Merge tag 'nfs-for-4.8-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
[cascardo/linux.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/file.h>
42 #include <linux/string.h>
43 #include <linux/ratelimit.h>
44 #include <linux/printk.h>
45 #include <linux/slab.h>
46 #include <linux/sunrpc/clnt.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69
70 #include "nfs4trace.h"
71
72 #define NFSDBG_FACILITY         NFSDBG_PROC
73
74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
76
77 /* file attributes which can be mapped to nfs attributes */
78 #define NFS4_VALID_ATTRS (ATTR_MODE \
79         | ATTR_UID \
80         | ATTR_GID \
81         | ATTR_SIZE \
82         | ATTR_ATIME \
83         | ATTR_MTIME \
84         | ATTR_CTIME \
85         | ATTR_ATIME_SET \
86         | ATTR_MTIME_SET)
87
88 struct nfs4_opendata;
89 static int _nfs4_proc_open(struct nfs4_opendata *data);
90 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
91 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
92 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
93 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
95 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
96                             struct nfs_fattr *fattr, struct iattr *sattr,
97                             struct nfs4_state *state, struct nfs4_label *ilabel,
98                             struct nfs4_label *olabel);
99 #ifdef CONFIG_NFS_V4_1
100 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
101                 struct rpc_cred *);
102 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
103                 struct rpc_cred *);
104 #endif
105
106 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
107 static inline struct nfs4_label *
108 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
109         struct iattr *sattr, struct nfs4_label *label)
110 {
111         int err;
112
113         if (label == NULL)
114                 return NULL;
115
116         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
117                 return NULL;
118
119         err = security_dentry_init_security(dentry, sattr->ia_mode,
120                                 &dentry->d_name, (void **)&label->label, &label->len);
121         if (err == 0)
122                 return label;
123
124         return NULL;
125 }
126 static inline void
127 nfs4_label_release_security(struct nfs4_label *label)
128 {
129         if (label)
130                 security_release_secctx(label->label, label->len);
131 }
132 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
133 {
134         if (label)
135                 return server->attr_bitmask;
136
137         return server->attr_bitmask_nl;
138 }
139 #else
140 static inline struct nfs4_label *
141 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
142         struct iattr *sattr, struct nfs4_label *l)
143 { return NULL; }
144 static inline void
145 nfs4_label_release_security(struct nfs4_label *label)
146 { return; }
147 static inline u32 *
148 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
149 { return server->attr_bitmask; }
150 #endif
151
152 /* Prevent leaks of NFSv4 errors into userland */
153 static int nfs4_map_errors(int err)
154 {
155         if (err >= -1000)
156                 return err;
157         switch (err) {
158         case -NFS4ERR_RESOURCE:
159         case -NFS4ERR_LAYOUTTRYLATER:
160         case -NFS4ERR_RECALLCONFLICT:
161                 return -EREMOTEIO;
162         case -NFS4ERR_WRONGSEC:
163         case -NFS4ERR_WRONG_CRED:
164                 return -EPERM;
165         case -NFS4ERR_BADOWNER:
166         case -NFS4ERR_BADNAME:
167                 return -EINVAL;
168         case -NFS4ERR_SHARE_DENIED:
169                 return -EACCES;
170         case -NFS4ERR_MINOR_VERS_MISMATCH:
171                 return -EPROTONOSUPPORT;
172         case -NFS4ERR_FILE_OPEN:
173                 return -EBUSY;
174         default:
175                 dprintk("%s could not handle NFSv4 error %d\n",
176                                 __func__, -err);
177                 break;
178         }
179         return -EIO;
180 }
181
182 /*
183  * This is our standard bitmap for GETATTR requests.
184  */
185 const u32 nfs4_fattr_bitmap[3] = {
186         FATTR4_WORD0_TYPE
187         | FATTR4_WORD0_CHANGE
188         | FATTR4_WORD0_SIZE
189         | FATTR4_WORD0_FSID
190         | FATTR4_WORD0_FILEID,
191         FATTR4_WORD1_MODE
192         | FATTR4_WORD1_NUMLINKS
193         | FATTR4_WORD1_OWNER
194         | FATTR4_WORD1_OWNER_GROUP
195         | FATTR4_WORD1_RAWDEV
196         | FATTR4_WORD1_SPACE_USED
197         | FATTR4_WORD1_TIME_ACCESS
198         | FATTR4_WORD1_TIME_METADATA
199         | FATTR4_WORD1_TIME_MODIFY
200         | FATTR4_WORD1_MOUNTED_ON_FILEID,
201 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
202         FATTR4_WORD2_SECURITY_LABEL
203 #endif
204 };
205
206 static const u32 nfs4_pnfs_open_bitmap[3] = {
207         FATTR4_WORD0_TYPE
208         | FATTR4_WORD0_CHANGE
209         | FATTR4_WORD0_SIZE
210         | FATTR4_WORD0_FSID
211         | FATTR4_WORD0_FILEID,
212         FATTR4_WORD1_MODE
213         | FATTR4_WORD1_NUMLINKS
214         | FATTR4_WORD1_OWNER
215         | FATTR4_WORD1_OWNER_GROUP
216         | FATTR4_WORD1_RAWDEV
217         | FATTR4_WORD1_SPACE_USED
218         | FATTR4_WORD1_TIME_ACCESS
219         | FATTR4_WORD1_TIME_METADATA
220         | FATTR4_WORD1_TIME_MODIFY,
221         FATTR4_WORD2_MDSTHRESHOLD
222 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
223         | FATTR4_WORD2_SECURITY_LABEL
224 #endif
225 };
226
227 static const u32 nfs4_open_noattr_bitmap[3] = {
228         FATTR4_WORD0_TYPE
229         | FATTR4_WORD0_CHANGE
230         | FATTR4_WORD0_FILEID,
231 };
232
233 const u32 nfs4_statfs_bitmap[3] = {
234         FATTR4_WORD0_FILES_AVAIL
235         | FATTR4_WORD0_FILES_FREE
236         | FATTR4_WORD0_FILES_TOTAL,
237         FATTR4_WORD1_SPACE_AVAIL
238         | FATTR4_WORD1_SPACE_FREE
239         | FATTR4_WORD1_SPACE_TOTAL
240 };
241
242 const u32 nfs4_pathconf_bitmap[3] = {
243         FATTR4_WORD0_MAXLINK
244         | FATTR4_WORD0_MAXNAME,
245         0
246 };
247
248 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
249                         | FATTR4_WORD0_MAXREAD
250                         | FATTR4_WORD0_MAXWRITE
251                         | FATTR4_WORD0_LEASE_TIME,
252                         FATTR4_WORD1_TIME_DELTA
253                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
254                         FATTR4_WORD2_LAYOUT_BLKSIZE
255                         | FATTR4_WORD2_CLONE_BLKSIZE
256 };
257
258 const u32 nfs4_fs_locations_bitmap[3] = {
259         FATTR4_WORD0_TYPE
260         | FATTR4_WORD0_CHANGE
261         | FATTR4_WORD0_SIZE
262         | FATTR4_WORD0_FSID
263         | FATTR4_WORD0_FILEID
264         | FATTR4_WORD0_FS_LOCATIONS,
265         FATTR4_WORD1_MODE
266         | FATTR4_WORD1_NUMLINKS
267         | FATTR4_WORD1_OWNER
268         | FATTR4_WORD1_OWNER_GROUP
269         | FATTR4_WORD1_RAWDEV
270         | FATTR4_WORD1_SPACE_USED
271         | FATTR4_WORD1_TIME_ACCESS
272         | FATTR4_WORD1_TIME_METADATA
273         | FATTR4_WORD1_TIME_MODIFY
274         | FATTR4_WORD1_MOUNTED_ON_FILEID,
275 };
276
277 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
278                 struct nfs4_readdir_arg *readdir)
279 {
280         __be32 *start, *p;
281
282         if (cookie > 2) {
283                 readdir->cookie = cookie;
284                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
285                 return;
286         }
287
288         readdir->cookie = 0;
289         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
290         if (cookie == 2)
291                 return;
292         
293         /*
294          * NFSv4 servers do not return entries for '.' and '..'
295          * Therefore, we fake these entries here.  We let '.'
296          * have cookie 0 and '..' have cookie 1.  Note that
297          * when talking to the server, we always send cookie 0
298          * instead of 1 or 2.
299          */
300         start = p = kmap_atomic(*readdir->pages);
301         
302         if (cookie == 0) {
303                 *p++ = xdr_one;                                  /* next */
304                 *p++ = xdr_zero;                   /* cookie, first word */
305                 *p++ = xdr_one;                   /* cookie, second word */
306                 *p++ = xdr_one;                             /* entry len */
307                 memcpy(p, ".\0\0\0", 4);                        /* entry */
308                 p++;
309                 *p++ = xdr_one;                         /* bitmap length */
310                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
311                 *p++ = htonl(8);              /* attribute buffer length */
312                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
313         }
314         
315         *p++ = xdr_one;                                  /* next */
316         *p++ = xdr_zero;                   /* cookie, first word */
317         *p++ = xdr_two;                   /* cookie, second word */
318         *p++ = xdr_two;                             /* entry len */
319         memcpy(p, "..\0\0", 4);                         /* entry */
320         p++;
321         *p++ = xdr_one;                         /* bitmap length */
322         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
323         *p++ = htonl(8);              /* attribute buffer length */
324         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
325
326         readdir->pgbase = (char *)p - (char *)start;
327         readdir->count -= readdir->pgbase;
328         kunmap_atomic(start);
329 }
330
331 static long nfs4_update_delay(long *timeout)
332 {
333         long ret;
334         if (!timeout)
335                 return NFS4_POLL_RETRY_MAX;
336         if (*timeout <= 0)
337                 *timeout = NFS4_POLL_RETRY_MIN;
338         if (*timeout > NFS4_POLL_RETRY_MAX)
339                 *timeout = NFS4_POLL_RETRY_MAX;
340         ret = *timeout;
341         *timeout <<= 1;
342         return ret;
343 }
344
345 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
346 {
347         int res = 0;
348
349         might_sleep();
350
351         freezable_schedule_timeout_killable_unsafe(
352                 nfs4_update_delay(timeout));
353         if (fatal_signal_pending(current))
354                 res = -ERESTARTSYS;
355         return res;
356 }
357
358 /* This is the error handling routine for processes that are allowed
359  * to sleep.
