                         goto out_restart;
6580         }
6581         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6582         data->rpc_status = task->tk_status;
6583         return;
6584 out_restart:
6585         task->tk_status = 0;
6586         rpc_restart_call_prepare(task);
6587 }
6588
6589 static void nfs4_delegreturn_release(void *calldata)
6590 {
6591         struct nfs4_delegreturndata *data = calldata;
6592         struct inode *inode = data->inode;
6593
6594         if (data->lr.roc)
6595                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6596                                  data->res.lr_ret);
6597         if (inode) {
6598                 nfs4_fattr_set_prechange(&data->fattr,
6599                                          inode_peek_iversion_raw(inode));
6600                 nfs_refresh_inode(inode, &data->fattr);
6601                 nfs_iput_and_deactive(inode);
6602         }
6603         kfree(calldata);
6604 }
6605
6606 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6607 {
6608         struct nfs4_delegreturndata *d_data;
6609         struct pnfs_layout_hdr *lo;
6610
6611         d_data = data;
6612
6613         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6614                 nfs4_sequence_done(task, &d_data->res.seq_res);
6615                 return;
6616         }
6617
6618         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6619         if (lo && !pnfs_layout_is_valid(lo)) {
6620                 d_data->args.lr_args = NULL;
6621                 d_data->res.lr_res = NULL;
6622         }
6623
6624         nfs4_setup_sequence(d_data->res.server->nfs_client,
6625                         &d_data->args.seq_args,
6626                         &d_data->res.seq_res,
6627                         task);
6628 }
6629
6630 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6631         .rpc_call_prepare = nfs4_delegreturn_prepare,
6632         .rpc_call_done = nfs4_delegreturn_done,
6633         .rpc_release = nfs4_delegreturn_release,
6634 };
6635
6636 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6637 {
6638         struct nfs4_delegreturndata *data;
6639         struct nfs_server *server = NFS_SERVER(inode);
6640         struct rpc_task *task;
6641         struct rpc_message msg = {
6642                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6643                 .rpc_cred = cred,
6644         };
6645         struct rpc_task_setup task_setup_data = {
6646                 .rpc_client = server->client,
6647                 .rpc_message = &msg,
6648                 .callback_ops = &nfs4_delegreturn_ops,
6649                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6650         };
6651         int status = 0;
6652
6653         if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6654                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6655
6656         data = kzalloc(sizeof(*data), GFP_KERNEL);
6657         if (data == NULL)
6658                 return -ENOMEM;
6659
6660         nfs4_state_protect(server->nfs_client,
6661                         NFS_SP4_MACH_CRED_CLEANUP,
6662                         &task_setup_data.rpc_client, &msg);
6663
6664         data->args.fhandle = &data->fh;
6665         data->args.stateid = &data->stateid;
6666         nfs4_bitmask_set(data->args.bitmask_store,
6667                          server->cache_consistency_bitmask, inode, 0);
6668         data->args.bitmask = data->args.bitmask_store;
6669         nfs_copy_fh(&data->fh, NFS_FH(inode));
6670         nfs4_stateid_copy(&data->stateid, stateid);
6671         data->res.fattr = &data->fattr;
6672         data->res.server = server;
6673         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6674         data->lr.arg.ld_private = &data->lr.ld_private;
6675         nfs_fattr_init(data->res.fattr);
6676         data->timestamp = jiffies;
6677         data->rpc_status = 0;
6678         data->inode = nfs_igrab_and_active(inode);
6679         if (data->inode || issync) {
6680                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6681                                         cred);
6682                 if (data->lr.roc) {
6683                         data->args.lr_args = &data->lr.arg;
6684                         data->res.lr_res = &data->lr.res;
6685                 }
6686         }
6687
6688         if (!data->inode)
6689                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6690                                    1);
6691         else
6692                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6693                                    0);
6694         task_setup_data.callback_data = data;
6695         msg.rpc_argp = &data->args;
6696         msg.rpc_resp = &data->res;
6697         task = rpc_run_task(&task_setup_data);
6698         if (IS_ERR(task))
6699                 return PTR_ERR(task);
6700         if (!issync)
6701                 goto out;
6702         status = rpc_wait_for_completion_task(task);
6703         if (status != 0)
6704                 goto out;
6705         status = data->rpc_status;
6706 out:
6707         rpc_put_task(task);
6708         return status;
6709 }
6710
6711 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6712 {
6713         struct nfs_server *server = NFS_SERVER(inode);
6714         struct nfs4_exception exception = { };
6715         int err;
6716         do {
6717                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6718                 trace_nfs4_delegreturn(inode, stateid, err);
6719                 switch (err) {
6720                         case -NFS4ERR_STALE_STATEID:
6721                         case -NFS4ERR_EXPIRED:
6722                         case 0:
6723                                 return 0;
6724                 }
6725                 err = nfs4_handle_exception(server, err, &exception);
6726         } while (exception.retry);
6727         return err;
6728 }
6729
6730 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6731 {
6732         struct inode *inode = state->inode;
6733         struct nfs_server *server = NFS_SERVER(inode);
6734         struct nfs_client *clp = server->nfs_client;
6735         struct nfs_lockt_args arg = {
6736                 .fh = NFS_FH(inode),
6737                 .fl = request,
6738         };
6739         struct nfs_lockt_res res = {
6740                 .denied = request,
6741         };
6742         struct rpc_message msg = {
6743                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6744                 .rpc_argp       = &arg,
6745                 .rpc_resp       = &res,
6746                 .rpc_cred       = state->owner->so_cred,
6747         };
6748         struct nfs4_lock_state *lsp;
6749         int status;
6750
6751         arg.lock_owner.clientid = clp->cl_clientid;
6752         status = nfs4_set_lock_state(state, request);
6753         if (status != 0)
6754                 goto out;
6755         lsp = request->fl_u.nfs4_fl.owner;
6756         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6757         arg.lock_owner.s_dev = server->s_dev;
6758         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6759         switch (status) {
6760                 case 0:
6761                         request->fl_type = F_UNLCK;
6762                         break;
6763                 case -NFS4ERR_DENIED:
6764                         status = 0;
6765         }
6766         request->fl_ops->fl_release_private(request);
6767         request->fl_ops = NULL;
6768 out:
6769         return status;
6770 }
6771
6772 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6773 {
6774         struct nfs4_exception exception = {
6775                 .interruptible = true,
6776         };
6777         int err;
6778
6779         do {
6780                 err = _nfs4_proc_getlk(state, cmd, request);
6781                 trace_nfs4_get_lock(request, state, cmd, err);
6782                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6783                                 &exception);
6784         } while (exception.retry);
6785         return err;
6786 }
6787
6788 /*
6789  * Update the seqid of a lock stateid after receiving
6790  * NFS4ERR_OLD_STATEID
6791  */
6792 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6793                 struct nfs4_lock_state *lsp)
6794 {
6795         struct nfs4_state *state = lsp->ls_state;
6796         bool ret = false;
6797
6798         spin_lock(&state->state_lock);
6799         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6800                 goto out;
6801         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6802                 nfs4_stateid_seqid_inc(dst);
6803         else
6804                 dst->seqid = lsp->ls_stateid.seqid;
6805         ret = true;
6806 out:
6807         spin_unlock(&state->state_lock);
6808         return ret;
6809 }
6810
6811 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6812                 struct nfs4_lock_state *lsp)
6813 {
6814         struct nfs4_state *state = lsp->ls_state;
6815         bool ret;
6816
6817         spin_lock(&state->state_lock);
6818         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6819         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6820         spin_unlock(&state->state_lock);
6821         return ret;
6822 }
6823
6824 struct nfs4_unlockdata {
6825         struct nfs_locku_args arg;
6826         struct nfs_locku_res res;
6827         struct nfs4_lock_state *lsp;
6828         struct nfs_open_context *ctx;
6829         struct nfs_lock_context *l_ctx;
6830         struct file_lock fl;
6831         struct nfs_server *server;
6832         unsigned long timestamp;
6833 };
6834
6835 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6836                 struct nfs_open_context *ctx,
6837                 struct nfs4_lock_state *lsp,
6838                 struct nfs_seqid *seqid)
6839 {
6840         struct nfs4_unlockdata *p;
6841         struct nfs4_state *state = lsp->ls_state;
6842         struct inode *inode = state->inode;
6843
6844         p = kzalloc(sizeof(*p), GFP_KERNEL);
6845         if (p == NULL)
6846                 return NULL;
6847         p->arg.fh = NFS_FH(inode);
6848         p->arg.fl = &p->fl;
6849         p->arg.seqid = seqid;
6850         p->res.seqid = seqid;
6851         p->lsp = lsp;
6852         /* Ensure we don't close file until we're done freeing locks! */
6853         p->ctx = get_nfs_open_context(ctx);
6854         p->l_ctx = nfs_get_lock_context(ctx);
6855         locks_init_lock(&p->fl);
6856         locks_copy_lock(&p->fl, fl);
6857         p->server = NFS_SERVER(inode);
6858         spin_lock(&state->state_lock);
6859         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6860         spin_unlock(&state->state_lock);
6861         return p;
6862 }
6863
6864 static void nfs4_locku_release_calldata(void *data)
6865 {
6866         struct nfs4_unlockdata *calldata = data;
6867         nfs_free_seqid(calldata->arg.seqid);
6868         nfs4_put_lock_state(calldata->lsp);
6869         nfs_put_lock_context(calldata->l_ctx);
6870         put_nfs_open_context(calldata->ctx);
6871         kfree(calldata);
6872 }
6873
6874 static void nfs4_locku_done(struct rpc_task *task, void *data)
6875 {
6876         struct nfs4_unlockdata *calldata = data;
6877         struct nfs4_exception exception = {
6878                 .inode = calldata->lsp->ls_state->inode,
6879                 .stateid = &calldata->arg.stateid,
6880         };
6881
6882         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6883                 return;
6884         switch (task->tk_status) {
6885                 case 0:
6886                         renew_lease(calldata->server, calldata->timestamp);
6887                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6888                         if (nfs4_update_lock_stateid(calldata->lsp,
6889                                         &calldata->res.stateid))
6890                                 break;
6891                         fallthrough;
6892                 case -NFS4ERR_ADMIN_REVOKED:
6893                 case -NFS4ERR_EXPIRED:
6894                         nfs4_free_revoked_stateid(calldata->server,
6895                                         &calldata->arg.stateid,
6896                                         task->tk_msg.rpc_cred);
6897                         fallthrough;
6898                 case -NFS4ERR_BAD_STATEID:
6899                 case -NFS4ERR_STALE_STATEID:
6900                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6901                                                 calldata->lsp))
6902                                 rpc_restart_call_prepare(task);
6903                         break;
6904                 case -NFS4ERR_OLD_STATEID:
6905                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6906                                                 calldata->lsp))
6907                                 rpc_restart_call_prepare(task);
6908                         break;
6909                 default:
6910                         task->tk_status = nfs4_async_handle_exception(task,
6911                                         calldata->server, task->tk_status,
6912                                         &exception);
6913                         if (exception.retry)
6914                                 rpc_restart_call_prepare(task);
6915         }
6916         nfs_release_seqid(calldata->arg.seqid);
6917 }
6918
6919 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6920 {
6921         struct nfs4_unlockdata *calldata = data;
6922
6923         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6924                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6925                 return;
6926
6927         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6928                 goto out_wait;
6929         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6930                 /* Note: exit _without_ running nfs4_locku_done */
6931                 goto out_no_action;
6932         }
6933         calldata->timestamp = jiffies;
6934         if (nfs4_setup_sequence(calldata->server->nfs_client,
6935                                 &calldata->arg.seq_args,
6936                                 &calldata->res.seq_res,
6937                                 task) != 0)
6938                 nfs_release_seqid(calldata->arg.seqid);
6939         return;
6940 out_no_action:
6941         task->tk_action = NULL;
6942 out_wait:
6943         nfs4_sequence_done(task, &calldata->res.seq_res);
6944 }
6945
6946 static const struct rpc_call_ops nfs4_locku_ops = {
6947         .rpc_call_prepare = nfs4_locku_prepare,
6948         .rpc_call_done = nfs4_locku_done,
6949         .rpc_release = nfs4_locku_release_calldata,
6950 };
6951
6952 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6953                 struct nfs_open_context *ctx,
6954                 struct nfs4_lock_state *lsp,
6955                 struct nfs_seqid *seqid)
6956 {
6957         struct nfs4_unlockdata *data;
6958         struct rpc_message msg = {
6959                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6960                 .rpc_cred = ctx->cred,
6961         };
6962         struct rpc_task_setup task_setup_data = {
6963                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6964                 .rpc_message = &msg,
6965                 .callback_ops = &nfs4_locku_ops,
6966                 .workqueue = nfsiod_workqueue,
6967                 .flags = RPC_TASK_ASYNC,
6968         };
6969
6970         if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
6971                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6972
6973         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6974                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6975
6976         /* Ensure this is an unlock - when canceling a lock, the
6977          * canceled lock is passed in, and it won't be an unlock.
