Merge tag 'block-5.19-2022-07-21' of git://git.kernel.dk/linux-block
[linux-2.6-microblaze.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57 #include <linux/iversion.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "sysfs.h"
67 #include "nfs4idmap.h"
68 #include "nfs4session.h"
69 #include "fscache.h"
70 #include "nfs42.h"
71
72 #include "nfs4trace.h"
73
74 #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         if (!page)
4016                 return -ENOMEM;
4017         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4018         if (!locations)
4019                 goto out_free;
4020         locations->fattr = nfs_alloc_fattr();
4021         if (!locations->fattr)
4022                 goto out_free_2;
4023
4024         status = nfs4_proc_get_locations(server, fhandle, locations, page,
4025                                          cred);
4026         if (status)
4027                 goto out_free_3;
4028
4029         for (i = 0; i < locations->nlocations; i++)
4030                 test_fs_location_for_trunking(&locations->locations[i], clp,
4031                                               server);
4032 out_free_3:
4033         kfree(locations->fattr);
4034 out_free_2:
4035         kfree(locations);
4036 out_free:
4037         __free_page(page);
4038         return status;
4039 }
4040
4041 static int nfs4_discover_trunking(struct nfs_server *server,
4042                                   struct nfs_fh *fhandle)
4043 {
4044         struct nfs4_exception exception = {
4045                 .interruptible = true,
4046         };
4047         struct nfs_client *clp = server->nfs_client;
4048         int err = 0;
4049
4050         if (!nfs4_has_session(clp))
4051                 goto out;
4052         do {
4053                 err = nfs4_handle_exception(server,
4054                                 _nfs4_discover_trunking(server, fhandle),
4055                                 &exception);
4056         } while (exception.retry);
4057 out:
4058         return err;
4059 }
4060
4061 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4062                 struct nfs_fsinfo *info)
4063 {
4064         u32 bitmask[3];
4065         struct nfs4_lookup_root_arg args = {
4066                 .bitmask = bitmask,
4067         };
4068         struct nfs4_lookup_res res = {
4069                 .server = server,
4070                 .fattr = info->fattr,
4071                 .fh = fhandle,
4072         };
4073         struct rpc_message msg = {
4074                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4075                 .rpc_argp = &args,
4076                 .rpc_resp = &res,
4077         };
4078
4079         bitmask[0] = nfs4_fattr_bitmap[0];
4080         bitmask[1] = nfs4_fattr_bitmap[1];
4081         /*
4082          * Process the label in the upcoming getfattr
4083          */
4084         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4085
4086         nfs_fattr_init(info->fattr);
4087         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4088 }
4089
4090 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4091                 struct nfs_fsinfo *info)
4092 {
4093         struct nfs4_exception exception = {
4094                 .interruptible = true,
4095         };
4096         int err;
4097         do {
4098                 err = _nfs4_lookup_root(server, fhandle, info);
4099                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4100                 switch (err) {
4101                 case 0:
4102                 case -NFS4ERR_WRONGSEC:
4103                         goto out;
4104                 default:
4105                         err = nfs4_handle_exception(server, err, &exception);
4106                 }
4107         } while (exception.retry);
4108 out:
4109         return err;
4110 }
4111
4112 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4113                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4114 {
4115         struct rpc_auth_create_args auth_args = {
4116                 .pseudoflavor = flavor,
4117         };
4118         struct rpc_auth *auth;
4119
4120         auth = rpcauth_create(&auth_args, server->client);
4121         if (IS_ERR(auth))
4122                 return -EACCES;
4123         return nfs4_lookup_root(server, fhandle, info);
4124 }
4125
4126 /*
4127  * Retry pseudoroot lookup with various security flavors.  We do this when:
4128  *
4129  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4130  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4131  *
4132  * Returns zero on success, or a negative NFS4ERR value, or a
4133  * negative errno value.
4134  */
4135 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4136                               struct nfs_fsinfo *info)
4137 {
4138         /* Per 3530bis 15.33.5 */
4139         static const rpc_authflavor_t flav_array[] = {
4140                 RPC_AUTH_GSS_KRB5P,
4141                 RPC_AUTH_GSS_KRB5I,
4142                 RPC_AUTH_GSS_KRB5,
4143                 RPC_AUTH_UNIX,                  /* courtesy */
4144                 RPC_AUTH_NULL,
4145         };
4146         int status = -EPERM;
4147         size_t i;
4148
4149         if (server->auth_info.flavor_len > 0) {
4150                 /* try each flavor specified by user */
4151                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4152                         status = nfs4_lookup_root_sec(server, fhandle, info,
4153                                                 server->auth_info.flavors[i]);
4154                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4155                                 continue;
4156                         break;
4157                 }
4158         } else {
4159                 /* no flavors specified by user, try default list */
4160                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4161                         status = nfs4_lookup_root_sec(server, fhandle, info,
4162                                                       flav_array[i]);
4163                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4164                                 continue;
4165                         break;
4166                 }
4167         }
4168
4169         /*
4170          * -EACCES could mean that the user doesn't have correct permissions
4171          * to access the mount.  It could also mean that we tried to mount
4172          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4173          * existing mount programs don't handle -EACCES very well so it should
4174          * be mapped to -EPERM instead.
4175          */
4176         if (status == -EACCES)
4177                 status = -EPERM;
4178         return status;
4179 }
4180
4181 /**
4182  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4183  * @server: initialized nfs_server handle
4184  * @fhandle: we fill in the pseudo-fs root file handle
4185  * @info: we fill in an FSINFO struct
4186  * @auth_probe: probe the auth flavours
4187  *
4188  * Returns zero on success, or a negative errno.
4189  */
4190 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4191                          struct nfs_fsinfo *info,
4192                          bool auth_probe)
4193 {
4194         int status = 0;
4195
4196         if (!auth_probe)
4197                 status = nfs4_lookup_root(server, fhandle, info);
4198
4199         if (auth_probe || status == NFS4ERR_WRONGSEC)
4200                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4201                                 fhandle, info);
4202
4203         if (status == 0)
4204                 status = nfs4_server_capabilities(server, fhandle);
4205         if (status == 0)
4206                 status = nfs4_do_fsinfo(server, fhandle, info);
4207
4208         return nfs4_map_errors(status);
4209 }
4210
4211 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4212                               struct nfs_fsinfo *info)
4213 {
4214         int error;
4215         struct nfs_fattr *fattr = info->fattr;
4216
4217         error = nfs4_server_capabilities(server, mntfh);
4218         if (error < 0) {
4219                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4220                 return error;
4221         }
4222
4223         error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4224         if (error < 0) {
4225                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4226                 goto out;
4227         }
4228
4229         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4230             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4231                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4232
4233 out:
4234         return error;
4235 }
4236
4237 /*
4238  * Get locations and (maybe) other attributes of a referral.
4239  * Note that we'll actually follow the referral later when
4240  * we detect fsid mismatch in inode revalidation
4241  */
4242 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4243                              const struct qstr *name, struct nfs_fattr *fattr,
4244                              struct nfs_fh *fhandle)
4245 {
4246         int status = -ENOMEM;
4247         struct page *page = NULL;
4248         struct nfs4_fs_locations *locations = NULL;
4249
4250         page = alloc_page(GFP_KERNEL);
4251         if (page == NULL)
4252                 goto out;
4253         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4254         if (locations == NULL)
4255                 goto out;
4256
4257         locations->fattr = fattr;
4258
4259         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4260         if (status != 0)
4261                 goto out;
4262
4263         /*
4264          * If the fsid didn't change, this is a migration event, not a
4265          * referral.  Cause us to drop into the exception handler, which
4266          * will kick off migration recovery.
4267          */
4268         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4269                 dprintk("%s: server did not return a different fsid for"
4270                         " a referral at %s\n", __func__, name->name);
4271                 status = -NFS4ERR_MOVED;
4272                 goto out;
4273         }
4274         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4275         nfs_fixup_referral_attributes(fattr);
4276         memset(fhandle, 0, sizeof(struct nfs_fh));
4277 out:
4278         if (page)
4279                 __free_page(page);
4280         kfree(locations);
4281         return status;
4282 }
4283
4284 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4285                                 struct nfs_fattr *fattr, struct inode *inode)
4286 {
4287         __u32 bitmask[NFS4_BITMASK_SZ];
4288         struct nfs4_getattr_arg args = {
4289                 .fh = fhandle,
4290                 .bitmask = bitmask,
4291         };
4292         struct nfs4_getattr_res res = {
4293                 .fattr = fattr,
4294                 .server = server,
4295         };
4296         struct rpc_message msg = {
4297                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4298                 .rpc_argp = &args,
4299                 .rpc_resp = &res,
4300         };
4301         unsigned short task_flags = 0;
4302
4303         if (nfs4_has_session(server->nfs_client))
4304                 task_flags = RPC_TASK_MOVEABLE;
4305
4306         /* Is this is an attribute revalidation, subject to softreval? */
4307         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4308                 task_flags |= RPC_TASK_TIMEOUT;
4309
4310         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4311         nfs_fattr_init(fattr);
4312         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4313         return nfs4_do_call_sync(server->client, server, &msg,
4314                         &args.seq_args, &res.seq_res, task_flags);
4315 }
4316
4317 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4318                                 struct nfs_fattr *fattr, struct inode *inode)
4319 {
4320         struct nfs4_exception exception = {
4321                 .interruptible = true,
4322         };
4323         int err;
4324         do {
4325                 err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4326                 trace_nfs4_getattr(server, fhandle, fattr, err);
4327                 err = nfs4_handle_exception(server, err,
4328                                 &exception);
4329         } while (exception.retry);
4330         return err;
4331 }
4332
4333 /* 
4334  * The file is not closed if it is opened due to the a request to change
4335  * the size of the file. The open call will not be needed once the
4336  * VFS layer lookup-intents are implemented.
4337  *
4338  * Close is called when the inode is destroyed.
4339  * If we haven't opened the file for O_WRONLY, we
4340  * need to in the size_change case to obtain a stateid.
4341  *
4342  * Got race?
4343  * Because OPEN is always done by name in nfsv4, it is
4344  * possible that we opened a different file by the same
4345  * name.  We can recognize this race condition, but we
4346  * can't do anything about it besides returning an error.
4347  *
4348  * This will be fixed with VFS changes (lookup-intent).
4349  */
4350 static int
4351 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4352                   struct iattr *sattr)
4353 {
4354         struct inode *inode = d_inode(dentry);
4355         const struct cred *cred = NULL;
4356         struct nfs_open_context *ctx = NULL;
4357         int status;
4358
4359         if (pnfs_ld_layoutret_on_setattr(inode) &&
4360             sattr->ia_valid & ATTR_SIZE &&
4361             sattr->ia_size < i_size_read(inode))
4362                 pnfs_commit_and_return_layout(inode);
4363
4364         nfs_fattr_init(fattr);
4365         
4366         /* Deal with open(O_TRUNC) */
4367         if (sattr->ia_valid & ATTR_OPEN)
4368                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4369
4370         /* Optimization: if the end result is no change, don't RPC */
4371         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4372                 return 0;
4373
4374         /* Search for an existing open(O_WRITE) file */
4375         if (sattr->ia_valid & ATTR_FILE) {
4376
4377                 ctx = nfs_file_open_context(sattr->ia_file);
4378                 if (ctx)
4379                         cred = ctx->cred;
4380         }
4381
4382         /* Return any delegations if we're going to change ACLs */
4383         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4384                 nfs4_inode_make_writeable(inode);
4385
4386         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4387         if (status == 0) {
4388                 nfs_setattr_update_inode(inode, sattr, fattr);
4389                 nfs_setsecurity(inode, fattr);
4390         }
4391         return status;
4392 }
4393
4394 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4395                 struct dentry *dentry, struct nfs_fh *fhandle,
4396                 struct nfs_fattr *fattr)
4397 {
4398         struct nfs_server *server = NFS_SERVER(dir);
4399         int                    status;
4400         struct nfs4_lookup_arg args = {
4401                 .bitmask = server->attr_bitmask,
4402                 .dir_fh = NFS_FH(dir),
4403                 .name = &dentry->d_name,
4404         };
4405         struct nfs4_lookup_res res = {
4406                 .server = server,
4407                 .fattr = fattr,
4408                 .fh = fhandle,
4409         };
4410         struct rpc_message msg = {
4411                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4412                 .rpc_argp = &args,
4413                 .rpc_resp = &res,
4414         };
4415         unsigned short task_flags = 0;
4416
4417         if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4418                 task_flags = RPC_TASK_MOVEABLE;
4419
4420         /* Is this is an attribute revalidation, subject to softreval? */
4421         if (nfs_lookup_is_soft_revalidate(dentry))
4422                 task_flags |= RPC_TASK_TIMEOUT;
4423
4424         args.bitmask = nfs4_bitmask(server, fattr->label);
4425
4426         nfs_fattr_init(fattr);
4427
4428         dprintk("NFS call  lookup %pd2\n", dentry);
4429         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4430         status = nfs4_do_call_sync(clnt, server, &msg,
4431                         &args.seq_args, &res.seq_res, task_flags);
4432         dprintk("NFS reply lookup: %d\n", status);
4433         return status;
4434 }
4435
4436 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4437 {
4438         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4439                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4440         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4441         fattr->nlink = 2;
4442 }
4443
4444 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4445                                    struct dentry *dentry, struct nfs_fh *fhandle,
4446                                    struct nfs_fattr *fattr)
4447 {
4448         struct nfs4_exception exception = {
4449                 .interruptible = true,
4450         };
4451         struct rpc_clnt *client = *clnt;
4452         const struct qstr *name = &dentry->d_name;
4453         int err;
4454         do {
4455                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4456                 trace_nfs4_lookup(dir, name, err);
4457                 switch (err) {
4458                 case -NFS4ERR_BADNAME:
4459                         err = -ENOENT;
4460                         goto out;
4461                 case -NFS4ERR_MOVED:
4462                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4463                         if (err == -NFS4ERR_MOVED)
4464                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4465                         goto out;
4466                 case -NFS4ERR_WRONGSEC:
4467                         err = -EPERM;
4468                         if (client != *clnt)
4469                                 goto out;
4470                         client = nfs4_negotiate_security(client, dir, name);
4471                         if (IS_ERR(client))
4472                                 return PTR_ERR(client);
4473
4474                         exception.retry = 1;
4475                         break;
4476                 default:
4477                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4478                 }
4479         } while (exception.retry);
4480
4481 out:
4482         if (err == 0)
4483                 *clnt = client;
4484         else if (client != *clnt)
4485                 rpc_shutdown_client(client);
4486
4487         return err;
4488 }
4489
4490 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4491                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4492 {
4493         int status;
4494         struct rpc_clnt *client = NFS_CLIENT(dir);
4495
4496         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4497         if (client != NFS_CLIENT(dir)) {
4498                 rpc_shutdown_client(client);
4499                 nfs_fixup_secinfo_attributes(fattr);
4500         }
4501         return status;
4502 }
4503
4504 struct rpc_clnt *
4505 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4506                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4507 {
4508         struct rpc_clnt *client = NFS_CLIENT(dir);
4509         int status;
4510
4511         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4512         if (status < 0)
4513                 return ERR_PTR(status);
4514         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4515 }
4516
4517 static int _nfs4_proc_lookupp(struct inode *inode,
4518                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4519 {
4520         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4521         struct nfs_server *server = NFS_SERVER(inode);
4522         int                    status;
4523         struct nfs4_lookupp_arg args = {
4524                 .bitmask = server->attr_bitmask,
4525                 .fh = NFS_FH(inode),
4526         };
4527         struct nfs4_lookupp_res res = {
4528                 .server = server,
4529                 .fattr = fattr,
4530                 .fh = fhandle,
4531         };
4532         struct rpc_message msg = {
4533                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4534                 .rpc_argp = &args,
4535                 .rpc_resp = &res,
4536         };
4537         unsigned short task_flags = 0;
4538
4539         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4540                 task_flags |= RPC_TASK_TIMEOUT;
4541
4542         args.bitmask = nfs4_bitmask(server, fattr->label);
4543
4544         nfs_fattr_init(fattr);
4545
4546         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4547         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4548                                 &res.seq_res, task_flags);
4549         dprintk("NFS reply lookupp: %d\n", status);
4550         return status;
4551 }
4552
4553 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4554                              struct nfs_fattr *fattr)
4555 {
4556         struct nfs4_exception exception = {
4557                 .interruptible = true,
4558         };
4559         int err;
4560         do {
4561                 err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4562                 trace_nfs4_lookupp(inode, err);
4563                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4564                                 &exception);
4565         } while (exception.retry);
4566         return err;
4567 }
4568
4569 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4570                              const struct cred *cred)
4571 {
4572         struct nfs_server *server = NFS_SERVER(inode);
4573         struct nfs4_accessargs args = {
4574                 .fh = NFS_FH(inode),
4575                 .access = entry->mask,
4576         };
4577         struct nfs4_accessres res = {
4578                 .server = server,
4579         };
4580         struct rpc_message msg = {
4581                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4582                 .rpc_argp = &args,
4583                 .rpc_resp = &res,
4584                 .rpc_cred = cred,
4585         };
4586         int status = 0;
4587
4588         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4589                 res.fattr = nfs_alloc_fattr();
4590                 if (res.fattr == NULL)
4591                         return -ENOMEM;
4592                 args.bitmask = server->cache_consistency_bitmask;
4593         }
4594         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4595         if (!status) {
4596                 nfs_access_set_mask(entry, res.access);
4597                 if (res.fattr)
4598                         nfs_refresh_inode(inode, res.fattr);
4599         }
4600         nfs_free_fattr(res.fattr);
4601         return status;
4602 }
4603
4604 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4605                             const struct cred *cred)
4606 {
4607         struct nfs4_exception exception = {
4608                 .interruptible = true,
4609         };
4610         int err;
4611         do {
4612                 err = _nfs4_proc_access(inode, entry, cred);
4613                 trace_nfs4_access(inode, err);
4614                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4615                                 &exception);
4616         } while (exception.retry);
4617         return err;
4618 }
4619
4620 /*
4621  * TODO: For the time being, we don't try to get any attributes
4622  * along with any of the zero-copy operations READ, READDIR,
4623  * READLINK, WRITE.
4624  *
4625  * In the case of the first three, we want to put the GETATTR
4626  * after the read-type operation -- this is because it is hard
4627  * to predict the length of a GETATTR response in v4, and thus
4628  * align the READ data correctly.  This means that the GETATTR
4629  * may end up partially falling into the page cache, and we should
4630  * shift it into the 'tail' of the xdr_buf before processing.
4631  * To do this efficiently, we need to know the total length
4632  * of data received, which doesn't seem to be available outside
4633  * of the RPC layer.
4634  *
4635  * In the case of WRITE, we also want to put the GETATTR after
4636  * the operation -- in this case because we want to make sure
4637  * we get the post-operation mtime and size.
4638  *
4639  * Both of these changes to the XDR layer would in fact be quite
4640  * minor, but I decided to leave them for a subsequent patch.