360  */
361 static int nfs4_do_handle_exception(struct nfs_server *server,
362                 int errorcode, struct nfs4_exception *exception)
363 {
364         struct nfs_client *clp = server->nfs_client;
365         struct nfs4_state *state = exception->state;
366         const nfs4_stateid *stateid = exception->stateid;
367         struct inode *inode = exception->inode;
368         int ret = errorcode;
369
370         exception->delay = 0;
371         exception->recovering = 0;
372         exception->retry = 0;
373         switch(errorcode) {
374                 case 0:
375                         return 0;
376                 case -NFS4ERR_OPENMODE:
377                 case -NFS4ERR_DELEG_REVOKED:
378                 case -NFS4ERR_ADMIN_REVOKED:
379                 case -NFS4ERR_BAD_STATEID:
380                         if (inode) {
381                                 int err;
382
383                                 err = nfs_async_inode_return_delegation(inode,
384                                                 stateid);
385                                 if (err == 0)
386                                         goto wait_on_recovery;
387                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
388                                         exception->retry = 1;
389                                         break;
390                                 }
391                         }
392                         if (state == NULL)
393                                 break;
394                         ret = nfs4_schedule_stateid_recovery(server, state);
395                         if (ret < 0)
396                                 break;
397                         goto wait_on_recovery;
398                 case -NFS4ERR_EXPIRED:
399                         if (state != NULL) {
400                                 ret = nfs4_schedule_stateid_recovery(server, state);
401                                 if (ret < 0)
402                                         break;
403                         }
404                 case -NFS4ERR_STALE_STATEID:
405                 case -NFS4ERR_STALE_CLIENTID:
406                         nfs4_schedule_lease_recovery(clp);
407                         goto wait_on_recovery;
408                 case -NFS4ERR_MOVED:
409                         ret = nfs4_schedule_migration_recovery(server);
410                         if (ret < 0)
411                                 break;
412                         goto wait_on_recovery;
413                 case -NFS4ERR_LEASE_MOVED:
414                         nfs4_schedule_lease_moved_recovery(clp);
415                         goto wait_on_recovery;
416 #if defined(CONFIG_NFS_V4_1)
417                 case -NFS4ERR_BADSESSION:
418                 case -NFS4ERR_BADSLOT:
419                 case -NFS4ERR_BAD_HIGH_SLOT:
420                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
421                 case -NFS4ERR_DEADSESSION:
422                 case -NFS4ERR_SEQ_FALSE_RETRY:
423                 case -NFS4ERR_SEQ_MISORDERED:
424                         dprintk("%s ERROR: %d Reset session\n", __func__,
425                                 errorcode);
426                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
427                         goto wait_on_recovery;
428 #endif /* defined(CONFIG_NFS_V4_1) */
429                 case -NFS4ERR_FILE_OPEN:
430                         if (exception->timeout > HZ) {
431                                 /* We have retried a decent amount, time to
432                                  * fail
433                                  */
434                                 ret = -EBUSY;
435                                 break;
436                         }
437                 case -NFS4ERR_DELAY:
438                         nfs_inc_server_stats(server, NFSIOS_DELAY);
439                 case -NFS4ERR_GRACE:
440                 case -NFS4ERR_LAYOUTTRYLATER:
441                 case -NFS4ERR_RECALLCONFLICT:
442                         exception->delay = 1;
443                         return 0;
444
445                 case -NFS4ERR_RETRY_UNCACHED_REP:
446                 case -NFS4ERR_OLD_STATEID:
447                         exception->retry = 1;
448                         break;
449                 case -NFS4ERR_BADOWNER:
450                         /* The following works around a Linux server bug! */
451                 case -NFS4ERR_BADNAME:
452                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
453                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
454                                 exception->retry = 1;
455                                 printk(KERN_WARNING "NFS: v4 server %s "
456                                                 "does not accept raw "
457                                                 "uid/gids. "
458                                                 "Reenabling the idmapper.\n",
459                                                 server->nfs_client->cl_hostname);
460                         }
461         }
462         /* We failed to handle the error */
463         return nfs4_map_errors(ret);
464 wait_on_recovery:
465         exception->recovering = 1;
466         return 0;
467 }
468
469 /* This is the error handling routine for processes that are allowed
470  * to sleep.
471  */
472 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
473 {
474         struct nfs_client *clp = server->nfs_client;
475         int ret;
476
477         ret = nfs4_do_handle_exception(server, errorcode, exception);
478         if (exception->delay) {
479                 ret = nfs4_delay(server->client, &exception->timeout);
480                 goto out_retry;
481         }
482         if (exception->recovering) {
483                 ret = nfs4_wait_clnt_recover(clp);
484                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
485                         return -EIO;
486                 goto out_retry;
487         }
488         return ret;
489 out_retry:
490         if (ret == 0)
491                 exception->retry = 1;
492         return ret;
493 }
494
495 static int
496 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
497                 int errorcode, struct nfs4_exception *exception)
498 {
499         struct nfs_client *clp = server->nfs_client;
500         int ret;
501
502         ret = nfs4_do_handle_exception(server, errorcode, exception);
503         if (exception->delay) {
504                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
505                 goto out_retry;
506         }
507         if (exception->recovering) {
508                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
509                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
510                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
511                 goto out_retry;
512         }
513         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
514                 ret = -EIO;
515         return ret;
516 out_retry:
517         if (ret == 0)
518                 exception->retry = 1;
519         return ret;
520 }
521
522 static int
523 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
524                         struct nfs4_state *state, long *timeout)
525 {
526         struct nfs4_exception exception = {
527                 .state = state,
528         };
529
530         if (task->tk_status >= 0)
531                 return 0;
532         if (timeout)
533                 exception.timeout = *timeout;
534         task->tk_status = nfs4_async_handle_exception(task, server,
535                         task->tk_status,
536                         &exception);
537         if (exception.delay && timeout)
538                 *timeout = exception.timeout;
539         if (exception.retry)
540                 return -EAGAIN;
541         return 0;
542 }
543
544 /*
545  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
546  * or 'false' otherwise.
547  */
548 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
549 {
550         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
551
552         if (flavor == RPC_AUTH_GSS_KRB5I ||
553             flavor == RPC_AUTH_GSS_KRB5P)
554                 return true;
555
556         return false;
557 }
558
559 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
560 {
561         spin_lock(&clp->cl_lock);
562         if (time_before(clp->cl_last_renewal,timestamp))
563                 clp->cl_last_renewal = timestamp;
564         spin_unlock(&clp->cl_lock);
565 }
566
567 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
568 {
569         struct nfs_client *clp = server->nfs_client;
570
571         if (!nfs4_has_session(clp))
572                 do_renew_lease(clp, timestamp);
573 }
574
575 struct nfs4_call_sync_data {
576         const struct nfs_server *seq_server;
577         struct nfs4_sequence_args *seq_args;
578         struct nfs4_sequence_res *seq_res;
579 };
580
581 void nfs4_init_sequence(struct nfs4_sequence_args *args,
582                         struct nfs4_sequence_res *res, int cache_reply)
583 {
584         args->sa_slot = NULL;
585         args->sa_cache_this = cache_reply;
586         args->sa_privileged = 0;
587
588         res->sr_slot = NULL;
589 }
590
591 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
592 {
593         args->sa_privileged = 1;
594 }
595
596 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
597                          struct nfs4_sequence_args *args,
598                          struct nfs4_sequence_res *res,
599                          struct rpc_task *task)
600 {
601         struct nfs4_slot *slot;
602
603         /* slot already allocated? */
604         if (res->sr_slot != NULL)
605                 goto out_start;
606
607         spin_lock(&tbl->slot_tbl_lock);
608         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
609                 goto out_sleep;
610
611         slot = nfs4_alloc_slot(tbl);
612         if (IS_ERR(slot)) {
613                 if (slot == ERR_PTR(-ENOMEM))
614                         task->tk_timeout = HZ >> 2;
615                 goto out_sleep;
616         }
617         spin_unlock(&tbl->slot_tbl_lock);
618
619         args->sa_slot = slot;
620         res->sr_slot = slot;
621
622 out_start:
623         rpc_call_start(task);
624         return 0;
625
626 out_sleep:
627         if (args->sa_privileged)
628                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
629                                 NULL, RPC_PRIORITY_PRIVILEGED);
630         else
631                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
632         spin_unlock(&tbl->slot_tbl_lock);
633         return -EAGAIN;
634 }
635 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
636
637 static int nfs40_sequence_done(struct rpc_task *task,
638                                struct nfs4_sequence_res *res)
639 {
640         struct nfs4_slot *slot = res->sr_slot;
641         struct nfs4_slot_table *tbl;
642
643         if (slot == NULL)
644                 goto out;
645
646         tbl = slot->table;
647         spin_lock(&tbl->slot_tbl_lock);
648         if (!nfs41_wake_and_assign_slot(tbl, slot))
649                 nfs4_free_slot(tbl, slot);
650         spin_unlock(&tbl->slot_tbl_lock);
651
652         res->sr_slot = NULL;
653 out:
654         return 1;
655 }
656
657 #if defined(CONFIG_NFS_V4_1)
658
659 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
660 {
661         struct nfs4_session *session;
662         struct nfs4_slot_table *tbl;
663         struct nfs4_slot *slot = res->sr_slot;
664         bool send_new_highest_used_slotid = false;
665
666         tbl = slot->table;
667         session = tbl->session;
668
669         spin_lock(&tbl->slot_tbl_lock);
670         /* Be nice to the server: try to ensure that the last transmitted
671          * value for highest_user_slotid <= target_highest_slotid
672          */
673         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
674                 send_new_highest_used_slotid = true;
675
676         if (nfs41_wake_and_assign_slot(tbl, slot)) {
677                 send_new_highest_used_slotid = false;
678                 goto out_unlock;
679         }
680         nfs4_free_slot(tbl, slot);
681
682         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
683                 send_new_highest_used_slotid = false;
684 out_unlock:
685         spin_unlock(&tbl->slot_tbl_lock);
686         res->sr_slot = NULL;
687         if (send_new_highest_used_slotid)
688                 nfs41_notify_server(session->clp);
689 }
690
691 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
692 {
693         struct nfs4_session *session;
694         struct nfs4_slot *slot = res->sr_slot;
695         struct nfs_client *clp;
696         bool interrupted = false;
697         int ret = 1;
698
699         if (slot == NULL)
700                 goto out_noaction;
701         /* don't increment the sequence number if the task wasn't sent */
702         if (!RPC_WAS_SENT(task))
703                 goto out;
704
705         session = slot->table->session;
706
707         if (slot->interrupted) {
708                 slot->interrupted = 0;
709                 interrupted = true;
710         }
711
712         trace_nfs4_sequence_done(session, res);
713         /* Check the SEQUENCE operation status */
714         switch (res->sr_status) {
715         case 0:
716                 /* Update the slot's sequence and clientid lease timer */
717                 ++slot->seq_nr;
718                 clp = session->clp;
719                 do_renew_lease(clp, res->sr_timestamp);
720                 /* Check sequence flags */
721                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
722                 nfs41_update_target_slotid(slot->table, slot, res);
723                 break;
724         case 1:
725                 /*
726                  * sr_status remains 1 if an RPC level error occurred.