6978          */
6979         fl->fl_type = F_UNLCK;
6980         if (fl->fl_flags & FL_CLOSE)
6981                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6982
6983         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6984         if (data == NULL) {
6985                 nfs_free_seqid(seqid);
6986                 return ERR_PTR(-ENOMEM);
6987         }
6988
6989         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6990         msg.rpc_argp = &data->arg;
6991         msg.rpc_resp = &data->res;
6992         task_setup_data.callback_data = data;
6993         return rpc_run_task(&task_setup_data);
6994 }
6995
6996 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6997 {
6998         struct inode *inode = state->inode;
6999         struct nfs4_state_owner *sp = state->owner;
7000         struct nfs_inode *nfsi = NFS_I(inode);
7001         struct nfs_seqid *seqid;
7002         struct nfs4_lock_state *lsp;
7003         struct rpc_task *task;
7004         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7005         int status = 0;
7006         unsigned char fl_flags = request->fl_flags;
7007
7008         status = nfs4_set_lock_state(state, request);
7009         /* Unlock _before_ we do the RPC call */
7010         request->fl_flags |= FL_EXISTS;
7011         /* Exclude nfs_delegation_claim_locks() */
7012         mutex_lock(&sp->so_delegreturn_mutex);
7013         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7014         down_read(&nfsi->rwsem);
7015         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7016                 up_read(&nfsi->rwsem);
7017                 mutex_unlock(&sp->so_delegreturn_mutex);
7018                 goto out;
7019         }
7020         up_read(&nfsi->rwsem);
7021         mutex_unlock(&sp->so_delegreturn_mutex);
7022         if (status != 0)
7023                 goto out;
7024         /* Is this a delegated lock? */
7025         lsp = request->fl_u.nfs4_fl.owner;
7026         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7027                 goto out;
7028         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7029         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7030         status = -ENOMEM;
7031         if (IS_ERR(seqid))
7032                 goto out;
7033         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
7034         status = PTR_ERR(task);
7035         if (IS_ERR(task))
7036                 goto out;
7037         status = rpc_wait_for_completion_task(task);
7038         rpc_put_task(task);
7039 out:
7040         request->fl_flags = fl_flags;
7041         trace_nfs4_unlock(request, state, F_SETLK, status);
7042         return status;
7043 }
7044
7045 struct nfs4_lockdata {
7046         struct nfs_lock_args arg;
7047         struct nfs_lock_res res;
7048         struct nfs4_lock_state *lsp;
7049         struct nfs_open_context *ctx;
7050         struct file_lock fl;
7051         unsigned long timestamp;
7052         int rpc_status;
7053         int cancelled;
7054         struct nfs_server *server;
7055 };
7056
7057 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7058                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7059                 gfp_t gfp_mask)
7060 {
7061         struct nfs4_lockdata *p;
7062         struct inode *inode = lsp->ls_state->inode;
7063         struct nfs_server *server = NFS_SERVER(inode);
7064         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7065
7066         p = kzalloc(sizeof(*p), gfp_mask);
7067         if (p == NULL)
7068                 return NULL;
7069
7070         p->arg.fh = NFS_FH(inode);
7071         p->arg.fl = &p->fl;
7072         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7073         if (IS_ERR(p->arg.open_seqid))
7074                 goto out_free;
7075         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7076         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7077         if (IS_ERR(p->arg.lock_seqid))
7078                 goto out_free_seqid;
7079         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7080         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7081         p->arg.lock_owner.s_dev = server->s_dev;
7082         p->res.lock_seqid = p->arg.lock_seqid;
7083         p->lsp = lsp;
7084         p->server = server;
7085         p->ctx = get_nfs_open_context(ctx);
7086         locks_init_lock(&p->fl);
7087         locks_copy_lock(&p->fl, fl);
7088         return p;
7089 out_free_seqid:
7090         nfs_free_seqid(p->arg.open_seqid);
7091 out_free:
7092         kfree(p);
7093         return NULL;
7094 }
7095
7096 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7097 {
7098         struct nfs4_lockdata *data = calldata;
7099         struct nfs4_state *state = data->lsp->ls_state;
7100
7101         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7102                 goto out_wait;
7103         /* Do we need to do an open_to_lock_owner? */
7104         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7105                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7106                         goto out_release_lock_seqid;
7107                 }
7108                 nfs4_stateid_copy(&data->arg.open_stateid,
7109                                 &state->open_stateid);
7110                 data->arg.new_lock_owner = 1;
7111                 data->res.open_seqid = data->arg.open_seqid;
7112         } else {
7113                 data->arg.new_lock_owner = 0;
7114                 nfs4_stateid_copy(&data->arg.lock_stateid,
7115                                 &data->lsp->ls_stateid);
7116         }
7117         if (!nfs4_valid_open_stateid(state)) {
7118                 data->rpc_status = -EBADF;
7119                 task->tk_action = NULL;
7120                 goto out_release_open_seqid;
7121         }
7122         data->timestamp = jiffies;
7123         if (nfs4_setup_sequence(data->server->nfs_client,
7124                                 &data->arg.seq_args,
7125                                 &data->res.seq_res,
7126                                 task) == 0)
7127                 return;
7128 out_release_open_seqid:
7129         nfs_release_seqid(data->arg.open_seqid);
7130 out_release_lock_seqid:
7131         nfs_release_seqid(data->arg.lock_seqid);
7132 out_wait:
7133         nfs4_sequence_done(task, &data->res.seq_res);
7134         dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7135 }
7136
7137 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7138 {
7139         struct nfs4_lockdata *data = calldata;
7140         struct nfs4_lock_state *lsp = data->lsp;
7141         struct nfs_server *server = NFS_SERVER(d_inode(data->ctx->dentry));
7142
7143         if (!nfs4_sequence_done(task, &data->res.seq_res))
7144                 return;
7145
7146         data->rpc_status = task->tk_status;
7147         switch (task->tk_status) {
7148         case 0:
7149                 renew_lease(server, data->timestamp);
7150                 if (data->arg.new_lock && !data->cancelled) {
7151                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7152                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7153                                 goto out_restart;
7154                 }
7155                 if (data->arg.new_lock_owner != 0) {
7156                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
7157                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7158                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7159                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7160                         goto out_restart;
7161                 break;
7162         case -NFS4ERR_BAD_STATEID:
7163         case -NFS4ERR_OLD_STATEID:
7164         case -NFS4ERR_STALE_STATEID:
7165         case -NFS4ERR_EXPIRED:
7166                 if (data->arg.new_lock_owner != 0) {
7167                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7168                                                 &lsp->ls_state->open_stateid))
7169                                 goto out_restart;
7170                         else if (nfs4_async_handle_error(task, server, lsp->ls_state, NULL) == -EAGAIN)
7171                                 goto out_restart;
7172                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7173                                                 &lsp->ls_stateid))
7174                                 goto out_restart;
7175         }
7176 out_done:
7177         dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7178         return;
7179 out_restart:
7180         if (!data->cancelled)
7181                 rpc_restart_call_prepare(task);
7182         goto out_done;
7183 }
7184
7185 static void nfs4_lock_release(void *calldata)
7186 {
7187         struct nfs4_lockdata *data = calldata;
7188
7189         nfs_free_seqid(data->arg.open_seqid);
7190         if (data->cancelled && data->rpc_status == 0) {
7191                 struct rpc_task *task;
7192                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7193                                 data->arg.lock_seqid);
7194                 if (!IS_ERR(task))
7195                         rpc_put_task_async(task);
7196                 dprintk("%s: cancelling lock!\n", __func__);
7197         } else
7198                 nfs_free_seqid(data->arg.lock_seqid);
7199         nfs4_put_lock_state(data->lsp);
7200         put_nfs_open_context(data->ctx);
7201         kfree(data);
7202 }
7203
7204 static const struct rpc_call_ops nfs4_lock_ops = {
7205         .rpc_call_prepare = nfs4_lock_prepare,
7206         .rpc_call_done = nfs4_lock_done,
7207         .rpc_release = nfs4_lock_release,
7208 };
7209
7210 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7211 {
7212         switch (error) {
7213         case -NFS4ERR_ADMIN_REVOKED:
7214         case -NFS4ERR_EXPIRED:
7215         case -NFS4ERR_BAD_STATEID:
7216                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7217                 if (new_lock_owner != 0 ||
7218                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7219                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7220                 break;
7221         case -NFS4ERR_STALE_STATEID:
7222                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7223                 nfs4_schedule_lease_recovery(server->nfs_client);
7224         }
7225 }
7226
7227 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7228 {
7229         struct nfs4_lockdata *data;
7230         struct rpc_task *task;
7231         struct rpc_message msg = {
7232                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7233                 .rpc_cred = state->owner->so_cred,
7234         };
7235         struct rpc_task_setup task_setup_data = {
7236                 .rpc_client = NFS_CLIENT(state->inode),
7237                 .rpc_message = &msg,
7238                 .callback_ops = &nfs4_lock_ops,
7239                 .workqueue = nfsiod_workqueue,
7240                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7241         };
7242         int ret;
7243
7244         if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7245                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
7246
7247         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7248                                    fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7249         if (data == NULL)
7250                 return -ENOMEM;
7251         if (IS_SETLKW(cmd))
7252                 data->arg.block = 1;
7253         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7254                                 recovery_type > NFS_LOCK_NEW);
7255         msg.rpc_argp = &data->arg;
7256         msg.rpc_resp = &data->res;
7257         task_setup_data.callback_data = data;
7258         if (recovery_type > NFS_LOCK_NEW) {
7259                 if (recovery_type == NFS_LOCK_RECLAIM)
7260                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7261         } else
7262                 data->arg.new_lock = 1;
7263         task = rpc_run_task(&task_setup_data);
7264         if (IS_ERR(task))
7265                 return PTR_ERR(task);
7266         ret = rpc_wait_for_completion_task(task);
7267         if (ret == 0) {
7268                 ret = data->rpc_status;
7269                 if (ret)
7270                         nfs4_handle_setlk_error(data->server, data->lsp,
7271                                         data->arg.new_lock_owner, ret);
7272         } else
7273                 data->cancelled = true;
7274         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7275         rpc_put_task(task);
7276         dprintk("%s: ret = %d\n", __func__, ret);
7277         return ret;
7278 }
7279
7280 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7281 {
7282         struct nfs_server *server = NFS_SERVER(state->inode);
7283         struct nfs4_exception exception = {
7284                 .inode = state->inode,
7285         };
7286         int err;
7287
7288         do {
7289                 /* Cache the lock if possible... */
7290                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7291                         return 0;
7292                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7293                 if (err != -NFS4ERR_DELAY)
7294                         break;
7295                 nfs4_handle_exception(server, err, &exception);
7296         } while (exception.retry);
7297         return err;
7298 }
7299
7300 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7301 {
7302         struct nfs_server *server = NFS_SERVER(state->inode);
7303         struct nfs4_exception exception = {
7304                 .inode = state->inode,
7305         };
7306         int err;
7307
7308         err = nfs4_set_lock_state(state, request);
7309         if (err != 0)
7310                 return err;
7311         if (!recover_lost_locks) {
7312                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7313                 return 0;
7314         }
7315         do {
7316                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7317                         return 0;
7318                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7319                 switch (err) {
7320                 default:
7321                         goto out;
7322                 case -NFS4ERR_GRACE:
7323                 case -NFS4ERR_DELAY:
7324                         nfs4_handle_exception(server, err, &exception);
7325                         err = 0;
7326                 }
7327         } while (exception.retry);
7328 out:
7329         return err;
7330 }
7331
7332 #if defined(CONFIG_NFS_V4_1)
7333 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7334 {
7335         struct nfs4_lock_state *lsp;
7336         int status;
7337
7338         status = nfs4_set_lock_state(state, request);
7339         if (status != 0)
7340                 return status;
7341         lsp = request->fl_u.nfs4_fl.owner;
7342         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7343             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7344                 return 0;
7345         return nfs4_lock_expired(state, request);
7346 }
7347 #endif
7348
7349 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7350 {
7351         struct nfs_inode *nfsi = NFS_I(state->inode);
7352         struct nfs4_state_owner *sp = state->owner;
7353         unsigned char fl_flags = request->fl_flags;
7354         int status;
7355
7356         request->fl_flags |= FL_ACCESS;
7357         status = locks_lock_inode_wait(state->inode, request);
7358         if (status < 0)
7359                 goto out;
7360         mutex_lock(&sp->so_delegreturn_mutex);
7361         down_read(&nfsi->rwsem);
7362         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7363                 /* Yes: cache locks! */
7364                 /* ...but avoid races with delegation recall... */
7365                 request->fl_flags = fl_flags & ~FL_SLEEP;
7366                 status = locks_lock_inode_wait(state->inode, request);
7367                 up_read(&nfsi->rwsem);
7368                 mutex_unlock(&sp->so_delegreturn_mutex);
7369                 goto out;
7370         }
7371         up_read(&nfsi->rwsem);
7372         mutex_unlock(&sp->so_delegreturn_mutex);
7373         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7374 out:
7375         request->fl_flags = fl_flags;
7376         return status;
7377 }
7378
7379 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7380 {
7381         struct nfs4_exception exception = {
7382                 .state = state,
7383                 .inode = state->inode,
7384                 .interruptible = true,
7385         };
7386         int err;
7387
7388         do {
7389                 err = _nfs4_proc_setlk(state, cmd, request);
7390                 if (err == -NFS4ERR_DENIED)
7391                         err = -EAGAIN;
7392                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7393                                 err, &exception);
7394         } while (exception.retry);
7395         return err;
7396 }
7397
7398 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7399 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7400
7401 static int
7402 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7403                         struct file_lock *request)
7404 {
7405         int             status = -ERESTARTSYS;
7406         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7407
7408         while(!signalled()) {
7409                 status = nfs4_proc_setlk(state, cmd, request);
7410                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7411                         break;
7412                 __set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
7413                 schedule_timeout(timeout);
7414                 timeout *= 2;
7415                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7416                 status = -ERESTARTSYS;
7417         }
7418         return status;
7419 }
7420
7421 #ifdef CONFIG_NFS_V4_1
7422 struct nfs4_lock_waiter {
7423         struct inode            *inode;
7424         struct nfs_lowner       owner;
7425         wait_queue_entry_t      wait;
7426 };
7427
7428 static int
7429 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7430 {
7431         struct nfs4_lock_waiter *waiter =
7432                 container_of(wait, struct nfs4_lock_waiter, wait);
7433
7434         /* NULL key means to wake up everyone */
7435         if (key) {
7436                 struct cb_notify_lock_args      *cbnl = key;
7437                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7438                                                 *wowner = &waiter->owner;
7439
7440                 /* Only wake if the callback was for the same owner. */
7441                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7442                         return 0;
7443
7444                 /* Make sure it's for the right inode */
7445                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7446                         return 0;
7447         }
7448
7449         return woken_wake_function(wait, mode, flags, key);
7450 }
7451
7452 static int
7453 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7454 {
7455         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7456         struct nfs_server *server = NFS_SERVER(state->inode);
7457         struct nfs_client *clp = server->nfs_client;
7458         wait_queue_head_t *q = &clp->cl_lock_waitq;
7459         struct nfs4_lock_waiter waiter = {
7460                 .inode = state->inode,
7461                 .owner = { .clientid = clp->cl_clientid,
7462                            .id = lsp->ls_seqid.owner_id,
7463                            .s_dev = server->s_dev },
7464         };
7465         int status;
7466
7467         /* Don't bother with waitqueue if we don't expect a callback */
7468         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7469                 return nfs4_retry_setlk_simple(state, cmd, request);
7470
7471         init_wait(&waiter.wait);
7472         waiter.wait.func = nfs4_wake_lock_waiter;
7473         add_wait_queue(q, &waiter.wait);
7474
7475         do {
7476                 status = nfs4_proc_setlk(state, cmd, request);
7477                 if (status != -EAGAIN || IS_SETLK(cmd))
7478                         break;
7479
7480                 status = -ERESTARTSYS;
7481                 wait_woken(&waiter.wait, TASK_INTERRUPTIBLE|TASK_FREEZABLE,
7482                            NFS4_LOCK_MAXTIMEOUT);
7483         } while (!signalled());
7484
7485         remove_wait_queue(q, &waiter.wait);
7486
7487         return status;
7488 }
7489 #else /* !CONFIG_NFS_V4_1 */
7490 static inline int
7491 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7492 {
7493         return nfs4_retry_setlk_simple(state, cmd, request);
7494 }
7495 #endif
7496
7497 static int
7498 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7499 {
7500         struct nfs_open_context *ctx;
7501         struct nfs4_state *state;
7502         int status;
7503
7504         /* verify open state */
7505         ctx = nfs_file_open_context(filp);
7506         state = ctx->state;
7507
7508         if (IS_GETLK(cmd)) {
7509                 if (state != NULL)
7510                         return nfs4_proc_getlk(state, F_GETLK, request);
7511                 return 0;
7512         }
7513
7514         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7515                 return -EINVAL;
7516
7517         if (request->fl_type == F_UNLCK) {
7518                 if (state != NULL)
7519                         return nfs4_proc_unlck(state, cmd, request);
7520                 return 0;
7521         }
7522
7523         if (state == NULL)
7524                 return -ENOLCK;
7525
7526         if ((request->fl_flags & FL_POSIX) &&
7527             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7528                 return -ENOLCK;
7529
7530         /*
7531          * Don't rely on the VFS having checked the file open mode,
7532          * since it won't do this for flock() locks.