4641  */
4642 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4643                 unsigned int pgbase, unsigned int pglen)
4644 {
4645         struct nfs4_readlink args = {
4646                 .fh       = NFS_FH(inode),
4647                 .pgbase   = pgbase,
4648                 .pglen    = pglen,
4649                 .pages    = &page,
4650         };
4651         struct nfs4_readlink_res res;
4652         struct rpc_message msg = {
4653                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4654                 .rpc_argp = &args,
4655                 .rpc_resp = &res,
4656         };
4657
4658         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4659 }
4660
4661 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4662                 unsigned int pgbase, unsigned int pglen)
4663 {
4664         struct nfs4_exception exception = {
4665                 .interruptible = true,
4666         };
4667         int err;
4668         do {
4669                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4670                 trace_nfs4_readlink(inode, err);
4671                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4672                                 &exception);
4673         } while (exception.retry);
4674         return err;
4675 }
4676
4677 /*
4678  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4679  */
4680 static int
4681 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4682                  int flags)
4683 {
4684         struct nfs_server *server = NFS_SERVER(dir);
4685         struct nfs4_label l, *ilabel = NULL;
4686         struct nfs_open_context *ctx;
4687         struct nfs4_state *state;
4688         int status = 0;
4689
4690         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4691         if (IS_ERR(ctx))
4692                 return PTR_ERR(ctx);
4693
4694         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4695
4696         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4697                 sattr->ia_mode &= ~current_umask();
4698         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4699         if (IS_ERR(state)) {
4700                 status = PTR_ERR(state);
4701                 goto out;
4702         }
4703 out:
4704         nfs4_label_release_security(ilabel);
4705         put_nfs_open_context(ctx);
4706         return status;
4707 }
4708
4709 static int
4710 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4711 {
4712         struct nfs_server *server = NFS_SERVER(dir);
4713         struct nfs_removeargs args = {
4714                 .fh = NFS_FH(dir),
4715                 .name = *name,
4716         };
4717         struct nfs_removeres res = {
4718                 .server = server,
4719         };
4720         struct rpc_message msg = {
4721                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4722                 .rpc_argp = &args,
4723                 .rpc_resp = &res,
4724         };
4725         unsigned long timestamp = jiffies;
4726         int status;
4727
4728         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4729         if (status == 0) {
4730                 spin_lock(&dir->i_lock);
4731                 /* Removing a directory decrements nlink in the parent */
4732                 if (ftype == NF4DIR && dir->i_nlink > 2)
4733                         nfs4_dec_nlink_locked(dir);
4734                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4735                                               NFS_INO_INVALID_DATA);
4736                 spin_unlock(&dir->i_lock);
4737         }
4738         return status;
4739 }
4740
4741 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4742 {
4743         struct nfs4_exception exception = {
4744                 .interruptible = true,
4745         };
4746         struct inode *inode = d_inode(dentry);
4747         int err;
4748
4749         if (inode) {
4750                 if (inode->i_nlink == 1)
4751                         nfs4_inode_return_delegation(inode);
4752                 else
4753                         nfs4_inode_make_writeable(inode);
4754         }
4755         do {
4756                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4757                 trace_nfs4_remove(dir, &dentry->d_name, err);
4758                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4759                                 &exception);
4760         } while (exception.retry);
4761         return err;
4762 }
4763
4764 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4765 {
4766         struct nfs4_exception exception = {
4767                 .interruptible = true,
4768         };
4769         int err;
4770
4771         do {
4772                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4773                 trace_nfs4_remove(dir, name, err);
4774                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4775                                 &exception);
4776         } while (exception.retry);
4777         return err;
4778 }
4779
4780 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4781                 struct dentry *dentry,
4782                 struct inode *inode)
4783 {
4784         struct nfs_removeargs *args = msg->rpc_argp;
4785         struct nfs_removeres *res = msg->rpc_resp;
4786
4787         res->server = NFS_SB(dentry->d_sb);
4788         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4789         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4790
4791         nfs_fattr_init(res->dir_attr);
4792
4793         if (inode) {
4794                 nfs4_inode_return_delegation(inode);
4795                 nfs_d_prune_case_insensitive_aliases(inode);
4796         }
4797 }
4798
4799 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4800 {
4801         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4802                         &data->args.seq_args,
4803                         &data->res.seq_res,
4804                         task);
4805 }
4806
4807 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4808 {
4809         struct nfs_unlinkdata *data = task->tk_calldata;
4810         struct nfs_removeres *res = &data->res;
4811
4812         if (!nfs4_sequence_done(task, &res->seq_res))
4813                 return 0;
4814         if (nfs4_async_handle_error(task, res->server, NULL,
4815                                     &data->timeout) == -EAGAIN)
4816                 return 0;
4817         if (task->tk_status == 0)
4818                 nfs4_update_changeattr(dir, &res->cinfo,
4819                                 res->dir_attr->time_start,
4820                                 NFS_INO_INVALID_DATA);
4821         return 1;
4822 }
4823
4824 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4825                 struct dentry *old_dentry,
4826                 struct dentry *new_dentry)
4827 {
4828         struct nfs_renameargs *arg = msg->rpc_argp;
4829         struct nfs_renameres *res = msg->rpc_resp;
4830         struct inode *old_inode = d_inode(old_dentry);
4831         struct inode *new_inode = d_inode(new_dentry);
4832
4833         if (old_inode)
4834                 nfs4_inode_make_writeable(old_inode);
4835         if (new_inode)
4836                 nfs4_inode_return_delegation(new_inode);
4837         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4838         res->server = NFS_SB(old_dentry->d_sb);
4839         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4840 }
4841
4842 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4843 {
4844         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4845                         &data->args.seq_args,
4846                         &data->res.seq_res,
4847                         task);
4848 }
4849
4850 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4851                                  struct inode *new_dir)
4852 {
4853         struct nfs_renamedata *data = task->tk_calldata;
4854         struct nfs_renameres *res = &data->res;
4855
4856         if (!nfs4_sequence_done(task, &res->seq_res))
4857                 return 0;
4858         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4859                 return 0;
4860
4861         if (task->tk_status == 0) {
4862                 nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4863                 if (new_dir != old_dir) {
4864                         /* Note: If we moved a directory, nlink will change */
4865                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4866                                         res->old_fattr->time_start,
4867                                         NFS_INO_INVALID_NLINK |
4868                                             NFS_INO_INVALID_DATA);
4869                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4870                                         res->new_fattr->time_start,
4871                                         NFS_INO_INVALID_NLINK |
4872                                             NFS_INO_INVALID_DATA);
4873                 } else
4874                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4875                                         res->old_fattr->time_start,
4876                                         NFS_INO_INVALID_DATA);
4877         }
4878         return 1;
4879 }
4880
4881 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4882 {
4883         struct nfs_server *server = NFS_SERVER(inode);
4884         __u32 bitmask[NFS4_BITMASK_SZ];
4885         struct nfs4_link_arg arg = {
4886                 .fh     = NFS_FH(inode),
4887                 .dir_fh = NFS_FH(dir),
4888                 .name   = name,
4889                 .bitmask = bitmask,
4890         };
4891         struct nfs4_link_res res = {
4892                 .server = server,
4893         };
4894         struct rpc_message msg = {
4895                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4896                 .rpc_argp = &arg,
4897                 .rpc_resp = &res,
4898         };
4899         int status = -ENOMEM;
4900
4901         res.fattr = nfs_alloc_fattr_with_label(server);
4902         if (res.fattr == NULL)
4903                 goto out;
4904
4905         nfs4_inode_make_writeable(inode);
4906         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4907                                 NFS_INO_INVALID_CHANGE);
4908         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4909         if (!status) {
4910                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4911                                        NFS_INO_INVALID_DATA);
4912                 nfs4_inc_nlink(inode);
4913                 status = nfs_post_op_update_inode(inode, res.fattr);
4914                 if (!status)
4915                         nfs_setsecurity(inode, res.fattr);
4916         }
4917
4918 out:
4919         nfs_free_fattr(res.fattr);
4920         return status;
4921 }
4922
4923 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4924 {
4925         struct nfs4_exception exception = {
4926                 .interruptible = true,
4927         };
4928         int err;
4929         do {
4930                 err = nfs4_handle_exception(NFS_SERVER(inode),
4931                                 _nfs4_proc_link(inode, dir, name),
4932                                 &exception);
4933         } while (exception.retry);
4934         return err;
4935 }
4936
4937 struct nfs4_createdata {
4938         struct rpc_message msg;
4939         struct nfs4_create_arg arg;
4940         struct nfs4_create_res res;
4941         struct nfs_fh fh;
4942         struct nfs_fattr fattr;
4943 };
4944
4945 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4946                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4947 {
4948         struct nfs4_createdata *data;
4949
4950         data = kzalloc(sizeof(*data), GFP_KERNEL);
4951         if (data != NULL) {
4952                 struct nfs_server *server = NFS_SERVER(dir);
4953
4954                 data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4955                 if (IS_ERR(data->fattr.label))
4956                         goto out_free;
4957
4958                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4959                 data->msg.rpc_argp = &data->arg;
4960                 data->msg.rpc_resp = &data->res;
4961                 data->arg.dir_fh = NFS_FH(dir);
4962                 data->arg.server = server;
4963                 data->arg.name = name;
4964                 data->arg.attrs = sattr;
4965                 data->arg.ftype = ftype;
4966                 data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4967                 data->arg.umask = current_umask();
4968                 data->res.server = server;
4969                 data->res.fh = &data->fh;
4970                 data->res.fattr = &data->fattr;
4971                 nfs_fattr_init(data->res.fattr);
4972         }
4973         return data;
4974 out_free:
4975         kfree(data);
4976         return NULL;
4977 }
4978
4979 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4980 {
4981         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4982                                     &data->arg.seq_args, &data->res.seq_res, 1);
4983         if (status == 0) {
4984                 spin_lock(&dir->i_lock);
4985                 /* Creating a directory bumps nlink in the parent */
4986                 if (data->arg.ftype == NF4DIR)
4987                         nfs4_inc_nlink_locked(dir);
4988                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4989                                               data->res.fattr->time_start,
4990                                               NFS_INO_INVALID_DATA);
4991                 spin_unlock(&dir->i_lock);
4992                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
4993         }
4994         return status;
4995 }
4996
4997 static void nfs4_free_createdata(struct nfs4_createdata *data)
4998 {
4999         nfs4_label_free(data->fattr.label);
5000         kfree(data);
5001 }
5002
5003 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5004                 struct page *page, unsigned int len, struct iattr *sattr,
5005                 struct nfs4_label *label)
5006 {
5007         struct nfs4_createdata *data;
5008         int status = -ENAMETOOLONG;
5009
5010         if (len > NFS4_MAXPATHLEN)
5011                 goto out;
5012
5013         status = -ENOMEM;
5014         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5015         if (data == NULL)
5016                 goto out;
5017
5018         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5019         data->arg.u.symlink.pages = &page;
5020         data->arg.u.symlink.len = len;
5021         data->arg.label = label;
5022         
5023         status = nfs4_do_create(dir, dentry, data);
5024
5025         nfs4_free_createdata(data);
5026 out:
5027         return status;
5028 }
5029
5030 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5031                 struct page *page, unsigned int len, struct iattr *sattr)
5032 {
5033         struct nfs4_exception exception = {
5034                 .interruptible = true,
5035         };
5036         struct nfs4_label l, *label = NULL;
5037         int err;
5038
5039         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5040
5041         do {
5042                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
5043                 trace_nfs4_symlink(dir, &dentry->d_name, err);
5044                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5045                                 &exception);
5046         } while (exception.retry);
5047
5048         nfs4_label_release_security(label);
5049         return err;
5050 }
5051
5052 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5053                 struct iattr *sattr, struct nfs4_label *label)
5054 {
5055         struct nfs4_createdata *data;
5056         int status = -ENOMEM;
5057
5058         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5059         if (data == NULL)
5060                 goto out;
5061
5062         data->arg.label = label;
5063         status = nfs4_do_create(dir, dentry, data);
5064
5065         nfs4_free_createdata(data);
5066 out:
5067         return status;
5068 }
5069
5070 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5071                 struct iattr *sattr)
5072 {
5073         struct nfs_server *server = NFS_SERVER(dir);
5074         struct nfs4_exception exception = {
5075                 .interruptible = true,
5076         };
5077         struct nfs4_label l, *label = NULL;
5078         int err;
5079
5080         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5081
5082         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5083                 sattr->ia_mode &= ~current_umask();
5084         do {
5085                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5086                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
5087                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5088                                 &exception);
5089         } while (exception.retry);
5090         nfs4_label_release_security(label);
5091
5092         return err;
5093 }
5094
5095 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5096                               struct nfs_readdir_res *nr_res)
5097 {
5098         struct inode            *dir = d_inode(nr_arg->dentry);
5099         struct nfs_server       *server = NFS_SERVER(dir);
5100         struct nfs4_readdir_arg args = {
5101                 .fh = NFS_FH(dir),
5102                 .pages = nr_arg->pages,
5103                 .pgbase = 0,
5104                 .count = nr_arg->page_len,
5105                 .plus = nr_arg->plus,
5106         };
5107         struct nfs4_readdir_res res;
5108         struct rpc_message msg = {
5109                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5110                 .rpc_argp = &args,
5111                 .rpc_resp = &res,
5112                 .rpc_cred = nr_arg->cred,
5113         };
5114         int                     status;
5115
5116         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5117                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5118         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5119                 args.bitmask = server->attr_bitmask_nl;
5120         else
5121                 args.bitmask = server->attr_bitmask;
5122
5123         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5124         res.pgbase = args.pgbase;
5125         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5126                         &res.seq_res, 0);
5127         if (status >= 0) {
5128                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5129                 status += args.pgbase;
5130         }
5131
5132         nfs_invalidate_atime(dir);
5133
5134         dprintk("%s: returns %d\n", __func__, status);
5135         return status;
5136 }
5137
5138 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5139                              struct nfs_readdir_res *res)
5140 {
5141         struct nfs4_exception exception = {
5142                 .interruptible = true,
5143         };
5144         int err;
5145         do {
5146                 err = _nfs4_proc_readdir(arg, res);
5147                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5148                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5149                                             err, &exception);
5150         } while (exception.retry);
5151         return err;
5152 }
5153
5154 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5155                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5156 {
5157         struct nfs4_createdata *data;
5158         int mode = sattr->ia_mode;
5159         int status = -ENOMEM;
5160
5161         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5162         if (data == NULL)
5163                 goto out;
5164
5165         if (S_ISFIFO(mode))
5166                 data->arg.ftype = NF4FIFO;
5167         else if (S_ISBLK(mode)) {
5168                 data->arg.ftype = NF4BLK;
5169                 data->arg.u.device.specdata1 = MAJOR(rdev);
5170                 data->arg.u.device.specdata2 = MINOR(rdev);
5171         }
5172         else if (S_ISCHR(mode)) {
5173                 data->arg.ftype = NF4CHR;
5174                 data->arg.u.device.specdata1 = MAJOR(rdev);
5175                 data->arg.u.device.specdata2 = MINOR(rdev);
5176         } else if (!S_ISSOCK(mode)) {
5177                 status = -EINVAL;
5178                 goto out_free;
5179         }
5180
5181         data->arg.label = label;
5182         status = nfs4_do_create(dir, dentry, data);
5183 out_free:
5184         nfs4_free_createdata(data);
5185 out:
5186         return status;
5187 }
5188
5189 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5190                 struct iattr *sattr, dev_t rdev)
5191 {
5192         struct nfs_server *server = NFS_SERVER(dir);
5193         struct nfs4_exception exception = {
5194                 .interruptible = true,
5195         };
5196         struct nfs4_label l, *label = NULL;
5197         int err;
5198
5199         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5200
5201         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5202                 sattr->ia_mode &= ~current_umask();
5203         do {
5204                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5205                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5206                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5207                                 &exception);
5208         } while (exception.retry);
5209
5210         nfs4_label_release_security(label);
5211
5212         return err;
5213 }
5214
5215 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5216                  struct nfs_fsstat *fsstat)
5217 {
5218         struct nfs4_statfs_arg args = {
5219                 .fh = fhandle,
5220                 .bitmask = server->attr_bitmask,
5221         };
5222         struct nfs4_statfs_res res = {
5223                 .fsstat = fsstat,
5224         };
5225         struct rpc_message msg = {
5226                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5227                 .rpc_argp = &args,
5228                 .rpc_resp = &res,
5229         };
5230
5231         nfs_fattr_init(fsstat->fattr);
5232         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5233 }
5234
5235 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5236 {
5237         struct nfs4_exception exception = {
5238                 .interruptible = true,
5239         };
5240         int err;
5241         do {
5242                 err = nfs4_handle_exception(server,
5243                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5244                                 &exception);
5245         } while (exception.retry);
5246         return err;
5247 }
5248
5249 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5250                 struct nfs_fsinfo *fsinfo)
5251 {
5252         struct nfs4_fsinfo_arg args = {
5253                 .fh = fhandle,
5254                 .bitmask = server->attr_bitmask,
5255         };
5256         struct nfs4_fsinfo_res res = {
5257                 .fsinfo = fsinfo,
5258         };
5259         struct rpc_message msg = {
5260                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5261                 .rpc_argp = &args,
5262                 .rpc_resp = &res,
5263         };
5264
5265         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5266 }
5267
5268 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5269 {
5270         struct nfs4_exception exception = {
5271                 .interruptible = true,
5272         };
5273         int err;
5274
5275         do {
5276                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5277                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5278                 if (err == 0) {
5279                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5280                         break;
5281                 }
5282                 err = nfs4_handle_exception(server, err, &exception);
5283         } while (exception.retry);
5284         return err;
5285 }
5286
5287 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5288 {
5289         int error;
5290
5291         nfs_fattr_init(fsinfo->fattr);
5292         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5293         if (error == 0) {
5294                 /* block layout checks this! */
5295                 server->pnfs_blksize = fsinfo->blksize;
5296                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5297         }
5298
5299         return error;
5300 }
5301
5302 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5303                 struct nfs_pathconf *pathconf)
5304 {
5305         struct nfs4_pathconf_arg args = {
5306                 .fh = fhandle,
5307                 .bitmask = server->attr_bitmask,
5308         };
5309         struct nfs4_pathconf_res res = {
5310                 .pathconf = pathconf,
5311         };
5312         struct rpc_message msg = {
5313                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5314                 .rpc_argp = &args,
5315                 .rpc_resp = &res,
5316         };
5317
5318         /* None of the pathconf attributes are mandatory to implement */
5319         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5320                 memset(pathconf, 0, sizeof(*pathconf));
5321                 return 0;
5322         }
5323
5324         nfs_fattr_init(pathconf->fattr);
5325         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5326 }
5327
5328 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5329                 struct nfs_pathconf *pathconf)
5330 {
5331         struct nfs4_exception exception = {
5332                 .interruptible = true,
5333         };
5334         int err;
5335
5336         do {
5337                 err = nfs4_handle_exception(server,
5338                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5339                                 &exception);
5340         } while (exception.retry);
5341         return err;
5342 }
5343
5344 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5345                 const struct nfs_open_context *ctx,
5346                 const struct nfs_lock_context *l_ctx,
5347                 fmode_t fmode)
5348 {
5349         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5350 }
5351 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5352
5353 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5354                 const struct nfs_open_context *ctx,
5355                 const struct nfs_lock_context *l_ctx,
5356                 fmode_t fmode)
5357 {
5358         nfs4_stateid _current_stateid;
5359
5360         /* If the current stateid represents a lost lock, then exit */
5361         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5362                 return true;
5363         return nfs4_stateid_match(stateid, &_current_stateid);
5364 }
5365
5366 static bool nfs4_error_stateid_expired(int err)
5367 {
5368         switch (err) {
5369         case -NFS4ERR_DELEG_REVOKED:
5370         case -NFS4ERR_ADMIN_REVOKED:
5371         case -NFS4ERR_BAD_STATEID:
5372         case -NFS4ERR_STALE_STATEID:
5373         case -NFS4ERR_OLD_STATEID:
5374         case -NFS4ERR_OPENMODE:
5375         case -NFS4ERR_EXPIRED:
5376                 return true;
5377         }
5378         return false;
5379 }
5380
5381 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5382 {
5383         struct nfs_server *server = NFS_SERVER(hdr->inode);
5384
5385         trace_nfs4_read(hdr, task->tk_status);
5386         if (task->tk_status < 0) {
5387                 struct nfs4_exception exception = {
5388                         .inode = hdr->inode,
5389                         .state = hdr->args.context->state,
5390                         .stateid = &hdr->args.stateid,
5391                 };
5392                 task->tk_status = nfs4_async_handle_exception(task,
5393                                 server, task->tk_status, &exception);
5394                 if (exception.retry) {
5395                         rpc_restart_call_prepare(task);
5396                         return -EAGAIN;
5397                 }
5398         }
5399
5400         if (task->tk_status > 0)
5401                 renew_lease(server, hdr->timestamp);
5402         return 0;
5403 }
5404
5405 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5406                 struct nfs_pgio_args *args)
5407 {
5408
5409         if (!nfs4_error_stateid_expired(task->tk_status) ||
5410                 nfs4_stateid_is_current(&args->stateid,
5411                                 args->context,
5412                                 args->lock_context,
5413                                 FMODE_READ))
5414                 return false;
5415         rpc_restart_call_prepare(task);
5416         return true;
5417 }
5418
5419 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5420                                          struct nfs_pgio_header *hdr)
5421 {
5422         struct nfs_server *server = NFS_SERVER(hdr->inode);
5423         struct rpc_message *msg = &task->tk_msg;
5424
5425         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5426             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5427                 server->caps &= ~NFS_CAP_READ_PLUS;
5428                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5429                 rpc_restart_call_prepare(task);
5430                 return true;
5431         }
5432         return false;
5433 }
5434
5435 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5436 {
5437         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5438                 return -EAGAIN;
5439         if (nfs4_read_stateid_changed(task, &hdr->args))
5440                 return -EAGAIN;
5441         if (nfs4_read_plus_not_supported(task, hdr))
5442                 return -EAGAIN;
5443         if (task->tk_status > 0)
5444                 nfs_invalidate_atime(hdr->inode);
5445         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5446                                     nfs4_read_done_cb(task, hdr);
5447 }
5448
5449 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5450 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5451                                     struct rpc_message *msg)
5452 {
5453         /* Note: We don't use READ_PLUS with pNFS yet */
5454         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5455                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5456 }
5457 #else
5458 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5459                                     struct rpc_message *msg)
5460 {
5461 }
5462 #endif /* CONFIG_NFS_V4_2 */
5463
5464 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5465                                  struct rpc_message *msg)
5466 {
5467         hdr->timestamp   = jiffies;
5468         if (!hdr->pgio_done_cb)
5469                 hdr->pgio_done_cb = nfs4_read_done_cb;
5470         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5471         nfs42_read_plus_support(hdr, msg);
5472         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5473 }
5474
5475 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5476                                       struct nfs_pgio_header *hdr)
5477 {
5478         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5479                         &hdr->args.seq_args,
5480                         &hdr->res.seq_res,
5481                         task))
5482                 return 0;
5483         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5484                                 hdr->args.lock_context,
5485                                 hdr->rw_mode) == -EIO)
5486                 return -EIO;
5487         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5488                 return -EIO;
5489         return 0;
5490 }
5491
5492 static int nfs4_write_done_cb(struct rpc_task *task,
5493                               struct nfs_pgio_header *hdr)
5494 {
5495         struct inode *inode = hdr->inode;
5496
5497         trace_nfs4_write(hdr, task->tk_status);
5498         if (task->tk_status < 0) {
5499                 struct nfs4_exception exception = {
5500                         .inode = hdr->inode,
5501                         .state = hdr->args.context->state,
5502                         .stateid = &hdr->args.stateid,
5503                 };
5504                 task->tk_status = nfs4_async_handle_exception(task,
5505                                 NFS_SERVER(inode), task->tk_status,
5506                                 &exception);
5507                 if (exception.retry) {
5508                         rpc_restart_call_prepare(task);
5509                         return -EAGAIN;
5510                 }
5511         }
5512         if (task->tk_status >= 0) {
5513                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5514                 nfs_writeback_update_inode(hdr);
5515         }
5516         return 0;
5517 }
5518
5519 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5520                 struct nfs_pgio_args *args)
5521 {
5522
5523         if (!nfs4_error_stateid_expired(task->tk_status) ||
5524                 nfs4_stateid_is_current(&args->stateid,
5525                                 args->context,
5526                                 args->lock_context,
5527                                 FMODE_WRITE))
5528                 return false;
5529         rpc_restart_call_prepare(task);
5530         return true;
5531 }
5532
5533 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5534 {
5535         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5536                 return -EAGAIN;
5537         if (nfs4_write_stateid_changed(task, &hdr->args))
5538                 return -EAGAIN;
5539         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5540                 nfs4_write_done_cb(task, hdr);
5541 }
5542
5543 static
5544 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5545 {
5546         /* Don't request attributes for pNFS or O_DIRECT writes */
5547         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5548                 return false;
5549         /* Otherwise, request attributes if and only if we don't hold
5550          * a delegation
5551          */
5552         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5553 }
5554
5555 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5556                       struct inode *inode, unsigned long cache_validity)
5557 {
5558         struct nfs_server *server = NFS_SERVER(inode);
5559         unsigned int i;
5560
5561         memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5562         cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5563
5564         if (cache_validity & NFS_INO_INVALID_CHANGE)
5565                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5566         if (cache_validity & NFS_INO_INVALID_ATIME)
5567                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5568         if (cache_validity & NFS_INO_INVALID_MODE)
5569                 bitmask[1] |= FATTR4_WORD1_MODE;
5570         if (cache_validity & NFS_INO_INVALID_OTHER)
5571                 bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5572         if (cache_validity & NFS_INO_INVALID_NLINK)
5573                 bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5574         if (cache_validity & NFS_INO_INVALID_CTIME)
5575                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5576         if (cache_validity & NFS_INO_INVALID_MTIME)
5577                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5578         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5579                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5580
5581         if (cache_validity & NFS_INO_INVALID_SIZE)
5582                 bitmask[0] |= FATTR4_WORD0_SIZE;
5583
5584         for (i = 0; i < NFS4_BITMASK_SZ; i++)
5585                 bitmask[i] &= server->attr_bitmask[i];
5586 }
5587
5588 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5589                                   struct rpc_message *msg,
5590                                   struct rpc_clnt **clnt)
5591 {
5592         struct nfs_server *server = NFS_SERVER(hdr->inode);
5593
5594         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5595                 hdr->args.bitmask = NULL;
5596                 hdr->res.fattr = NULL;
5597         } else {
5598                 nfs4_bitmask_set(hdr->args.bitmask_store,
5599                                  server->cache_consistency_bitmask,
5600                                  hdr->inode, NFS_INO_INVALID_BLOCKS);
5601                 hdr->args.bitmask = hdr->args.bitmask_store;
5602         }
5603
5604         if (!hdr->pgio_done_cb)
5605                 hdr->pgio_done_cb = nfs4_write_done_cb;
5606         hdr->res.server = server;
5607         hdr->timestamp   = jiffies;
5608
5609         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5610         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5611         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5612 }
5613
5614 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5615 {
5616         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5617                         &data->args.seq_args,
5618                         &data->res.seq_res,
5619                         task);
5620 }
5621
5622 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5623 {
5624         struct inode *inode = data->inode;
5625
5626         trace_nfs4_commit(data, task->tk_status);
5627         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5628                                     NULL, NULL) == -EAGAIN) {
5629                 rpc_restart_call_prepare(task);
5630                 return -EAGAIN;
5631         }
5632         return 0;
5633 }
5634
5635 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5636 {
5637         if (!nfs4_sequence_done(task, &data->res.seq_res))
5638                 return -EAGAIN;
5639         return data->commit_done_cb(task, data);
5640 }
5641
5642 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5643                                    struct rpc_clnt **clnt)
5644 {
5645         struct nfs_server *server = NFS_SERVER(data->inode);
5646
5647         if (data->commit_done_cb == NULL)
5648                 data->commit_done_cb = nfs4_commit_done_cb;
5649         data->res.server = server;
5650         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5651         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5652         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5653 }
5654
5655 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5656                                 struct nfs_commitres *res)
5657 {
5658         struct inode *dst_inode = file_inode(dst);
5659         struct nfs_server *server = NFS_SERVER(dst_inode);
5660         struct rpc_message msg = {
5661                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5662                 .rpc_argp = args,
5663                 .rpc_resp = res,
5664         };
5665
5666         args->fh = NFS_FH(dst_inode);
5667         return nfs4_call_sync(server->client, server, &msg,
5668                         &args->seq_args, &res->seq_res, 1);
5669 }
5670
5671 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5672 {
5673         struct nfs_commitargs args = {
5674                 .offset = offset,
5675                 .count = count,
5676         };
5677         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5678         struct nfs4_exception exception = { };
5679         int status;
5680
5681         do {
5682                 status = _nfs4_proc_commit(dst, &args, res);
5683                 status = nfs4_handle_exception(dst_server, status, &exception);
5684         } while (exception.retry);
5685
5686         return status;
5687 }
5688
5689 struct nfs4_renewdata {
5690         struct nfs_client       *client;
5691         unsigned long           timestamp;
5692 };
5693
5694 /*
5695  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5696  * standalone procedure for queueing an asynchronous RENEW.
5697  */
5698 static void nfs4_renew_release(void *calldata)
5699 {
5700         struct nfs4_renewdata *data = calldata;
5701         struct nfs_client *clp = data->client;
5702
5703         if (refcount_read(&clp->cl_count) > 1)
5704                 nfs4_schedule_state_renewal(clp);
5705         nfs_put_client(clp);
5706         kfree(data);
5707 }
5708
5709 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5710 {
5711         struct nfs4_renewdata *data = calldata;
5712         struct nfs_client *clp = data->client;
5713         unsigned long timestamp = data->timestamp;
5714
5715         trace_nfs4_renew_async(clp, task->tk_status);
5716         switch (task->tk_status) {
5717         case 0:
5718                 break;
5719         case -NFS4ERR_LEASE_MOVED:
5720                 nfs4_schedule_lease_moved_recovery(clp);
5721                 break;
5722         default:
5723                 /* Unless we're shutting down, schedule state recovery! */
5724                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5725                         return;
5726                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5727                         nfs4_schedule_lease_recovery(clp);
5728                         return;
5729                 }
5730                 nfs4_schedule_path_down_recovery(clp);
5731         }
5732         do_renew_lease(clp, timestamp);
5733 }
5734
5735 static const struct rpc_call_ops nfs4_renew_ops = {
5736         .rpc_call_done = nfs4_renew_done,
5737         .rpc_release = nfs4_renew_release,
5738 };
5739
5740 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5741 {
5742         struct rpc_message msg = {
5743                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5744                 .rpc_argp       = clp,
5745                 .rpc_cred       = cred,
5746         };
5747         struct nfs4_renewdata *data;
5748
5749         if (renew_flags == 0)
5750                 return 0;
5751         if (!refcount_inc_not_zero(&clp->cl_count))
5752                 return -EIO;
5753         data = kmalloc(sizeof(*data), GFP_NOFS);
5754         if (data == NULL) {
5755                 nfs_put_client(clp);
5756                 return -ENOMEM;
5757         }
5758         data->client = clp;
5759         data->timestamp = jiffies;
5760         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5761                         &nfs4_renew_ops, data);
5762 }
5763
5764 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5765 {
5766         struct rpc_message msg = {
5767                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5768                 .rpc_argp       = clp,
5769                 .rpc_cred       = cred,
5770         };
5771         unsigned long now = jiffies;
5772         int status;
5773
5774         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5775         if (status < 0)
5776                 return status;
5777         do_renew_lease(clp, now);
5778         return 0;
5779 }
5780
5781 static bool nfs4_server_supports_acls(const struct nfs_server *server,
5782                                       enum nfs4_acl_type type)
5783 {
5784         switch (type) {
5785         default:
5786                 return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5787         case NFS4ACL_DACL:
5788                 return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5789         case NFS4ACL_SACL:
5790                 return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5791         }
5792 }
5793
5794 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5795  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5796  * the stack.
