fix short copy handling in copy_mc_pipe_to_iter()
[sfrench/cifs-2.6.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         freezable_schedule_timeout_killable_unsafe(
420                 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         freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
431         if (!signal_pending(current))
432                 return 0;
433         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
434 }
435
436 static int nfs4_delay(long *timeout, bool interruptible)
437 {
438         if (interruptible)
439                 return nfs4_delay_interruptible(timeout);
440         return nfs4_delay_killable(timeout);
441 }
442
443 static const nfs4_stateid *
444 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
445 {
446         if (!stateid)
447                 return NULL;
448         switch (stateid->type) {
449         case NFS4_OPEN_STATEID_TYPE:
450         case NFS4_LOCK_STATEID_TYPE:
451         case NFS4_DELEGATION_STATEID_TYPE:
452                 return stateid;
453         default:
454                 break;
455         }
456         return NULL;
457 }
458
459 /* This is the error handling routine for processes that are allowed
460  * to sleep.
461  */
462 static int nfs4_do_handle_exception(struct nfs_server *server,
463                 int errorcode, struct nfs4_exception *exception)
464 {
465         struct nfs_client *clp = server->nfs_client;
466         struct nfs4_state *state = exception->state;
467         const nfs4_stateid *stateid;
468         struct inode *inode = exception->inode;
469         int ret = errorcode;
470
471         exception->delay = 0;
472         exception->recovering = 0;
473         exception->retry = 0;
474
475         stateid = nfs4_recoverable_stateid(exception->stateid);
476         if (stateid == NULL && state != NULL)
477                 stateid = nfs4_recoverable_stateid(&state->stateid);
478
479         switch(errorcode) {
480                 case 0:
481                         return 0;
482                 case -NFS4ERR_BADHANDLE:
483                 case -ESTALE:
484                         if (inode != NULL && S_ISREG(inode->i_mode))
485                                 pnfs_destroy_layout(NFS_I(inode));
486                         break;
487                 case -NFS4ERR_DELEG_REVOKED:
488                 case -NFS4ERR_ADMIN_REVOKED:
489                 case -NFS4ERR_EXPIRED:
490                 case -NFS4ERR_BAD_STATEID:
491                 case -NFS4ERR_PARTNER_NO_AUTH:
492                         if (inode != NULL && stateid != NULL) {
493                                 nfs_inode_find_state_and_recover(inode,
494                                                 stateid);
495                                 goto wait_on_recovery;
496                         }
497                         fallthrough;
498                 case -NFS4ERR_OPENMODE:
499                         if (inode) {
500                                 int err;
501
502                                 err = nfs_async_inode_return_delegation(inode,
503                                                 stateid);
504                                 if (err == 0)
505                                         goto wait_on_recovery;
506                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
507                                         exception->retry = 1;
508                                         break;
509                                 }
510                         }
511                         if (state == NULL)
512                                 break;
513                         ret = nfs4_schedule_stateid_recovery(server, state);
514                         if (ret < 0)
515                                 break;
516                         goto wait_on_recovery;
517                 case -NFS4ERR_STALE_STATEID:
518                 case -NFS4ERR_STALE_CLIENTID:
519                         nfs4_schedule_lease_recovery(clp);
520                         goto wait_on_recovery;
521                 case -NFS4ERR_MOVED:
522                         ret = nfs4_schedule_migration_recovery(server);
523                         if (ret < 0)
524                                 break;
525                         goto wait_on_recovery;
526                 case -NFS4ERR_LEASE_MOVED:
527                         nfs4_schedule_lease_moved_recovery(clp);
528                         goto wait_on_recovery;
529 #if defined(CONFIG_NFS_V4_1)
530                 case -NFS4ERR_BADSESSION:
531                 case -NFS4ERR_BADSLOT:
532                 case -NFS4ERR_BAD_HIGH_SLOT:
533                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
534                 case -NFS4ERR_DEADSESSION:
535                 case -NFS4ERR_SEQ_FALSE_RETRY:
536                 case -NFS4ERR_SEQ_MISORDERED:
537                         /* Handled in nfs41_sequence_process() */
538                         goto wait_on_recovery;
539 #endif /* defined(CONFIG_NFS_V4_1) */
540                 case -NFS4ERR_FILE_OPEN:
541                         if (exception->timeout > HZ) {
542                                 /* We have retried a decent amount, time to
543                                  * fail
544                                  */
545                                 ret = -EBUSY;
546                                 break;
547                         }
548                         fallthrough;
549                 case -NFS4ERR_DELAY:
550                         nfs_inc_server_stats(server, NFSIOS_DELAY);
551                         fallthrough;
552                 case -NFS4ERR_GRACE:
553                 case -NFS4ERR_LAYOUTTRYLATER:
554                 case -NFS4ERR_RECALLCONFLICT:
555                         exception->delay = 1;
556                         return 0;
557
558                 case -NFS4ERR_RETRY_UNCACHED_REP:
559                 case -NFS4ERR_OLD_STATEID:
560                         exception->retry = 1;
561                         break;
562                 case -NFS4ERR_BADOWNER:
563                         /* The following works around a Linux server bug! */
564                 case -NFS4ERR_BADNAME:
565                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
566                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
567                                 exception->retry = 1;
568                                 printk(KERN_WARNING "NFS: v4 server %s "
569                                                 "does not accept raw "
570                                                 "uid/gids. "
571                                                 "Reenabling the idmapper.\n",
572                                                 server->nfs_client->cl_hostname);
573                         }
574         }
575         /* We failed to handle the error */
576         return nfs4_map_errors(ret);
577 wait_on_recovery:
578         exception->recovering = 1;
579         return 0;
580 }
581
582 /* This is the error handling routine for processes that are allowed
583  * to sleep.
584  */
585 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
586 {
587         struct nfs_client *clp = server->nfs_client;
588         int ret;
589
590         ret = nfs4_do_handle_exception(server, errorcode, exception);
591         if (exception->delay) {
592                 ret = nfs4_delay(&exception->timeout,
593                                 exception->interruptible);
594                 goto out_retry;
595         }
596         if (exception->recovering) {
597                 if (exception->task_is_privileged)
598                         return -EDEADLOCK;
599                 ret = nfs4_wait_clnt_recover(clp);
600                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
601                         return -EIO;
602                 goto out_retry;
603         }
604         return ret;
605 out_retry:
606         if (ret == 0)
607                 exception->retry = 1;
608         return ret;
609 }
610
611 static int
612 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
613                 int errorcode, struct nfs4_exception *exception)
614 {
615         struct nfs_client *clp = server->nfs_client;
616         int ret;
617
618         ret = nfs4_do_handle_exception(server, errorcode, exception);
619         if (exception->delay) {
620                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
621                 goto out_retry;
622         }
623         if (exception->recovering) {
624                 if (exception->task_is_privileged)
625                         return -EDEADLOCK;
626                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
627                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
628                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
629                 goto out_retry;
630         }
631         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
632                 ret = -EIO;
633         return ret;
634 out_retry:
635         if (ret == 0) {
636                 exception->retry = 1;
637                 /*
638                  * For NFS4ERR_MOVED, the client transport will need to
639                  * be recomputed after migration recovery has completed.
640                  */
641                 if (errorcode == -NFS4ERR_MOVED)
642                         rpc_task_release_transport(task);
643         }
644         return ret;
645 }
646
647 int
648 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
649                         struct nfs4_state *state, long *timeout)
650 {
651         struct nfs4_exception exception = {
652                 .state = state,
653         };
654
655         if (task->tk_status >= 0)
656                 return 0;
657         if (timeout)
658                 exception.timeout = *timeout;
659         task->tk_status = nfs4_async_handle_exception(task, server,
660                         task->tk_status,
661                         &exception);
662         if (exception.delay && timeout)
663                 *timeout = exception.timeout;
664         if (exception.retry)
665                 return -EAGAIN;
666         return 0;
667 }
668
669 /*
670  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
671  * or 'false' otherwise.
672  */
673 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
674 {
675         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
676         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
677 }
678
679 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
680 {
681         spin_lock(&clp->cl_lock);
682         if (time_before(clp->cl_last_renewal,timestamp))
683                 clp->cl_last_renewal = timestamp;
684         spin_unlock(&clp->cl_lock);
685 }
686
687 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
688 {
689         struct nfs_client *clp = server->nfs_client;
690
691         if (!nfs4_has_session(clp))
692                 do_renew_lease(clp, timestamp);
693 }
694
695 struct nfs4_call_sync_data {
696         const struct nfs_server *seq_server;
697         struct nfs4_sequence_args *seq_args;
698         struct nfs4_sequence_res *seq_res;
699 };
700
701 void nfs4_init_sequence(struct nfs4_sequence_args *args,
702                         struct nfs4_sequence_res *res, int cache_reply,
703                         int privileged)
704 {
705         args->sa_slot = NULL;
706         args->sa_cache_this = cache_reply;
707         args->sa_privileged = privileged;
708
709         res->sr_slot = NULL;
710 }
711
712 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
713 {
714         struct nfs4_slot *slot = res->sr_slot;
715         struct nfs4_slot_table *tbl;
716
717         tbl = slot->table;
718         spin_lock(&tbl->slot_tbl_lock);
719         if (!nfs41_wake_and_assign_slot(tbl, slot))
720                 nfs4_free_slot(tbl, slot);
721         spin_unlock(&tbl->slot_tbl_lock);
722
723         res->sr_slot = NULL;
724 }
725
726 static int nfs40_sequence_done(struct rpc_task *task,
727                                struct nfs4_sequence_res *res)
728 {
729         if (res->sr_slot != NULL)
730                 nfs40_sequence_free_slot(res);
731         return 1;
732 }
733
734 #if defined(CONFIG_NFS_V4_1)
735
736 static void nfs41_release_slot(struct nfs4_slot *slot)
737 {
738         struct nfs4_session *session;
739         struct nfs4_slot_table *tbl;
740         bool send_new_highest_used_slotid = false;
741
742         if (!slot)
743                 return;
744         tbl = slot->table;
745         session = tbl->session;
746
747         /* Bump the slot sequence number */
748         if (slot->seq_done)
749                 slot->seq_nr++;
750         slot->seq_done = 0;
751
752         spin_lock(&tbl->slot_tbl_lock);
753         /* Be nice to the server: try to ensure that the last transmitted
754          * value for highest_user_slotid <= target_highest_slotid
755          */
756         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
757                 send_new_highest_used_slotid = true;
758
759         if (nfs41_wake_and_assign_slot(tbl, slot)) {
760                 send_new_highest_used_slotid = false;
761                 goto out_unlock;
762         }
763         nfs4_free_slot(tbl, slot);
764
765         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
766                 send_new_highest_used_slotid = false;
767 out_unlock:
768         spin_unlock(&tbl->slot_tbl_lock);
769         if (send_new_highest_used_slotid)
770                 nfs41_notify_server(session->clp);
771         if (waitqueue_active(&tbl->slot_waitq))
772                 wake_up_all(&tbl->slot_waitq);
773 }
774
775 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
776 {
777         nfs41_release_slot(res->sr_slot);
778         res->sr_slot = NULL;
779 }
780
781 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
782                 u32 seqnr)
783 {
784         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
785                 slot->seq_nr_highest_sent = seqnr;
786 }
787 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
788                 u32 seqnr)
789 {
790         slot->seq_nr_highest_sent = 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                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
858                 goto out_retry;
859         case -NFS4ERR_RETRY_UNCACHED_REP:
860         case -NFS4ERR_SEQ_FALSE_RETRY:
861                 /*
862                  * The server thinks we tried to replay a request.
863                  * Retry the call after bumping the sequence ID.
864                  */
865                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
866                 goto retry_new_seq;
867         case -NFS4ERR_BADSLOT:
868                 /*
869                  * The slot id we used was probably retired. Try again
870                  * using a different slot id.
871                  */
872                 if (slot->slot_nr < slot->table->target_highest_slotid)
873                         goto session_recover;
874                 goto retry_nowait;
875         case -NFS4ERR_SEQ_MISORDERED:
876                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
877                 /*
878                  * Were one or more calls using this slot interrupted?
879                  * If the server never received the request, then our
880                  * transmitted slot sequence number may be too high. However,
881                  * if the server did receive the request then it might
882                  * accidentally give us a reply with a mismatched operation.
883                  * We can sort this out by sending a lone sequence operation
884                  * to the server on the same slot.
885                  */
886                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
887                         slot->seq_nr--;
888                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
889                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
890                                 res->sr_slot = NULL;
891                         }
892                         goto retry_nowait;
893                 }
894                 /*
895                  * RFC5661:
896                  * A retry might be sent while the original request is
897                  * still in progress on the replier. The replier SHOULD
898                  * deal with the issue by returning NFS4ERR_DELAY as the
899                  * reply to SEQUENCE or CB_SEQUENCE operation, but
900                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
901                  *
902                  * Restart the search after a delay.
903                  */
904                 slot->seq_nr = slot->seq_nr_highest_sent;
905                 goto out_retry;
906         case -NFS4ERR_BADSESSION:
907         case -NFS4ERR_DEADSESSION:
908         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
909                 goto session_recover;
910         default:
911                 /* Just update the slot sequence no. */
912                 slot->seq_done = 1;
913         }
914 out:
915         /* The session may be reset by one of the error handlers. */
916         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
917 out_noaction:
918         return ret;
919 session_recover:
920         nfs4_schedule_session_recovery(session, status);
921         dprintk("%s ERROR: %d Reset session\n", __func__, status);
922         nfs41_sequence_free_slot(res);
923         goto out;
924 retry_new_seq:
925         ++slot->seq_nr;
926 retry_nowait:
927         if (rpc_restart_call_prepare(task)) {
928                 nfs41_sequence_free_slot(res);
929                 task->tk_status = 0;
930                 ret = 0;
931         }
932         goto out;
933 out_retry:
934         if (!rpc_restart_call(task))
935                 goto out;
936         rpc_delay(task, NFS4_POLL_RETRY_MAX);
937         return 0;
938 }
939
940 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
941 {
942         if (!nfs41_sequence_process(task, res))
943                 return 0;
944         if (res->sr_slot != NULL)
945                 nfs41_sequence_free_slot(res);
946         return 1;
947
948 }
949 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
950
951 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
952 {
953         if (res->sr_slot == NULL)
954                 return 1;
955         if (res->sr_slot->table->session != NULL)
956                 return nfs41_sequence_process(task, res);
957         return nfs40_sequence_done(task, res);
958 }
959
960 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
961 {
962         if (res->sr_slot != NULL) {
963                 if (res->sr_slot->table->session != NULL)
964                         nfs41_sequence_free_slot(res);
965                 else
966                         nfs40_sequence_free_slot(res);
967         }
968 }
969
970 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
971 {
972         if (res->sr_slot == NULL)
973                 return 1;
974         if (!res->sr_slot->table->session)
975                 return nfs40_sequence_done(task, res);
976         return nfs41_sequence_done(task, res);
977 }
978 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
979
980 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
981 {
982         struct nfs4_call_sync_data *data = calldata;
983
984         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
985
986         nfs4_setup_sequence(data->seq_server->nfs_client,
987                             data->seq_args, data->seq_res, task);
988 }
989
990 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
991 {
992         struct nfs4_call_sync_data *data = calldata;
993
994         nfs41_sequence_done(task, data->seq_res);
995 }
996
997 static const struct rpc_call_ops nfs41_call_sync_ops = {
998         .rpc_call_prepare = nfs41_call_sync_prepare,
999         .rpc_call_done = nfs41_call_sync_done,
1000 };
1001
1002 #else   /* !CONFIG_NFS_V4_1 */
1003
1004 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1005 {
1006         return nfs40_sequence_done(task, res);
1007 }
1008
1009 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1010 {
1011         if (res->sr_slot != NULL)
1012                 nfs40_sequence_free_slot(res);
1013 }
1014
1015 int nfs4_sequence_done(struct rpc_task *task,
1016                        struct nfs4_sequence_res *res)
1017 {
1018         return nfs40_sequence_done(task, res);
1019 }
1020 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1021
1022 #endif  /* !CONFIG_NFS_V4_1 */
1023
1024 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1025 {
1026         res->sr_timestamp = jiffies;
1027         res->sr_status_flags = 0;
1028         res->sr_status = 1;
1029 }
1030
1031 static
1032 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1033                 struct nfs4_sequence_res *res,
1034                 struct nfs4_slot *slot)
1035 {
1036         if (!slot)
1037                 return;
1038         slot->privileged = args->sa_privileged ? 1 : 0;
1039         args->sa_slot = slot;
1040
1041         res->sr_slot = slot;
1042 }
1043
1044 int nfs4_setup_sequence(struct nfs_client *client,
1045                         struct nfs4_sequence_args *args,
1046                         struct nfs4_sequence_res *res,
1047                         struct rpc_task *task)
1048 {
1049         struct nfs4_session *session = nfs4_get_session(client);
1050         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1051         struct nfs4_slot *slot;
1052
1053         /* slot already allocated? */
1054         if (res->sr_slot != NULL)
1055                 goto out_start;
1056
1057         if (session)
1058                 tbl = &session->fc_slot_table;
1059
1060         spin_lock(&tbl->slot_tbl_lock);
1061         /* The state manager will wait until the slot table is empty */
1062         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1063                 goto out_sleep;
1064
1065         slot = nfs4_alloc_slot(tbl);
1066         if (IS_ERR(slot)) {
1067                 if (slot == ERR_PTR(-ENOMEM))
1068                         goto out_sleep_timeout;
1069                 goto out_sleep;
1070         }
1071         spin_unlock(&tbl->slot_tbl_lock);
1072
1073         nfs4_sequence_attach_slot(args, res, slot);
1074
1075         trace_nfs4_setup_sequence(session, args);
1076 out_start:
1077         nfs41_sequence_res_init(res);
1078         rpc_call_start(task);
1079         return 0;
1080 out_sleep_timeout:
1081         /* Try again in 1/4 second */
1082         if (args->sa_privileged)
1083                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1084                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1085         else
1086                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1087                                 NULL, jiffies + (HZ >> 2));
1088         spin_unlock(&tbl->slot_tbl_lock);
1089         return -EAGAIN;
1090 out_sleep:
1091         if (args->sa_privileged)
1092                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1093                                 RPC_PRIORITY_PRIVILEGED);
1094         else
1095                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1096         spin_unlock(&tbl->slot_tbl_lock);
1097         return -EAGAIN;
1098 }
1099 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1100
1101 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1102 {
1103         struct nfs4_call_sync_data *data = calldata;
1104         nfs4_setup_sequence(data->seq_server->nfs_client,
1105                                 data->seq_args, data->seq_res, task);
1106 }
1107
1108 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1109 {
1110         struct nfs4_call_sync_data *data = calldata;
1111         nfs4_sequence_done(task, data->seq_res);
1112 }
1113
1114 static const struct rpc_call_ops nfs40_call_sync_ops = {
1115         .rpc_call_prepare = nfs40_call_sync_prepare,
1116         .rpc_call_done = nfs40_call_sync_done,
1117 };
1118
1119 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1120 {
1121         int ret;
1122         struct rpc_task *task;
1123
1124         task = rpc_run_task(task_setup);
1125         if (IS_ERR(task))
1126                 return PTR_ERR(task);
1127
1128         ret = task->tk_status;
1129         rpc_put_task(task);
1130         return ret;
1131 }
1132
1133 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1134                              struct nfs_server *server,
1135                              struct rpc_message *msg,
1136                              struct nfs4_sequence_args *args,
1137                              struct nfs4_sequence_res *res,
1138                              unsigned short task_flags)
1139 {
1140         struct nfs_client *clp = server->nfs_client;
1141         struct nfs4_call_sync_data data = {
1142                 .seq_server = server,
1143                 .seq_args = args,
1144                 .seq_res = res,
1145         };
1146         struct rpc_task_setup task_setup = {
1147                 .rpc_client = clnt,
1148                 .rpc_message = msg,
1149                 .callback_ops = clp->cl_mvops->call_sync_ops,
1150                 .callback_data = &data,
1151                 .flags = task_flags,
1152         };
1153
1154         return nfs4_call_sync_custom(&task_setup);
1155 }
1156
1157 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1158                                    struct nfs_server *server,
1159                                    struct rpc_message *msg,
1160                                    struct nfs4_sequence_args *args,
1161                                    struct nfs4_sequence_res *res)
1162 {
1163         unsigned short task_flags = 0;
1164
1165         if (server->caps & NFS_CAP_MOVEABLE)
1166                 task_flags = RPC_TASK_MOVEABLE;
1167         return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1168 }
1169
1170
1171 int nfs4_call_sync(struct rpc_clnt *clnt,
1172                    struct nfs_server *server,
1173                    struct rpc_message *msg,
1174                    struct nfs4_sequence_args *args,
1175                    struct nfs4_sequence_res *res,
1176                    int cache_reply)
1177 {
1178         nfs4_init_sequence(args, res, cache_reply, 0);
1179         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1180 }
1181
1182 static void
1183 nfs4_inc_nlink_locked(struct inode *inode)
1184 {
1185         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1186                                              NFS_INO_INVALID_CTIME |
1187                                              NFS_INO_INVALID_NLINK);
1188         inc_nlink(inode);
1189 }
1190
1191 static void
1192 nfs4_inc_nlink(struct inode *inode)
1193 {
1194         spin_lock(&inode->i_lock);
1195         nfs4_inc_nlink_locked(inode);
1196         spin_unlock(&inode->i_lock);
1197 }
1198
1199 static void
1200 nfs4_dec_nlink_locked(struct inode *inode)
1201 {
1202         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1203                                              NFS_INO_INVALID_CTIME |
1204                                              NFS_INO_INVALID_NLINK);
1205         drop_nlink(inode);
1206 }
1207
1208 static void
1209 nfs4_update_changeattr_locked(struct inode *inode,
1210                 struct nfs4_change_info *cinfo,
1211                 unsigned long timestamp, unsigned long cache_validity)
1212 {
1213         struct nfs_inode *nfsi = NFS_I(inode);
1214         u64 change_attr = inode_peek_iversion_raw(inode);
1215
1216         cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1217         if (S_ISDIR(inode->i_mode))
1218                 cache_validity |= NFS_INO_INVALID_DATA;
1219
1220         switch (NFS_SERVER(inode)->change_attr_type) {
1221         case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1222                 if (cinfo->after == change_attr)
1223                         goto out;
1224                 break;
1225         default:
1226                 if ((s64)(change_attr - cinfo->after) >= 0)
1227                         goto out;
1228         }
1229
1230         inode_set_iversion_raw(inode, cinfo->after);
1231         if (!cinfo->atomic || cinfo->before != change_attr) {
1232                 if (S_ISDIR(inode->i_mode))
1233                         nfs_force_lookup_revalidate(inode);
1234
1235                 if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1236                         cache_validity |=
1237                                 NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1238                                 NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1239                                 NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1240                                 NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR;
1241                 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1242         }
1243         nfsi->attrtimeo_timestamp = jiffies;
1244         nfsi->read_cache_jiffies = timestamp;
1245         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1246         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1247 out:
1248         nfs_set_cache_invalid(inode, cache_validity);
1249 }
1250
1251 void
1252 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1253                 unsigned long timestamp, unsigned long cache_validity)
1254 {
1255         spin_lock(&dir->i_lock);
1256         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1257         spin_unlock(&dir->i_lock);
1258 }
1259
1260 struct nfs4_open_createattrs {
1261         struct nfs4_label *label;
1262         struct iattr *sattr;
1263         const __u32 verf[2];
1264 };
1265
1266 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1267                 int err, struct nfs4_exception *exception)
1268 {
1269         if (err != -EINVAL)
1270                 return false;
1271         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1272                 return false;
1273         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1274         exception->retry = 1;
1275         return true;
1276 }
1277
1278 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1279 {
1280          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1281 }
1282
1283 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1284 {
1285         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1286
1287         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1288 }
1289
1290 static u32
1291 nfs4_map_atomic_open_share(struct nfs_server *server,
1292                 fmode_t fmode, int openflags)
1293 {
1294         u32 res = 0;
1295
1296         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1297         case FMODE_READ:
1298                 res = NFS4_SHARE_ACCESS_READ;
1299                 break;
1300         case FMODE_WRITE:
1301                 res = NFS4_SHARE_ACCESS_WRITE;
1302                 break;
1303         case FMODE_READ|FMODE_WRITE:
1304                 res = NFS4_SHARE_ACCESS_BOTH;
1305         }
1306         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1307                 goto out;
1308         /* Want no delegation if we're using O_DIRECT */
1309         if (openflags & O_DIRECT)
1310                 res |= NFS4_SHARE_WANT_NO_DELEG;
1311 out:
1312         return res;
1313 }
1314
1315 static enum open_claim_type4
1316 nfs4_map_atomic_open_claim(struct nfs_server *server,
1317                 enum open_claim_type4 claim)
1318 {
1319         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1320                 return claim;
1321         switch (claim) {
1322         default:
1323                 return claim;
1324         case NFS4_OPEN_CLAIM_FH:
1325                 return NFS4_OPEN_CLAIM_NULL;
1326         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1327                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1328         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1329                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1330         }
1331 }
1332
1333 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1334 {
1335         p->o_res.f_attr = &p->f_attr;
1336         p->o_res.seqid = p->o_arg.seqid;
1337         p->c_res.seqid = p->c_arg.seqid;
1338         p->o_res.server = p->o_arg.server;
1339         p->o_res.access_request = p->o_arg.access;
1340         nfs_fattr_init(&p->f_attr);
1341         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1342 }
1343
1344 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1345                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1346                 const struct nfs4_open_createattrs *c,
1347                 enum open_claim_type4 claim,
1348                 gfp_t gfp_mask)
1349 {
1350         struct dentry *parent = dget_parent(dentry);
1351         struct inode *dir = d_inode(parent);
1352         struct nfs_server *server = NFS_SERVER(dir);
1353         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1354         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1355         struct nfs4_opendata *p;
1356
1357         p = kzalloc(sizeof(*p), gfp_mask);
1358         if (p == NULL)
1359                 goto err;
1360
1361         p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1362         if (IS_ERR(p->f_attr.label))
1363                 goto err_free_p;
1364
1365         p->a_label = nfs4_label_alloc(server, gfp_mask);
1366         if (IS_ERR(p->a_label))
1367                 goto err_free_f;
1368
1369         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1370         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1371         if (IS_ERR(p->o_arg.seqid))
1372                 goto err_free_label;
1373         nfs_sb_active(dentry->d_sb);
1374         p->dentry = dget(dentry);
1375         p->dir = parent;
1376         p->owner = sp;
1377         atomic_inc(&sp->so_count);
1378         p->o_arg.open_flags = flags;
1379         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1380         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1381         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1382                         fmode, flags);
1383         if (flags & O_CREAT) {
1384                 p->o_arg.umask = current_umask();
1385                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1386                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1387                         p->o_arg.u.attrs = &p->attrs;
1388                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1389
1390                         memcpy(p->o_arg.u.verifier.data, c->verf,
1391                                         sizeof(p->o_arg.u.verifier.data));
1392                 }
1393         }
1394         /* ask server to check for all possible rights as results
1395          * are cached */
1396         switch (p->o_arg.claim) {
1397         default:
1398                 break;
1399         case NFS4_OPEN_CLAIM_NULL:
1400         case NFS4_OPEN_CLAIM_FH:
1401                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1402                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1403                                   NFS4_ACCESS_EXECUTE |
1404                                   nfs_access_xattr_mask(server);
1405         }
1406         p->o_arg.clientid = server->nfs_client->cl_clientid;
1407         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1408         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1409         p->o_arg.name = &dentry->d_name;
1410         p->o_arg.server = server;
1411         p->o_arg.bitmask = nfs4_bitmask(server, label);
1412         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1413         switch (p->o_arg.claim) {
1414         case NFS4_OPEN_CLAIM_NULL:
1415         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1416         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1417                 p->o_arg.fh = NFS_FH(dir);
1418                 break;
1419         case NFS4_OPEN_CLAIM_PREVIOUS:
1420         case NFS4_OPEN_CLAIM_FH:
1421         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1422         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1423                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1424         }
1425         p->c_arg.fh = &p->o_res.fh;
1426         p->c_arg.stateid = &p->o_res.stateid;
1427         p->c_arg.seqid = p->o_arg.seqid;
1428         nfs4_init_opendata_res(p);
1429         kref_init(&p->kref);
1430         return p;
1431
1432 err_free_label:
1433         nfs4_label_free(p->a_label);
1434 err_free_f:
1435         nfs4_label_free(p->f_attr.label);
1436 err_free_p:
1437         kfree(p);
1438 err:
1439         dput(parent);
1440         return NULL;
1441 }
1442
1443 static void nfs4_opendata_free(struct kref *kref)
1444 {
1445         struct nfs4_opendata *p = container_of(kref,
1446                         struct nfs4_opendata, kref);
1447         struct super_block *sb = p->dentry->d_sb;
1448
1449         nfs4_lgopen_release(p->lgp);
1450         nfs_free_seqid(p->o_arg.seqid);
1451         nfs4_sequence_free_slot(&p->o_res.seq_res);
1452         if (p->state != NULL)
1453                 nfs4_put_open_state(p->state);
1454         nfs4_put_state_owner(p->owner);
1455
1456         nfs4_label_free(p->a_label);
1457         nfs4_label_free(p->f_attr.label);
1458
1459         dput(p->dir);
1460         dput(p->dentry);
1461         nfs_sb_deactive(sb);
1462         nfs_fattr_free_names(&p->f_attr);
1463         kfree(p->f_attr.mdsthreshold);
1464         kfree(p);
1465 }
1466
1467 static void nfs4_opendata_put(struct nfs4_opendata *p)
1468 {
1469         if (p != NULL)
1470                 kref_put(&p->kref, nfs4_opendata_free);
1471 }
1472
1473 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1474                 fmode_t fmode)
1475 {
1476         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1477         case FMODE_READ|FMODE_WRITE:
1478                 return state->n_rdwr != 0;
1479         case FMODE_WRITE:
1480                 return state->n_wronly != 0;
1481         case FMODE_READ:
1482                 return state->n_rdonly != 0;
1483         }
1484         WARN_ON_ONCE(1);
1485         return false;
1486 }
1487
1488 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1489                 int open_mode, enum open_claim_type4 claim)
1490 {
1491         int ret = 0;
1492
1493         if (open_mode & (O_EXCL|O_TRUNC))
1494                 goto out;
1495         switch (claim) {
1496         case NFS4_OPEN_CLAIM_NULL:
1497         case NFS4_OPEN_CLAIM_FH:
1498                 goto out;
1499         default:
1500                 break;
1501         }
1502         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1503                 case FMODE_READ:
1504                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1505                                 && state->n_rdonly != 0;
1506                         break;
1507                 case FMODE_WRITE:
1508                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1509                                 && state->n_wronly != 0;
1510                         break;
1511                 case FMODE_READ|FMODE_WRITE:
1512                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1513                                 && state->n_rdwr != 0;
1514         }
1515 out:
1516         return ret;
1517 }
1518
1519 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1520                 enum open_claim_type4 claim)
1521 {
1522         if (delegation == NULL)
1523                 return 0;
1524         if ((delegation->type & fmode) != fmode)
1525                 return 0;
1526         switch (claim) {
1527         case NFS4_OPEN_CLAIM_NULL:
1528         case NFS4_OPEN_CLAIM_FH:
1529                 break;
1530         case NFS4_OPEN_CLAIM_PREVIOUS:
1531                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1532                         break;
1533                 fallthrough;
1534         default:
1535                 return 0;
1536         }
1537         nfs_mark_delegation_referenced(delegation);
1538         return 1;
1539 }
1540
1541 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1542 {
1543         switch (fmode) {
1544                 case FMODE_WRITE:
1545                         state->n_wronly++;
1546                         break;
1547                 case FMODE_READ:
1548                         state->n_rdonly++;
1549                         break;
1550                 case FMODE_READ|FMODE_WRITE:
1551                         state->n_rdwr++;
1552         }
1553         nfs4_state_set_mode_locked(state, state->state | fmode);
1554 }
1555
1556 #ifdef CONFIG_NFS_V4_1
1557 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1558 {
1559         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1560                 return true;
1561         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1562                 return true;
1563         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1564                 return true;
1565         return false;
1566 }
1567 #endif /* CONFIG_NFS_V4_1 */
1568
1569 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1570 {
1571         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1572                 wake_up_all(&state->waitq);
1573 }
1574
1575 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1576 {
1577         struct nfs_client *clp = state->owner->so_server->nfs_client;
1578         bool need_recover = false;
1579
1580         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1581                 need_recover = true;
1582         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1583                 need_recover = true;
1584         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1585                 need_recover = true;
1586         if (need_recover)
1587                 nfs4_state_mark_reclaim_nograce(clp, state);
1588 }
1589
1590 /*
1591  * Check for whether or not the caller may update the open stateid
1592  * to the value passed in by stateid.
1593  *
1594  * Note: This function relies heavily on the server implementing
1595  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1596  * correctly.
1597  * i.e. The stateid seqids have to be initialised to 1, and
1598  * are then incremented on every state transition.
1599  */
1600 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1601                 const nfs4_stateid *stateid)
1602 {
1603         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1604                 /* The common case - we're updating to a new sequence number */
1605                 if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1606                         if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1607                                 return true;
1608                         return false;
1609                 }
1610                 /* The server returned a new stateid */
1611         }
1612         /* This is the first OPEN in this generation */
1613         if (stateid->seqid == cpu_to_be32(1))
1614                 return true;
1615         return false;
1616 }
1617
1618 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1619 {
1620         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1621                 return;
1622         if (state->n_wronly)
1623                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1624         if (state->n_rdonly)
1625                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1626         if (state->n_rdwr)
1627                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1628         set_bit(NFS_OPEN_STATE, &state->flags);
1629 }
1630
1631 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1632                 nfs4_stateid *stateid, fmode_t fmode)
1633 {
1634         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1635         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1636         case FMODE_WRITE:
1637                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1638                 break;
1639         case FMODE_READ:
1640                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1641                 break;
1642         case 0:
1643                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1644                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1645                 clear_bit(NFS_OPEN_STATE, &state->flags);
1646         }
1647         if (stateid == NULL)
1648                 return;
1649         /* Handle OPEN+OPEN_DOWNGRADE races */
1650         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1651             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1652                 nfs_resync_open_stateid_locked(state);
1653                 goto out;
1654         }
1655         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1656                 nfs4_stateid_copy(&state->stateid, stateid);
1657         nfs4_stateid_copy(&state->open_stateid, stateid);
1658         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1659 out:
1660         nfs_state_log_update_open_stateid(state);
1661 }
1662
1663 static void nfs_clear_open_stateid(struct nfs4_state *state,
1664         nfs4_stateid *arg_stateid,
1665         nfs4_stateid *stateid, fmode_t fmode)
1666 {
1667         write_seqlock(&state->seqlock);
1668         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1669         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1670                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1671         write_sequnlock(&state->seqlock);
1672         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1673                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1674 }
1675
1676 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1677                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1678         __must_hold(&state->owner->so_lock)
1679         __must_hold(&state->seqlock)
1680         __must_hold(RCU)
1681
1682 {
1683         DEFINE_WAIT(wait);
1684         int status = 0;
1685         for (;;) {
1686
1687                 if (nfs_stateid_is_sequential(state, stateid))
1688                         break;
1689
1690                 if (status)
1691                         break;
1692                 /* Rely on seqids for serialisation with NFSv4.0 */
1693                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1694                         break;
1695
1696                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1697                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1698                 /*
1699                  * Ensure we process the state changes in the same order
1700                  * in which the server processed them by delaying the
1701                  * update of the stateid until we are in sequence.