7533          */
7534         switch (request->fl_type) {
7535         case F_RDLCK:
7536                 if (!(filp->f_mode & FMODE_READ))
7537                         return -EBADF;
7538                 break;
7539         case F_WRLCK:
7540                 if (!(filp->f_mode & FMODE_WRITE))
7541                         return -EBADF;
7542         }
7543
7544         status = nfs4_set_lock_state(state, request);
7545         if (status != 0)
7546                 return status;
7547
7548         return nfs4_retry_setlk(state, cmd, request);
7549 }
7550
7551 static int nfs4_delete_lease(struct file *file, void **priv)
7552 {
7553         return generic_setlease(file, F_UNLCK, NULL, priv);
7554 }
7555
7556 static int nfs4_add_lease(struct file *file, long arg, struct file_lock **lease,
7557                           void **priv)
7558 {
7559         struct inode *inode = file_inode(file);
7560         fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7561         int ret;
7562
7563         /* No delegation, no lease */
7564         if (!nfs4_have_delegation(inode, type))
7565                 return -EAGAIN;
7566         ret = generic_setlease(file, arg, lease, priv);
7567         if (ret || nfs4_have_delegation(inode, type))
7568                 return ret;
7569         /* We raced with a delegation return */
7570         nfs4_delete_lease(file, priv);
7571         return -EAGAIN;
7572 }
7573
7574 int nfs4_proc_setlease(struct file *file, long arg, struct file_lock **lease,
7575                        void **priv)
7576 {
7577         switch (arg) {
7578         case F_RDLCK:
7579         case F_WRLCK:
7580                 return nfs4_add_lease(file, arg, lease, priv);
7581         case F_UNLCK:
7582                 return nfs4_delete_lease(file, priv);
7583         default:
7584                 return -EINVAL;
7585         }
7586 }
7587
7588 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7589 {
7590         struct nfs_server *server = NFS_SERVER(state->inode);
7591         int err;
7592
7593         err = nfs4_set_lock_state(state, fl);
7594         if (err != 0)
7595                 return err;
7596         do {
7597                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7598                 if (err != -NFS4ERR_DELAY)
7599                         break;
7600                 ssleep(1);
7601         } while (err == -NFS4ERR_DELAY);
7602         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7603 }
7604
7605 struct nfs_release_lockowner_data {
7606         struct nfs4_lock_state *lsp;
7607         struct nfs_server *server;
7608         struct nfs_release_lockowner_args args;
7609         struct nfs_release_lockowner_res res;
7610         unsigned long timestamp;
7611 };
7612
7613 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7614 {
7615         struct nfs_release_lockowner_data *data = calldata;
7616         struct nfs_server *server = data->server;
7617         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7618                            &data->res.seq_res, task);
7619         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7620         data->timestamp = jiffies;
7621 }
7622
7623 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7624 {
7625         struct nfs_release_lockowner_data *data = calldata;
7626         struct nfs_server *server = data->server;
7627
7628         nfs40_sequence_done(task, &data->res.seq_res);
7629
7630         switch (task->tk_status) {
7631         case 0:
7632                 renew_lease(server, data->timestamp);
7633                 break;
7634         case -NFS4ERR_STALE_CLIENTID:
7635         case -NFS4ERR_EXPIRED:
7636                 nfs4_schedule_lease_recovery(server->nfs_client);
7637                 break;
7638         case -NFS4ERR_LEASE_MOVED:
7639         case -NFS4ERR_DELAY:
7640                 if (nfs4_async_handle_error(task, server,
7641                                             NULL, NULL) == -EAGAIN)
7642                         rpc_restart_call_prepare(task);
7643         }
7644 }
7645
7646 static void nfs4_release_lockowner_release(void *calldata)
7647 {
7648         struct nfs_release_lockowner_data *data = calldata;
7649         nfs4_free_lock_state(data->server, data->lsp);
7650         kfree(calldata);
7651 }
7652
7653 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7654         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7655         .rpc_call_done = nfs4_release_lockowner_done,
7656         .rpc_release = nfs4_release_lockowner_release,
7657 };
7658
7659 static void
7660 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7661 {
7662         struct nfs_release_lockowner_data *data;
7663         struct rpc_message msg = {
7664                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7665         };
7666
7667         if (server->nfs_client->cl_mvops->minor_version != 0)
7668                 return;
7669
7670         data = kmalloc(sizeof(*data), GFP_KERNEL);
7671         if (!data)
7672                 return;
7673         data->lsp = lsp;
7674         data->server = server;
7675         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7676         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7677         data->args.lock_owner.s_dev = server->s_dev;
7678
7679         msg.rpc_argp = &data->args;
7680         msg.rpc_resp = &data->res;
7681         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7682         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7683 }
7684
7685 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7686
7687 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7688                                    struct user_namespace *mnt_userns,
7689                                    struct dentry *unused, struct inode *inode,
7690                                    const char *key, const void *buf,
7691                                    size_t buflen, int flags)
7692 {
7693         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7694 }
7695
7696 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7697                                    struct dentry *unused, struct inode *inode,
7698                                    const char *key, void *buf, size_t buflen)
7699 {
7700         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7701 }
7702
7703 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7704 {
7705         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7706 }
7707
7708 #if defined(CONFIG_NFS_V4_1)
7709 #define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7710
7711 static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7712                                     struct user_namespace *mnt_userns,
7713                                     struct dentry *unused, struct inode *inode,
7714                                     const char *key, const void *buf,
7715                                     size_t buflen, int flags)
7716 {
7717         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7718 }
7719
7720 static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7721                                     struct dentry *unused, struct inode *inode,
7722                                     const char *key, void *buf, size_t buflen)
7723 {
7724         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7725 }
7726
7727 static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7728 {
7729         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7730 }
7731
7732 #define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7733
7734 static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7735                                     struct user_namespace *mnt_userns,
7736                                     struct dentry *unused, struct inode *inode,
7737                                     const char *key, const void *buf,
7738                                     size_t buflen, int flags)
7739 {
7740         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7741 }
7742
7743 static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7744                                     struct dentry *unused, struct inode *inode,
7745                                     const char *key, void *buf, size_t buflen)
7746 {
7747         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7748 }
7749
7750 static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7751 {
7752         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7753 }
7754
7755 #endif
7756
7757 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7758
7759 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7760                                      struct user_namespace *mnt_userns,
7761                                      struct dentry *unused, struct inode *inode,
7762                                      const char *key, const void *buf,
7763                                      size_t buflen, int flags)
7764 {
7765         if (security_ismaclabel(key))
7766                 return nfs4_set_security_label(inode, buf, buflen);
7767
7768         return -EOPNOTSUPP;
7769 }
7770
7771 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7772                                      struct dentry *unused, struct inode *inode,
7773                                      const char *key, void *buf, size_t buflen)
7774 {
7775         if (security_ismaclabel(key))
7776                 return nfs4_get_security_label(inode, buf, buflen);
7777         return -EOPNOTSUPP;
7778 }
7779
7780 static ssize_t
7781 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7782 {
7783         int len = 0;
7784
7785         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7786                 len = security_inode_listsecurity(inode, list, list_len);
7787                 if (len >= 0 && list_len && len > list_len)
7788                         return -ERANGE;
7789         }
7790         return len;
7791 }
7792
7793 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7794         .prefix = XATTR_SECURITY_PREFIX,
7795         .get    = nfs4_xattr_get_nfs4_label,
7796         .set    = nfs4_xattr_set_nfs4_label,
7797 };
7798
7799 #else
7800
7801 static ssize_t
7802 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7803 {
7804         return 0;
7805 }
7806
7807 #endif
7808
7809 #ifdef CONFIG_NFS_V4_2
7810 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7811                                     struct user_namespace *mnt_userns,
7812                                     struct dentry *unused, struct inode *inode,
7813                                     const char *key, const void *buf,
7814                                     size_t buflen, int flags)
7815 {
7816         u32 mask;
7817         int ret;
7818
7819         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7820                 return -EOPNOTSUPP;
7821
7822         /*
7823          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7824          * flags right now. Handling of xattr operations use the normal
7825          * file read/write permissions.
7826          *
7827          * Just in case the server has other ideas (which RFC 8276 allows),
7828          * do a cached access check for the XA* flags to possibly avoid
7829          * doing an RPC and getting EACCES back.
7830          */
7831         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7832                 if (!(mask & NFS_ACCESS_XAWRITE))
7833                         return -EACCES;
7834         }
7835
7836         if (buf == NULL) {
7837                 ret = nfs42_proc_removexattr(inode, key);
7838                 if (!ret)
7839                         nfs4_xattr_cache_remove(inode, key);
7840         } else {
7841                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7842                 if (!ret)
7843                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7844         }
7845
7846         return ret;
7847 }
7848
7849 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7850                                     struct dentry *unused, struct inode *inode,
7851                                     const char *key, void *buf, size_t buflen)
7852 {
7853         u32 mask;
7854         ssize_t ret;
7855
7856         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7857                 return -EOPNOTSUPP;
7858
7859         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7860                 if (!(mask & NFS_ACCESS_XAREAD))
7861                         return -EACCES;
7862         }
7863
7864         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7865         if (ret)
7866                 return ret;
7867
7868         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7869         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7870                 return ret;
7871
7872         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7873
7874         return ret;
7875 }
7876
7877 static ssize_t
7878 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7879 {
7880         u64 cookie;
7881         bool eof;
7882         ssize_t ret, size;
7883         char *buf;
7884         size_t buflen;
7885         u32 mask;
7886
7887         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7888                 return 0;
7889
7890         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7891                 if (!(mask & NFS_ACCESS_XALIST))
7892                         return 0;
7893         }
7894
7895         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7896         if (ret)
7897                 return ret;
7898
7899         ret = nfs4_xattr_cache_list(inode, list, list_len);
7900         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7901                 return ret;
7902
7903         cookie = 0;
7904         eof = false;
7905         buflen = list_len ? list_len : XATTR_LIST_MAX;
7906         buf = list_len ? list : NULL;
7907         size = 0;
7908
7909         while (!eof) {
7910                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7911                     &cookie, &eof);
7912                 if (ret < 0)
7913                         return ret;
7914
7915                 if (list_len) {
7916                         buf += ret;
7917                         buflen -= ret;
7918                 }
7919                 size += ret;
7920         }
7921
7922         if (list_len)
7923                 nfs4_xattr_cache_set_list(inode, list, size);
7924
7925         return size;
7926 }
7927
7928 #else
7929
7930 static ssize_t
7931 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7932 {
7933         return 0;
7934 }
7935 #endif /* CONFIG_NFS_V4_2 */
7936
7937 /*
7938  * nfs_fhget will use either the mounted_on_fileid or the fileid
7939  */
7940 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7941 {
7942         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7943                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7944               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7945               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7946                 return;
7947
7948         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7949                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7950         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7951         fattr->nlink = 2;
7952 }
7953
7954 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7955                                    const struct qstr *name,
7956                                    struct nfs4_fs_locations *fs_locations,
7957                                    struct page *page)
7958 {
7959         struct nfs_server *server = NFS_SERVER(dir);
7960         u32 bitmask[3];
7961         struct nfs4_fs_locations_arg args = {
7962                 .dir_fh = NFS_FH(dir),
7963                 .name = name,
7964                 .page = page,
7965                 .bitmask = bitmask,
7966         };
7967         struct nfs4_fs_locations_res res = {
7968                 .fs_locations = fs_locations,
7969         };
7970         struct rpc_message msg = {
7971                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7972                 .rpc_argp = &args,
7973                 .rpc_resp = &res,
7974         };
7975         int status;
7976
7977         dprintk("%s: start\n", __func__);
7978
7979         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7980         bitmask[1] = nfs4_fattr_bitmap[1];
7981
7982         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7983          * is not supported */
7984         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7985                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7986         else
7987                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7988
7989         nfs_fattr_init(fs_locations->fattr);
7990         fs_locations->server = server;
7991         fs_locations->nlocations = 0;
7992         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7993         dprintk("%s: returned status = %d\n", __func__, status);
7994         return status;
7995 }
7996
7997 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7998                            const struct qstr *name,
7999                            struct nfs4_fs_locations *fs_locations,
8000                            struct page *page)
8001 {
8002         struct nfs4_exception exception = {
8003                 .interruptible = true,
8004         };
8005         int err;
8006         do {
8007                 err = _nfs4_proc_fs_locations(client, dir, name,
8008                                 fs_locations, page);
8009                 trace_nfs4_get_fs_locations(dir, name, err);
8010                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8011                                 &exception);
8012         } while (exception.retry);
8013         return err;
8014 }
8015
8016 /*
8017  * This operation also signals the server that this client is
8018  * performing migration recovery.  The server can stop returning
8019  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
8020  * appended to this compound to identify the client ID which is
8021  * performing recovery.
8022  */
8023 static int _nfs40_proc_get_locations(struct nfs_server *server,
8024                                      struct nfs_fh *fhandle,
8025                                      struct nfs4_fs_locations *locations,
8026                                      struct page *page, const struct cred *cred)
8027 {
8028         struct rpc_clnt *clnt = server->client;
8029         u32 bitmask[2] = {
8030                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8031         };
8032         struct nfs4_fs_locations_arg args = {
8033                 .clientid       = server->nfs_client->cl_clientid,
8034                 .fh             = fhandle,
8035                 .page           = page,
8036                 .bitmask        = bitmask,
8037                 .migration      = 1,            /* skip LOOKUP */
8038                 .renew          = 1,            /* append RENEW */
8039         };
8040         struct nfs4_fs_locations_res res = {
8041                 .fs_locations   = locations,
8042                 .migration      = 1,
8043                 .renew          = 1,
8044         };
8045         struct rpc_message msg = {
8046                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8047                 .rpc_argp       = &args,
8048                 .rpc_resp       = &res,
8049                 .rpc_cred       = cred,
8050         };
8051         unsigned long now = jiffies;
8052         int status;
8053
8054         nfs_fattr_init(locations->fattr);
8055         locations->server = server;
8056         locations->nlocations = 0;
8057
8058         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8059         status = nfs4_call_sync_sequence(clnt, server, &msg,
8060                                         &args.seq_args, &res.seq_res);
8061         if (status)
8062                 return status;
8063
8064         renew_lease(server, now);
8065         return 0;
8066 }
8067
8068 #ifdef CONFIG_NFS_V4_1
8069
8070 /*
8071  * This operation also signals the server that this client is
8072  * performing migration recovery.  The server can stop asserting
8073  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8074  * performing this operation is identified in the SEQUENCE
8075  * operation in this compound.
8076  *
8077  * When the client supports GETATTR(fs_locations_info), it can
8078  * be plumbed in here.
8079  */
8080 static int _nfs41_proc_get_locations(struct nfs_server *server,
8081                                      struct nfs_fh *fhandle,
8082                                      struct nfs4_fs_locations *locations,
8083                                      struct page *page, const struct cred *cred)
8084 {
8085         struct rpc_clnt *clnt = server->client;
8086         u32 bitmask[2] = {
8087                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8088         };
8089         struct nfs4_fs_locations_arg args = {
8090                 .fh             = fhandle,
8091                 .page           = page,
8092                 .bitmask        = bitmask,
8093                 .migration      = 1,            /* skip LOOKUP */
8094         };
8095         struct nfs4_fs_locations_res res = {
8096                 .fs_locations   = locations,
8097                 .migration      = 1,
8098         };
8099         struct rpc_message msg = {
8100                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8101                 .rpc_argp       = &args,
8102                 .rpc_resp       = &res,
8103                 .rpc_cred       = cred,
8104         };
8105         struct nfs4_call_sync_data data = {
8106                 .seq_server = server,
8107                 .seq_args = &args.seq_args,
8108                 .seq_res = &res.seq_res,
8109         };
8110         struct rpc_task_setup task_setup_data = {
8111                 .rpc_client = clnt,
8112                 .rpc_message = &msg,
8113                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
8114                 .callback_data = &data,
8115                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8116         };
8117         int status;
8118
8119         nfs_fattr_init(locations->fattr);
8120         locations->server = server;
8121         locations->nlocations = 0;
8122
8123         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8124         status = nfs4_call_sync_custom(&task_setup_data);
8125         if (status == NFS4_OK &&
8126             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8127                 status = -NFS4ERR_LEASE_MOVED;
8128         return status;
8129 }
8130
8131 #endif  /* CONFIG_NFS_V4_1 */
8132
8133 /**
8134  * nfs4_proc_get_locations - discover locations for a migrated FSID
8135  * @server: pointer to nfs_server to process
8136  * @fhandle: pointer to the kernel NFS client file handle
8137  * @locations: result of query
8138  * @page: buffer
8139  * @cred: credential to use for this operation
8140  *
8141  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8142  * operation failed, or a negative errno if a local error occurred.
8143  *
8144  * On success, "locations" is filled in, but if the server has
8145  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8146  * asserted.
8147  *
8148  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8149  * from this client that require migration recovery.