5797  */
5798 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5799
5800 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5801                 struct page **pages)
5802 {
5803         struct page *newpage, **spages;
5804         int rc = 0;
5805         size_t len;
5806         spages = pages;
5807
5808         do {
5809                 len = min_t(size_t, PAGE_SIZE, buflen);
5810                 newpage = alloc_page(GFP_KERNEL);
5811
5812                 if (newpage == NULL)
5813                         goto unwind;
5814                 memcpy(page_address(newpage), buf, len);
5815                 buf += len;
5816                 buflen -= len;
5817                 *pages++ = newpage;
5818                 rc++;
5819         } while (buflen != 0);
5820
5821         return rc;
5822
5823 unwind:
5824         for(; rc > 0; rc--)
5825                 __free_page(spages[rc-1]);
5826         return -ENOMEM;
5827 }
5828
5829 struct nfs4_cached_acl {
5830         enum nfs4_acl_type type;
5831         int cached;
5832         size_t len;
5833         char data[];
5834 };
5835
5836 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5837 {
5838         struct nfs_inode *nfsi = NFS_I(inode);
5839
5840         spin_lock(&inode->i_lock);
5841         kfree(nfsi->nfs4_acl);
5842         nfsi->nfs4_acl = acl;
5843         spin_unlock(&inode->i_lock);
5844 }
5845
5846 static void nfs4_zap_acl_attr(struct inode *inode)
5847 {
5848         nfs4_set_cached_acl(inode, NULL);
5849 }
5850
5851 static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5852                                     size_t buflen, enum nfs4_acl_type type)
5853 {
5854         struct nfs_inode *nfsi = NFS_I(inode);
5855         struct nfs4_cached_acl *acl;
5856         int ret = -ENOENT;
5857
5858         spin_lock(&inode->i_lock);
5859         acl = nfsi->nfs4_acl;
5860         if (acl == NULL)
5861                 goto out;
5862         if (acl->type != type)
5863                 goto out;
5864         if (buf == NULL) /* user is just asking for length */
5865                 goto out_len;
5866         if (acl->cached == 0)
5867                 goto out;
5868         ret = -ERANGE; /* see getxattr(2) man page */
5869         if (acl->len > buflen)
5870                 goto out;
5871         memcpy(buf, acl->data, acl->len);
5872 out_len:
5873         ret = acl->len;
5874 out:
5875         spin_unlock(&inode->i_lock);
5876         return ret;
5877 }
5878
5879 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
5880                                   size_t pgbase, size_t acl_len,
5881                                   enum nfs4_acl_type type)
5882 {
5883         struct nfs4_cached_acl *acl;
5884         size_t buflen = sizeof(*acl) + acl_len;
5885
5886         if (buflen <= PAGE_SIZE) {
5887                 acl = kmalloc(buflen, GFP_KERNEL);
5888                 if (acl == NULL)
5889                         goto out;
5890                 acl->cached = 1;
5891                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5892         } else {
5893                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5894                 if (acl == NULL)
5895                         goto out;
5896                 acl->cached = 0;
5897         }
5898         acl->type = type;
5899         acl->len = acl_len;
5900 out:
5901         nfs4_set_cached_acl(inode, acl);
5902 }
5903
5904 /*
5905  * The getxattr API returns the required buffer length when called with a
5906  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5907  * the required buf.  On a NULL buf, we send a page of data to the server
5908  * guessing that the ACL request can be serviced by a page. If so, we cache
5909  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5910  * the cache. If not so, we throw away the page, and cache the required
5911  * length. The next getxattr call will then produce another round trip to
5912  * the server, this time with the input buf of the required size.
5913  */
5914 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
5915                                        size_t buflen, enum nfs4_acl_type type)
5916 {
5917         struct page **pages;
5918         struct nfs_getaclargs args = {
5919                 .fh = NFS_FH(inode),
5920                 .acl_type = type,
5921                 .acl_len = buflen,
5922         };
5923         struct nfs_getaclres res = {
5924                 .acl_type = type,
5925                 .acl_len = buflen,
5926         };
5927         struct rpc_message msg = {
5928                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5929                 .rpc_argp = &args,
5930                 .rpc_resp = &res,
5931         };
5932         unsigned int npages;
5933         int ret = -ENOMEM, i;
5934         struct nfs_server *server = NFS_SERVER(inode);
5935
5936         if (buflen == 0)
5937                 buflen = server->rsize;
5938
5939         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5940         pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
5941         if (!pages)
5942                 return -ENOMEM;
5943
5944         args.acl_pages = pages;
5945
5946         for (i = 0; i < npages; i++) {
5947                 pages[i] = alloc_page(GFP_KERNEL);
5948                 if (!pages[i])
5949                         goto out_free;
5950         }
5951
5952         /* for decoding across pages */
5953         res.acl_scratch = alloc_page(GFP_KERNEL);
5954         if (!res.acl_scratch)
5955                 goto out_free;
5956
5957         args.acl_len = npages * PAGE_SIZE;
5958
5959         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5960                 __func__, buf, buflen, npages, args.acl_len);
5961         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5962                              &msg, &args.seq_args, &res.seq_res, 0);
5963         if (ret)
5964                 goto out_free;
5965
5966         /* Handle the case where the passed-in buffer is too short */
5967         if (res.acl_flags & NFS4_ACL_TRUNC) {
5968                 /* Did the user only issue a request for the acl length? */
5969                 if (buf == NULL)
5970                         goto out_ok;
5971                 ret = -ERANGE;
5972                 goto out_free;
5973         }
5974         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
5975                               type);
5976         if (buf) {
5977                 if (res.acl_len > buflen) {
5978                         ret = -ERANGE;
5979                         goto out_free;
5980                 }
5981                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5982         }
5983 out_ok:
5984         ret = res.acl_len;
5985 out_free:
5986         for (i = 0; i < npages; i++)
5987                 if (pages[i])
5988                         __free_page(pages[i]);
5989         if (res.acl_scratch)
5990                 __free_page(res.acl_scratch);
5991         kfree(pages);
5992         return ret;
5993 }
5994
5995 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
5996                                      size_t buflen, enum nfs4_acl_type type)
5997 {
5998         struct nfs4_exception exception = {
5999                 .interruptible = true,
6000         };
6001         ssize_t ret;
6002         do {
6003                 ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
6004                 trace_nfs4_get_acl(inode, ret);
6005                 if (ret >= 0)
6006                         break;
6007                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6008         } while (exception.retry);
6009         return ret;
6010 }
6011
6012 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6013                                  enum nfs4_acl_type type)
6014 {
6015         struct nfs_server *server = NFS_SERVER(inode);
6016         int ret;
6017
6018         if (!nfs4_server_supports_acls(server, type))
6019                 return -EOPNOTSUPP;
6020         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6021         if (ret < 0)
6022                 return ret;
6023         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6024                 nfs_zap_acl_cache(inode);
6025         ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6026         if (ret != -ENOENT)
6027                 /* -ENOENT is returned if there is no ACL or if there is an ACL
6028                  * but no cached acl data, just the acl length */
6029                 return ret;
6030         return nfs4_get_acl_uncached(inode, buf, buflen, type);
6031 }
6032
6033 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6034                                size_t buflen, enum nfs4_acl_type type)
6035 {
6036         struct nfs_server *server = NFS_SERVER(inode);
6037         struct page *pages[NFS4ACL_MAXPAGES];
6038         struct nfs_setaclargs arg = {
6039                 .fh = NFS_FH(inode),
6040                 .acl_type = type,
6041                 .acl_len = buflen,
6042                 .acl_pages = pages,
6043         };
6044         struct nfs_setaclres res;
6045         struct rpc_message msg = {
6046                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6047                 .rpc_argp       = &arg,
6048                 .rpc_resp       = &res,
6049         };
6050         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6051         int ret, i;
6052
6053         /* You can't remove system.nfs4_acl: */
6054         if (buflen == 0)
6055                 return -EINVAL;
6056         if (!nfs4_server_supports_acls(server, type))
6057                 return -EOPNOTSUPP;
6058         if (npages > ARRAY_SIZE(pages))
6059                 return -ERANGE;
6060         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6061         if (i < 0)
6062                 return i;
6063         nfs4_inode_make_writeable(inode);
6064         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6065
6066         /*
6067          * Free each page after tx, so the only ref left is
6068          * held by the network stack
6069          */
6070         for (; i > 0; i--)
6071                 put_page(pages[i-1]);
6072
6073         /*
6074          * Acl update can result in inode attribute update.
6075          * so mark the attribute cache invalid.
6076          */
6077         spin_lock(&inode->i_lock);
6078         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6079                                              NFS_INO_INVALID_CTIME |
6080                                              NFS_INO_REVAL_FORCED);
6081         spin_unlock(&inode->i_lock);
6082         nfs_access_zap_cache(inode);
6083         nfs_zap_acl_cache(inode);
6084         return ret;
6085 }
6086
6087 static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6088                              size_t buflen, enum nfs4_acl_type type)
6089 {
6090         struct nfs4_exception exception = { };
6091         int err;
6092         do {
6093                 err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6094                 trace_nfs4_set_acl(inode, err);
6095                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6096                         /*
6097                          * no need to retry since the kernel
6098                          * isn't involved in encoding the ACEs.
6099                          */
6100                         err = -EINVAL;
6101                         break;
6102                 }
6103                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6104                                 &exception);
6105         } while (exception.retry);
6106         return err;
6107 }
6108
6109 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6110 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6111                                         size_t buflen)
6112 {
6113         struct nfs_server *server = NFS_SERVER(inode);
6114         struct nfs4_label label = {0, 0, buflen, buf};
6115
6116         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6117         struct nfs_fattr fattr = {
6118                 .label = &label,
6119         };
6120         struct nfs4_getattr_arg arg = {
6121                 .fh             = NFS_FH(inode),
6122                 .bitmask        = bitmask,
6123         };
6124         struct nfs4_getattr_res res = {
6125                 .fattr          = &fattr,
6126                 .server         = server,
6127         };
6128         struct rpc_message msg = {
6129                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6130                 .rpc_argp       = &arg,
6131                 .rpc_resp       = &res,
6132         };
6133         int ret;
6134
6135         nfs_fattr_init(&fattr);
6136
6137         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6138         if (ret)
6139                 return ret;
6140         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6141                 return -ENOENT;
6142         return label.len;
6143 }
6144
6145 static int nfs4_get_security_label(struct inode *inode, void *buf,
6146                                         size_t buflen)
6147 {
6148         struct nfs4_exception exception = {
6149                 .interruptible = true,
6150         };
6151         int err;
6152
6153         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6154                 return -EOPNOTSUPP;
6155
6156         do {
6157                 err = _nfs4_get_security_label(inode, buf, buflen);
6158                 trace_nfs4_get_security_label(inode, err);
6159                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6160                                 &exception);
6161         } while (exception.retry);
6162         return err;
6163 }
6164
6165 static int _nfs4_do_set_security_label(struct inode *inode,
6166                 struct nfs4_label *ilabel,
6167                 struct nfs_fattr *fattr)
6168 {
6169
6170         struct iattr sattr = {0};
6171         struct nfs_server *server = NFS_SERVER(inode);
6172         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6173         struct nfs_setattrargs arg = {
6174                 .fh             = NFS_FH(inode),
6175                 .iap            = &sattr,
6176                 .server         = server,
6177                 .bitmask        = bitmask,
6178                 .label          = ilabel,
6179         };
6180         struct nfs_setattrres res = {
6181                 .fattr          = fattr,
6182                 .server         = server,
6183         };
6184         struct rpc_message msg = {
6185                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6186                 .rpc_argp       = &arg,
6187                 .rpc_resp       = &res,
6188         };
6189         int status;
6190
6191         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6192
6193         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6194         if (status)
6195                 dprintk("%s failed: %d\n", __func__, status);
6196
6197         return status;
6198 }
6199
6200 static int nfs4_do_set_security_label(struct inode *inode,
6201                 struct nfs4_label *ilabel,
6202                 struct nfs_fattr *fattr)
6203 {
6204         struct nfs4_exception exception = { };
6205         int err;
6206
6207         do {
6208                 err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6209                 trace_nfs4_set_security_label(inode, err);
6210                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6211                                 &exception);
6212         } while (exception.retry);
6213         return err;
6214 }
6215
6216 static int
6217 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6218 {
6219         struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6220         struct nfs_fattr *fattr;
6221         int status;
6222
6223         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6224                 return -EOPNOTSUPP;
6225
6226         fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6227         if (fattr == NULL)
6228                 return -ENOMEM;
6229
6230         status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6231         if (status == 0)
6232                 nfs_setsecurity(inode, fattr);
6233
6234         return status;
6235 }
6236 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6237
6238
6239 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6240                                     nfs4_verifier *bootverf)
6241 {
6242         __be32 verf[2];
6243
6244         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6245                 /* An impossible timestamp guarantees this value
6246                  * will never match a generated boot time. */
6247                 verf[0] = cpu_to_be32(U32_MAX);
6248                 verf[1] = cpu_to_be32(U32_MAX);
6249         } else {
6250                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6251                 u64 ns = ktime_to_ns(nn->boot_time);
6252
6253                 verf[0] = cpu_to_be32(ns >> 32);
6254                 verf[1] = cpu_to_be32(ns);
6255         }
6256         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6257 }
6258
6259 static size_t
6260 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6261 {
6262         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6263         struct nfs_netns_client *nn_clp = nn->nfs_client;
6264         const char *id;
6265
6266         buf[0] = '\0';
6267
6268         if (nn_clp) {
6269                 rcu_read_lock();
6270                 id = rcu_dereference(nn_clp->identifier);
6271                 if (id)
6272                         strscpy(buf, id, buflen);
6273                 rcu_read_unlock();
6274         }
6275
6276         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6277                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6278
6279         return strlen(buf);
6280 }
6281
6282 static int
6283 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6284 {
6285         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6286         size_t buflen;
6287         size_t len;
6288         char *str;
6289
6290         if (clp->cl_owner_id != NULL)
6291                 return 0;
6292
6293         rcu_read_lock();
6294         len = 14 +
6295                 strlen(clp->cl_rpcclient->cl_nodename) +
6296                 1 +
6297                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6298                 1;
6299         rcu_read_unlock();
6300
6301         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6302         if (buflen)
6303                 len += buflen + 1;
6304
6305         if (len > NFS4_OPAQUE_LIMIT + 1)
6306                 return -EINVAL;
6307
6308         /*
6309          * Since this string is allocated at mount time, and held until the
6310          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6311          * about a memory-reclaim deadlock.
6312          */
6313         str = kmalloc(len, GFP_KERNEL);
6314         if (!str)
6315                 return -ENOMEM;
6316
6317         rcu_read_lock();
6318         if (buflen)
6319                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6320                           clp->cl_rpcclient->cl_nodename, buf,
6321                           rpc_peeraddr2str(clp->cl_rpcclient,
6322                                            RPC_DISPLAY_ADDR));
6323         else
6324                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6325                           clp->cl_rpcclient->cl_nodename,
6326                           rpc_peeraddr2str(clp->cl_rpcclient,
6327                                            RPC_DISPLAY_ADDR));
6328         rcu_read_unlock();
6329
6330         clp->cl_owner_id = str;
6331         return 0;
6332 }
6333
6334 static int
6335 nfs4_init_uniform_client_string(struct nfs_client *clp)
6336 {
6337         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6338         size_t buflen;
6339         size_t len;
6340         char *str;
6341
6342         if (clp->cl_owner_id != NULL)
6343                 return 0;
6344
6345         len = 10 + 10 + 1 + 10 + 1 +
6346                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6347
6348         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6349         if (buflen)
6350                 len += buflen + 1;
6351
6352         if (len > NFS4_OPAQUE_LIMIT + 1)
6353                 return -EINVAL;
6354
6355         /*
6356          * Since this string is allocated at mount time, and held until the
6357          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6358          * about a memory-reclaim deadlock.
6359          */
6360         str = kmalloc(len, GFP_KERNEL);
6361         if (!str)
6362                 return -ENOMEM;
6363
6364         if (buflen)
6365                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6366                           clp->rpc_ops->version, clp->cl_minorversion,
6367                           buf, clp->cl_rpcclient->cl_nodename);
6368         else
6369                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6370                           clp->rpc_ops->version, clp->cl_minorversion,
6371                           clp->cl_rpcclient->cl_nodename);
6372         clp->cl_owner_id = str;
6373         return 0;
6374 }
6375
6376 /*
6377  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6378  * services.  Advertise one based on the address family of the
6379  * clientaddr.