1702                  */
1703                 write_sequnlock(&state->seqlock);
1704                 spin_unlock(&state->owner->so_lock);
1705                 rcu_read_unlock();
1706                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1707
1708                 if (!fatal_signal_pending(current)) {
1709                         if (schedule_timeout(5*HZ) == 0)
1710                                 status = -EAGAIN;
1711                         else
1712                                 status = 0;
1713                 } else
1714                         status = -EINTR;
1715                 finish_wait(&state->waitq, &wait);
1716                 rcu_read_lock();
1717                 spin_lock(&state->owner->so_lock);
1718                 write_seqlock(&state->seqlock);
1719         }
1720
1721         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1722             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1723                 nfs4_stateid_copy(freeme, &state->open_stateid);
1724                 nfs_test_and_clear_all_open_stateid(state);
1725         }
1726
1727         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1728                 nfs4_stateid_copy(&state->stateid, stateid);
1729         nfs4_stateid_copy(&state->open_stateid, stateid);
1730         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1731         nfs_state_log_update_open_stateid(state);
1732 }
1733
1734 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1735                 const nfs4_stateid *open_stateid,
1736                 fmode_t fmode,
1737                 nfs4_stateid *freeme)
1738 {
1739         /*
1740          * Protect the call to nfs4_state_set_mode_locked and
1741          * serialise the stateid update
1742          */
1743         write_seqlock(&state->seqlock);
1744         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1745         switch (fmode) {
1746         case FMODE_READ:
1747                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1748                 break;
1749         case FMODE_WRITE:
1750                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1751                 break;
1752         case FMODE_READ|FMODE_WRITE:
1753                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1754         }
1755         set_bit(NFS_OPEN_STATE, &state->flags);
1756         write_sequnlock(&state->seqlock);
1757 }
1758
1759 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1760 {
1761         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1762         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1763         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1764         clear_bit(NFS_OPEN_STATE, &state->flags);
1765 }
1766
1767 static void nfs_state_set_delegation(struct nfs4_state *state,
1768                 const nfs4_stateid *deleg_stateid,
1769                 fmode_t fmode)
1770 {
1771         /*
1772          * Protect the call to nfs4_state_set_mode_locked and
1773          * serialise the stateid update
1774          */
1775         write_seqlock(&state->seqlock);
1776         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1777         set_bit(NFS_DELEGATED_STATE, &state->flags);
1778         write_sequnlock(&state->seqlock);
1779 }
1780
1781 static void nfs_state_clear_delegation(struct nfs4_state *state)
1782 {
1783         write_seqlock(&state->seqlock);
1784         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1785         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1786         write_sequnlock(&state->seqlock);
1787 }
1788
1789 int update_open_stateid(struct nfs4_state *state,
1790                 const nfs4_stateid *open_stateid,
1791                 const nfs4_stateid *delegation,
1792                 fmode_t fmode)
1793 {
1794         struct nfs_server *server = NFS_SERVER(state->inode);
1795         struct nfs_client *clp = server->nfs_client;
1796         struct nfs_inode *nfsi = NFS_I(state->inode);
1797         struct nfs_delegation *deleg_cur;
1798         nfs4_stateid freeme = { };
1799         int ret = 0;
1800
1801         fmode &= (FMODE_READ|FMODE_WRITE);
1802
1803         rcu_read_lock();
1804         spin_lock(&state->owner->so_lock);
1805         if (open_stateid != NULL) {
1806                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1807                 ret = 1;
1808         }
1809
1810         deleg_cur = nfs4_get_valid_delegation(state->inode);
1811         if (deleg_cur == NULL)
1812                 goto no_delegation;
1813
1814         spin_lock(&deleg_cur->lock);
1815         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1816            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1817             (deleg_cur->type & fmode) != fmode)
1818                 goto no_delegation_unlock;
1819
1820         if (delegation == NULL)
1821                 delegation = &deleg_cur->stateid;
1822         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1823                 goto no_delegation_unlock;
1824
1825         nfs_mark_delegation_referenced(deleg_cur);
1826         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1827         ret = 1;
1828 no_delegation_unlock:
1829         spin_unlock(&deleg_cur->lock);
1830 no_delegation:
1831         if (ret)
1832                 update_open_stateflags(state, fmode);
1833         spin_unlock(&state->owner->so_lock);
1834         rcu_read_unlock();
1835
1836         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1837                 nfs4_schedule_state_manager(clp);
1838         if (freeme.type != 0)
1839                 nfs4_test_and_free_stateid(server, &freeme,
1840                                 state->owner->so_cred);
1841
1842         return ret;
1843 }
1844
1845 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1846                 const nfs4_stateid *stateid)
1847 {
1848         struct nfs4_state *state = lsp->ls_state;
1849         bool ret = false;
1850
1851         spin_lock(&state->state_lock);
1852         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1853                 goto out_noupdate;
1854         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1855                 goto out_noupdate;
1856         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1857         ret = true;
1858 out_noupdate:
1859         spin_unlock(&state->state_lock);
1860         return ret;
1861 }
1862
1863 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1864 {
1865         struct nfs_delegation *delegation;
1866
1867         fmode &= FMODE_READ|FMODE_WRITE;
1868         rcu_read_lock();
1869         delegation = nfs4_get_valid_delegation(inode);
1870         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1871                 rcu_read_unlock();
1872                 return;
1873         }
1874         rcu_read_unlock();
1875         nfs4_inode_return_delegation(inode);
1876 }
1877
1878 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1879 {
1880         struct nfs4_state *state = opendata->state;
1881         struct nfs_delegation *delegation;
1882         int open_mode = opendata->o_arg.open_flags;
1883         fmode_t fmode = opendata->o_arg.fmode;
1884         enum open_claim_type4 claim = opendata->o_arg.claim;
1885         nfs4_stateid stateid;
1886         int ret = -EAGAIN;
1887
1888         for (;;) {
1889                 spin_lock(&state->owner->so_lock);
1890                 if (can_open_cached(state, fmode, open_mode, claim)) {
1891                         update_open_stateflags(state, fmode);
1892                         spin_unlock(&state->owner->so_lock);
1893                         goto out_return_state;
1894                 }
1895                 spin_unlock(&state->owner->so_lock);
1896                 rcu_read_lock();
1897                 delegation = nfs4_get_valid_delegation(state->inode);
1898                 if (!can_open_delegated(delegation, fmode, claim)) {
1899                         rcu_read_unlock();
1900                         break;
1901                 }
1902                 /* Save the delegation */
1903                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1904                 rcu_read_unlock();
1905                 nfs_release_seqid(opendata->o_arg.seqid);
1906                 if (!opendata->is_recover) {
1907                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1908                         if (ret != 0)
1909                                 goto out;
1910                 }
1911                 ret = -EAGAIN;
1912
1913                 /* Try to update the stateid using the delegation */
1914                 if (update_open_stateid(state, NULL, &stateid, fmode))
1915                         goto out_return_state;
1916         }
1917 out:
1918         return ERR_PTR(ret);
1919 out_return_state:
1920         refcount_inc(&state->count);
1921         return state;
1922 }
1923
1924 static void
1925 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1926 {
1927         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1928         struct nfs_delegation *delegation;
1929         int delegation_flags = 0;
1930
1931         rcu_read_lock();
1932         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1933         if (delegation)
1934                 delegation_flags = delegation->flags;
1935         rcu_read_unlock();
1936         switch (data->o_arg.claim) {
1937         default:
1938                 break;
1939         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1940         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1941                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1942                                    "returning a delegation for "
1943                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1944                                    clp->cl_hostname);
1945                 return;
1946         }
1947         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1948                 nfs_inode_set_delegation(state->inode,
1949                                 data->owner->so_cred,
1950                                 data->o_res.delegation_type,
1951                                 &data->o_res.delegation,
1952                                 data->o_res.pagemod_limit);
1953         else
1954                 nfs_inode_reclaim_delegation(state->inode,
1955                                 data->owner->so_cred,
1956                                 data->o_res.delegation_type,
1957                                 &data->o_res.delegation,
1958                                 data->o_res.pagemod_limit);
1959
1960         if (data->o_res.do_recall)
1961                 nfs_async_inode_return_delegation(state->inode,
1962                                                   &data->o_res.delegation);
1963 }
1964
1965 /*
1966  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1967  * and update the nfs4_state.
1968  */
1969 static struct nfs4_state *
1970 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1971 {
1972         struct inode *inode = data->state->inode;
1973         struct nfs4_state *state = data->state;
1974         int ret;
1975
1976         if (!data->rpc_done) {
1977                 if (data->rpc_status)
1978                         return ERR_PTR(data->rpc_status);
1979                 /* cached opens have already been processed */
1980                 goto update;
1981         }
1982
1983         ret = nfs_refresh_inode(inode, &data->f_attr);
1984         if (ret)
1985                 return ERR_PTR(ret);
1986
1987         if (data->o_res.delegation_type != 0)
1988                 nfs4_opendata_check_deleg(data, state);
1989 update:
1990         if (!update_open_stateid(state, &data->o_res.stateid,
1991                                 NULL, data->o_arg.fmode))
1992                 return ERR_PTR(-EAGAIN);
1993         refcount_inc(&state->count);
1994
1995         return state;
1996 }
1997
1998 static struct inode *
1999 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2000 {
2001         struct inode *inode;
2002
2003         switch (data->o_arg.claim) {
2004         case NFS4_OPEN_CLAIM_NULL:
2005         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2006         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2007                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2008                         return ERR_PTR(-EAGAIN);
2009                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2010                                 &data->f_attr);
2011                 break;
2012         default:
2013                 inode = d_inode(data->dentry);
2014                 ihold(inode);
2015                 nfs_refresh_inode(inode, &data->f_attr);
2016         }
2017         return inode;
2018 }
2019
2020 static struct nfs4_state *
2021 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2022 {
2023         struct nfs4_state *state;
2024         struct inode *inode;
2025
2026         inode = nfs4_opendata_get_inode(data);
2027         if (IS_ERR(inode))
2028                 return ERR_CAST(inode);
2029         if (data->state != NULL && data->state->inode == inode) {
2030                 state = data->state;
2031                 refcount_inc(&state->count);
2032         } else
2033                 state = nfs4_get_open_state(inode, data->owner);
2034         iput(inode);
2035         if (state == NULL)
2036                 state = ERR_PTR(-ENOMEM);
2037         return state;
2038 }
2039
2040 static struct nfs4_state *
2041 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2042 {
2043         struct nfs4_state *state;
2044
2045         if (!data->rpc_done) {
2046                 state = nfs4_try_open_cached(data);
2047                 trace_nfs4_cached_open(data->state);
2048                 goto out;
2049         }
2050
2051         state = nfs4_opendata_find_nfs4_state(data);
2052         if (IS_ERR(state))
2053                 goto out;
2054
2055         if (data->o_res.delegation_type != 0)
2056                 nfs4_opendata_check_deleg(data, state);
2057         if (!update_open_stateid(state, &data->o_res.stateid,
2058                                 NULL, data->o_arg.fmode)) {
2059                 nfs4_put_open_state(state);
2060                 state = ERR_PTR(-EAGAIN);
2061         }
2062 out:
2063         nfs_release_seqid(data->o_arg.seqid);
2064         return state;
2065 }
2066
2067 static struct nfs4_state *
2068 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2069 {
2070         struct nfs4_state *ret;
2071
2072         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2073                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2074         else
2075                 ret = _nfs4_opendata_to_nfs4_state(data);
2076         nfs4_sequence_free_slot(&data->o_res.seq_res);
2077         return ret;
2078 }
2079
2080 static struct nfs_open_context *
2081 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2082 {
2083         struct nfs_inode *nfsi = NFS_I(state->inode);
2084         struct nfs_open_context *ctx;
2085
2086         rcu_read_lock();
2087         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2088                 if (ctx->state != state)
2089                         continue;
2090                 if ((ctx->mode & mode) != mode)
2091                         continue;
2092                 if (!get_nfs_open_context(ctx))
2093                         continue;
2094                 rcu_read_unlock();
2095                 return ctx;
2096         }
2097         rcu_read_unlock();
2098         return ERR_PTR(-ENOENT);
2099 }
2100
2101 static struct nfs_open_context *
2102 nfs4_state_find_open_context(struct nfs4_state *state)
2103 {
2104         struct nfs_open_context *ctx;
2105
2106         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2107         if (!IS_ERR(ctx))
2108                 return ctx;
2109         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2110         if (!IS_ERR(ctx))
2111                 return ctx;
2112         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2113 }
2114
2115 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2116                 struct nfs4_state *state, enum open_claim_type4 claim)
2117 {
2118         struct nfs4_opendata *opendata;
2119
2120         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2121                         NULL, claim, GFP_NOFS);
2122         if (opendata == NULL)
2123                 return ERR_PTR(-ENOMEM);
2124         opendata->state = state;
2125         refcount_inc(&state->count);
2126         return opendata;
2127 }
2128
2129 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2130                 fmode_t fmode)
2131 {
2132         struct nfs4_state *newstate;
2133         int ret;
2134
2135         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2136                 return 0;
2137         opendata->o_arg.open_flags = 0;
2138         opendata->o_arg.fmode = fmode;
2139         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2140                         NFS_SB(opendata->dentry->d_sb),
2141                         fmode, 0);
2142         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2143         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2144         nfs4_init_opendata_res(opendata);
2145         ret = _nfs4_recover_proc_open(opendata);
2146         if (ret != 0)
2147                 return ret; 
2148         newstate = nfs4_opendata_to_nfs4_state(opendata);
2149         if (IS_ERR(newstate))
2150                 return PTR_ERR(newstate);
2151         if (newstate != opendata->state)
2152                 ret = -ESTALE;
2153         nfs4_close_state(newstate, fmode);
2154         return ret;
2155 }
2156
2157 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2158 {
2159         int ret;
2160
2161         /* memory barrier prior to reading state->n_* */
2162         smp_rmb();
2163         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2164         if (ret != 0)
2165                 return ret;
2166         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2167         if (ret != 0)
2168                 return ret;
2169         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2170         if (ret != 0)
2171                 return ret;
2172         /*
2173          * We may have performed cached opens for all three recoveries.
2174          * Check if we need to update the current stateid.
2175          */
2176         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2177             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2178                 write_seqlock(&state->seqlock);
2179                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2180                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2181                 write_sequnlock(&state->seqlock);
2182         }
2183         return 0;
2184 }
2185
2186 /*
2187  * OPEN_RECLAIM:
2188  *      reclaim state on the server after a reboot.
2189  */
2190 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2191 {
2192         struct nfs_delegation *delegation;
2193         struct nfs4_opendata *opendata;
2194         fmode_t delegation_type = 0;
2195         int status;
2196
2197         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2198                         NFS4_OPEN_CLAIM_PREVIOUS);
2199         if (IS_ERR(opendata))
2200                 return PTR_ERR(opendata);
2201         rcu_read_lock();
2202         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2203         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2204                 delegation_type = delegation->type;
2205         rcu_read_unlock();
2206         opendata->o_arg.u.delegation_type = delegation_type;
2207         status = nfs4_open_recover(opendata, state);
2208         nfs4_opendata_put(opendata);
2209         return status;
2210 }
2211
2212 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2213 {
2214         struct nfs_server *server = NFS_SERVER(state->inode);
2215         struct nfs4_exception exception = { };
2216         int err;
2217         do {
2218                 err = _nfs4_do_open_reclaim(ctx, state);
2219                 trace_nfs4_open_reclaim(ctx, 0, err);
2220                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2221                         continue;
2222                 if (err != -NFS4ERR_DELAY)
2223                         break;
2224                 nfs4_handle_exception(server, err, &exception);
2225         } while (exception.retry);
2226         return err;
2227 }
2228
2229 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2230 {
2231         struct nfs_open_context *ctx;
2232         int ret;
2233
2234         ctx = nfs4_state_find_open_context(state);
2235         if (IS_ERR(ctx))
2236                 return -EAGAIN;
2237         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2238         nfs_state_clear_open_state_flags(state);
2239         ret = nfs4_do_open_reclaim(ctx, state);
2240         put_nfs_open_context(ctx);
2241         return ret;
2242 }
2243
2244 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)
2245 {
2246         switch (err) {
2247                 default:
2248                         printk(KERN_ERR "NFS: %s: unhandled error "
2249                                         "%d.\n", __func__, err);
2250                         fallthrough;
2251                 case 0:
2252                 case -ENOENT:
2253                 case -EAGAIN:
2254                 case -ESTALE:
2255                 case -ETIMEDOUT:
2256                         break;
2257                 case -NFS4ERR_BADSESSION:
2258                 case -NFS4ERR_BADSLOT:
2259                 case -NFS4ERR_BAD_HIGH_SLOT:
2260                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2261                 case -NFS4ERR_DEADSESSION:
2262                         return -EAGAIN;
2263                 case -NFS4ERR_STALE_CLIENTID:
2264                 case -NFS4ERR_STALE_STATEID:
2265                         /* Don't recall a delegation if it was lost */
2266                         nfs4_schedule_lease_recovery(server->nfs_client);
2267                         return -EAGAIN;
2268                 case -NFS4ERR_MOVED:
2269                         nfs4_schedule_migration_recovery(server);
2270                         return -EAGAIN;
2271                 case -NFS4ERR_LEASE_MOVED:
2272                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2273                         return -EAGAIN;
2274                 case -NFS4ERR_DELEG_REVOKED:
2275                 case -NFS4ERR_ADMIN_REVOKED:
2276                 case -NFS4ERR_EXPIRED:
2277                 case -NFS4ERR_BAD_STATEID:
2278                 case -NFS4ERR_OPENMODE:
2279                         nfs_inode_find_state_and_recover(state->inode,
2280                                         stateid);
2281                         nfs4_schedule_stateid_recovery(server, state);
2282                         return -EAGAIN;
2283                 case -NFS4ERR_DELAY:
2284                 case -NFS4ERR_GRACE:
2285                         ssleep(1);
2286                         return -EAGAIN;
2287                 case -ENOMEM:
2288                 case -NFS4ERR_DENIED:
2289                         if (fl) {
2290                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2291                                 if (lsp)
2292                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2293                         }
2294                         return 0;
2295         }
2296         return err;
2297 }
2298
2299 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2300                 struct nfs4_state *state, const nfs4_stateid *stateid)
2301 {
2302         struct nfs_server *server = NFS_SERVER(state->inode);
2303         struct nfs4_opendata *opendata;
2304         int err = 0;
2305
2306         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2307                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2308         if (IS_ERR(opendata))
2309                 return PTR_ERR(opendata);
2310         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2311         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2312                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2313                 if (err)
2314                         goto out;
2315         }
2316         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2317                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2318                 if (err)
2319                         goto out;
2320         }
2321         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2322                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2323                 if (err)
2324                         goto out;
2325         }
2326         nfs_state_clear_delegation(state);
2327 out:
2328         nfs4_opendata_put(opendata);
2329         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2330 }
2331
2332 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2333 {
2334         struct nfs4_opendata *data = calldata;
2335
2336         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2337                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2338 }
2339
2340 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2341 {
2342         struct nfs4_opendata *data = calldata;
2343
2344         nfs40_sequence_done(task, &data->c_res.seq_res);
2345
2346         data->rpc_status = task->tk_status;
2347         if (data->rpc_status == 0) {
2348                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2349                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2350                 renew_lease(data->o_res.server, data->timestamp);
2351                 data->rpc_done = true;
2352         }
2353 }
2354
2355 static void nfs4_open_confirm_release(void *calldata)
2356 {
2357         struct nfs4_opendata *data = calldata;
2358         struct nfs4_state *state = NULL;
2359
2360         /* If this request hasn't been cancelled, do nothing */
2361         if (!data->cancelled)
2362                 goto out_free;
2363         /* In case of error, no cleanup! */
2364         if (!data->rpc_done)
2365                 goto out_free;
2366         state = nfs4_opendata_to_nfs4_state(data);
2367         if (!IS_ERR(state))
2368                 nfs4_close_state(state, data->o_arg.fmode);
2369 out_free:
2370         nfs4_opendata_put(data);
2371 }
2372
2373 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2374         .rpc_call_prepare = nfs4_open_confirm_prepare,
2375         .rpc_call_done = nfs4_open_confirm_done,
2376         .rpc_release = nfs4_open_confirm_release,
2377 };
2378
2379 /*
2380  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2381  */
2382 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2383 {
2384         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2385         struct rpc_task *task;
2386         struct  rpc_message msg = {
2387                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2388                 .rpc_argp = &data->c_arg,
2389                 .rpc_resp = &data->c_res,
2390                 .rpc_cred = data->owner->so_cred,
2391         };
2392         struct rpc_task_setup task_setup_data = {
2393                 .rpc_client = server->client,
2394                 .rpc_message = &msg,
2395                 .callback_ops = &nfs4_open_confirm_ops,
2396                 .callback_data = data,
2397                 .workqueue = nfsiod_workqueue,
2398                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2399         };
2400         int status;
2401
2402         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2403                                 data->is_recover);
2404         kref_get(&data->kref);
2405         data->rpc_done = false;
2406         data->rpc_status = 0;
2407         data->timestamp = jiffies;
2408         task = rpc_run_task(&task_setup_data);
2409         if (IS_ERR(task))
2410                 return PTR_ERR(task);
2411         status = rpc_wait_for_completion_task(task);
2412         if (status != 0) {
2413                 data->cancelled = true;
2414                 smp_wmb();
2415         } else
2416                 status = data->rpc_status;
2417         rpc_put_task(task);
2418         return status;
2419 }
2420
2421 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2422 {
2423         struct nfs4_opendata *data = calldata;
2424         struct nfs4_state_owner *sp = data->owner;
2425         struct nfs_client *clp = sp->so_server->nfs_client;
2426         enum open_claim_type4 claim = data->o_arg.claim;
2427
2428         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2429                 goto out_wait;
2430         /*
2431          * Check if we still need to send an OPEN call, or if we can use
2432          * a delegation instead.
2433          */
2434         if (data->state != NULL) {
2435                 struct nfs_delegation *delegation;
2436
2437                 if (can_open_cached(data->state, data->o_arg.fmode,
2438                                         data->o_arg.open_flags, claim))
2439                         goto out_no_action;
2440                 rcu_read_lock();
2441                 delegation = nfs4_get_valid_delegation(data->state->inode);
2442                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2443                         goto unlock_no_action;
2444                 rcu_read_unlock();
2445         }
2446         /* Update client id. */
2447         data->o_arg.clientid = clp->cl_clientid;
2448         switch (claim) {
2449         default:
2450                 break;
2451         case NFS4_OPEN_CLAIM_PREVIOUS:
2452         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2453         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2454                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2455                 fallthrough;
2456         case NFS4_OPEN_CLAIM_FH:
2457                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2458         }
2459         data->timestamp = jiffies;
2460         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2461                                 &data->o_arg.seq_args,
2462                                 &data->o_res.seq_res,
2463                                 task) != 0)
2464                 nfs_release_seqid(data->o_arg.seqid);
2465
2466         /* Set the create mode (note dependency on the session type) */
2467         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2468         if (data->o_arg.open_flags & O_EXCL) {
2469                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2470                 if (clp->cl_mvops->minor_version == 0) {
2471                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2472                         /* don't put an ACCESS op in OPEN compound if O_EXCL,
2473                          * because ACCESS will return permission denied for
2474                          * all bits until close */
2475                         data->o_res.access_request = data->o_arg.access = 0;
2476                 } else if (nfs4_has_persistent_session(clp))
2477                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2478         }
2479         return;
2480 unlock_no_action:
2481         trace_nfs4_cached_open(data->state);
2482         rcu_read_unlock();
2483 out_no_action:
2484         task->tk_action = NULL;
2485 out_wait:
2486         nfs4_sequence_done(task, &data->o_res.seq_res);
2487 }
2488
2489 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2490 {
2491         struct nfs4_opendata *data = calldata;
2492
2493         data->rpc_status = task->tk_status;
2494
2495         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2496                 return;
2497
2498         if (task->tk_status == 0) {
2499                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2500                         switch (data->o_res.f_attr->mode & S_IFMT) {
2501                         case S_IFREG:
2502                                 break;
2503                         case S_IFLNK:
2504                                 data->rpc_status = -ELOOP;
2505                                 break;
2506                         case S_IFDIR:
2507                                 data->rpc_status = -EISDIR;
2508                                 break;
2509                         default:
2510                                 data->rpc_status = -ENOTDIR;
2511                         }
2512                 }
2513                 renew_lease(data->o_res.server, data->timestamp);
2514                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2515                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2516         }
2517         data->rpc_done = true;
2518 }
2519
2520 static void nfs4_open_release(void *calldata)
2521 {
2522         struct nfs4_opendata *data = calldata;
2523         struct nfs4_state *state = NULL;
2524
2525         /* If this request hasn't been cancelled, do nothing */
2526         if (!data->cancelled)
2527                 goto out_free;
2528         /* In case of error, no cleanup! */
2529         if (data->rpc_status != 0 || !data->rpc_done)
2530                 goto out_free;
2531         /* In case we need an open_confirm, no cleanup! */
2532         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2533                 goto out_free;
2534         state = nfs4_opendata_to_nfs4_state(data);
2535         if (!IS_ERR(state))
2536                 nfs4_close_state(state, data->o_arg.fmode);
2537 out_free:
2538         nfs4_opendata_put(data);
2539 }
2540
2541 static const struct rpc_call_ops nfs4_open_ops = {
2542         .rpc_call_prepare = nfs4_open_prepare,
2543         .rpc_call_done = nfs4_open_done,
2544         .rpc_release = nfs4_open_release,
2545 };
2546
2547 static int nfs4_run_open_task(struct nfs4_opendata *data,
2548                               struct nfs_open_context *ctx)
2549 {
2550         struct inode *dir = d_inode(data->dir);
2551         struct nfs_server *server = NFS_SERVER(dir);
2552         struct nfs_openargs *o_arg = &data->o_arg;
2553         struct nfs_openres *o_res = &data->o_res;
2554         struct rpc_task *task;
2555         struct rpc_message msg = {
2556                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2557                 .rpc_argp = o_arg,
2558                 .rpc_resp = o_res,
2559                 .rpc_cred = data->owner->so_cred,
2560         };
2561         struct rpc_task_setup task_setup_data = {
2562                 .rpc_client = server->client,
2563                 .rpc_message = &msg,
2564                 .callback_ops = &nfs4_open_ops,
2565                 .callback_data = data,
2566                 .workqueue = nfsiod_workqueue,
2567                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2568         };
2569         int status;
2570
2571         if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
2572                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
2573
2574         kref_get(&data->kref);
2575         data->rpc_done = false;
2576         data->rpc_status = 0;
2577         data->cancelled = false;
2578         data->is_recover = false;
2579         if (!ctx) {
2580                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2581                 data->is_recover = true;
2582                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2583         } else {
2584                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2585                 pnfs_lgopen_prepare(data, ctx);
2586         }
2587         task = rpc_run_task(&task_setup_data);
2588         if (IS_ERR(task))
2589                 return PTR_ERR(task);
2590         status = rpc_wait_for_completion_task(task);
2591         if (status != 0) {
2592                 data->cancelled = true;
2593                 smp_wmb();
2594         } else
2595                 status = data->rpc_status;
2596         rpc_put_task(task);
2597
2598         return status;
2599 }
2600
2601 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2602 {
2603         struct inode *dir = d_inode(data->dir);
2604         struct nfs_openres *o_res = &data->o_res;
2605         int status;
2606
2607         status = nfs4_run_open_task(data, NULL);
2608         if (status != 0 || !data->rpc_done)
2609                 return status;
2610
2611         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2612
2613         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2614                 status = _nfs4_proc_open_confirm(data);
2615
2616         return status;
2617 }
2618
2619 /*
2620  * Additional permission checks in order to distinguish between an
2621  * open for read, and an open for execute. This works around the
2622  * fact that NFSv4 OPEN treats read and execute permissions as being
2623  * the same.
2624  * Note that in the non-execute case, we want to turn off permission
2625  * checking if we just created a new file (POSIX open() semantics).
2626  */
2627 static int nfs4_opendata_access(const struct cred *cred,
2628                                 struct nfs4_opendata *opendata,
2629                                 struct nfs4_state *state, fmode_t fmode,
2630                                 int openflags)
2631 {
2632         struct nfs_access_entry cache;
2633         u32 mask, flags;
2634
2635         /* access call failed or for some reason the server doesn't
2636          * support any access modes -- defer access call until later */
2637         if (opendata->o_res.access_supported == 0)
2638                 return 0;
2639
2640         mask = 0;
2641         /*
2642          * Use openflags to check for exec, because fmode won't
2643          * always have FMODE_EXEC set when file open for exec.
2644          */
2645         if (openflags & __FMODE_EXEC) {
2646                 /* ONLY check for exec rights */
2647                 if (S_ISDIR(state->inode->i_mode))
2648                         mask = NFS4_ACCESS_LOOKUP;
2649                 else
2650                         mask = NFS4_ACCESS_EXECUTE;
2651         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2652                 mask = NFS4_ACCESS_READ;
2653
2654         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2655         nfs_access_add_cache(state->inode, &cache, cred);
2656
2657         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2658         if ((mask & ~cache.mask & flags) == 0)
2659                 return 0;
2660
2661         return -EACCES;
2662 }
2663
2664 /*
2665  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2666  */
2667 static int _nfs4_proc_open(struct nfs4_opendata *data,
2668                            struct nfs_open_context *ctx)
2669 {
2670         struct inode *dir = d_inode(data->dir);
2671         struct nfs_server *server = NFS_SERVER(dir);
2672         struct nfs_openargs *o_arg = &data->o_arg;
2673         struct nfs_openres *o_res = &data->o_res;
2674         int status;
2675
2676         status = nfs4_run_open_task(data, ctx);
2677         if (!data->rpc_done)
2678                 return status;
2679         if (status != 0) {
2680                 if (status == -NFS4ERR_BADNAME &&
2681                                 !(o_arg->open_flags & O_CREAT))
2682                         return -ENOENT;
2683                 return status;
2684         }
2685
2686         nfs_fattr_map_and_free_names(server, &data->f_attr);
2687
2688         if (o_arg->open_flags & O_CREAT) {
2689                 if (o_arg->open_flags & O_EXCL)
2690                         data->file_created = true;
2691                 else if (o_res->cinfo.before != o_res->cinfo.after)
2692                         data->file_created = true;
2693                 if (data->file_created ||
2694                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2695                         nfs4_update_changeattr(dir, &o_res->cinfo,
2696                                         o_res->f_attr->time_start,
2697                                         NFS_INO_INVALID_DATA);
2698         }
2699         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2700                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2701         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2702                 status = _nfs4_proc_open_confirm(data);
2703                 if (status != 0)
2704                         return status;
2705         }
2706         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2707                 nfs4_sequence_free_slot(&o_res->seq_res);
2708                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, NULL);
2709         }
2710         return 0;
2711 }
2712
2713 /*
2714  * OPEN_EXPIRED:
2715  *      reclaim state on the server after a network partition.