8150  */
8151 int nfs4_proc_get_locations(struct nfs_server *server,
8152                             struct nfs_fh *fhandle,
8153                             struct nfs4_fs_locations *locations,
8154                             struct page *page, const struct cred *cred)
8155 {
8156         struct nfs_client *clp = server->nfs_client;
8157         const struct nfs4_mig_recovery_ops *ops =
8158                                         clp->cl_mvops->mig_recovery_ops;
8159         struct nfs4_exception exception = {
8160                 .interruptible = true,
8161         };
8162         int status;
8163
8164         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8165                 (unsigned long long)server->fsid.major,
8166                 (unsigned long long)server->fsid.minor,
8167                 clp->cl_hostname);
8168         nfs_display_fhandle(fhandle, __func__);
8169
8170         do {
8171                 status = ops->get_locations(server, fhandle, locations, page,
8172                                             cred);
8173                 if (status != -NFS4ERR_DELAY)
8174                         break;
8175                 nfs4_handle_exception(server, status, &exception);
8176         } while (exception.retry);
8177         return status;
8178 }
8179
8180 /*
8181  * This operation also signals the server that this client is
8182  * performing "lease moved" recovery.  The server can stop
8183  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8184  * is appended to this compound to identify the client ID which is
8185  * performing recovery.
8186  */
8187 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8188 {
8189         struct nfs_server *server = NFS_SERVER(inode);
8190         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8191         struct rpc_clnt *clnt = server->client;
8192         struct nfs4_fsid_present_arg args = {
8193                 .fh             = NFS_FH(inode),
8194                 .clientid       = clp->cl_clientid,
8195                 .renew          = 1,            /* append RENEW */
8196         };
8197         struct nfs4_fsid_present_res res = {
8198                 .renew          = 1,
8199         };
8200         struct rpc_message msg = {
8201                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8202                 .rpc_argp       = &args,
8203                 .rpc_resp       = &res,
8204                 .rpc_cred       = cred,
8205         };
8206         unsigned long now = jiffies;
8207         int status;
8208
8209         res.fh = nfs_alloc_fhandle();
8210         if (res.fh == NULL)
8211                 return -ENOMEM;
8212
8213         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8214         status = nfs4_call_sync_sequence(clnt, server, &msg,
8215                                                 &args.seq_args, &res.seq_res);
8216         nfs_free_fhandle(res.fh);
8217         if (status)
8218                 return status;
8219
8220         do_renew_lease(clp, now);
8221         return 0;
8222 }
8223
8224 #ifdef CONFIG_NFS_V4_1
8225
8226 /*
8227  * This operation also signals the server that this client is
8228  * performing "lease moved" recovery.  The server can stop asserting
8229  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8230  * this operation is identified in the SEQUENCE operation in this
8231  * compound.
8232  */
8233 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8234 {
8235         struct nfs_server *server = NFS_SERVER(inode);
8236         struct rpc_clnt *clnt = server->client;
8237         struct nfs4_fsid_present_arg args = {
8238                 .fh             = NFS_FH(inode),
8239         };
8240         struct nfs4_fsid_present_res res = {
8241         };
8242         struct rpc_message msg = {
8243                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8244                 .rpc_argp       = &args,
8245                 .rpc_resp       = &res,
8246                 .rpc_cred       = cred,
8247         };
8248         int status;
8249
8250         res.fh = nfs_alloc_fhandle();
8251         if (res.fh == NULL)
8252                 return -ENOMEM;
8253
8254         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8255         status = nfs4_call_sync_sequence(clnt, server, &msg,
8256                                                 &args.seq_args, &res.seq_res);
8257         nfs_free_fhandle(res.fh);
8258         if (status == NFS4_OK &&
8259             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8260                 status = -NFS4ERR_LEASE_MOVED;
8261         return status;
8262 }
8263
8264 #endif  /* CONFIG_NFS_V4_1 */
8265
8266 /**
8267  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8268  * @inode: inode on FSID to check
8269  * @cred: credential to use for this operation
8270  *
8271  * Server indicates whether the FSID is present, moved, or not
8272  * recognized.  This operation is necessary to clear a LEASE_MOVED
8273  * condition for this client ID.
8274  *
8275  * Returns NFS4_OK if the FSID is present on this server,
8276  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8277  *  NFS4ERR code if some error occurred on the server, or a
8278  *  negative errno if a local failure occurred.
8279  */
8280 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8281 {
8282         struct nfs_server *server = NFS_SERVER(inode);
8283         struct nfs_client *clp = server->nfs_client;
8284         const struct nfs4_mig_recovery_ops *ops =
8285                                         clp->cl_mvops->mig_recovery_ops;
8286         struct nfs4_exception exception = {
8287                 .interruptible = true,
8288         };
8289         int status;
8290
8291         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8292                 (unsigned long long)server->fsid.major,
8293                 (unsigned long long)server->fsid.minor,
8294                 clp->cl_hostname);
8295         nfs_display_fhandle(NFS_FH(inode), __func__);
8296
8297         do {
8298                 status = ops->fsid_present(inode, cred);
8299                 if (status != -NFS4ERR_DELAY)
8300                         break;
8301                 nfs4_handle_exception(server, status, &exception);
8302         } while (exception.retry);
8303         return status;
8304 }
8305
8306 /*
8307  * If 'use_integrity' is true and the state managment nfs_client
8308  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8309  * and the machine credential as per RFC3530bis and RFC5661 Security
8310  * Considerations sections. Otherwise, just use the user cred with the
8311  * filesystem's rpc_client.
8312  */
8313 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8314 {
8315         int status;
8316         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8317         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8318         struct nfs4_secinfo_arg args = {
8319                 .dir_fh = NFS_FH(dir),
8320                 .name   = name,
8321         };
8322         struct nfs4_secinfo_res res = {
8323                 .flavors     = flavors,
8324         };
8325         struct rpc_message msg = {
8326                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8327                 .rpc_argp = &args,
8328                 .rpc_resp = &res,
8329         };
8330         struct nfs4_call_sync_data data = {
8331                 .seq_server = NFS_SERVER(dir),
8332                 .seq_args = &args.seq_args,
8333                 .seq_res = &res.seq_res,
8334         };
8335         struct rpc_task_setup task_setup = {
8336                 .rpc_client = clnt,
8337                 .rpc_message = &msg,
8338                 .callback_ops = clp->cl_mvops->call_sync_ops,
8339                 .callback_data = &data,
8340                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8341         };
8342         const struct cred *cred = NULL;
8343
8344         if (use_integrity) {
8345                 clnt = clp->cl_rpcclient;
8346                 task_setup.rpc_client = clnt;
8347
8348                 cred = nfs4_get_clid_cred(clp);
8349                 msg.rpc_cred = cred;
8350         }
8351
8352         dprintk("NFS call  secinfo %s\n", name->name);
8353
8354         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8355         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8356         status = nfs4_call_sync_custom(&task_setup);
8357
8358         dprintk("NFS reply  secinfo: %d\n", status);
8359
8360         put_cred(cred);
8361         return status;
8362 }
8363
8364 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8365                       struct nfs4_secinfo_flavors *flavors)
8366 {
8367         struct nfs4_exception exception = {
8368                 .interruptible = true,
8369         };
8370         int err;
8371         do {
8372                 err = -NFS4ERR_WRONGSEC;
8373
8374                 /* try to use integrity protection with machine cred */
8375                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8376                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8377
8378                 /*
8379                  * if unable to use integrity protection, or SECINFO with
8380                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8381                  * disallowed by spec, but exists in deployed servers) use
8382                  * the current filesystem's rpc_client and the user cred.
8383                  */
8384                 if (err == -NFS4ERR_WRONGSEC)
8385                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8386
8387                 trace_nfs4_secinfo(dir, name, err);
8388                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8389                                 &exception);
8390         } while (exception.retry);
8391         return err;
8392 }
8393
8394 #ifdef CONFIG_NFS_V4_1
8395 /*
8396  * Check the exchange flags returned by the server for invalid flags, having
8397  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8398  * DS flags set.
8399  */
8400 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8401 {
8402         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8403                 goto out_inval;
8404         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8405                 goto out_inval;
8406         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8407             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8408                 goto out_inval;
8409         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8410                 goto out_inval;
8411         return NFS_OK;
8412 out_inval:
8413         return -NFS4ERR_INVAL;
8414 }
8415
8416 static bool
8417 nfs41_same_server_scope(struct nfs41_server_scope *a,
8418                         struct nfs41_server_scope *b)
8419 {
8420         if (a->server_scope_sz != b->server_scope_sz)
8421                 return false;
8422         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8423 }
8424
8425 static void
8426 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8427 {
8428         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8429         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8430         struct nfs_client *clp = args->client;
8431
8432         switch (task->tk_status) {
8433         case -NFS4ERR_BADSESSION:
8434         case -NFS4ERR_DEADSESSION:
8435                 nfs4_schedule_session_recovery(clp->cl_session,
8436                                 task->tk_status);
8437                 return;
8438         }
8439         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8440                         res->dir != NFS4_CDFS4_BOTH) {
8441                 rpc_task_close_connection(task);
8442                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8443                         rpc_restart_call(task);
8444         }
8445 }
8446
8447 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8448         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8449 };
8450
8451 /*
8452  * nfs4_proc_bind_one_conn_to_session()
8453  *
8454  * The 4.1 client currently uses the same TCP connection for the
8455  * fore and backchannel.
8456  */
8457 static
8458 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8459                 struct rpc_xprt *xprt,
8460                 struct nfs_client *clp,
8461                 const struct cred *cred)
8462 {
8463         int status;
8464         struct nfs41_bind_conn_to_session_args args = {
8465                 .client = clp,
8466                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8467                 .retries = 0,
8468         };
8469         struct nfs41_bind_conn_to_session_res res;
8470         struct rpc_message msg = {
8471                 .rpc_proc =
8472                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8473                 .rpc_argp = &args,
8474                 .rpc_resp = &res,
8475                 .rpc_cred = cred,
8476         };
8477         struct rpc_task_setup task_setup_data = {
8478                 .rpc_client = clnt,
8479                 .rpc_xprt = xprt,
8480                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8481                 .rpc_message = &msg,
8482                 .flags = RPC_TASK_TIMEOUT,
8483         };
8484         struct rpc_task *task;
8485
8486         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8487         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8488                 args.dir = NFS4_CDFC4_FORE;
8489
8490         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8491         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8492                 args.dir = NFS4_CDFC4_FORE;
8493
8494         task = rpc_run_task(&task_setup_data);
8495         if (!IS_ERR(task)) {
8496                 status = task->tk_status;
8497                 rpc_put_task(task);
8498         } else
8499                 status = PTR_ERR(task);
8500         trace_nfs4_bind_conn_to_session(clp, status);
8501         if (status == 0) {
8502                 if (memcmp(res.sessionid.data,
8503                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8504                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8505                         return -EIO;
8506                 }
8507                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8508                         dprintk("NFS: %s: Unexpected direction from server\n",
8509                                 __func__);
8510                         return -EIO;
8511                 }
8512                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8513                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8514                                 __func__);
8515                         return -EIO;
8516                 }
8517         }
8518
8519         return status;
8520 }
8521
8522 struct rpc_bind_conn_calldata {
8523         struct nfs_client *clp;
8524         const struct cred *cred;
8525 };
8526
8527 static int
8528 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8529                 struct rpc_xprt *xprt,
8530                 void *calldata)
8531 {
8532         struct rpc_bind_conn_calldata *p = calldata;
8533
8534         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8535 }
8536
8537 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8538 {
8539         struct rpc_bind_conn_calldata data = {
8540                 .clp = clp,
8541                 .cred = cred,
8542         };
8543         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8544                         nfs4_proc_bind_conn_to_session_callback, &data);
8545 }
8546
8547 /*
8548  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8549  * and operations we'd like to see to enable certain features in the allow map
8550  */
8551 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8552         .how = SP4_MACH_CRED,
8553         .enforce.u.words = {
8554                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8555                       1 << (OP_EXCHANGE_ID - 32) |
8556                       1 << (OP_CREATE_SESSION - 32) |
8557                       1 << (OP_DESTROY_SESSION - 32) |
8558                       1 << (OP_DESTROY_CLIENTID - 32)
8559         },
8560         .allow.u.words = {
8561                 [0] = 1 << (OP_CLOSE) |
8562                       1 << (OP_OPEN_DOWNGRADE) |
8563                       1 << (OP_LOCKU) |
8564                       1 << (OP_DELEGRETURN) |
8565                       1 << (OP_COMMIT),
8566                 [1] = 1 << (OP_SECINFO - 32) |
8567                       1 << (OP_SECINFO_NO_NAME - 32) |
8568                       1 << (OP_LAYOUTRETURN - 32) |
8569                       1 << (OP_TEST_STATEID - 32) |
8570                       1 << (OP_FREE_STATEID - 32) |
8571                       1 << (OP_WRITE - 32)
8572         }
8573 };
8574
8575 /*
8576  * Select the state protection mode for client `clp' given the server results
8577  * from exchange_id in `sp'.
8578  *
8579  * Returns 0 on success, negative errno otherwise.
8580  */
8581 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8582                                  struct nfs41_state_protection *sp)
8583 {
8584         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8585                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8586                       1 << (OP_EXCHANGE_ID - 32) |
8587                       1 << (OP_CREATE_SESSION - 32) |
8588                       1 << (OP_DESTROY_SESSION - 32) |
8589                       1 << (OP_DESTROY_CLIENTID - 32)
8590         };
8591         unsigned long flags = 0;
8592         unsigned int i;
8593         int ret = 0;
8594
8595         if (sp->how == SP4_MACH_CRED) {
8596                 /* Print state protect result */
8597                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8598                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8599                         if (test_bit(i, sp->enforce.u.longs))
8600                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8601                         if (test_bit(i, sp->allow.u.longs))
8602                                 dfprintk(MOUNT, "  allow op %d\n", i);
8603                 }
8604
8605                 /* make sure nothing is on enforce list that isn't supported */
8606                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8607                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8608                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8609                                 ret = -EINVAL;
8610                                 goto out;
8611                         }
8612                 }
8613
8614                 /*
8615                  * Minimal mode - state operations are allowed to use machine
8616                  * credential.  Note this already happens by default, so the
8617                  * client doesn't have to do anything more than the negotiation.
8618                  *
8619                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8620                  *       we're already using the machine cred for exchange_id
8621                  *       and will never use a different cred.
8622                  */
8623                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8624                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8625                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8626                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8627                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8628                         dfprintk(MOUNT, "  minimal mode enabled\n");
8629                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8630                 } else {
8631                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8632                         ret = -EINVAL;
8633                         goto out;
8634                 }
8635
8636                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8637                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8638                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8639                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8640                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8641                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8642                 }
8643
8644                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8645                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8646                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8647                 }
8648
8649                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8650                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8651                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8652                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8653                 }
8654
8655                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8656                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8657                         dfprintk(MOUNT, "  stateid mode enabled\n");
8658                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8659                 }
8660
8661                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8662                         dfprintk(MOUNT, "  write mode enabled\n");
8663                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8664                 }
8665
8666                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8667                         dfprintk(MOUNT, "  commit mode enabled\n");
8668                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8669                 }
8670         }
8671 out:
8672         clp->cl_sp4_flags = flags;
8673         return ret;
8674 }
8675
8676 struct nfs41_exchange_id_data {
8677         struct nfs41_exchange_id_res res;
8678         struct nfs41_exchange_id_args args;
8679 };
8680
8681 static void nfs4_exchange_id_release(void *data)
8682 {
8683         struct nfs41_exchange_id_data *cdata =
8684                                         (struct nfs41_exchange_id_data *)data;
8685
8686         nfs_put_client(cdata->args.client);
8687         kfree(cdata->res.impl_id);
8688         kfree(cdata->res.server_scope);
8689         kfree(cdata->res.server_owner);
8690         kfree(cdata);
8691 }
8692
8693 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8694         .rpc_release = nfs4_exchange_id_release,
8695 };
8696
8697 /*
8698  * _nfs4_proc_exchange_id()
8699  *
8700  * Wrapper for EXCHANGE_ID operation.