6380  */
6381 static unsigned int
6382 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6383 {
6384         if (strchr(clp->cl_ipaddr, ':') != NULL)
6385                 return scnprintf(buf, len, "tcp6");
6386         else
6387                 return scnprintf(buf, len, "tcp");
6388 }
6389
6390 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6391 {
6392         struct nfs4_setclientid *sc = calldata;
6393
6394         if (task->tk_status == 0)
6395                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6396 }
6397
6398 static const struct rpc_call_ops nfs4_setclientid_ops = {
6399         .rpc_call_done = nfs4_setclientid_done,
6400 };
6401
6402 /**
6403  * nfs4_proc_setclientid - Negotiate client ID
6404  * @clp: state data structure
6405  * @program: RPC program for NFSv4 callback service
6406  * @port: IP port number for NFS4 callback service
6407  * @cred: credential to use for this call
6408  * @res: where to place the result
6409  *
6410  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6411  */
6412 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6413                 unsigned short port, const struct cred *cred,
6414                 struct nfs4_setclientid_res *res)
6415 {
6416         nfs4_verifier sc_verifier;
6417         struct nfs4_setclientid setclientid = {
6418                 .sc_verifier = &sc_verifier,
6419                 .sc_prog = program,
6420                 .sc_clnt = clp,
6421         };
6422         struct rpc_message msg = {
6423                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6424                 .rpc_argp = &setclientid,
6425                 .rpc_resp = res,
6426                 .rpc_cred = cred,
6427         };
6428         struct rpc_task_setup task_setup_data = {
6429                 .rpc_client = clp->cl_rpcclient,
6430                 .rpc_message = &msg,
6431                 .callback_ops = &nfs4_setclientid_ops,
6432                 .callback_data = &setclientid,
6433                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6434         };
6435         unsigned long now = jiffies;
6436         int status;
6437
6438         /* nfs_client_id4 */
6439         nfs4_init_boot_verifier(clp, &sc_verifier);
6440
6441         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6442                 status = nfs4_init_uniform_client_string(clp);
6443         else
6444                 status = nfs4_init_nonuniform_client_string(clp);
6445
6446         if (status)
6447                 goto out;
6448
6449         /* cb_client4 */
6450         setclientid.sc_netid_len =
6451                                 nfs4_init_callback_netid(clp,
6452                                                 setclientid.sc_netid,
6453                                                 sizeof(setclientid.sc_netid));
6454         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6455                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6456                                 clp->cl_ipaddr, port >> 8, port & 255);
6457
6458         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6459                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6460                 clp->cl_owner_id);
6461
6462         status = nfs4_call_sync_custom(&task_setup_data);
6463         if (setclientid.sc_cred) {
6464                 kfree(clp->cl_acceptor);
6465                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6466                 put_rpccred(setclientid.sc_cred);
6467         }
6468
6469         if (status == 0)
6470                 do_renew_lease(clp, now);
6471 out:
6472         trace_nfs4_setclientid(clp, status);
6473         dprintk("NFS reply setclientid: %d\n", status);
6474         return status;
6475 }
6476
6477 /**
6478  * nfs4_proc_setclientid_confirm - Confirm client ID
6479  * @clp: state data structure
6480  * @arg: result of a previous SETCLIENTID
6481  * @cred: credential to use for this call
6482  *
6483  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6484  */
6485 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6486                 struct nfs4_setclientid_res *arg,
6487                 const struct cred *cred)
6488 {
6489         struct rpc_message msg = {
6490                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6491                 .rpc_argp = arg,
6492                 .rpc_cred = cred,
6493         };
6494         int status;
6495
6496         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6497                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6498                 clp->cl_clientid);
6499         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6500                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6501         trace_nfs4_setclientid_confirm(clp, status);
6502         dprintk("NFS reply setclientid_confirm: %d\n", status);
6503         return status;
6504 }
6505
6506 struct nfs4_delegreturndata {
6507         struct nfs4_delegreturnargs args;
6508         struct nfs4_delegreturnres res;
6509         struct nfs_fh fh;
6510         nfs4_stateid stateid;
6511         unsigned long timestamp;
6512         struct {
6513                 struct nfs4_layoutreturn_args arg;
6514                 struct nfs4_layoutreturn_res res;
6515                 struct nfs4_xdr_opaque_data ld_private;
6516                 u32 roc_barrier;
6517                 bool roc;
6518         } lr;
6519         struct nfs_fattr fattr;
6520         int rpc_status;
6521         struct inode *inode;
6522 };
6523
6524 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6525 {
6526         struct nfs4_delegreturndata *data = calldata;
6527         struct nfs4_exception exception = {
6528                 .inode = data->inode,
6529                 .stateid = &data->stateid,
6530                 .task_is_privileged = data->args.seq_args.sa_privileged,
6531         };
6532
6533         if (!nfs4_sequence_done(task, &data->res.seq_res))
6534                 return;
6535
6536         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6537
6538         /* Handle Layoutreturn errors */
6539         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6540                           &data->res.lr_ret) == -EAGAIN)
6541                 goto out_restart;
6542
6543         switch (task->tk_status) {
6544         case 0:
6545                 renew_lease(data->res.server, data->timestamp);
6546                 break;
6547         case -NFS4ERR_ADMIN_REVOKED:
6548         case -NFS4ERR_DELEG_REVOKED:
6549         case -NFS4ERR_EXPIRED:
6550                 nfs4_free_revoked_stateid(data->res.server,
6551                                 data->args.stateid,
6552                                 task->tk_msg.rpc_cred);
6553                 fallthrough;
6554         case -NFS4ERR_BAD_STATEID:
6555         case -NFS4ERR_STALE_STATEID:
6556         case -ETIMEDOUT:
6557                 task->tk_status = 0;
6558                 break;
6559         case -NFS4ERR_OLD_STATEID:
6560                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6561                         nfs4_stateid_seqid_inc(&data->stateid);
6562                 if (data->args.bitmask) {
6563                         data->args.bitmask = NULL;
6564                         data->res.fattr = NULL;
6565                 }
6566                 goto out_restart;
6567         case -NFS4ERR_ACCESS:
6568                 if (data->args.bitmask) {
6569                         data->args.bitmask = NULL;
6570                         data->res.fattr = NULL;
6571                         goto out_restart;
6572                 }
6573                 fallthrough;
6574         default:
6575                 task->tk_status = nfs4_async_handle_exception(task,
6576                                 data->res.server, task->tk_status,
6577                                 &exception);
6578                 if (exception.retry)
6579                         goto out_restart;
6580         }
6581         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6582         data->rpc_status = task->tk_status;
6583         return;
6584 out_restart:
6585         task->tk_status = 0;
6586         rpc_restart_call_prepare(task);
6587 }
6588
6589 static void nfs4_delegreturn_release(void *calldata)
6590 {
6591         struct nfs4_delegreturndata *data = calldata;
6592         struct inode *inode = data->inode;
6593
6594         if (data->lr.roc)
6595                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6596                                  data->res.lr_ret);
6597         if (inode) {
6598                 nfs4_fattr_set_prechange(&data->fattr,
6599                                          inode_peek_iversion_raw(inode));
6600                 nfs_refresh_inode(inode, &data->fattr);
6601                 nfs_iput_and_deactive(inode);
6602         }
6603         kfree(calldata);
6604 }
6605
6606 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6607 {
6608         struct nfs4_delegreturndata *d_data;
6609         struct pnfs_layout_hdr *lo;
6610
6611         d_data = (struct nfs4_delegreturndata *)data;
6612
6613         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6614                 nfs4_sequence_done(task, &d_data->res.seq_res);
6615                 return;
6616         }
6617
6618         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6619         if (lo && !pnfs_layout_is_valid(lo)) {
6620                 d_data->args.lr_args = NULL;
6621                 d_data->res.lr_res = NULL;
6622         }
6623
6624         nfs4_setup_sequence(d_data->res.server->nfs_client,
6625                         &d_data->args.seq_args,
6626                         &d_data->res.seq_res,
6627                         task);
6628 }
6629
6630 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6631         .rpc_call_prepare = nfs4_delegreturn_prepare,
6632         .rpc_call_done = nfs4_delegreturn_done,
6633         .rpc_release = nfs4_delegreturn_release,
6634 };
6635
6636 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6637 {
6638         struct nfs4_delegreturndata *data;
6639         struct nfs_server *server = NFS_SERVER(inode);
6640         struct rpc_task *task;
6641         struct rpc_message msg = {
6642                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6643                 .rpc_cred = cred,
6644         };
6645         struct rpc_task_setup task_setup_data = {
6646                 .rpc_client = server->client,
6647                 .rpc_message = &msg,
6648                 .callback_ops = &nfs4_delegreturn_ops,
6649                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6650         };
6651         int status = 0;
6652
6653         if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6654                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6655
6656         data = kzalloc(sizeof(*data), GFP_KERNEL);
6657         if (data == NULL)
6658                 return -ENOMEM;
6659
6660         nfs4_state_protect(server->nfs_client,
6661                         NFS_SP4_MACH_CRED_CLEANUP,
6662                         &task_setup_data.rpc_client, &msg);
6663
6664         data->args.fhandle = &data->fh;
6665         data->args.stateid = &data->stateid;
6666         nfs4_bitmask_set(data->args.bitmask_store,
6667                          server->cache_consistency_bitmask, inode, 0);
6668         data->args.bitmask = data->args.bitmask_store;
6669         nfs_copy_fh(&data->fh, NFS_FH(inode));
6670         nfs4_stateid_copy(&data->stateid, stateid);
6671         data->res.fattr = &data->fattr;
6672         data->res.server = server;
6673         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6674         data->lr.arg.ld_private = &data->lr.ld_private;
6675         nfs_fattr_init(data->res.fattr);
6676         data->timestamp = jiffies;
6677         data->rpc_status = 0;
6678         data->inode = nfs_igrab_and_active(inode);
6679         if (data->inode || issync) {
6680                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6681                                         cred);
6682                 if (data->lr.roc) {
6683                         data->args.lr_args = &data->lr.arg;
6684                         data->res.lr_res = &data->lr.res;
6685                 }
6686         }
6687
6688         if (!data->inode)
6689                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6690                                    1);
6691         else
6692                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6693                                    0);
6694         task_setup_data.callback_data = data;
6695         msg.rpc_argp = &data->args;
6696         msg.rpc_resp = &data->res;
6697         task = rpc_run_task(&task_setup_data);
6698         if (IS_ERR(task))
6699                 return PTR_ERR(task);
6700         if (!issync)
6701                 goto out;
6702         status = rpc_wait_for_completion_task(task);
6703         if (status != 0)
6704                 goto out;
6705         status = data->rpc_status;
6706 out:
6707         rpc_put_task(task);
6708         return status;
6709 }
6710
6711 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6712 {
6713         struct nfs_server *server = NFS_SERVER(inode);
6714         struct nfs4_exception exception = { };
6715         int err;
6716         do {
6717                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6718                 trace_nfs4_delegreturn(inode, stateid, err);
6719                 switch (err) {
6720                         case -NFS4ERR_STALE_STATEID:
6721                         case -NFS4ERR_EXPIRED:
6722                         case 0:
6723                                 return 0;
6724                 }
6725                 err = nfs4_handle_exception(server, err, &exception);
6726         } while (exception.retry);
6727         return err;
6728 }
6729
6730 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6731 {
6732         struct inode *inode = state->inode;
6733         struct nfs_server *server = NFS_SERVER(inode);
6734         struct nfs_client *clp = server->nfs_client;
6735         struct nfs_lockt_args arg = {
6736                 .fh = NFS_FH(inode),
6737                 .fl = request,
6738         };
6739         struct nfs_lockt_res res = {
6740                 .denied = request,
6741         };
6742         struct rpc_message msg = {
6743                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6744                 .rpc_argp       = &arg,
6745                 .rpc_resp       = &res,
6746                 .rpc_cred       = state->owner->so_cred,
6747         };
6748         struct nfs4_lock_state *lsp;
6749         int status;
6750
6751         arg.lock_owner.clientid = clp->cl_clientid;
6752         status = nfs4_set_lock_state(state, request);
6753         if (status != 0)
6754                 goto out;
6755         lsp = request->fl_u.nfs4_fl.owner;
6756         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6757         arg.lock_owner.s_dev = server->s_dev;
6758         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6759         switch (status) {
6760                 case 0:
6761                         request->fl_type = F_UNLCK;
6762                         break;
6763                 case -NFS4ERR_DENIED:
6764                         status = 0;
6765         }
6766         request->fl_ops->fl_release_private(request);
6767         request->fl_ops = NULL;
6768 out:
6769         return status;
6770 }
6771
6772 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6773 {
6774         struct nfs4_exception exception = {
6775                 .interruptible = true,
6776         };
6777         int err;
6778
6779         do {
6780                 err = _nfs4_proc_getlk(state, cmd, request);
6781                 trace_nfs4_get_lock(request, state, cmd, err);
6782                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6783                                 &exception);
6784         } while (exception.retry);
6785         return err;
6786 }
6787
6788 /*
6789  * Update the seqid of a lock stateid after receiving
6790  * NFS4ERR_OLD_STATEID
6791  */
6792 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6793                 struct nfs4_lock_state *lsp)
6794 {
6795         struct nfs4_state *state = lsp->ls_state;
6796         bool ret = false;
6797
6798         spin_lock(&state->state_lock);
6799         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6800                 goto out;
6801         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6802                 nfs4_stateid_seqid_inc(dst);
6803         else
6804                 dst->seqid = lsp->ls_stateid.seqid;
6805         ret = true;
6806 out:
6807         spin_unlock(&state->state_lock);
6808         return ret;
6809 }
6810
6811 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6812                 struct nfs4_lock_state *lsp)
6813 {
6814         struct nfs4_state *state = lsp->ls_state;
6815         bool ret;
6816
6817         spin_lock(&state->state_lock);
6818         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6819         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6820         spin_unlock(&state->state_lock);
6821         return ret;
6822 }
6823
6824 struct nfs4_unlockdata {
6825         struct nfs_locku_args arg;
6826         struct nfs_locku_res res;
6827         struct nfs4_lock_state *lsp;
6828         struct nfs_open_context *ctx;
6829         struct nfs_lock_context *l_ctx;
6830         struct file_lock fl;
6831         struct nfs_server *server;
6832         unsigned long timestamp;
6833 };
6834
6835 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6836                 struct nfs_open_context *ctx,
6837                 struct nfs4_lock_state *lsp,
6838                 struct nfs_seqid *seqid)
6839 {
6840         struct nfs4_unlockdata *p;
6841         struct nfs4_state *state = lsp->ls_state;
6842         struct inode *inode = state->inode;
6843
6844         p = kzalloc(sizeof(*p), GFP_KERNEL);
6845         if (p == NULL)
6846                 return NULL;
6847         p->arg.fh = NFS_FH(inode);
6848         p->arg.fl = &p->fl;
6849         p->arg.seqid = seqid;
6850         p->res.seqid = seqid;
6851         p->lsp = lsp;
6852         /* Ensure we don't close file until we're done freeing locks! */
6853         p->ctx = get_nfs_open_context(ctx);
6854         p->l_ctx = nfs_get_lock_context(ctx);
6855         locks_init_lock(&p->fl);
6856         locks_copy_lock(&p->fl, fl);
6857         p->server = NFS_SERVER(inode);
6858         spin_lock(&state->state_lock);
6859         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6860         spin_unlock(&state->state_lock);
6861         return p;
6862 }
6863
6864 static void nfs4_locku_release_calldata(void *data)
6865 {
6866         struct nfs4_unlockdata *calldata = data;
6867         nfs_free_seqid(calldata->arg.seqid);
6868         nfs4_put_lock_state(calldata->lsp);
6869         nfs_put_lock_context(calldata->l_ctx);
6870         put_nfs_open_context(calldata->ctx);
6871         kfree(calldata);
6872 }
6873
6874 static void nfs4_locku_done(struct rpc_task *task, void *data)
6875 {
6876         struct nfs4_unlockdata *calldata = data;
6877         struct nfs4_exception exception = {
6878                 .inode = calldata->lsp->ls_state->inode,
6879                 .stateid = &calldata->arg.stateid,
6880         };
6881
6882         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6883                 return;
6884         switch (task->tk_status) {
6885                 case 0:
6886                         renew_lease(calldata->server, calldata->timestamp);
6887                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6888                         if (nfs4_update_lock_stateid(calldata->lsp,
6889                                         &calldata->res.stateid))
6890                                 break;
6891                         fallthrough;
6892                 case -NFS4ERR_ADMIN_REVOKED:
6893                 case -NFS4ERR_EXPIRED:
6894                         nfs4_free_revoked_stateid(calldata->server,
6895                                         &calldata->arg.stateid,
6896                                         task->tk_msg.rpc_cred);
6897                         fallthrough;
6898                 case -NFS4ERR_BAD_STATEID:
6899                 case -NFS4ERR_STALE_STATEID:
6900                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6901                                                 calldata->lsp))
6902                                 rpc_restart_call_prepare(task);
6903                         break;
6904                 case -NFS4ERR_OLD_STATEID:
6905                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6906                                                 calldata->lsp))
6907                                 rpc_restart_call_prepare(task);
6908                         break;
6909                 default:
6910                         task->tk_status = nfs4_async_handle_exception(task,
6911                                         calldata->server, task->tk_status,
6912                                         &exception);
6913                         if (exception.retry)
6914                                 rpc_restart_call_prepare(task);
6915         }
6916         nfs_release_seqid(calldata->arg.seqid);
6917 }
6918
6919 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6920 {
6921         struct nfs4_unlockdata *calldata = data;
6922
6923         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6924                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6925                 return;
6926
6927         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6928                 goto out_wait;
6929         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6930                 /* Note: exit _without_ running nfs4_locku_done */
6931                 goto out_no_action;
6932         }
6933         calldata->timestamp = jiffies;
6934         if (nfs4_setup_sequence(calldata->server->nfs_client,
6935                                 &calldata->arg.seq_args,
6936                                 &calldata->res.seq_res,
6937                                 task) != 0)
6938                 nfs_release_seqid(calldata->arg.seqid);
6939         return;
6940 out_no_action:
6941         task->tk_action = NULL;
6942 out_wait:
6943         nfs4_sequence_done(task, &calldata->res.seq_res);
6944 }
6945
6946 static const struct rpc_call_ops nfs4_locku_ops = {
6947         .rpc_call_prepare = nfs4_locku_prepare,
6948         .rpc_call_done = nfs4_locku_done,
6949         .rpc_release = nfs4_locku_release_calldata,
6950 };
6951
6952 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6953                 struct nfs_open_context *ctx,
6954                 struct nfs4_lock_state *lsp,
6955                 struct nfs_seqid *seqid)
6956 {
6957         struct nfs4_unlockdata *data;
6958         struct rpc_message msg = {
6959                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6960                 .rpc_cred = ctx->cred,
6961         };
6962         struct rpc_task_setup task_setup_data = {
6963                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6964                 .rpc_message = &msg,
6965                 .callback_ops = &nfs4_locku_ops,
6966                 .workqueue = nfsiod_workqueue,
6967                 .flags = RPC_TASK_ASYNC,
6968         };
6969
6970         if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
6971                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6972
6973         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6974                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6975
6976         /* Ensure this is an unlock - when canceling a lock, the
6977          * canceled lock is passed in, and it won't be an unlock.
6978          */
6979         fl->fl_type = F_UNLCK;
6980         if (fl->fl_flags & FL_CLOSE)
6981                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6982
6983         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6984         if (data == NULL) {
6985                 nfs_free_seqid(seqid);
6986                 return ERR_PTR(-ENOMEM);
6987         }
6988
6989         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6990         msg.rpc_argp = &data->arg;
6991         msg.rpc_resp = &data->res;
6992         task_setup_data.callback_data = data;
6993         return rpc_run_task(&task_setup_data);
6994 }
6995
6996 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6997 {
6998         struct inode *inode = state->inode;
6999         struct nfs4_state_owner *sp = state->owner;
7000         struct nfs_inode *nfsi = NFS_I(inode);
7001         struct nfs_seqid *seqid;
7002         struct nfs4_lock_state *lsp;
7003         struct rpc_task *task;
7004         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7005         int status = 0;
7006         unsigned char fl_flags = request->fl_flags;
7007
7008         status = nfs4_set_lock_state(state, request);
7009         /* Unlock _before_ we do the RPC call */
7010         request->fl_flags |= FL_EXISTS;
7011         /* Exclude nfs_delegation_claim_locks() */
7012         mutex_lock(&sp->so_delegreturn_mutex);
7013         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7014         down_read(&nfsi->rwsem);
7015         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7016                 up_read(&nfsi->rwsem);
7017                 mutex_unlock(&sp->so_delegreturn_mutex);
7018                 goto out;
7019         }
7020         up_read(&nfsi->rwsem);
7021         mutex_unlock(&sp->so_delegreturn_mutex);
7022         if (status != 0)
7023                 goto out;
7024         /* Is this a delegated lock? */
7025         lsp = request->fl_u.nfs4_fl.owner;
7026         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7027                 goto out;
7028         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7029         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7030         status = -ENOMEM;
7031         if (IS_ERR(seqid))
7032                 goto out;
7033         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
7034         status = PTR_ERR(task);
7035         if (IS_ERR(task))
7036                 goto out;
7037         status = rpc_wait_for_completion_task(task);
7038         rpc_put_task(task);
7039 out:
7040         request->fl_flags = fl_flags;
7041         trace_nfs4_unlock(request, state, F_SETLK, status);
7042         return status;
7043 }
7044
7045 struct nfs4_lockdata {
7046         struct nfs_lock_args arg;
7047         struct nfs_lock_res res;
7048         struct nfs4_lock_state *lsp;
7049         struct nfs_open_context *ctx;
7050         struct file_lock fl;
7051         unsigned long timestamp;
7052         int rpc_status;
7053         int cancelled;
7054         struct nfs_server *server;
7055 };
7056
7057 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7058                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7059                 gfp_t gfp_mask)
7060 {
7061         struct nfs4_lockdata *p;
7062         struct inode *inode = lsp->ls_state->inode;
7063         struct nfs_server *server = NFS_SERVER(inode);
7064         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7065
7066         p = kzalloc(sizeof(*p), gfp_mask);
7067         if (p == NULL)
7068                 return NULL;
7069
7070         p->arg.fh = NFS_FH(inode);
7071         p->arg.fl = &p->fl;
7072         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7073         if (IS_ERR(p->arg.open_seqid))
7074                 goto out_free;
7075         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7076         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7077         if (IS_ERR(p->arg.lock_seqid))
7078                 goto out_free_seqid;
7079         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7080         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7081         p->arg.lock_owner.s_dev = server->s_dev;
7082         p->res.lock_seqid = p->arg.lock_seqid;
7083         p->lsp = lsp;
7084         p->server = server;
7085         p->ctx = get_nfs_open_context(ctx);
7086         locks_init_lock(&p->fl);
7087         locks_copy_lock(&p->fl, fl);
7088         return p;
7089 out_free_seqid:
7090         nfs_free_seqid(p->arg.open_seqid);
7091 out_free:
7092         kfree(p);
7093         return NULL;
7094 }
7095
7096 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7097 {
7098         struct nfs4_lockdata *data = calldata;
7099         struct nfs4_state *state = data->lsp->ls_state;
7100
7101         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7102                 goto out_wait;
7103         /* Do we need to do an open_to_lock_owner? */
7104         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7105                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7106                         goto out_release_lock_seqid;
7107                 }
7108                 nfs4_stateid_copy(&data->arg.open_stateid,
7109                                 &state->open_stateid);
7110                 data->arg.new_lock_owner = 1;
7111                 data->res.open_seqid = data->arg.open_seqid;
7112         } else {
7113                 data->arg.new_lock_owner = 0;
7114                 nfs4_stateid_copy(&data->arg.lock_stateid,
7115                                 &data->lsp->ls_stateid);
7116         }
7117         if (!nfs4_valid_open_stateid(state)) {
7118                 data->rpc_status = -EBADF;
7119                 task->tk_action = NULL;
7120                 goto out_release_open_seqid;
7121         }
7122         data->timestamp = jiffies;
7123         if (nfs4_setup_sequence(data->server->nfs_client,
7124                                 &data->arg.seq_args,
7125                                 &data->res.seq_res,
7126                                 task) == 0)
7127                 return;
7128 out_release_open_seqid:
7129         nfs_release_seqid(data->arg.open_seqid);
7130 out_release_lock_seqid:
7131         nfs_release_seqid(data->arg.lock_seqid);
7132 out_wait:
7133         nfs4_sequence_done(task, &data->res.seq_res);
7134         dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7135 }
7136
7137 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7138 {
7139         struct nfs4_lockdata *data = calldata;
7140         struct nfs4_lock_state *lsp = data->lsp;
7141
7142         if (!nfs4_sequence_done(task, &data->res.seq_res))
7143                 return;
7144
7145         data->rpc_status = task->tk_status;
7146         switch (task->tk_status) {
7147         case 0:
7148                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7149                                 data->timestamp);
7150                 if (data->arg.new_lock && !data->cancelled) {
7151                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7152                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7153                                 goto out_restart;
7154                 }
7155                 if (data->arg.new_lock_owner != 0) {
7156                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
7157                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7158                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7159                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7160                         goto out_restart;
7161                 break;
7162         case -NFS4ERR_BAD_STATEID:
7163         case -NFS4ERR_OLD_STATEID:
7164         case -NFS4ERR_STALE_STATEID:
7165         case -NFS4ERR_EXPIRED:
7166                 if (data->arg.new_lock_owner != 0) {
7167                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7168                                                 &lsp->ls_state->open_stateid))
7169                                 goto out_restart;
7170                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7171                                                 &lsp->ls_stateid))
7172                                 goto out_restart;
7173         }
7174 out_done:
7175         dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7176         return;
7177 out_restart:
7178         if (!data->cancelled)
7179                 rpc_restart_call_prepare(task);
7180         goto out_done;
7181 }
7182
7183 static void nfs4_lock_release(void *calldata)
7184 {
7185         struct nfs4_lockdata *data = calldata;
7186
7187         nfs_free_seqid(data->arg.open_seqid);
7188         if (data->cancelled && data->rpc_status == 0) {
7189                 struct rpc_task *task;
7190                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7191                                 data->arg.lock_seqid);
7192                 if (!IS_ERR(task))
7193                         rpc_put_task_async(task);
7194                 dprintk("%s: cancelling lock!\n", __func__);
7195         } else
7196                 nfs_free_seqid(data->arg.lock_seqid);
7197         nfs4_put_lock_state(data->lsp);
7198         put_nfs_open_context(data->ctx);
7199         kfree(data);
7200 }
7201
7202 static const struct rpc_call_ops nfs4_lock_ops = {
7203         .rpc_call_prepare = nfs4_lock_prepare,
7204         .rpc_call_done = nfs4_lock_done,
7205         .rpc_release = nfs4_lock_release,
7206 };
7207
7208 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7209 {
7210         switch (error) {
7211         case -NFS4ERR_ADMIN_REVOKED:
7212         case -NFS4ERR_EXPIRED:
7213         case -NFS4ERR_BAD_STATEID:
7214                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7215                 if (new_lock_owner != 0 ||
7216                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7217                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7218                 break;
7219         case -NFS4ERR_STALE_STATEID:
7220                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7221                 nfs4_schedule_lease_recovery(server->nfs_client);
7222         }
7223 }
7224
7225 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7226 {
7227         struct nfs4_lockdata *data;
7228         struct rpc_task *task;
7229         struct rpc_message msg = {
7230                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7231                 .rpc_cred = state->owner->so_cred,
7232         };
7233         struct rpc_task_setup task_setup_data = {
7234                 .rpc_client = NFS_CLIENT(state->inode),
7235                 .rpc_message = &msg,
7236                 .callback_ops = &nfs4_lock_ops,
7237                 .workqueue = nfsiod_workqueue,
7238                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7239         };
7240         int ret;
7241
7242         if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7243                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
7244
7245         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7246                                    fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7247         if (data == NULL)
7248                 return -ENOMEM;
7249         if (IS_SETLKW(cmd))
7250                 data->arg.block = 1;
7251         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7252                                 recovery_type > NFS_LOCK_NEW);
7253         msg.rpc_argp = &data->arg;
7254         msg.rpc_resp = &data->res;
7255         task_setup_data.callback_data = data;
7256         if (recovery_type > NFS_LOCK_NEW) {
7257                 if (recovery_type == NFS_LOCK_RECLAIM)
7258                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7259         } else
7260                 data->arg.new_lock = 1;
7261         task = rpc_run_task(&task_setup_data);
7262         if (IS_ERR(task))
7263                 return PTR_ERR(task);
7264         ret = rpc_wait_for_completion_task(task);
7265         if (ret == 0) {
7266                 ret = data->rpc_status;
7267                 if (ret)
7268                         nfs4_handle_setlk_error(data->server, data->lsp,
7269                                         data->arg.new_lock_owner, ret);
7270         } else
7271                 data->cancelled = true;
7272         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7273         rpc_put_task(task);
7274         dprintk("%s: ret = %d\n", __func__, ret);
7275         return ret;
7276 }
7277
7278 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7279 {
7280         struct nfs_server *server = NFS_SERVER(state->inode);
7281         struct nfs4_exception exception = {
7282                 .inode = state->inode,
7283         };
7284         int err;
7285
7286         do {
7287                 /* Cache the lock if possible... */
7288                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7289                         return 0;
7290                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7291                 if (err != -NFS4ERR_DELAY)
7292                         break;
7293                 nfs4_handle_exception(server, err, &exception);
7294         } while (exception.retry);
7295         return err;
7296 }
7297
7298 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7299 {
7300         struct nfs_server *server = NFS_SERVER(state->inode);
7301         struct nfs4_exception exception = {
7302                 .inode = state->inode,
7303         };
7304         int err;
7305
7306         err = nfs4_set_lock_state(state, request);
7307         if (err != 0)
7308                 return err;
7309         if (!recover_lost_locks) {
7310                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7311                 return 0;
7312         }
7313         do {
7314                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7315                         return 0;
7316                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7317                 switch (err) {
7318                 default:
7319                         goto out;
7320                 case -NFS4ERR_GRACE:
7321                 case -NFS4ERR_DELAY:
7322                         nfs4_handle_exception(server, err, &exception);
7323                         err = 0;
7324                 }
7325         } while (exception.retry);
7326 out:
7327         return err;
7328 }
7329
7330 #if defined(CONFIG_NFS_V4_1)
7331 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7332 {
7333         struct nfs4_lock_state *lsp;
7334         int status;
7335
7336         status = nfs4_set_lock_state(state, request);
7337         if (status != 0)
7338                 return status;
7339         lsp = request->fl_u.nfs4_fl.owner;
7340         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7341             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7342                 return 0;
7343         return nfs4_lock_expired(state, request);
7344 }
7345 #endif
7346
7347 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7348 {
7349         struct nfs_inode *nfsi = NFS_I(state->inode);
7350         struct nfs4_state_owner *sp = state->owner;
7351         unsigned char fl_flags = request->fl_flags;
7352         int status;
7353
7354         request->fl_flags |= FL_ACCESS;
7355         status = locks_lock_inode_wait(state->inode, request);
7356         if (status < 0)
7357                 goto out;
7358         mutex_lock(&sp->so_delegreturn_mutex);
7359         down_read(&nfsi->rwsem);
7360         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7361                 /* Yes: cache locks! */
7362                 /* ...but avoid races with delegation recall... */
7363                 request->fl_flags = fl_flags & ~FL_SLEEP;
7364                 status = locks_lock_inode_wait(state->inode, request);
7365                 up_read(&nfsi->rwsem);
7366                 mutex_unlock(&sp->so_delegreturn_mutex);
7367                 goto out;
7368         }
7369         up_read(&nfsi->rwsem);
7370         mutex_unlock(&sp->so_delegreturn_mutex);
7371         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7372 out:
7373         request->fl_flags = fl_flags;
7374         return status;
7375 }
7376
7377 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7378 {
7379         struct nfs4_exception exception = {
7380                 .state = state,
7381                 .inode = state->inode,
7382                 .interruptible = true,
7383         };
7384         int err;
7385
7386         do {
7387                 err = _nfs4_proc_setlk(state, cmd, request);
7388                 if (err == -NFS4ERR_DENIED)
7389                         err = -EAGAIN;
7390                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7391                                 err, &exception);
7392         } while (exception.retry);
7393         return err;
7394 }
7395
7396 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7397 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7398
7399 static int
7400 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7401                         struct file_lock *request)
7402 {
7403         int             status = -ERESTARTSYS;
7404         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7405
7406         while(!signalled()) {
7407                 status = nfs4_proc_setlk(state, cmd, request);
7408                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7409                         break;
7410                 freezable_schedule_timeout_interruptible(timeout);
7411                 timeout *= 2;
7412                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7413                 status = -ERESTARTSYS;
7414         }
7415         return status;
7416 }
7417
7418 #ifdef CONFIG_NFS_V4_1
7419 struct nfs4_lock_waiter {
7420         struct inode            *inode;
7421         struct nfs_lowner       owner;
7422         wait_queue_entry_t      wait;
7423 };
7424
7425 static int
7426 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7427 {
7428         struct nfs4_lock_waiter *waiter =
7429                 container_of(wait, struct nfs4_lock_waiter, wait);
7430
7431         /* NULL key means to wake up everyone */
7432         if (key) {
7433                 struct cb_notify_lock_args      *cbnl = key;
7434                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7435                                                 *wowner = &waiter->owner;
7436
7437                 /* Only wake if the callback was for the same owner. */
7438                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7439                         return 0;
7440
7441                 /* Make sure it's for the right inode */
7442                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7443                         return 0;
7444         }
7445
7446         return woken_wake_function(wait, mode, flags, key);
7447 }
7448
7449 static int
7450 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7451 {
7452         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7453         struct nfs_server *server = NFS_SERVER(state->inode);
7454         struct nfs_client *clp = server->nfs_client;
7455         wait_queue_head_t *q = &clp->cl_lock_waitq;
7456         struct nfs4_lock_waiter waiter = {
7457                 .inode = state->inode,
7458                 .owner = { .clientid = clp->cl_clientid,
7459                            .id = lsp->ls_seqid.owner_id,
7460                            .s_dev = server->s_dev },
7461         };
7462         int status;
7463
7464         /* Don't bother with waitqueue if we don't expect a callback */
7465         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7466                 return nfs4_retry_setlk_simple(state, cmd, request);
7467
7468         init_wait(&waiter.wait);
7469         waiter.wait.func = nfs4_wake_lock_waiter;
7470         add_wait_queue(q, &waiter.wait);
7471
7472         do {
7473                 status = nfs4_proc_setlk(state, cmd, request);
7474                 if (status != -EAGAIN || IS_SETLK(cmd))
7475                         break;
7476
7477                 status = -ERESTARTSYS;
7478                 freezer_do_not_count();
7479                 wait_woken(&waiter.wait, TASK_INTERRUPTIBLE,
7480                            NFS4_LOCK_MAXTIMEOUT);
7481                 freezer_count();
7482         } while (!signalled());
7483
7484         remove_wait_queue(q, &waiter.wait);
7485
7486         return status;
7487 }
7488 #else /* !CONFIG_NFS_V4_1 */
7489 static inline int
7490 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7491 {
7492         return nfs4_retry_setlk_simple(state, cmd, request);
7493 }
7494 #endif
7495
7496 static int
7497 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7498 {
7499         struct nfs_open_context *ctx;
7500         struct nfs4_state *state;
7501         int status;
7502
7503         /* verify open state */
7504         ctx = nfs_file_open_context(filp);
7505         state = ctx->state;
7506
7507         if (IS_GETLK(cmd)) {
7508                 if (state != NULL)
7509                         return nfs4_proc_getlk(state, F_GETLK, request);
7510                 return 0;
7511         }
7512
7513         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7514                 return -EINVAL;
7515
7516         if (request->fl_type == F_UNLCK) {
7517                 if (state != NULL)
7518                         return nfs4_proc_unlck(state, cmd, request);
7519                 return 0;
7520         }
7521
7522         if (state == NULL)
7523                 return -ENOLCK;
7524
7525         if ((request->fl_flags & FL_POSIX) &&
7526             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7527                 return -ENOLCK;
7528
7529         /*
7530          * Don't rely on the VFS having checked the file open mode,
7531          * since it won't do this for flock() locks.