2716  *      Assumes caller holds the appropriate lock
2717  */
2718 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2719 {
2720         struct nfs4_opendata *opendata;
2721         int ret;
2722
2723         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2724                         NFS4_OPEN_CLAIM_FH);
2725         if (IS_ERR(opendata))
2726                 return PTR_ERR(opendata);
2727         ret = nfs4_open_recover(opendata, state);
2728         if (ret == -ESTALE)
2729                 d_drop(ctx->dentry);
2730         nfs4_opendata_put(opendata);
2731         return ret;
2732 }
2733
2734 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2735 {
2736         struct nfs_server *server = NFS_SERVER(state->inode);
2737         struct nfs4_exception exception = { };
2738         int err;
2739
2740         do {
2741                 err = _nfs4_open_expired(ctx, state);
2742                 trace_nfs4_open_expired(ctx, 0, err);
2743                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2744                         continue;
2745                 switch (err) {
2746                 default:
2747                         goto out;
2748                 case -NFS4ERR_GRACE:
2749                 case -NFS4ERR_DELAY:
2750                         nfs4_handle_exception(server, err, &exception);
2751                         err = 0;
2752                 }
2753         } while (exception.retry);
2754 out:
2755         return err;
2756 }
2757
2758 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2759 {
2760         struct nfs_open_context *ctx;
2761         int ret;
2762
2763         ctx = nfs4_state_find_open_context(state);
2764         if (IS_ERR(ctx))
2765                 return -EAGAIN;
2766         ret = nfs4_do_open_expired(ctx, state);
2767         put_nfs_open_context(ctx);
2768         return ret;
2769 }
2770
2771 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2772                 const nfs4_stateid *stateid)
2773 {
2774         nfs_remove_bad_delegation(state->inode, stateid);
2775         nfs_state_clear_delegation(state);
2776 }
2777
2778 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2779 {
2780         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2781                 nfs_finish_clear_delegation_stateid(state, NULL);
2782 }
2783
2784 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2785 {
2786         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2787         nfs40_clear_delegation_stateid(state);
2788         nfs_state_clear_open_state_flags(state);
2789         return nfs4_open_expired(sp, state);
2790 }
2791
2792 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2793                 nfs4_stateid *stateid,
2794                 const struct cred *cred)
2795 {
2796         return -NFS4ERR_BAD_STATEID;
2797 }
2798
2799 #if defined(CONFIG_NFS_V4_1)
2800 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2801                 nfs4_stateid *stateid,
2802                 const struct cred *cred)
2803 {
2804         int status;
2805
2806         switch (stateid->type) {
2807         default:
2808                 break;
2809         case NFS4_INVALID_STATEID_TYPE:
2810         case NFS4_SPECIAL_STATEID_TYPE:
2811                 return -NFS4ERR_BAD_STATEID;
2812         case NFS4_REVOKED_STATEID_TYPE:
2813                 goto out_free;
2814         }
2815
2816         status = nfs41_test_stateid(server, stateid, cred);
2817         switch (status) {
2818         case -NFS4ERR_EXPIRED:
2819         case -NFS4ERR_ADMIN_REVOKED:
2820         case -NFS4ERR_DELEG_REVOKED:
2821                 break;
2822         default:
2823                 return status;
2824         }
2825 out_free:
2826         /* Ack the revoked state to the server */
2827         nfs41_free_stateid(server, stateid, cred, true);
2828         return -NFS4ERR_EXPIRED;
2829 }
2830
2831 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2832 {
2833         struct nfs_server *server = NFS_SERVER(state->inode);
2834         nfs4_stateid stateid;
2835         struct nfs_delegation *delegation;
2836         const struct cred *cred = NULL;
2837         int status, ret = NFS_OK;
2838
2839         /* Get the delegation credential for use by test/free_stateid */
2840         rcu_read_lock();
2841         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2842         if (delegation == NULL) {
2843                 rcu_read_unlock();
2844                 nfs_state_clear_delegation(state);
2845                 return NFS_OK;
2846         }
2847
2848         spin_lock(&delegation->lock);
2849         nfs4_stateid_copy(&stateid, &delegation->stateid);
2850
2851         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2852                                 &delegation->flags)) {
2853                 spin_unlock(&delegation->lock);
2854                 rcu_read_unlock();
2855                 return NFS_OK;
2856         }
2857
2858         if (delegation->cred)
2859                 cred = get_cred(delegation->cred);
2860         spin_unlock(&delegation->lock);
2861         rcu_read_unlock();
2862         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2863         trace_nfs4_test_delegation_stateid(state, NULL, status);
2864         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2865                 nfs_finish_clear_delegation_stateid(state, &stateid);
2866         else
2867                 ret = status;
2868
2869         put_cred(cred);
2870         return ret;
2871 }
2872
2873 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2874 {
2875         nfs4_stateid tmp;
2876
2877         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2878             nfs4_copy_delegation_stateid(state->inode, state->state,
2879                                 &tmp, NULL) &&
2880             nfs4_stateid_match_other(&state->stateid, &tmp))
2881                 nfs_state_set_delegation(state, &tmp, state->state);
2882         else
2883                 nfs_state_clear_delegation(state);
2884 }
2885
2886 /**
2887  * nfs41_check_expired_locks - possibly free a lock stateid
2888  *
2889  * @state: NFSv4 state for an inode
2890  *
2891  * Returns NFS_OK if recovery for this stateid is now finished.
2892  * Otherwise a negative NFS4ERR value is returned.
2893  */
2894 static int nfs41_check_expired_locks(struct nfs4_state *state)
2895 {
2896         int status, ret = NFS_OK;
2897         struct nfs4_lock_state *lsp, *prev = NULL;
2898         struct nfs_server *server = NFS_SERVER(state->inode);
2899
2900         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2901                 goto out;
2902
2903         spin_lock(&state->state_lock);
2904         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2905                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2906                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2907
2908                         refcount_inc(&lsp->ls_count);
2909                         spin_unlock(&state->state_lock);
2910
2911                         nfs4_put_lock_state(prev);
2912                         prev = lsp;
2913
2914                         status = nfs41_test_and_free_expired_stateid(server,
2915                                         &lsp->ls_stateid,
2916                                         cred);
2917                         trace_nfs4_test_lock_stateid(state, lsp, status);
2918                         if (status == -NFS4ERR_EXPIRED ||
2919                             status == -NFS4ERR_BAD_STATEID) {
2920                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2921                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2922                                 if (!recover_lost_locks)
2923                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2924                         } else if (status != NFS_OK) {
2925                                 ret = status;
2926                                 nfs4_put_lock_state(prev);
2927                                 goto out;
2928                         }
2929                         spin_lock(&state->state_lock);
2930                 }
2931         }
2932         spin_unlock(&state->state_lock);
2933         nfs4_put_lock_state(prev);
2934 out:
2935         return ret;
2936 }
2937
2938 /**
2939  * nfs41_check_open_stateid - possibly free an open stateid
2940  *
2941  * @state: NFSv4 state for an inode
2942  *
2943  * Returns NFS_OK if recovery for this stateid is now finished.
2944  * Otherwise a negative NFS4ERR value is returned.
2945  */
2946 static int nfs41_check_open_stateid(struct nfs4_state *state)
2947 {
2948         struct nfs_server *server = NFS_SERVER(state->inode);
2949         nfs4_stateid *stateid = &state->open_stateid;
2950         const struct cred *cred = state->owner->so_cred;
2951         int status;
2952
2953         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2954                 return -NFS4ERR_BAD_STATEID;
2955         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2956         trace_nfs4_test_open_stateid(state, NULL, status);
2957         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2958                 nfs_state_clear_open_state_flags(state);
2959                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2960                 return status;
2961         }
2962         if (nfs_open_stateid_recover_openmode(state))
2963                 return -NFS4ERR_OPENMODE;
2964         return NFS_OK;
2965 }
2966
2967 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2968 {
2969         int status;
2970
2971         status = nfs41_check_delegation_stateid(state);
2972         if (status != NFS_OK)
2973                 return status;
2974         nfs41_delegation_recover_stateid(state);
2975
2976         status = nfs41_check_expired_locks(state);
2977         if (status != NFS_OK)
2978                 return status;
2979         status = nfs41_check_open_stateid(state);
2980         if (status != NFS_OK)
2981                 status = nfs4_open_expired(sp, state);
2982         return status;
2983 }
2984 #endif
2985
2986 /*
2987  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2988  * fields corresponding to attributes that were used to store the verifier.
2989  * Make sure we clobber those fields in the later setattr call
2990  */
2991 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2992                                 struct iattr *sattr, struct nfs4_label **label)
2993 {
2994         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2995         __u32 attrset[3];
2996         unsigned ret;
2997         unsigned i;
2998
2999         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3000                 attrset[i] = opendata->o_res.attrset[i];
3001                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3002                         attrset[i] &= ~bitmask[i];
3003         }
3004
3005         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3006                 sattr->ia_valid : 0;
3007
3008         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3009                 if (sattr->ia_valid & ATTR_ATIME_SET)
3010                         ret |= ATTR_ATIME_SET;
3011                 else
3012                         ret |= ATTR_ATIME;
3013         }
3014
3015         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3016                 if (sattr->ia_valid & ATTR_MTIME_SET)
3017                         ret |= ATTR_MTIME_SET;
3018                 else
3019                         ret |= ATTR_MTIME;
3020         }
3021
3022         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3023                 *label = NULL;
3024         return ret;
3025 }
3026
3027 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3028                 int flags, struct nfs_open_context *ctx)
3029 {
3030         struct nfs4_state_owner *sp = opendata->owner;
3031         struct nfs_server *server = sp->so_server;
3032         struct dentry *dentry;
3033         struct nfs4_state *state;
3034         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3035         struct inode *dir = d_inode(opendata->dir);
3036         unsigned long dir_verifier;
3037         unsigned int seq;
3038         int ret;
3039
3040         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3041         dir_verifier = nfs_save_change_attribute(dir);
3042
3043         ret = _nfs4_proc_open(opendata, ctx);
3044         if (ret != 0)
3045                 goto out;
3046
3047         state = _nfs4_opendata_to_nfs4_state(opendata);
3048         ret = PTR_ERR(state);
3049         if (IS_ERR(state))
3050                 goto out;
3051         ctx->state = state;
3052         if (server->caps & NFS_CAP_POSIX_LOCK)
3053                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3054         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3055                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3056         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_PRESERVE_UNLINKED)
3057                 set_bit(NFS_INO_PRESERVE_UNLINKED, &NFS_I(state->inode)->flags);
3058
3059         dentry = opendata->dentry;
3060         if (d_really_is_negative(dentry)) {
3061                 struct dentry *alias;
3062                 d_drop(dentry);
3063                 alias = d_exact_alias(dentry, state->inode);
3064                 if (!alias)
3065                         alias = d_splice_alias(igrab(state->inode), dentry);
3066                 /* d_splice_alias() can't fail here - it's a non-directory */
3067                 if (alias) {
3068                         dput(ctx->dentry);
3069                         ctx->dentry = dentry = alias;
3070                 }
3071         }
3072
3073         switch(opendata->o_arg.claim) {
3074         default:
3075                 break;
3076         case NFS4_OPEN_CLAIM_NULL:
3077         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3078         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3079                 if (!opendata->rpc_done)
3080                         break;
3081                 if (opendata->o_res.delegation_type != 0)
3082                         dir_verifier = nfs_save_change_attribute(dir);
3083                 nfs_set_verifier(dentry, dir_verifier);
3084         }
3085
3086         /* Parse layoutget results before we check for access */
3087         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3088
3089         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3090                         acc_mode, flags);
3091         if (ret != 0)
3092                 goto out;
3093
3094         if (d_inode(dentry) == state->inode) {
3095                 nfs_inode_attach_open_context(ctx);
3096                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3097                         nfs4_schedule_stateid_recovery(server, state);
3098         }
3099
3100 out:
3101         if (opendata->lgp) {
3102                 nfs4_lgopen_release(opendata->lgp);
3103                 opendata->lgp = NULL;
3104         }
3105         if (!opendata->cancelled)
3106                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
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 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 = &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         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4016         if (page == NULL || locations == NULL)
4017                 goto out;
4018
4019         status = nfs4_proc_get_locations(server, fhandle, locations, page,
4020                                          cred);
4021         if (status)
4022                 goto out;
4023
4024         for (i = 0; i < locations->nlocations; i++)
4025                 test_fs_location_for_trunking(&locations->locations[i], clp,
4026                                               server);
4027 out:
4028         if (page)
4029                 __free_page(page);
4030         kfree(locations);
4031         return status;
4032 }
4033
4034 static int nfs4_discover_trunking(struct nfs_server *server,
4035                                   struct nfs_fh *fhandle)
4036 {
4037         struct nfs4_exception exception = {
4038                 .interruptible = true,
4039         };
4040         struct nfs_client *clp = server->nfs_client;
4041         int err = 0;
4042
4043         if (!nfs4_has_session(clp))
4044                 goto out;
4045         do {
4046                 err = nfs4_handle_exception(server,
4047                                 _nfs4_discover_trunking(server, fhandle),
4048                                 &exception);
4049         } while (exception.retry);
4050 out:
4051         return err;
4052 }
4053
4054 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4055                 struct nfs_fsinfo *info)
4056 {
4057         u32 bitmask[3];
4058         struct nfs4_lookup_root_arg args = {
4059                 .bitmask = bitmask,
4060         };
4061         struct nfs4_lookup_res res = {
4062                 .server = server,
4063                 .fattr = info->fattr,
4064                 .fh = fhandle,
4065         };
4066         struct rpc_message msg = {
4067                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4068                 .rpc_argp = &args,
4069                 .rpc_resp = &res,
4070         };
4071
4072         bitmask[0] = nfs4_fattr_bitmap[0];
4073         bitmask[1] = nfs4_fattr_bitmap[1];
4074         /*
4075          * Process the label in the upcoming getfattr
4076          */
4077         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4078
4079         nfs_fattr_init(info->fattr);
4080         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4081 }
4082
4083 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4084                 struct nfs_fsinfo *info)
4085 {
4086         struct nfs4_exception exception = {
4087                 .interruptible = true,
4088         };
4089         int err;
4090         do {
4091                 err = _nfs4_lookup_root(server, fhandle, info);
4092                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4093                 switch (err) {
4094                 case 0:
4095                 case -NFS4ERR_WRONGSEC:
4096                         goto out;
4097                 default:
4098                         err = nfs4_handle_exception(server, err, &exception);
4099                 }
4100         } while (exception.retry);
4101 out:
4102         return err;
4103 }
4104
4105 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4106                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4107 {
4108         struct rpc_auth_create_args auth_args = {
4109                 .pseudoflavor = flavor,
4110         };
4111         struct rpc_auth *auth;
4112
4113         auth = rpcauth_create(&auth_args, server->client);
4114         if (IS_ERR(auth))
4115                 return -EACCES;
4116         return nfs4_lookup_root(server, fhandle, info);
4117 }
4118
4119 /*
4120  * Retry pseudoroot lookup with various security flavors.  We do this when:
4121  *
4122  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4123  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4124  *
4125  * Returns zero on success, or a negative NFS4ERR value, or a
4126  * negative errno value.
4127  */
4128 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4129                               struct nfs_fsinfo *info)
4130 {
4131         /* Per 3530bis 15.33.5 */
4132         static const rpc_authflavor_t flav_array[] = {
4133                 RPC_AUTH_GSS_KRB5P,
4134                 RPC_AUTH_GSS_KRB5I,
4135                 RPC_AUTH_GSS_KRB5,
4136                 RPC_AUTH_UNIX,                  /* courtesy */
4137                 RPC_AUTH_NULL,
4138         };
4139         int status = -EPERM;
4140         size_t i;
4141
4142         if (server->auth_info.flavor_len > 0) {
4143                 /* try each flavor specified by user */
4144                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4145                         status = nfs4_lookup_root_sec(server, fhandle, info,
4146                                                 server->auth_info.flavors[i]);
4147                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4148                                 continue;
4149                         break;
4150                 }
4151         } else {
4152                 /* no flavors specified by user, try default list */
4153                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4154                         status = nfs4_lookup_root_sec(server, fhandle, info,
4155                                                       flav_array[i]);
4156                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4157                                 continue;
4158                         break;
4159                 }
4160         }
4161
4162         /*
4163          * -EACCES could mean that the user doesn't have correct permissions
4164          * to access the mount.  It could also mean that we tried to mount
4165          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4166          * existing mount programs don't handle -EACCES very well so it should
4167          * be mapped to -EPERM instead.
4168          */
4169         if (status == -EACCES)
4170                 status = -EPERM;
4171         return status;
4172 }
4173
4174 /**
4175  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4176  * @server: initialized nfs_server handle
4177  * @fhandle: we fill in the pseudo-fs root file handle
4178  * @info: we fill in an FSINFO struct
4179  * @auth_probe: probe the auth flavours
4180  *
4181  * Returns zero on success, or a negative errno.
4182  */
4183 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4184                          struct nfs_fsinfo *info,
4185                          bool auth_probe)
4186 {
4187         int status = 0;
4188
4189         if (!auth_probe)
4190                 status = nfs4_lookup_root(server, fhandle, info);
4191
4192         if (auth_probe || status == NFS4ERR_WRONGSEC)
4193                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4194                                 fhandle, info);
4195
4196         if (status == 0)
4197                 status = nfs4_server_capabilities(server, fhandle);
4198         if (status == 0)
4199                 status = nfs4_do_fsinfo(server, fhandle, info);
4200
4201         return nfs4_map_errors(status);
4202 }
4203
4204 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4205                               struct nfs_fsinfo *info)
4206 {
4207         int error;
4208         struct nfs_fattr *fattr = info->fattr;
4209
4210         error = nfs4_server_capabilities(server, mntfh);
4211         if (error < 0) {
4212                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4213                 return error;
4214         }
4215
4216         error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4217         if (error < 0) {
4218                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4219                 goto out;
4220         }
4221
4222         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4223             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4224                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4225
4226 out:
4227         return error;
4228 }
4229
4230 /*
4231  * Get locations and (maybe) other attributes of a referral.
4232  * Note that we'll actually follow the referral later when
4233  * we detect fsid mismatch in inode revalidation
4234  */
4235 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4236                              const struct qstr *name, struct nfs_fattr *fattr,
4237                              struct nfs_fh *fhandle)
4238 {
4239         int status = -ENOMEM;
4240         struct page *page = NULL;
4241         struct nfs4_fs_locations *locations = NULL;
4242
4243         page = alloc_page(GFP_KERNEL);
4244         if (page == NULL)
4245                 goto out;
4246         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4247         if (locations == NULL)
4248                 goto out;
4249
4250         locations->fattr = fattr;
4251
4252         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4253         if (status != 0)
4254                 goto out;
4255
4256         /*
4257          * If the fsid didn't change, this is a migration event, not a
4258          * referral.  Cause us to drop into the exception handler, which
4259          * will kick off migration recovery.
4260          */
4261         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4262                 dprintk("%s: server did not return a different fsid for"
4263                         " a referral at %s\n", __func__, name->name);
4264                 status = -NFS4ERR_MOVED;
4265                 goto out;
4266         }
4267         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4268         nfs_fixup_referral_attributes(fattr);
4269         memset(fhandle, 0, sizeof(struct nfs_fh));
4270 out:
4271         if (page)
4272                 __free_page(page);
4273         kfree(locations);
4274         return status;
4275 }
4276
4277 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4278                                 struct nfs_fattr *fattr, struct inode *inode)
4279 {
4280         __u32 bitmask[NFS4_BITMASK_SZ];
4281         struct nfs4_getattr_arg args = {
4282                 .fh = fhandle,
4283                 .bitmask = bitmask,
4284         };
4285         struct nfs4_getattr_res res = {
4286                 .fattr = fattr,
4287                 .server = server,
4288         };
4289         struct rpc_message msg = {
4290                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4291                 .rpc_argp = &args,
4292                 .rpc_resp = &res,
4293         };
4294         unsigned short task_flags = 0;
4295
4296         if (nfs4_has_session(server->nfs_client))
4297                 task_flags = RPC_TASK_MOVEABLE;
4298
4299         /* Is this is an attribute revalidation, subject to softreval? */
4300         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4301                 task_flags |= RPC_TASK_TIMEOUT;
4302
4303         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4304         nfs_fattr_init(fattr);
4305         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4306         return nfs4_do_call_sync(server->client, server, &msg,
4307                         &args.seq_args, &res.seq_res, task_flags);
4308 }
4309
4310 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4311                                 struct nfs_fattr *fattr, struct inode *inode)
4312 {
4313         struct nfs4_exception exception = {
4314                 .interruptible = true,
4315         };
4316         int err;
4317         do {
4318                 err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4319                 trace_nfs4_getattr(server, fhandle, fattr, err);
4320                 err = nfs4_handle_exception(server, err,
4321                                 &exception);
4322         } while (exception.retry);
4323         return err;
4324 }
4325
4326 /* 
4327  * The file is not closed if it is opened due to the a request to change
4328  * the size of the file. The open call will not be needed once the
4329  * VFS layer lookup-intents are implemented.
4330  *
4331  * Close is called when the inode is destroyed.
4332  * If we haven't opened the file for O_WRONLY, we
4333  * need to in the size_change case to obtain a stateid.
4334  *
4335  * Got race?
4336  * Because OPEN is always done by name in nfsv4, it is
4337  * possible that we opened a different file by the same
4338  * name.  We can recognize this race condition, but we
4339  * can't do anything about it besides returning an error.
4340  *
4341  * This will be fixed with VFS changes (lookup-intent).
4342  */
4343 static int
4344 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4345                   struct iattr *sattr)
4346 {
4347         struct inode *inode = d_inode(dentry);
4348         const struct cred *cred = NULL;
4349         struct nfs_open_context *ctx = NULL;
4350         int status;
4351
4352         if (pnfs_ld_layoutret_on_setattr(inode) &&
4353             sattr->ia_valid & ATTR_SIZE &&
4354             sattr->ia_size < i_size_read(inode))
4355                 pnfs_commit_and_return_layout(inode);
4356
4357         nfs_fattr_init(fattr);
4358         
4359         /* Deal with open(O_TRUNC) */
4360         if (sattr->ia_valid & ATTR_OPEN)
4361                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4362
4363         /* Optimization: if the end result is no change, don't RPC */
4364         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4365                 return 0;
4366
4367         /* Search for an existing open(O_WRITE) file */
4368         if (sattr->ia_valid & ATTR_FILE) {
4369
4370                 ctx = nfs_file_open_context(sattr->ia_file);
4371                 if (ctx)
4372                         cred = ctx->cred;
4373         }
4374
4375         /* Return any delegations if we're going to change ACLs */
4376         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4377                 nfs4_inode_make_writeable(inode);
4378
4379         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4380         if (status == 0) {
4381                 nfs_setattr_update_inode(inode, sattr, fattr);
4382                 nfs_setsecurity(inode, fattr);
4383         }
4384         return status;
4385 }
4386
4387 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4388                 struct dentry *dentry, struct nfs_fh *fhandle,
4389                 struct nfs_fattr *fattr)
4390 {
4391         struct nfs_server *server = NFS_SERVER(dir);
4392         int                    status;
4393         struct nfs4_lookup_arg args = {
4394                 .bitmask = server->attr_bitmask,
4395                 .dir_fh = NFS_FH(dir),
4396                 .name = &dentry->d_name,
4397         };
4398         struct nfs4_lookup_res res = {
4399                 .server = server,
4400                 .fattr = fattr,
4401                 .fh = fhandle,
4402         };
4403         struct rpc_message msg = {
4404                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4405                 .rpc_argp = &args,
4406                 .rpc_resp = &res,
4407         };
4408         unsigned short task_flags = 0;
4409
4410         if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4411                 task_flags = RPC_TASK_MOVEABLE;
4412
4413         /* Is this is an attribute revalidation, subject to softreval? */
4414         if (nfs_lookup_is_soft_revalidate(dentry))
4415                 task_flags |= RPC_TASK_TIMEOUT;
4416
4417         args.bitmask = nfs4_bitmask(server, fattr->label);
4418
4419         nfs_fattr_init(fattr);
4420
4421         dprintk("NFS call  lookup %pd2\n", dentry);
4422         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4423         status = nfs4_do_call_sync(clnt, server, &msg,
4424                         &args.seq_args, &res.seq_res, task_flags);
4425         dprintk("NFS reply lookup: %d\n", status);
4426         return status;
4427 }
4428
4429 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4430 {
4431         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4432                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4433         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4434         fattr->nlink = 2;
4435 }
4436
4437 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4438                                    struct dentry *dentry, struct nfs_fh *fhandle,
4439                                    struct nfs_fattr *fattr)
4440 {
4441         struct nfs4_exception exception = {
4442                 .interruptible = true,
4443         };
4444         struct rpc_clnt *client = *clnt;
4445         const struct qstr *name = &dentry->d_name;
4446         int err;
4447         do {
4448                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4449                 trace_nfs4_lookup(dir, name, err);
4450                 switch (err) {
4451                 case -NFS4ERR_BADNAME:
4452                         err = -ENOENT;
4453                         goto out;
4454                 case -NFS4ERR_MOVED:
4455                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4456                         if (err == -NFS4ERR_MOVED)
4457                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4458                         goto out;
4459                 case -NFS4ERR_WRONGSEC:
4460                         err = -EPERM;
4461                         if (client != *clnt)
4462                                 goto out;
4463                         client = nfs4_negotiate_security(client, dir, name);
4464                         if (IS_ERR(client))
4465                                 return PTR_ERR(client);
4466
4467                         exception.retry = 1;
4468                         break;
4469                 default:
4470                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4471                 }
4472         } while (exception.retry);
4473
4474 out:
4475         if (err == 0)
4476                 *clnt = client;
4477         else if (client != *clnt)
4478                 rpc_shutdown_client(client);
4479
4480         return err;
4481 }
4482
4483 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4484                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4485 {
4486         int status;
4487         struct rpc_clnt *client = NFS_CLIENT(dir);
4488
4489         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4490         if (client != NFS_CLIENT(dir)) {
4491                 rpc_shutdown_client(client);
4492                 nfs_fixup_secinfo_attributes(fattr);
4493         }
4494         return status;
4495 }
4496
4497 struct rpc_clnt *
4498 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4499                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4500 {
4501         struct rpc_clnt *client = NFS_CLIENT(dir);
4502         int status;
4503
4504         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4505         if (status < 0)
4506                 return ERR_PTR(status);
4507         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4508 }
4509
4510 static int _nfs4_proc_lookupp(struct inode *inode,
4511                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4512 {
4513         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4514         struct nfs_server *server = NFS_SERVER(inode);
4515         int                    status;
4516         struct nfs4_lookupp_arg args = {
4517                 .bitmask = server->attr_bitmask,
4518                 .fh = NFS_FH(inode),
4519         };
4520         struct nfs4_lookupp_res res = {
4521                 .server = server,
4522                 .fattr = fattr,
4523                 .fh = fhandle,
4524         };
4525         struct rpc_message msg = {
4526                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4527                 .rpc_argp = &args,
4528                 .rpc_resp = &res,
4529         };
4530         unsigned short task_flags = 0;
4531
4532         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4533                 task_flags |= RPC_TASK_TIMEOUT;
4534
4535         args.bitmask = nfs4_bitmask(server, fattr->label);
4536
4537         nfs_fattr_init(fattr);
4538
4539         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4540         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4541                                 &res.seq_res, task_flags);
4542         dprintk("NFS reply lookupp: %d\n", status);
4543         return status;
4544 }
4545
4546 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4547                              struct nfs_fattr *fattr)
4548 {
4549         struct nfs4_exception exception = {
4550                 .interruptible = true,
4551         };
4552         int err;
4553         do {
4554                 err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4555                 trace_nfs4_lookupp(inode, err);
4556                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4557                                 &exception);
4558         } while (exception.retry);
4559         return err;
4560 }
4561
4562 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4563                              const struct cred *cred)
4564 {
4565         struct nfs_server *server = NFS_SERVER(inode);
4566         struct nfs4_accessargs args = {
4567                 .fh = NFS_FH(inode),
4568                 .access = entry->mask,
4569         };
4570         struct nfs4_accessres res = {
4571                 .server = server,
4572         };
4573         struct rpc_message msg = {
4574                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4575                 .rpc_argp = &args,
4576                 .rpc_resp = &res,
4577                 .rpc_cred = cred,
4578         };
4579         int status = 0;
4580
4581         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4582                 res.fattr = nfs_alloc_fattr();
4583                 if (res.fattr == NULL)
4584                         return -ENOMEM;
4585                 args.bitmask = server->cache_consistency_bitmask;
4586         }
4587         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4588         if (!status) {
4589                 nfs_access_set_mask(entry, res.access);
4590                 if (res.fattr)
4591                         nfs_refresh_inode(inode, res.fattr);
4592         }
4593         nfs_free_fattr(res.fattr);
4594         return status;
4595 }
4596
4597 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4598                             const struct cred *cred)
4599 {
4600         struct nfs4_exception exception = {
4601                 .interruptible = true,
4602         };
4603         int err;
4604         do {
4605                 err = _nfs4_proc_access(inode, entry, cred);
4606                 trace_nfs4_access(inode, err);
4607                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4608                                 &exception);
4609         } while (exception.retry);
4610         return err;
4611 }
4612
4613 /*
4614  * TODO: For the time being, we don't try to get any attributes
4615  * along with any of the zero-copy operations READ, READDIR,
4616  * READLINK, WRITE.
4617  *
4618  * In the case of the first three, we want to put the GETATTR
4619  * after the read-type operation -- this is because it is hard
4620  * to predict the length of a GETATTR response in v4, and thus
4621  * align the READ data correctly.  This means that the GETATTR
4622  * may end up partially falling into the page cache, and we should
4623  * shift it into the 'tail' of the xdr_buf before processing.
4624  * To do this efficiently, we need to know the total length
4625  * of data received, which doesn't seem to be available outside
4626  * of the RPC layer.
4627  *
4628  * In the case of WRITE, we also want to put the GETATTR after
4629  * the operation -- in this case because we want to make sure
4630  * we get the post-operation mtime and size.
4631  *
4632  * Both of these changes to the XDR layer would in fact be quite
4633  * minor, but I decided to leave them for a subsequent patch.