727                  * The server may or may not have processed the sequence
728                  * operation..
729                  * Mark the slot as having hosted an interrupted RPC call.
730                  */
731                 slot->interrupted = 1;
732                 goto out;
733         case -NFS4ERR_DELAY:
734                 /* The server detected a resend of the RPC call and
735                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
736                  * of RFC5661.
737                  */
738                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
739                         __func__,
740                         slot->slot_nr,
741                         slot->seq_nr);
742                 goto out_retry;
743         case -NFS4ERR_BADSLOT:
744                 /*
745                  * The slot id we used was probably retired. Try again
746                  * using a different slot id.
747                  */
748                 goto retry_nowait;
749         case -NFS4ERR_SEQ_MISORDERED:
750                 /*
751                  * Was the last operation on this sequence interrupted?
752                  * If so, retry after bumping the sequence number.
753                  */
754                 if (interrupted) {
755                         ++slot->seq_nr;
756                         goto retry_nowait;
757                 }
758                 /*
759                  * Could this slot have been previously retired?
760                  * If so, then the server may be expecting seq_nr = 1!
761                  */
762                 if (slot->seq_nr != 1) {
763                         slot->seq_nr = 1;
764                         goto retry_nowait;
765                 }
766                 break;
767         case -NFS4ERR_SEQ_FALSE_RETRY:
768                 ++slot->seq_nr;
769                 goto retry_nowait;
770         default:
771                 /* Just update the slot sequence no. */
772                 ++slot->seq_nr;
773         }
774 out:
775         /* The session may be reset by one of the error handlers. */
776         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
777         nfs41_sequence_free_slot(res);
778 out_noaction:
779         return ret;
780 retry_nowait:
781         if (rpc_restart_call_prepare(task)) {
782                 task->tk_status = 0;
783                 ret = 0;
784         }
785         goto out;
786 out_retry:
787         if (!rpc_restart_call(task))
788                 goto out;
789         rpc_delay(task, NFS4_POLL_RETRY_MAX);
790         return 0;
791 }
792 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
793
794 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
795 {
796         if (res->sr_slot == NULL)
797                 return 1;
798         if (!res->sr_slot->table->session)
799                 return nfs40_sequence_done(task, res);
800         return nfs41_sequence_done(task, res);
801 }
802 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
803
804 int nfs41_setup_sequence(struct nfs4_session *session,
805                                 struct nfs4_sequence_args *args,
806                                 struct nfs4_sequence_res *res,
807                                 struct rpc_task *task)
808 {
809         struct nfs4_slot *slot;
810         struct nfs4_slot_table *tbl;
811
812         dprintk("--> %s\n", __func__);
813         /* slot already allocated? */
814         if (res->sr_slot != NULL)
815                 goto out_success;
816
817         tbl = &session->fc_slot_table;
818
819         task->tk_timeout = 0;
820
821         spin_lock(&tbl->slot_tbl_lock);
822         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
823             !args->sa_privileged) {
824                 /* The state manager will wait until the slot table is empty */
825                 dprintk("%s session is draining\n", __func__);
826                 goto out_sleep;
827         }
828
829         slot = nfs4_alloc_slot(tbl);
830         if (IS_ERR(slot)) {
831                 /* If out of memory, try again in 1/4 second */
832                 if (slot == ERR_PTR(-ENOMEM))
833                         task->tk_timeout = HZ >> 2;
834                 dprintk("<-- %s: no free slots\n", __func__);
835                 goto out_sleep;
836         }
837         spin_unlock(&tbl->slot_tbl_lock);
838
839         args->sa_slot = slot;
840
841         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
842                         slot->slot_nr, slot->seq_nr);
843
844         res->sr_slot = slot;
845         res->sr_timestamp = jiffies;
846         res->sr_status_flags = 0;
847         /*
848          * sr_status is only set in decode_sequence, and so will remain
849          * set to 1 if an rpc level failure occurs.
850          */
851         res->sr_status = 1;
852         trace_nfs4_setup_sequence(session, args);
853 out_success:
854         rpc_call_start(task);
855         return 0;
856 out_sleep:
857         /* Privileged tasks are queued with top priority */
858         if (args->sa_privileged)
859                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
860                                 NULL, RPC_PRIORITY_PRIVILEGED);
861         else
862                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
863         spin_unlock(&tbl->slot_tbl_lock);
864         return -EAGAIN;
865 }
866 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
867
868 static int nfs4_setup_sequence(const struct nfs_server *server,
869                                struct nfs4_sequence_args *args,
870                                struct nfs4_sequence_res *res,
871                                struct rpc_task *task)
872 {
873         struct nfs4_session *session = nfs4_get_session(server);
874         int ret = 0;
875
876         if (!session)
877                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
878                                             args, res, task);
879
880         dprintk("--> %s clp %p session %p sr_slot %u\n",
881                 __func__, session->clp, session, res->sr_slot ?
882                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
883
884         ret = nfs41_setup_sequence(session, args, res, task);
885
886         dprintk("<-- %s status=%d\n", __func__, ret);
887         return ret;
888 }
889
890 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
891 {
892         struct nfs4_call_sync_data *data = calldata;
893         struct nfs4_session *session = nfs4_get_session(data->seq_server);
894
895         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
896
897         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
898 }
899
900 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
901 {
902         struct nfs4_call_sync_data *data = calldata;
903
904         nfs41_sequence_done(task, data->seq_res);
905 }
906
907 static const struct rpc_call_ops nfs41_call_sync_ops = {
908         .rpc_call_prepare = nfs41_call_sync_prepare,
909         .rpc_call_done = nfs41_call_sync_done,
910 };
911
912 #else   /* !CONFIG_NFS_V4_1 */
913
914 static int nfs4_setup_sequence(const struct nfs_server *server,
915                                struct nfs4_sequence_args *args,
916                                struct nfs4_sequence_res *res,
917                                struct rpc_task *task)
918 {
919         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
920                                     args, res, task);
921 }
922
923 int nfs4_sequence_done(struct rpc_task *task,
924                        struct nfs4_sequence_res *res)
925 {
926         return nfs40_sequence_done(task, res);
927 }
928 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
929
930 #endif  /* !CONFIG_NFS_V4_1 */
931
932 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
933 {
934         struct nfs4_call_sync_data *data = calldata;
935         nfs4_setup_sequence(data->seq_server,
936                                 data->seq_args, data->seq_res, task);
937 }
938
939 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
940 {
941         struct nfs4_call_sync_data *data = calldata;
942         nfs4_sequence_done(task, data->seq_res);
943 }
944
945 static const struct rpc_call_ops nfs40_call_sync_ops = {
946         .rpc_call_prepare = nfs40_call_sync_prepare,
947         .rpc_call_done = nfs40_call_sync_done,
948 };
949
950 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
951                                    struct nfs_server *server,
952                                    struct rpc_message *msg,
953                                    struct nfs4_sequence_args *args,
954                                    struct nfs4_sequence_res *res)
955 {
956         int ret;
957         struct rpc_task *task;
958         struct nfs_client *clp = server->nfs_client;
959         struct nfs4_call_sync_data data = {
960                 .seq_server = server,
961                 .seq_args = args,
962                 .seq_res = res,
963         };
964         struct rpc_task_setup task_setup = {
965                 .rpc_client = clnt,
966                 .rpc_message = msg,
967                 .callback_ops = clp->cl_mvops->call_sync_ops,
968                 .callback_data = &data
969         };
970
971         task = rpc_run_task(&task_setup);
972         if (IS_ERR(task))
973                 ret = PTR_ERR(task);
974         else {
975                 ret = task->tk_status;
976                 rpc_put_task(task);
977         }
978         return ret;
979 }
980
981 int nfs4_call_sync(struct rpc_clnt *clnt,
982                    struct nfs_server *server,
983                    struct rpc_message *msg,
984                    struct nfs4_sequence_args *args,
985                    struct nfs4_sequence_res *res,
986                    int cache_reply)
987 {
988         nfs4_init_sequence(args, res, cache_reply);
989         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
990 }
991
992 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
993 {
994         struct nfs_inode *nfsi = NFS_I(dir);
995
996         spin_lock(&dir->i_lock);
997         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
998         if (!cinfo->atomic || cinfo->before != dir->i_version)
999                 nfs_force_lookup_revalidate(dir);
1000         dir->i_version = cinfo->after;
1001         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1002         nfs_fscache_invalidate(dir);
1003         spin_unlock(&dir->i_lock);
1004 }
1005
1006 struct nfs4_opendata {
1007         struct kref kref;
1008         struct nfs_openargs o_arg;
1009         struct nfs_openres o_res;
1010         struct nfs_open_confirmargs c_arg;
1011         struct nfs_open_confirmres c_res;
1012         struct nfs4_string owner_name;
1013         struct nfs4_string group_name;
1014         struct nfs4_label *a_label;
1015         struct nfs_fattr f_attr;
1016         struct nfs4_label *f_label;
1017         struct dentry *dir;
1018         struct dentry *dentry;
1019         struct nfs4_state_owner *owner;
1020         struct nfs4_state *state;
1021         struct iattr attrs;
1022         unsigned long timestamp;
1023         unsigned int rpc_done : 1;
1024         unsigned int file_created : 1;
1025         unsigned int is_recover : 1;
1026         int rpc_status;
1027         int cancelled;
1028 };
1029
1030 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1031                 int err, struct nfs4_exception *exception)
1032 {
1033         if (err != -EINVAL)
1034                 return false;
1035         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1036                 return false;
1037         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1038         exception->retry = 1;
1039         return true;
1040 }
1041
1042 static u32
1043 nfs4_map_atomic_open_share(struct nfs_server *server,
1044                 fmode_t fmode, int openflags)
1045 {
1046         u32 res = 0;
1047
1048         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1049         case FMODE_READ:
1050                 res = NFS4_SHARE_ACCESS_READ;
1051                 break;
1052         case FMODE_WRITE:
1053                 res = NFS4_SHARE_ACCESS_WRITE;
1054                 break;
1055         case FMODE_READ|FMODE_WRITE:
1056                 res = NFS4_SHARE_ACCESS_BOTH;
1057         }
1058         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1059                 goto out;
1060         /* Want no delegation if we're using O_DIRECT */
1061         if (openflags & O_DIRECT)
1062                 res |= NFS4_SHARE_WANT_NO_DELEG;
1063 out:
1064         return res;
1065 }
1066
1067 static enum open_claim_type4
1068 nfs4_map_atomic_open_claim(struct nfs_server *server,
1069                 enum open_claim_type4 claim)
1070 {
1071         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1072                 return claim;
1073         switch (claim) {
1074         default:
1075                 return claim;
1076         case NFS4_OPEN_CLAIM_FH:
1077                 return NFS4_OPEN_CLAIM_NULL;
1078         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1079                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1080         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1081                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1082         }