7532          */
7533         switch (request->fl_type) {
7534         case F_RDLCK:
7535                 if (!(filp->f_mode & FMODE_READ))
7536                         return -EBADF;
7537                 break;
7538         case F_WRLCK:
7539                 if (!(filp->f_mode & FMODE_WRITE))
7540                         return -EBADF;
7541         }
7542
7543         status = nfs4_set_lock_state(state, request);
7544         if (status != 0)
7545                 return status;
7546
7547         return nfs4_retry_setlk(state, cmd, request);
7548 }
7549
7550 static int nfs4_delete_lease(struct file *file, void **priv)
7551 {
7552         return generic_setlease(file, F_UNLCK, NULL, priv);
7553 }
7554
7555 static int nfs4_add_lease(struct file *file, long arg, struct file_lock **lease,
7556                           void **priv)
7557 {
7558         struct inode *inode = file_inode(file);
7559         fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7560         int ret;
7561
7562         /* No delegation, no lease */
7563         if (!nfs4_have_delegation(inode, type))
7564                 return -EAGAIN;
7565         ret = generic_setlease(file, arg, lease, priv);
7566         if (ret || nfs4_have_delegation(inode, type))
7567                 return ret;
7568         /* We raced with a delegation return */
7569         nfs4_delete_lease(file, priv);
7570         return -EAGAIN;
7571 }
7572
7573 int nfs4_proc_setlease(struct file *file, long arg, struct file_lock **lease,
7574                        void **priv)
7575 {
7576         switch (arg) {
7577         case F_RDLCK:
7578         case F_WRLCK:
7579                 return nfs4_add_lease(file, arg, lease, priv);
7580         case F_UNLCK:
7581                 return nfs4_delete_lease(file, priv);
7582         default:
7583                 return -EINVAL;
7584         }
7585 }
7586
7587 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7588 {
7589         struct nfs_server *server = NFS_SERVER(state->inode);
7590         int err;
7591
7592         err = nfs4_set_lock_state(state, fl);
7593         if (err != 0)
7594                 return err;
7595         do {
7596                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7597                 if (err != -NFS4ERR_DELAY)
7598                         break;
7599                 ssleep(1);
7600         } while (err == -NFS4ERR_DELAY);
7601         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7602 }
7603
7604 struct nfs_release_lockowner_data {
7605         struct nfs4_lock_state *lsp;
7606         struct nfs_server *server;
7607         struct nfs_release_lockowner_args args;
7608         struct nfs_release_lockowner_res res;
7609         unsigned long timestamp;
7610 };
7611
7612 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7613 {
7614         struct nfs_release_lockowner_data *data = calldata;
7615         struct nfs_server *server = data->server;
7616         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7617                            &data->res.seq_res, task);
7618         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7619         data->timestamp = jiffies;
7620 }
7621
7622 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7623 {
7624         struct nfs_release_lockowner_data *data = calldata;
7625         struct nfs_server *server = data->server;
7626
7627         nfs40_sequence_done(task, &data->res.seq_res);
7628
7629         switch (task->tk_status) {
7630         case 0:
7631                 renew_lease(server, data->timestamp);
7632                 break;
7633         case -NFS4ERR_STALE_CLIENTID:
7634         case -NFS4ERR_EXPIRED:
7635                 nfs4_schedule_lease_recovery(server->nfs_client);
7636                 break;
7637         case -NFS4ERR_LEASE_MOVED:
7638         case -NFS4ERR_DELAY:
7639                 if (nfs4_async_handle_error(task, server,
7640                                             NULL, NULL) == -EAGAIN)
7641                         rpc_restart_call_prepare(task);
7642         }
7643 }
7644
7645 static void nfs4_release_lockowner_release(void *calldata)
7646 {
7647         struct nfs_release_lockowner_data *data = calldata;
7648         nfs4_free_lock_state(data->server, data->lsp);
7649         kfree(calldata);
7650 }
7651
7652 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7653         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7654         .rpc_call_done = nfs4_release_lockowner_done,
7655         .rpc_release = nfs4_release_lockowner_release,
7656 };
7657
7658 static void
7659 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7660 {
7661         struct nfs_release_lockowner_data *data;
7662         struct rpc_message msg = {
7663                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7664         };
7665
7666         if (server->nfs_client->cl_mvops->minor_version != 0)
7667                 return;
7668
7669         data = kmalloc(sizeof(*data), GFP_KERNEL);
7670         if (!data)
7671                 return;
7672         data->lsp = lsp;
7673         data->server = server;
7674         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7675         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7676         data->args.lock_owner.s_dev = server->s_dev;
7677
7678         msg.rpc_argp = &data->args;
7679         msg.rpc_resp = &data->res;
7680         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7681         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7682 }
7683
7684 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7685
7686 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7687                                    struct user_namespace *mnt_userns,
7688                                    struct dentry *unused, struct inode *inode,
7689                                    const char *key, const void *buf,
7690                                    size_t buflen, int flags)
7691 {
7692         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7693 }
7694
7695 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7696                                    struct dentry *unused, struct inode *inode,
7697                                    const char *key, void *buf, size_t buflen)
7698 {
7699         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7700 }
7701
7702 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7703 {
7704         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7705 }
7706
7707 #if defined(CONFIG_NFS_V4_1)
7708 #define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7709
7710 static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7711                                     struct user_namespace *mnt_userns,
7712                                     struct dentry *unused, struct inode *inode,
7713                                     const char *key, const void *buf,
7714                                     size_t buflen, int flags)
7715 {
7716         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7717 }
7718
7719 static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7720                                     struct dentry *unused, struct inode *inode,
7721                                     const char *key, void *buf, size_t buflen)
7722 {
7723         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7724 }
7725
7726 static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7727 {
7728         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7729 }
7730
7731 #define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7732
7733 static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7734                                     struct user_namespace *mnt_userns,
7735                                     struct dentry *unused, struct inode *inode,
7736                                     const char *key, const void *buf,
7737                                     size_t buflen, int flags)
7738 {
7739         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7740 }
7741
7742 static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7743                                     struct dentry *unused, struct inode *inode,
7744                                     const char *key, void *buf, size_t buflen)
7745 {
7746         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7747 }
7748
7749 static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7750 {
7751         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7752 }
7753
7754 #endif
7755
7756 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7757
7758 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7759                                      struct user_namespace *mnt_userns,
7760                                      struct dentry *unused, struct inode *inode,
7761                                      const char *key, const void *buf,
7762                                      size_t buflen, int flags)
7763 {
7764         if (security_ismaclabel(key))
7765                 return nfs4_set_security_label(inode, buf, buflen);
7766
7767         return -EOPNOTSUPP;
7768 }
7769
7770 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7771                                      struct dentry *unused, struct inode *inode,
7772                                      const char *key, void *buf, size_t buflen)
7773 {
7774         if (security_ismaclabel(key))
7775                 return nfs4_get_security_label(inode, buf, buflen);
7776         return -EOPNOTSUPP;
7777 }
7778
7779 static ssize_t
7780 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7781 {
7782         int len = 0;
7783
7784         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7785                 len = security_inode_listsecurity(inode, list, list_len);
7786                 if (len >= 0 && list_len && len > list_len)
7787                         return -ERANGE;
7788         }
7789         return len;
7790 }
7791
7792 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7793         .prefix = XATTR_SECURITY_PREFIX,
7794         .get    = nfs4_xattr_get_nfs4_label,
7795         .set    = nfs4_xattr_set_nfs4_label,
7796 };
7797
7798 #else
7799
7800 static ssize_t
7801 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7802 {
7803         return 0;
7804 }
7805
7806 #endif
7807
7808 #ifdef CONFIG_NFS_V4_2
7809 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7810                                     struct user_namespace *mnt_userns,
7811                                     struct dentry *unused, struct inode *inode,
7812                                     const char *key, const void *buf,
7813                                     size_t buflen, int flags)
7814 {
7815         u32 mask;
7816         int ret;
7817
7818         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7819                 return -EOPNOTSUPP;
7820
7821         /*
7822          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7823          * flags right now. Handling of xattr operations use the normal
7824          * file read/write permissions.
7825          *
7826          * Just in case the server has other ideas (which RFC 8276 allows),
7827          * do a cached access check for the XA* flags to possibly avoid
7828          * doing an RPC and getting EACCES back.
7829          */
7830         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7831                 if (!(mask & NFS_ACCESS_XAWRITE))
7832                         return -EACCES;
7833         }
7834
7835         if (buf == NULL) {
7836                 ret = nfs42_proc_removexattr(inode, key);
7837                 if (!ret)
7838                         nfs4_xattr_cache_remove(inode, key);
7839         } else {
7840                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7841                 if (!ret)
7842                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7843         }
7844
7845         return ret;
7846 }
7847
7848 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7849                                     struct dentry *unused, struct inode *inode,
7850                                     const char *key, void *buf, size_t buflen)
7851 {
7852         u32 mask;
7853         ssize_t ret;
7854
7855         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7856                 return -EOPNOTSUPP;
7857
7858         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7859                 if (!(mask & NFS_ACCESS_XAREAD))
7860                         return -EACCES;
7861         }
7862
7863         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7864         if (ret)
7865                 return ret;
7866
7867         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7868         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7869                 return ret;
7870
7871         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7872
7873         return ret;
7874 }
7875
7876 static ssize_t
7877 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7878 {
7879         u64 cookie;
7880         bool eof;
7881         ssize_t ret, size;
7882         char *buf;
7883         size_t buflen;
7884         u32 mask;
7885
7886         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7887                 return 0;
7888
7889         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7890                 if (!(mask & NFS_ACCESS_XALIST))
7891                         return 0;
7892         }
7893
7894         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7895         if (ret)
7896                 return ret;
7897
7898         ret = nfs4_xattr_cache_list(inode, list, list_len);
7899         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7900                 return ret;
7901
7902         cookie = 0;
7903         eof = false;
7904         buflen = list_len ? list_len : XATTR_LIST_MAX;
7905         buf = list_len ? list : NULL;
7906         size = 0;
7907
7908         while (!eof) {
7909                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7910                     &cookie, &eof);
7911                 if (ret < 0)
7912                         return ret;
7913
7914                 if (list_len) {
7915                         buf += ret;
7916                         buflen -= ret;
7917                 }
7918                 size += ret;
7919         }
7920
7921         if (list_len)
7922                 nfs4_xattr_cache_set_list(inode, list, size);
7923
7924         return size;
7925 }
7926
7927 #else
7928
7929 static ssize_t
7930 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7931 {
7932         return 0;
7933 }
7934 #endif /* CONFIG_NFS_V4_2 */
7935
7936 /*
7937  * nfs_fhget will use either the mounted_on_fileid or the fileid
7938  */
7939 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7940 {
7941         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7942                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7943               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7944               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7945                 return;
7946
7947         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7948                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7949         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7950         fattr->nlink = 2;
7951 }
7952
7953 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7954                                    const struct qstr *name,
7955                                    struct nfs4_fs_locations *fs_locations,
7956                                    struct page *page)
7957 {
7958         struct nfs_server *server = NFS_SERVER(dir);
7959         u32 bitmask[3];
7960         struct nfs4_fs_locations_arg args = {
7961                 .dir_fh = NFS_FH(dir),
7962                 .name = name,
7963                 .page = page,
7964                 .bitmask = bitmask,
7965         };
7966         struct nfs4_fs_locations_res res = {
7967                 .fs_locations = fs_locations,
7968         };
7969         struct rpc_message msg = {
7970                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7971                 .rpc_argp = &args,
7972                 .rpc_resp = &res,
7973         };
7974         int status;
7975
7976         dprintk("%s: start\n", __func__);
7977
7978         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7979         bitmask[1] = nfs4_fattr_bitmap[1];
7980
7981         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7982          * is not supported */
7983         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7984                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7985         else
7986                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7987
7988         nfs_fattr_init(fs_locations->fattr);
7989         fs_locations->server = server;
7990         fs_locations->nlocations = 0;
7991         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7992         dprintk("%s: returned status = %d\n", __func__, status);
7993         return status;
7994 }
7995
7996 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7997                            const struct qstr *name,
7998                            struct nfs4_fs_locations *fs_locations,
7999                            struct page *page)
8000 {
8001         struct nfs4_exception exception = {
8002                 .interruptible = true,
8003         };
8004         int err;
8005         do {
8006                 err = _nfs4_proc_fs_locations(client, dir, name,
8007                                 fs_locations, page);
8008                 trace_nfs4_get_fs_locations(dir, name, err);
8009                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8010                                 &exception);
8011         } while (exception.retry);
8012         return err;
8013 }
8014
8015 /*
8016  * This operation also signals the server that this client is
8017  * performing migration recovery.  The server can stop returning
8018  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
8019  * appended to this compound to identify the client ID which is
8020  * performing recovery.
8021  */
8022 static int _nfs40_proc_get_locations(struct nfs_server *server,
8023                                      struct nfs_fh *fhandle,
8024                                      struct nfs4_fs_locations *locations,
8025                                      struct page *page, const struct cred *cred)
8026 {
8027         struct rpc_clnt *clnt = server->client;
8028         u32 bitmask[2] = {
8029                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8030         };
8031         struct nfs4_fs_locations_arg args = {
8032                 .clientid       = server->nfs_client->cl_clientid,
8033                 .fh             = fhandle,
8034                 .page           = page,
8035                 .bitmask        = bitmask,
8036                 .migration      = 1,            /* skip LOOKUP */
8037                 .renew          = 1,            /* append RENEW */
8038         };
8039         struct nfs4_fs_locations_res res = {
8040                 .fs_locations   = locations,
8041                 .migration      = 1,
8042                 .renew          = 1,
8043         };
8044         struct rpc_message msg = {
8045                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8046                 .rpc_argp       = &args,
8047                 .rpc_resp       = &res,
8048                 .rpc_cred       = cred,
8049         };
8050         unsigned long now = jiffies;
8051         int status;
8052
8053         nfs_fattr_init(locations->fattr);
8054         locations->server = server;
8055         locations->nlocations = 0;
8056
8057         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8058         status = nfs4_call_sync_sequence(clnt, server, &msg,
8059                                         &args.seq_args, &res.seq_res);
8060         if (status)
8061                 return status;
8062
8063         renew_lease(server, now);
8064         return 0;
8065 }
8066
8067 #ifdef CONFIG_NFS_V4_1
8068
8069 /*
8070  * This operation also signals the server that this client is
8071  * performing migration recovery.  The server can stop asserting
8072  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8073  * performing this operation is identified in the SEQUENCE
8074  * operation in this compound.
8075  *
8076  * When the client supports GETATTR(fs_locations_info), it can
8077  * be plumbed in here.
8078  */
8079 static int _nfs41_proc_get_locations(struct nfs_server *server,
8080                                      struct nfs_fh *fhandle,
8081                                      struct nfs4_fs_locations *locations,
8082                                      struct page *page, const struct cred *cred)
8083 {
8084         struct rpc_clnt *clnt = server->client;
8085         u32 bitmask[2] = {
8086                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8087         };
8088         struct nfs4_fs_locations_arg args = {
8089                 .fh             = fhandle,
8090                 .page           = page,
8091                 .bitmask        = bitmask,
8092                 .migration      = 1,            /* skip LOOKUP */
8093         };
8094         struct nfs4_fs_locations_res res = {
8095                 .fs_locations   = locations,
8096                 .migration      = 1,
8097         };
8098         struct rpc_message msg = {
8099                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8100                 .rpc_argp       = &args,
8101                 .rpc_resp       = &res,
8102                 .rpc_cred       = cred,
8103         };
8104         struct nfs4_call_sync_data data = {
8105                 .seq_server = server,
8106                 .seq_args = &args.seq_args,
8107                 .seq_res = &res.seq_res,
8108         };
8109         struct rpc_task_setup task_setup_data = {
8110                 .rpc_client = clnt,
8111                 .rpc_message = &msg,
8112                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
8113                 .callback_data = &data,
8114                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8115         };
8116         int status;
8117
8118         nfs_fattr_init(locations->fattr);
8119         locations->server = server;
8120         locations->nlocations = 0;
8121
8122         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8123         status = nfs4_call_sync_custom(&task_setup_data);
8124         if (status == NFS4_OK &&
8125             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8126                 status = -NFS4ERR_LEASE_MOVED;
8127         return status;
8128 }
8129
8130 #endif  /* CONFIG_NFS_V4_1 */
8131
8132 /**
8133  * nfs4_proc_get_locations - discover locations for a migrated FSID
8134  * @server: pointer to nfs_server to process
8135  * @fhandle: pointer to the kernel NFS client file handle
8136  * @locations: result of query
8137  * @page: buffer
8138  * @cred: credential to use for this operation
8139  *
8140  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8141  * operation failed, or a negative errno if a local error occurred.
8142  *
8143  * On success, "locations" is filled in, but if the server has
8144  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8145  * asserted.
8146  *
8147  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8148  * from this client that require migration recovery.
8149  */
8150 int nfs4_proc_get_locations(struct nfs_server *server,
8151                             struct nfs_fh *fhandle,
8152                             struct nfs4_fs_locations *locations,
8153                             struct page *page, const struct cred *cred)
8154 {
8155         struct nfs_client *clp = server->nfs_client;
8156         const struct nfs4_mig_recovery_ops *ops =
8157                                         clp->cl_mvops->mig_recovery_ops;
8158         struct nfs4_exception exception = {
8159                 .interruptible = true,
8160         };
8161         int status;
8162
8163         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8164                 (unsigned long long)server->fsid.major,
8165                 (unsigned long long)server->fsid.minor,
8166                 clp->cl_hostname);
8167         nfs_display_fhandle(fhandle, __func__);
8168
8169         do {
8170                 status = ops->get_locations(server, fhandle, locations, page,
8171                                             cred);
8172                 if (status != -NFS4ERR_DELAY)
8173                         break;
8174                 nfs4_handle_exception(server, status, &exception);
8175         } while (exception.retry);
8176         return status;
8177 }
8178
8179 /*
8180  * This operation also signals the server that this client is
8181  * performing "lease moved" recovery.  The server can stop
8182  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8183  * is appended to this compound to identify the client ID which is
8184  * performing recovery.
8185  */
8186 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8187 {
8188         struct nfs_server *server = NFS_SERVER(inode);
8189         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8190         struct rpc_clnt *clnt = server->client;
8191         struct nfs4_fsid_present_arg args = {
8192                 .fh             = NFS_FH(inode),
8193                 .clientid       = clp->cl_clientid,
8194                 .renew          = 1,            /* append RENEW */
8195         };
8196         struct nfs4_fsid_present_res res = {
8197                 .renew          = 1,
8198         };
8199         struct rpc_message msg = {
8200                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8201                 .rpc_argp       = &args,
8202                 .rpc_resp       = &res,
8203                 .rpc_cred       = cred,
8204         };
8205         unsigned long now = jiffies;
8206         int status;
8207
8208         res.fh = nfs_alloc_fhandle();
8209         if (res.fh == NULL)
8210                 return -ENOMEM;
8211
8212         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8213         status = nfs4_call_sync_sequence(clnt, server, &msg,
8214                                                 &args.seq_args, &res.seq_res);
8215         nfs_free_fhandle(res.fh);
8216         if (status)
8217                 return status;
8218
8219         do_renew_lease(clp, now);
8220         return 0;
8221 }
8222
8223 #ifdef CONFIG_NFS_V4_1
8224
8225 /*
8226  * This operation also signals the server that this client is
8227  * performing "lease moved" recovery.  The server can stop asserting
8228  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8229  * this operation is identified in the SEQUENCE operation in this
8230  * compound.
8231  */
8232 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8233 {
8234         struct nfs_server *server = NFS_SERVER(inode);
8235         struct rpc_clnt *clnt = server->client;
8236         struct nfs4_fsid_present_arg args = {
8237                 .fh             = NFS_FH(inode),
8238         };
8239         struct nfs4_fsid_present_res res = {
8240         };
8241         struct rpc_message msg = {
8242                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8243                 .rpc_argp       = &args,
8244                 .rpc_resp       = &res,
8245                 .rpc_cred       = cred,
8246         };
8247         int status;
8248
8249         res.fh = nfs_alloc_fhandle();
8250         if (res.fh == NULL)
8251                 return -ENOMEM;
8252
8253         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8254         status = nfs4_call_sync_sequence(clnt, server, &msg,
8255                                                 &args.seq_args, &res.seq_res);
8256         nfs_free_fhandle(res.fh);
8257         if (status == NFS4_OK &&
8258             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8259                 status = -NFS4ERR_LEASE_MOVED;
8260         return status;
8261 }
8262
8263 #endif  /* CONFIG_NFS_V4_1 */
8264
8265 /**
8266  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8267  * @inode: inode on FSID to check
8268  * @cred: credential to use for this operation
8269  *
8270  * Server indicates whether the FSID is present, moved, or not
8271  * recognized.  This operation is necessary to clear a LEASE_MOVED
8272  * condition for this client ID.