8701  */
8702 static struct rpc_task *
8703 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8704                         u32 sp4_how, struct rpc_xprt *xprt)
8705 {
8706         struct rpc_message msg = {
8707                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8708                 .rpc_cred = cred,
8709         };
8710         struct rpc_task_setup task_setup_data = {
8711                 .rpc_client = clp->cl_rpcclient,
8712                 .callback_ops = &nfs4_exchange_id_call_ops,
8713                 .rpc_message = &msg,
8714                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8715         };
8716         struct nfs41_exchange_id_data *calldata;
8717         int status;
8718
8719         if (!refcount_inc_not_zero(&clp->cl_count))
8720                 return ERR_PTR(-EIO);
8721
8722         status = -ENOMEM;
8723         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8724         if (!calldata)
8725                 goto out;
8726
8727         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8728
8729         status = nfs4_init_uniform_client_string(clp);
8730         if (status)
8731                 goto out_calldata;
8732
8733         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8734                                                 GFP_NOFS);
8735         status = -ENOMEM;
8736         if (unlikely(calldata->res.server_owner == NULL))
8737                 goto out_calldata;
8738
8739         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8740                                         GFP_NOFS);
8741         if (unlikely(calldata->res.server_scope == NULL))
8742                 goto out_server_owner;
8743
8744         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8745         if (unlikely(calldata->res.impl_id == NULL))
8746                 goto out_server_scope;
8747
8748         switch (sp4_how) {
8749         case SP4_NONE:
8750                 calldata->args.state_protect.how = SP4_NONE;
8751                 break;
8752
8753         case SP4_MACH_CRED:
8754                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8755                 break;
8756
8757         default:
8758                 /* unsupported! */
8759                 WARN_ON_ONCE(1);
8760                 status = -EINVAL;
8761                 goto out_impl_id;
8762         }
8763         if (xprt) {
8764                 task_setup_data.rpc_xprt = xprt;
8765                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8766                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8767                                 sizeof(calldata->args.verifier.data));
8768         }
8769         calldata->args.client = clp;
8770         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8771         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8772 #ifdef CONFIG_NFS_V4_1_MIGRATION
8773         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8774 #endif
8775         msg.rpc_argp = &calldata->args;
8776         msg.rpc_resp = &calldata->res;
8777         task_setup_data.callback_data = calldata;
8778
8779         return rpc_run_task(&task_setup_data);
8780
8781 out_impl_id:
8782         kfree(calldata->res.impl_id);
8783 out_server_scope:
8784         kfree(calldata->res.server_scope);
8785 out_server_owner:
8786         kfree(calldata->res.server_owner);
8787 out_calldata:
8788         kfree(calldata);
8789 out:
8790         nfs_put_client(clp);
8791         return ERR_PTR(status);
8792 }
8793
8794 /*
8795  * _nfs4_proc_exchange_id()
8796  *
8797  * Wrapper for EXCHANGE_ID operation.
8798  */
8799 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8800                         u32 sp4_how)
8801 {
8802         struct rpc_task *task;
8803         struct nfs41_exchange_id_args *argp;
8804         struct nfs41_exchange_id_res *resp;
8805         unsigned long now = jiffies;
8806         int status;
8807
8808         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8809         if (IS_ERR(task))
8810                 return PTR_ERR(task);
8811
8812         argp = task->tk_msg.rpc_argp;
8813         resp = task->tk_msg.rpc_resp;
8814         status = task->tk_status;
8815         if (status  != 0)
8816                 goto out;
8817
8818         status = nfs4_check_cl_exchange_flags(resp->flags,
8819                         clp->cl_mvops->minor_version);
8820         if (status  != 0)
8821                 goto out;
8822
8823         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8824         if (status != 0)
8825                 goto out;
8826
8827         do_renew_lease(clp, now);
8828
8829         clp->cl_clientid = resp->clientid;
8830         clp->cl_exchange_flags = resp->flags;
8831         clp->cl_seqid = resp->seqid;
8832         /* Client ID is not confirmed */
8833         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8834                 clear_bit(NFS4_SESSION_ESTABLISHED,
8835                           &clp->cl_session->session_state);
8836
8837         if (clp->cl_serverscope != NULL &&
8838             !nfs41_same_server_scope(clp->cl_serverscope,
8839                                 resp->server_scope)) {
8840                 dprintk("%s: server_scope mismatch detected\n",
8841                         __func__);
8842                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8843         }
8844
8845         swap(clp->cl_serverowner, resp->server_owner);
8846         swap(clp->cl_serverscope, resp->server_scope);
8847         swap(clp->cl_implid, resp->impl_id);
8848
8849         /* Save the EXCHANGE_ID verifier session trunk tests */
8850         memcpy(clp->cl_confirm.data, argp->verifier.data,
8851                sizeof(clp->cl_confirm.data));
8852 out:
8853         trace_nfs4_exchange_id(clp, status);
8854         rpc_put_task(task);
8855         return status;
8856 }
8857
8858 /*
8859  * nfs4_proc_exchange_id()
8860  *
8861  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8862  *
8863  * Since the clientid has expired, all compounds using sessions
8864  * associated with the stale clientid will be returning
8865  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8866  * be in some phase of session reset.
8867  *
8868  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8869  */
8870 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8871 {
8872         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8873         int status;
8874
8875         /* try SP4_MACH_CRED if krb5i/p */
8876         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8877             authflavor == RPC_AUTH_GSS_KRB5P) {
8878                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8879                 if (!status)
8880                         return 0;
8881         }
8882
8883         /* try SP4_NONE */
8884         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8885 }
8886
8887 /**
8888  * nfs4_test_session_trunk
8889  *
8890  * This is an add_xprt_test() test function called from
8891  * rpc_clnt_setup_test_and_add_xprt.
8892  *
8893  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8894  * and is dereferrenced in nfs4_exchange_id_release
8895  *
8896  * Upon success, add the new transport to the rpc_clnt
8897  *
8898  * @clnt: struct rpc_clnt to get new transport
8899  * @xprt: the rpc_xprt to test
8900  * @data: call data for _nfs4_proc_exchange_id.
8901  */
8902 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8903                             void *data)
8904 {
8905         struct nfs4_add_xprt_data *adata = data;
8906         struct rpc_task *task;
8907         int status;
8908
8909         u32 sp4_how;
8910
8911         dprintk("--> %s try %s\n", __func__,
8912                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8913
8914         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8915
8916         /* Test connection for session trunking. Async exchange_id call */
8917         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8918         if (IS_ERR(task))
8919                 return;
8920
8921         status = task->tk_status;
8922         if (status == 0)
8923                 status = nfs4_detect_session_trunking(adata->clp,
8924                                 task->tk_msg.rpc_resp, xprt);
8925
8926         if (status == 0)
8927                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8928         else if (rpc_clnt_xprt_switch_has_addr(clnt,
8929                                 (struct sockaddr *)&xprt->addr))
8930                 rpc_clnt_xprt_switch_remove_xprt(clnt, xprt);
8931
8932         rpc_put_task(task);
8933 }
8934 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8935
8936 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8937                 const struct cred *cred)
8938 {
8939         struct rpc_message msg = {
8940                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8941                 .rpc_argp = clp,
8942                 .rpc_cred = cred,
8943         };
8944         int status;
8945
8946         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8947                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8948         trace_nfs4_destroy_clientid(clp, status);
8949         if (status)
8950                 dprintk("NFS: Got error %d from the server %s on "
8951                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8952         return status;
8953 }
8954
8955 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8956                 const struct cred *cred)
8957 {
8958         unsigned int loop;
8959         int ret;
8960
8961         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8962                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8963                 switch (ret) {
8964                 case -NFS4ERR_DELAY:
8965                 case -NFS4ERR_CLIENTID_BUSY:
8966                         ssleep(1);
8967                         break;
8968                 default:
8969                         return ret;
8970                 }
8971         }
8972         return 0;
8973 }
8974
8975 int nfs4_destroy_clientid(struct nfs_client *clp)
8976 {
8977         const struct cred *cred;
8978         int ret = 0;
8979
8980         if (clp->cl_mvops->minor_version < 1)
8981                 goto out;
8982         if (clp->cl_exchange_flags == 0)
8983                 goto out;
8984         if (clp->cl_preserve_clid)
8985                 goto out;
8986         cred = nfs4_get_clid_cred(clp);
8987         ret = nfs4_proc_destroy_clientid(clp, cred);
8988         put_cred(cred);
8989         switch (ret) {
8990         case 0:
8991         case -NFS4ERR_STALE_CLIENTID:
8992                 clp->cl_exchange_flags = 0;
8993         }
8994 out:
8995         return ret;
8996 }
8997
8998 #endif /* CONFIG_NFS_V4_1 */
8999
9000 struct nfs4_get_lease_time_data {
9001         struct nfs4_get_lease_time_args *args;
9002         struct nfs4_get_lease_time_res *res;
9003         struct nfs_client *clp;
9004 };
9005
9006 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
9007                                         void *calldata)
9008 {
9009         struct nfs4_get_lease_time_data *data =
9010                         (struct nfs4_get_lease_time_data *)calldata;
9011
9012         /* just setup sequence, do not trigger session recovery
9013            since we're invoked within one */
9014         nfs4_setup_sequence(data->clp,
9015                         &data->args->la_seq_args,
9016                         &data->res->lr_seq_res,
9017                         task);
9018 }
9019
9020 /*
9021  * Called from nfs4_state_manager thread for session setup, so don't recover
9022  * from sequence operation or clientid errors.
9023  */
9024 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9025 {
9026         struct nfs4_get_lease_time_data *data =
9027                         (struct nfs4_get_lease_time_data *)calldata;
9028
9029         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9030                 return;
9031         switch (task->tk_status) {
9032         case -NFS4ERR_DELAY:
9033         case -NFS4ERR_GRACE:
9034                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
9035                 task->tk_status = 0;
9036                 fallthrough;
9037         case -NFS4ERR_RETRY_UNCACHED_REP:
9038                 rpc_restart_call_prepare(task);
9039                 return;
9040         }
9041 }
9042
9043 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9044         .rpc_call_prepare = nfs4_get_lease_time_prepare,
9045         .rpc_call_done = nfs4_get_lease_time_done,
9046 };
9047
9048 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9049 {
9050         struct nfs4_get_lease_time_args args;
9051         struct nfs4_get_lease_time_res res = {
9052                 .lr_fsinfo = fsinfo,
9053         };
9054         struct nfs4_get_lease_time_data data = {
9055                 .args = &args,
9056                 .res = &res,
9057                 .clp = clp,
9058         };
9059         struct rpc_message msg = {
9060                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9061                 .rpc_argp = &args,
9062                 .rpc_resp = &res,
9063         };
9064         struct rpc_task_setup task_setup = {
9065                 .rpc_client = clp->cl_rpcclient,
9066                 .rpc_message = &msg,
9067                 .callback_ops = &nfs4_get_lease_time_ops,
9068                 .callback_data = &data,
9069                 .flags = RPC_TASK_TIMEOUT,
9070         };
9071
9072         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
9073         return nfs4_call_sync_custom(&task_setup);
9074 }
9075
9076 #ifdef CONFIG_NFS_V4_1
9077
9078 /*
9079  * Initialize the values to be used by the client in CREATE_SESSION
9080  * If nfs4_init_session set the fore channel request and response sizes,
9081  * use them.
9082  *
9083  * Set the back channel max_resp_sz_cached to zero to force the client to
9084  * always set csa_cachethis to FALSE because the current implementation
9085  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9086  */
9087 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9088                                     struct rpc_clnt *clnt)
9089 {
9090         unsigned int max_rqst_sz, max_resp_sz;
9091         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9092         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9093
9094         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9095         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9096
9097         /* Fore channel attributes */
9098         args->fc_attrs.max_rqst_sz = max_rqst_sz;
9099         args->fc_attrs.max_resp_sz = max_resp_sz;
9100         args->fc_attrs.max_ops = NFS4_MAX_OPS;
9101         args->fc_attrs.max_reqs = max_session_slots;
9102
9103         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9104                 "max_ops=%u max_reqs=%u\n",
9105                 __func__,
9106                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9107                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9108
9109         /* Back channel attributes */
9110         args->bc_attrs.max_rqst_sz = max_bc_payload;
9111         args->bc_attrs.max_resp_sz = max_bc_payload;
9112         args->bc_attrs.max_resp_sz_cached = 0;
9113         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9114         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9115         if (args->bc_attrs.max_reqs > max_bc_slots)
9116                 args->bc_attrs.max_reqs = max_bc_slots;
9117
9118         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9119                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9120                 __func__,
9121                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9122                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9123                 args->bc_attrs.max_reqs);
9124 }
9125
9126 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9127                 struct nfs41_create_session_res *res)
9128 {
9129         struct nfs4_channel_attrs *sent = &args->fc_attrs;
9130         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9131
9132         if (rcvd->max_resp_sz > sent->max_resp_sz)
9133                 return -EINVAL;
9134         /*
9135          * Our requested max_ops is the minimum we need; we're not
9136          * prepared to break up compounds into smaller pieces than that.
9137          * So, no point even trying to continue if the server won't
9138          * cooperate:
9139          */
9140         if (rcvd->max_ops < sent->max_ops)
9141                 return -EINVAL;
9142         if (rcvd->max_reqs == 0)
9143                 return -EINVAL;
9144         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9145                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9146         return 0;
9147 }
9148
9149 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9150                 struct nfs41_create_session_res *res)
9151 {
9152         struct nfs4_channel_attrs *sent = &args->bc_attrs;
9153         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9154
9155         if (!(res->flags & SESSION4_BACK_CHAN))
9156                 goto out;
9157         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9158                 return -EINVAL;
9159         if (rcvd->max_resp_sz < sent->max_resp_sz)
9160                 return -EINVAL;
9161         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9162                 return -EINVAL;
9163         if (rcvd->max_ops > sent->max_ops)
9164                 return -EINVAL;
9165         if (rcvd->max_reqs > sent->max_reqs)
9166                 return -EINVAL;
9167 out:
9168         return 0;
9169 }
9170
9171 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9172                                      struct nfs41_create_session_res *res)
9173 {
9174         int ret;
9175
9176         ret = nfs4_verify_fore_channel_attrs(args, res);
9177         if (ret)
9178                 return ret;
9179         return nfs4_verify_back_channel_attrs(args, res);
9180 }
9181
9182 static void nfs4_update_session(struct nfs4_session *session,
9183                 struct nfs41_create_session_res *res)
9184 {
9185         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9186         /* Mark client id and session as being confirmed */
9187         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9188         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9189         session->flags = res->flags;
9190         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9191         if (res->flags & SESSION4_BACK_CHAN)
9192                 memcpy(&session->bc_attrs, &res->bc_attrs,
9193                                 sizeof(session->bc_attrs));
9194 }
9195
9196 static int _nfs4_proc_create_session(struct nfs_client *clp,
9197                 const struct cred *cred)
9198 {
9199         struct nfs4_session *session = clp->cl_session;
9200         struct nfs41_create_session_args args = {
9201                 .client = clp,
9202                 .clientid = clp->cl_clientid,
9203                 .seqid = clp->cl_seqid,
9204                 .cb_program = NFS4_CALLBACK,
9205         };
9206         struct nfs41_create_session_res res;
9207
9208         struct rpc_message msg = {
9209                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9210                 .rpc_argp = &args,
9211                 .rpc_resp = &res,
9212                 .rpc_cred = cred,
9213         };
9214         int status;
9215
9216         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9217         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9218
9219         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9220                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9221         trace_nfs4_create_session(clp, status);
9222
9223         switch (status) {
9224         case -NFS4ERR_STALE_CLIENTID:
9225         case -NFS4ERR_DELAY:
9226         case -ETIMEDOUT:
9227         case -EACCES:
9228         case -EAGAIN:
9229                 goto out;
9230         }
9231
9232         clp->cl_seqid++;
9233         if (!status) {
9234                 /* Verify the session's negotiated channel_attrs values */
9235                 status = nfs4_verify_channel_attrs(&args, &res);
9236                 /* Increment the clientid slot sequence id */
9237                 if (status)
9238                         goto out;
9239                 nfs4_update_session(session, &res);
9240         }
9241 out:
9242         return status;
9243 }
9244
9245 /*
9246  * Issues a CREATE_SESSION operation to the server.