4634  */
4635 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4636                 unsigned int pgbase, unsigned int pglen)
4637 {
4638         struct nfs4_readlink args = {
4639                 .fh       = NFS_FH(inode),
4640                 .pgbase   = pgbase,
4641                 .pglen    = pglen,
4642                 .pages    = &page,
4643         };
4644         struct nfs4_readlink_res res;
4645         struct rpc_message msg = {
4646                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4647                 .rpc_argp = &args,
4648                 .rpc_resp = &res,
4649         };
4650
4651         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4652 }
4653
4654 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4655                 unsigned int pgbase, unsigned int pglen)
4656 {
4657         struct nfs4_exception exception = {
4658                 .interruptible = true,
4659         };
4660         int err;
4661         do {
4662                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4663                 trace_nfs4_readlink(inode, err);
4664                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4665                                 &exception);
4666         } while (exception.retry);
4667         return err;
4668 }
4669
4670 /*
4671  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4672  */
4673 static int
4674 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4675                  int flags)
4676 {
4677         struct nfs_server *server = NFS_SERVER(dir);
4678         struct nfs4_label l, *ilabel = NULL;
4679         struct nfs_open_context *ctx;
4680         struct nfs4_state *state;
4681         int status = 0;
4682
4683         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4684         if (IS_ERR(ctx))
4685                 return PTR_ERR(ctx);
4686
4687         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4688
4689         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4690                 sattr->ia_mode &= ~current_umask();
4691         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4692         if (IS_ERR(state)) {
4693                 status = PTR_ERR(state);
4694                 goto out;
4695         }
4696 out:
4697         nfs4_label_release_security(ilabel);
4698         put_nfs_open_context(ctx);
4699         return status;
4700 }
4701
4702 static int
4703 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4704 {
4705         struct nfs_server *server = NFS_SERVER(dir);
4706         struct nfs_removeargs args = {
4707                 .fh = NFS_FH(dir),
4708                 .name = *name,
4709         };
4710         struct nfs_removeres res = {
4711                 .server = server,
4712         };
4713         struct rpc_message msg = {
4714                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4715                 .rpc_argp = &args,
4716                 .rpc_resp = &res,
4717         };
4718         unsigned long timestamp = jiffies;
4719         int status;
4720
4721         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4722         if (status == 0) {
4723                 spin_lock(&dir->i_lock);
4724                 /* Removing a directory decrements nlink in the parent */
4725                 if (ftype == NF4DIR && dir->i_nlink > 2)
4726                         nfs4_dec_nlink_locked(dir);
4727                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4728                                               NFS_INO_INVALID_DATA);
4729                 spin_unlock(&dir->i_lock);
4730         }
4731         return status;
4732 }
4733
4734 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4735 {
4736         struct nfs4_exception exception = {
4737                 .interruptible = true,
4738         };
4739         struct inode *inode = d_inode(dentry);
4740         int err;
4741
4742         if (inode) {
4743                 if (inode->i_nlink == 1)
4744                         nfs4_inode_return_delegation(inode);
4745                 else
4746                         nfs4_inode_make_writeable(inode);
4747         }
4748         do {
4749                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4750                 trace_nfs4_remove(dir, &dentry->d_name, err);
4751                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4752                                 &exception);
4753         } while (exception.retry);
4754         return err;
4755 }
4756
4757 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4758 {
4759         struct nfs4_exception exception = {
4760                 .interruptible = true,
4761         };
4762         int err;
4763
4764         do {
4765                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4766                 trace_nfs4_remove(dir, name, err);
4767                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4768                                 &exception);
4769         } while (exception.retry);
4770         return err;
4771 }
4772
4773 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4774                 struct dentry *dentry,
4775                 struct inode *inode)
4776 {
4777         struct nfs_removeargs *args = msg->rpc_argp;
4778         struct nfs_removeres *res = msg->rpc_resp;
4779
4780         res->server = NFS_SB(dentry->d_sb);
4781         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4782         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4783
4784         nfs_fattr_init(res->dir_attr);
4785
4786         if (inode) {
4787                 nfs4_inode_return_delegation(inode);
4788                 nfs_d_prune_case_insensitive_aliases(inode);
4789         }
4790 }
4791
4792 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4793 {
4794         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4795                         &data->args.seq_args,
4796                         &data->res.seq_res,
4797                         task);
4798 }
4799
4800 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4801 {
4802         struct nfs_unlinkdata *data = task->tk_calldata;
4803         struct nfs_removeres *res = &data->res;
4804
4805         if (!nfs4_sequence_done(task, &res->seq_res))
4806                 return 0;
4807         if (nfs4_async_handle_error(task, res->server, NULL,
4808                                     &data->timeout) == -EAGAIN)
4809                 return 0;
4810         if (task->tk_status == 0)
4811                 nfs4_update_changeattr(dir, &res->cinfo,
4812                                 res->dir_attr->time_start,
4813                                 NFS_INO_INVALID_DATA);
4814         return 1;
4815 }
4816
4817 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4818                 struct dentry *old_dentry,
4819                 struct dentry *new_dentry)
4820 {
4821         struct nfs_renameargs *arg = msg->rpc_argp;
4822         struct nfs_renameres *res = msg->rpc_resp;
4823         struct inode *old_inode = d_inode(old_dentry);
4824         struct inode *new_inode = d_inode(new_dentry);
4825
4826         if (old_inode)
4827                 nfs4_inode_make_writeable(old_inode);
4828         if (new_inode)
4829                 nfs4_inode_return_delegation(new_inode);
4830         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4831         res->server = NFS_SB(old_dentry->d_sb);
4832         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4833 }
4834
4835 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4836 {
4837         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4838                         &data->args.seq_args,
4839                         &data->res.seq_res,
4840                         task);
4841 }
4842
4843 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4844                                  struct inode *new_dir)
4845 {
4846         struct nfs_renamedata *data = task->tk_calldata;
4847         struct nfs_renameres *res = &data->res;
4848
4849         if (!nfs4_sequence_done(task, &res->seq_res))
4850                 return 0;
4851         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4852                 return 0;
4853
4854         if (task->tk_status == 0) {
4855                 nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4856                 if (new_dir != old_dir) {
4857                         /* Note: If we moved a directory, nlink will change */
4858                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4859                                         res->old_fattr->time_start,
4860                                         NFS_INO_INVALID_NLINK |
4861                                             NFS_INO_INVALID_DATA);
4862                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4863                                         res->new_fattr->time_start,
4864                                         NFS_INO_INVALID_NLINK |
4865                                             NFS_INO_INVALID_DATA);
4866                 } else
4867                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4868                                         res->old_fattr->time_start,
4869                                         NFS_INO_INVALID_DATA);
4870         }
4871         return 1;
4872 }
4873
4874 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4875 {
4876         struct nfs_server *server = NFS_SERVER(inode);
4877         __u32 bitmask[NFS4_BITMASK_SZ];
4878         struct nfs4_link_arg arg = {
4879                 .fh     = NFS_FH(inode),
4880                 .dir_fh = NFS_FH(dir),
4881                 .name   = name,
4882                 .bitmask = bitmask,
4883         };
4884         struct nfs4_link_res res = {
4885                 .server = server,
4886         };
4887         struct rpc_message msg = {
4888                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4889                 .rpc_argp = &arg,
4890                 .rpc_resp = &res,
4891         };
4892         int status = -ENOMEM;
4893
4894         res.fattr = nfs_alloc_fattr_with_label(server);
4895         if (res.fattr == NULL)
4896                 goto out;
4897
4898         nfs4_inode_make_writeable(inode);
4899         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4900                                 NFS_INO_INVALID_CHANGE);
4901         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4902         if (!status) {
4903                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4904                                        NFS_INO_INVALID_DATA);
4905                 nfs4_inc_nlink(inode);
4906                 status = nfs_post_op_update_inode(inode, res.fattr);
4907                 if (!status)
4908                         nfs_setsecurity(inode, res.fattr);
4909         }
4910
4911 out:
4912         nfs_free_fattr(res.fattr);
4913         return status;
4914 }
4915
4916 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4917 {
4918         struct nfs4_exception exception = {
4919                 .interruptible = true,
4920         };
4921         int err;
4922         do {
4923                 err = nfs4_handle_exception(NFS_SERVER(inode),
4924                                 _nfs4_proc_link(inode, dir, name),
4925                                 &exception);
4926         } while (exception.retry);
4927         return err;
4928 }
4929
4930 struct nfs4_createdata {
4931         struct rpc_message msg;
4932         struct nfs4_create_arg arg;
4933         struct nfs4_create_res res;
4934         struct nfs_fh fh;
4935         struct nfs_fattr fattr;
4936 };
4937
4938 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4939                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4940 {
4941         struct nfs4_createdata *data;
4942
4943         data = kzalloc(sizeof(*data), GFP_KERNEL);
4944         if (data != NULL) {
4945                 struct nfs_server *server = NFS_SERVER(dir);
4946
4947                 data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4948                 if (IS_ERR(data->fattr.label))
4949                         goto out_free;
4950
4951                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4952                 data->msg.rpc_argp = &data->arg;
4953                 data->msg.rpc_resp = &data->res;
4954                 data->arg.dir_fh = NFS_FH(dir);
4955                 data->arg.server = server;
4956                 data->arg.name = name;
4957                 data->arg.attrs = sattr;
4958                 data->arg.ftype = ftype;
4959                 data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4960                 data->arg.umask = current_umask();
4961                 data->res.server = server;
4962                 data->res.fh = &data->fh;
4963                 data->res.fattr = &data->fattr;
4964                 nfs_fattr_init(data->res.fattr);
4965         }
4966         return data;
4967 out_free:
4968         kfree(data);
4969         return NULL;
4970 }
4971
4972 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4973 {
4974         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4975                                     &data->arg.seq_args, &data->res.seq_res, 1);
4976         if (status == 0) {
4977                 spin_lock(&dir->i_lock);
4978                 /* Creating a directory bumps nlink in the parent */
4979                 if (data->arg.ftype == NF4DIR)
4980                         nfs4_inc_nlink_locked(dir);
4981                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4982                                               data->res.fattr->time_start,
4983                                               NFS_INO_INVALID_DATA);
4984                 spin_unlock(&dir->i_lock);
4985                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
4986         }
4987         return status;
4988 }
4989
4990 static void nfs4_free_createdata(struct nfs4_createdata *data)
4991 {
4992         nfs4_label_free(data->fattr.label);
4993         kfree(data);
4994 }
4995
4996 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4997                 struct page *page, unsigned int len, struct iattr *sattr,
4998                 struct nfs4_label *label)
4999 {
5000         struct nfs4_createdata *data;
5001         int status = -ENAMETOOLONG;
5002
5003         if (len > NFS4_MAXPATHLEN)
5004                 goto out;
5005
5006         status = -ENOMEM;
5007         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5008         if (data == NULL)
5009                 goto out;
5010
5011         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5012         data->arg.u.symlink.pages = &page;
5013         data->arg.u.symlink.len = len;
5014         data->arg.label = label;
5015         
5016         status = nfs4_do_create(dir, dentry, data);
5017
5018         nfs4_free_createdata(data);
5019 out:
5020         return status;
5021 }
5022
5023 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5024                 struct page *page, unsigned int len, struct iattr *sattr)
5025 {
5026         struct nfs4_exception exception = {
5027                 .interruptible = true,
5028         };
5029         struct nfs4_label l, *label = NULL;
5030         int err;
5031
5032         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5033
5034         do {
5035                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
5036                 trace_nfs4_symlink(dir, &dentry->d_name, err);
5037                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5038                                 &exception);
5039         } while (exception.retry);
5040
5041         nfs4_label_release_security(label);
5042         return err;
5043 }
5044
5045 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5046                 struct iattr *sattr, struct nfs4_label *label)
5047 {
5048         struct nfs4_createdata *data;
5049         int status = -ENOMEM;
5050
5051         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5052         if (data == NULL)
5053                 goto out;
5054
5055         data->arg.label = label;
5056         status = nfs4_do_create(dir, dentry, data);
5057
5058         nfs4_free_createdata(data);
5059 out:
5060         return status;
5061 }
5062
5063 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5064                 struct iattr *sattr)
5065 {
5066         struct nfs_server *server = NFS_SERVER(dir);
5067         struct nfs4_exception exception = {
5068                 .interruptible = true,
5069         };
5070         struct nfs4_label l, *label = NULL;
5071         int err;
5072
5073         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5074
5075         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5076                 sattr->ia_mode &= ~current_umask();
5077         do {
5078                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5079                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
5080                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5081                                 &exception);
5082         } while (exception.retry);
5083         nfs4_label_release_security(label);
5084
5085         return err;
5086 }
5087
5088 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5089                               struct nfs_readdir_res *nr_res)
5090 {
5091         struct inode            *dir = d_inode(nr_arg->dentry);
5092         struct nfs_server       *server = NFS_SERVER(dir);
5093         struct nfs4_readdir_arg args = {
5094                 .fh = NFS_FH(dir),
5095                 .pages = nr_arg->pages,
5096                 .pgbase = 0,
5097                 .count = nr_arg->page_len,
5098                 .plus = nr_arg->plus,
5099         };
5100         struct nfs4_readdir_res res;
5101         struct rpc_message msg = {
5102                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5103                 .rpc_argp = &args,
5104                 .rpc_resp = &res,
5105                 .rpc_cred = nr_arg->cred,
5106         };
5107         int                     status;
5108
5109         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5110                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5111         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5112                 args.bitmask = server->attr_bitmask_nl;
5113         else
5114                 args.bitmask = server->attr_bitmask;
5115
5116         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5117         res.pgbase = args.pgbase;
5118         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5119                         &res.seq_res, 0);
5120         if (status >= 0) {
5121                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5122                 status += args.pgbase;
5123         }
5124
5125         nfs_invalidate_atime(dir);
5126
5127         dprintk("%s: returns %d\n", __func__, status);
5128         return status;
5129 }
5130
5131 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5132                              struct nfs_readdir_res *res)
5133 {
5134         struct nfs4_exception exception = {
5135                 .interruptible = true,
5136         };
5137         int err;
5138         do {
5139                 err = _nfs4_proc_readdir(arg, res);
5140                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5141                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5142                                             err, &exception);
5143         } while (exception.retry);
5144         return err;
5145 }
5146
5147 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5148                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5149 {
5150         struct nfs4_createdata *data;
5151         int mode = sattr->ia_mode;
5152         int status = -ENOMEM;
5153
5154         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5155         if (data == NULL)
5156                 goto out;
5157
5158         if (S_ISFIFO(mode))
5159                 data->arg.ftype = NF4FIFO;
5160         else if (S_ISBLK(mode)) {
5161                 data->arg.ftype = NF4BLK;
5162                 data->arg.u.device.specdata1 = MAJOR(rdev);
5163                 data->arg.u.device.specdata2 = MINOR(rdev);
5164         }
5165         else if (S_ISCHR(mode)) {
5166                 data->arg.ftype = NF4CHR;
5167                 data->arg.u.device.specdata1 = MAJOR(rdev);
5168                 data->arg.u.device.specdata2 = MINOR(rdev);
5169         } else if (!S_ISSOCK(mode)) {
5170                 status = -EINVAL;
5171                 goto out_free;
5172         }
5173
5174         data->arg.label = label;
5175         status = nfs4_do_create(dir, dentry, data);
5176 out_free:
5177         nfs4_free_createdata(data);
5178 out:
5179         return status;
5180 }
5181
5182 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5183                 struct iattr *sattr, dev_t rdev)
5184 {
5185         struct nfs_server *server = NFS_SERVER(dir);
5186         struct nfs4_exception exception = {
5187                 .interruptible = true,
5188         };
5189         struct nfs4_label l, *label = NULL;
5190         int err;
5191
5192         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5193
5194         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5195                 sattr->ia_mode &= ~current_umask();
5196         do {
5197                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5198                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5199                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5200                                 &exception);
5201         } while (exception.retry);
5202
5203         nfs4_label_release_security(label);
5204
5205         return err;
5206 }
5207
5208 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5209                  struct nfs_fsstat *fsstat)
5210 {
5211         struct nfs4_statfs_arg args = {
5212                 .fh = fhandle,
5213                 .bitmask = server->attr_bitmask,
5214         };
5215         struct nfs4_statfs_res res = {
5216                 .fsstat = fsstat,
5217         };
5218         struct rpc_message msg = {
5219                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5220                 .rpc_argp = &args,
5221                 .rpc_resp = &res,
5222         };
5223
5224         nfs_fattr_init(fsstat->fattr);
5225         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5226 }
5227
5228 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5229 {
5230         struct nfs4_exception exception = {
5231                 .interruptible = true,
5232         };
5233         int err;
5234         do {
5235                 err = nfs4_handle_exception(server,
5236                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5237                                 &exception);
5238         } while (exception.retry);
5239         return err;
5240 }
5241
5242 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5243                 struct nfs_fsinfo *fsinfo)
5244 {
5245         struct nfs4_fsinfo_arg args = {
5246                 .fh = fhandle,
5247                 .bitmask = server->attr_bitmask,
5248         };
5249         struct nfs4_fsinfo_res res = {
5250                 .fsinfo = fsinfo,
5251         };
5252         struct rpc_message msg = {
5253                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5254                 .rpc_argp = &args,
5255                 .rpc_resp = &res,
5256         };
5257
5258         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5259 }
5260
5261 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5262 {
5263         struct nfs4_exception exception = {
5264                 .interruptible = true,
5265         };
5266         int err;
5267
5268         do {
5269                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5270                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5271                 if (err == 0) {
5272                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5273                         break;
5274                 }
5275                 err = nfs4_handle_exception(server, err, &exception);
5276         } while (exception.retry);
5277         return err;
5278 }
5279
5280 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5281 {
5282         int error;
5283
5284         nfs_fattr_init(fsinfo->fattr);
5285         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5286         if (error == 0) {
5287                 /* block layout checks this! */
5288                 server->pnfs_blksize = fsinfo->blksize;
5289                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5290         }
5291
5292         return error;
5293 }
5294
5295 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5296                 struct nfs_pathconf *pathconf)
5297 {
5298         struct nfs4_pathconf_arg args = {
5299                 .fh = fhandle,
5300                 .bitmask = server->attr_bitmask,
5301         };
5302         struct nfs4_pathconf_res res = {
5303                 .pathconf = pathconf,
5304         };
5305         struct rpc_message msg = {
5306                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5307                 .rpc_argp = &args,
5308                 .rpc_resp = &res,
5309         };
5310
5311         /* None of the pathconf attributes are mandatory to implement */
5312         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5313                 memset(pathconf, 0, sizeof(*pathconf));
5314                 return 0;
5315         }
5316
5317         nfs_fattr_init(pathconf->fattr);
5318         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5319 }
5320
5321 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5322                 struct nfs_pathconf *pathconf)
5323 {
5324         struct nfs4_exception exception = {
5325                 .interruptible = true,
5326         };
5327         int err;
5328
5329         do {
5330                 err = nfs4_handle_exception(server,
5331                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5332                                 &exception);
5333         } while (exception.retry);
5334         return err;
5335 }
5336
5337 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5338                 const struct nfs_open_context *ctx,
5339                 const struct nfs_lock_context *l_ctx,
5340                 fmode_t fmode)
5341 {
5342         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5343 }
5344 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5345
5346 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5347                 const struct nfs_open_context *ctx,
5348                 const struct nfs_lock_context *l_ctx,
5349                 fmode_t fmode)
5350 {
5351         nfs4_stateid _current_stateid;
5352
5353         /* If the current stateid represents a lost lock, then exit */
5354         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5355                 return true;
5356         return nfs4_stateid_match(stateid, &_current_stateid);
5357 }
5358
5359 static bool nfs4_error_stateid_expired(int err)
5360 {
5361         switch (err) {
5362         case -NFS4ERR_DELEG_REVOKED:
5363         case -NFS4ERR_ADMIN_REVOKED:
5364         case -NFS4ERR_BAD_STATEID:
5365         case -NFS4ERR_STALE_STATEID:
5366         case -NFS4ERR_OLD_STATEID:
5367         case -NFS4ERR_OPENMODE:
5368         case -NFS4ERR_EXPIRED:
5369                 return true;
5370         }
5371         return false;
5372 }
5373
5374 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5375 {
5376         struct nfs_server *server = NFS_SERVER(hdr->inode);
5377
5378         trace_nfs4_read(hdr, task->tk_status);
5379         if (task->tk_status < 0) {
5380                 struct nfs4_exception exception = {
5381                         .inode = hdr->inode,
5382                         .state = hdr->args.context->state,
5383                         .stateid = &hdr->args.stateid,
5384                 };
5385                 task->tk_status = nfs4_async_handle_exception(task,
5386                                 server, task->tk_status, &exception);
5387                 if (exception.retry) {
5388                         rpc_restart_call_prepare(task);
5389                         return -EAGAIN;
5390                 }
5391         }
5392
5393         if (task->tk_status > 0)
5394                 renew_lease(server, hdr->timestamp);
5395         return 0;
5396 }
5397
5398 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5399                 struct nfs_pgio_args *args)
5400 {
5401
5402         if (!nfs4_error_stateid_expired(task->tk_status) ||
5403                 nfs4_stateid_is_current(&args->stateid,
5404                                 args->context,
5405                                 args->lock_context,
5406                                 FMODE_READ))
5407                 return false;
5408         rpc_restart_call_prepare(task);
5409         return true;
5410 }
5411
5412 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5413                                          struct nfs_pgio_header *hdr)
5414 {
5415         struct nfs_server *server = NFS_SERVER(hdr->inode);
5416         struct rpc_message *msg = &task->tk_msg;
5417
5418         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5419             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5420                 server->caps &= ~NFS_CAP_READ_PLUS;
5421                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5422                 rpc_restart_call_prepare(task);
5423                 return true;
5424         }
5425         return false;
5426 }
5427
5428 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5429 {
5430         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5431                 return -EAGAIN;
5432         if (nfs4_read_stateid_changed(task, &hdr->args))
5433                 return -EAGAIN;
5434         if (nfs4_read_plus_not_supported(task, hdr))
5435                 return -EAGAIN;
5436         if (task->tk_status > 0)
5437                 nfs_invalidate_atime(hdr->inode);
5438         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5439                                     nfs4_read_done_cb(task, hdr);
5440 }
5441
5442 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5443 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5444                                     struct rpc_message *msg)
5445 {
5446         /* Note: We don't use READ_PLUS with pNFS yet */
5447         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5448                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5449 }
5450 #else
5451 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5452                                     struct rpc_message *msg)
5453 {
5454 }
5455 #endif /* CONFIG_NFS_V4_2 */
5456
5457 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5458                                  struct rpc_message *msg)
5459 {
5460         hdr->timestamp   = jiffies;
5461         if (!hdr->pgio_done_cb)
5462                 hdr->pgio_done_cb = nfs4_read_done_cb;
5463         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5464         nfs42_read_plus_support(hdr, msg);
5465         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5466 }
5467
5468 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5469                                       struct nfs_pgio_header *hdr)
5470 {
5471         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5472                         &hdr->args.seq_args,
5473                         &hdr->res.seq_res,
5474                         task))
5475                 return 0;
5476         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5477                                 hdr->args.lock_context,
5478                                 hdr->rw_mode) == -EIO)
5479                 return -EIO;
5480         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5481                 return -EIO;
5482         return 0;
5483 }
5484
5485 static int nfs4_write_done_cb(struct rpc_task *task,
5486                               struct nfs_pgio_header *hdr)
5487 {
5488         struct inode *inode = hdr->inode;
5489
5490         trace_nfs4_write(hdr, task->tk_status);
5491         if (task->tk_status < 0) {
5492                 struct nfs4_exception exception = {
5493                         .inode = hdr->inode,
5494                         .state = hdr->args.context->state,
5495                         .stateid = &hdr->args.stateid,
5496                 };
5497                 task->tk_status = nfs4_async_handle_exception(task,
5498                                 NFS_SERVER(inode), task->tk_status,
5499                                 &exception);
5500                 if (exception.retry) {
5501                         rpc_restart_call_prepare(task);
5502                         return -EAGAIN;
5503                 }
5504         }
5505         if (task->tk_status >= 0) {
5506                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5507                 nfs_writeback_update_inode(hdr);
5508         }
5509         return 0;
5510 }
5511
5512 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5513                 struct nfs_pgio_args *args)
5514 {
5515
5516         if (!nfs4_error_stateid_expired(task->tk_status) ||
5517                 nfs4_stateid_is_current(&args->stateid,
5518                                 args->context,
5519                                 args->lock_context,
5520                                 FMODE_WRITE))
5521                 return false;
5522         rpc_restart_call_prepare(task);
5523         return true;
5524 }
5525
5526 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5527 {
5528         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5529                 return -EAGAIN;
5530         if (nfs4_write_stateid_changed(task, &hdr->args))
5531                 return -EAGAIN;
5532         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5533                 nfs4_write_done_cb(task, hdr);
5534 }
5535
5536 static
5537 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5538 {
5539         /* Don't request attributes for pNFS or O_DIRECT writes */
5540         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5541                 return false;
5542         /* Otherwise, request attributes if and only if we don't hold
5543          * a delegation
5544          */
5545         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5546 }
5547
5548 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5549                       struct inode *inode, unsigned long cache_validity)
5550 {
5551         struct nfs_server *server = NFS_SERVER(inode);
5552         unsigned int i;
5553
5554         memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5555         cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5556
5557         if (cache_validity & NFS_INO_INVALID_CHANGE)
5558                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5559         if (cache_validity & NFS_INO_INVALID_ATIME)
5560                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5561         if (cache_validity & NFS_INO_INVALID_MODE)
5562                 bitmask[1] |= FATTR4_WORD1_MODE;
5563         if (cache_validity & NFS_INO_INVALID_OTHER)
5564                 bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5565         if (cache_validity & NFS_INO_INVALID_NLINK)
5566                 bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5567         if (cache_validity & NFS_INO_INVALID_CTIME)
5568                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5569         if (cache_validity & NFS_INO_INVALID_MTIME)
5570                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5571         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5572                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5573
5574         if (cache_validity & NFS_INO_INVALID_SIZE)
5575                 bitmask[0] |= FATTR4_WORD0_SIZE;
5576
5577         for (i = 0; i < NFS4_BITMASK_SZ; i++)
5578                 bitmask[i] &= server->attr_bitmask[i];
5579 }
5580
5581 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5582                                   struct rpc_message *msg,
5583                                   struct rpc_clnt **clnt)
5584 {
5585         struct nfs_server *server = NFS_SERVER(hdr->inode);
5586
5587         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5588                 hdr->args.bitmask = NULL;
5589                 hdr->res.fattr = NULL;
5590         } else {
5591                 nfs4_bitmask_set(hdr->args.bitmask_store,
5592                                  server->cache_consistency_bitmask,
5593                                  hdr->inode, NFS_INO_INVALID_BLOCKS);
5594                 hdr->args.bitmask = hdr->args.bitmask_store;
5595         }
5596
5597         if (!hdr->pgio_done_cb)
5598                 hdr->pgio_done_cb = nfs4_write_done_cb;
5599         hdr->res.server = server;
5600         hdr->timestamp   = jiffies;
5601
5602         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5603         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5604         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5605 }
5606
5607 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5608 {
5609         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5610                         &data->args.seq_args,
5611                         &data->res.seq_res,
5612                         task);
5613 }
5614
5615 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5616 {
5617         struct inode *inode = data->inode;
5618
5619         trace_nfs4_commit(data, task->tk_status);
5620         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5621                                     NULL, NULL) == -EAGAIN) {
5622                 rpc_restart_call_prepare(task);
5623                 return -EAGAIN;
5624         }
5625         return 0;
5626 }
5627
5628 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5629 {
5630         if (!nfs4_sequence_done(task, &data->res.seq_res))
5631                 return -EAGAIN;
5632         return data->commit_done_cb(task, data);
5633 }
5634
5635 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5636                                    struct rpc_clnt **clnt)
5637 {
5638         struct nfs_server *server = NFS_SERVER(data->inode);
5639
5640         if (data->commit_done_cb == NULL)
5641                 data->commit_done_cb = nfs4_commit_done_cb;
5642         data->res.server = server;
5643         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5644         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5645         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5646 }
5647
5648 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5649                                 struct nfs_commitres *res)
5650 {
5651         struct inode *dst_inode = file_inode(dst);
5652         struct nfs_server *server = NFS_SERVER(dst_inode);
5653         struct rpc_message msg = {
5654                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5655                 .rpc_argp = args,
5656                 .rpc_resp = res,
5657         };
5658
5659         args->fh = NFS_FH(dst_inode);
5660         return nfs4_call_sync(server->client, server, &msg,
5661                         &args->seq_args, &res->seq_res, 1);
5662 }
5663
5664 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5665 {
5666         struct nfs_commitargs args = {
5667                 .offset = offset,
5668                 .count = count,
5669         };
5670         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5671         struct nfs4_exception exception = { };
5672         int status;
5673
5674         do {
5675                 status = _nfs4_proc_commit(dst, &args, res);
5676                 status = nfs4_handle_exception(dst_server, status, &exception);
5677         } while (exception.retry);
5678
5679         return status;
5680 }
5681
5682 struct nfs4_renewdata {
5683         struct nfs_client       *client;
5684         unsigned long           timestamp;
5685 };
5686
5687 /*
5688  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5689  * standalone procedure for queueing an asynchronous RENEW.
5690  */
5691 static void nfs4_renew_release(void *calldata)
5692 {
5693         struct nfs4_renewdata *data = calldata;
5694         struct nfs_client *clp = data->client;
5695
5696         if (refcount_read(&clp->cl_count) > 1)
5697                 nfs4_schedule_state_renewal(clp);
5698         nfs_put_client(clp);
5699         kfree(data);
5700 }
5701
5702 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5703 {
5704         struct nfs4_renewdata *data = calldata;
5705         struct nfs_client *clp = data->client;
5706         unsigned long timestamp = data->timestamp;
5707
5708         trace_nfs4_renew_async(clp, task->tk_status);
5709         switch (task->tk_status) {
5710         case 0:
5711                 break;
5712         case -NFS4ERR_LEASE_MOVED:
5713                 nfs4_schedule_lease_moved_recovery(clp);
5714                 break;
5715         default:
5716                 /* Unless we're shutting down, schedule state recovery! */
5717                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5718                         return;
5719                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5720                         nfs4_schedule_lease_recovery(clp);
5721                         return;
5722                 }
5723                 nfs4_schedule_path_down_recovery(clp);
5724         }
5725         do_renew_lease(clp, timestamp);
5726 }
5727
5728 static const struct rpc_call_ops nfs4_renew_ops = {
5729         .rpc_call_done = nfs4_renew_done,
5730         .rpc_release = nfs4_renew_release,
5731 };
5732
5733 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5734 {
5735         struct rpc_message msg = {
5736                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5737                 .rpc_argp       = clp,
5738                 .rpc_cred       = cred,
5739         };
5740         struct nfs4_renewdata *data;
5741
5742         if (renew_flags == 0)
5743                 return 0;
5744         if (!refcount_inc_not_zero(&clp->cl_count))
5745                 return -EIO;
5746         data = kmalloc(sizeof(*data), GFP_NOFS);
5747         if (data == NULL) {
5748                 nfs_put_client(clp);
5749                 return -ENOMEM;
5750         }
5751         data->client = clp;
5752         data->timestamp = jiffies;
5753         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5754                         &nfs4_renew_ops, data);
5755 }
5756
5757 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5758 {
5759         struct rpc_message msg = {
5760                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5761                 .rpc_argp       = clp,
5762                 .rpc_cred       = cred,
5763         };
5764         unsigned long now = jiffies;
5765         int status;
5766
5767         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5768         if (status < 0)
5769                 return status;
5770         do_renew_lease(clp, now);
5771         return 0;
5772 }
5773
5774 static bool nfs4_server_supports_acls(const struct nfs_server *server,
5775                                       enum nfs4_acl_type type)
5776 {
5777         switch (type) {
5778         default:
5779                 return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5780         case NFS4ACL_DACL:
5781                 return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5782         case NFS4ACL_SACL:
5783                 return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5784         }
5785 }
5786
5787 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5788  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5789  * the stack.
5790  */
5791 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5792
5793 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5794                 struct page **pages)
5795 {
5796         struct page *newpage, **spages;
5797         int rc = 0;
5798         size_t len;
5799         spages = pages;
5800
5801         do {
5802                 len = min_t(size_t, PAGE_SIZE, buflen);
5803                 newpage = alloc_page(GFP_KERNEL);
5804
5805                 if (newpage == NULL)
5806                         goto unwind;
5807                 memcpy(page_address(newpage), buf, len);
5808                 buf += len;
5809                 buflen -= len;
5810                 *pages++ = newpage;
5811                 rc++;
5812         } while (buflen != 0);
5813
5814         return rc;
5815
5816 unwind:
5817         for(; rc > 0; rc--)
5818                 __free_page(spages[rc-1]);
5819         return -ENOMEM;
5820 }
5821
5822 struct nfs4_cached_acl {
5823         enum nfs4_acl_type type;
5824         int cached;
5825         size_t len;
5826         char data[];
5827 };
5828
5829 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5830 {
5831         struct nfs_inode *nfsi = NFS_I(inode);
5832
5833         spin_lock(&inode->i_lock);
5834         kfree(nfsi->nfs4_acl);
5835         nfsi->nfs4_acl = acl;
5836         spin_unlock(&inode->i_lock);
5837 }
5838
5839 static void nfs4_zap_acl_attr(struct inode *inode)
5840 {
5841         nfs4_set_cached_acl(inode, NULL);
5842 }
5843
5844 static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5845                                     size_t buflen, enum nfs4_acl_type type)
5846 {
5847         struct nfs_inode *nfsi = NFS_I(inode);
5848         struct nfs4_cached_acl *acl;
5849         int ret = -ENOENT;
5850
5851         spin_lock(&inode->i_lock);
5852         acl = nfsi->nfs4_acl;
5853         if (acl == NULL)
5854                 goto out;
5855         if (acl->type != type)
5856                 goto out;
5857         if (buf == NULL) /* user is just asking for length */
5858                 goto out_len;
5859         if (acl->cached == 0)
5860                 goto out;
5861         ret = -ERANGE; /* see getxattr(2) man page */
5862         if (acl->len > buflen)
5863                 goto out;
5864         memcpy(buf, acl->data, acl->len);
5865 out_len:
5866         ret = acl->len;
5867 out:
5868         spin_unlock(&inode->i_lock);
5869         return ret;
5870 }
5871
5872 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
5873                                   size_t pgbase, size_t acl_len,
5874                                   enum nfs4_acl_type type)
5875 {
5876         struct nfs4_cached_acl *acl;
5877         size_t buflen = sizeof(*acl) + acl_len;
5878
5879         if (buflen <= PAGE_SIZE) {
5880                 acl = kmalloc(buflen, GFP_KERNEL);
5881                 if (acl == NULL)
5882                         goto out;
5883                 acl->cached = 1;
5884                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5885         } else {
5886                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5887                 if (acl == NULL)
5888                         goto out;
5889                 acl->cached = 0;
5890         }
5891         acl->type = type;
5892         acl->len = acl_len;
5893 out:
5894         nfs4_set_cached_acl(inode, acl);
5895 }
5896
5897 /*
5898  * The getxattr API returns the required buffer length when called with a
5899  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5900  * the required buf.  On a NULL buf, we send a page of data to the server
5901  * guessing that the ACL request can be serviced by a page. If so, we cache
5902  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5903  * the cache. If not so, we throw away the page, and cache the required
5904  * length. The next getxattr call will then produce another round trip to
5905  * the server, this time with the input buf of the required size.
5906  */
5907 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
5908                                        size_t buflen, enum nfs4_acl_type type)
5909 {
5910         struct page **pages;
5911         struct nfs_getaclargs args = {
5912                 .fh = NFS_FH(inode),
5913                 .acl_type = type,
5914                 .acl_len = buflen,
5915         };
5916         struct nfs_getaclres res = {
5917                 .acl_type = type,
5918                 .acl_len = buflen,
5919         };
5920         struct rpc_message msg = {
5921                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5922                 .rpc_argp = &args,
5923                 .rpc_resp = &res,
5924         };
5925         unsigned int npages;
5926         int ret = -ENOMEM, i;
5927         struct nfs_server *server = NFS_SERVER(inode);
5928
5929         if (buflen == 0)
5930                 buflen = server->rsize;
5931
5932         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5933         pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
5934         if (!pages)
5935                 return -ENOMEM;
5936
5937         args.acl_pages = pages;
5938
5939         for (i = 0; i < npages; i++) {
5940                 pages[i] = alloc_page(GFP_KERNEL);
5941                 if (!pages[i])
5942                         goto out_free;
5943         }
5944
5945         /* for decoding across pages */
5946         res.acl_scratch = alloc_page(GFP_KERNEL);
5947         if (!res.acl_scratch)
5948                 goto out_free;
5949
5950         args.acl_len = npages * PAGE_SIZE;
5951
5952         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5953                 __func__, buf, buflen, npages, args.acl_len);
5954         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5955                              &msg, &args.seq_args, &res.seq_res, 0);
5956         if (ret)
5957                 goto out_free;
5958
5959         /* Handle the case where the passed-in buffer is too short */
5960         if (res.acl_flags & NFS4_ACL_TRUNC) {
5961                 /* Did the user only issue a request for the acl length? */
5962                 if (buf == NULL)
5963                         goto out_ok;
5964                 ret = -ERANGE;
5965                 goto out_free;
5966         }
5967         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
5968                               type);
5969         if (buf) {
5970                 if (res.acl_len > buflen) {
5971                         ret = -ERANGE;
5972                         goto out_free;
5973                 }
5974                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5975         }
5976 out_ok:
5977         ret = res.acl_len;
5978 out_free:
5979         for (i = 0; i < npages; i++)
5980                 if (pages[i])
5981                         __free_page(pages[i]);
5982         if (res.acl_scratch)
5983                 __free_page(res.acl_scratch);
5984         kfree(pages);
5985         return ret;
5986 }
5987
5988 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
5989                                      size_t buflen, enum nfs4_acl_type type)
5990 {
5991         struct nfs4_exception exception = {
5992                 .interruptible = true,
5993         };
5994         ssize_t ret;
5995         do {
5996                 ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
5997                 trace_nfs4_get_acl(inode, ret);
5998                 if (ret >= 0)
5999                         break;
6000                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6001         } while (exception.retry);
6002         return ret;
6003 }
6004
6005 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6006                                  enum nfs4_acl_type type)
6007 {
6008         struct nfs_server *server = NFS_SERVER(inode);
6009         int ret;
6010
6011         if (!nfs4_server_supports_acls(server, type))
6012                 return -EOPNOTSUPP;
6013         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6014         if (ret < 0)
6015                 return ret;
6016         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6017                 nfs_zap_acl_cache(inode);
6018         ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6019         if (ret != -ENOENT)
6020                 /* -ENOENT is returned if there is no ACL or if there is an ACL
6021                  * but no cached acl data, just the acl length */
6022                 return ret;
6023         return nfs4_get_acl_uncached(inode, buf, buflen, type);
6024 }
6025
6026 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6027                                size_t buflen, enum nfs4_acl_type type)
6028 {
6029         struct nfs_server *server = NFS_SERVER(inode);
6030         struct page *pages[NFS4ACL_MAXPAGES];
6031         struct nfs_setaclargs arg = {
6032                 .fh = NFS_FH(inode),
6033                 .acl_type = type,
6034                 .acl_len = buflen,
6035                 .acl_pages = pages,
6036         };
6037         struct nfs_setaclres res;
6038         struct rpc_message msg = {
6039                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6040                 .rpc_argp       = &arg,
6041                 .rpc_resp       = &res,
6042         };
6043         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6044         int ret, i;
6045
6046         /* You can't remove system.nfs4_acl: */
6047         if (buflen == 0)
6048                 return -EINVAL;
6049         if (!nfs4_server_supports_acls(server, type))
6050                 return -EOPNOTSUPP;
6051         if (npages > ARRAY_SIZE(pages))
6052                 return -ERANGE;
6053         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6054         if (i < 0)
6055                 return i;
6056         nfs4_inode_make_writeable(inode);
6057         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6058
6059         /*
6060          * Free each page after tx, so the only ref left is
6061          * held by the network stack
6062          */
6063         for (; i > 0; i--)
6064                 put_page(pages[i-1]);
6065
6066         /*
6067          * Acl update can result in inode attribute update.
6068          * so mark the attribute cache invalid.
6069          */
6070         spin_lock(&inode->i_lock);
6071         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6072                                              NFS_INO_INVALID_CTIME |
6073                                              NFS_INO_REVAL_FORCED);
6074         spin_unlock(&inode->i_lock);
6075         nfs_access_zap_cache(inode);
6076         nfs_zap_acl_cache(inode);
6077         return ret;
6078 }
6079
6080 static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6081                              size_t buflen, enum nfs4_acl_type type)
6082 {
6083         struct nfs4_exception exception = { };
6084         int err;
6085         do {
6086                 err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6087                 trace_nfs4_set_acl(inode, err);
6088                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6089                         /*
6090                          * no need to retry since the kernel
6091                          * isn't involved in encoding the ACEs.