8273  *
8274  * Returns NFS4_OK if the FSID is present on this server,
8275  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8276  *  NFS4ERR code if some error occurred on the server, or a
8277  *  negative errno if a local failure occurred.
8278  */
8279 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8280 {
8281         struct nfs_server *server = NFS_SERVER(inode);
8282         struct nfs_client *clp = server->nfs_client;
8283         const struct nfs4_mig_recovery_ops *ops =
8284                                         clp->cl_mvops->mig_recovery_ops;
8285         struct nfs4_exception exception = {
8286                 .interruptible = true,
8287         };
8288         int status;
8289
8290         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8291                 (unsigned long long)server->fsid.major,
8292                 (unsigned long long)server->fsid.minor,
8293                 clp->cl_hostname);
8294         nfs_display_fhandle(NFS_FH(inode), __func__);
8295
8296         do {
8297                 status = ops->fsid_present(inode, cred);
8298                 if (status != -NFS4ERR_DELAY)
8299                         break;
8300                 nfs4_handle_exception(server, status, &exception);
8301         } while (exception.retry);
8302         return status;
8303 }
8304
8305 /*
8306  * If 'use_integrity' is true and the state managment nfs_client
8307  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8308  * and the machine credential as per RFC3530bis and RFC5661 Security
8309  * Considerations sections. Otherwise, just use the user cred with the
8310  * filesystem's rpc_client.
8311  */
8312 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8313 {
8314         int status;
8315         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8316         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8317         struct nfs4_secinfo_arg args = {
8318                 .dir_fh = NFS_FH(dir),
8319                 .name   = name,
8320         };
8321         struct nfs4_secinfo_res res = {
8322                 .flavors     = flavors,
8323         };
8324         struct rpc_message msg = {
8325                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8326                 .rpc_argp = &args,
8327                 .rpc_resp = &res,
8328         };
8329         struct nfs4_call_sync_data data = {
8330                 .seq_server = NFS_SERVER(dir),
8331                 .seq_args = &args.seq_args,
8332                 .seq_res = &res.seq_res,
8333         };
8334         struct rpc_task_setup task_setup = {
8335                 .rpc_client = clnt,
8336                 .rpc_message = &msg,
8337                 .callback_ops = clp->cl_mvops->call_sync_ops,
8338                 .callback_data = &data,
8339                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8340         };
8341         const struct cred *cred = NULL;
8342
8343         if (use_integrity) {
8344                 clnt = clp->cl_rpcclient;
8345                 task_setup.rpc_client = clnt;
8346
8347                 cred = nfs4_get_clid_cred(clp);
8348                 msg.rpc_cred = cred;
8349         }
8350
8351         dprintk("NFS call  secinfo %s\n", name->name);
8352
8353         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8354         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8355         status = nfs4_call_sync_custom(&task_setup);
8356
8357         dprintk("NFS reply  secinfo: %d\n", status);
8358
8359         put_cred(cred);
8360         return status;
8361 }
8362
8363 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8364                       struct nfs4_secinfo_flavors *flavors)
8365 {
8366         struct nfs4_exception exception = {
8367                 .interruptible = true,
8368         };
8369         int err;
8370         do {
8371                 err = -NFS4ERR_WRONGSEC;
8372
8373                 /* try to use integrity protection with machine cred */
8374                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8375                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8376
8377                 /*
8378                  * if unable to use integrity protection, or SECINFO with
8379                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8380                  * disallowed by spec, but exists in deployed servers) use
8381                  * the current filesystem's rpc_client and the user cred.
8382                  */
8383                 if (err == -NFS4ERR_WRONGSEC)
8384                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8385
8386                 trace_nfs4_secinfo(dir, name, err);
8387                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8388                                 &exception);
8389         } while (exception.retry);
8390         return err;
8391 }
8392
8393 #ifdef CONFIG_NFS_V4_1
8394 /*
8395  * Check the exchange flags returned by the server for invalid flags, having
8396  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8397  * DS flags set.
8398  */
8399 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8400 {
8401         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8402                 goto out_inval;
8403         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8404                 goto out_inval;
8405         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8406             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8407                 goto out_inval;
8408         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8409                 goto out_inval;
8410         return NFS_OK;
8411 out_inval:
8412         return -NFS4ERR_INVAL;
8413 }
8414
8415 static bool
8416 nfs41_same_server_scope(struct nfs41_server_scope *a,
8417                         struct nfs41_server_scope *b)
8418 {
8419         if (a->server_scope_sz != b->server_scope_sz)
8420                 return false;
8421         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8422 }
8423
8424 static void
8425 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8426 {
8427         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8428         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8429         struct nfs_client *clp = args->client;
8430
8431         switch (task->tk_status) {
8432         case -NFS4ERR_BADSESSION:
8433         case -NFS4ERR_DEADSESSION:
8434                 nfs4_schedule_session_recovery(clp->cl_session,
8435                                 task->tk_status);
8436                 return;
8437         }
8438         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8439                         res->dir != NFS4_CDFS4_BOTH) {
8440                 rpc_task_close_connection(task);
8441                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8442                         rpc_restart_call(task);
8443         }
8444 }
8445
8446 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8447         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8448 };
8449
8450 /*
8451  * nfs4_proc_bind_one_conn_to_session()
8452  *
8453  * The 4.1 client currently uses the same TCP connection for the
8454  * fore and backchannel.
8455  */
8456 static
8457 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8458                 struct rpc_xprt *xprt,
8459                 struct nfs_client *clp,
8460                 const struct cred *cred)
8461 {
8462         int status;
8463         struct nfs41_bind_conn_to_session_args args = {
8464                 .client = clp,
8465                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8466                 .retries = 0,
8467         };
8468         struct nfs41_bind_conn_to_session_res res;
8469         struct rpc_message msg = {
8470                 .rpc_proc =
8471                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8472                 .rpc_argp = &args,
8473                 .rpc_resp = &res,
8474                 .rpc_cred = cred,
8475         };
8476         struct rpc_task_setup task_setup_data = {
8477                 .rpc_client = clnt,
8478                 .rpc_xprt = xprt,
8479                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8480                 .rpc_message = &msg,
8481                 .flags = RPC_TASK_TIMEOUT,
8482         };
8483         struct rpc_task *task;
8484
8485         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8486         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8487                 args.dir = NFS4_CDFC4_FORE;
8488
8489         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8490         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8491                 args.dir = NFS4_CDFC4_FORE;
8492
8493         task = rpc_run_task(&task_setup_data);
8494         if (!IS_ERR(task)) {
8495                 status = task->tk_status;
8496                 rpc_put_task(task);
8497         } else
8498                 status = PTR_ERR(task);
8499         trace_nfs4_bind_conn_to_session(clp, status);
8500         if (status == 0) {
8501                 if (memcmp(res.sessionid.data,
8502                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8503                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8504                         return -EIO;
8505                 }
8506                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8507                         dprintk("NFS: %s: Unexpected direction from server\n",
8508                                 __func__);
8509                         return -EIO;
8510                 }
8511                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8512                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8513                                 __func__);
8514                         return -EIO;
8515                 }
8516         }
8517
8518         return status;
8519 }
8520
8521 struct rpc_bind_conn_calldata {
8522         struct nfs_client *clp;
8523         const struct cred *cred;
8524 };
8525
8526 static int
8527 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8528                 struct rpc_xprt *xprt,
8529                 void *calldata)
8530 {
8531         struct rpc_bind_conn_calldata *p = calldata;
8532
8533         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8534 }
8535
8536 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8537 {
8538         struct rpc_bind_conn_calldata data = {
8539                 .clp = clp,
8540                 .cred = cred,
8541         };
8542         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8543                         nfs4_proc_bind_conn_to_session_callback, &data);
8544 }
8545
8546 /*
8547  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8548  * and operations we'd like to see to enable certain features in the allow map
8549  */
8550 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8551         .how = SP4_MACH_CRED,
8552         .enforce.u.words = {
8553                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8554                       1 << (OP_EXCHANGE_ID - 32) |
8555                       1 << (OP_CREATE_SESSION - 32) |
8556                       1 << (OP_DESTROY_SESSION - 32) |
8557                       1 << (OP_DESTROY_CLIENTID - 32)
8558         },
8559         .allow.u.words = {
8560                 [0] = 1 << (OP_CLOSE) |
8561                       1 << (OP_OPEN_DOWNGRADE) |
8562                       1 << (OP_LOCKU) |
8563                       1 << (OP_DELEGRETURN) |
8564                       1 << (OP_COMMIT),
8565                 [1] = 1 << (OP_SECINFO - 32) |
8566                       1 << (OP_SECINFO_NO_NAME - 32) |
8567                       1 << (OP_LAYOUTRETURN - 32) |
8568                       1 << (OP_TEST_STATEID - 32) |
8569                       1 << (OP_FREE_STATEID - 32) |
8570                       1 << (OP_WRITE - 32)
8571         }
8572 };
8573
8574 /*
8575  * Select the state protection mode for client `clp' given the server results
8576  * from exchange_id in `sp'.
8577  *
8578  * Returns 0 on success, negative errno otherwise.
8579  */
8580 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8581                                  struct nfs41_state_protection *sp)
8582 {
8583         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8584                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8585                       1 << (OP_EXCHANGE_ID - 32) |
8586                       1 << (OP_CREATE_SESSION - 32) |
8587                       1 << (OP_DESTROY_SESSION - 32) |
8588                       1 << (OP_DESTROY_CLIENTID - 32)
8589         };
8590         unsigned long flags = 0;
8591         unsigned int i;
8592         int ret = 0;
8593
8594         if (sp->how == SP4_MACH_CRED) {
8595                 /* Print state protect result */
8596                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8597                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8598                         if (test_bit(i, sp->enforce.u.longs))
8599                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8600                         if (test_bit(i, sp->allow.u.longs))
8601                                 dfprintk(MOUNT, "  allow op %d\n", i);
8602                 }
8603
8604                 /* make sure nothing is on enforce list that isn't supported */
8605                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8606                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8607                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8608                                 ret = -EINVAL;
8609                                 goto out;
8610                         }
8611                 }
8612
8613                 /*
8614                  * Minimal mode - state operations are allowed to use machine
8615                  * credential.  Note this already happens by default, so the
8616                  * client doesn't have to do anything more than the negotiation.
8617                  *
8618                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8619                  *       we're already using the machine cred for exchange_id
8620                  *       and will never use a different cred.
8621                  */
8622                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8623                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8624                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8625                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8626                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8627                         dfprintk(MOUNT, "  minimal mode enabled\n");
8628                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8629                 } else {
8630                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8631                         ret = -EINVAL;
8632                         goto out;
8633                 }
8634
8635                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8636                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8637                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8638                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8639                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8640                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8641                 }
8642
8643                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8644                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8645                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8646                 }
8647
8648                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8649                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8650                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8651                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8652                 }
8653
8654                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8655                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8656                         dfprintk(MOUNT, "  stateid mode enabled\n");
8657                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8658                 }
8659
8660                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8661                         dfprintk(MOUNT, "  write mode enabled\n");
8662                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8663                 }
8664
8665                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8666                         dfprintk(MOUNT, "  commit mode enabled\n");
8667                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8668                 }
8669         }
8670 out:
8671         clp->cl_sp4_flags = flags;
8672         return ret;
8673 }
8674
8675 struct nfs41_exchange_id_data {
8676         struct nfs41_exchange_id_res res;
8677         struct nfs41_exchange_id_args args;
8678 };
8679
8680 static void nfs4_exchange_id_release(void *data)
8681 {
8682         struct nfs41_exchange_id_data *cdata =
8683                                         (struct nfs41_exchange_id_data *)data;
8684
8685         nfs_put_client(cdata->args.client);
8686         kfree(cdata->res.impl_id);
8687         kfree(cdata->res.server_scope);
8688         kfree(cdata->res.server_owner);
8689         kfree(cdata);
8690 }
8691
8692 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8693         .rpc_release = nfs4_exchange_id_release,
8694 };
8695
8696 /*
8697  * _nfs4_proc_exchange_id()
8698  *
8699  * Wrapper for EXCHANGE_ID operation.
8700  */
8701 static struct rpc_task *
8702 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8703                         u32 sp4_how, struct rpc_xprt *xprt)
8704 {
8705         struct rpc_message msg = {
8706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8707                 .rpc_cred = cred,
8708         };
8709         struct rpc_task_setup task_setup_data = {
8710                 .rpc_client = clp->cl_rpcclient,
8711                 .callback_ops = &nfs4_exchange_id_call_ops,
8712                 .rpc_message = &msg,
8713                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8714         };
8715         struct nfs41_exchange_id_data *calldata;
8716         int status;
8717
8718         if (!refcount_inc_not_zero(&clp->cl_count))
8719                 return ERR_PTR(-EIO);
8720
8721         status = -ENOMEM;
8722         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8723         if (!calldata)
8724                 goto out;
8725
8726         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8727
8728         status = nfs4_init_uniform_client_string(clp);
8729         if (status)
8730                 goto out_calldata;
8731
8732         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8733                                                 GFP_NOFS);
8734         status = -ENOMEM;
8735         if (unlikely(calldata->res.server_owner == NULL))
8736                 goto out_calldata;
8737
8738         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8739                                         GFP_NOFS);
8740         if (unlikely(calldata->res.server_scope == NULL))
8741                 goto out_server_owner;
8742
8743         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8744         if (unlikely(calldata->res.impl_id == NULL))
8745                 goto out_server_scope;
8746
8747         switch (sp4_how) {
8748         case SP4_NONE:
8749                 calldata->args.state_protect.how = SP4_NONE;
8750                 break;
8751
8752         case SP4_MACH_CRED:
8753                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8754                 break;
8755
8756         default:
8757                 /* unsupported! */
8758                 WARN_ON_ONCE(1);
8759                 status = -EINVAL;
8760                 goto out_impl_id;
8761         }
8762         if (xprt) {
8763                 task_setup_data.rpc_xprt = xprt;
8764                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8765                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8766                                 sizeof(calldata->args.verifier.data));
8767         }
8768         calldata->args.client = clp;
8769         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8770         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8771 #ifdef CONFIG_NFS_V4_1_MIGRATION
8772         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8773 #endif
8774         msg.rpc_argp = &calldata->args;
8775         msg.rpc_resp = &calldata->res;
8776         task_setup_data.callback_data = calldata;
8777
8778         return rpc_run_task(&task_setup_data);
8779
8780 out_impl_id:
8781         kfree(calldata->res.impl_id);
8782 out_server_scope:
8783         kfree(calldata->res.server_scope);
8784 out_server_owner:
8785         kfree(calldata->res.server_owner);
8786 out_calldata:
8787         kfree(calldata);
8788 out:
8789         nfs_put_client(clp);
8790         return ERR_PTR(status);
8791 }
8792
8793 /*
8794  * _nfs4_proc_exchange_id()
8795  *
8796  * Wrapper for EXCHANGE_ID operation.
8797  */
8798 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8799                         u32 sp4_how)
8800 {
8801         struct rpc_task *task;
8802         struct nfs41_exchange_id_args *argp;
8803         struct nfs41_exchange_id_res *resp;
8804         unsigned long now = jiffies;
8805         int status;
8806
8807         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8808         if (IS_ERR(task))
8809                 return PTR_ERR(task);
8810
8811         argp = task->tk_msg.rpc_argp;
8812         resp = task->tk_msg.rpc_resp;
8813         status = task->tk_status;
8814         if (status  != 0)
8815                 goto out;
8816
8817         status = nfs4_check_cl_exchange_flags(resp->flags,
8818                         clp->cl_mvops->minor_version);
8819         if (status  != 0)
8820                 goto out;
8821
8822         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8823         if (status != 0)
8824                 goto out;
8825
8826         do_renew_lease(clp, now);
8827
8828         clp->cl_clientid = resp->clientid;
8829         clp->cl_exchange_flags = resp->flags;
8830         clp->cl_seqid = resp->seqid;
8831         /* Client ID is not confirmed */
8832         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8833                 clear_bit(NFS4_SESSION_ESTABLISHED,
8834                           &clp->cl_session->session_state);
8835
8836         if (clp->cl_serverscope != NULL &&
8837             !nfs41_same_server_scope(clp->cl_serverscope,
8838                                 resp->server_scope)) {
8839                 dprintk("%s: server_scope mismatch detected\n",
8840                         __func__);
8841                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8842         }
8843
8844         swap(clp->cl_serverowner, resp->server_owner);
8845         swap(clp->cl_serverscope, resp->server_scope);
8846         swap(clp->cl_implid, resp->impl_id);
8847
8848         /* Save the EXCHANGE_ID verifier session trunk tests */
8849         memcpy(clp->cl_confirm.data, argp->verifier.data,
8850                sizeof(clp->cl_confirm.data));
8851 out:
8852         trace_nfs4_exchange_id(clp, status);
8853         rpc_put_task(task);
8854         return status;
8855 }
8856
8857 /*
8858  * nfs4_proc_exchange_id()
8859  *
8860  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8861  *
8862  * Since the clientid has expired, all compounds using sessions
8863  * associated with the stale clientid will be returning
8864  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8865  * be in some phase of session reset.
8866  *
8867  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8868  */
8869 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8870 {
8871         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8872         int status;
8873
8874         /* try SP4_MACH_CRED if krb5i/p */
8875         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8876             authflavor == RPC_AUTH_GSS_KRB5P) {
8877                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8878                 if (!status)
8879                         return 0;
8880         }
8881
8882         /* try SP4_NONE */
8883         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8884 }
8885
8886 /**
8887  * nfs4_test_session_trunk
8888  *
8889  * This is an add_xprt_test() test function called from
8890  * rpc_clnt_setup_test_and_add_xprt.
8891  *
8892  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8893  * and is dereferrenced in nfs4_exchange_id_release
8894  *
8895  * Upon success, add the new transport to the rpc_clnt
8896  *
8897  * @clnt: struct rpc_clnt to get new transport
8898  * @xprt: the rpc_xprt to test
8899  * @data: call data for _nfs4_proc_exchange_id.
8900  */
8901 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8902                             void *data)
8903 {
8904         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8905         struct rpc_task *task;
8906         int status;
8907
8908         u32 sp4_how;
8909
8910         dprintk("--> %s try %s\n", __func__,
8911                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8912
8913         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8914
8915         /* Test connection for session trunking. Async exchange_id call */
8916         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8917         if (IS_ERR(task))
8918                 return;
8919
8920         status = task->tk_status;
8921         if (status == 0)
8922                 status = nfs4_detect_session_trunking(adata->clp,
8923                                 task->tk_msg.rpc_resp, xprt);
8924
8925         if (status == 0)
8926                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8927
8928         rpc_put_task(task);
8929 }
8930 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8931
8932 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8933                 const struct cred *cred)
8934 {
8935         struct rpc_message msg = {
8936                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8937                 .rpc_argp = clp,
8938                 .rpc_cred = cred,
8939         };
8940         int status;
8941
8942         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8943                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8944         trace_nfs4_destroy_clientid(clp, status);
8945         if (status)
8946                 dprintk("NFS: Got error %d from the server %s on "
8947                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8948         return status;
8949 }
8950
8951 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8952                 const struct cred *cred)
8953 {
8954         unsigned int loop;
8955         int ret;
8956
8957         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8958                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8959                 switch (ret) {
8960                 case -NFS4ERR_DELAY:
8961                 case -NFS4ERR_CLIENTID_BUSY:
8962                         ssleep(1);
8963                         break;
8964                 default:
8965                         return ret;
8966                 }
8967         }
8968         return 0;
8969 }
8970
8971 int nfs4_destroy_clientid(struct nfs_client *clp)
8972 {
8973         const struct cred *cred;
8974         int ret = 0;
8975
8976         if (clp->cl_mvops->minor_version < 1)
8977                 goto out;
8978         if (clp->cl_exchange_flags == 0)
8979                 goto out;
8980         if (clp->cl_preserve_clid)
8981                 goto out;
8982         cred = nfs4_get_clid_cred(clp);
8983         ret = nfs4_proc_destroy_clientid(clp, cred);
8984         put_cred(cred);
8985         switch (ret) {
8986         case 0:
8987         case -NFS4ERR_STALE_CLIENTID:
8988                 clp->cl_exchange_flags = 0;
8989         }
8990 out:
8991         return ret;
8992 }
8993
8994 #endif /* CONFIG_NFS_V4_1 */
8995
8996 struct nfs4_get_lease_time_data {
8997         struct nfs4_get_lease_time_args *args;
8998         struct nfs4_get_lease_time_res *res;
8999         struct nfs_client *clp;
9000 };
9001
9002 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
9003                                         void *calldata)
9004 {
9005         struct nfs4_get_lease_time_data *data =
9006                         (struct nfs4_get_lease_time_data *)calldata;
9007
9008         /* just setup sequence, do not trigger session recovery
9009            since we're invoked within one */
9010         nfs4_setup_sequence(data->clp,
9011                         &data->args->la_seq_args,
9012                         &data->res->lr_seq_res,
9013                         task);
9014 }
9015
9016 /*
9017  * Called from nfs4_state_manager thread for session setup, so don't recover
9018  * from sequence operation or clientid errors.
9019  */
9020 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9021 {
9022         struct nfs4_get_lease_time_data *data =
9023                         (struct nfs4_get_lease_time_data *)calldata;
9024
9025         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9026                 return;
9027         switch (task->tk_status) {
9028         case -NFS4ERR_DELAY:
9029         case -NFS4ERR_GRACE:
9030                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
9031                 task->tk_status = 0;
9032                 fallthrough;
9033         case -NFS4ERR_RETRY_UNCACHED_REP:
9034                 rpc_restart_call_prepare(task);
9035                 return;
9036         }
9037 }
9038
9039 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9040         .rpc_call_prepare = nfs4_get_lease_time_prepare,
9041         .rpc_call_done = nfs4_get_lease_time_done,
9042 };
9043
9044 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9045 {
9046         struct nfs4_get_lease_time_args args;
9047         struct nfs4_get_lease_time_res res = {
9048                 .lr_fsinfo = fsinfo,
9049         };
9050         struct nfs4_get_lease_time_data data = {
9051                 .args = &args,
9052                 .res = &res,
9053                 .clp = clp,
9054         };
9055         struct rpc_message msg = {
9056                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9057                 .rpc_argp = &args,
9058                 .rpc_resp = &res,
9059         };
9060         struct rpc_task_setup task_setup = {
9061                 .rpc_client = clp->cl_rpcclient,
9062                 .rpc_message = &msg,
9063                 .callback_ops = &nfs4_get_lease_time_ops,
9064                 .callback_data = &data,
9065                 .flags = RPC_TASK_TIMEOUT,
9066         };
9067
9068         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
9069         return nfs4_call_sync_custom(&task_setup);
9070 }
9071
9072 #ifdef CONFIG_NFS_V4_1
9073
9074 /*
9075  * Initialize the values to be used by the client in CREATE_SESSION
9076  * If nfs4_init_session set the fore channel request and response sizes,
9077  * use them.
9078  *
9079  * Set the back channel max_resp_sz_cached to zero to force the client to
9080  * always set csa_cachethis to FALSE because the current implementation
9081  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9082  */
9083 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9084                                     struct rpc_clnt *clnt)
9085 {
9086         unsigned int max_rqst_sz, max_resp_sz;
9087         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9088         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9089
9090         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9091         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9092
9093         /* Fore channel attributes */
9094         args->fc_attrs.max_rqst_sz = max_rqst_sz;
9095         args->fc_attrs.max_resp_sz = max_resp_sz;
9096         args->fc_attrs.max_ops = NFS4_MAX_OPS;
9097         args->fc_attrs.max_reqs = max_session_slots;
9098
9099         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9100                 "max_ops=%u max_reqs=%u\n",
9101                 __func__,
9102                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9103                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9104
9105         /* Back channel attributes */
9106         args->bc_attrs.max_rqst_sz = max_bc_payload;
9107         args->bc_attrs.max_resp_sz = max_bc_payload;
9108         args->bc_attrs.max_resp_sz_cached = 0;
9109         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9110         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9111         if (args->bc_attrs.max_reqs > max_bc_slots)
9112                 args->bc_attrs.max_reqs = max_bc_slots;
9113
9114         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9115                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9116                 __func__,
9117                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9118                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9119                 args->bc_attrs.max_reqs);
9120 }
9121
9122 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9123                 struct nfs41_create_session_res *res)
9124 {
9125         struct nfs4_channel_attrs *sent = &args->fc_attrs;
9126         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9127
9128         if (rcvd->max_resp_sz > sent->max_resp_sz)
9129                 return -EINVAL;
9130         /*
9131          * Our requested max_ops is the minimum we need; we're not
9132          * prepared to break up compounds into smaller pieces than that.