9247  * It is the responsibility of the caller to verify the session is
9248  * expired before calling this routine.
9249  */
9250 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9251 {
9252         int status;
9253         unsigned *ptr;
9254         struct nfs4_session *session = clp->cl_session;
9255         struct nfs4_add_xprt_data xprtdata = {
9256                 .clp = clp,
9257         };
9258         struct rpc_add_xprt_test rpcdata = {
9259                 .add_xprt_test = clp->cl_mvops->session_trunk,
9260                 .data = &xprtdata,
9261         };
9262
9263         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9264
9265         status = _nfs4_proc_create_session(clp, cred);
9266         if (status)
9267                 goto out;
9268
9269         /* Init or reset the session slot tables */
9270         status = nfs4_setup_session_slot_tables(session);
9271         dprintk("slot table setup returned %d\n", status);
9272         if (status)
9273                 goto out;
9274
9275         ptr = (unsigned *)&session->sess_id.data[0];
9276         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9277                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9278         rpc_clnt_probe_trunked_xprts(clp->cl_rpcclient, &rpcdata);
9279 out:
9280         return status;
9281 }
9282
9283 /*
9284  * Issue the over-the-wire RPC DESTROY_SESSION.
9285  * The caller must serialize access to this routine.
9286  */
9287 int nfs4_proc_destroy_session(struct nfs4_session *session,
9288                 const struct cred *cred)
9289 {
9290         struct rpc_message msg = {
9291                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9292                 .rpc_argp = session,
9293                 .rpc_cred = cred,
9294         };
9295         int status = 0;
9296
9297         /* session is still being setup */
9298         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9299                 return 0;
9300
9301         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9302                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9303         trace_nfs4_destroy_session(session->clp, status);
9304
9305         if (status)
9306                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9307                         "Session has been destroyed regardless...\n", status);
9308         rpc_clnt_manage_trunked_xprts(session->clp->cl_rpcclient);
9309         return status;
9310 }
9311
9312 /*
9313  * Renew the cl_session lease.
9314  */
9315 struct nfs4_sequence_data {
9316         struct nfs_client *clp;
9317         struct nfs4_sequence_args args;
9318         struct nfs4_sequence_res res;
9319 };
9320
9321 static void nfs41_sequence_release(void *data)
9322 {
9323         struct nfs4_sequence_data *calldata = data;
9324         struct nfs_client *clp = calldata->clp;
9325
9326         if (refcount_read(&clp->cl_count) > 1)
9327                 nfs4_schedule_state_renewal(clp);
9328         nfs_put_client(clp);
9329         kfree(calldata);
9330 }
9331
9332 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9333 {
9334         switch(task->tk_status) {
9335         case -NFS4ERR_DELAY:
9336                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9337                 return -EAGAIN;
9338         default:
9339                 nfs4_schedule_lease_recovery(clp);
9340         }
9341         return 0;
9342 }
9343
9344 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9345 {
9346         struct nfs4_sequence_data *calldata = data;
9347         struct nfs_client *clp = calldata->clp;
9348
9349         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9350                 return;
9351
9352         trace_nfs4_sequence(clp, task->tk_status);
9353         if (task->tk_status < 0) {
9354                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9355                 if (refcount_read(&clp->cl_count) == 1)
9356                         return;
9357
9358                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9359                         rpc_restart_call_prepare(task);
9360                         return;
9361                 }
9362         }
9363         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9364 }
9365
9366 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9367 {
9368         struct nfs4_sequence_data *calldata = data;
9369         struct nfs_client *clp = calldata->clp;
9370         struct nfs4_sequence_args *args;
9371         struct nfs4_sequence_res *res;
9372
9373         args = task->tk_msg.rpc_argp;
9374         res = task->tk_msg.rpc_resp;
9375
9376         nfs4_setup_sequence(clp, args, res, task);
9377 }
9378
9379 static const struct rpc_call_ops nfs41_sequence_ops = {
9380         .rpc_call_done = nfs41_sequence_call_done,
9381         .rpc_call_prepare = nfs41_sequence_prepare,
9382         .rpc_release = nfs41_sequence_release,
9383 };
9384
9385 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9386                 const struct cred *cred,
9387                 struct nfs4_slot *slot,
9388                 bool is_privileged)
9389 {
9390         struct nfs4_sequence_data *calldata;
9391         struct rpc_message msg = {
9392                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9393                 .rpc_cred = cred,
9394         };
9395         struct rpc_task_setup task_setup_data = {
9396                 .rpc_client = clp->cl_rpcclient,
9397                 .rpc_message = &msg,
9398                 .callback_ops = &nfs41_sequence_ops,
9399                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9400         };
9401         struct rpc_task *ret;
9402
9403         ret = ERR_PTR(-EIO);
9404         if (!refcount_inc_not_zero(&clp->cl_count))
9405                 goto out_err;
9406
9407         ret = ERR_PTR(-ENOMEM);
9408         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
9409         if (calldata == NULL)
9410                 goto out_put_clp;
9411         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9412         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9413         msg.rpc_argp = &calldata->args;
9414         msg.rpc_resp = &calldata->res;
9415         calldata->clp = clp;
9416         task_setup_data.callback_data = calldata;
9417
9418         ret = rpc_run_task(&task_setup_data);
9419         if (IS_ERR(ret))
9420                 goto out_err;
9421         return ret;
9422 out_put_clp:
9423         nfs_put_client(clp);
9424 out_err:
9425         nfs41_release_slot(slot);
9426         return ret;
9427 }
9428
9429 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9430 {
9431         struct rpc_task *task;
9432         int ret = 0;
9433
9434         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9435                 return -EAGAIN;
9436         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9437         if (IS_ERR(task))
9438                 ret = PTR_ERR(task);
9439         else
9440                 rpc_put_task_async(task);
9441         dprintk("<-- %s status=%d\n", __func__, ret);
9442         return ret;
9443 }
9444
9445 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9446 {
9447         struct rpc_task *task;
9448         int ret;
9449
9450         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9451         if (IS_ERR(task)) {
9452                 ret = PTR_ERR(task);
9453                 goto out;
9454         }
9455         ret = rpc_wait_for_completion_task(task);
9456         if (!ret)
9457                 ret = task->tk_status;
9458         rpc_put_task(task);
9459 out:
9460         dprintk("<-- %s status=%d\n", __func__, ret);
9461         return ret;
9462 }
9463
9464 struct nfs4_reclaim_complete_data {
9465         struct nfs_client *clp;
9466         struct nfs41_reclaim_complete_args arg;
9467         struct nfs41_reclaim_complete_res res;
9468 };
9469
9470 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9471 {
9472         struct nfs4_reclaim_complete_data *calldata = data;
9473
9474         nfs4_setup_sequence(calldata->clp,
9475                         &calldata->arg.seq_args,
9476                         &calldata->res.seq_res,
9477                         task);
9478 }
9479
9480 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9481 {
9482         switch(task->tk_status) {
9483         case 0:
9484                 wake_up_all(&clp->cl_lock_waitq);
9485                 fallthrough;
9486         case -NFS4ERR_COMPLETE_ALREADY:
9487         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9488                 break;
9489         case -NFS4ERR_DELAY:
9490                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9491                 fallthrough;
9492         case -NFS4ERR_RETRY_UNCACHED_REP:
9493         case -EACCES:
9494                 dprintk("%s: failed to reclaim complete error %d for server %s, retrying\n",
9495                         __func__, task->tk_status, clp->cl_hostname);
9496                 return -EAGAIN;
9497         case -NFS4ERR_BADSESSION:
9498         case -NFS4ERR_DEADSESSION:
9499         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9500                 break;
9501         default:
9502                 nfs4_schedule_lease_recovery(clp);
9503         }
9504         return 0;
9505 }
9506
9507 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9508 {
9509         struct nfs4_reclaim_complete_data *calldata = data;
9510         struct nfs_client *clp = calldata->clp;
9511         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9512
9513         if (!nfs41_sequence_done(task, res))
9514                 return;
9515
9516         trace_nfs4_reclaim_complete(clp, task->tk_status);
9517         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9518                 rpc_restart_call_prepare(task);
9519                 return;
9520         }
9521 }
9522
9523 static void nfs4_free_reclaim_complete_data(void *data)
9524 {
9525         struct nfs4_reclaim_complete_data *calldata = data;
9526
9527         kfree(calldata);
9528 }
9529
9530 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9531         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9532         .rpc_call_done = nfs4_reclaim_complete_done,
9533         .rpc_release = nfs4_free_reclaim_complete_data,
9534 };
9535
9536 /*
9537  * Issue a global reclaim complete.
9538  */
9539 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9540                 const struct cred *cred)
9541 {
9542         struct nfs4_reclaim_complete_data *calldata;
9543         struct rpc_message msg = {
9544                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9545                 .rpc_cred = cred,
9546         };
9547         struct rpc_task_setup task_setup_data = {
9548                 .rpc_client = clp->cl_rpcclient,
9549                 .rpc_message = &msg,
9550                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9551                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9552         };
9553         int status = -ENOMEM;
9554
9555         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9556         if (calldata == NULL)
9557                 goto out;
9558         calldata->clp = clp;
9559         calldata->arg.one_fs = 0;
9560
9561         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9562         msg.rpc_argp = &calldata->arg;
9563         msg.rpc_resp = &calldata->res;
9564         task_setup_data.callback_data = calldata;
9565         status = nfs4_call_sync_custom(&task_setup_data);
9566 out:
9567         dprintk("<-- %s status=%d\n", __func__, status);
9568         return status;
9569 }
9570
9571 static void
9572 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9573 {
9574         struct nfs4_layoutget *lgp = calldata;
9575         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9576
9577         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9578                                 &lgp->res.seq_res, task);
9579 }
9580
9581 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9582 {
9583         struct nfs4_layoutget *lgp = calldata;
9584
9585         nfs41_sequence_process(task, &lgp->res.seq_res);
9586 }
9587
9588 static int
9589 nfs4_layoutget_handle_exception(struct rpc_task *task,
9590                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9591 {
9592         struct inode *inode = lgp->args.inode;
9593         struct nfs_server *server = NFS_SERVER(inode);
9594         struct pnfs_layout_hdr *lo = lgp->lo;
9595         int nfs4err = task->tk_status;
9596         int err, status = 0;
9597         LIST_HEAD(head);
9598
9599         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9600
9601         nfs4_sequence_free_slot(&lgp->res.seq_res);
9602
9603         switch (nfs4err) {
9604         case 0:
9605                 goto out;
9606
9607         /*
9608          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9609          * on the file. set tk_status to -ENODATA to tell upper layer to
9610          * retry go inband.
9611          */
9612         case -NFS4ERR_LAYOUTUNAVAILABLE:
9613                 status = -ENODATA;
9614                 goto out;
9615         /*
9616          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9617          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9618          */
9619         case -NFS4ERR_BADLAYOUT:
9620                 status = -EOVERFLOW;
9621                 goto out;
9622         /*
9623          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9624          * (or clients) writing to the same RAID stripe except when
9625          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9626          *
9627          * Treat it like we would RECALLCONFLICT -- we retry for a little
9628          * while, and then eventually give up.