1083 }
1084
1085 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1086 {
1087         p->o_res.f_attr = &p->f_attr;
1088         p->o_res.f_label = p->f_label;
1089         p->o_res.seqid = p->o_arg.seqid;
1090         p->c_res.seqid = p->c_arg.seqid;
1091         p->o_res.server = p->o_arg.server;
1092         p->o_res.access_request = p->o_arg.access;
1093         nfs_fattr_init(&p->f_attr);
1094         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1095 }
1096
1097 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1098                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1099                 const struct iattr *attrs,
1100                 struct nfs4_label *label,
1101                 enum open_claim_type4 claim,
1102                 gfp_t gfp_mask)
1103 {
1104         struct dentry *parent = dget_parent(dentry);
1105         struct inode *dir = d_inode(parent);
1106         struct nfs_server *server = NFS_SERVER(dir);
1107         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1108         struct nfs4_opendata *p;
1109
1110         p = kzalloc(sizeof(*p), gfp_mask);
1111         if (p == NULL)
1112                 goto err;
1113
1114         p->f_label = nfs4_label_alloc(server, gfp_mask);
1115         if (IS_ERR(p->f_label))
1116                 goto err_free_p;
1117
1118         p->a_label = nfs4_label_alloc(server, gfp_mask);
1119         if (IS_ERR(p->a_label))
1120                 goto err_free_f;
1121
1122         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1123         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1124         if (IS_ERR(p->o_arg.seqid))
1125                 goto err_free_label;
1126         nfs_sb_active(dentry->d_sb);
1127         p->dentry = dget(dentry);
1128         p->dir = parent;
1129         p->owner = sp;
1130         atomic_inc(&sp->so_count);
1131         p->o_arg.open_flags = flags;
1132         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1133         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1134                         fmode, flags);
1135         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1136          * will return permission denied for all bits until close */
1137         if (!(flags & O_EXCL)) {
1138                 /* ask server to check for all possible rights as results
1139                  * are cached */
1140                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1141                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1142         }
1143         p->o_arg.clientid = server->nfs_client->cl_clientid;
1144         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1145         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1146         p->o_arg.name = &dentry->d_name;
1147         p->o_arg.server = server;
1148         p->o_arg.bitmask = nfs4_bitmask(server, label);
1149         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1150         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1151         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1152         switch (p->o_arg.claim) {
1153         case NFS4_OPEN_CLAIM_NULL:
1154         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1155         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1156                 p->o_arg.fh = NFS_FH(dir);
1157                 break;
1158         case NFS4_OPEN_CLAIM_PREVIOUS:
1159         case NFS4_OPEN_CLAIM_FH:
1160         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1161         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1162                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1163         }
1164         if (attrs != NULL && attrs->ia_valid != 0) {
1165                 __u32 verf[2];
1166
1167                 p->o_arg.u.attrs = &p->attrs;
1168                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1169
1170                 verf[0] = jiffies;
1171                 verf[1] = current->pid;
1172                 memcpy(p->o_arg.u.verifier.data, verf,
1173                                 sizeof(p->o_arg.u.verifier.data));
1174         }
1175         p->c_arg.fh = &p->o_res.fh;
1176         p->c_arg.stateid = &p->o_res.stateid;
1177         p->c_arg.seqid = p->o_arg.seqid;
1178         nfs4_init_opendata_res(p);
1179         kref_init(&p->kref);
1180         return p;
1181
1182 err_free_label:
1183         nfs4_label_free(p->a_label);
1184 err_free_f:
1185         nfs4_label_free(p->f_label);
1186 err_free_p:
1187         kfree(p);
1188 err:
1189         dput(parent);
1190         return NULL;
1191 }
1192
1193 static void nfs4_opendata_free(struct kref *kref)
1194 {
1195         struct nfs4_opendata *p = container_of(kref,
1196                         struct nfs4_opendata, kref);
1197         struct super_block *sb = p->dentry->d_sb;
1198
1199         nfs_free_seqid(p->o_arg.seqid);
1200         if (p->state != NULL)
1201                 nfs4_put_open_state(p->state);
1202         nfs4_put_state_owner(p->owner);
1203
1204         nfs4_label_free(p->a_label);
1205         nfs4_label_free(p->f_label);
1206
1207         dput(p->dir);
1208         dput(p->dentry);
1209         nfs_sb_deactive(sb);
1210         nfs_fattr_free_names(&p->f_attr);
1211         kfree(p->f_attr.mdsthreshold);
1212         kfree(p);
1213 }
1214
1215 static void nfs4_opendata_put(struct nfs4_opendata *p)
1216 {
1217         if (p != NULL)
1218                 kref_put(&p->kref, nfs4_opendata_free);
1219 }
1220
1221 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1222 {
1223         int ret;
1224
1225         ret = rpc_wait_for_completion_task(task);
1226         return ret;
1227 }
1228
1229 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1230                 fmode_t fmode)
1231 {
1232         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1233         case FMODE_READ|FMODE_WRITE:
1234                 return state->n_rdwr != 0;
1235         case FMODE_WRITE:
1236                 return state->n_wronly != 0;
1237         case FMODE_READ:
1238                 return state->n_rdonly != 0;
1239         }
1240         WARN_ON_ONCE(1);
1241         return false;
1242 }
1243
1244 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1245 {
1246         int ret = 0;
1247
1248         if (open_mode & (O_EXCL|O_TRUNC))
1249                 goto out;
1250         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1251                 case FMODE_READ:
1252                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1253                                 && state->n_rdonly != 0;
1254                         break;
1255                 case FMODE_WRITE:
1256                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1257                                 && state->n_wronly != 0;
1258                         break;
1259                 case FMODE_READ|FMODE_WRITE:
1260                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1261                                 && state->n_rdwr != 0;
1262         }
1263 out:
1264         return ret;
1265 }
1266
1267 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1268                 enum open_claim_type4 claim)
1269 {
1270         if (delegation == NULL)
1271                 return 0;
1272         if ((delegation->type & fmode) != fmode)
1273                 return 0;
1274         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1275                 return 0;
1276         switch (claim) {
1277         case NFS4_OPEN_CLAIM_NULL:
1278         case NFS4_OPEN_CLAIM_FH:
1279                 break;
1280         case NFS4_OPEN_CLAIM_PREVIOUS:
1281                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1282                         break;
1283         default:
1284                 return 0;
1285         }
1286         nfs_mark_delegation_referenced(delegation);
1287         return 1;
1288 }
1289
1290 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1291 {
1292         switch (fmode) {
1293                 case FMODE_WRITE:
1294                         state->n_wronly++;
1295                         break;
1296                 case FMODE_READ:
1297                         state->n_rdonly++;
1298                         break;
1299                 case FMODE_READ|FMODE_WRITE:
1300                         state->n_rdwr++;
1301         }
1302         nfs4_state_set_mode_locked(state, state->state | fmode);
1303 }
1304
1305 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1306 {
1307         struct nfs_client *clp = state->owner->so_server->nfs_client;
1308         bool need_recover = false;
1309
1310         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1311                 need_recover = true;
1312         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1313                 need_recover = true;
1314         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1315                 need_recover = true;
1316         if (need_recover)
1317                 nfs4_state_mark_reclaim_nograce(clp, state);
1318 }
1319
1320 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1321                 nfs4_stateid *stateid)
1322 {
1323         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1324                 return true;
1325         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1326                 nfs_test_and_clear_all_open_stateid(state);
1327                 return true;
1328         }
1329         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1330                 return true;
1331         return false;
1332 }
1333
1334 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1335 {
1336         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1337                 return;
1338         if (state->n_wronly)
1339                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1340         if (state->n_rdonly)
1341                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1342         if (state->n_rdwr)
1343                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1344         set_bit(NFS_OPEN_STATE, &state->flags);
1345 }
1346
1347 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1348                 nfs4_stateid *arg_stateid,
1349                 nfs4_stateid *stateid, fmode_t fmode)
1350 {
1351         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1352         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1353         case FMODE_WRITE:
1354                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1355                 break;
1356         case FMODE_READ:
1357                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1358                 break;
1359         case 0:
1360                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1361                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1362                 clear_bit(NFS_OPEN_STATE, &state->flags);
1363         }
1364         if (stateid == NULL)
1365                 return;
1366         /* Handle races with OPEN */
1367         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1368             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1369             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1370                 nfs_resync_open_stateid_locked(state);
1371                 return;
1372         }
1373         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1374                 nfs4_stateid_copy(&state->stateid, stateid);
1375         nfs4_stateid_copy(&state->open_stateid, stateid);
1376 }
1377
1378 static void nfs_clear_open_stateid(struct nfs4_state *state,
1379         nfs4_stateid *arg_stateid,
1380         nfs4_stateid *stateid, fmode_t fmode)
1381 {
1382         write_seqlock(&state->seqlock);
1383         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1384         write_sequnlock(&state->seqlock);
1385         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1386                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1387 }
1388
1389 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1390 {
1391         switch (fmode) {
1392                 case FMODE_READ:
1393                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1394                         break;
1395                 case FMODE_WRITE:
1396                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1397                         break;
1398                 case FMODE_READ|FMODE_WRITE:
1399                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1400         }
1401         if (!nfs_need_update_open_stateid(state, stateid))
1402                 return;
1403         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1404                 nfs4_stateid_copy(&state->stateid, stateid);
1405         nfs4_stateid_copy(&state->open_stateid, stateid);
1406 }
1407
1408 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1409 {
1410         /*
1411          * Protect the call to nfs4_state_set_mode_locked and
1412          * serialise the stateid update
1413          */
1414         spin_lock(&state->owner->so_lock);
1415         write_seqlock(&state->seqlock);
1416         if (deleg_stateid != NULL) {