6092                          */
6093                         err = -EINVAL;
6094                         break;
6095                 }
6096                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6097                                 &exception);
6098         } while (exception.retry);
6099         return err;
6100 }
6101
6102 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6103 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6104                                         size_t buflen)
6105 {
6106         struct nfs_server *server = NFS_SERVER(inode);
6107         struct nfs4_label label = {0, 0, buflen, buf};
6108
6109         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6110         struct nfs_fattr fattr = {
6111                 .label = &label,
6112         };
6113         struct nfs4_getattr_arg arg = {
6114                 .fh             = NFS_FH(inode),
6115                 .bitmask        = bitmask,
6116         };
6117         struct nfs4_getattr_res res = {
6118                 .fattr          = &fattr,
6119                 .server         = server,
6120         };
6121         struct rpc_message msg = {
6122                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6123                 .rpc_argp       = &arg,
6124                 .rpc_resp       = &res,
6125         };
6126         int ret;
6127
6128         nfs_fattr_init(&fattr);
6129
6130         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6131         if (ret)
6132                 return ret;
6133         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6134                 return -ENOENT;
6135         return label.len;
6136 }
6137
6138 static int nfs4_get_security_label(struct inode *inode, void *buf,
6139                                         size_t buflen)
6140 {
6141         struct nfs4_exception exception = {
6142                 .interruptible = true,
6143         };
6144         int err;
6145
6146         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6147                 return -EOPNOTSUPP;
6148
6149         do {
6150                 err = _nfs4_get_security_label(inode, buf, buflen);
6151                 trace_nfs4_get_security_label(inode, err);
6152                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6153                                 &exception);
6154         } while (exception.retry);
6155         return err;
6156 }
6157
6158 static int _nfs4_do_set_security_label(struct inode *inode,
6159                 struct nfs4_label *ilabel,
6160                 struct nfs_fattr *fattr)
6161 {
6162
6163         struct iattr sattr = {0};
6164         struct nfs_server *server = NFS_SERVER(inode);
6165         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6166         struct nfs_setattrargs arg = {
6167                 .fh             = NFS_FH(inode),
6168                 .iap            = &sattr,
6169                 .server         = server,
6170                 .bitmask        = bitmask,
6171                 .label          = ilabel,
6172         };
6173         struct nfs_setattrres res = {
6174                 .fattr          = fattr,
6175                 .server         = server,
6176         };
6177         struct rpc_message msg = {
6178                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6179                 .rpc_argp       = &arg,
6180                 .rpc_resp       = &res,
6181         };
6182         int status;
6183
6184         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6185
6186         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6187         if (status)
6188                 dprintk("%s failed: %d\n", __func__, status);
6189
6190         return status;
6191 }
6192
6193 static int nfs4_do_set_security_label(struct inode *inode,
6194                 struct nfs4_label *ilabel,
6195                 struct nfs_fattr *fattr)
6196 {
6197         struct nfs4_exception exception = { };
6198         int err;
6199
6200         do {
6201                 err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6202                 trace_nfs4_set_security_label(inode, err);
6203                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6204                                 &exception);
6205         } while (exception.retry);
6206         return err;
6207 }
6208
6209 static int
6210 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6211 {
6212         struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6213         struct nfs_fattr *fattr;
6214         int status;
6215
6216         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6217                 return -EOPNOTSUPP;
6218
6219         fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6220         if (fattr == NULL)
6221                 return -ENOMEM;
6222
6223         status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6224         if (status == 0)
6225                 nfs_setsecurity(inode, fattr);
6226
6227         return status;
6228 }
6229 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6230
6231
6232 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6233                                     nfs4_verifier *bootverf)
6234 {
6235         __be32 verf[2];
6236
6237         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6238                 /* An impossible timestamp guarantees this value
6239                  * will never match a generated boot time. */
6240                 verf[0] = cpu_to_be32(U32_MAX);
6241                 verf[1] = cpu_to_be32(U32_MAX);
6242         } else {
6243                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6244                 u64 ns = ktime_to_ns(nn->boot_time);
6245
6246                 verf[0] = cpu_to_be32(ns >> 32);
6247                 verf[1] = cpu_to_be32(ns);
6248         }
6249         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6250 }
6251
6252 static size_t
6253 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6254 {
6255         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6256         struct nfs_netns_client *nn_clp = nn->nfs_client;
6257         const char *id;
6258
6259         buf[0] = '\0';
6260
6261         if (nn_clp) {
6262                 rcu_read_lock();
6263                 id = rcu_dereference(nn_clp->identifier);
6264                 if (id)
6265                         strscpy(buf, id, buflen);
6266                 rcu_read_unlock();
6267         }
6268
6269         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6270                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6271
6272         return strlen(buf);
6273 }
6274
6275 static int
6276 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6277 {
6278         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6279         size_t buflen;
6280         size_t len;
6281         char *str;
6282
6283         if (clp->cl_owner_id != NULL)
6284                 return 0;
6285
6286         rcu_read_lock();
6287         len = 14 +
6288                 strlen(clp->cl_rpcclient->cl_nodename) +
6289                 1 +
6290                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6291                 1;
6292         rcu_read_unlock();
6293
6294         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6295         if (buflen)
6296                 len += buflen + 1;
6297
6298         if (len > NFS4_OPAQUE_LIMIT + 1)
6299                 return -EINVAL;
6300
6301         /*
6302          * Since this string is allocated at mount time, and held until the
6303          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6304          * about a memory-reclaim deadlock.
6305          */
6306         str = kmalloc(len, GFP_KERNEL);
6307         if (!str)
6308                 return -ENOMEM;
6309
6310         rcu_read_lock();
6311         if (buflen)
6312                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6313                           clp->cl_rpcclient->cl_nodename, buf,
6314                           rpc_peeraddr2str(clp->cl_rpcclient,
6315                                            RPC_DISPLAY_ADDR));
6316         else
6317                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6318                           clp->cl_rpcclient->cl_nodename,
6319                           rpc_peeraddr2str(clp->cl_rpcclient,
6320                                            RPC_DISPLAY_ADDR));
6321         rcu_read_unlock();
6322
6323         clp->cl_owner_id = str;
6324         return 0;
6325 }
6326
6327 static int
6328 nfs4_init_uniform_client_string(struct nfs_client *clp)
6329 {
6330         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6331         size_t buflen;
6332         size_t len;
6333         char *str;
6334
6335         if (clp->cl_owner_id != NULL)
6336                 return 0;
6337
6338         len = 10 + 10 + 1 + 10 + 1 +
6339                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6340
6341         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6342         if (buflen)
6343                 len += buflen + 1;
6344
6345         if (len > NFS4_OPAQUE_LIMIT + 1)
6346                 return -EINVAL;
6347
6348         /*
6349          * Since this string is allocated at mount time, and held until the
6350          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6351          * about a memory-reclaim deadlock.
6352          */
6353         str = kmalloc(len, GFP_KERNEL);
6354         if (!str)
6355                 return -ENOMEM;
6356
6357         if (buflen)
6358                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6359                           clp->rpc_ops->version, clp->cl_minorversion,
6360                           buf, clp->cl_rpcclient->cl_nodename);
6361         else
6362                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6363                           clp->rpc_ops->version, clp->cl_minorversion,
6364                           clp->cl_rpcclient->cl_nodename);
6365         clp->cl_owner_id = str;
6366         return 0;
6367 }
6368
6369 /*
6370  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6371  * services.  Advertise one based on the address family of the
6372  * clientaddr.
6373  */
6374 static unsigned int
6375 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6376 {
6377         if (strchr(clp->cl_ipaddr, ':') != NULL)
6378                 return scnprintf(buf, len, "tcp6");
6379         else
6380                 return scnprintf(buf, len, "tcp");
6381 }
6382
6383 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6384 {
6385         struct nfs4_setclientid *sc = calldata;
6386
6387         if (task->tk_status == 0)
6388                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6389 }
6390
6391 static const struct rpc_call_ops nfs4_setclientid_ops = {
6392         .rpc_call_done = nfs4_setclientid_done,
6393 };
6394
6395 /**
6396  * nfs4_proc_setclientid - Negotiate client ID
6397  * @clp: state data structure
6398  * @program: RPC program for NFSv4 callback service
6399  * @port: IP port number for NFS4 callback service
6400  * @cred: credential to use for this call
6401  * @res: where to place the result
6402  *
6403  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6404  */
6405 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6406                 unsigned short port, const struct cred *cred,
6407                 struct nfs4_setclientid_res *res)
6408 {
6409         nfs4_verifier sc_verifier;
6410         struct nfs4_setclientid setclientid = {
6411                 .sc_verifier = &sc_verifier,
6412                 .sc_prog = program,
6413                 .sc_clnt = clp,
6414         };
6415         struct rpc_message msg = {
6416                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6417                 .rpc_argp = &setclientid,
6418                 .rpc_resp = res,
6419                 .rpc_cred = cred,
6420         };
6421         struct rpc_task_setup task_setup_data = {
6422                 .rpc_client = clp->cl_rpcclient,
6423                 .rpc_message = &msg,
6424                 .callback_ops = &nfs4_setclientid_ops,
6425                 .callback_data = &setclientid,
6426                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6427         };
6428         unsigned long now = jiffies;
6429         int status;
6430
6431         /* nfs_client_id4 */
6432         nfs4_init_boot_verifier(clp, &sc_verifier);
6433
6434         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6435                 status = nfs4_init_uniform_client_string(clp);
6436         else
6437                 status = nfs4_init_nonuniform_client_string(clp);
6438
6439         if (status)
6440                 goto out;
6441
6442         /* cb_client4 */
6443         setclientid.sc_netid_len =
6444                                 nfs4_init_callback_netid(clp,
6445                                                 setclientid.sc_netid,
6446                                                 sizeof(setclientid.sc_netid));
6447         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6448                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6449                                 clp->cl_ipaddr, port >> 8, port & 255);
6450
6451         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6452                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6453                 clp->cl_owner_id);
6454
6455         status = nfs4_call_sync_custom(&task_setup_data);
6456         if (setclientid.sc_cred) {
6457                 kfree(clp->cl_acceptor);
6458                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6459                 put_rpccred(setclientid.sc_cred);
6460         }
6461
6462         if (status == 0)
6463                 do_renew_lease(clp, now);
6464 out:
6465         trace_nfs4_setclientid(clp, status);
6466         dprintk("NFS reply setclientid: %d\n", status);
6467         return status;
6468 }
6469
6470 /**
6471  * nfs4_proc_setclientid_confirm - Confirm client ID
6472  * @clp: state data structure
6473  * @arg: result of a previous SETCLIENTID
6474  * @cred: credential to use for this call
6475  *
6476  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6477  */
6478 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6479                 struct nfs4_setclientid_res *arg,
6480                 const struct cred *cred)
6481 {
6482         struct rpc_message msg = {
6483                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6484                 .rpc_argp = arg,
6485                 .rpc_cred = cred,
6486         };
6487         int status;
6488
6489         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6490                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6491                 clp->cl_clientid);
6492         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6493                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6494         trace_nfs4_setclientid_confirm(clp, status);
6495         dprintk("NFS reply setclientid_confirm: %d\n", status);
6496         return status;
6497 }
6498
6499 struct nfs4_delegreturndata {
6500         struct nfs4_delegreturnargs args;
6501         struct nfs4_delegreturnres res;
6502         struct nfs_fh fh;
6503         nfs4_stateid stateid;
6504         unsigned long timestamp;
6505         struct {
6506                 struct nfs4_layoutreturn_args arg;
6507                 struct nfs4_layoutreturn_res res;
6508                 struct nfs4_xdr_opaque_data ld_private;
6509                 u32 roc_barrier;
6510                 bool roc;
6511         } lr;
6512         struct nfs_fattr fattr;
6513         int rpc_status;
6514         struct inode *inode;
6515 };
6516
6517 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6518 {
6519         struct nfs4_delegreturndata *data = calldata;
6520         struct nfs4_exception exception = {
6521                 .inode = data->inode,
6522                 .stateid = &data->stateid,
6523                 .task_is_privileged = data->args.seq_args.sa_privileged,
6524         };
6525
6526         if (!nfs4_sequence_done(task, &data->res.seq_res))
6527                 return;
6528
6529         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6530
6531         /* Handle Layoutreturn errors */
6532         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6533                           &data->res.lr_ret) == -EAGAIN)
6534                 goto out_restart;
6535
6536         switch (task->tk_status) {
6537         case 0:
6538                 renew_lease(data->res.server, data->timestamp);
6539                 break;
6540         case -NFS4ERR_ADMIN_REVOKED:
6541         case -NFS4ERR_DELEG_REVOKED:
6542         case -NFS4ERR_EXPIRED:
6543                 nfs4_free_revoked_stateid(data->res.server,
6544                                 data->args.stateid,
6545                                 task->tk_msg.rpc_cred);
6546                 fallthrough;
6547         case -NFS4ERR_BAD_STATEID:
6548         case -NFS4ERR_STALE_STATEID:
6549         case -ETIMEDOUT:
6550                 task->tk_status = 0;
6551                 break;
6552         case -NFS4ERR_OLD_STATEID:
6553                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6554                         nfs4_stateid_seqid_inc(&data->stateid);
6555                 if (data->args.bitmask) {
6556                         data->args.bitmask = NULL;
6557                         data->res.fattr = NULL;
6558                 }
6559                 goto out_restart;
6560         case -NFS4ERR_ACCESS:
6561                 if (data->args.bitmask) {
6562                         data->args.bitmask = NULL;
6563                         data->res.fattr = NULL;
6564                         goto out_restart;
6565                 }
6566                 fallthrough;
6567         default:
6568                 task->tk_status = nfs4_async_handle_exception(task,
6569                                 data->res.server, task->tk_status,
6570                                 &exception);
6571                 if (exception.retry)
6572                         goto out_restart;
6573         }
6574         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6575         data->rpc_status = task->tk_status;
6576         return;
6577 out_restart:
6578         task->tk_status = 0;
6579         rpc_restart_call_prepare(task);
6580 }
6581
6582 static void nfs4_delegreturn_release(void *calldata)
6583 {
6584         struct nfs4_delegreturndata *data = calldata;
6585         struct inode *inode = data->inode;
6586
6587         if (data->lr.roc)
6588                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6589                                  data->res.lr_ret);
6590         if (inode) {
6591                 nfs4_fattr_set_prechange(&data->fattr,
6592                                          inode_peek_iversion_raw(inode));
6593                 nfs_refresh_inode(inode, &data->fattr);
6594                 nfs_iput_and_deactive(inode);
6595         }
6596         kfree(calldata);
6597 }
6598
6599 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6600 {
6601         struct nfs4_delegreturndata *d_data;
6602         struct pnfs_layout_hdr *lo;
6603
6604         d_data = (struct nfs4_delegreturndata *)data;
6605
6606         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6607                 nfs4_sequence_done(task, &d_data->res.seq_res);
6608                 return;
6609         }
6610
6611         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6612         if (lo && !pnfs_layout_is_valid(lo)) {
6613                 d_data->args.lr_args = NULL;
6614                 d_data->res.lr_res = NULL;
6615         }
6616
6617         nfs4_setup_sequence(d_data->res.server->nfs_client,
6618                         &d_data->args.seq_args,
6619                         &d_data->res.seq_res,
6620                         task);
6621 }
6622
6623 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6624         .rpc_call_prepare = nfs4_delegreturn_prepare,
6625         .rpc_call_done = nfs4_delegreturn_done,
6626         .rpc_release = nfs4_delegreturn_release,
6627 };
6628
6629 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6630 {
6631         struct nfs4_delegreturndata *data;
6632         struct nfs_server *server = NFS_SERVER(inode);
6633         struct rpc_task *task;
6634         struct rpc_message msg = {
6635                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6636                 .rpc_cred = cred,
6637         };
6638         struct rpc_task_setup task_setup_data = {
6639                 .rpc_client = server->client,
6640                 .rpc_message = &msg,
6641                 .callback_ops = &nfs4_delegreturn_ops,
6642                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6643         };
6644         int status = 0;
6645
6646         if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6647                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6648
6649         data = kzalloc(sizeof(*data), GFP_KERNEL);
6650         if (data == NULL)
6651                 return -ENOMEM;
6652
6653         nfs4_state_protect(server->nfs_client,
6654                         NFS_SP4_MACH_CRED_CLEANUP,
6655                         &task_setup_data.rpc_client, &msg);
6656
6657         data->args.fhandle = &data->fh;
6658         data->args.stateid = &data->stateid;
6659         nfs4_bitmask_set(data->args.bitmask_store,
6660                          server->cache_consistency_bitmask, inode, 0);
6661         data->args.bitmask = data->args.bitmask_store;
6662         nfs_copy_fh(&data->fh, NFS_FH(inode));
6663         nfs4_stateid_copy(&data->stateid, stateid);
6664         data->res.fattr = &data->fattr;
6665         data->res.server = server;
6666         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6667         data->lr.arg.ld_private = &data->lr.ld_private;
6668         nfs_fattr_init(data->res.fattr);
6669         data->timestamp = jiffies;
6670         data->rpc_status = 0;
6671         data->inode = nfs_igrab_and_active(inode);
6672         if (data->inode || issync) {
6673                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6674                                         cred);
6675                 if (data->lr.roc) {
6676                         data->args.lr_args = &data->lr.arg;
6677                         data->res.lr_res = &data->lr.res;
6678                 }
6679         }
6680
6681         if (!data->inode)
6682                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6683                                    1);
6684         else
6685                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6686                                    0);
6687         task_setup_data.callback_data = data;
6688         msg.rpc_argp = &data->args;
6689         msg.rpc_resp = &data->res;
6690         task = rpc_run_task(&task_setup_data);
6691         if (IS_ERR(task))
6692                 return PTR_ERR(task);
6693         if (!issync)
6694                 goto out;
6695         status = rpc_wait_for_completion_task(task);
6696         if (status != 0)
6697                 goto out;
6698         status = data->rpc_status;
6699 out:
6700         rpc_put_task(task);
6701         return status;
6702 }
6703
6704 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6705 {
6706         struct nfs_server *server = NFS_SERVER(inode);
6707         struct nfs4_exception exception = { };
6708         int err;
6709         do {
6710                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6711                 trace_nfs4_delegreturn(inode, stateid, err);
6712                 switch (err) {
6713                         case -NFS4ERR_STALE_STATEID:
6714                         case -NFS4ERR_EXPIRED:
6715                         case 0:
6716                                 return 0;
6717                 }
6718                 err = nfs4_handle_exception(server, err, &exception);
6719         } while (exception.retry);
6720         return err;
6721 }
6722
6723 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6724 {
6725         struct inode *inode = state->inode;
6726         struct nfs_server *server = NFS_SERVER(inode);
6727         struct nfs_client *clp = server->nfs_client;
6728         struct nfs_lockt_args arg = {
6729                 .fh = NFS_FH(inode),
6730                 .fl = request,
6731         };
6732         struct nfs_lockt_res res = {
6733                 .denied = request,
6734         };
6735         struct rpc_message msg = {
6736                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6737                 .rpc_argp       = &arg,
6738                 .rpc_resp       = &res,
6739                 .rpc_cred       = state->owner->so_cred,
6740         };
6741         struct nfs4_lock_state *lsp;
6742         int status;
6743
6744         arg.lock_owner.clientid = clp->cl_clientid;
6745         status = nfs4_set_lock_state(state, request);
6746         if (status != 0)
6747                 goto out;
6748         lsp = request->fl_u.nfs4_fl.owner;
6749         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6750         arg.lock_owner.s_dev = server->s_dev;
6751         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6752         switch (status) {
6753                 case 0:
6754                         request->fl_type = F_UNLCK;
6755                         break;
6756                 case -NFS4ERR_DENIED:
6757                         status = 0;
6758         }
6759         request->fl_ops->fl_release_private(request);
6760         request->fl_ops = NULL;
6761 out:
6762         return status;
6763 }
6764
6765 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6766 {
6767         struct nfs4_exception exception = {
6768                 .interruptible = true,
6769         };
6770         int err;
6771
6772         do {
6773                 err = _nfs4_proc_getlk(state, cmd, request);
6774                 trace_nfs4_get_lock(request, state, cmd, err);
6775                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6776                                 &exception);
6777         } while (exception.retry);
6778         return err;
6779 }
6780
6781 /*
6782  * Update the seqid of a lock stateid after receiving
6783  * NFS4ERR_OLD_STATEID
6784  */
6785 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6786                 struct nfs4_lock_state *lsp)
6787 {
6788         struct nfs4_state *state = lsp->ls_state;
6789         bool ret = false;
6790
6791         spin_lock(&state->state_lock);
6792         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6793                 goto out;
6794         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6795                 nfs4_stateid_seqid_inc(dst);
6796         else
6797                 dst->seqid = lsp->ls_stateid.seqid;
6798         ret = true;
6799 out:
6800         spin_unlock(&state->state_lock);
6801         return ret;
6802 }
6803
6804 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6805                 struct nfs4_lock_state *lsp)
6806 {
6807         struct nfs4_state *state = lsp->ls_state;
6808         bool ret;
6809
6810         spin_lock(&state->state_lock);
6811         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6812         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6813         spin_unlock(&state->state_lock);
6814         return ret;
6815 }
6816
6817 struct nfs4_unlockdata {
6818         struct nfs_locku_args arg;
6819         struct nfs_locku_res res;
6820         struct nfs4_lock_state *lsp;
6821         struct nfs_open_context *ctx;
6822         struct nfs_lock_context *l_ctx;
6823         struct file_lock fl;
6824         struct nfs_server *server;
6825         unsigned long timestamp;
6826 };
6827
6828 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6829                 struct nfs_open_context *ctx,
6830                 struct nfs4_lock_state *lsp,
6831                 struct nfs_seqid *seqid)
6832 {
6833         struct nfs4_unlockdata *p;
6834         struct nfs4_state *state = lsp->ls_state;
6835         struct inode *inode = state->inode;
6836
6837         p = kzalloc(sizeof(*p), GFP_KERNEL);
6838         if (p == NULL)
6839                 return NULL;
6840         p->arg.fh = NFS_FH(inode);
6841         p->arg.fl = &p->fl;
6842         p->arg.seqid = seqid;
6843         p->res.seqid = seqid;
6844         p->lsp = lsp;
6845         /* Ensure we don't close file until we're done freeing locks! */
6846         p->ctx = get_nfs_open_context(ctx);
6847         p->l_ctx = nfs_get_lock_context(ctx);
6848         locks_init_lock(&p->fl);
6849         locks_copy_lock(&p->fl, fl);
6850         p->server = NFS_SERVER(inode);
6851         spin_lock(&state->state_lock);
6852         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6853         spin_unlock(&state->state_lock);
6854         return p;
6855 }
6856
6857 static void nfs4_locku_release_calldata(void *data)
6858 {
6859         struct nfs4_unlockdata *calldata = data;
6860         nfs_free_seqid(calldata->arg.seqid);
6861         nfs4_put_lock_state(calldata->lsp);
6862         nfs_put_lock_context(calldata->l_ctx);
6863         put_nfs_open_context(calldata->ctx);
6864         kfree(calldata);
6865 }
6866
6867 static void nfs4_locku_done(struct rpc_task *task, void *data)
6868 {
6869         struct nfs4_unlockdata *calldata = data;
6870         struct nfs4_exception exception = {
6871                 .inode = calldata->lsp->ls_state->inode,
6872                 .stateid = &calldata->arg.stateid,
6873         };
6874
6875         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6876                 return;
6877         switch (task->tk_status) {
6878                 case 0:
6879                         renew_lease(calldata->server, calldata->timestamp);
6880                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6881                         if (nfs4_update_lock_stateid(calldata->lsp,
6882                                         &calldata->res.stateid))
6883                                 break;
6884                         fallthrough;
6885                 case -NFS4ERR_ADMIN_REVOKED:
6886                 case -NFS4ERR_EXPIRED:
6887                         nfs4_free_revoked_stateid(calldata->server,
6888                                         &calldata->arg.stateid,
6889                                         task->tk_msg.rpc_cred);
6890                         fallthrough;
6891                 case -NFS4ERR_BAD_STATEID:
6892                 case -NFS4ERR_STALE_STATEID:
6893                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6894                                                 calldata->lsp))
6895                                 rpc_restart_call_prepare(task);
6896                         break;
6897                 case -NFS4ERR_OLD_STATEID:
6898                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6899                                                 calldata->lsp))
6900                                 rpc_restart_call_prepare(task);
6901                         break;
6902                 default:
6903                         task->tk_status = nfs4_async_handle_exception(task,
6904                                         calldata->server, task->tk_status,
6905                                         &exception);
6906                         if (exception.retry)
6907                                 rpc_restart_call_prepare(task);
6908         }
6909         nfs_release_seqid(calldata->arg.seqid);
6910 }
6911
6912 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6913 {
6914         struct nfs4_unlockdata *calldata = data;
6915
6916         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6917                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6918                 return;
6919
6920         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6921                 goto out_wait;
6922         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6923                 /* Note: exit _without_ running nfs4_locku_done */
6924                 goto out_no_action;
6925         }
6926         calldata->timestamp = jiffies;
6927         if (nfs4_setup_sequence(calldata->server->nfs_client,
6928                                 &calldata->arg.seq_args,
6929                                 &calldata->res.seq_res,
6930                                 task) != 0)
6931                 nfs_release_seqid(calldata->arg.seqid);
6932         return;
6933 out_no_action:
6934         task->tk_action = NULL;
6935 out_wait:
6936         nfs4_sequence_done(task, &calldata->res.seq_res);
6937 }
6938
6939 static const struct rpc_call_ops nfs4_locku_ops = {
6940         .rpc_call_prepare = nfs4_locku_prepare,
6941         .rpc_call_done = nfs4_locku_done,
6942         .rpc_release = nfs4_locku_release_calldata,
6943 };
6944
6945 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6946                 struct nfs_open_context *ctx,
6947                 struct nfs4_lock_state *lsp,
6948                 struct nfs_seqid *seqid)
6949 {
6950         struct nfs4_unlockdata *data;
6951         struct rpc_message msg = {
6952                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6953                 .rpc_cred = ctx->cred,
6954         };
6955         struct rpc_task_setup task_setup_data = {
6956                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6957                 .rpc_message = &msg,
6958                 .callback_ops = &nfs4_locku_ops,
6959                 .workqueue = nfsiod_workqueue,
6960                 .flags = RPC_TASK_ASYNC,
6961         };
6962
6963         if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
6964                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6965
6966         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6967                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6968
6969         /* Ensure this is an unlock - when canceling a lock, the
6970          * canceled lock is passed in, and it won't be an unlock.