9133          * So, no point even trying to continue if the server won't
9134          * cooperate:
9135          */
9136         if (rcvd->max_ops < sent->max_ops)
9137                 return -EINVAL;
9138         if (rcvd->max_reqs == 0)
9139                 return -EINVAL;
9140         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9141                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9142         return 0;
9143 }
9144
9145 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9146                 struct nfs41_create_session_res *res)
9147 {
9148         struct nfs4_channel_attrs *sent = &args->bc_attrs;
9149         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9150
9151         if (!(res->flags & SESSION4_BACK_CHAN))
9152                 goto out;
9153         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9154                 return -EINVAL;
9155         if (rcvd->max_resp_sz < sent->max_resp_sz)
9156                 return -EINVAL;
9157         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9158                 return -EINVAL;
9159         if (rcvd->max_ops > sent->max_ops)
9160                 return -EINVAL;
9161         if (rcvd->max_reqs > sent->max_reqs)
9162                 return -EINVAL;
9163 out:
9164         return 0;
9165 }
9166
9167 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9168                                      struct nfs41_create_session_res *res)
9169 {
9170         int ret;
9171
9172         ret = nfs4_verify_fore_channel_attrs(args, res);
9173         if (ret)
9174                 return ret;
9175         return nfs4_verify_back_channel_attrs(args, res);
9176 }
9177
9178 static void nfs4_update_session(struct nfs4_session *session,
9179                 struct nfs41_create_session_res *res)
9180 {
9181         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9182         /* Mark client id and session as being confirmed */
9183         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9184         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9185         session->flags = res->flags;
9186         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9187         if (res->flags & SESSION4_BACK_CHAN)
9188                 memcpy(&session->bc_attrs, &res->bc_attrs,
9189                                 sizeof(session->bc_attrs));
9190 }
9191
9192 static int _nfs4_proc_create_session(struct nfs_client *clp,
9193                 const struct cred *cred)
9194 {
9195         struct nfs4_session *session = clp->cl_session;
9196         struct nfs41_create_session_args args = {
9197                 .client = clp,
9198                 .clientid = clp->cl_clientid,
9199                 .seqid = clp->cl_seqid,
9200                 .cb_program = NFS4_CALLBACK,
9201         };
9202         struct nfs41_create_session_res res;
9203
9204         struct rpc_message msg = {
9205                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9206                 .rpc_argp = &args,
9207                 .rpc_resp = &res,
9208                 .rpc_cred = cred,
9209         };
9210         int status;
9211
9212         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9213         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9214
9215         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9216                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9217         trace_nfs4_create_session(clp, status);
9218
9219         switch (status) {
9220         case -NFS4ERR_STALE_CLIENTID:
9221         case -NFS4ERR_DELAY:
9222         case -ETIMEDOUT:
9223         case -EACCES:
9224         case -EAGAIN:
9225                 goto out;
9226         }
9227
9228         clp->cl_seqid++;
9229         if (!status) {
9230                 /* Verify the session's negotiated channel_attrs values */
9231                 status = nfs4_verify_channel_attrs(&args, &res);
9232                 /* Increment the clientid slot sequence id */
9233                 if (status)
9234                         goto out;
9235                 nfs4_update_session(session, &res);
9236         }
9237 out:
9238         return status;
9239 }
9240
9241 /*
9242  * Issues a CREATE_SESSION operation to the server.
9243  * It is the responsibility of the caller to verify the session is
9244  * expired before calling this routine.
9245  */
9246 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9247 {
9248         int status;
9249         unsigned *ptr;
9250         struct nfs4_session *session = clp->cl_session;
9251
9252         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9253
9254         status = _nfs4_proc_create_session(clp, cred);
9255         if (status)
9256                 goto out;
9257
9258         /* Init or reset the session slot tables */
9259         status = nfs4_setup_session_slot_tables(session);
9260         dprintk("slot table setup returned %d\n", status);
9261         if (status)
9262                 goto out;
9263
9264         ptr = (unsigned *)&session->sess_id.data[0];
9265         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9266                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9267 out:
9268         return status;
9269 }
9270
9271 /*
9272  * Issue the over-the-wire RPC DESTROY_SESSION.
9273  * The caller must serialize access to this routine.
9274  */
9275 int nfs4_proc_destroy_session(struct nfs4_session *session,
9276                 const struct cred *cred)
9277 {
9278         struct rpc_message msg = {
9279                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9280                 .rpc_argp = session,
9281                 .rpc_cred = cred,
9282         };
9283         int status = 0;
9284
9285         /* session is still being setup */
9286         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9287                 return 0;
9288
9289         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9290                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9291         trace_nfs4_destroy_session(session->clp, status);
9292
9293         if (status)
9294                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9295                         "Session has been destroyed regardless...\n", status);
9296         return status;
9297 }
9298
9299 /*
9300  * Renew the cl_session lease.
9301  */
9302 struct nfs4_sequence_data {
9303         struct nfs_client *clp;
9304         struct nfs4_sequence_args args;
9305         struct nfs4_sequence_res res;
9306 };
9307
9308 static void nfs41_sequence_release(void *data)
9309 {
9310         struct nfs4_sequence_data *calldata = data;
9311         struct nfs_client *clp = calldata->clp;
9312
9313         if (refcount_read(&clp->cl_count) > 1)
9314                 nfs4_schedule_state_renewal(clp);
9315         nfs_put_client(clp);
9316         kfree(calldata);
9317 }
9318
9319 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9320 {
9321         switch(task->tk_status) {
9322         case -NFS4ERR_DELAY:
9323                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9324                 return -EAGAIN;
9325         default:
9326                 nfs4_schedule_lease_recovery(clp);
9327         }
9328         return 0;
9329 }
9330
9331 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9332 {
9333         struct nfs4_sequence_data *calldata = data;
9334         struct nfs_client *clp = calldata->clp;
9335
9336         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9337                 return;
9338
9339         trace_nfs4_sequence(clp, task->tk_status);
9340         if (task->tk_status < 0) {
9341                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9342                 if (refcount_read(&clp->cl_count) == 1)
9343                         return;
9344
9345                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9346                         rpc_restart_call_prepare(task);
9347                         return;
9348                 }
9349         }
9350         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9351 }
9352
9353 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9354 {
9355         struct nfs4_sequence_data *calldata = data;
9356         struct nfs_client *clp = calldata->clp;
9357         struct nfs4_sequence_args *args;
9358         struct nfs4_sequence_res *res;
9359
9360         args = task->tk_msg.rpc_argp;
9361         res = task->tk_msg.rpc_resp;
9362
9363         nfs4_setup_sequence(clp, args, res, task);
9364 }
9365
9366 static const struct rpc_call_ops nfs41_sequence_ops = {
9367         .rpc_call_done = nfs41_sequence_call_done,
9368         .rpc_call_prepare = nfs41_sequence_prepare,
9369         .rpc_release = nfs41_sequence_release,
9370 };
9371
9372 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9373                 const struct cred *cred,
9374                 struct nfs4_slot *slot,
9375                 bool is_privileged)
9376 {
9377         struct nfs4_sequence_data *calldata;
9378         struct rpc_message msg = {
9379                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9380                 .rpc_cred = cred,
9381         };
9382         struct rpc_task_setup task_setup_data = {
9383                 .rpc_client = clp->cl_rpcclient,
9384                 .rpc_message = &msg,
9385                 .callback_ops = &nfs41_sequence_ops,
9386                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9387         };
9388         struct rpc_task *ret;
9389
9390         ret = ERR_PTR(-EIO);
9391         if (!refcount_inc_not_zero(&clp->cl_count))
9392                 goto out_err;
9393
9394         ret = ERR_PTR(-ENOMEM);
9395         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
9396         if (calldata == NULL)
9397                 goto out_put_clp;
9398         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9399         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9400         msg.rpc_argp = &calldata->args;
9401         msg.rpc_resp = &calldata->res;
9402         calldata->clp = clp;
9403         task_setup_data.callback_data = calldata;
9404
9405         ret = rpc_run_task(&task_setup_data);
9406         if (IS_ERR(ret))
9407                 goto out_err;
9408         return ret;
9409 out_put_clp:
9410         nfs_put_client(clp);
9411 out_err:
9412         nfs41_release_slot(slot);
9413         return ret;
9414 }
9415
9416 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9417 {
9418         struct rpc_task *task;
9419         int ret = 0;
9420
9421         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9422                 return -EAGAIN;
9423         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9424         if (IS_ERR(task))
9425                 ret = PTR_ERR(task);
9426         else
9427                 rpc_put_task_async(task);
9428         dprintk("<-- %s status=%d\n", __func__, ret);
9429         return ret;
9430 }
9431
9432 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9433 {
9434         struct rpc_task *task;
9435         int ret;
9436
9437         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9438         if (IS_ERR(task)) {
9439                 ret = PTR_ERR(task);
9440                 goto out;
9441         }
9442         ret = rpc_wait_for_completion_task(task);
9443         if (!ret)
9444                 ret = task->tk_status;
9445         rpc_put_task(task);
9446 out:
9447         dprintk("<-- %s status=%d\n", __func__, ret);
9448         return ret;
9449 }
9450
9451 struct nfs4_reclaim_complete_data {
9452         struct nfs_client *clp;
9453         struct nfs41_reclaim_complete_args arg;
9454         struct nfs41_reclaim_complete_res res;
9455 };
9456
9457 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9458 {
9459         struct nfs4_reclaim_complete_data *calldata = data;
9460
9461         nfs4_setup_sequence(calldata->clp,
9462                         &calldata->arg.seq_args,
9463                         &calldata->res.seq_res,
9464                         task);
9465 }
9466
9467 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9468 {
9469         switch(task->tk_status) {
9470         case 0:
9471                 wake_up_all(&clp->cl_lock_waitq);
9472                 fallthrough;
9473         case -NFS4ERR_COMPLETE_ALREADY:
9474         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9475                 break;
9476         case -NFS4ERR_DELAY:
9477                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9478                 fallthrough;
9479         case -NFS4ERR_RETRY_UNCACHED_REP:
9480                 return -EAGAIN;
9481         case -NFS4ERR_BADSESSION:
9482         case -NFS4ERR_DEADSESSION:
9483         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9484                 break;
9485         default:
9486                 nfs4_schedule_lease_recovery(clp);
9487         }
9488         return 0;
9489 }
9490
9491 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9492 {
9493         struct nfs4_reclaim_complete_data *calldata = data;
9494         struct nfs_client *clp = calldata->clp;
9495         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9496
9497         if (!nfs41_sequence_done(task, res))
9498                 return;
9499
9500         trace_nfs4_reclaim_complete(clp, task->tk_status);
9501         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9502                 rpc_restart_call_prepare(task);
9503                 return;
9504         }
9505 }
9506
9507 static void nfs4_free_reclaim_complete_data(void *data)
9508 {
9509         struct nfs4_reclaim_complete_data *calldata = data;
9510
9511         kfree(calldata);
9512 }
9513
9514 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9515         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9516         .rpc_call_done = nfs4_reclaim_complete_done,
9517         .rpc_release = nfs4_free_reclaim_complete_data,
9518 };
9519
9520 /*
9521  * Issue a global reclaim complete.
9522  */
9523 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9524                 const struct cred *cred)
9525 {
9526         struct nfs4_reclaim_complete_data *calldata;
9527         struct rpc_message msg = {
9528                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9529                 .rpc_cred = cred,
9530         };
9531         struct rpc_task_setup task_setup_data = {
9532                 .rpc_client = clp->cl_rpcclient,
9533                 .rpc_message = &msg,
9534                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9535                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9536         };
9537         int status = -ENOMEM;
9538
9539         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9540         if (calldata == NULL)
9541                 goto out;
9542         calldata->clp = clp;
9543         calldata->arg.one_fs = 0;
9544
9545         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9546         msg.rpc_argp = &calldata->arg;
9547         msg.rpc_resp = &calldata->res;
9548         task_setup_data.callback_data = calldata;
9549         status = nfs4_call_sync_custom(&task_setup_data);
9550 out:
9551         dprintk("<-- %s status=%d\n", __func__, status);
9552         return status;
9553 }
9554
9555 static void
9556 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9557 {
9558         struct nfs4_layoutget *lgp = calldata;
9559         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9560
9561         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9562                                 &lgp->res.seq_res, task);
9563 }
9564
9565 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9566 {
9567         struct nfs4_layoutget *lgp = calldata;
9568
9569         nfs41_sequence_process(task, &lgp->res.seq_res);
9570 }
9571
9572 static int
9573 nfs4_layoutget_handle_exception(struct rpc_task *task,
9574                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9575 {
9576         struct inode *inode = lgp->args.inode;
9577         struct nfs_server *server = NFS_SERVER(inode);
9578         struct pnfs_layout_hdr *lo = lgp->lo;
9579         int nfs4err = task->tk_status;
9580         int err, status = 0;
9581         LIST_HEAD(head);
9582
9583         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9584
9585         nfs4_sequence_free_slot(&lgp->res.seq_res);
9586
9587         switch (nfs4err) {
9588         case 0:
9589                 goto out;
9590
9591         /*
9592          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9593          * on the file. set tk_status to -ENODATA to tell upper layer to
9594          * retry go inband.
9595          */
9596         case -NFS4ERR_LAYOUTUNAVAILABLE:
9597                 status = -ENODATA;
9598                 goto out;
9599         /*
9600          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9601          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9602          */
9603         case -NFS4ERR_BADLAYOUT:
9604                 status = -EOVERFLOW;
9605                 goto out;
9606         /*
9607          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9608          * (or clients) writing to the same RAID stripe except when
9609          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9610          *
9611          * Treat it like we would RECALLCONFLICT -- we retry for a little
9612          * while, and then eventually give up.
9613          */
9614         case -NFS4ERR_LAYOUTTRYLATER:
9615                 if (lgp->args.minlength == 0) {
9616                         status = -EOVERFLOW;
9617                         goto out;
9618                 }
9619                 status = -EBUSY;
9620                 break;
9621         case -NFS4ERR_RECALLCONFLICT:
9622                 status = -ERECALLCONFLICT;
9623                 break;
9624         case -NFS4ERR_DELEG_REVOKED:
9625         case -NFS4ERR_ADMIN_REVOKED:
9626         case -NFS4ERR_EXPIRED:
9627         case -NFS4ERR_BAD_STATEID:
9628                 exception->timeout = 0;
9629                 spin_lock(&inode->i_lock);
9630                 /* If the open stateid was bad, then recover it. */
9631                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9632                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9633                         spin_unlock(&inode->i_lock);
9634                         exception->state = lgp->args.ctx->state;
9635                         exception->stateid = &lgp->args.stateid;
9636                         break;
9637                 }
9638
9639                 /*
9640                  * Mark the bad layout state as invalid, then retry
9641                  */
9642                 pnfs_mark_layout_stateid_invalid(lo, &head);
9643                 spin_unlock(&inode->i_lock);
9644                 nfs_commit_inode(inode, 0);
9645                 pnfs_free_lseg_list(&head);
9646                 status = -EAGAIN;
9647                 goto out;
9648         }
9649
9650         err = nfs4_handle_exception(server, nfs4err, exception);
9651         if (!status) {
9652                 if (exception->retry)
9653                         status = -EAGAIN;
9654                 else
9655                         status = err;
9656         }
9657 out:
9658         return status;
9659 }
9660
9661 size_t max_response_pages(struct nfs_server *server)
9662 {
9663         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9664         return nfs_page_array_len(0, max_resp_sz);
9665 }
9666
9667 static void nfs4_layoutget_release(void *calldata)
9668 {
9669         struct nfs4_layoutget *lgp = calldata;
9670
9671         nfs4_sequence_free_slot(&lgp->res.seq_res);
9672         pnfs_layoutget_free(lgp);
9673 }
9674
9675 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9676         .rpc_call_prepare = nfs4_layoutget_prepare,
9677         .rpc_call_done = nfs4_layoutget_done,
9678         .rpc_release = nfs4_layoutget_release,
9679 };
9680
9681 struct pnfs_layout_segment *
9682 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9683 {
9684         struct inode *inode = lgp->args.inode;
9685         struct nfs_server *server = NFS_SERVER(inode);
9686         struct rpc_task *task;
9687         struct rpc_message msg = {
9688                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9689                 .rpc_argp = &lgp->args,
9690                 .rpc_resp = &lgp->res,
9691                 .rpc_cred = lgp->cred,
9692         };
9693         struct rpc_task_setup task_setup_data = {
9694                 .rpc_client = server->client,
9695                 .rpc_message = &msg,
9696                 .callback_ops = &nfs4_layoutget_call_ops,
9697                 .callback_data = lgp,
9698                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9699                          RPC_TASK_MOVEABLE,
9700         };
9701         struct pnfs_layout_segment *lseg = NULL;
9702         struct nfs4_exception exception = {
9703                 .inode = inode,
9704                 .timeout = *timeout,
9705         };
9706         int status = 0;
9707
9708         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9709
9710         task = rpc_run_task(&task_setup_data);
9711         if (IS_ERR(task))
9712                 return ERR_CAST(task);
9713
9714         status = rpc_wait_for_completion_task(task);
9715         if (status != 0)
9716                 goto out;
9717
9718         if (task->tk_status < 0) {
9719                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9720                 *timeout = exception.timeout;
9721         } else if (lgp->res.layoutp->len == 0) {
9722                 status = -EAGAIN;
9723                 *timeout = nfs4_update_delay(&exception.timeout);
9724         } else
9725                 lseg = pnfs_layout_process(lgp);
9726 out:
9727         trace_nfs4_layoutget(lgp->args.ctx,
9728                         &lgp->args.range,
9729                         &lgp->res.range,
9730                         &lgp->res.stateid,
9731                         status);
9732
9733         rpc_put_task(task);
9734         dprintk("<-- %s status=%d\n", __func__, status);
9735         if (status)
9736                 return ERR_PTR(status);
9737         return lseg;
9738 }
9739
9740 static void
9741 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9742 {
9743         struct nfs4_layoutreturn *lrp = calldata;
9744
9745         nfs4_setup_sequence(lrp->clp,
9746                         &lrp->args.seq_args,
9747                         &lrp->res.seq_res,
9748                         task);
9749         if (!pnfs_layout_is_valid(lrp->args.layout))
9750                 rpc_exit(task, 0);
9751 }
9752
9753 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9754 {
9755         struct nfs4_layoutreturn *lrp = calldata;
9756         struct nfs_server *server;
9757
9758         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9759                 return;
9760
9761         /*
9762          * Was there an RPC level error? Assume the call succeeded,
9763          * and that we need to release the layout
9764          */
9765         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9766                 lrp->res.lrs_present = 0;
9767                 return;
9768         }
9769
9770         server = NFS_SERVER(lrp->args.inode);
9771         switch (task->tk_status) {
9772         case -NFS4ERR_OLD_STATEID:
9773                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9774                                         &lrp->args.range,
9775                                         lrp->args.inode))
9776                         goto out_restart;
9777                 fallthrough;
9778         default:
9779                 task->tk_status = 0;
9780                 fallthrough;
9781         case 0:
9782                 break;
9783         case -NFS4ERR_DELAY:
9784                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9785                         break;
9786                 goto out_restart;
9787         }
9788         return;
9789 out_restart:
9790         task->tk_status = 0;
9791         nfs4_sequence_free_slot(&lrp->res.seq_res);
9792         rpc_restart_call_prepare(task);
9793 }
9794
9795 static void nfs4_layoutreturn_release(void *calldata)
9796 {
9797         struct nfs4_layoutreturn *lrp = calldata;
9798         struct pnfs_layout_hdr *lo = lrp->args.layout;
9799
9800         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9801                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9802         nfs4_sequence_free_slot(&lrp->res.seq_res);
9803         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9804                 lrp->ld_private.ops->free(&lrp->ld_private);
9805         pnfs_put_layout_hdr(lrp->args.layout);
9806         nfs_iput_and_deactive(lrp->inode);
9807         put_cred(lrp->cred);
9808         kfree(calldata);
9809 }
9810
9811 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9812         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9813         .rpc_call_done = nfs4_layoutreturn_done,
9814         .rpc_release = nfs4_layoutreturn_release,
9815 };
9816
9817 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9818 {
9819         struct rpc_task *task;
9820         struct rpc_message msg = {
9821                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9822                 .rpc_argp = &lrp->args,
9823                 .rpc_resp = &lrp->res,
9824                 .rpc_cred = lrp->cred,
9825         };
9826         struct rpc_task_setup task_setup_data = {
9827                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9828                 .rpc_message = &msg,
9829                 .callback_ops = &nfs4_layoutreturn_call_ops,
9830                 .callback_data = lrp,
9831                 .flags = RPC_TASK_MOVEABLE,
9832         };
9833         int status = 0;
9834
9835         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9836                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9837                         &task_setup_data.rpc_client, &msg);
9838
9839         lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9840         if (!sync) {
9841                 if (!lrp->inode) {
9842                         nfs4_layoutreturn_release(lrp);
9843                         return -EAGAIN;
9844                 }
9845                 task_setup_data.flags |= RPC_TASK_ASYNC;
9846         }
9847         if (!lrp->inode)
9848                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9849                                    1);
9850         else
9851                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9852                                    0);
9853         task = rpc_run_task(&task_setup_data);
9854         if (IS_ERR(task))
9855                 return PTR_ERR(task);
9856         if (sync)
9857                 status = task->tk_status;
9858         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9859         dprintk("<-- %s status=%d\n", __func__, status);
9860         rpc_put_task(task);
9861         return status;
9862 }
9863
9864 static int
9865 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9866                 struct pnfs_device *pdev,
9867                 const struct cred *cred)
9868 {
9869         struct nfs4_getdeviceinfo_args args = {
9870                 .pdev = pdev,
9871                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9872                         NOTIFY_DEVICEID4_DELETE,
9873         };
9874         struct nfs4_getdeviceinfo_res res = {
9875                 .pdev = pdev,
9876         };
9877         struct rpc_message msg = {
9878                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9879                 .rpc_argp = &args,
9880                 .rpc_resp = &res,
9881                 .rpc_cred = cred,
9882         };
9883         int status;
9884
9885         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9886         if (res.notification & ~args.notify_types)
9887                 dprintk("%s: unsupported notification\n", __func__);
9888         if (res.notification != args.notify_types)
9889                 pdev->nocache = 1;
9890
9891         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9892
9893         dprintk("<-- %s status=%d\n", __func__, status);
9894
9895         return status;
9896 }
9897
9898 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9899                 struct pnfs_device *pdev,
9900                 const struct cred *cred)
9901 {
9902         struct nfs4_exception exception = { };
9903         int err;
9904
9905         do {
9906                 err = nfs4_handle_exception(server,
9907                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9908                                         &exception);
9909         } while (exception.retry);
9910         return err;
9911 }
9912 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9913
9914 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9915 {
9916         struct nfs4_layoutcommit_data *data = calldata;
9917         struct nfs_server *server = NFS_SERVER(data->args.inode);
9918
9919         nfs4_setup_sequence(server->nfs_client,
9920                         &data->args.seq_args,
9921                         &data->res.seq_res,
9922                         task);
9923 }
9924
9925 static void
9926 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9927 {
9928         struct nfs4_layoutcommit_data *data = calldata;
9929         struct nfs_server *server = NFS_SERVER(data->args.inode);
9930
9931         if (!nfs41_sequence_done(task, &data->res.seq_res))
9932                 return;
9933
9934         switch (task->tk_status) { /* Just ignore these failures */
9935         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9936         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9937         case -NFS4ERR_BADLAYOUT:     /* no layout */
9938         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9939                 task->tk_status = 0;
9940                 break;
9941         case 0:
9942                 break;
9943         default:
9944                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9945                         rpc_restart_call_prepare(task);
9946                         return;
9947                 }
9948         }
9949 }
9950
9951 static void nfs4_layoutcommit_release(void *calldata)
9952 {
9953         struct nfs4_layoutcommit_data *data = calldata;
9954
9955         pnfs_cleanup_layoutcommit(data);
9956         nfs_post_op_update_inode_force_wcc(data->args.inode,
9957                                            data->res.fattr);
9958         put_cred(data->cred);
9959         nfs_iput_and_deactive(data->inode);
9960         kfree(data);
9961 }
9962
9963 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9964         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9965         .rpc_call_done = nfs4_layoutcommit_done,
9966         .rpc_release = nfs4_layoutcommit_release,
9967 };
9968
9969 int
9970 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9971 {
9972         struct rpc_message msg = {
9973                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9974                 .rpc_argp = &data->args,
9975                 .rpc_resp = &data->res,
9976                 .rpc_cred = data->cred,
9977         };
9978         struct rpc_task_setup task_setup_data = {
9979                 .task = &data->task,
9980                 .rpc_client = NFS_CLIENT(data->args.inode),
9981                 .rpc_message = &msg,
9982                 .callback_ops = &nfs4_layoutcommit_ops,
9983                 .callback_data = data,
9984                 .flags = RPC_TASK_MOVEABLE,
9985         };
9986         struct rpc_task *task;
9987         int status = 0;
9988
9989         dprintk("NFS: initiating layoutcommit call. sync %d "
9990                 "lbw: %llu inode %lu\n", sync,
9991                 data->args.lastbytewritten,
9992                 data->args.inode->i_ino);
9993
9994         if (!sync) {
9995                 data->inode = nfs_igrab_and_active(data->args.inode);
9996                 if (data->inode == NULL) {
9997                         nfs4_layoutcommit_release(data);
9998                         return -EAGAIN;
9999                 }
10000                 task_setup_data.flags = RPC_TASK_ASYNC;
10001         }
10002         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
10003         task = rpc_run_task(&task_setup_data);
10004         if (IS_ERR(task))
10005                 return PTR_ERR(task);
10006         if (sync)
10007                 status = task->tk_status;
10008         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10009         dprintk("%s: status %d\n", __func__, status);
10010         rpc_put_task(task);
10011         return status;
10012 }
10013
10014 /*
10015  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10016  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10017  */
10018 static int
10019 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10020                     struct nfs_fsinfo *info,
10021                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
10022 {
10023         struct nfs41_secinfo_no_name_args args = {
10024                 .style = SECINFO_STYLE_CURRENT_FH,
10025         };
10026         struct nfs4_secinfo_res res = {
10027                 .flavors = flavors,
10028         };
10029         struct rpc_message msg = {
10030                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10031                 .rpc_argp = &args,
10032                 .rpc_resp = &res,
10033         };
10034         struct nfs4_call_sync_data data = {
10035                 .seq_server = server,
10036                 .seq_args = &args.seq_args,
10037                 .seq_res = &res.seq_res,
10038         };
10039         struct rpc_task_setup task_setup = {
10040                 .rpc_client = server->client,
10041                 .rpc_message = &msg,
10042                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
10043                 .callback_data = &data,
10044                 .flags = RPC_TASK_NO_ROUND_ROBIN,
10045         };
10046         const struct cred *cred = NULL;
10047         int status;
10048
10049         if (use_integrity) {
10050                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
10051
10052                 cred = nfs4_get_clid_cred(server->nfs_client);
10053                 msg.rpc_cred = cred;
10054         }
10055
10056         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
10057         status = nfs4_call_sync_custom(&task_setup);
10058         dprintk("<-- %s status=%d\n", __func__, status);
10059
10060         put_cred(cred);
10061
10062         return status;
10063 }
10064
10065 static int
10066 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10067                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
10068 {
10069         struct nfs4_exception exception = {
10070                 .interruptible = true,
10071         };
10072         int err;
10073         do {
10074                 /* first try using integrity protection */
10075                 err = -NFS4ERR_WRONGSEC;
10076
10077                 /* try to use integrity protection with machine cred */
10078                 if (_nfs4_is_integrity_protected(server->nfs_client))
10079                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10080                                                           flavors, true);
10081
10082                 /*
10083                  * if unable to use integrity protection, or SECINFO with
10084                  * integrity protection returns NFS4ERR_WRONGSEC (which is
10085                  * disallowed by spec, but exists in deployed servers) use
10086                  * the current filesystem's rpc_client and the user cred.