1417                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1418                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1419         }
1420         if (open_stateid != NULL)
1421                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1422         write_sequnlock(&state->seqlock);
1423         update_open_stateflags(state, fmode);
1424         spin_unlock(&state->owner->so_lock);
1425 }
1426
1427 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1428 {
1429         struct nfs_inode *nfsi = NFS_I(state->inode);
1430         struct nfs_delegation *deleg_cur;
1431         int ret = 0;
1432
1433         fmode &= (FMODE_READ|FMODE_WRITE);
1434
1435         rcu_read_lock();
1436         deleg_cur = rcu_dereference(nfsi->delegation);
1437         if (deleg_cur == NULL)
1438                 goto no_delegation;
1439
1440         spin_lock(&deleg_cur->lock);
1441         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1442            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1443             (deleg_cur->type & fmode) != fmode)
1444                 goto no_delegation_unlock;
1445
1446         if (delegation == NULL)
1447                 delegation = &deleg_cur->stateid;
1448         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1449                 goto no_delegation_unlock;
1450
1451         nfs_mark_delegation_referenced(deleg_cur);
1452         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1453         ret = 1;
1454 no_delegation_unlock:
1455         spin_unlock(&deleg_cur->lock);
1456 no_delegation:
1457         rcu_read_unlock();
1458
1459         if (!ret && open_stateid != NULL) {
1460                 __update_open_stateid(state, open_stateid, NULL, fmode);
1461                 ret = 1;
1462         }
1463         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1464                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1465
1466         return ret;
1467 }
1468
1469 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1470                 const nfs4_stateid *stateid)
1471 {
1472         struct nfs4_state *state = lsp->ls_state;
1473         bool ret = false;
1474
1475         spin_lock(&state->state_lock);
1476         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1477                 goto out_noupdate;
1478         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1479                 goto out_noupdate;
1480         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1481         ret = true;
1482 out_noupdate:
1483         spin_unlock(&state->state_lock);
1484         return ret;
1485 }
1486
1487 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1488 {
1489         struct nfs_delegation *delegation;
1490
1491         rcu_read_lock();
1492         delegation = rcu_dereference(NFS_I(inode)->delegation);
1493         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1494                 rcu_read_unlock();
1495                 return;
1496         }
1497         rcu_read_unlock();
1498         nfs4_inode_return_delegation(inode);
1499 }
1500
1501 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1502 {
1503         struct nfs4_state *state = opendata->state;
1504         struct nfs_inode *nfsi = NFS_I(state->inode);
1505         struct nfs_delegation *delegation;
1506         int open_mode = opendata->o_arg.open_flags;
1507         fmode_t fmode = opendata->o_arg.fmode;
1508         enum open_claim_type4 claim = opendata->o_arg.claim;
1509         nfs4_stateid stateid;
1510         int ret = -EAGAIN;
1511
1512         for (;;) {
1513                 spin_lock(&state->owner->so_lock);
1514                 if (can_open_cached(state, fmode, open_mode)) {
1515                         update_open_stateflags(state, fmode);
1516                         spin_unlock(&state->owner->so_lock);
1517                         goto out_return_state;
1518                 }
1519                 spin_unlock(&state->owner->so_lock);
1520                 rcu_read_lock();
1521                 delegation = rcu_dereference(nfsi->delegation);
1522                 if (!can_open_delegated(delegation, fmode, claim)) {
1523                         rcu_read_unlock();
1524                         break;
1525                 }
1526                 /* Save the delegation */
1527                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1528                 rcu_read_unlock();
1529                 nfs_release_seqid(opendata->o_arg.seqid);
1530                 if (!opendata->is_recover) {
1531                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1532                         if (ret != 0)
1533                                 goto out;
1534                 }
1535                 ret = -EAGAIN;
1536
1537                 /* Try to update the stateid using the delegation */
1538                 if (update_open_stateid(state, NULL, &stateid, fmode))
1539                         goto out_return_state;
1540         }
1541 out:
1542         return ERR_PTR(ret);
1543 out_return_state:
1544         atomic_inc(&state->count);
1545         return state;
1546 }
1547
1548 static void
1549 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1550 {
1551         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1552         struct nfs_delegation *delegation;
1553         int delegation_flags = 0;
1554
1555         rcu_read_lock();
1556         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1557         if (delegation)
1558                 delegation_flags = delegation->flags;
1559         rcu_read_unlock();
1560         switch (data->o_arg.claim) {
1561         default:
1562                 break;
1563         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1564         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1565                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1566                                    "returning a delegation for "
1567                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1568                                    clp->cl_hostname);
1569                 return;
1570         }
1571         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1572                 nfs_inode_set_delegation(state->inode,
1573                                          data->owner->so_cred,
1574                                          &data->o_res);
1575         else
1576                 nfs_inode_reclaim_delegation(state->inode,
1577                                              data->owner->so_cred,
1578                                              &data->o_res);
1579 }
1580
1581 /*
1582  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1583  * and update the nfs4_state.
1584  */
1585 static struct nfs4_state *
1586 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1587 {
1588         struct inode *inode = data->state->inode;
1589         struct nfs4_state *state = data->state;
1590         int ret;
1591
1592         if (!data->rpc_done) {
1593                 if (data->rpc_status) {
1594                         ret = data->rpc_status;
1595                         goto err;
1596                 }
1597                 /* cached opens have already been processed */
1598                 goto update;
1599         }
1600
1601         ret = nfs_refresh_inode(inode, &data->f_attr);
1602         if (ret)
1603                 goto err;
1604
1605         if (data->o_res.delegation_type != 0)
1606                 nfs4_opendata_check_deleg(data, state);
1607 update:
1608         update_open_stateid(state, &data->o_res.stateid, NULL,
1609                             data->o_arg.fmode);
1610         atomic_inc(&state->count);
1611
1612         return state;
1613 err:
1614         return ERR_PTR(ret);
1615
1616 }
1617
1618 static struct nfs4_state *
1619 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1620 {
1621         struct inode *inode;
1622         struct nfs4_state *state = NULL;
1623         int ret;
1624
1625         if (!data->rpc_done) {
1626                 state = nfs4_try_open_cached(data);
1627                 trace_nfs4_cached_open(data->state);
1628                 goto out;
1629         }
1630
1631         ret = -EAGAIN;
1632         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1633                 goto err;
1634         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1635         ret = PTR_ERR(inode);
1636         if (IS_ERR(inode))
1637                 goto err;
1638         ret = -ENOMEM;
1639         state = nfs4_get_open_state(inode, data->owner);
1640         if (state == NULL)
1641                 goto err_put_inode;
1642         if (data->o_res.delegation_type != 0)
1643                 nfs4_opendata_check_deleg(data, state);
1644         update_open_stateid(state, &data->o_res.stateid, NULL,
1645                         data->o_arg.fmode);
1646         iput(inode);
1647 out:
1648         nfs_release_seqid(data->o_arg.seqid);
1649         return state;
1650 err_put_inode:
1651         iput(inode);
1652 err:
1653         return ERR_PTR(ret);
1654 }
1655
1656 static struct nfs4_state *
1657 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1658 {
1659         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1660                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1661         return _nfs4_opendata_to_nfs4_state(data);
1662 }
1663
1664 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1665 {
1666         struct nfs_inode *nfsi = NFS_I(state->inode);
1667         struct nfs_open_context *ctx;
1668
1669         spin_lock(&state->inode->i_lock);
1670         list_for_each_entry(ctx, &nfsi->open_files, list) {
1671                 if (ctx->state != state)
1672                         continue;
1673                 get_nfs_open_context(ctx);
1674                 spin_unlock(&state->inode->i_lock);
1675                 return ctx;
1676         }
1677         spin_unlock(&state->inode->i_lock);
1678         return ERR_PTR(-ENOENT);
1679 }
1680
1681 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1682                 struct nfs4_state *state, enum open_claim_type4 claim)
1683 {
1684         struct nfs4_opendata *opendata;
1685
1686         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1687                         NULL, NULL, claim, GFP_NOFS);
1688         if (opendata == NULL)
1689                 return ERR_PTR(-ENOMEM);
1690         opendata->state = state;
1691         atomic_inc(&state->count);
1692         return opendata;
1693 }
1694
1695 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1696                 fmode_t fmode)
1697 {
1698         struct nfs4_state *newstate;
1699         int ret;
1700
1701         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1702                 return 0;
1703         opendata->o_arg.open_flags = 0;
1704         opendata->o_arg.fmode = fmode;
1705         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1706                         NFS_SB(opendata->dentry->d_sb),
1707                         fmode, 0);
1708         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1709         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1710         nfs4_init_opendata_res(opendata);
1711         ret = _nfs4_recover_proc_open(opendata);
1712         if (ret != 0)
1713                 return ret; 
1714         newstate = nfs4_opendata_to_nfs4_state(opendata);
1715         if (IS_ERR(newstate))
1716                 return PTR_ERR(newstate);
1717         if (newstate != opendata->state)
1718                 ret = -ESTALE;
1719         nfs4_close_state(newstate, fmode);
1720         return ret;
1721 }
1722
1723 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1724 {
1725         int ret;
1726
1727         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1728         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1729         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1730         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1731         /* memory barrier prior to reading state->n_* */
1732         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1733         clear_bit(NFS_OPEN_STATE, &state->flags);
1734         smp_rmb();
1735         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1736         if (ret != 0)
1737                 return ret;
1738         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1739         if (ret != 0)
1740                 return ret;
1741         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1742         if (ret != 0)
1743                 return ret;
1744         /*
1745          * We may have performed cached opens for all three recoveries.
1746          * Check if we need to update the current stateid.
1747          */
1748         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1749             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1750                 write_seqlock(&state->seqlock);
1751                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1752                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1753                 write_sequnlock(&state->seqlock);
1754         }
1755         return 0;
1756 }
1757
1758 /*
1759  * OPEN_RECLAIM:
1760  *      reclaim state on the server after a reboot.