6971          */
6972         fl->fl_type = F_UNLCK;
6973         if (fl->fl_flags & FL_CLOSE)
6974                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6975
6976         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6977         if (data == NULL) {
6978                 nfs_free_seqid(seqid);
6979                 return ERR_PTR(-ENOMEM);
6980         }
6981
6982         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6983         msg.rpc_argp = &data->arg;
6984         msg.rpc_resp = &data->res;
6985         task_setup_data.callback_data = data;
6986         return rpc_run_task(&task_setup_data);
6987 }
6988
6989 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6990 {
6991         struct inode *inode = state->inode;
6992         struct nfs4_state_owner *sp = state->owner;
6993         struct nfs_inode *nfsi = NFS_I(inode);
6994         struct nfs_seqid *seqid;
6995         struct nfs4_lock_state *lsp;
6996         struct rpc_task *task;
6997         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6998         int status = 0;
6999         unsigned char fl_flags = request->fl_flags;
7000
7001         status = nfs4_set_lock_state(state, request);
7002         /* Unlock _before_ we do the RPC call */
7003         request->fl_flags |= FL_EXISTS;
7004         /* Exclude nfs_delegation_claim_locks() */
7005         mutex_lock(&sp->so_delegreturn_mutex);
7006         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7007         down_read(&nfsi->rwsem);
7008         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7009                 up_read(&nfsi->rwsem);
7010                 mutex_unlock(&sp->so_delegreturn_mutex);
7011                 goto out;
7012         }
7013         up_read(&nfsi->rwsem);
7014         mutex_unlock(&sp->so_delegreturn_mutex);
7015         if (status != 0)
7016                 goto out;
7017         /* Is this a delegated lock? */
7018         lsp = request->fl_u.nfs4_fl.owner;
7019         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7020                 goto out;
7021         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7022         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7023         status = -ENOMEM;
7024         if (IS_ERR(seqid))
7025                 goto out;
7026         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
7027         status = PTR_ERR(task);
7028         if (IS_ERR(task))
7029                 goto out;
7030         status = rpc_wait_for_completion_task(task);
7031         rpc_put_task(task);
7032 out:
7033         request->fl_flags = fl_flags;
7034         trace_nfs4_unlock(request, state, F_SETLK, status);
7035         return status;
7036 }
7037
7038 struct nfs4_lockdata {
7039         struct nfs_lock_args arg;
7040         struct nfs_lock_res res;
7041         struct nfs4_lock_state *lsp;
7042         struct nfs_open_context *ctx;
7043         struct file_lock fl;
7044         unsigned long timestamp;
7045         int rpc_status;
7046         int cancelled;
7047         struct nfs_server *server;
7048 };
7049
7050 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7051                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7052                 gfp_t gfp_mask)
7053 {
7054         struct nfs4_lockdata *p;
7055         struct inode *inode = lsp->ls_state->inode;
7056         struct nfs_server *server = NFS_SERVER(inode);
7057         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7058
7059         p = kzalloc(sizeof(*p), gfp_mask);
7060         if (p == NULL)
7061                 return NULL;
7062
7063         p->arg.fh = NFS_FH(inode);
7064         p->arg.fl = &p->fl;
7065         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7066         if (IS_ERR(p->arg.open_seqid))
7067                 goto out_free;
7068         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7069         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7070         if (IS_ERR(p->arg.lock_seqid))
7071                 goto out_free_seqid;
7072         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7073         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7074         p->arg.lock_owner.s_dev = server->s_dev;
7075         p->res.lock_seqid = p->arg.lock_seqid;
7076         p->lsp = lsp;
7077         p->server = server;
7078         p->ctx = get_nfs_open_context(ctx);
7079         locks_init_lock(&p->fl);
7080         locks_copy_lock(&p->fl, fl);
7081         return p;
7082 out_free_seqid:
7083         nfs_free_seqid(p->arg.open_seqid);
7084 out_free:
7085         kfree(p);
7086         return NULL;
7087 }
7088
7089 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7090 {
7091         struct nfs4_lockdata *data = calldata;
7092         struct nfs4_state *state = data->lsp->ls_state;
7093
7094         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7095                 goto out_wait;
7096         /* Do we need to do an open_to_lock_owner? */
7097         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7098                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7099                         goto out_release_lock_seqid;
7100                 }
7101                 nfs4_stateid_copy(&data->arg.open_stateid,
7102                                 &state->open_stateid);
7103                 data->arg.new_lock_owner = 1;
7104                 data->res.open_seqid = data->arg.open_seqid;
7105         } else {
7106                 data->arg.new_lock_owner = 0;
7107                 nfs4_stateid_copy(&data->arg.lock_stateid,
7108                                 &data->lsp->ls_stateid);
7109         }
7110         if (!nfs4_valid_open_stateid(state)) {
7111                 data->rpc_status = -EBADF;
7112                 task->tk_action = NULL;
7113                 goto out_release_open_seqid;
7114         }
7115         data->timestamp = jiffies;
7116         if (nfs4_setup_sequence(data->server->nfs_client,
7117                                 &data->arg.seq_args,
7118                                 &data->res.seq_res,
7119                                 task) == 0)
7120                 return;
7121 out_release_open_seqid:
7122         nfs_release_seqid(data->arg.open_seqid);
7123 out_release_lock_seqid:
7124         nfs_release_seqid(data->arg.lock_seqid);
7125 out_wait:
7126         nfs4_sequence_done(task, &data->res.seq_res);
7127         dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7128 }
7129
7130 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7131 {
7132         struct nfs4_lockdata *data = calldata;
7133         struct nfs4_lock_state *lsp = data->lsp;
7134
7135         if (!nfs4_sequence_done(task, &data->res.seq_res))
7136                 return;
7137
7138         data->rpc_status = task->tk_status;
7139         switch (task->tk_status) {
7140         case 0:
7141                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7142                                 data->timestamp);
7143                 if (data->arg.new_lock && !data->cancelled) {
7144                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7145                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7146                                 goto out_restart;
7147                 }
7148                 if (data->arg.new_lock_owner != 0) {
7149                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
7150                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7151                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7152                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7153                         goto out_restart;
7154                 break;
7155         case -NFS4ERR_BAD_STATEID:
7156         case -NFS4ERR_OLD_STATEID:
7157         case -NFS4ERR_STALE_STATEID:
7158         case -NFS4ERR_EXPIRED:
7159                 if (data->arg.new_lock_owner != 0) {
7160                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7161                                                 &lsp->ls_state->open_stateid))
7162                                 goto out_restart;
7163                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7164                                                 &lsp->ls_stateid))
7165                                 goto out_restart;
7166         }
7167 out_done:
7168         dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7169         return;
7170 out_restart:
7171         if (!data->cancelled)
7172                 rpc_restart_call_prepare(task);
7173         goto out_done;
7174 }
7175
7176 static void nfs4_lock_release(void *calldata)
7177 {
7178         struct nfs4_lockdata *data = calldata;
7179
7180         nfs_free_seqid(data->arg.open_seqid);
7181         if (data->cancelled && data->rpc_status == 0) {
7182                 struct rpc_task *task;
7183                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7184                                 data->arg.lock_seqid);
7185                 if (!IS_ERR(task))
7186                         rpc_put_task_async(task);
7187                 dprintk("%s: cancelling lock!\n", __func__);
7188         } else
7189                 nfs_free_seqid(data->arg.lock_seqid);
7190         nfs4_put_lock_state(data->lsp);
7191         put_nfs_open_context(data->ctx);
7192         kfree(data);
7193 }
7194
7195 static const struct rpc_call_ops nfs4_lock_ops = {
7196         .rpc_call_prepare = nfs4_lock_prepare,
7197         .rpc_call_done = nfs4_lock_done,
7198         .rpc_release = nfs4_lock_release,
7199 };
7200
7201 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7202 {
7203         switch (error) {
7204         case -NFS4ERR_ADMIN_REVOKED:
7205         case -NFS4ERR_EXPIRED:
7206         case -NFS4ERR_BAD_STATEID:
7207                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7208                 if (new_lock_owner != 0 ||
7209                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7210                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7211                 break;
7212         case -NFS4ERR_STALE_STATEID:
7213                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7214                 nfs4_schedule_lease_recovery(server->nfs_client);
7215         }
7216 }
7217
7218 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7219 {
7220         struct nfs4_lockdata *data;
7221         struct rpc_task *task;
7222         struct rpc_message msg = {
7223                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7224                 .rpc_cred = state->owner->so_cred,
7225         };
7226         struct rpc_task_setup task_setup_data = {
7227                 .rpc_client = NFS_CLIENT(state->inode),
7228                 .rpc_message = &msg,
7229                 .callback_ops = &nfs4_lock_ops,
7230                 .workqueue = nfsiod_workqueue,
7231                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7232         };
7233         int ret;
7234
7235         if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7236                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
7237
7238         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7239                                    fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7240         if (data == NULL)
7241                 return -ENOMEM;
7242         if (IS_SETLKW(cmd))
7243                 data->arg.block = 1;
7244         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7245                                 recovery_type > NFS_LOCK_NEW);
7246         msg.rpc_argp = &data->arg;
7247         msg.rpc_resp = &data->res;
7248         task_setup_data.callback_data = data;
7249         if (recovery_type > NFS_LOCK_NEW) {
7250                 if (recovery_type == NFS_LOCK_RECLAIM)
7251                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7252         } else
7253                 data->arg.new_lock = 1;
7254         task = rpc_run_task(&task_setup_data);
7255         if (IS_ERR(task))
7256                 return PTR_ERR(task);
7257         ret = rpc_wait_for_completion_task(task);
7258         if (ret == 0) {
7259                 ret = data->rpc_status;
7260                 if (ret)
7261                         nfs4_handle_setlk_error(data->server, data->lsp,
7262                                         data->arg.new_lock_owner, ret);
7263         } else
7264                 data->cancelled = true;
7265         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7266         rpc_put_task(task);
7267         dprintk("%s: ret = %d\n", __func__, ret);
7268         return ret;
7269 }
7270
7271 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7272 {
7273         struct nfs_server *server = NFS_SERVER(state->inode);
7274         struct nfs4_exception exception = {
7275                 .inode = state->inode,
7276         };
7277         int err;
7278
7279         do {
7280                 /* Cache the lock if possible... */
7281                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7282                         return 0;
7283                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7284                 if (err != -NFS4ERR_DELAY)
7285                         break;
7286                 nfs4_handle_exception(server, err, &exception);
7287         } while (exception.retry);
7288         return err;
7289 }
7290
7291 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7292 {
7293         struct nfs_server *server = NFS_SERVER(state->inode);
7294         struct nfs4_exception exception = {
7295                 .inode = state->inode,
7296         };
7297         int err;
7298
7299         err = nfs4_set_lock_state(state, request);
7300         if (err != 0)
7301                 return err;
7302         if (!recover_lost_locks) {
7303                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7304                 return 0;
7305         }
7306         do {
7307                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7308                         return 0;
7309                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7310                 switch (err) {
7311                 default:
7312                         goto out;
7313                 case -NFS4ERR_GRACE:
7314                 case -NFS4ERR_DELAY:
7315                         nfs4_handle_exception(server, err, &exception);
7316                         err = 0;
7317                 }
7318         } while (exception.retry);
7319 out:
7320         return err;
7321 }
7322
7323 #if defined(CONFIG_NFS_V4_1)
7324 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7325 {
7326         struct nfs4_lock_state *lsp;
7327         int status;
7328
7329         status = nfs4_set_lock_state(state, request);
7330         if (status != 0)
7331                 return status;
7332         lsp = request->fl_u.nfs4_fl.owner;
7333         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7334             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7335                 return 0;
7336         return nfs4_lock_expired(state, request);
7337 }
7338 #endif
7339
7340 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7341 {
7342         struct nfs_inode *nfsi = NFS_I(state->inode);
7343         struct nfs4_state_owner *sp = state->owner;
7344         unsigned char fl_flags = request->fl_flags;
7345         int status;
7346
7347         request->fl_flags |= FL_ACCESS;
7348         status = locks_lock_inode_wait(state->inode, request);
7349         if (status < 0)
7350                 goto out;
7351         mutex_lock(&sp->so_delegreturn_mutex);
7352         down_read(&nfsi->rwsem);
7353         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7354                 /* Yes: cache locks! */
7355                 /* ...but avoid races with delegation recall... */
7356                 request->fl_flags = fl_flags & ~FL_SLEEP;
7357                 status = locks_lock_inode_wait(state->inode, request);
7358                 up_read(&nfsi->rwsem);
7359                 mutex_unlock(&sp->so_delegreturn_mutex);
7360                 goto out;
7361         }
7362         up_read(&nfsi->rwsem);
7363         mutex_unlock(&sp->so_delegreturn_mutex);
7364         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7365 out:
7366         request->fl_flags = fl_flags;
7367         return status;
7368 }
7369
7370 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7371 {
7372         struct nfs4_exception exception = {
7373                 .state = state,
7374                 .inode = state->inode,
7375                 .interruptible = true,
7376         };
7377         int err;
7378
7379         do {
7380                 err = _nfs4_proc_setlk(state, cmd, request);
7381                 if (err == -NFS4ERR_DENIED)
7382                         err = -EAGAIN;
7383                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7384                                 err, &exception);
7385         } while (exception.retry);
7386         return err;
7387 }
7388
7389 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7390 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7391
7392 static int
7393 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7394                         struct file_lock *request)
7395 {
7396         int             status = -ERESTARTSYS;
7397         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7398
7399         while(!signalled()) {
7400                 status = nfs4_proc_setlk(state, cmd, request);
7401                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7402                         break;
7403                 freezable_schedule_timeout_interruptible(timeout);
7404                 timeout *= 2;
7405                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7406                 status = -ERESTARTSYS;
7407         }
7408         return status;
7409 }
7410
7411 #ifdef CONFIG_NFS_V4_1
7412 struct nfs4_lock_waiter {
7413         struct inode            *inode;
7414         struct nfs_lowner       owner;
7415         wait_queue_entry_t      wait;
7416 };
7417
7418 static int
7419 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7420 {
7421         struct nfs4_lock_waiter *waiter =
7422                 container_of(wait, struct nfs4_lock_waiter, wait);
7423
7424         /* NULL key means to wake up everyone */
7425         if (key) {
7426                 struct cb_notify_lock_args      *cbnl = key;
7427                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7428                                                 *wowner = &waiter->owner;
7429
7430                 /* Only wake if the callback was for the same owner. */
7431                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7432                         return 0;
7433
7434                 /* Make sure it's for the right inode */
7435                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7436                         return 0;
7437         }
7438
7439         return woken_wake_function(wait, mode, flags, key);
7440 }
7441
7442 static int
7443 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7444 {
7445         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7446         struct nfs_server *server = NFS_SERVER(state->inode);
7447         struct nfs_client *clp = server->nfs_client;
7448         wait_queue_head_t *q = &clp->cl_lock_waitq;
7449         struct nfs4_lock_waiter waiter = {
7450                 .inode = state->inode,
7451                 .owner = { .clientid = clp->cl_clientid,
7452                            .id = lsp->ls_seqid.owner_id,
7453                            .s_dev = server->s_dev },
7454         };
7455         int status;
7456
7457         /* Don't bother with waitqueue if we don't expect a callback */
7458         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7459                 return nfs4_retry_setlk_simple(state, cmd, request);
7460
7461         init_wait(&waiter.wait);
7462         waiter.wait.func = nfs4_wake_lock_waiter;
7463         add_wait_queue(q, &waiter.wait);
7464
7465         do {
7466                 status = nfs4_proc_setlk(state, cmd, request);
7467                 if (status != -EAGAIN || IS_SETLK(cmd))
7468                         break;
7469
7470                 status = -ERESTARTSYS;
7471                 freezer_do_not_count();
7472                 wait_woken(&waiter.wait, TASK_INTERRUPTIBLE,
7473                            NFS4_LOCK_MAXTIMEOUT);
7474                 freezer_count();
7475         } while (!signalled());
7476
7477         remove_wait_queue(q, &waiter.wait);
7478
7479         return status;
7480 }
7481 #else /* !CONFIG_NFS_V4_1 */
7482 static inline int
7483 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7484 {
7485         return nfs4_retry_setlk_simple(state, cmd, request);
7486 }
7487 #endif
7488
7489 static int
7490 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7491 {
7492         struct nfs_open_context *ctx;
7493         struct nfs4_state *state;
7494         int status;
7495
7496         /* verify open state */
7497         ctx = nfs_file_open_context(filp);
7498         state = ctx->state;
7499
7500         if (IS_GETLK(cmd)) {
7501                 if (state != NULL)
7502                         return nfs4_proc_getlk(state, F_GETLK, request);
7503                 return 0;
7504         }
7505
7506         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7507                 return -EINVAL;
7508
7509         if (request->fl_type == F_UNLCK) {
7510                 if (state != NULL)
7511                         return nfs4_proc_unlck(state, cmd, request);
7512                 return 0;
7513         }
7514
7515         if (state == NULL)
7516                 return -ENOLCK;
7517
7518         if ((request->fl_flags & FL_POSIX) &&
7519             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7520                 return -ENOLCK;
7521
7522         /*
7523          * Don't rely on the VFS having checked the file open mode,
7524          * since it won't do this for flock() locks.
7525          */
7526         switch (request->fl_type) {
7527         case F_RDLCK:
7528                 if (!(filp->f_mode & FMODE_READ))
7529                         return -EBADF;
7530                 break;
7531         case F_WRLCK:
7532                 if (!(filp->f_mode & FMODE_WRITE))
7533                         return -EBADF;
7534         }
7535
7536         status = nfs4_set_lock_state(state, request);
7537         if (status != 0)
7538                 return status;
7539
7540         return nfs4_retry_setlk(state, cmd, request);
7541 }
7542
7543 static int nfs4_delete_lease(struct file *file, void **priv)
7544 {
7545         return generic_setlease(file, F_UNLCK, NULL, priv);
7546 }
7547
7548 static int nfs4_add_lease(struct file *file, long arg, struct file_lock **lease,
7549                           void **priv)
7550 {
7551         struct inode *inode = file_inode(file);
7552         fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7553         int ret;
7554
7555         /* No delegation, no lease */
7556         if (!nfs4_have_delegation(inode, type))
7557                 return -EAGAIN;
7558         ret = generic_setlease(file, arg, lease, priv);
7559         if (ret || nfs4_have_delegation(inode, type))
7560                 return ret;
7561         /* We raced with a delegation return */
7562         nfs4_delete_lease(file, priv);
7563         return -EAGAIN;
7564 }
7565
7566 int nfs4_proc_setlease(struct file *file, long arg, struct file_lock **lease,
7567                        void **priv)
7568 {
7569         switch (arg) {
7570         case F_RDLCK:
7571         case F_WRLCK:
7572                 return nfs4_add_lease(file, arg, lease, priv);
7573         case F_UNLCK:
7574                 return nfs4_delete_lease(file, priv);
7575         default:
7576                 return -EINVAL;
7577         }
7578 }
7579
7580 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7581 {
7582         struct nfs_server *server = NFS_SERVER(state->inode);
7583         int err;
7584
7585         err = nfs4_set_lock_state(state, fl);
7586         if (err != 0)
7587                 return err;
7588         do {
7589                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7590                 if (err != -NFS4ERR_DELAY)
7591                         break;
7592                 ssleep(1);
7593         } while (err == -NFS4ERR_DELAY);
7594         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7595 }
7596
7597 struct nfs_release_lockowner_data {
7598         struct nfs4_lock_state *lsp;
7599         struct nfs_server *server;
7600         struct nfs_release_lockowner_args args;
7601         struct nfs_release_lockowner_res res;
7602         unsigned long timestamp;
7603 };
7604
7605 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7606 {
7607         struct nfs_release_lockowner_data *data = calldata;
7608         struct nfs_server *server = data->server;
7609         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7610                            &data->res.seq_res, task);
7611         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7612         data->timestamp = jiffies;
7613 }
7614
7615 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7616 {
7617         struct nfs_release_lockowner_data *data = calldata;
7618         struct nfs_server *server = data->server;
7619
7620         nfs40_sequence_done(task, &data->res.seq_res);
7621
7622         switch (task->tk_status) {
7623         case 0:
7624                 renew_lease(server, data->timestamp);
7625                 break;
7626         case -NFS4ERR_STALE_CLIENTID:
7627         case -NFS4ERR_EXPIRED:
7628                 nfs4_schedule_lease_recovery(server->nfs_client);
7629                 break;
7630         case -NFS4ERR_LEASE_MOVED:
7631         case -NFS4ERR_DELAY:
7632                 if (nfs4_async_handle_error(task, server,
7633                                             NULL, NULL) == -EAGAIN)
7634                         rpc_restart_call_prepare(task);
7635         }
7636 }
7637
7638 static void nfs4_release_lockowner_release(void *calldata)
7639 {
7640         struct nfs_release_lockowner_data *data = calldata;
7641         nfs4_free_lock_state(data->server, data->lsp);
7642         kfree(calldata);
7643 }
7644
7645 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7646         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7647         .rpc_call_done = nfs4_release_lockowner_done,
7648         .rpc_release = nfs4_release_lockowner_release,
7649 };
7650
7651 static void
7652 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7653 {
7654         struct nfs_release_lockowner_data *data;
7655         struct rpc_message msg = {
7656                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7657         };
7658
7659         if (server->nfs_client->cl_mvops->minor_version != 0)
7660                 return;
7661
7662         data = kmalloc(sizeof(*data), GFP_KERNEL);
7663         if (!data)
7664                 return;
7665         data->lsp = lsp;
7666         data->server = server;
7667         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7668         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7669         data->args.lock_owner.s_dev = server->s_dev;
7670
7671         msg.rpc_argp = &data->args;
7672         msg.rpc_resp = &data->res;
7673         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7674         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7675 }
7676
7677 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7678
7679 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7680                                    struct user_namespace *mnt_userns,
7681                                    struct dentry *unused, struct inode *inode,
7682                                    const char *key, const void *buf,
7683                                    size_t buflen, int flags)
7684 {
7685         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7686 }
7687
7688 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7689                                    struct dentry *unused, struct inode *inode,
7690                                    const char *key, void *buf, size_t buflen)
7691 {
7692         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7693 }
7694
7695 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7696 {
7697         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7698 }
7699
7700 #if defined(CONFIG_NFS_V4_1)
7701 #define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7702
7703 static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7704                                     struct user_namespace *mnt_userns,
7705                                     struct dentry *unused, struct inode *inode,
7706                                     const char *key, const void *buf,
7707                                     size_t buflen, int flags)
7708 {
7709         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7710 }
7711
7712 static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7713                                     struct dentry *unused, struct inode *inode,
7714                                     const char *key, void *buf, size_t buflen)
7715 {
7716         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7717 }
7718
7719 static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7720 {
7721         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7722 }
7723
7724 #define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7725
7726 static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7727                                     struct user_namespace *mnt_userns,
7728                                     struct dentry *unused, struct inode *inode,
7729                                     const char *key, const void *buf,
7730                                     size_t buflen, int flags)
7731 {
7732         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7733 }
7734
7735 static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7736                                     struct dentry *unused, struct inode *inode,
7737                                     const char *key, void *buf, size_t buflen)
7738 {
7739         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7740 }
7741
7742 static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7743 {
7744         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7745 }
7746
7747 #endif
7748
7749 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7750
7751 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7752                                      struct user_namespace *mnt_userns,
7753                                      struct dentry *unused, struct inode *inode,
7754                                      const char *key, const void *buf,
7755                                      size_t buflen, int flags)
7756 {
7757         if (security_ismaclabel(key))
7758                 return nfs4_set_security_label(inode, buf, buflen);
7759
7760         return -EOPNOTSUPP;
7761 }
7762
7763 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7764                                      struct dentry *unused, struct inode *inode,
7765                                      const char *key, void *buf, size_t buflen)
7766 {
7767         if (security_ismaclabel(key))
7768                 return nfs4_get_security_label(inode, buf, buflen);
7769         return -EOPNOTSUPP;
7770 }
7771
7772 static ssize_t
7773 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7774 {
7775         int len = 0;
7776
7777         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7778                 len = security_inode_listsecurity(inode, list, list_len);
7779                 if (len >= 0 && list_len && len > list_len)
7780                         return -ERANGE;
7781         }
7782         return len;
7783 }
7784
7785 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7786         .prefix = XATTR_SECURITY_PREFIX,
7787         .get    = nfs4_xattr_get_nfs4_label,
7788         .set    = nfs4_xattr_set_nfs4_label,
7789 };
7790
7791 #else
7792
7793 static ssize_t
7794 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7795 {
7796         return 0;
7797 }
7798
7799 #endif
7800
7801 #ifdef CONFIG_NFS_V4_2
7802 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7803                                     struct user_namespace *mnt_userns,
7804                                     struct dentry *unused, struct inode *inode,
7805                                     const char *key, const void *buf,
7806                                     size_t buflen, int flags)
7807 {
7808         u32 mask;
7809         int ret;
7810
7811         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7812                 return -EOPNOTSUPP;
7813
7814         /*
7815          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7816          * flags right now. Handling of xattr operations use the normal
7817          * file read/write permissions.
7818          *
7819          * Just in case the server has other ideas (which RFC 8276 allows),
7820          * do a cached access check for the XA* flags to possibly avoid
7821          * doing an RPC and getting EACCES back.
7822          */
7823         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7824                 if (!(mask & NFS_ACCESS_XAWRITE))
7825                         return -EACCES;
7826         }
7827
7828         if (buf == NULL) {
7829                 ret = nfs42_proc_removexattr(inode, key);
7830                 if (!ret)
7831                         nfs4_xattr_cache_remove(inode, key);
7832         } else {
7833                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7834                 if (!ret)
7835                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7836         }
7837
7838         return ret;
7839 }
7840
7841 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7842                                     struct dentry *unused, struct inode *inode,
7843                                     const char *key, void *buf, size_t buflen)
7844 {
7845         u32 mask;
7846         ssize_t ret;
7847
7848         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7849                 return -EOPNOTSUPP;
7850
7851         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7852                 if (!(mask & NFS_ACCESS_XAREAD))
7853                         return -EACCES;
7854         }
7855
7856         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7857         if (ret)
7858                 return ret;
7859
7860         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7861         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7862                 return ret;
7863
7864         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7865
7866         return ret;
7867 }
7868
7869 static ssize_t
7870 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7871 {
7872         u64 cookie;
7873         bool eof;
7874         ssize_t ret, size;
7875         char *buf;
7876         size_t buflen;
7877         u32 mask;
7878
7879         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7880                 return 0;
7881
7882         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7883                 if (!(mask & NFS_ACCESS_XALIST))
7884                         return 0;
7885         }
7886
7887         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7888         if (ret)
7889                 return ret;
7890
7891         ret = nfs4_xattr_cache_list(inode, list, list_len);
7892         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7893                 return ret;
7894
7895         cookie = 0;
7896         eof = false;
7897         buflen = list_len ? list_len : XATTR_LIST_MAX;
7898         buf = list_len ? list : NULL;
7899         size = 0;
7900
7901         while (!eof) {
7902                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7903                     &cookie, &eof);
7904                 if (ret < 0)
7905                         return ret;
7906
7907                 if (list_len) {
7908                         buf += ret;
7909                         buflen -= ret;
7910                 }
7911                 size += ret;
7912         }
7913
7914         if (list_len)
7915                 nfs4_xattr_cache_set_list(inode, list, size);
7916
7917         return size;
7918 }
7919
7920 #else
7921
7922 static ssize_t
7923 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7924 {
7925         return 0;
7926 }
7927 #endif /* CONFIG_NFS_V4_2 */
7928
7929 /*
7930  * nfs_fhget will use either the mounted_on_fileid or the fileid
7931  */
7932 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7933 {
7934         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7935                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7936               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7937               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7938                 return;
7939
7940         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7941                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7942         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7943         fattr->nlink = 2;
7944 }
7945
7946 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7947                                    const struct qstr *name,
7948                                    struct nfs4_fs_locations *fs_locations,
7949                                    struct page *page)
7950 {
7951         struct nfs_server *server = NFS_SERVER(dir);
7952         u32 bitmask[3];
7953         struct nfs4_fs_locations_arg args = {
7954                 .dir_fh = NFS_FH(dir),
7955                 .name = name,
7956                 .page = page,
7957                 .bitmask = bitmask,
7958         };
7959         struct nfs4_fs_locations_res res = {
7960                 .fs_locations = fs_locations,
7961         };
7962         struct rpc_message msg = {
7963                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7964                 .rpc_argp = &args,
7965                 .rpc_resp = &res,
7966         };
7967         int status;
7968
7969         dprintk("%s: start\n", __func__);
7970
7971         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7972         bitmask[1] = nfs4_fattr_bitmap[1];
7973
7974         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7975          * is not supported */
7976         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7977                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7978         else
7979                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7980
7981         nfs_fattr_init(fs_locations->fattr);
7982         fs_locations->server = server;
7983         fs_locations->nlocations = 0;
7984         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7985         dprintk("%s: returned status = %d\n", __func__, status);
7986         return status;
7987 }
7988
7989 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7990                            const struct qstr *name,
7991                            struct nfs4_fs_locations *fs_locations,
7992                            struct page *page)
7993 {
7994         struct nfs4_exception exception = {
7995                 .interruptible = true,
7996         };
7997         int err;
7998         do {
7999                 err = _nfs4_proc_fs_locations(client, dir, name,
8000                                 fs_locations, page);
8001                 trace_nfs4_get_fs_locations(dir, name, err);
8002                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8003                                 &exception);
8004         } while (exception.retry);
8005         return err;
8006 }
8007
8008 /*
8009  * This operation also signals the server that this client is
8010  * performing migration recovery.  The server can stop returning
8011  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
8012  * appended to this compound to identify the client ID which is
8013  * performing recovery.
8014  */
8015 static int _nfs40_proc_get_locations(struct nfs_server *server,
8016                                      struct nfs_fh *fhandle,
8017                                      struct nfs4_fs_locations *locations,
8018                                      struct page *page, const struct cred *cred)
8019 {
8020         struct rpc_clnt *clnt = server->client;
8021         u32 bitmask[2] = {
8022                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8023         };
8024         struct nfs4_fs_locations_arg args = {
8025                 .clientid       = server->nfs_client->cl_clientid,
8026                 .fh             = fhandle,
8027                 .page           = page,
8028                 .bitmask        = bitmask,
8029                 .migration      = 1,            /* skip LOOKUP */
8030                 .renew          = 1,            /* append RENEW */
8031         };
8032         struct nfs4_fs_locations_res res = {
8033                 .fs_locations   = locations,
8034                 .migration      = 1,
8035                 .renew          = 1,
8036         };
8037         struct rpc_message msg = {
8038                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8039                 .rpc_argp       = &args,
8040                 .rpc_resp       = &res,
8041                 .rpc_cred       = cred,
8042         };
8043         unsigned long now = jiffies;
8044         int status;
8045
8046         nfs_fattr_init(locations->fattr);
8047         locations->server = server;
8048         locations->nlocations = 0;
8049
8050         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8051         status = nfs4_call_sync_sequence(clnt, server, &msg,
8052                                         &args.seq_args, &res.seq_res);
8053         if (status)
8054                 return status;
8055
8056         renew_lease(server, now);
8057         return 0;
8058 }
8059
8060 #ifdef CONFIG_NFS_V4_1
8061
8062 /*
8063  * This operation also signals the server that this client is
8064  * performing migration recovery.  The server can stop asserting
8065  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8066  * performing this operation is identified in the SEQUENCE
8067  * operation in this compound.
8068  *
8069  * When the client supports GETATTR(fs_locations_info), it can
8070  * be plumbed in here.
8071  */
8072 static int _nfs41_proc_get_locations(struct nfs_server *server,
8073                                      struct nfs_fh *fhandle,
8074                                      struct nfs4_fs_locations *locations,
8075                                      struct page *page, const struct cred *cred)
8076 {
8077         struct rpc_clnt *clnt = server->client;
8078         u32 bitmask[2] = {
8079                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8080         };
8081         struct nfs4_fs_locations_arg args = {
8082                 .fh             = fhandle,
8083                 .page           = page,
8084                 .bitmask        = bitmask,
8085                 .migration      = 1,            /* skip LOOKUP */
8086         };
8087         struct nfs4_fs_locations_res res = {
8088                 .fs_locations   = locations,
8089                 .migration      = 1,
8090         };
8091         struct rpc_message msg = {
8092                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8093                 .rpc_argp       = &args,
8094                 .rpc_resp       = &res,
8095                 .rpc_cred       = cred,
8096         };
8097         struct nfs4_call_sync_data data = {
8098                 .seq_server = server,
8099                 .seq_args = &args.seq_args,
8100                 .seq_res = &res.seq_res,
8101         };
8102         struct rpc_task_setup task_setup_data = {
8103                 .rpc_client = clnt,
8104                 .rpc_message = &msg,
8105                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
8106                 .callback_data = &data,
8107                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8108         };
8109         int status;
8110
8111         nfs_fattr_init(locations->fattr);
8112         locations->server = server;
8113         locations->nlocations = 0;
8114
8115         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8116         status = nfs4_call_sync_custom(&task_setup_data);
8117         if (status == NFS4_OK &&
8118             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8119                 status = -NFS4ERR_LEASE_MOVED;
8120         return status;
8121 }
8122
8123 #endif  /* CONFIG_NFS_V4_1 */
8124
8125 /**
8126  * nfs4_proc_get_locations - discover locations for a migrated FSID
8127  * @server: pointer to nfs_server to process
8128  * @fhandle: pointer to the kernel NFS client file handle
8129  * @locations: result of query
8130  * @page: buffer
8131  * @cred: credential to use for this operation
8132  *
8133  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8134  * operation failed, or a negative errno if a local error occurred.
8135  *
8136  * On success, "locations" is filled in, but if the server has
8137  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8138  * asserted.
8139  *
8140  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8141  * from this client that require migration recovery.
8142  */
8143 int nfs4_proc_get_locations(struct nfs_server *server,
8144                             struct nfs_fh *fhandle,
8145                             struct nfs4_fs_locations *locations,
8146                             struct page *page, const struct cred *cred)
8147 {
8148         struct nfs_client *clp = server->nfs_client;
8149         const struct nfs4_mig_recovery_ops *ops =
8150                                         clp->cl_mvops->mig_recovery_ops;
8151         struct nfs4_exception exception = {
8152                 .interruptible = true,
8153         };
8154         int status;
8155
8156         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8157                 (unsigned long long)server->fsid.major,
8158                 (unsigned long long)server->fsid.minor,
8159                 clp->cl_hostname);
8160         nfs_display_fhandle(fhandle, __func__);
8161
8162         do {
8163                 status = ops->get_locations(server, fhandle, locations, page,
8164                                             cred);
8165                 if (status != -NFS4ERR_DELAY)
8166                         break;
8167                 nfs4_handle_exception(server, status, &exception);
8168         } while (exception.retry);
8169         return status;
8170 }
8171
8172 /*
8173  * This operation also signals the server that this client is
8174  * performing "lease moved" recovery.  The server can stop
8175  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8176  * is appended to this compound to identify the client ID which is
8177  * performing recovery.
8178  */
8179 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8180 {
8181         struct nfs_server *server = NFS_SERVER(inode);
8182         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8183         struct rpc_clnt *clnt = server->client;
8184         struct nfs4_fsid_present_arg args = {
8185                 .fh             = NFS_FH(inode),
8186                 .clientid       = clp->cl_clientid,
8187                 .renew          = 1,            /* append RENEW */
8188         };
8189         struct nfs4_fsid_present_res res = {
8190                 .renew          = 1,
8191         };
8192         struct rpc_message msg = {
8193                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8194                 .rpc_argp       = &args,
8195                 .rpc_resp       = &res,
8196                 .rpc_cred       = cred,
8197         };
8198         unsigned long now = jiffies;
8199         int status;
8200
8201         res.fh = nfs_alloc_fhandle();
8202         if (res.fh == NULL)
8203                 return -ENOMEM;
8204
8205         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8206         status = nfs4_call_sync_sequence(clnt, server, &msg,
8207                                                 &args.seq_args, &res.seq_res);
8208         nfs_free_fhandle(res.fh);
8209         if (status)
8210                 return status;
8211
8212         do_renew_lease(clp, now);
8213         return 0;
8214 }
8215
8216 #ifdef CONFIG_NFS_V4_1
8217
8218 /*
8219  * This operation also signals the server that this client is
8220  * performing "lease moved" recovery.  The server can stop asserting
8221  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8222  * this operation is identified in the SEQUENCE operation in this
8223  * compound.
8224  */
8225 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8226 {
8227         struct nfs_server *server = NFS_SERVER(inode);
8228         struct rpc_clnt *clnt = server->client;
8229         struct nfs4_fsid_present_arg args = {
8230                 .fh             = NFS_FH(inode),
8231         };
8232         struct nfs4_fsid_present_res res = {
8233         };
8234         struct rpc_message msg = {
8235                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8236                 .rpc_argp       = &args,
8237                 .rpc_resp       = &res,
8238                 .rpc_cred       = cred,
8239         };
8240         int status;
8241
8242         res.fh = nfs_alloc_fhandle();
8243         if (res.fh == NULL)
8244                 return -ENOMEM;
8245
8246         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8247         status = nfs4_call_sync_sequence(clnt, server, &msg,
8248                                                 &args.seq_args, &res.seq_res);
8249         nfs_free_fhandle(res.fh);
8250         if (status == NFS4_OK &&
8251             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8252                 status = -NFS4ERR_LEASE_MOVED;
8253         return status;
8254 }
8255
8256 #endif  /* CONFIG_NFS_V4_1 */
8257
8258 /**
8259  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8260  * @inode: inode on FSID to check
8261  * @cred: credential to use for this operation
8262  *
8263  * Server indicates whether the FSID is present, moved, or not
8264  * recognized.  This operation is necessary to clear a LEASE_MOVED
8265  * condition for this client ID.
8266  *
8267  * Returns NFS4_OK if the FSID is present on this server,
8268  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8269  *  NFS4ERR code if some error occurred on the server, or a
8270  *  negative errno if a local failure occurred.
8271  */
8272 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8273 {
8274         struct nfs_server *server = NFS_SERVER(inode);
8275         struct nfs_client *clp = server->nfs_client;
8276         const struct nfs4_mig_recovery_ops *ops =
8277                                         clp->cl_mvops->mig_recovery_ops;
8278         struct nfs4_exception exception = {
8279                 .interruptible = true,
8280         };
8281         int status;
8282
8283         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8284                 (unsigned long long)server->fsid.major,
8285                 (unsigned long long)server->fsid.minor,
8286                 clp->cl_hostname);
8287         nfs_display_fhandle(NFS_FH(inode), __func__);
8288
8289         do {
8290                 status = ops->fsid_present(inode, cred);
8291                 if (status != -NFS4ERR_DELAY)
8292                         break;
8293                 nfs4_handle_exception(server, status, &exception);
8294         } while (exception.retry);
8295         return status;
8296 }
8297
8298 /*
8299  * If 'use_integrity' is true and the state managment nfs_client
8300  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8301  * and the machine credential as per RFC3530bis and RFC5661 Security
8302  * Considerations sections. Otherwise, just use the user cred with the
8303  * filesystem's rpc_client.
8304  */
8305 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8306 {
8307         int status;
8308         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8309         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8310         struct nfs4_secinfo_arg args = {
8311                 .dir_fh = NFS_FH(dir),
8312                 .name   = name,
8313         };
8314         struct nfs4_secinfo_res res = {
8315                 .flavors     = flavors,
8316         };
8317         struct rpc_message msg = {
8318                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8319                 .rpc_argp = &args,
8320                 .rpc_resp = &res,
8321         };
8322         struct nfs4_call_sync_data data = {
8323                 .seq_server = NFS_SERVER(dir),
8324                 .seq_args = &args.seq_args,
8325                 .seq_res = &res.seq_res,
8326         };
8327         struct rpc_task_setup task_setup = {
8328                 .rpc_client = clnt,
8329                 .rpc_message = &msg,
8330                 .callback_ops = clp->cl_mvops->call_sync_ops,
8331                 .callback_data = &data,
8332                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8333         };
8334         const struct cred *cred = NULL;
8335
8336         if (use_integrity) {
8337                 clnt = clp->cl_rpcclient;
8338                 task_setup.rpc_client = clnt;
8339
8340                 cred = nfs4_get_clid_cred(clp);
8341                 msg.rpc_cred = cred;
8342         }
8343
8344         dprintk("NFS call  secinfo %s\n", name->name);
8345
8346         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8347         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8348         status = nfs4_call_sync_custom(&task_setup);
8349
8350         dprintk("NFS reply  secinfo: %d\n", status);
8351
8352         put_cred(cred);
8353         return status;
8354 }
8355
8356 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8357                       struct nfs4_secinfo_flavors *flavors)
8358 {
8359         struct nfs4_exception exception = {
8360                 .interruptible = true,
8361         };
8362         int err;
8363         do {
8364                 err = -NFS4ERR_WRONGSEC;
8365
8366                 /* try to use integrity protection with machine cred */
8367                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8368                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8369
8370                 /*
8371                  * if unable to use integrity protection, or SECINFO with
8372                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8373                  * disallowed by spec, but exists in deployed servers) use
8374                  * the current filesystem's rpc_client and the user cred.
8375                  */
8376                 if (err == -NFS4ERR_WRONGSEC)
8377                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8378
8379                 trace_nfs4_secinfo(dir, name, err);
8380                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8381                                 &exception);
8382         } while (exception.retry);
8383         return err;
8384 }
8385
8386 #ifdef CONFIG_NFS_V4_1
8387 /*
8388  * Check the exchange flags returned by the server for invalid flags, having
8389  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8390  * DS flags set.
8391  */
8392 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8393 {
8394         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8395                 goto out_inval;
8396         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8397                 goto out_inval;
8398         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8399             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8400                 goto out_inval;
8401         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8402                 goto out_inval;
8403         return NFS_OK;
8404 out_inval:
8405         return -NFS4ERR_INVAL;
8406 }
8407
8408 static bool
8409 nfs41_same_server_scope(struct nfs41_server_scope *a,
8410                         struct nfs41_server_scope *b)
8411 {
8412         if (a->server_scope_sz != b->server_scope_sz)
8413                 return false;
8414         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8415 }
8416
8417 static void
8418 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8419 {
8420         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8421         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8422         struct nfs_client *clp = args->client;
8423
8424         switch (task->tk_status) {
8425         case -NFS4ERR_BADSESSION:
8426         case -NFS4ERR_DEADSESSION:
8427                 nfs4_schedule_session_recovery(clp->cl_session,
8428                                 task->tk_status);
8429                 return;
8430         }
8431         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8432                         res->dir != NFS4_CDFS4_BOTH) {
8433                 rpc_task_close_connection(task);
8434                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8435                         rpc_restart_call(task);
8436         }
8437 }
8438
8439 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8440         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8441 };
8442
8443 /*
8444  * nfs4_proc_bind_one_conn_to_session()
8445  *
8446  * The 4.1 client currently uses the same TCP connection for the
8447  * fore and backchannel.