10087                  */
10088                 if (err == -NFS4ERR_WRONGSEC)
10089                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10090                                                           flavors, false);
10091
10092                 switch (err) {
10093                 case 0:
10094                 case -NFS4ERR_WRONGSEC:
10095                 case -ENOTSUPP:
10096                         goto out;
10097                 default:
10098                         err = nfs4_handle_exception(server, err, &exception);
10099                 }
10100         } while (exception.retry);
10101 out:
10102         return err;
10103 }
10104
10105 static int
10106 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10107                     struct nfs_fsinfo *info)
10108 {
10109         int err;
10110         struct page *page;
10111         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10112         struct nfs4_secinfo_flavors *flavors;
10113         struct nfs4_secinfo4 *secinfo;
10114         int i;
10115
10116         page = alloc_page(GFP_KERNEL);
10117         if (!page) {
10118                 err = -ENOMEM;
10119                 goto out;
10120         }
10121
10122         flavors = page_address(page);
10123         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10124
10125         /*
10126          * Fall back on "guess and check" method if
10127          * the server doesn't support SECINFO_NO_NAME
10128          */
10129         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10130                 err = nfs4_find_root_sec(server, fhandle, info);
10131                 goto out_freepage;
10132         }
10133         if (err)
10134                 goto out_freepage;
10135
10136         for (i = 0; i < flavors->num_flavors; i++) {
10137                 secinfo = &flavors->flavors[i];
10138
10139                 switch (secinfo->flavor) {
10140                 case RPC_AUTH_NULL:
10141                 case RPC_AUTH_UNIX:
10142                 case RPC_AUTH_GSS:
10143                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10144                                         &secinfo->flavor_info);
10145                         break;
10146                 default:
10147                         flavor = RPC_AUTH_MAXFLAVOR;
10148                         break;
10149                 }
10150
10151                 if (!nfs_auth_info_match(&server->auth_info, flavor))
10152                         flavor = RPC_AUTH_MAXFLAVOR;
10153
10154                 if (flavor != RPC_AUTH_MAXFLAVOR) {
10155                         err = nfs4_lookup_root_sec(server, fhandle,
10156                                                    info, flavor);
10157                         if (!err)
10158                                 break;
10159                 }
10160         }
10161
10162         if (flavor == RPC_AUTH_MAXFLAVOR)
10163                 err = -EPERM;
10164
10165 out_freepage:
10166         put_page(page);
10167         if (err == -EACCES)
10168                 return -EPERM;
10169 out:
10170         return err;
10171 }
10172
10173 static int _nfs41_test_stateid(struct nfs_server *server,
10174                 nfs4_stateid *stateid,
10175                 const struct cred *cred)
10176 {
10177         int status;
10178         struct nfs41_test_stateid_args args = {
10179                 .stateid = stateid,
10180         };
10181         struct nfs41_test_stateid_res res;
10182         struct rpc_message msg = {
10183                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10184                 .rpc_argp = &args,
10185                 .rpc_resp = &res,
10186                 .rpc_cred = cred,
10187         };
10188         struct rpc_clnt *rpc_client = server->client;
10189
10190         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10191                 &rpc_client, &msg);
10192
10193         dprintk("NFS call  test_stateid %p\n", stateid);
10194         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10195         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10196                         &args.seq_args, &res.seq_res);
10197         if (status != NFS_OK) {
10198                 dprintk("NFS reply test_stateid: failed, %d\n", status);
10199                 return status;
10200         }
10201         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10202         return -res.status;
10203 }
10204
10205 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10206                 int err, struct nfs4_exception *exception)
10207 {
10208         exception->retry = 0;
10209         switch(err) {
10210         case -NFS4ERR_DELAY:
10211         case -NFS4ERR_RETRY_UNCACHED_REP:
10212                 nfs4_handle_exception(server, err, exception);
10213                 break;
10214         case -NFS4ERR_BADSESSION:
10215         case -NFS4ERR_BADSLOT:
10216         case -NFS4ERR_BAD_HIGH_SLOT:
10217         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10218         case -NFS4ERR_DEADSESSION:
10219                 nfs4_do_handle_exception(server, err, exception);
10220         }
10221 }
10222
10223 /**
10224  * nfs41_test_stateid - perform a TEST_STATEID operation
10225  *
10226  * @server: server / transport on which to perform the operation
10227  * @stateid: state ID to test
10228  * @cred: credential
10229  *
10230  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10231  * Otherwise a negative NFS4ERR value is returned if the operation
10232  * failed or the state ID is not currently valid.
10233  */
10234 static int nfs41_test_stateid(struct nfs_server *server,
10235                 nfs4_stateid *stateid,
10236                 const struct cred *cred)
10237 {
10238         struct nfs4_exception exception = {
10239                 .interruptible = true,
10240         };
10241         int err;
10242         do {
10243                 err = _nfs41_test_stateid(server, stateid, cred);
10244                 nfs4_handle_delay_or_session_error(server, err, &exception);
10245         } while (exception.retry);
10246         return err;
10247 }
10248
10249 struct nfs_free_stateid_data {
10250         struct nfs_server *server;
10251         struct nfs41_free_stateid_args args;
10252         struct nfs41_free_stateid_res res;
10253 };
10254
10255 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10256 {
10257         struct nfs_free_stateid_data *data = calldata;
10258         nfs4_setup_sequence(data->server->nfs_client,
10259                         &data->args.seq_args,
10260                         &data->res.seq_res,
10261                         task);
10262 }
10263
10264 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10265 {
10266         struct nfs_free_stateid_data *data = calldata;
10267
10268         nfs41_sequence_done(task, &data->res.seq_res);
10269
10270         switch (task->tk_status) {
10271         case -NFS4ERR_DELAY:
10272                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10273                         rpc_restart_call_prepare(task);
10274         }
10275 }
10276
10277 static void nfs41_free_stateid_release(void *calldata)
10278 {
10279         struct nfs_free_stateid_data *data = calldata;
10280         struct nfs_client *clp = data->server->nfs_client;
10281
10282         nfs_put_client(clp);
10283         kfree(calldata);
10284 }
10285
10286 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10287         .rpc_call_prepare = nfs41_free_stateid_prepare,
10288         .rpc_call_done = nfs41_free_stateid_done,
10289         .rpc_release = nfs41_free_stateid_release,
10290 };
10291
10292 /**
10293  * nfs41_free_stateid - perform a FREE_STATEID operation
10294  *
10295  * @server: server / transport on which to perform the operation
10296  * @stateid: state ID to release
10297  * @cred: credential
10298  * @privileged: set to true if this call needs to be privileged
10299  *
10300  * Note: this function is always asynchronous.
10301  */
10302 static int nfs41_free_stateid(struct nfs_server *server,
10303                 const nfs4_stateid *stateid,
10304                 const struct cred *cred,
10305                 bool privileged)
10306 {
10307         struct rpc_message msg = {
10308                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10309                 .rpc_cred = cred,
10310         };
10311         struct rpc_task_setup task_setup = {
10312                 .rpc_client = server->client,
10313                 .rpc_message = &msg,
10314                 .callback_ops = &nfs41_free_stateid_ops,
10315                 .flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10316         };
10317         struct nfs_free_stateid_data *data;
10318         struct rpc_task *task;
10319         struct nfs_client *clp = server->nfs_client;
10320
10321         if (!refcount_inc_not_zero(&clp->cl_count))
10322                 return -EIO;
10323
10324         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10325                 &task_setup.rpc_client, &msg);
10326
10327         dprintk("NFS call  free_stateid %p\n", stateid);
10328         data = kmalloc(sizeof(*data), GFP_KERNEL);
10329         if (!data)
10330                 return -ENOMEM;
10331         data->server = server;
10332         nfs4_stateid_copy(&data->args.stateid, stateid);
10333
10334         task_setup.callback_data = data;
10335
10336         msg.rpc_argp = &data->args;
10337         msg.rpc_resp = &data->res;
10338         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10339         task = rpc_run_task(&task_setup);
10340         if (IS_ERR(task))
10341                 return PTR_ERR(task);
10342         rpc_put_task(task);
10343         return 0;
10344 }
10345
10346 static void
10347 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10348 {
10349         const struct cred *cred = lsp->ls_state->owner->so_cred;
10350
10351         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10352         nfs4_free_lock_state(server, lsp);
10353 }
10354
10355 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10356                 const nfs4_stateid *s2)
10357 {
10358         if (s1->type != s2->type)
10359                 return false;
10360
10361         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10362                 return false;
10363
10364         if (s1->seqid == s2->seqid)
10365                 return true;
10366
10367         return s1->seqid == 0 || s2->seqid == 0;
10368 }
10369
10370 #endif /* CONFIG_NFS_V4_1 */
10371
10372 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10373                 const nfs4_stateid *s2)
10374 {
10375         return nfs4_stateid_match(s1, s2);
10376 }
10377
10378
10379 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10380         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10381         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10382         .recover_open   = nfs4_open_reclaim,
10383         .recover_lock   = nfs4_lock_reclaim,
10384         .establish_clid = nfs4_init_clientid,
10385         .detect_trunking = nfs40_discover_server_trunking,
10386 };
10387
10388 #if defined(CONFIG_NFS_V4_1)
10389 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10390         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10391         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10392         .recover_open   = nfs4_open_reclaim,
10393         .recover_lock   = nfs4_lock_reclaim,
10394         .establish_clid = nfs41_init_clientid,
10395         .reclaim_complete = nfs41_proc_reclaim_complete,
10396         .detect_trunking = nfs41_discover_server_trunking,
10397 };
10398 #endif /* CONFIG_NFS_V4_1 */
10399
10400 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10401         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10402         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10403         .recover_open   = nfs40_open_expired,
10404         .recover_lock   = nfs4_lock_expired,
10405         .establish_clid = nfs4_init_clientid,
10406 };
10407
10408 #if defined(CONFIG_NFS_V4_1)
10409 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10410         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10411         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10412         .recover_open   = nfs41_open_expired,
10413         .recover_lock   = nfs41_lock_expired,
10414         .establish_clid = nfs41_init_clientid,
10415 };
10416 #endif /* CONFIG_NFS_V4_1 */
10417
10418 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10419         .sched_state_renewal = nfs4_proc_async_renew,
10420         .get_state_renewal_cred = nfs4_get_renew_cred,
10421         .renew_lease = nfs4_proc_renew,
10422 };
10423
10424 #if defined(CONFIG_NFS_V4_1)
10425 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10426         .sched_state_renewal = nfs41_proc_async_sequence,
10427         .get_state_renewal_cred = nfs4_get_machine_cred,
10428         .renew_lease = nfs4_proc_sequence,
10429 };
10430 #endif
10431
10432 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10433         .get_locations = _nfs40_proc_get_locations,
10434         .fsid_present = _nfs40_proc_fsid_present,
10435 };
10436
10437 #if defined(CONFIG_NFS_V4_1)
10438 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10439         .get_locations = _nfs41_proc_get_locations,
10440         .fsid_present = _nfs41_proc_fsid_present,
10441 };
10442 #endif  /* CONFIG_NFS_V4_1 */
10443
10444 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10445         .minor_version = 0,
10446         .init_caps = NFS_CAP_READDIRPLUS
10447                 | NFS_CAP_ATOMIC_OPEN
10448                 | NFS_CAP_POSIX_LOCK,
10449         .init_client = nfs40_init_client,
10450         .shutdown_client = nfs40_shutdown_client,
10451         .match_stateid = nfs4_match_stateid,
10452         .find_root_sec = nfs4_find_root_sec,
10453         .free_lock_state = nfs4_release_lockowner,
10454         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10455         .alloc_seqid = nfs_alloc_seqid,
10456         .call_sync_ops = &nfs40_call_sync_ops,
10457         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10458         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10459         .state_renewal_ops = &nfs40_state_renewal_ops,
10460         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10461 };
10462
10463 #if defined(CONFIG_NFS_V4_1)
10464 static struct nfs_seqid *
10465 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10466 {
10467         return NULL;
10468 }
10469
10470 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10471         .minor_version = 1,
10472         .init_caps = NFS_CAP_READDIRPLUS
10473                 | NFS_CAP_ATOMIC_OPEN
10474                 | NFS_CAP_POSIX_LOCK
10475                 | NFS_CAP_STATEID_NFSV41
10476                 | NFS_CAP_ATOMIC_OPEN_V1
10477                 | NFS_CAP_LGOPEN
10478                 | NFS_CAP_MOVEABLE,
10479         .init_client = nfs41_init_client,
10480         .shutdown_client = nfs41_shutdown_client,
10481         .match_stateid = nfs41_match_stateid,
10482         .find_root_sec = nfs41_find_root_sec,
10483         .free_lock_state = nfs41_free_lock_state,
10484         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10485         .alloc_seqid = nfs_alloc_no_seqid,
10486         .session_trunk = nfs4_test_session_trunk,
10487         .call_sync_ops = &nfs41_call_sync_ops,
10488         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10489         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10490         .state_renewal_ops = &nfs41_state_renewal_ops,
10491         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10492 };
10493 #endif
10494
10495 #if defined(CONFIG_NFS_V4_2)
10496 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10497         .minor_version = 2,
10498         .init_caps = NFS_CAP_READDIRPLUS
10499                 | NFS_CAP_ATOMIC_OPEN
10500                 | NFS_CAP_POSIX_LOCK
10501                 | NFS_CAP_STATEID_NFSV41
10502                 | NFS_CAP_ATOMIC_OPEN_V1
10503                 | NFS_CAP_LGOPEN
10504                 | NFS_CAP_ALLOCATE
10505                 | NFS_CAP_COPY
10506                 | NFS_CAP_OFFLOAD_CANCEL
10507                 | NFS_CAP_COPY_NOTIFY
10508                 | NFS_CAP_DEALLOCATE
10509                 | NFS_CAP_SEEK
10510                 | NFS_CAP_LAYOUTSTATS
10511                 | NFS_CAP_CLONE
10512                 | NFS_CAP_LAYOUTERROR
10513                 | NFS_CAP_READ_PLUS
10514                 | NFS_CAP_MOVEABLE,
10515         .init_client = nfs41_init_client,
10516         .shutdown_client = nfs41_shutdown_client,
10517         .match_stateid = nfs41_match_stateid,
10518         .find_root_sec = nfs41_find_root_sec,
10519         .free_lock_state = nfs41_free_lock_state,
10520         .call_sync_ops = &nfs41_call_sync_ops,
10521         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10522         .alloc_seqid = nfs_alloc_no_seqid,
10523         .session_trunk = nfs4_test_session_trunk,
10524         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10525         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10526         .state_renewal_ops = &nfs41_state_renewal_ops,
10527         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10528 };
10529 #endif
10530
10531 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10532         [0] = &nfs_v4_0_minor_ops,
10533 #if defined(CONFIG_NFS_V4_1)
10534         [1] = &nfs_v4_1_minor_ops,
10535 #endif
10536 #if defined(CONFIG_NFS_V4_2)
10537         [2] = &nfs_v4_2_minor_ops,
10538 #endif
10539 };
10540
10541 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10542 {
10543         ssize_t error, error2, error3;
10544
10545         error = generic_listxattr(dentry, list, size);
10546         if (error < 0)
10547                 return error;
10548         if (list) {
10549                 list += error;
10550                 size -= error;
10551         }
10552
10553         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10554         if (error2 < 0)
10555                 return error2;
10556
10557         if (list) {
10558                 list += error2;
10559                 size -= error2;
10560         }
10561
10562         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10563         if (error3 < 0)
10564                 return error3;
10565
10566         return error + error2 + error3;
10567 }
10568
10569 static void nfs4_enable_swap(struct inode *inode)
10570 {
10571         /* The state manager thread must always be running.
10572          * It will notice the client is a swapper, and stay put.
10573          */
10574         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10575
10576         nfs4_schedule_state_manager(clp);
10577 }
10578
10579 static void nfs4_disable_swap(struct inode *inode)
10580 {
10581         /* The state manager thread will now exit once it is
10582          * woken.
10583          */
10584         wake_up_var(&NFS_SERVER(inode)->nfs_client->cl_state);
10585 }
10586
10587 static const struct inode_operations nfs4_dir_inode_operations = {
10588         .create         = nfs_create,
10589         .lookup         = nfs_lookup,
10590         .atomic_open    = nfs_atomic_open,
10591         .link           = nfs_link,
10592         .unlink         = nfs_unlink,
10593         .symlink        = nfs_symlink,
10594         .mkdir          = nfs_mkdir,
10595         .rmdir          = nfs_rmdir,
10596         .mknod          = nfs_mknod,
10597         .rename         = nfs_rename,
10598         .permission     = nfs_permission,
10599         .getattr        = nfs_getattr,
10600         .setattr        = nfs_setattr,
10601         .listxattr      = nfs4_listxattr,
10602 };
10603
10604 static const struct inode_operations nfs4_file_inode_operations = {
10605         .permission     = nfs_permission,
10606         .getattr        = nfs_getattr,
10607         .setattr        = nfs_setattr,
10608         .listxattr      = nfs4_listxattr,
10609 };
10610
10611 const struct nfs_rpc_ops nfs_v4_clientops = {
10612         .version        = 4,                    /* protocol version */
10613         .dentry_ops     = &nfs4_dentry_operations,
10614         .dir_inode_ops  = &nfs4_dir_inode_operations,
10615         .file_inode_ops = &nfs4_file_inode_operations,
10616         .file_ops       = &nfs4_file_operations,
10617         .getroot        = nfs4_proc_get_root,
10618         .submount       = nfs4_submount,
10619         .try_get_tree   = nfs4_try_get_tree,
10620         .getattr        = nfs4_proc_getattr,
10621         .setattr        = nfs4_proc_setattr,
10622         .lookup         = nfs4_proc_lookup,
10623         .lookupp        = nfs4_proc_lookupp,
10624         .access         = nfs4_proc_access,
10625         .readlink       = nfs4_proc_readlink,
10626         .create         = nfs4_proc_create,
10627         .remove         = nfs4_proc_remove,
10628         .unlink_setup   = nfs4_proc_unlink_setup,
10629         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10630         .unlink_done    = nfs4_proc_unlink_done,
10631         .rename_setup   = nfs4_proc_rename_setup,
10632         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10633         .rename_done    = nfs4_proc_rename_done,
10634         .link           = nfs4_proc_link,
10635         .symlink        = nfs4_proc_symlink,
10636         .mkdir          = nfs4_proc_mkdir,
10637         .rmdir          = nfs4_proc_rmdir,
10638         .readdir        = nfs4_proc_readdir,
10639         .mknod          = nfs4_proc_mknod,
10640         .statfs         = nfs4_proc_statfs,
10641         .fsinfo         = nfs4_proc_fsinfo,
10642         .pathconf       = nfs4_proc_pathconf,
10643         .set_capabilities = nfs4_server_capabilities,
10644         .decode_dirent  = nfs4_decode_dirent,
10645         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10646         .read_setup     = nfs4_proc_read_setup,
10647         .read_done      = nfs4_read_done,
10648         .write_setup    = nfs4_proc_write_setup,
10649         .write_done     = nfs4_write_done,
10650         .commit_setup   = nfs4_proc_commit_setup,
10651         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10652         .commit_done    = nfs4_commit_done,
10653         .lock           = nfs4_proc_lock,
10654         .clear_acl_cache = nfs4_zap_acl_attr,
10655         .close_context  = nfs4_close_context,
10656         .open_context   = nfs4_atomic_open,
10657         .have_delegation = nfs4_have_delegation,
10658         .alloc_client   = nfs4_alloc_client,
10659         .init_client    = nfs4_init_client,
10660         .free_client    = nfs4_free_client,
10661         .create_server  = nfs4_create_server,
10662         .clone_server   = nfs_clone_server,
10663         .discover_trunking = nfs4_discover_trunking,
10664         .enable_swap    = nfs4_enable_swap,
10665         .disable_swap   = nfs4_disable_swap,
10666 };
10667
10668 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10669         .name   = XATTR_NAME_NFSV4_ACL,
10670         .list   = nfs4_xattr_list_nfs4_acl,
10671         .get    = nfs4_xattr_get_nfs4_acl,
10672         .set    = nfs4_xattr_set_nfs4_acl,
10673 };
10674
10675 #if defined(CONFIG_NFS_V4_1)
10676 static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10677         .name   = XATTR_NAME_NFSV4_DACL,
10678         .list   = nfs4_xattr_list_nfs4_dacl,
10679         .get    = nfs4_xattr_get_nfs4_dacl,
10680         .set    = nfs4_xattr_set_nfs4_dacl,
10681 };
10682
10683 static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10684         .name   = XATTR_NAME_NFSV4_SACL,
10685         .list   = nfs4_xattr_list_nfs4_sacl,
10686         .get    = nfs4_xattr_get_nfs4_sacl,
10687         .set    = nfs4_xattr_set_nfs4_sacl,
10688 };
10689 #endif
10690
10691 #ifdef CONFIG_NFS_V4_2
10692 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10693         .prefix = XATTR_USER_PREFIX,
10694         .get    = nfs4_xattr_get_nfs4_user,
10695         .set    = nfs4_xattr_set_nfs4_user,
10696 };
10697 #endif
10698
10699 const struct xattr_handler *nfs4_xattr_handlers[] = {
10700         &nfs4_xattr_nfs4_acl_handler,
10701 #if defined(CONFIG_NFS_V4_1)
10702         &nfs4_xattr_nfs4_dacl_handler,
10703         &nfs4_xattr_nfs4_sacl_handler,
10704 #endif
10705 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10706         &nfs4_xattr_nfs4_label_handler,
10707 #endif
10708 #ifdef CONFIG_NFS_V4_2
10709         &nfs4_xattr_nfs4_user_handler,
10710 #endif
10711         NULL
10712 };