1761  */
1762 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1763 {
1764         struct nfs_delegation *delegation;
1765         struct nfs4_opendata *opendata;
1766         fmode_t delegation_type = 0;
1767         int status;
1768
1769         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1770                         NFS4_OPEN_CLAIM_PREVIOUS);
1771         if (IS_ERR(opendata))
1772                 return PTR_ERR(opendata);
1773         rcu_read_lock();
1774         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1775         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1776                 delegation_type = delegation->type;
1777         rcu_read_unlock();
1778         opendata->o_arg.u.delegation_type = delegation_type;
1779         status = nfs4_open_recover(opendata, state);
1780         nfs4_opendata_put(opendata);
1781         return status;
1782 }
1783
1784 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1785 {
1786         struct nfs_server *server = NFS_SERVER(state->inode);
1787         struct nfs4_exception exception = { };
1788         int err;
1789         do {
1790                 err = _nfs4_do_open_reclaim(ctx, state);
1791                 trace_nfs4_open_reclaim(ctx, 0, err);
1792                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1793                         continue;
1794                 if (err != -NFS4ERR_DELAY)
1795                         break;
1796                 nfs4_handle_exception(server, err, &exception);
1797         } while (exception.retry);
1798         return err;
1799 }
1800
1801 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1802 {
1803         struct nfs_open_context *ctx;
1804         int ret;
1805
1806         ctx = nfs4_state_find_open_context(state);
1807         if (IS_ERR(ctx))
1808                 return -EAGAIN;
1809         ret = nfs4_do_open_reclaim(ctx, state);
1810         put_nfs_open_context(ctx);
1811         return ret;
1812 }
1813
1814 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1815 {
1816         switch (err) {
1817                 default:
1818                         printk(KERN_ERR "NFS: %s: unhandled error "
1819                                         "%d.\n", __func__, err);
1820                 case 0:
1821                 case -ENOENT:
1822                 case -EAGAIN:
1823                 case -ESTALE:
1824                         break;
1825                 case -NFS4ERR_BADSESSION:
1826                 case -NFS4ERR_BADSLOT:
1827                 case -NFS4ERR_BAD_HIGH_SLOT:
1828                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1829                 case -NFS4ERR_DEADSESSION:
1830                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1831                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1832                         return -EAGAIN;
1833                 case -NFS4ERR_STALE_CLIENTID:
1834                 case -NFS4ERR_STALE_STATEID:
1835                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1836                 case -NFS4ERR_EXPIRED:
1837                         /* Don't recall a delegation if it was lost */
1838                         nfs4_schedule_lease_recovery(server->nfs_client);
1839                         return -EAGAIN;
1840                 case -NFS4ERR_MOVED:
1841                         nfs4_schedule_migration_recovery(server);
1842                         return -EAGAIN;
1843                 case -NFS4ERR_LEASE_MOVED:
1844                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1845                         return -EAGAIN;
1846                 case -NFS4ERR_DELEG_REVOKED:
1847                 case -NFS4ERR_ADMIN_REVOKED:
1848                 case -NFS4ERR_BAD_STATEID:
1849                 case -NFS4ERR_OPENMODE:
1850                         nfs_inode_find_state_and_recover(state->inode,
1851                                         stateid);
1852                         nfs4_schedule_stateid_recovery(server, state);
1853                         return -EAGAIN;
1854                 case -NFS4ERR_DELAY:
1855                 case -NFS4ERR_GRACE:
1856                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1857                         ssleep(1);
1858                         return -EAGAIN;
1859                 case -ENOMEM:
1860                 case -NFS4ERR_DENIED:
1861                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1862                         return 0;
1863         }
1864         return err;
1865 }
1866
1867 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1868                 struct nfs4_state *state, const nfs4_stateid *stateid,
1869                 fmode_t type)
1870 {
1871         struct nfs_server *server = NFS_SERVER(state->inode);
1872         struct nfs4_opendata *opendata;
1873         int err = 0;
1874
1875         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1876                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1877         if (IS_ERR(opendata))
1878                 return PTR_ERR(opendata);
1879         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1880         write_seqlock(&state->seqlock);
1881         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1882         write_sequnlock(&state->seqlock);
1883         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1884         switch (type & (FMODE_READ|FMODE_WRITE)) {
1885         case FMODE_READ|FMODE_WRITE:
1886         case FMODE_WRITE:
1887                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1888                 if (err)
1889                         break;
1890                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1891                 if (err)
1892                         break;
1893         case FMODE_READ:
1894                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1895         }
1896         nfs4_opendata_put(opendata);
1897         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1898 }
1899
1900 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1901 {
1902         struct nfs4_opendata *data = calldata;
1903
1904         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1905                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1906 }
1907
1908 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1909 {
1910         struct nfs4_opendata *data = calldata;
1911
1912         nfs40_sequence_done(task, &data->c_res.seq_res);
1913
1914         data->rpc_status = task->tk_status;
1915         if (data->rpc_status == 0) {
1916                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1917                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1918                 renew_lease(data->o_res.server, data->timestamp);
1919                 data->rpc_done = 1;
1920         }
1921 }
1922
1923 static void nfs4_open_confirm_release(void *calldata)
1924 {
1925         struct nfs4_opendata *data = calldata;
1926         struct nfs4_state *state = NULL;
1927
1928         /* If this request hasn't been cancelled, do nothing */
1929         if (data->cancelled == 0)
1930                 goto out_free;
1931         /* In case of error, no cleanup! */
1932         if (!data->rpc_done)
1933                 goto out_free;
1934         state = nfs4_opendata_to_nfs4_state(data);
1935         if (!IS_ERR(state))
1936                 nfs4_close_state(state, data->o_arg.fmode);
1937 out_free:
1938         nfs4_opendata_put(data);
1939 }
1940
1941 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1942         .rpc_call_prepare = nfs4_open_confirm_prepare,
1943         .rpc_call_done = nfs4_open_confirm_done,
1944         .rpc_release = nfs4_open_confirm_release,
1945 };
1946
1947 /*
1948  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1949  */
1950 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1951 {
1952         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1953         struct rpc_task *task;
1954         struct  rpc_message msg = {
1955                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1956                 .rpc_argp = &data->c_arg,
1957                 .rpc_resp = &data->c_res,
1958                 .rpc_cred = data->owner->so_cred,
1959         };
1960         struct rpc_task_setup task_setup_data = {
1961                 .rpc_client = server->client,
1962                 .rpc_message = &msg,
1963                 .callback_ops = &nfs4_open_confirm_ops,
1964                 .callback_data = data,
1965                 .workqueue = nfsiod_workqueue,
1966                 .flags = RPC_TASK_ASYNC,
1967         };
1968         int status;
1969
1970         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1971         kref_get(&data->kref);
1972         data->rpc_done = 0;
1973         data->rpc_status = 0;
1974         data->timestamp = jiffies;
1975         if (data->is_recover)
1976                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1977         task = rpc_run_task(&task_setup_data);
1978         if (IS_ERR(task))
1979                 return PTR_ERR(task);
1980         status = nfs4_wait_for_completion_rpc_task(task);
1981         if (status != 0) {
1982                 data->cancelled = 1;
1983                 smp_wmb();
1984         } else
1985                 status = data->rpc_status;
1986         rpc_put_task(task);
1987         return status;
1988 }
1989
1990 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1991 {
1992         struct nfs4_opendata *data = calldata;
1993         struct nfs4_state_owner *sp = data->owner;
1994         struct nfs_client *clp = sp->so_server->nfs_client;
1995         enum open_claim_type4 claim = data->o_arg.claim;
1996
1997         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1998                 goto out_wait;
1999         /*
2000          * Check if we still need to send an OPEN call, or if we can use
2001          * a delegation instead.
2002          */
2003         if (data->state != NULL) {
2004                 struct nfs_delegation *delegation;
2005
2006                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2007                         goto out_no_action;
2008                 rcu_read_lock();
2009                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2010                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2011                         goto unlock_no_action;
2012                 rcu_read_unlock();
2013         }
2014         /* Update client id. */
2015         data->o_arg.clientid = clp->cl_clientid;
2016         switch (claim) {
2017         default:
2018                 break;
2019         case NFS4_OPEN_CLAIM_PREVIOUS:
2020         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2021         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2022                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2023         case NFS4_OPEN_CLAIM_FH:
2024                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2025                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2026         }
2027         data->timestamp = jiffies;
2028         if (nfs4_setup_sequence(data->o_arg.server,
2029                                 &data->o_arg.seq_args,
2030                                 &data->o_res.seq_res,
2031                                 task) != 0)
2032                 nfs_release_seqid(data->o_arg.seqid);
2033
2034         /* Set the create mode (note dependency on the session type) */
2035         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2036         if (data->o_arg.open_flags & O_EXCL) {
2037                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2038                 if (nfs4_has_persistent_session(clp))
2039                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2040                 else if (clp->cl_mvops->minor_version > 0)
2041                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2042         }
2043         return;
2044 unlock_no_action:
2045         trace_nfs4_cached_open(data->state);
2046         rcu_read_unlock();
2047 out_no_action:
2048         task->tk_action = NULL;
2049 out_wait:
2050         nfs4_sequence_done(task, &data->o_res.seq_res);
2051 }
2052
2053 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2054 {
2055         struct nfs4_opendata *data = calldata;
2056
2057         data->rpc_status = task->tk_status;
2058
2059         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2060                 return;
2061
2062         if (task->tk_status == 0) {
2063                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2064                         switch (data->o_res.f_attr->mode & S_IFMT) {
2065                         case S_IFREG:
2066                                 break;
2067                         case S_IFLNK:
2068                                 data->rpc_status = -ELOOP;
2069                                 break;
2070                         case S_IFDIR:
2071                                 data->rpc_status = -EISDIR;
2072                                 break;
2073                         default:
2074                                 data->rpc_status = -ENOTDIR;
2075                         }
2076                 }
2077                 renew_lease(data->o_res.server, data->timestamp);
2078                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2079                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2080         }
2081         data->rpc_done = 1;
2082 }
2083
2084 static void nfs4_open_release(void *calldata)
2085 {
2086         struct nfs4_opendata *data = calldata;
2087         struct nfs4_state *state = NULL;
2088
2089         /* If this request hasn't been cancelled, do nothing */
2090         if (data->cancelled == 0)
2091                 goto out_free;
2092         /* In case of error, no cleanup! */
2093         if (data->rpc_status != 0 || !data->rpc_done)
2094                 goto out_free;
2095         /* In case we need an open_confirm, no cleanup! */
2096         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2097                 goto out_free;
2098         state = nfs4_opendata_to_nfs4_state(data);
2099         if (!IS_ERR(state))
2100                 nfs4_close_state(state, data->o_arg.fmode);
2101 out_free:
2102         nfs4_opendata_put(data);
2103 }
2104
2105 static const struct rpc_call_ops nfs4_open_ops = {
2106         .rpc_call_prepare = nfs4_open_prepare,
2107         .rpc_call_done = nfs4_open_done,
2108         .rpc_release = nfs4_open_release,
2109 };
2110
2111 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2112 {
2113         struct inode *dir = d_inode(data->dir);
2114         struct nfs_server *server = NFS_SERVER(dir);
2115         struct nfs_openargs *o_arg = &data->o_arg;
2116         struct nfs_openres *o_res = &data->o_res;
2117         struct rpc_task *task;
2118         struct rpc_message msg = {
2119                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2120                 .rpc_argp = o_arg,
2121                 .rpc_resp = o_res,
2122                 .rpc_cred = data->owner->so_cred,
2123         };
2124         struct rpc_task_setup task_setup_data = {
2125                 .rpc_client = server->client,
2126                 .rpc_message = &msg,
2127                 .callback_ops = &nfs4_open_ops,
2128                 .callback_data = data,
2129                 .workqueue = nfsiod_workqueue,
2130                 .flags = RPC_TASK_ASYNC,
2131         };
2132         int status;
2133
2134         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2135         kref_get(&data->kref);
2136         data->rpc_done = 0;
2137         data->rpc_status = 0;
2138         data->cancelled = 0;
2139         data->is_recover = 0;
2140         if (isrecover) {
2141                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2142                 data->is_recover = 1;
2143         }
2144         task = rpc_run_task(&task_setup_data);
2145         if (IS_ERR(task))
2146                 return PTR_ERR(task);
2147         status = nfs4_wait_for_completion_rpc_task(task);
2148         if (status != 0) {
2149                 data->cancelled = 1;
2150                 smp_wmb();
2151         } else
2152                 status = data->rpc_status;
2153         rpc_put_task(task);
2154
2155         return status;
2156 }
2157
2158 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2159 {
2160         struct inode *dir = d_inode(data->dir);
2161         struct nfs_openres *o_res = &data->o_res;
2162         int status;
2163
2164         status = nfs4_run_open_task(data, 1);
2165         if (status != 0 || !data->rpc_done)
2166                 return status;
2167
2168         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2169
2170         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2171                 status = _nfs4_proc_open_confirm(data);
2172                 if (status != 0)
2173                         return status;
2174         }
2175
2176         return status;
2177 }
2178
2179 /*
2180  * Additional permission checks in order to distinguish between an
2181  * open for read, and an open for execute. This works around the
2182  * fact that NFSv4 OPEN treats read and execute permissions as being
2183  * the same.