8448  */
8449 static
8450 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8451                 struct rpc_xprt *xprt,
8452                 struct nfs_client *clp,
8453                 const struct cred *cred)
8454 {
8455         int status;
8456         struct nfs41_bind_conn_to_session_args args = {
8457                 .client = clp,
8458                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8459                 .retries = 0,
8460         };
8461         struct nfs41_bind_conn_to_session_res res;
8462         struct rpc_message msg = {
8463                 .rpc_proc =
8464                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8465                 .rpc_argp = &args,
8466                 .rpc_resp = &res,
8467                 .rpc_cred = cred,
8468         };
8469         struct rpc_task_setup task_setup_data = {
8470                 .rpc_client = clnt,
8471                 .rpc_xprt = xprt,
8472                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8473                 .rpc_message = &msg,
8474                 .flags = RPC_TASK_TIMEOUT,
8475         };
8476         struct rpc_task *task;
8477
8478         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8479         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8480                 args.dir = NFS4_CDFC4_FORE;
8481
8482         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8483         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8484                 args.dir = NFS4_CDFC4_FORE;
8485
8486         task = rpc_run_task(&task_setup_data);
8487         if (!IS_ERR(task)) {
8488                 status = task->tk_status;
8489                 rpc_put_task(task);
8490         } else
8491                 status = PTR_ERR(task);
8492         trace_nfs4_bind_conn_to_session(clp, status);
8493         if (status == 0) {
8494                 if (memcmp(res.sessionid.data,
8495                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8496                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8497                         return -EIO;
8498                 }
8499                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8500                         dprintk("NFS: %s: Unexpected direction from server\n",
8501                                 __func__);
8502                         return -EIO;
8503                 }
8504                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8505                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8506                                 __func__);
8507                         return -EIO;
8508                 }
8509         }
8510
8511         return status;
8512 }
8513
8514 struct rpc_bind_conn_calldata {
8515         struct nfs_client *clp;
8516         const struct cred *cred;
8517 };
8518
8519 static int
8520 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8521                 struct rpc_xprt *xprt,
8522                 void *calldata)
8523 {
8524         struct rpc_bind_conn_calldata *p = calldata;
8525
8526         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8527 }
8528
8529 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8530 {
8531         struct rpc_bind_conn_calldata data = {
8532                 .clp = clp,
8533                 .cred = cred,
8534         };
8535         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8536                         nfs4_proc_bind_conn_to_session_callback, &data);
8537 }
8538
8539 /*
8540  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8541  * and operations we'd like to see to enable certain features in the allow map
8542  */
8543 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8544         .how = SP4_MACH_CRED,
8545         .enforce.u.words = {
8546                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8547                       1 << (OP_EXCHANGE_ID - 32) |
8548                       1 << (OP_CREATE_SESSION - 32) |
8549                       1 << (OP_DESTROY_SESSION - 32) |
8550                       1 << (OP_DESTROY_CLIENTID - 32)
8551         },
8552         .allow.u.words = {
8553                 [0] = 1 << (OP_CLOSE) |
8554                       1 << (OP_OPEN_DOWNGRADE) |
8555                       1 << (OP_LOCKU) |
8556                       1 << (OP_DELEGRETURN) |
8557                       1 << (OP_COMMIT),
8558                 [1] = 1 << (OP_SECINFO - 32) |
8559                       1 << (OP_SECINFO_NO_NAME - 32) |
8560                       1 << (OP_LAYOUTRETURN - 32) |
8561                       1 << (OP_TEST_STATEID - 32) |
8562                       1 << (OP_FREE_STATEID - 32) |
8563                       1 << (OP_WRITE - 32)
8564         }
8565 };
8566
8567 /*
8568  * Select the state protection mode for client `clp' given the server results
8569  * from exchange_id in `sp'.
8570  *
8571  * Returns 0 on success, negative errno otherwise.
8572  */
8573 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8574                                  struct nfs41_state_protection *sp)
8575 {
8576         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8577                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8578                       1 << (OP_EXCHANGE_ID - 32) |
8579                       1 << (OP_CREATE_SESSION - 32) |
8580                       1 << (OP_DESTROY_SESSION - 32) |
8581                       1 << (OP_DESTROY_CLIENTID - 32)
8582         };
8583         unsigned long flags = 0;
8584         unsigned int i;
8585         int ret = 0;
8586
8587         if (sp->how == SP4_MACH_CRED) {
8588                 /* Print state protect result */
8589                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8590                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8591                         if (test_bit(i, sp->enforce.u.longs))
8592                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8593                         if (test_bit(i, sp->allow.u.longs))
8594                                 dfprintk(MOUNT, "  allow op %d\n", i);
8595                 }
8596
8597                 /* make sure nothing is on enforce list that isn't supported */
8598                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8599                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8600                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8601                                 ret = -EINVAL;
8602                                 goto out;
8603                         }
8604                 }
8605
8606                 /*
8607                  * Minimal mode - state operations are allowed to use machine
8608                  * credential.  Note this already happens by default, so the
8609                  * client doesn't have to do anything more than the negotiation.
8610                  *
8611                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8612                  *       we're already using the machine cred for exchange_id
8613                  *       and will never use a different cred.
8614                  */
8615                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8616                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8617                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8618                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8619                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8620                         dfprintk(MOUNT, "  minimal mode enabled\n");
8621                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8622                 } else {
8623                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8624                         ret = -EINVAL;
8625                         goto out;
8626                 }
8627
8628                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8629                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8630                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8631                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8632                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8633                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8634                 }
8635
8636                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8637                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8638                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8639                 }
8640
8641                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8642                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8643                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8644                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8645                 }
8646
8647                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8648                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8649                         dfprintk(MOUNT, "  stateid mode enabled\n");
8650                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8651                 }
8652
8653                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8654                         dfprintk(MOUNT, "  write mode enabled\n");
8655                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8656                 }
8657
8658                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8659                         dfprintk(MOUNT, "  commit mode enabled\n");
8660                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8661                 }
8662         }
8663 out:
8664         clp->cl_sp4_flags = flags;
8665         return ret;
8666 }
8667
8668 struct nfs41_exchange_id_data {
8669         struct nfs41_exchange_id_res res;
8670         struct nfs41_exchange_id_args args;
8671 };
8672
8673 static void nfs4_exchange_id_release(void *data)
8674 {
8675         struct nfs41_exchange_id_data *cdata =
8676                                         (struct nfs41_exchange_id_data *)data;
8677
8678         nfs_put_client(cdata->args.client);
8679         kfree(cdata->res.impl_id);
8680         kfree(cdata->res.server_scope);
8681         kfree(cdata->res.server_owner);
8682         kfree(cdata);
8683 }
8684
8685 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8686         .rpc_release = nfs4_exchange_id_release,
8687 };
8688
8689 /*
8690  * _nfs4_proc_exchange_id()
8691  *
8692  * Wrapper for EXCHANGE_ID operation.
8693  */
8694 static struct rpc_task *
8695 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8696                         u32 sp4_how, struct rpc_xprt *xprt)
8697 {
8698         struct rpc_message msg = {
8699                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8700                 .rpc_cred = cred,
8701         };
8702         struct rpc_task_setup task_setup_data = {
8703                 .rpc_client = clp->cl_rpcclient,
8704                 .callback_ops = &nfs4_exchange_id_call_ops,
8705                 .rpc_message = &msg,
8706                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8707         };
8708         struct nfs41_exchange_id_data *calldata;
8709         int status;
8710
8711         if (!refcount_inc_not_zero(&clp->cl_count))
8712                 return ERR_PTR(-EIO);
8713
8714         status = -ENOMEM;
8715         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8716         if (!calldata)
8717                 goto out;
8718
8719         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8720
8721         status = nfs4_init_uniform_client_string(clp);
8722         if (status)
8723                 goto out_calldata;
8724
8725         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8726                                                 GFP_NOFS);
8727         status = -ENOMEM;
8728         if (unlikely(calldata->res.server_owner == NULL))
8729                 goto out_calldata;
8730
8731         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8732                                         GFP_NOFS);
8733         if (unlikely(calldata->res.server_scope == NULL))
8734                 goto out_server_owner;
8735
8736         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8737         if (unlikely(calldata->res.impl_id == NULL))
8738                 goto out_server_scope;
8739
8740         switch (sp4_how) {
8741         case SP4_NONE:
8742                 calldata->args.state_protect.how = SP4_NONE;
8743                 break;
8744
8745         case SP4_MACH_CRED:
8746                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8747                 break;
8748
8749         default:
8750                 /* unsupported! */
8751                 WARN_ON_ONCE(1);
8752                 status = -EINVAL;
8753                 goto out_impl_id;
8754         }
8755         if (xprt) {
8756                 task_setup_data.rpc_xprt = xprt;
8757                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8758                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8759                                 sizeof(calldata->args.verifier.data));
8760         }
8761         calldata->args.client = clp;
8762         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8763         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8764 #ifdef CONFIG_NFS_V4_1_MIGRATION
8765         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8766 #endif
8767         msg.rpc_argp = &calldata->args;
8768         msg.rpc_resp = &calldata->res;
8769         task_setup_data.callback_data = calldata;
8770
8771         return rpc_run_task(&task_setup_data);
8772
8773 out_impl_id:
8774         kfree(calldata->res.impl_id);
8775 out_server_scope:
8776         kfree(calldata->res.server_scope);
8777 out_server_owner:
8778         kfree(calldata->res.server_owner);
8779 out_calldata:
8780         kfree(calldata);
8781 out:
8782         nfs_put_client(clp);
8783         return ERR_PTR(status);
8784 }
8785
8786 /*
8787  * _nfs4_proc_exchange_id()
8788  *
8789  * Wrapper for EXCHANGE_ID operation.
8790  */
8791 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8792                         u32 sp4_how)
8793 {
8794         struct rpc_task *task;
8795         struct nfs41_exchange_id_args *argp;
8796         struct nfs41_exchange_id_res *resp;
8797         unsigned long now = jiffies;
8798         int status;
8799
8800         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8801         if (IS_ERR(task))
8802                 return PTR_ERR(task);
8803
8804         argp = task->tk_msg.rpc_argp;
8805         resp = task->tk_msg.rpc_resp;
8806         status = task->tk_status;
8807         if (status  != 0)
8808                 goto out;
8809
8810         status = nfs4_check_cl_exchange_flags(resp->flags,
8811                         clp->cl_mvops->minor_version);
8812         if (status  != 0)
8813                 goto out;
8814
8815         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8816         if (status != 0)
8817                 goto out;
8818
8819         do_renew_lease(clp, now);
8820
8821         clp->cl_clientid = resp->clientid;
8822         clp->cl_exchange_flags = resp->flags;
8823         clp->cl_seqid = resp->seqid;
8824         /* Client ID is not confirmed */
8825         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8826                 clear_bit(NFS4_SESSION_ESTABLISHED,
8827                           &clp->cl_session->session_state);
8828
8829         if (clp->cl_serverscope != NULL &&
8830             !nfs41_same_server_scope(clp->cl_serverscope,
8831                                 resp->server_scope)) {
8832                 dprintk("%s: server_scope mismatch detected\n",
8833                         __func__);
8834                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8835         }
8836
8837         swap(clp->cl_serverowner, resp->server_owner);
8838         swap(clp->cl_serverscope, resp->server_scope);
8839         swap(clp->cl_implid, resp->impl_id);
8840
8841         /* Save the EXCHANGE_ID verifier session trunk tests */
8842         memcpy(clp->cl_confirm.data, argp->verifier.data,
8843                sizeof(clp->cl_confirm.data));
8844 out:
8845         trace_nfs4_exchange_id(clp, status);
8846         rpc_put_task(task);
8847         return status;
8848 }
8849
8850 /*
8851  * nfs4_proc_exchange_id()
8852  *
8853  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8854  *
8855  * Since the clientid has expired, all compounds using sessions
8856  * associated with the stale clientid will be returning
8857  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8858  * be in some phase of session reset.
8859  *
8860  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8861  */
8862 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8863 {
8864         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8865         int status;
8866
8867         /* try SP4_MACH_CRED if krb5i/p */
8868         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8869             authflavor == RPC_AUTH_GSS_KRB5P) {
8870                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8871                 if (!status)
8872                         return 0;
8873         }
8874
8875         /* try SP4_NONE */
8876         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8877 }
8878
8879 /**
8880  * nfs4_test_session_trunk
8881  *
8882  * This is an add_xprt_test() test function called from
8883  * rpc_clnt_setup_test_and_add_xprt.
8884  *
8885  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8886  * and is dereferrenced in nfs4_exchange_id_release
8887  *
8888  * Upon success, add the new transport to the rpc_clnt
8889  *
8890  * @clnt: struct rpc_clnt to get new transport
8891  * @xprt: the rpc_xprt to test
8892  * @data: call data for _nfs4_proc_exchange_id.
8893  */
8894 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8895                             void *data)
8896 {
8897         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8898         struct rpc_task *task;
8899         int status;
8900
8901         u32 sp4_how;
8902
8903         dprintk("--> %s try %s\n", __func__,
8904                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8905
8906         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8907
8908         /* Test connection for session trunking. Async exchange_id call */
8909         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8910         if (IS_ERR(task))
8911                 return;
8912
8913         status = task->tk_status;
8914         if (status == 0)
8915                 status = nfs4_detect_session_trunking(adata->clp,
8916                                 task->tk_msg.rpc_resp, xprt);
8917
8918         if (status == 0)
8919                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8920
8921         rpc_put_task(task);
8922 }
8923 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8924
8925 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8926                 const struct cred *cred)
8927 {
8928         struct rpc_message msg = {
8929                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8930                 .rpc_argp = clp,
8931                 .rpc_cred = cred,
8932         };
8933         int status;
8934
8935         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8936                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8937         trace_nfs4_destroy_clientid(clp, status);
8938         if (status)
8939                 dprintk("NFS: Got error %d from the server %s on "
8940                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8941         return status;
8942 }
8943
8944 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8945                 const struct cred *cred)
8946 {
8947         unsigned int loop;
8948         int ret;
8949
8950         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8951                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8952                 switch (ret) {
8953                 case -NFS4ERR_DELAY:
8954                 case -NFS4ERR_CLIENTID_BUSY:
8955                         ssleep(1);
8956                         break;
8957                 default:
8958                         return ret;
8959                 }
8960         }
8961         return 0;
8962 }
8963
8964 int nfs4_destroy_clientid(struct nfs_client *clp)
8965 {
8966         const struct cred *cred;
8967         int ret = 0;
8968
8969         if (clp->cl_mvops->minor_version < 1)
8970                 goto out;
8971         if (clp->cl_exchange_flags == 0)
8972                 goto out;
8973         if (clp->cl_preserve_clid)
8974                 goto out;
8975         cred = nfs4_get_clid_cred(clp);
8976         ret = nfs4_proc_destroy_clientid(clp, cred);
8977         put_cred(cred);
8978         switch (ret) {
8979         case 0:
8980         case -NFS4ERR_STALE_CLIENTID:
8981                 clp->cl_exchange_flags = 0;
8982         }
8983 out:
8984         return ret;
8985 }
8986
8987 #endif /* CONFIG_NFS_V4_1 */
8988
8989 struct nfs4_get_lease_time_data {
8990         struct nfs4_get_lease_time_args *args;
8991         struct nfs4_get_lease_time_res *res;
8992         struct nfs_client *clp;
8993 };
8994
8995 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8996                                         void *calldata)
8997 {
8998         struct nfs4_get_lease_time_data *data =
8999                         (struct nfs4_get_lease_time_data *)calldata;
9000
9001         /* just setup sequence, do not trigger session recovery
9002            since we're invoked within one */
9003         nfs4_setup_sequence(data->clp,
9004                         &data->args->la_seq_args,
9005                         &data->res->lr_seq_res,
9006                         task);
9007 }
9008
9009 /*
9010  * Called from nfs4_state_manager thread for session setup, so don't recover
9011  * from sequence operation or clientid errors.
9012  */
9013 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9014 {
9015         struct nfs4_get_lease_time_data *data =
9016                         (struct nfs4_get_lease_time_data *)calldata;
9017
9018         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9019                 return;
9020         switch (task->tk_status) {
9021         case -NFS4ERR_DELAY:
9022         case -NFS4ERR_GRACE:
9023                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
9024                 task->tk_status = 0;
9025                 fallthrough;
9026         case -NFS4ERR_RETRY_UNCACHED_REP:
9027                 rpc_restart_call_prepare(task);
9028                 return;
9029         }
9030 }
9031
9032 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9033         .rpc_call_prepare = nfs4_get_lease_time_prepare,
9034         .rpc_call_done = nfs4_get_lease_time_done,
9035 };
9036
9037 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9038 {
9039         struct nfs4_get_lease_time_args args;
9040         struct nfs4_get_lease_time_res res = {
9041                 .lr_fsinfo = fsinfo,
9042         };
9043         struct nfs4_get_lease_time_data data = {
9044                 .args = &args,
9045                 .res = &res,
9046                 .clp = clp,
9047         };
9048         struct rpc_message msg = {
9049                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9050                 .rpc_argp = &args,
9051                 .rpc_resp = &res,
9052         };
9053         struct rpc_task_setup task_setup = {
9054                 .rpc_client = clp->cl_rpcclient,
9055                 .rpc_message = &msg,
9056                 .callback_ops = &nfs4_get_lease_time_ops,
9057                 .callback_data = &data,
9058                 .flags = RPC_TASK_TIMEOUT,
9059         };
9060
9061         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
9062         return nfs4_call_sync_custom(&task_setup);
9063 }
9064
9065 #ifdef CONFIG_NFS_V4_1
9066
9067 /*
9068  * Initialize the values to be used by the client in CREATE_SESSION
9069  * If nfs4_init_session set the fore channel request and response sizes,
9070  * use them.
9071  *
9072  * Set the back channel max_resp_sz_cached to zero to force the client to
9073  * always set csa_cachethis to FALSE because the current implementation
9074  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9075  */
9076 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9077                                     struct rpc_clnt *clnt)
9078 {
9079         unsigned int max_rqst_sz, max_resp_sz;
9080         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9081         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9082
9083         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9084         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9085
9086         /* Fore channel attributes */
9087         args->fc_attrs.max_rqst_sz = max_rqst_sz;
9088         args->fc_attrs.max_resp_sz = max_resp_sz;
9089         args->fc_attrs.max_ops = NFS4_MAX_OPS;
9090         args->fc_attrs.max_reqs = max_session_slots;
9091
9092         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9093                 "max_ops=%u max_reqs=%u\n",
9094                 __func__,
9095                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9096                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9097
9098         /* Back channel attributes */
9099         args->bc_attrs.max_rqst_sz = max_bc_payload;
9100         args->bc_attrs.max_resp_sz = max_bc_payload;
9101         args->bc_attrs.max_resp_sz_cached = 0;
9102         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9103         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9104         if (args->bc_attrs.max_reqs > max_bc_slots)
9105                 args->bc_attrs.max_reqs = max_bc_slots;
9106
9107         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9108                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9109                 __func__,
9110                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9111                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9112                 args->bc_attrs.max_reqs);
9113 }
9114
9115 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9116                 struct nfs41_create_session_res *res)
9117 {
9118         struct nfs4_channel_attrs *sent = &args->fc_attrs;
9119         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9120
9121         if (rcvd->max_resp_sz > sent->max_resp_sz)
9122                 return -EINVAL;
9123         /*
9124          * Our requested max_ops is the minimum we need; we're not
9125          * prepared to break up compounds into smaller pieces than that.
9126          * So, no point even trying to continue if the server won't
9127          * cooperate:
9128          */
9129         if (rcvd->max_ops < sent->max_ops)
9130                 return -EINVAL;
9131         if (rcvd->max_reqs == 0)
9132                 return -EINVAL;
9133         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9134                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9135         return 0;
9136 }
9137
9138 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9139                 struct nfs41_create_session_res *res)
9140 {
9141         struct nfs4_channel_attrs *sent = &args->bc_attrs;
9142         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9143
9144         if (!(res->flags & SESSION4_BACK_CHAN))
9145                 goto out;
9146         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9147                 return -EINVAL;
9148         if (rcvd->max_resp_sz < sent->max_resp_sz)
9149                 return -EINVAL;
9150         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9151                 return -EINVAL;
9152         if (rcvd->max_ops > sent->max_ops)
9153                 return -EINVAL;
9154         if (rcvd->max_reqs > sent->max_reqs)
9155                 return -EINVAL;
9156 out:
9157         return 0;
9158 }
9159
9160 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9161                                      struct nfs41_create_session_res *res)
9162 {
9163         int ret;
9164
9165         ret = nfs4_verify_fore_channel_attrs(args, res);
9166         if (ret)
9167                 return ret;
9168         return nfs4_verify_back_channel_attrs(args, res);
9169 }
9170
9171 static void nfs4_update_session(struct nfs4_session *session,
9172                 struct nfs41_create_session_res *res)
9173 {
9174         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9175         /* Mark client id and session as being confirmed */
9176         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9177         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9178         session->flags = res->flags;
9179         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9180         if (res->flags & SESSION4_BACK_CHAN)
9181                 memcpy(&session->bc_attrs, &res->bc_attrs,
9182                                 sizeof(session->bc_attrs));
9183 }
9184
9185 static int _nfs4_proc_create_session(struct nfs_client *clp,
9186                 const struct cred *cred)
9187 {
9188         struct nfs4_session *session = clp->cl_session;
9189         struct nfs41_create_session_args args = {
9190                 .client = clp,
9191                 .clientid = clp->cl_clientid,
9192                 .seqid = clp->cl_seqid,
9193                 .cb_program = NFS4_CALLBACK,
9194         };
9195         struct nfs41_create_session_res res;
9196
9197         struct rpc_message msg = {
9198                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9199                 .rpc_argp = &args,
9200                 .rpc_resp = &res,
9201                 .rpc_cred = cred,
9202         };
9203         int status;
9204
9205         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9206         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9207
9208         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9209                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9210         trace_nfs4_create_session(clp, status);
9211
9212         switch (status) {
9213         case -NFS4ERR_STALE_CLIENTID:
9214         case -NFS4ERR_DELAY:
9215         case -ETIMEDOUT:
9216         case -EACCES:
9217         case -EAGAIN:
9218                 goto out;
9219         }
9220
9221         clp->cl_seqid++;
9222         if (!status) {
9223                 /* Verify the session's negotiated channel_attrs values */
9224                 status = nfs4_verify_channel_attrs(&args, &res);
9225                 /* Increment the clientid slot sequence id */
9226                 if (status)
9227                         goto out;
9228                 nfs4_update_session(session, &res);
9229         }
9230 out:
9231         return status;
9232 }
9233
9234 /*
9235  * Issues a CREATE_SESSION operation to the server.
9236  * It is the responsibility of the caller to verify the session is
9237  * expired before calling this routine.
9238  */
9239 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9240 {
9241         int status;
9242         unsigned *ptr;
9243         struct nfs4_session *session = clp->cl_session;
9244
9245         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9246
9247         status = _nfs4_proc_create_session(clp, cred);
9248         if (status)
9249                 goto out;
9250
9251         /* Init or reset the session slot tables */
9252         status = nfs4_setup_session_slot_tables(session);
9253         dprintk("slot table setup returned %d\n", status);
9254         if (status)
9255                 goto out;
9256
9257         ptr = (unsigned *)&session->sess_id.data[0];
9258         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9259                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9260 out:
9261         return status;
9262 }
9263
9264 /*
9265  * Issue the over-the-wire RPC DESTROY_SESSION.
9266  * The caller must serialize access to this routine.
9267  */
9268 int nfs4_proc_destroy_session(struct nfs4_session *session,
9269                 const struct cred *cred)
9270 {
9271         struct rpc_message msg = {
9272                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9273                 .rpc_argp = session,
9274                 .rpc_cred = cred,
9275         };
9276         int status = 0;
9277
9278         /* session is still being setup */
9279         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9280                 return 0;
9281
9282         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9283                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9284         trace_nfs4_destroy_session(session->clp, status);
9285
9286         if (status)
9287                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9288                         "Session has been destroyed regardless...\n", status);
9289         return status;
9290 }
9291
9292 /*
9293  * Renew the cl_session lease.
9294  */
9295 struct nfs4_sequence_data {
9296         struct nfs_client *clp;
9297         struct nfs4_sequence_args args;
9298         struct nfs4_sequence_res res;
9299 };
9300
9301 static void nfs41_sequence_release(void *data)
9302 {
9303         struct nfs4_sequence_data *calldata = data;
9304         struct nfs_client *clp = calldata->clp;
9305
9306         if (refcount_read(&clp->cl_count) > 1)
9307                 nfs4_schedule_state_renewal(clp);
9308         nfs_put_client(clp);
9309         kfree(calldata);
9310 }
9311
9312 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9313 {
9314         switch(task->tk_status) {
9315         case -NFS4ERR_DELAY:
9316                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9317                 return -EAGAIN;
9318         default:
9319                 nfs4_schedule_lease_recovery(clp);
9320         }
9321         return 0;
9322 }
9323
9324 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9325 {
9326         struct nfs4_sequence_data *calldata = data;
9327         struct nfs_client *clp = calldata->clp;
9328
9329         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9330                 return;
9331
9332         trace_nfs4_sequence(clp, task->tk_status);
9333         if (task->tk_status < 0) {
9334                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9335                 if (refcount_read(&clp->cl_count) == 1)
9336                         return;
9337
9338                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9339                         rpc_restart_call_prepare(task);
9340                         return;
9341                 }
9342         }
9343         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9344 }
9345
9346 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9347 {
9348         struct nfs4_sequence_data *calldata = data;
9349         struct nfs_client *clp = calldata->clp;
9350         struct nfs4_sequence_args *args;
9351         struct nfs4_sequence_res *res;
9352
9353         args = task->tk_msg.rpc_argp;
9354         res = task->tk_msg.rpc_resp;
9355
9356         nfs4_setup_sequence(clp, args, res, task);
9357 }
9358
9359 static const struct rpc_call_ops nfs41_sequence_ops = {
9360         .rpc_call_done = nfs41_sequence_call_done,
9361         .rpc_call_prepare = nfs41_sequence_prepare,
9362         .rpc_release = nfs41_sequence_release,
9363 };
9364
9365 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9366                 const struct cred *cred,
9367                 struct nfs4_slot *slot,
9368                 bool is_privileged)
9369 {
9370         struct nfs4_sequence_data *calldata;
9371         struct rpc_message msg = {
9372                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9373                 .rpc_cred = cred,
9374         };
9375         struct rpc_task_setup task_setup_data = {
9376                 .rpc_client = clp->cl_rpcclient,
9377                 .rpc_message = &msg,
9378                 .callback_ops = &nfs41_sequence_ops,
9379                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9380         };
9381         struct rpc_task *ret;
9382
9383         ret = ERR_PTR(-EIO);
9384         if (!refcount_inc_not_zero(&clp->cl_count))
9385                 goto out_err;
9386
9387         ret = ERR_PTR(-ENOMEM);
9388         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
9389         if (calldata == NULL)
9390                 goto out_put_clp;
9391         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9392         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9393         msg.rpc_argp = &calldata->args;
9394         msg.rpc_resp = &calldata->res;
9395         calldata->clp = clp;
9396         task_setup_data.callback_data = calldata;
9397
9398         ret = rpc_run_task(&task_setup_data);
9399         if (IS_ERR(ret))
9400                 goto out_err;
9401         return ret;
9402 out_put_clp:
9403         nfs_put_client(clp);
9404 out_err:
9405         nfs41_release_slot(slot);
9406         return ret;
9407 }
9408
9409 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9410 {
9411         struct rpc_task *task;
9412         int ret = 0;
9413
9414         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9415                 return -EAGAIN;
9416         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9417         if (IS_ERR(task))
9418                 ret = PTR_ERR(task);
9419         else
9420                 rpc_put_task_async(task);
9421         dprintk("<-- %s status=%d\n", __func__, ret);
9422         return ret;
9423 }
9424
9425 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9426 {
9427         struct rpc_task *task;
9428         int ret;
9429
9430         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9431         if (IS_ERR(task)) {
9432                 ret = PTR_ERR(task);
9433                 goto out;
9434         }
9435         ret = rpc_wait_for_completion_task(task);
9436         if (!ret)
9437                 ret = task->tk_status;
9438         rpc_put_task(task);
9439 out:
9440         dprintk("<-- %s status=%d\n", __func__, ret);
9441         return ret;
9442 }
9443
9444 struct nfs4_reclaim_complete_data {
9445         struct nfs_client *clp;
9446         struct nfs41_reclaim_complete_args arg;
9447         struct nfs41_reclaim_complete_res res;
9448 };
9449
9450 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9451 {
9452         struct nfs4_reclaim_complete_data *calldata = data;
9453
9454         nfs4_setup_sequence(calldata->clp,
9455                         &calldata->arg.seq_args,
9456                         &calldata->res.seq_res,
9457                         task);
9458 }
9459
9460 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9461 {
9462         switch(task->tk_status) {
9463         case 0:
9464                 wake_up_all(&clp->cl_lock_waitq);
9465                 fallthrough;
9466         case -NFS4ERR_COMPLETE_ALREADY:
9467         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9468                 break;
9469         case -NFS4ERR_DELAY:
9470                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9471                 fallthrough;
9472         case -NFS4ERR_RETRY_UNCACHED_REP:
9473                 return -EAGAIN;
9474         case -NFS4ERR_BADSESSION:
9475         case -NFS4ERR_DEADSESSION:
9476         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9477                 break;
9478         default:
9479                 nfs4_schedule_lease_recovery(clp);
9480         }
9481         return 0;
9482 }
9483
9484 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9485 {
9486         struct nfs4_reclaim_complete_data *calldata = data;
9487         struct nfs_client *clp = calldata->clp;
9488         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9489
9490         if (!nfs41_sequence_done(task, res))
9491                 return;
9492
9493         trace_nfs4_reclaim_complete(clp, task->tk_status);
9494         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9495                 rpc_restart_call_prepare(task);
9496                 return;
9497         }
9498 }
9499
9500 static void nfs4_free_reclaim_complete_data(void *data)
9501 {
9502         struct nfs4_reclaim_complete_data *calldata = data;
9503
9504         kfree(calldata);
9505 }
9506
9507 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9508         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9509         .rpc_call_done = nfs4_reclaim_complete_done,
9510         .rpc_release = nfs4_free_reclaim_complete_data,
9511 };
9512
9513 /*
9514  * Issue a global reclaim complete.
9515  */
9516 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9517                 const struct cred *cred)
9518 {
9519         struct nfs4_reclaim_complete_data *calldata;
9520         struct rpc_message msg = {
9521                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9522                 .rpc_cred = cred,
9523         };
9524         struct rpc_task_setup task_setup_data = {
9525                 .rpc_client = clp->cl_rpcclient,
9526                 .rpc_message = &msg,
9527                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9528                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9529         };
9530         int status = -ENOMEM;
9531
9532         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9533         if (calldata == NULL)
9534                 goto out;
9535         calldata->clp = clp;
9536         calldata->arg.one_fs = 0;
9537
9538         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9539         msg.rpc_argp = &calldata->arg;
9540         msg.rpc_resp = &calldata->res;
9541         task_setup_data.callback_data = calldata;
9542         status = nfs4_call_sync_custom(&task_setup_data);
9543 out:
9544         dprintk("<-- %s status=%d\n", __func__, status);
9545         return status;
9546 }
9547
9548 static void
9549 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9550 {
9551         struct nfs4_layoutget *lgp = calldata;
9552         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9553
9554         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9555                                 &lgp->res.seq_res, task);
9556 }
9557
9558 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9559 {
9560         struct nfs4_layoutget *lgp = calldata;
9561
9562         nfs41_sequence_process(task, &lgp->res.seq_res);
9563 }
9564
9565 static int
9566 nfs4_layoutget_handle_exception(struct rpc_task *task,
9567                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9568 {
9569         struct inode *inode = lgp->args.inode;
9570         struct nfs_server *server = NFS_SERVER(inode);
9571         struct pnfs_layout_hdr *lo = lgp->lo;
9572         int nfs4err = task->tk_status;
9573         int err, status = 0;
9574         LIST_HEAD(head);
9575
9576         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9577
9578         nfs4_sequence_free_slot(&lgp->res.seq_res);
9579
9580         switch (nfs4err) {
9581         case 0:
9582                 goto out;
9583
9584         /*
9585          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9586          * on the file. set tk_status to -ENODATA to tell upper layer to
9587          * retry go inband.
9588          */
9589         case -NFS4ERR_LAYOUTUNAVAILABLE:
9590                 status = -ENODATA;
9591                 goto out;
9592         /*
9593          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9594          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9595          */
9596         case -NFS4ERR_BADLAYOUT:
9597                 status = -EOVERFLOW;
9598                 goto out;
9599         /*
9600          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9601          * (or clients) writing to the same RAID stripe except when
9602          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9603          *
9604          * Treat it like we would RECALLCONFLICT -- we retry for a little
9605          * while, and then eventually give up.