9629          */
9630         case -NFS4ERR_LAYOUTTRYLATER:
9631                 if (lgp->args.minlength == 0) {
9632                         status = -EOVERFLOW;
9633                         goto out;
9634                 }
9635                 status = -EBUSY;
9636                 break;
9637         case -NFS4ERR_RECALLCONFLICT:
9638                 status = -ERECALLCONFLICT;
9639                 break;
9640         case -NFS4ERR_DELEG_REVOKED:
9641         case -NFS4ERR_ADMIN_REVOKED:
9642         case -NFS4ERR_EXPIRED:
9643         case -NFS4ERR_BAD_STATEID:
9644                 exception->timeout = 0;
9645                 spin_lock(&inode->i_lock);
9646                 /* If the open stateid was bad, then recover it. */
9647                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9648                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9649                         spin_unlock(&inode->i_lock);
9650                         exception->state = lgp->args.ctx->state;
9651                         exception->stateid = &lgp->args.stateid;
9652                         break;
9653                 }
9654
9655                 /*
9656                  * Mark the bad layout state as invalid, then retry
9657                  */
9658                 pnfs_mark_layout_stateid_invalid(lo, &head);
9659                 spin_unlock(&inode->i_lock);
9660                 nfs_commit_inode(inode, 0);
9661                 pnfs_free_lseg_list(&head);
9662                 status = -EAGAIN;
9663                 goto out;
9664         }
9665
9666         err = nfs4_handle_exception(server, nfs4err, exception);
9667         if (!status) {
9668                 if (exception->retry)
9669                         status = -EAGAIN;
9670                 else
9671                         status = err;
9672         }
9673 out:
9674         return status;
9675 }
9676
9677 size_t max_response_pages(struct nfs_server *server)
9678 {
9679         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9680         return nfs_page_array_len(0, max_resp_sz);
9681 }
9682
9683 static void nfs4_layoutget_release(void *calldata)
9684 {
9685         struct nfs4_layoutget *lgp = calldata;
9686
9687         nfs4_sequence_free_slot(&lgp->res.seq_res);
9688         pnfs_layoutget_free(lgp);
9689 }
9690
9691 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9692         .rpc_call_prepare = nfs4_layoutget_prepare,
9693         .rpc_call_done = nfs4_layoutget_done,
9694         .rpc_release = nfs4_layoutget_release,
9695 };
9696
9697 struct pnfs_layout_segment *
9698 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9699 {
9700         struct inode *inode = lgp->args.inode;
9701         struct nfs_server *server = NFS_SERVER(inode);
9702         struct rpc_task *task;
9703         struct rpc_message msg = {
9704                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9705                 .rpc_argp = &lgp->args,
9706                 .rpc_resp = &lgp->res,
9707                 .rpc_cred = lgp->cred,
9708         };
9709         struct rpc_task_setup task_setup_data = {
9710                 .rpc_client = server->client,
9711                 .rpc_message = &msg,
9712                 .callback_ops = &nfs4_layoutget_call_ops,
9713                 .callback_data = lgp,
9714                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9715                          RPC_TASK_MOVEABLE,
9716         };
9717         struct pnfs_layout_segment *lseg = NULL;
9718         struct nfs4_exception exception = {
9719                 .inode = inode,
9720                 .timeout = *timeout,
9721         };
9722         int status = 0;
9723
9724         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9725
9726         task = rpc_run_task(&task_setup_data);
9727         if (IS_ERR(task))
9728                 return ERR_CAST(task);
9729
9730         status = rpc_wait_for_completion_task(task);
9731         if (status != 0)
9732                 goto out;
9733
9734         if (task->tk_status < 0) {
9735                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9736                 *timeout = exception.timeout;
9737         } else if (lgp->res.layoutp->len == 0) {
9738                 status = -EAGAIN;
9739                 *timeout = nfs4_update_delay(&exception.timeout);
9740         } else
9741                 lseg = pnfs_layout_process(lgp);
9742 out:
9743         trace_nfs4_layoutget(lgp->args.ctx,
9744                         &lgp->args.range,
9745                         &lgp->res.range,
9746                         &lgp->res.stateid,
9747                         status);
9748
9749         rpc_put_task(task);
9750         dprintk("<-- %s status=%d\n", __func__, status);
9751         if (status)
9752                 return ERR_PTR(status);
9753         return lseg;
9754 }
9755
9756 static void
9757 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9758 {
9759         struct nfs4_layoutreturn *lrp = calldata;
9760
9761         nfs4_setup_sequence(lrp->clp,
9762                         &lrp->args.seq_args,
9763                         &lrp->res.seq_res,
9764                         task);
9765         if (!pnfs_layout_is_valid(lrp->args.layout))
9766                 rpc_exit(task, 0);
9767 }
9768
9769 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9770 {
9771         struct nfs4_layoutreturn *lrp = calldata;
9772         struct nfs_server *server;
9773
9774         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9775                 return;
9776
9777         /*
9778          * Was there an RPC level error? Assume the call succeeded,
9779          * and that we need to release the layout
9780          */
9781         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9782                 lrp->res.lrs_present = 0;
9783                 return;
9784         }
9785
9786         server = NFS_SERVER(lrp->args.inode);
9787         switch (task->tk_status) {
9788         case -NFS4ERR_OLD_STATEID:
9789                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9790                                         &lrp->args.range,
9791                                         lrp->args.inode))
9792                         goto out_restart;
9793                 fallthrough;
9794         default:
9795                 task->tk_status = 0;
9796                 fallthrough;
9797         case 0:
9798                 break;
9799         case -NFS4ERR_DELAY:
9800                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9801                         break;
9802                 goto out_restart;
9803         }
9804         return;
9805 out_restart:
9806         task->tk_status = 0;
9807         nfs4_sequence_free_slot(&lrp->res.seq_res);
9808         rpc_restart_call_prepare(task);
9809 }
9810
9811 static void nfs4_layoutreturn_release(void *calldata)
9812 {
9813         struct nfs4_layoutreturn *lrp = calldata;
9814         struct pnfs_layout_hdr *lo = lrp->args.layout;
9815
9816         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9817                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9818         nfs4_sequence_free_slot(&lrp->res.seq_res);
9819         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9820                 lrp->ld_private.ops->free(&lrp->ld_private);
9821         pnfs_put_layout_hdr(lrp->args.layout);
9822         nfs_iput_and_deactive(lrp->inode);
9823         put_cred(lrp->cred);
9824         kfree(calldata);
9825 }
9826
9827 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9828         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9829         .rpc_call_done = nfs4_layoutreturn_done,
9830         .rpc_release = nfs4_layoutreturn_release,
9831 };
9832
9833 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9834 {
9835         struct rpc_task *task;
9836         struct rpc_message msg = {
9837                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9838                 .rpc_argp = &lrp->args,
9839                 .rpc_resp = &lrp->res,
9840                 .rpc_cred = lrp->cred,
9841         };
9842         struct rpc_task_setup task_setup_data = {
9843                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9844                 .rpc_message = &msg,
9845                 .callback_ops = &nfs4_layoutreturn_call_ops,
9846                 .callback_data = lrp,
9847                 .flags = RPC_TASK_MOVEABLE,
9848         };
9849         int status = 0;
9850
9851         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9852                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9853                         &task_setup_data.rpc_client, &msg);
9854
9855         lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9856         if (!sync) {
9857                 if (!lrp->inode) {
9858                         nfs4_layoutreturn_release(lrp);
9859                         return -EAGAIN;
9860                 }
9861                 task_setup_data.flags |= RPC_TASK_ASYNC;
9862         }
9863         if (!lrp->inode)
9864                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9865                                    1);
9866         else
9867                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9868                                    0);
9869         task = rpc_run_task(&task_setup_data);
9870         if (IS_ERR(task))
9871                 return PTR_ERR(task);
9872         if (sync)
9873                 status = task->tk_status;
9874         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9875         dprintk("<-- %s status=%d\n", __func__, status);
9876         rpc_put_task(task);
9877         return status;
9878 }
9879
9880 static int
9881 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9882                 struct pnfs_device *pdev,
9883                 const struct cred *cred)
9884 {
9885         struct nfs4_getdeviceinfo_args args = {
9886                 .pdev = pdev,
9887                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9888                         NOTIFY_DEVICEID4_DELETE,
9889         };
9890         struct nfs4_getdeviceinfo_res res = {
9891                 .pdev = pdev,
9892         };
9893         struct rpc_message msg = {
9894                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9895                 .rpc_argp = &args,
9896                 .rpc_resp = &res,
9897                 .rpc_cred = cred,
9898         };
9899         int status;
9900
9901         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9902         if (res.notification & ~args.notify_types)
9903                 dprintk("%s: unsupported notification\n", __func__);
9904         if (res.notification != args.notify_types)
9905                 pdev->nocache = 1;
9906
9907         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9908
9909         dprintk("<-- %s status=%d\n", __func__, status);
9910
9911         return status;
9912 }
9913
9914 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9915                 struct pnfs_device *pdev,
9916                 const struct cred *cred)
9917 {
9918         struct nfs4_exception exception = { };
9919         int err;
9920
9921         do {
9922                 err = nfs4_handle_exception(server,
9923                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9924                                         &exception);
9925         } while (exception.retry);
9926         return err;
9927 }
9928 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9929
9930 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9931 {
9932         struct nfs4_layoutcommit_data *data = calldata;
9933         struct nfs_server *server = NFS_SERVER(data->args.inode);
9934
9935         nfs4_setup_sequence(server->nfs_client,
9936                         &data->args.seq_args,
9937                         &data->res.seq_res,
9938                         task);
9939 }
9940
9941 static void
9942 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9943 {
9944         struct nfs4_layoutcommit_data *data = calldata;
9945         struct nfs_server *server = NFS_SERVER(data->args.inode);
9946
9947         if (!nfs41_sequence_done(task, &data->res.seq_res))
9948                 return;
9949
9950         switch (task->tk_status) { /* Just ignore these failures */
9951         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9952         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9953         case -NFS4ERR_BADLAYOUT:     /* no layout */
9954         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9955                 task->tk_status = 0;
9956                 break;
9957         case 0:
9958                 break;
9959         default:
9960                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9961                         rpc_restart_call_prepare(task);
9962                         return;
9963                 }
9964         }
9965 }
9966
9967 static void nfs4_layoutcommit_release(void *calldata)
9968 {
9969         struct nfs4_layoutcommit_data *data = calldata;
9970
9971         pnfs_cleanup_layoutcommit(data);
9972         nfs_post_op_update_inode_force_wcc(data->args.inode,
9973                                            data->res.fattr);
9974         put_cred(data->cred);
9975         nfs_iput_and_deactive(data->inode);
9976         kfree(data);
9977 }
9978
9979 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9980         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9981         .rpc_call_done = nfs4_layoutcommit_done,
9982         .rpc_release = nfs4_layoutcommit_release,
9983 };
9984
9985 int
9986 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9987 {
9988         struct rpc_message msg = {
9989                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9990                 .rpc_argp = &data->args,
9991                 .rpc_resp = &data->res,
9992                 .rpc_cred = data->cred,
9993         };
9994         struct rpc_task_setup task_setup_data = {
9995                 .task = &data->task,
9996                 .rpc_client = NFS_CLIENT(data->args.inode),
9997                 .rpc_message = &msg,
9998                 .callback_ops = &nfs4_layoutcommit_ops,
9999                 .callback_data = data,
10000                 .flags = RPC_TASK_MOVEABLE,
10001         };
10002         struct rpc_task *task;
10003         int status = 0;
10004
10005         dprintk("NFS: initiating layoutcommit call. sync %d "
10006                 "lbw: %llu inode %lu\n", sync,
10007                 data->args.lastbytewritten,
10008                 data->args.inode->i_ino);
10009
10010         if (!sync) {
10011                 data->inode = nfs_igrab_and_active(data->args.inode);
10012                 if (data->inode == NULL) {
10013                         nfs4_layoutcommit_release(data);
10014                         return -EAGAIN;
10015                 }
10016                 task_setup_data.flags = RPC_TASK_ASYNC;
10017         }
10018         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
10019         task = rpc_run_task(&task_setup_data);
10020         if (IS_ERR(task))
10021                 return PTR_ERR(task);
10022         if (sync)
10023                 status = task->tk_status;
10024         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10025         dprintk("%s: status %d\n", __func__, status);
10026         rpc_put_task(task);
10027         return status;
10028 }
10029
10030 /*
10031  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10032  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10033  */
10034 static int
10035 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10036                     struct nfs_fsinfo *info,
10037                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
10038 {
10039         struct nfs41_secinfo_no_name_args args = {
10040                 .style = SECINFO_STYLE_CURRENT_FH,
10041         };
10042         struct nfs4_secinfo_res res = {
10043                 .flavors = flavors,
10044         };
10045         struct rpc_message msg = {
10046                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10047                 .rpc_argp = &args,
10048                 .rpc_resp = &res,
10049         };
10050         struct nfs4_call_sync_data data = {
10051                 .seq_server = server,
10052                 .seq_args = &args.seq_args,
10053                 .seq_res = &res.seq_res,
10054         };
10055         struct rpc_task_setup task_setup = {
10056                 .rpc_client = server->client,
10057                 .rpc_message = &msg,
10058                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
10059                 .callback_data = &data,
10060                 .flags = RPC_TASK_NO_ROUND_ROBIN,
10061         };
10062         const struct cred *cred = NULL;
10063         int status;
10064
10065         if (use_integrity) {
10066                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
10067
10068                 cred = nfs4_get_clid_cred(server->nfs_client);
10069                 msg.rpc_cred = cred;
10070         }
10071
10072         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
10073         status = nfs4_call_sync_custom(&task_setup);
10074         dprintk("<-- %s status=%d\n", __func__, status);
10075
10076         put_cred(cred);
10077
10078         return status;
10079 }
10080
10081 static int
10082 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10083                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
10084 {
10085         struct nfs4_exception exception = {
10086                 .interruptible = true,
10087         };
10088         int err;
10089         do {
10090                 /* first try using integrity protection */
10091                 err = -NFS4ERR_WRONGSEC;
10092
10093                 /* try to use integrity protection with machine cred */
10094                 if (_nfs4_is_integrity_protected(server->nfs_client))
10095                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10096                                                           flavors, true);
10097
10098                 /*
10099                  * if unable to use integrity protection, or SECINFO with
10100                  * integrity protection returns NFS4ERR_WRONGSEC (which is
10101                  * disallowed by spec, but exists in deployed servers) use
10102                  * the current filesystem's rpc_client and the user cred.
10103                  */
10104                 if (err == -NFS4ERR_WRONGSEC)
10105                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10106                                                           flavors, false);
10107
10108                 switch (err) {
10109                 case 0:
10110                 case -NFS4ERR_WRONGSEC:
10111                 case -ENOTSUPP:
10112                         goto out;
10113                 default:
10114                         err = nfs4_handle_exception(server, err, &exception);
10115                 }
10116         } while (exception.retry);
10117 out:
10118         return err;
10119 }
10120
10121 static int
10122 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10123                     struct nfs_fsinfo *info)
10124 {
10125         int err;
10126         struct page *page;
10127         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10128         struct nfs4_secinfo_flavors *flavors;
10129         struct nfs4_secinfo4 *secinfo;
10130         int i;
10131
10132         page = alloc_page(GFP_KERNEL);
10133         if (!page) {
10134                 err = -ENOMEM;
10135                 goto out;
10136         }
10137
10138         flavors = page_address(page);
10139         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10140
10141         /*
10142          * Fall back on "guess and check" method if
10143          * the server doesn't support SECINFO_NO_NAME
10144          */
10145         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10146                 err = nfs4_find_root_sec(server, fhandle, info);
10147                 goto out_freepage;
10148         }
10149         if (err)
10150                 goto out_freepage;
10151
10152         for (i = 0; i < flavors->num_flavors; i++) {
10153                 secinfo = &flavors->flavors[i];
10154
10155                 switch (secinfo->flavor) {
10156                 case RPC_AUTH_NULL:
10157                 case RPC_AUTH_UNIX:
10158                 case RPC_AUTH_GSS:
10159                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10160                                         &secinfo->flavor_info);
10161                         break;
10162                 default:
10163                         flavor = RPC_AUTH_MAXFLAVOR;
10164                         break;
10165                 }
10166
10167                 if (!nfs_auth_info_match(&server->auth_info, flavor))
10168                         flavor = RPC_AUTH_MAXFLAVOR;
10169
10170                 if (flavor != RPC_AUTH_MAXFLAVOR) {
10171                         err = nfs4_lookup_root_sec(server, fhandle,
10172                                                    info, flavor);
10173                         if (!err)
10174                                 break;
10175                 }
10176         }
10177
10178         if (flavor == RPC_AUTH_MAXFLAVOR)
10179                 err = -EPERM;
10180
10181 out_freepage:
10182         put_page(page);
10183         if (err == -EACCES)
10184                 return -EPERM;
10185 out:
10186         return err;
10187 }
10188
10189 static int _nfs41_test_stateid(struct nfs_server *server,
10190                 nfs4_stateid *stateid,
10191                 const struct cred *cred)
10192 {
10193         int status;
10194         struct nfs41_test_stateid_args args = {
10195                 .stateid = stateid,
10196         };
10197         struct nfs41_test_stateid_res res;
10198         struct rpc_message msg = {
10199                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10200                 .rpc_argp = &args,
10201                 .rpc_resp = &res,
10202                 .rpc_cred = cred,
10203         };
10204         struct rpc_clnt *rpc_client = server->client;
10205
10206         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10207                 &rpc_client, &msg);
10208
10209         dprintk("NFS call  test_stateid %p\n", stateid);
10210         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10211         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10212                         &args.seq_args, &res.seq_res);
10213         if (status != NFS_OK) {
10214                 dprintk("NFS reply test_stateid: failed, %d\n", status);
10215                 return status;
10216         }
10217         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10218         return -res.status;
10219 }
10220
10221 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10222                 int err, struct nfs4_exception *exception)
10223 {
10224         exception->retry = 0;
10225         switch(err) {
10226         case -NFS4ERR_DELAY:
10227         case -NFS4ERR_RETRY_UNCACHED_REP:
10228                 nfs4_handle_exception(server, err, exception);
10229                 break;
10230         case -NFS4ERR_BADSESSION:
10231         case -NFS4ERR_BADSLOT:
10232         case -NFS4ERR_BAD_HIGH_SLOT:
10233         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10234         case -NFS4ERR_DEADSESSION:
10235                 nfs4_do_handle_exception(server, err, exception);
10236         }
10237 }
10238
10239 /**
10240  * nfs41_test_stateid - perform a TEST_STATEID operation
10241  *
10242  * @server: server / transport on which to perform the operation
10243  * @stateid: state ID to test
10244  * @cred: credential
10245  *
10246  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10247  * Otherwise a negative NFS4ERR value is returned if the operation
10248  * failed or the state ID is not currently valid.
10249  */
10250 static int nfs41_test_stateid(struct nfs_server *server,
10251                 nfs4_stateid *stateid,
10252                 const struct cred *cred)
10253 {
10254         struct nfs4_exception exception = {
10255                 .interruptible = true,
10256         };
10257         int err;
10258         do {
10259                 err = _nfs41_test_stateid(server, stateid, cred);
10260                 nfs4_handle_delay_or_session_error(server, err, &exception);
10261         } while (exception.retry);
10262         return err;
10263 }
10264
10265 struct nfs_free_stateid_data {
10266         struct nfs_server *server;
10267         struct nfs41_free_stateid_args args;
10268         struct nfs41_free_stateid_res res;
10269 };
10270
10271 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10272 {
10273         struct nfs_free_stateid_data *data = calldata;
10274         nfs4_setup_sequence(data->server->nfs_client,
10275                         &data->args.seq_args,
10276                         &data->res.seq_res,
10277                         task);
10278 }
10279
10280 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10281 {
10282         struct nfs_free_stateid_data *data = calldata;
10283
10284         nfs41_sequence_done(task, &data->res.seq_res);
10285
10286         switch (task->tk_status) {
10287         case -NFS4ERR_DELAY:
10288                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10289                         rpc_restart_call_prepare(task);
10290         }
10291 }
10292
10293 static void nfs41_free_stateid_release(void *calldata)
10294 {
10295         struct nfs_free_stateid_data *data = calldata;
10296         struct nfs_client *clp = data->server->nfs_client;
10297
10298         nfs_put_client(clp);
10299         kfree(calldata);
10300 }
10301
10302 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10303         .rpc_call_prepare = nfs41_free_stateid_prepare,
10304         .rpc_call_done = nfs41_free_stateid_done,
10305         .rpc_release = nfs41_free_stateid_release,
10306 };
10307
10308 /**
10309  * nfs41_free_stateid - perform a FREE_STATEID operation
10310  *
10311  * @server: server / transport on which to perform the operation
10312  * @stateid: state ID to release
10313  * @cred: credential
10314  * @privileged: set to true if this call needs to be privileged
10315  *
10316  * Note: this function is always asynchronous.