2184  * Note that in the non-execute case, we want to turn off permission
2185  * checking if we just created a new file (POSIX open() semantics).
2186  */
2187 static int nfs4_opendata_access(struct rpc_cred *cred,
2188                                 struct nfs4_opendata *opendata,
2189                                 struct nfs4_state *state, fmode_t fmode,
2190                                 int openflags)
2191 {
2192         struct nfs_access_entry cache;
2193         u32 mask;
2194
2195         /* access call failed or for some reason the server doesn't
2196          * support any access modes -- defer access call until later */
2197         if (opendata->o_res.access_supported == 0)
2198                 return 0;
2199
2200         mask = 0;
2201         /*
2202          * Use openflags to check for exec, because fmode won't
2203          * always have FMODE_EXEC set when file open for exec.
2204          */
2205         if (openflags & __FMODE_EXEC) {
2206                 /* ONLY check for exec rights */
2207                 mask = MAY_EXEC;
2208         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2209                 mask = MAY_READ;
2210
2211         cache.cred = cred;
2212         cache.jiffies = jiffies;
2213         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2214         nfs_access_add_cache(state->inode, &cache);
2215
2216         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2217                 return 0;
2218
2219         /* even though OPEN succeeded, access is denied. Close the file */
2220         nfs4_close_state(state, fmode);
2221         return -EACCES;
2222 }
2223
2224 /*
2225  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2226  */
2227 static int _nfs4_proc_open(struct nfs4_opendata *data)
2228 {
2229         struct inode *dir = d_inode(data->dir);
2230         struct nfs_server *server = NFS_SERVER(dir);
2231         struct nfs_openargs *o_arg = &data->o_arg;
2232         struct nfs_openres *o_res = &data->o_res;
2233         int status;
2234
2235         status = nfs4_run_open_task(data, 0);
2236         if (!data->rpc_done)
2237                 return status;
2238         if (status != 0) {
2239                 if (status == -NFS4ERR_BADNAME &&
2240                                 !(o_arg->open_flags & O_CREAT))
2241                         return -ENOENT;
2242                 return status;
2243         }
2244
2245         nfs_fattr_map_and_free_names(server, &data->f_attr);
2246
2247         if (o_arg->open_flags & O_CREAT) {
2248                 update_changeattr(dir, &o_res->cinfo);
2249                 if (o_arg->open_flags & O_EXCL)
2250                         data->file_created = 1;
2251                 else if (o_res->cinfo.before != o_res->cinfo.after)
2252                         data->file_created = 1;
2253         }
2254         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2255                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2256         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2257                 status = _nfs4_proc_open_confirm(data);
2258                 if (status != 0)
2259                         return status;
2260         }
2261         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2262                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2263         return 0;
2264 }
2265
2266 static int nfs4_recover_expired_lease(struct nfs_server *server)
2267 {
2268         return nfs4_client_recover_expired_lease(server->nfs_client);
2269 }
2270
2271 /*
2272  * OPEN_EXPIRED:
2273  *      reclaim state on the server after a network partition.
2274  *      Assumes caller holds the appropriate lock
2275  */
2276 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2277 {
2278         struct nfs4_opendata *opendata;
2279         int ret;
2280
2281         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2282                         NFS4_OPEN_CLAIM_FH);
2283         if (IS_ERR(opendata))
2284                 return PTR_ERR(opendata);
2285         ret = nfs4_open_recover(opendata, state);
2286         if (ret == -ESTALE)
2287                 d_drop(ctx->dentry);
2288         nfs4_opendata_put(opendata);
2289         return ret;
2290 }
2291
2292 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2293 {
2294         struct nfs_server *server = NFS_SERVER(state->inode);
2295         struct nfs4_exception exception = { };
2296         int err;
2297
2298         do {
2299                 err = _nfs4_open_expired(ctx, state);
2300                 trace_nfs4_open_expired(ctx, 0, err);
2301                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2302                         continue;
2303                 switch (err) {
2304                 default:
2305                         goto out;
2306                 case -NFS4ERR_GRACE:
2307                 case -NFS4ERR_DELAY:
2308                         nfs4_handle_exception(server, err, &exception);
2309                         err = 0;
2310                 }
2311         } while (exception.retry);
2312 out:
2313         return err;
2314 }
2315
2316 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2317 {
2318         struct nfs_open_context *ctx;
2319         int ret;
2320
2321         ctx = nfs4_state_find_open_context(state);
2322         if (IS_ERR(ctx))
2323                 return -EAGAIN;
2324         ret = nfs4_do_open_expired(ctx, state);
2325         put_nfs_open_context(ctx);
2326         return ret;
2327 }
2328
2329 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2330 {
2331         nfs_remove_bad_delegation(state->inode);
2332         write_seqlock(&state->seqlock);
2333         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2334         write_sequnlock(&state->seqlock);
2335         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2336 }
2337
2338 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2339 {
2340         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2341                 nfs_finish_clear_delegation_stateid(state);
2342 }
2343
2344 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2345 {
2346         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2347         nfs40_clear_delegation_stateid(state);
2348         return nfs4_open_expired(sp, state);
2349 }
2350
2351 #if defined(CONFIG_NFS_V4_1)
2352 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2353 {
2354         struct nfs_server *server = NFS_SERVER(state->inode);
2355         nfs4_stateid stateid;
2356         struct nfs_delegation *delegation;
2357         struct rpc_cred *cred;
2358         int status;
2359
2360         /* Get the delegation credential for use by test/free_stateid */
2361         rcu_read_lock();
2362         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2363         if (delegation == NULL) {
2364                 rcu_read_unlock();
2365                 return;
2366         }
2367
2368         nfs4_stateid_copy(&stateid, &delegation->stateid);
2369         cred = get_rpccred(delegation->cred);
2370         rcu_read_unlock();
2371         status = nfs41_test_stateid(server, &stateid, cred);
2372         trace_nfs4_test_delegation_stateid(state, NULL, status);
2373
2374         if (status != NFS_OK) {
2375                 /* Free the stateid unless the server explicitly
2376                  * informs us the stateid is unrecognized. */
2377                 if (status != -NFS4ERR_BAD_STATEID)
2378                         nfs41_free_stateid(server, &stateid, cred);
2379                 nfs_finish_clear_delegation_stateid(state);
2380         }
2381
2382         put_rpccred(cred);
2383 }
2384
2385 /**
2386  * nfs41_check_open_stateid - possibly free an open stateid
2387  *
2388  * @state: NFSv4 state for an inode
2389  *
2390  * Returns NFS_OK if recovery for this stateid is now finished.
2391  * Otherwise a negative NFS4ERR value is returned.
2392  */
2393 static int nfs41_check_open_stateid(struct nfs4_state *state)
2394 {
2395         struct nfs_server *server = NFS_SERVER(state->inode);
2396         nfs4_stateid *stateid = &state->open_stateid;
2397         struct rpc_cred *cred = state->owner->so_cred;
2398         int status;
2399
2400         /* If a state reset has been done, test_stateid is unneeded */
2401         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2402             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2403             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2404                 return -NFS4ERR_BAD_STATEID;
2405
2406         status = nfs41_test_stateid(server, stateid, cred);
2407         trace_nfs4_test_open_stateid(state, NULL, status);
2408         if (status != NFS_OK) {
2409                 /* Free the stateid unless the server explicitly
2410                  * informs us the stateid is unrecognized. */
2411                 if (status != -NFS4ERR_BAD_STATEID)
2412                         nfs41_free_stateid(server, stateid, cred);
2413
2414                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2415                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2416                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2417                 clear_bit(NFS_OPEN_STATE, &state->flags);
2418         }
2419         return status;
2420 }
2421
2422 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2423 {
2424         int status;
2425
2426         nfs41_check_delegation_stateid(state);
2427         status = nfs41_check_open_stateid(state);
2428         if (status != NFS_OK)
2429                 status = nfs4_open_expired(sp, state);
2430         return status;
2431 }
2432 #endif
2433
2434 /*
2435  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2436  * fields corresponding to attributes that were used to store the verifier.