9606          */
9607         case -NFS4ERR_LAYOUTTRYLATER:
9608                 if (lgp->args.minlength == 0) {
9609                         status = -EOVERFLOW;
9610                         goto out;
9611                 }
9612                 status = -EBUSY;
9613                 break;
9614         case -NFS4ERR_RECALLCONFLICT:
9615                 status = -ERECALLCONFLICT;
9616                 break;
9617         case -NFS4ERR_DELEG_REVOKED:
9618         case -NFS4ERR_ADMIN_REVOKED:
9619         case -NFS4ERR_EXPIRED:
9620         case -NFS4ERR_BAD_STATEID:
9621                 exception->timeout = 0;
9622                 spin_lock(&inode->i_lock);
9623                 /* If the open stateid was bad, then recover it. */
9624                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9625                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9626                         spin_unlock(&inode->i_lock);
9627                         exception->state = lgp->args.ctx->state;
9628                         exception->stateid = &lgp->args.stateid;
9629                         break;
9630                 }
9631
9632                 /*
9633                  * Mark the bad layout state as invalid, then retry
9634                  */
9635                 pnfs_mark_layout_stateid_invalid(lo, &head);
9636                 spin_unlock(&inode->i_lock);
9637                 nfs_commit_inode(inode, 0);
9638                 pnfs_free_lseg_list(&head);
9639                 status = -EAGAIN;
9640                 goto out;
9641         }
9642
9643         err = nfs4_handle_exception(server, nfs4err, exception);
9644         if (!status) {
9645                 if (exception->retry)
9646                         status = -EAGAIN;
9647                 else
9648                         status = err;
9649         }
9650 out:
9651         return status;
9652 }
9653
9654 size_t max_response_pages(struct nfs_server *server)
9655 {
9656         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9657         return nfs_page_array_len(0, max_resp_sz);
9658 }
9659
9660 static void nfs4_layoutget_release(void *calldata)
9661 {
9662         struct nfs4_layoutget *lgp = calldata;
9663
9664         nfs4_sequence_free_slot(&lgp->res.seq_res);
9665         pnfs_layoutget_free(lgp);
9666 }
9667
9668 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9669         .rpc_call_prepare = nfs4_layoutget_prepare,
9670         .rpc_call_done = nfs4_layoutget_done,
9671         .rpc_release = nfs4_layoutget_release,
9672 };
9673
9674 struct pnfs_layout_segment *
9675 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9676 {
9677         struct inode *inode = lgp->args.inode;
9678         struct nfs_server *server = NFS_SERVER(inode);
9679         struct rpc_task *task;
9680         struct rpc_message msg = {
9681                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9682                 .rpc_argp = &lgp->args,
9683                 .rpc_resp = &lgp->res,
9684                 .rpc_cred = lgp->cred,
9685         };
9686         struct rpc_task_setup task_setup_data = {
9687                 .rpc_client = server->client,
9688                 .rpc_message = &msg,
9689                 .callback_ops = &nfs4_layoutget_call_ops,
9690                 .callback_data = lgp,
9691                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9692                          RPC_TASK_MOVEABLE,
9693         };
9694         struct pnfs_layout_segment *lseg = NULL;
9695         struct nfs4_exception exception = {
9696                 .inode = inode,
9697                 .timeout = *timeout,
9698         };
9699         int status = 0;
9700
9701         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9702
9703         task = rpc_run_task(&task_setup_data);
9704         if (IS_ERR(task))
9705                 return ERR_CAST(task);
9706
9707         status = rpc_wait_for_completion_task(task);
9708         if (status != 0)
9709                 goto out;
9710
9711         if (task->tk_status < 0) {
9712                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9713                 *timeout = exception.timeout;
9714         } else if (lgp->res.layoutp->len == 0) {
9715                 status = -EAGAIN;
9716                 *timeout = nfs4_update_delay(&exception.timeout);
9717         } else
9718                 lseg = pnfs_layout_process(lgp);
9719 out:
9720         trace_nfs4_layoutget(lgp->args.ctx,
9721                         &lgp->args.range,
9722                         &lgp->res.range,
9723                         &lgp->res.stateid,
9724                         status);
9725
9726         rpc_put_task(task);
9727         dprintk("<-- %s status=%d\n", __func__, status);
9728         if (status)
9729                 return ERR_PTR(status);
9730         return lseg;
9731 }
9732
9733 static void
9734 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9735 {
9736         struct nfs4_layoutreturn *lrp = calldata;
9737
9738         nfs4_setup_sequence(lrp->clp,
9739                         &lrp->args.seq_args,
9740                         &lrp->res.seq_res,
9741                         task);
9742         if (!pnfs_layout_is_valid(lrp->args.layout))
9743                 rpc_exit(task, 0);
9744 }
9745
9746 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9747 {
9748         struct nfs4_layoutreturn *lrp = calldata;
9749         struct nfs_server *server;
9750
9751         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9752                 return;
9753
9754         /*
9755          * Was there an RPC level error? Assume the call succeeded,
9756          * and that we need to release the layout
9757          */
9758         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9759                 lrp->res.lrs_present = 0;
9760                 return;
9761         }
9762
9763         server = NFS_SERVER(lrp->args.inode);
9764         switch (task->tk_status) {
9765         case -NFS4ERR_OLD_STATEID:
9766                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9767                                         &lrp->args.range,
9768                                         lrp->args.inode))
9769                         goto out_restart;
9770                 fallthrough;
9771         default:
9772                 task->tk_status = 0;
9773                 fallthrough;
9774         case 0:
9775                 break;
9776         case -NFS4ERR_DELAY:
9777                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9778                         break;
9779                 goto out_restart;
9780         }
9781         return;
9782 out_restart:
9783         task->tk_status = 0;
9784         nfs4_sequence_free_slot(&lrp->res.seq_res);
9785         rpc_restart_call_prepare(task);
9786 }
9787
9788 static void nfs4_layoutreturn_release(void *calldata)
9789 {
9790         struct nfs4_layoutreturn *lrp = calldata;
9791         struct pnfs_layout_hdr *lo = lrp->args.layout;
9792
9793         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9794                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9795         nfs4_sequence_free_slot(&lrp->res.seq_res);
9796         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9797                 lrp->ld_private.ops->free(&lrp->ld_private);
9798         pnfs_put_layout_hdr(lrp->args.layout);
9799         nfs_iput_and_deactive(lrp->inode);
9800         put_cred(lrp->cred);
9801         kfree(calldata);
9802 }
9803
9804 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9805         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9806         .rpc_call_done = nfs4_layoutreturn_done,
9807         .rpc_release = nfs4_layoutreturn_release,
9808 };
9809
9810 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9811 {
9812         struct rpc_task *task;
9813         struct rpc_message msg = {
9814                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9815                 .rpc_argp = &lrp->args,
9816                 .rpc_resp = &lrp->res,
9817                 .rpc_cred = lrp->cred,
9818         };
9819         struct rpc_task_setup task_setup_data = {
9820                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9821                 .rpc_message = &msg,
9822                 .callback_ops = &nfs4_layoutreturn_call_ops,
9823                 .callback_data = lrp,
9824                 .flags = RPC_TASK_MOVEABLE,
9825         };
9826         int status = 0;
9827
9828         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9829                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9830                         &task_setup_data.rpc_client, &msg);
9831
9832         lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9833         if (!sync) {
9834                 if (!lrp->inode) {
9835                         nfs4_layoutreturn_release(lrp);
9836                         return -EAGAIN;
9837                 }
9838                 task_setup_data.flags |= RPC_TASK_ASYNC;
9839         }
9840         if (!lrp->inode)
9841                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9842                                    1);
9843         else
9844                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9845                                    0);
9846         task = rpc_run_task(&task_setup_data);
9847         if (IS_ERR(task))
9848                 return PTR_ERR(task);
9849         if (sync)
9850                 status = task->tk_status;
9851         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9852         dprintk("<-- %s status=%d\n", __func__, status);
9853         rpc_put_task(task);
9854         return status;
9855 }
9856
9857 static int
9858 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9859                 struct pnfs_device *pdev,
9860                 const struct cred *cred)
9861 {
9862         struct nfs4_getdeviceinfo_args args = {
9863                 .pdev = pdev,
9864                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9865                         NOTIFY_DEVICEID4_DELETE,
9866         };
9867         struct nfs4_getdeviceinfo_res res = {
9868                 .pdev = pdev,
9869         };
9870         struct rpc_message msg = {
9871                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9872                 .rpc_argp = &args,
9873                 .rpc_resp = &res,
9874                 .rpc_cred = cred,
9875         };
9876         int status;
9877
9878         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9879         if (res.notification & ~args.notify_types)
9880                 dprintk("%s: unsupported notification\n", __func__);
9881         if (res.notification != args.notify_types)
9882                 pdev->nocache = 1;
9883
9884         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9885
9886         dprintk("<-- %s status=%d\n", __func__, status);
9887
9888         return status;
9889 }
9890
9891 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9892                 struct pnfs_device *pdev,
9893                 const struct cred *cred)
9894 {
9895         struct nfs4_exception exception = { };
9896         int err;
9897
9898         do {
9899                 err = nfs4_handle_exception(server,
9900                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9901                                         &exception);
9902         } while (exception.retry);
9903         return err;
9904 }
9905 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9906
9907 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9908 {
9909         struct nfs4_layoutcommit_data *data = calldata;
9910         struct nfs_server *server = NFS_SERVER(data->args.inode);
9911
9912         nfs4_setup_sequence(server->nfs_client,
9913                         &data->args.seq_args,
9914                         &data->res.seq_res,
9915                         task);
9916 }
9917
9918 static void
9919 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9920 {
9921         struct nfs4_layoutcommit_data *data = calldata;
9922         struct nfs_server *server = NFS_SERVER(data->args.inode);
9923
9924         if (!nfs41_sequence_done(task, &data->res.seq_res))
9925                 return;
9926
9927         switch (task->tk_status) { /* Just ignore these failures */
9928         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9929         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9930         case -NFS4ERR_BADLAYOUT:     /* no layout */
9931         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9932                 task->tk_status = 0;
9933                 break;
9934         case 0:
9935                 break;
9936         default:
9937                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9938                         rpc_restart_call_prepare(task);
9939                         return;
9940                 }
9941         }
9942 }
9943
9944 static void nfs4_layoutcommit_release(void *calldata)
9945 {
9946         struct nfs4_layoutcommit_data *data = calldata;
9947
9948         pnfs_cleanup_layoutcommit(data);
9949         nfs_post_op_update_inode_force_wcc(data->args.inode,
9950                                            data->res.fattr);
9951         put_cred(data->cred);
9952         nfs_iput_and_deactive(data->inode);
9953         kfree(data);
9954 }
9955
9956 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9957         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9958         .rpc_call_done = nfs4_layoutcommit_done,
9959         .rpc_release = nfs4_layoutcommit_release,
9960 };
9961
9962 int
9963 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9964 {
9965         struct rpc_message msg = {
9966                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9967                 .rpc_argp = &data->args,
9968                 .rpc_resp = &data->res,
9969                 .rpc_cred = data->cred,
9970         };
9971         struct rpc_task_setup task_setup_data = {
9972                 .task = &data->task,
9973                 .rpc_client = NFS_CLIENT(data->args.inode),
9974                 .rpc_message = &msg,
9975                 .callback_ops = &nfs4_layoutcommit_ops,
9976                 .callback_data = data,
9977                 .flags = RPC_TASK_MOVEABLE,
9978         };
9979         struct rpc_task *task;
9980         int status = 0;
9981
9982         dprintk("NFS: initiating layoutcommit call. sync %d "
9983                 "lbw: %llu inode %lu\n", sync,
9984                 data->args.lastbytewritten,
9985                 data->args.inode->i_ino);
9986
9987         if (!sync) {
9988                 data->inode = nfs_igrab_and_active(data->args.inode);
9989                 if (data->inode == NULL) {
9990                         nfs4_layoutcommit_release(data);
9991                         return -EAGAIN;
9992                 }
9993                 task_setup_data.flags = RPC_TASK_ASYNC;
9994         }
9995         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9996         task = rpc_run_task(&task_setup_data);
9997         if (IS_ERR(task))
9998                 return PTR_ERR(task);
9999         if (sync)
10000                 status = task->tk_status;
10001         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10002         dprintk("%s: status %d\n", __func__, status);
10003         rpc_put_task(task);
10004         return status;
10005 }
10006
10007 /*
10008  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10009  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10010  */
10011 static int
10012 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10013                     struct nfs_fsinfo *info,
10014                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
10015 {
10016         struct nfs41_secinfo_no_name_args args = {
10017                 .style = SECINFO_STYLE_CURRENT_FH,
10018         };
10019         struct nfs4_secinfo_res res = {
10020                 .flavors = flavors,
10021         };
10022         struct rpc_message msg = {
10023                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10024                 .rpc_argp = &args,
10025                 .rpc_resp = &res,
10026         };
10027         struct nfs4_call_sync_data data = {
10028                 .seq_server = server,
10029                 .seq_args = &args.seq_args,
10030                 .seq_res = &res.seq_res,
10031         };
10032         struct rpc_task_setup task_setup = {
10033                 .rpc_client = server->client,
10034                 .rpc_message = &msg,
10035                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
10036                 .callback_data = &data,
10037                 .flags = RPC_TASK_NO_ROUND_ROBIN,
10038         };
10039         const struct cred *cred = NULL;
10040         int status;
10041
10042         if (use_integrity) {
10043                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
10044
10045                 cred = nfs4_get_clid_cred(server->nfs_client);
10046                 msg.rpc_cred = cred;
10047         }
10048
10049         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
10050         status = nfs4_call_sync_custom(&task_setup);
10051         dprintk("<-- %s status=%d\n", __func__, status);
10052
10053         put_cred(cred);
10054
10055         return status;
10056 }
10057
10058 static int
10059 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10060                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
10061 {
10062         struct nfs4_exception exception = {
10063                 .interruptible = true,
10064         };
10065         int err;
10066         do {
10067                 /* first try using integrity protection */
10068                 err = -NFS4ERR_WRONGSEC;
10069
10070                 /* try to use integrity protection with machine cred */
10071                 if (_nfs4_is_integrity_protected(server->nfs_client))
10072                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10073                                                           flavors, true);
10074
10075                 /*
10076                  * if unable to use integrity protection, or SECINFO with
10077                  * integrity protection returns NFS4ERR_WRONGSEC (which is
10078                  * disallowed by spec, but exists in deployed servers) use
10079                  * the current filesystem's rpc_client and the user cred.
10080                  */
10081                 if (err == -NFS4ERR_WRONGSEC)
10082                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10083                                                           flavors, false);
10084
10085                 switch (err) {
10086                 case 0:
10087                 case -NFS4ERR_WRONGSEC:
10088                 case -ENOTSUPP:
10089                         goto out;
10090                 default:
10091                         err = nfs4_handle_exception(server, err, &exception);
10092                 }
10093         } while (exception.retry);
10094 out:
10095         return err;
10096 }
10097
10098 static int
10099 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10100                     struct nfs_fsinfo *info)
10101 {
10102         int err;
10103         struct page *page;
10104         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10105         struct nfs4_secinfo_flavors *flavors;
10106         struct nfs4_secinfo4 *secinfo;
10107         int i;
10108
10109         page = alloc_page(GFP_KERNEL);
10110         if (!page) {
10111                 err = -ENOMEM;
10112                 goto out;
10113         }
10114
10115         flavors = page_address(page);
10116         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10117
10118         /*
10119          * Fall back on "guess and check" method if
10120          * the server doesn't support SECINFO_NO_NAME
10121          */
10122         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10123                 err = nfs4_find_root_sec(server, fhandle, info);
10124                 goto out_freepage;
10125         }
10126         if (err)
10127                 goto out_freepage;
10128
10129         for (i = 0; i < flavors->num_flavors; i++) {
10130                 secinfo = &flavors->flavors[i];
10131
10132                 switch (secinfo->flavor) {
10133                 case RPC_AUTH_NULL:
10134                 case RPC_AUTH_UNIX:
10135                 case RPC_AUTH_GSS:
10136                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10137                                         &secinfo->flavor_info);
10138                         break;
10139                 default:
10140                         flavor = RPC_AUTH_MAXFLAVOR;
10141                         break;
10142                 }
10143
10144                 if (!nfs_auth_info_match(&server->auth_info, flavor))
10145                         flavor = RPC_AUTH_MAXFLAVOR;
10146
10147                 if (flavor != RPC_AUTH_MAXFLAVOR) {
10148                         err = nfs4_lookup_root_sec(server, fhandle,
10149                                                    info, flavor);
10150                         if (!err)
10151                                 break;
10152                 }
10153         }
10154
10155         if (flavor == RPC_AUTH_MAXFLAVOR)
10156                 err = -EPERM;
10157
10158 out_freepage:
10159         put_page(page);
10160         if (err == -EACCES)
10161                 return -EPERM;
10162 out:
10163         return err;
10164 }
10165
10166 static int _nfs41_test_stateid(struct nfs_server *server,
10167                 nfs4_stateid *stateid,
10168                 const struct cred *cred)
10169 {
10170         int status;
10171         struct nfs41_test_stateid_args args = {
10172                 .stateid = stateid,
10173         };
10174         struct nfs41_test_stateid_res res;
10175         struct rpc_message msg = {
10176                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10177                 .rpc_argp = &args,
10178                 .rpc_resp = &res,
10179                 .rpc_cred = cred,
10180         };
10181         struct rpc_clnt *rpc_client = server->client;
10182
10183         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10184                 &rpc_client, &msg);
10185
10186         dprintk("NFS call  test_stateid %p\n", stateid);
10187         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10188         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10189                         &args.seq_args, &res.seq_res);
10190         if (status != NFS_OK) {
10191                 dprintk("NFS reply test_stateid: failed, %d\n", status);
10192                 return status;
10193         }
10194         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10195         return -res.status;
10196 }
10197
10198 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10199                 int err, struct nfs4_exception *exception)
10200 {
10201         exception->retry = 0;
10202         switch(err) {
10203         case -NFS4ERR_DELAY:
10204         case -NFS4ERR_RETRY_UNCACHED_REP:
10205                 nfs4_handle_exception(server, err, exception);
10206                 break;
10207         case -NFS4ERR_BADSESSION:
10208         case -NFS4ERR_BADSLOT:
10209         case -NFS4ERR_BAD_HIGH_SLOT:
10210         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10211         case -NFS4ERR_DEADSESSION:
10212                 nfs4_do_handle_exception(server, err, exception);
10213         }
10214 }
10215
10216 /**
10217  * nfs41_test_stateid - perform a TEST_STATEID operation
10218  *
10219  * @server: server / transport on which to perform the operation
10220  * @stateid: state ID to test
10221  * @cred: credential
10222  *
10223  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10224  * Otherwise a negative NFS4ERR value is returned if the operation
10225  * failed or the state ID is not currently valid.
10226  */
10227 static int nfs41_test_stateid(struct nfs_server *server,
10228                 nfs4_stateid *stateid,
10229                 const struct cred *cred)
10230 {
10231         struct nfs4_exception exception = {
10232                 .interruptible = true,
10233         };
10234         int err;
10235         do {
10236                 err = _nfs41_test_stateid(server, stateid, cred);
10237                 nfs4_handle_delay_or_session_error(server, err, &exception);
10238         } while (exception.retry);
10239         return err;
10240 }
10241
10242 struct nfs_free_stateid_data {
10243         struct nfs_server *server;
10244         struct nfs41_free_stateid_args args;
10245         struct nfs41_free_stateid_res res;
10246 };
10247
10248 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10249 {
10250         struct nfs_free_stateid_data *data = calldata;
10251         nfs4_setup_sequence(data->server->nfs_client,
10252                         &data->args.seq_args,
10253                         &data->res.seq_res,
10254                         task);
10255 }
10256
10257 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10258 {
10259         struct nfs_free_stateid_data *data = calldata;
10260
10261         nfs41_sequence_done(task, &data->res.seq_res);
10262
10263         switch (task->tk_status) {
10264         case -NFS4ERR_DELAY:
10265                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10266                         rpc_restart_call_prepare(task);
10267         }
10268 }
10269
10270 static void nfs41_free_stateid_release(void *calldata)
10271 {
10272         struct nfs_free_stateid_data *data = calldata;
10273         struct nfs_client *clp = data->server->nfs_client;
10274
10275         nfs_put_client(clp);
10276         kfree(calldata);
10277 }
10278
10279 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10280         .rpc_call_prepare = nfs41_free_stateid_prepare,
10281         .rpc_call_done = nfs41_free_stateid_done,
10282         .rpc_release = nfs41_free_stateid_release,
10283 };
10284
10285 /**
10286  * nfs41_free_stateid - perform a FREE_STATEID operation
10287  *
10288  * @server: server / transport on which to perform the operation
10289  * @stateid: state ID to release
10290  * @cred: credential
10291  * @privileged: set to true if this call needs to be privileged
10292  *
10293  * Note: this function is always asynchronous.
10294  */
10295 static int nfs41_free_stateid(struct nfs_server *server,
10296                 const nfs4_stateid *stateid,
10297                 const struct cred *cred,
10298                 bool privileged)
10299 {
10300         struct rpc_message msg = {
10301                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10302                 .rpc_cred = cred,
10303         };
10304         struct rpc_task_setup task_setup = {
10305                 .rpc_client = server->client,
10306                 .rpc_message = &msg,
10307                 .callback_ops = &nfs41_free_stateid_ops,
10308                 .flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10309         };
10310         struct nfs_free_stateid_data *data;
10311         struct rpc_task *task;
10312         struct nfs_client *clp = server->nfs_client;
10313
10314         if (!refcount_inc_not_zero(&clp->cl_count))
10315                 return -EIO;
10316
10317         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10318                 &task_setup.rpc_client, &msg);
10319
10320         dprintk("NFS call  free_stateid %p\n", stateid);
10321         data = kmalloc(sizeof(*data), GFP_KERNEL);
10322         if (!data)
10323                 return -ENOMEM;
10324         data->server = server;
10325         nfs4_stateid_copy(&data->args.stateid, stateid);
10326
10327         task_setup.callback_data = data;
10328
10329         msg.rpc_argp = &data->args;
10330         msg.rpc_resp = &data->res;
10331         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10332         task = rpc_run_task(&task_setup);
10333         if (IS_ERR(task))
10334                 return PTR_ERR(task);
10335         rpc_put_task(task);
10336         return 0;
10337 }
10338
10339 static void
10340 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10341 {
10342         const struct cred *cred = lsp->ls_state->owner->so_cred;
10343
10344         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10345         nfs4_free_lock_state(server, lsp);
10346 }
10347
10348 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10349                 const nfs4_stateid *s2)
10350 {
10351         if (s1->type != s2->type)
10352                 return false;
10353
10354         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10355                 return false;
10356
10357         if (s1->seqid == s2->seqid)
10358                 return true;
10359
10360         return s1->seqid == 0 || s2->seqid == 0;
10361 }
10362
10363 #endif /* CONFIG_NFS_V4_1 */
10364
10365 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10366                 const nfs4_stateid *s2)
10367 {
10368         return nfs4_stateid_match(s1, s2);
10369 }
10370
10371
10372 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10373         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10374         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10375         .recover_open   = nfs4_open_reclaim,
10376         .recover_lock   = nfs4_lock_reclaim,
10377         .establish_clid = nfs4_init_clientid,
10378         .detect_trunking = nfs40_discover_server_trunking,
10379 };
10380
10381 #if defined(CONFIG_NFS_V4_1)
10382 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10383         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10384         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10385         .recover_open   = nfs4_open_reclaim,
10386         .recover_lock   = nfs4_lock_reclaim,
10387         .establish_clid = nfs41_init_clientid,
10388         .reclaim_complete = nfs41_proc_reclaim_complete,
10389         .detect_trunking = nfs41_discover_server_trunking,
10390 };
10391 #endif /* CONFIG_NFS_V4_1 */
10392
10393 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10394         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10395         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10396         .recover_open   = nfs40_open_expired,
10397         .recover_lock   = nfs4_lock_expired,
10398         .establish_clid = nfs4_init_clientid,
10399 };
10400
10401 #if defined(CONFIG_NFS_V4_1)
10402 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10403         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10404         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10405         .recover_open   = nfs41_open_expired,
10406         .recover_lock   = nfs41_lock_expired,
10407         .establish_clid = nfs41_init_clientid,
10408 };
10409 #endif /* CONFIG_NFS_V4_1 */
10410
10411 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10412         .sched_state_renewal = nfs4_proc_async_renew,
10413         .get_state_renewal_cred = nfs4_get_renew_cred,
10414         .renew_lease = nfs4_proc_renew,
10415 };
10416
10417 #if defined(CONFIG_NFS_V4_1)
10418 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10419         .sched_state_renewal = nfs41_proc_async_sequence,
10420         .get_state_renewal_cred = nfs4_get_machine_cred,
10421         .renew_lease = nfs4_proc_sequence,
10422 };
10423 #endif
10424
10425 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10426         .get_locations = _nfs40_proc_get_locations,
10427         .fsid_present = _nfs40_proc_fsid_present,
10428 };
10429
10430 #if defined(CONFIG_NFS_V4_1)
10431 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10432         .get_locations = _nfs41_proc_get_locations,
10433         .fsid_present = _nfs41_proc_fsid_present,
10434 };
10435 #endif  /* CONFIG_NFS_V4_1 */
10436
10437 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10438         .minor_version = 0,
10439         .init_caps = NFS_CAP_READDIRPLUS
10440                 | NFS_CAP_ATOMIC_OPEN
10441                 | NFS_CAP_POSIX_LOCK,
10442         .init_client = nfs40_init_client,
10443         .shutdown_client = nfs40_shutdown_client,
10444         .match_stateid = nfs4_match_stateid,
10445         .find_root_sec = nfs4_find_root_sec,
10446         .free_lock_state = nfs4_release_lockowner,
10447         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10448         .alloc_seqid = nfs_alloc_seqid,
10449         .call_sync_ops = &nfs40_call_sync_ops,
10450         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10451         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10452         .state_renewal_ops = &nfs40_state_renewal_ops,
10453         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10454 };
10455
10456 #if defined(CONFIG_NFS_V4_1)
10457 static struct nfs_seqid *
10458 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10459 {
10460         return NULL;
10461 }
10462
10463 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10464         .minor_version = 1,
10465         .init_caps = NFS_CAP_READDIRPLUS
10466                 | NFS_CAP_ATOMIC_OPEN
10467                 | NFS_CAP_POSIX_LOCK
10468                 | NFS_CAP_STATEID_NFSV41
10469                 | NFS_CAP_ATOMIC_OPEN_V1
10470                 | NFS_CAP_LGOPEN
10471                 | NFS_CAP_MOVEABLE,
10472         .init_client = nfs41_init_client,
10473         .shutdown_client = nfs41_shutdown_client,
10474         .match_stateid = nfs41_match_stateid,
10475         .find_root_sec = nfs41_find_root_sec,
10476         .free_lock_state = nfs41_free_lock_state,
10477         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10478         .alloc_seqid = nfs_alloc_no_seqid,
10479         .session_trunk = nfs4_test_session_trunk,
10480         .call_sync_ops = &nfs41_call_sync_ops,
10481         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10482         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10483         .state_renewal_ops = &nfs41_state_renewal_ops,
10484         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10485 };
10486 #endif
10487
10488 #if defined(CONFIG_NFS_V4_2)
10489 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10490         .minor_version = 2,
10491         .init_caps = NFS_CAP_READDIRPLUS
10492                 | NFS_CAP_ATOMIC_OPEN
10493                 | NFS_CAP_POSIX_LOCK
10494                 | NFS_CAP_STATEID_NFSV41
10495                 | NFS_CAP_ATOMIC_OPEN_V1
10496                 | NFS_CAP_LGOPEN
10497                 | NFS_CAP_ALLOCATE
10498                 | NFS_CAP_COPY
10499                 | NFS_CAP_OFFLOAD_CANCEL
10500                 | NFS_CAP_COPY_NOTIFY
10501                 | NFS_CAP_DEALLOCATE
10502                 | NFS_CAP_SEEK
10503                 | NFS_CAP_LAYOUTSTATS
10504                 | NFS_CAP_CLONE
10505                 | NFS_CAP_LAYOUTERROR
10506                 | NFS_CAP_READ_PLUS
10507                 | NFS_CAP_MOVEABLE,
10508         .init_client = nfs41_init_client,
10509         .shutdown_client = nfs41_shutdown_client,
10510         .match_stateid = nfs41_match_stateid,
10511         .find_root_sec = nfs41_find_root_sec,
10512         .free_lock_state = nfs41_free_lock_state,
10513         .call_sync_ops = &nfs41_call_sync_ops,
10514         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10515         .alloc_seqid = nfs_alloc_no_seqid,
10516         .session_trunk = nfs4_test_session_trunk,
10517         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10518         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10519         .state_renewal_ops = &nfs41_state_renewal_ops,
10520         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10521 };
10522 #endif
10523
10524 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10525         [0] = &nfs_v4_0_minor_ops,
10526 #if defined(CONFIG_NFS_V4_1)
10527         [1] = &nfs_v4_1_minor_ops,
10528 #endif
10529 #if defined(CONFIG_NFS_V4_2)
10530         [2] = &nfs_v4_2_minor_ops,
10531 #endif
10532 };
10533
10534 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10535 {
10536         ssize_t error, error2, error3;
10537
10538         error = generic_listxattr(dentry, list, size);
10539         if (error < 0)
10540                 return error;
10541         if (list) {
10542                 list += error;
10543                 size -= error;
10544         }
10545
10546         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10547         if (error2 < 0)
10548                 return error2;
10549
10550         if (list) {
10551                 list += error2;
10552                 size -= error2;
10553         }
10554
10555         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10556         if (error3 < 0)
10557                 return error3;
10558
10559         return error + error2 + error3;
10560 }
10561
10562 static void nfs4_enable_swap(struct inode *inode)
10563 {
10564         /* The state manager thread must always be running.
10565          * It will notice the client is a swapper, and stay put.
10566          */
10567         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10568
10569         nfs4_schedule_state_manager(clp);
10570 }
10571
10572 static void nfs4_disable_swap(struct inode *inode)
10573 {
10574         /* The state manager thread will now exit once it is
10575          * woken.
10576          */
10577         wake_up_var(&NFS_SERVER(inode)->nfs_client->cl_state);
10578 }
10579
10580 static const struct inode_operations nfs4_dir_inode_operations = {
10581         .create         = nfs_create,
10582         .lookup         = nfs_lookup,
10583         .atomic_open    = nfs_atomic_open,
10584         .link           = nfs_link,
10585         .unlink         = nfs_unlink,
10586         .symlink        = nfs_symlink,
10587         .mkdir          = nfs_mkdir,
10588         .rmdir          = nfs_rmdir,
10589         .mknod          = nfs_mknod,
10590         .rename         = nfs_rename,
10591         .permission     = nfs_permission,
10592         .getattr        = nfs_getattr,
10593         .setattr        = nfs_setattr,
10594         .listxattr      = nfs4_listxattr,
10595 };
10596
10597 static const struct inode_operations nfs4_file_inode_operations = {
10598         .permission     = nfs_permission,
10599         .getattr        = nfs_getattr,
10600         .setattr        = nfs_setattr,
10601         .listxattr      = nfs4_listxattr,
10602 };
10603
10604 const struct nfs_rpc_ops nfs_v4_clientops = {
10605         .version        = 4,                    /* protocol version */
10606         .dentry_ops     = &nfs4_dentry_operations,
10607         .dir_inode_ops  = &nfs4_dir_inode_operations,
10608         .file_inode_ops = &nfs4_file_inode_operations,
10609         .file_ops       = &nfs4_file_operations,
10610         .getroot        = nfs4_proc_get_root,
10611         .submount       = nfs4_submount,
10612         .try_get_tree   = nfs4_try_get_tree,
10613         .getattr        = nfs4_proc_getattr,
10614         .setattr        = nfs4_proc_setattr,
10615         .lookup         = nfs4_proc_lookup,
10616         .lookupp        = nfs4_proc_lookupp,
10617         .access         = nfs4_proc_access,
10618         .readlink       = nfs4_proc_readlink,
10619         .create         = nfs4_proc_create,
10620         .remove         = nfs4_proc_remove,
10621         .unlink_setup   = nfs4_proc_unlink_setup,
10622         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10623         .unlink_done    = nfs4_proc_unlink_done,
10624         .rename_setup   = nfs4_proc_rename_setup,
10625         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10626         .rename_done    = nfs4_proc_rename_done,
10627         .link           = nfs4_proc_link,
10628         .symlink        = nfs4_proc_symlink,
10629         .mkdir          = nfs4_proc_mkdir,
10630         .rmdir          = nfs4_proc_rmdir,
10631         .readdir        = nfs4_proc_readdir,
10632         .mknod          = nfs4_proc_mknod,
10633         .statfs         = nfs4_proc_statfs,
10634         .fsinfo         = nfs4_proc_fsinfo,
10635         .pathconf       = nfs4_proc_pathconf,
10636         .set_capabilities = nfs4_server_capabilities,
10637         .decode_dirent  = nfs4_decode_dirent,
10638         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10639         .read_setup     = nfs4_proc_read_setup,
10640         .read_done      = nfs4_read_done,
10641         .write_setup    = nfs4_proc_write_setup,
10642         .write_done     = nfs4_write_done,
10643         .commit_setup   = nfs4_proc_commit_setup,
10644         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10645         .commit_done    = nfs4_commit_done,
10646         .lock           = nfs4_proc_lock,
10647         .clear_acl_cache = nfs4_zap_acl_attr,
10648         .close_context  = nfs4_close_context,
10649         .open_context   = nfs4_atomic_open,
10650         .have_delegation = nfs4_have_delegation,
10651         .alloc_client   = nfs4_alloc_client,
10652         .init_client    = nfs4_init_client,
10653         .free_client    = nfs4_free_client,
10654         .create_server  = nfs4_create_server,
10655         .clone_server   = nfs_clone_server,
10656         .discover_trunking = nfs4_discover_trunking,
10657         .enable_swap    = nfs4_enable_swap,
10658         .disable_swap   = nfs4_disable_swap,
10659 };
10660
10661 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10662         .name   = XATTR_NAME_NFSV4_ACL,
10663         .list   = nfs4_xattr_list_nfs4_acl,
10664         .get    = nfs4_xattr_get_nfs4_acl,
10665         .set    = nfs4_xattr_set_nfs4_acl,
10666 };
10667
10668 #if defined(CONFIG_NFS_V4_1)
10669 static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10670         .name   = XATTR_NAME_NFSV4_DACL,
10671         .list   = nfs4_xattr_list_nfs4_dacl,
10672         .get    = nfs4_xattr_get_nfs4_dacl,
10673         .set    = nfs4_xattr_set_nfs4_dacl,
10674 };
10675
10676 static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10677         .name   = XATTR_NAME_NFSV4_SACL,
10678         .list   = nfs4_xattr_list_nfs4_sacl,
10679         .get    = nfs4_xattr_get_nfs4_sacl,
10680         .set    = nfs4_xattr_set_nfs4_sacl,
10681 };
10682 #endif
10683
10684 #ifdef CONFIG_NFS_V4_2
10685 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10686         .prefix = XATTR_USER_PREFIX,
10687         .get    = nfs4_xattr_get_nfs4_user,
10688         .set    = nfs4_xattr_set_nfs4_user,
10689 };
10690 #endif
10691
10692 const struct xattr_handler *nfs4_xattr_handlers[] = {
10693         &nfs4_xattr_nfs4_acl_handler,
10694 #if defined(CONFIG_NFS_V4_1)
10695         &nfs4_xattr_nfs4_dacl_handler,
10696         &nfs4_xattr_nfs4_sacl_handler,
10697 #endif
10698 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10699         &nfs4_xattr_nfs4_label_handler,
10700 #endif
10701 #ifdef CONFIG_NFS_V4_2
10702         &nfs4_xattr_nfs4_user_handler,
10703 #endif
10704         NULL
10705 };