10317  */
10318 static int nfs41_free_stateid(struct nfs_server *server,
10319                 const nfs4_stateid *stateid,
10320                 const struct cred *cred,
10321                 bool privileged)
10322 {
10323         struct rpc_message msg = {
10324                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10325                 .rpc_cred = cred,
10326         };
10327         struct rpc_task_setup task_setup = {
10328                 .rpc_client = server->client,
10329                 .rpc_message = &msg,
10330                 .callback_ops = &nfs41_free_stateid_ops,
10331                 .flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10332         };
10333         struct nfs_free_stateid_data *data;
10334         struct rpc_task *task;
10335         struct nfs_client *clp = server->nfs_client;
10336
10337         if (!refcount_inc_not_zero(&clp->cl_count))
10338                 return -EIO;
10339
10340         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10341                 &task_setup.rpc_client, &msg);
10342
10343         dprintk("NFS call  free_stateid %p\n", stateid);
10344         data = kmalloc(sizeof(*data), GFP_KERNEL);
10345         if (!data)
10346                 return -ENOMEM;
10347         data->server = server;
10348         nfs4_stateid_copy(&data->args.stateid, stateid);
10349
10350         task_setup.callback_data = data;
10351
10352         msg.rpc_argp = &data->args;
10353         msg.rpc_resp = &data->res;
10354         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10355         task = rpc_run_task(&task_setup);
10356         if (IS_ERR(task))
10357                 return PTR_ERR(task);
10358         rpc_put_task(task);
10359         return 0;
10360 }
10361
10362 static void
10363 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10364 {
10365         const struct cred *cred = lsp->ls_state->owner->so_cred;
10366
10367         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10368         nfs4_free_lock_state(server, lsp);
10369 }
10370
10371 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10372                 const nfs4_stateid *s2)
10373 {
10374         if (s1->type != s2->type)
10375                 return false;
10376
10377         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10378                 return false;
10379
10380         if (s1->seqid == s2->seqid)
10381                 return true;
10382
10383         return s1->seqid == 0 || s2->seqid == 0;
10384 }
10385
10386 #endif /* CONFIG_NFS_V4_1 */
10387
10388 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10389                 const nfs4_stateid *s2)
10390 {
10391         return nfs4_stateid_match(s1, s2);
10392 }
10393
10394
10395 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10396         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10397         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10398         .recover_open   = nfs4_open_reclaim,
10399         .recover_lock   = nfs4_lock_reclaim,
10400         .establish_clid = nfs4_init_clientid,
10401         .detect_trunking = nfs40_discover_server_trunking,
10402 };
10403
10404 #if defined(CONFIG_NFS_V4_1)
10405 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10406         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10407         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10408         .recover_open   = nfs4_open_reclaim,
10409         .recover_lock   = nfs4_lock_reclaim,
10410         .establish_clid = nfs41_init_clientid,
10411         .reclaim_complete = nfs41_proc_reclaim_complete,
10412         .detect_trunking = nfs41_discover_server_trunking,
10413 };
10414 #endif /* CONFIG_NFS_V4_1 */
10415
10416 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10417         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10418         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10419         .recover_open   = nfs40_open_expired,
10420         .recover_lock   = nfs4_lock_expired,
10421         .establish_clid = nfs4_init_clientid,
10422 };
10423
10424 #if defined(CONFIG_NFS_V4_1)
10425 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10426         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10427         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10428         .recover_open   = nfs41_open_expired,
10429         .recover_lock   = nfs41_lock_expired,
10430         .establish_clid = nfs41_init_clientid,
10431 };
10432 #endif /* CONFIG_NFS_V4_1 */
10433
10434 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10435         .sched_state_renewal = nfs4_proc_async_renew,
10436         .get_state_renewal_cred = nfs4_get_renew_cred,
10437         .renew_lease = nfs4_proc_renew,
10438 };
10439
10440 #if defined(CONFIG_NFS_V4_1)
10441 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10442         .sched_state_renewal = nfs41_proc_async_sequence,
10443         .get_state_renewal_cred = nfs4_get_machine_cred,
10444         .renew_lease = nfs4_proc_sequence,
10445 };
10446 #endif
10447
10448 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10449         .get_locations = _nfs40_proc_get_locations,
10450         .fsid_present = _nfs40_proc_fsid_present,
10451 };
10452
10453 #if defined(CONFIG_NFS_V4_1)
10454 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10455         .get_locations = _nfs41_proc_get_locations,
10456         .fsid_present = _nfs41_proc_fsid_present,
10457 };
10458 #endif  /* CONFIG_NFS_V4_1 */
10459
10460 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10461         .minor_version = 0,
10462         .init_caps = NFS_CAP_READDIRPLUS
10463                 | NFS_CAP_ATOMIC_OPEN
10464                 | NFS_CAP_POSIX_LOCK,
10465         .init_client = nfs40_init_client,
10466         .shutdown_client = nfs40_shutdown_client,
10467         .match_stateid = nfs4_match_stateid,
10468         .find_root_sec = nfs4_find_root_sec,
10469         .free_lock_state = nfs4_release_lockowner,
10470         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10471         .alloc_seqid = nfs_alloc_seqid,
10472         .call_sync_ops = &nfs40_call_sync_ops,
10473         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10474         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10475         .state_renewal_ops = &nfs40_state_renewal_ops,
10476         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10477 };
10478
10479 #if defined(CONFIG_NFS_V4_1)
10480 static struct nfs_seqid *
10481 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10482 {
10483         return NULL;
10484 }
10485
10486 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10487         .minor_version = 1,
10488         .init_caps = NFS_CAP_READDIRPLUS
10489                 | NFS_CAP_ATOMIC_OPEN
10490                 | NFS_CAP_POSIX_LOCK
10491                 | NFS_CAP_STATEID_NFSV41
10492                 | NFS_CAP_ATOMIC_OPEN_V1
10493                 | NFS_CAP_LGOPEN
10494                 | NFS_CAP_MOVEABLE,
10495         .init_client = nfs41_init_client,
10496         .shutdown_client = nfs41_shutdown_client,
10497         .match_stateid = nfs41_match_stateid,
10498         .find_root_sec = nfs41_find_root_sec,
10499         .free_lock_state = nfs41_free_lock_state,
10500         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10501         .alloc_seqid = nfs_alloc_no_seqid,
10502         .session_trunk = nfs4_test_session_trunk,
10503         .call_sync_ops = &nfs41_call_sync_ops,
10504         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10505         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10506         .state_renewal_ops = &nfs41_state_renewal_ops,
10507         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10508 };
10509 #endif
10510
10511 #if defined(CONFIG_NFS_V4_2)
10512 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10513         .minor_version = 2,
10514         .init_caps = NFS_CAP_READDIRPLUS
10515                 | NFS_CAP_ATOMIC_OPEN
10516                 | NFS_CAP_POSIX_LOCK
10517                 | NFS_CAP_STATEID_NFSV41
10518                 | NFS_CAP_ATOMIC_OPEN_V1
10519                 | NFS_CAP_LGOPEN
10520                 | NFS_CAP_ALLOCATE
10521                 | NFS_CAP_COPY
10522                 | NFS_CAP_OFFLOAD_CANCEL
10523                 | NFS_CAP_COPY_NOTIFY
10524                 | NFS_CAP_DEALLOCATE
10525                 | NFS_CAP_SEEK
10526                 | NFS_CAP_LAYOUTSTATS
10527                 | NFS_CAP_CLONE
10528                 | NFS_CAP_LAYOUTERROR
10529                 | NFS_CAP_READ_PLUS
10530                 | NFS_CAP_MOVEABLE,
10531         .init_client = nfs41_init_client,
10532         .shutdown_client = nfs41_shutdown_client,
10533         .match_stateid = nfs41_match_stateid,
10534         .find_root_sec = nfs41_find_root_sec,
10535         .free_lock_state = nfs41_free_lock_state,
10536         .call_sync_ops = &nfs41_call_sync_ops,
10537         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10538         .alloc_seqid = nfs_alloc_no_seqid,
10539         .session_trunk = nfs4_test_session_trunk,
10540         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10541         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10542         .state_renewal_ops = &nfs41_state_renewal_ops,
10543         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10544 };
10545 #endif
10546
10547 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10548         [0] = &nfs_v4_0_minor_ops,
10549 #if defined(CONFIG_NFS_V4_1)
10550         [1] = &nfs_v4_1_minor_ops,
10551 #endif
10552 #if defined(CONFIG_NFS_V4_2)
10553         [2] = &nfs_v4_2_minor_ops,
10554 #endif
10555 };
10556
10557 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10558 {
10559         ssize_t error, error2, error3;
10560
10561         error = generic_listxattr(dentry, list, size);
10562         if (error < 0)
10563                 return error;
10564         if (list) {
10565                 list += error;
10566                 size -= error;
10567         }
10568
10569         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10570         if (error2 < 0)
10571                 return error2;
10572
10573         if (list) {
10574                 list += error2;
10575                 size -= error2;
10576         }
10577
10578         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10579         if (error3 < 0)
10580                 return error3;
10581
10582         return error + error2 + error3;
10583 }
10584
10585 static void nfs4_enable_swap(struct inode *inode)
10586 {
10587         /* The state manager thread must always be running.
10588          * It will notice the client is a swapper, and stay put.
10589          */
10590         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10591
10592         nfs4_schedule_state_manager(clp);
10593 }
10594
10595 static void nfs4_disable_swap(struct inode *inode)
10596 {
10597         /* The state manager thread will now exit once it is
10598          * woken.
10599          */
10600         wake_up_var(&NFS_SERVER(inode)->nfs_client->cl_state);
10601 }
10602
10603 static const struct inode_operations nfs4_dir_inode_operations = {
10604         .create         = nfs_create,
10605         .lookup         = nfs_lookup,
10606         .atomic_open    = nfs_atomic_open,
10607         .link           = nfs_link,
10608         .unlink         = nfs_unlink,
10609         .symlink        = nfs_symlink,
10610         .mkdir          = nfs_mkdir,
10611         .rmdir          = nfs_rmdir,
10612         .mknod          = nfs_mknod,
10613         .rename         = nfs_rename,
10614         .permission     = nfs_permission,
10615         .getattr        = nfs_getattr,
10616         .setattr        = nfs_setattr,
10617         .listxattr      = nfs4_listxattr,
10618 };
10619
10620 static const struct inode_operations nfs4_file_inode_operations = {
10621         .permission     = nfs_permission,
10622         .getattr        = nfs_getattr,
10623         .setattr        = nfs_setattr,
10624         .listxattr      = nfs4_listxattr,
10625 };
10626
10627 const struct nfs_rpc_ops nfs_v4_clientops = {
10628         .version        = 4,                    /* protocol version */
10629         .dentry_ops     = &nfs4_dentry_operations,
10630         .dir_inode_ops  = &nfs4_dir_inode_operations,
10631         .file_inode_ops = &nfs4_file_inode_operations,
10632         .file_ops       = &nfs4_file_operations,
10633         .getroot        = nfs4_proc_get_root,
10634         .submount       = nfs4_submount,
10635         .try_get_tree   = nfs4_try_get_tree,
10636         .getattr        = nfs4_proc_getattr,
10637         .setattr        = nfs4_proc_setattr,
10638         .lookup         = nfs4_proc_lookup,
10639         .lookupp        = nfs4_proc_lookupp,
10640         .access         = nfs4_proc_access,
10641         .readlink       = nfs4_proc_readlink,
10642         .create         = nfs4_proc_create,
10643         .remove         = nfs4_proc_remove,
10644         .unlink_setup   = nfs4_proc_unlink_setup,
10645         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10646         .unlink_done    = nfs4_proc_unlink_done,
10647         .rename_setup   = nfs4_proc_rename_setup,
10648         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10649         .rename_done    = nfs4_proc_rename_done,
10650         .link           = nfs4_proc_link,
10651         .symlink        = nfs4_proc_symlink,
10652         .mkdir          = nfs4_proc_mkdir,
10653         .rmdir          = nfs4_proc_rmdir,
10654         .readdir        = nfs4_proc_readdir,
10655         .mknod          = nfs4_proc_mknod,
10656         .statfs         = nfs4_proc_statfs,
10657         .fsinfo         = nfs4_proc_fsinfo,
10658         .pathconf       = nfs4_proc_pathconf,
10659         .set_capabilities = nfs4_server_capabilities,
10660         .decode_dirent  = nfs4_decode_dirent,
10661         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10662         .read_setup     = nfs4_proc_read_setup,
10663         .read_done      = nfs4_read_done,
10664         .write_setup    = nfs4_proc_write_setup,
10665         .write_done     = nfs4_write_done,
10666         .commit_setup   = nfs4_proc_commit_setup,
10667         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10668         .commit_done    = nfs4_commit_done,
10669         .lock           = nfs4_proc_lock,
10670         .clear_acl_cache = nfs4_zap_acl_attr,
10671         .close_context  = nfs4_close_context,
10672         .open_context   = nfs4_atomic_open,
10673         .have_delegation = nfs4_have_delegation,
10674         .alloc_client   = nfs4_alloc_client,
10675         .init_client    = nfs4_init_client,
10676         .free_client    = nfs4_free_client,
10677         .create_server  = nfs4_create_server,
10678         .clone_server   = nfs_clone_server,
10679         .discover_trunking = nfs4_discover_trunking,
10680         .enable_swap    = nfs4_enable_swap,
10681         .disable_swap   = nfs4_disable_swap,
10682 };
10683
10684 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10685         .name   = XATTR_NAME_NFSV4_ACL,
10686         .list   = nfs4_xattr_list_nfs4_acl,
10687         .get    = nfs4_xattr_get_nfs4_acl,
10688         .set    = nfs4_xattr_set_nfs4_acl,
10689 };
10690
10691 #if defined(CONFIG_NFS_V4_1)
10692 static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10693         .name   = XATTR_NAME_NFSV4_DACL,
10694         .list   = nfs4_xattr_list_nfs4_dacl,
10695         .get    = nfs4_xattr_get_nfs4_dacl,
10696         .set    = nfs4_xattr_set_nfs4_dacl,
10697 };
10698
10699 static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10700         .name   = XATTR_NAME_NFSV4_SACL,
10701         .list   = nfs4_xattr_list_nfs4_sacl,
10702         .get    = nfs4_xattr_get_nfs4_sacl,
10703         .set    = nfs4_xattr_set_nfs4_sacl,
10704 };
10705 #endif
10706
10707 #ifdef CONFIG_NFS_V4_2
10708 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10709         .prefix = XATTR_USER_PREFIX,
10710         .get    = nfs4_xattr_get_nfs4_user,
10711         .set    = nfs4_xattr_set_nfs4_user,
10712 };
10713 #endif
10714
10715 const struct xattr_handler *nfs4_xattr_handlers[] = {
10716         &nfs4_xattr_nfs4_acl_handler,
10717 #if defined(CONFIG_NFS_V4_1)
10718         &nfs4_xattr_nfs4_dacl_handler,
10719         &nfs4_xattr_nfs4_sacl_handler,
10720 #endif
10721 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10722         &nfs4_xattr_nfs4_label_handler,
10723 #endif
10724 #ifdef CONFIG_NFS_V4_2
10725         &nfs4_xattr_nfs4_user_handler,
10726 #endif
10727         NULL
10728 };