2437  * Make sure we clobber those fields in the later setattr call
2438  */
2439 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2440                                 struct iattr *sattr, struct nfs4_label **label)
2441 {
2442         const u32 *attrset = opendata->o_res.attrset;
2443
2444         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2445             !(sattr->ia_valid & ATTR_ATIME_SET))
2446                 sattr->ia_valid |= ATTR_ATIME;
2447
2448         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2449             !(sattr->ia_valid & ATTR_MTIME_SET))
2450                 sattr->ia_valid |= ATTR_MTIME;
2451
2452         /* Except MODE, it seems harmless of setting twice. */
2453         if ((attrset[1] & FATTR4_WORD1_MODE))
2454                 sattr->ia_valid &= ~ATTR_MODE;
2455
2456         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2457                 *label = NULL;
2458 }
2459
2460 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2461                 fmode_t fmode,
2462                 int flags,
2463                 struct nfs_open_context *ctx)
2464 {
2465         struct nfs4_state_owner *sp = opendata->owner;
2466         struct nfs_server *server = sp->so_server;
2467         struct dentry *dentry;
2468         struct nfs4_state *state;
2469         unsigned int seq;
2470         int ret;
2471
2472         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2473
2474         ret = _nfs4_proc_open(opendata);
2475         if (ret != 0)
2476                 goto out;
2477
2478         state = nfs4_opendata_to_nfs4_state(opendata);
2479         ret = PTR_ERR(state);
2480         if (IS_ERR(state))
2481                 goto out;
2482         if (server->caps & NFS_CAP_POSIX_LOCK)
2483                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2484
2485         dentry = opendata->dentry;
2486         if (d_really_is_negative(dentry)) {
2487                 struct dentry *alias;
2488                 d_drop(dentry);
2489                 alias = d_exact_alias(dentry, state->inode);
2490                 if (!alias)
2491                         alias = d_splice_alias(igrab(state->inode), dentry);
2492                 /* d_splice_alias() can't fail here - it's a non-directory */
2493                 if (alias) {
2494                         dput(ctx->dentry);
2495                         ctx->dentry = dentry = alias;
2496                 }
2497                 nfs_set_verifier(dentry,
2498                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2499         }
2500
2501         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2502         if (ret != 0)
2503                 goto out;
2504
2505         ctx->state = state;
2506         if (d_inode(dentry) == state->inode) {
2507                 nfs_inode_attach_open_context(ctx);
2508                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2509                         nfs4_schedule_stateid_recovery(server, state);
2510         }
2511 out:
2512         return ret;
2513 }
2514
2515 /*
2516  * Returns a referenced nfs4_state
2517  */
2518 static int _nfs4_do_open(struct inode *dir,
2519                         struct nfs_open_context *ctx,
2520                         int flags,
2521                         struct iattr *sattr,
2522                         struct nfs4_label *label,
2523                         int *opened)
2524 {
2525         struct nfs4_state_owner  *sp;
2526         struct nfs4_state     *state = NULL;
2527         struct nfs_server       *server = NFS_SERVER(dir);
2528         struct nfs4_opendata *opendata;
2529         struct dentry *dentry = ctx->dentry;
2530         struct rpc_cred *cred = ctx->cred;
2531         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2532         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2533         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2534         struct nfs4_label *olabel = NULL;
2535         int status;
2536
2537         /* Protect against reboot recovery conflicts */
2538         status = -ENOMEM;
2539         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2540         if (sp == NULL) {
2541                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2542                 goto out_err;
2543         }
2544         status = nfs4_recover_expired_lease(server);
2545         if (status != 0)
2546                 goto err_put_state_owner;
2547         if (d_really_is_positive(dentry))
2548                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2549         status = -ENOMEM;
2550         if (d_really_is_positive(dentry))
2551                 claim = NFS4_OPEN_CLAIM_FH;
2552         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2553                         label, claim, GFP_KERNEL);
2554         if (opendata == NULL)
2555                 goto err_put_state_owner;
2556
2557         if (label) {
2558                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2559                 if (IS_ERR(olabel)) {
2560                         status = PTR_ERR(olabel);
2561                         goto err_opendata_put;
2562                 }
2563         }
2564
2565         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2566                 if (!opendata->f_attr.mdsthreshold) {
2567                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2568                         if (!opendata->f_attr.mdsthreshold)
2569                                 goto err_free_label;
2570                 }
2571                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2572         }
2573         if (d_really_is_positive(dentry))
2574                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2575
2576         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2577         if (status != 0)
2578                 goto err_free_label;
2579         state = ctx->state;
2580
2581         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2582             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2583                 nfs4_exclusive_attrset(opendata, sattr, &label);
2584                 /*
2585                  * send create attributes which was not set by open
2586                  * with an extra setattr.
2587                  */
2588                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2589                         nfs_fattr_init(opendata->o_res.f_attr);
2590                         status = nfs4_do_setattr(state->inode, cred,
2591                                         opendata->o_res.f_attr, sattr,
2592                                         state, label, olabel);
2593                         if (status == 0) {
2594                                 nfs_setattr_update_inode(state->inode, sattr,
2595                                                 opendata->o_res.f_attr);
2596                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2597                         }
2598                 }
2599         }
2600         if (opened && opendata->file_created)
2601                 *opened |= FILE_CREATED;
2602
2603         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2604                 *ctx_th = opendata->f_attr.mdsthreshold;
2605                 opendata->f_attr.mdsthreshold = NULL;
2606         }
2607
2608         nfs4_label_free(olabel);
2609
2610         nfs4_opendata_put(opendata);
2611         nfs4_put_state_owner(sp);
2612         return 0;
2613 err_free_label:
2614         nfs4_label_free(olabel);
2615 err_opendata_put:
2616         nfs4_opendata_put(opendata);
2617 err_put_state_owner:
2618         nfs4_put_state_owner(sp);
2619 out_err:
2620         return status;
2621 }
2622
2623
2624 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2625                                         struct nfs_open_context *ctx,
2626                                         int flags,
2627                                         struct iattr *sattr,
2628                                         struct nfs4_label *label,
2629                                         int *opened)
2630 {
2631         struct nfs_server *server = NFS_SERVER(dir);
2632         struct nfs4_exception exception = { };
2633         struct nfs4_state *res;
2634         int status;
2635
2636         do {
2637                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2638                 res = ctx->state;
2639                 trace_nfs4_open_file(ctx, flags, status);
2640                 if (status == 0)
2641                         break;
2642                 /* NOTE: BAD_SEQID means the server and client disagree about the
2643                  * book-keeping w.r.t. state-changing operations
2644                  * (OPEN/CLOSE/LOCK/LOCKU...)
2645                  * It is actually a sign of a bug on the client or on the server.
2646                  *
2647                  * If we receive a BAD_SEQID error in the particular case of
2648                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2649                  * have unhashed the old state_owner for us, and that we can
2650                  * therefore safely retry using a new one. We should still warn
2651                  * the user though...
2652                  */
2653                 if (status == -NFS4ERR_BAD_SEQID) {
2654                         pr_warn_ratelimited("NFS: v4 server %s "
2655                                         " returned a bad sequence-id error!\n",
2656                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2657                         exception.retry = 1;
2658                         continue;
2659                 }
2660                 /*
2661                  * BAD_STATEID on OPEN means that the server cancelled our
2662                  * state before it received the OPEN_CONFIRM.
2663                  * Recover by retrying the request as per the discussion
2664                  * on Page 181 of RFC3530.
2665                  */
2666                 if (status == -NFS4ERR_BAD_STATEID) {
2667                         exception.retry = 1;
2668                         continue;
2669                 }
2670                 if (status == -EAGAIN) {
2671                         /* We must have found a delegation */
2672                         exception.retry = 1;
2673                         continue;
2674                 }
2675                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2676                         continue;
2677                 res = ERR_PTR(nfs4_handle_exception(server,
2678                                         status, &exception));
2679         } while (exception.retry);
2680         return res;
2681 }
2682
2683 static int _nfs4_do_setattr(struct inode *inode,
2684                             struct nfs_setattrargs *arg,
2685                             struct nfs_setattrres *res,
2686                             struct rpc_cred *cred,
2687                             struct nfs4_state *state)
2688 {
2689         struct nfs_server *server = NFS_SERVER(inode);
2690         struct rpc_message msg = {
2691                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2692                 .rpc_argp       = arg,
2693                 .rpc_resp       = res,
2694                 .rpc_cred       = cred,
2695         };
2696         struct rpc_cred *delegation_cred = NULL;
2697         unsigned long timestamp = jiffies;
2698         fmode_t fmode;
2699         bool truncate;
2700         int status;
2701
2702         nfs_fattr_init(res->fattr);
2703
2704         /* Servers should only apply open mode checks for file size changes */
2705         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2706         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2707
2708         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2709                 /* Use that stateid */
2710         } else if (truncate && state != NULL) {
2711                 struct nfs_lockowner lockowner = {
2712                         .l_owner = current->files,
2713                         .l_pid = current->tgid,
2714                 };
2715                 if (!nfs4_valid_open_stateid(state))
2716                         return -EBADF;
2717                 if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner,
2718                                 &arg->stateid, &delegation_cred) == -EIO)
2719                         return -EBADF;
2720         } else
2721                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2722         if (delegation_cred)
2723                 msg.rpc_cred = delegation_cred;
2724
2725         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2726
2727         put_rpccred(delegation_cred);
2728         if (status == 0 && state != NULL)
2729                 renew_lease(server, timestamp);
2730         trace_nfs4_setattr(inode, &arg->stateid, status);
2731         return status;
2732 }
2733
2734 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2735                            struct nfs_fattr *fattr, struct iattr *sattr,
2736                            struct nfs4_state *state, struct nfs4_label *ilabel,
2737                            struct nfs4_label *olabel)
2738 {
2739         struct nfs_server *server = NFS_SERVER(inode);
2740         struct nfs_setattrargs  arg = {
2741                 .fh             = NFS_FH(inode),
2742                 .iap            = sattr,
2743                 .server         = server,
2744                 .bitmask = server->attr_bitmask,
2745                 .label          = ilabel,
2746         };
2747         struct nfs_setattrres  res = {
2748                 .fattr          = fattr,
2749                 .label          = olabel,
2750                 .server         = server,
2751         };
2752         struct nfs4_exception exception = {
2753                 .state = state,
2754                 .inode = inode,
2755                 .stateid = &arg.stateid,
2756         };
2757         int err;
2758
2759         arg.bitmask = nfs4_bitmask(server, ilabel);
2760         if (ilabel)
2761                 arg.bitmask = nfs4_bitmask(server, olabel);
2762
2763         do {
2764                 err = _nfs4_do_setattr(inode, &arg, &res, cred, state);
2765                 switch (err) {
2766                 case -NFS4ERR_OPENMODE:
2767                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2768                                 pr_warn_once("NFSv4: server %s is incorrectly "
2769                                                 "applying open mode checks to "
2770                                                 "a SETATTR that is not "
2771                                                 "changing file size.\n",
2772                                                 server->nfs_client->cl_hostname);
2773                         }
2774                         if (state && !(state->state & FMODE_WRITE)) {
2775                                 err = -EBADF;
2776                                 if (sattr->ia_valid & ATTR_OPEN)
2777                                         err = -EACCES;
2778                                 goto out;
2779                         }
2780                 }
2781                 err = nfs4_handle_exception(server, err, &exception);
2782         } while (exception.retry);
2783 out:
2784         return err;
2785 }
2786
2787 static bool
2788 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2789 {
2790         if (inode == NULL || !nfs_have_layout(inode))
2791                 return false;
2792
2793         return pnfs_wait_on_layoutreturn(inode, task);
2794 }
2795
2796 struct nfs4_closedata {
2797         struct inode *inode;
2798         struct nfs4_state *state;
2799         struct nfs_closeargs arg;
2800         struct nfs_closeres res;
2801         struct nfs_fattr fattr;
2802         unsigned long timestamp;
2803         bool roc;
2804         u32 roc_barrier;
2805 };
2806
2807 static void nfs4_free_closedata(void *data)
2808 {
2809         struct nfs4_closedata *calldata = data;
2810         struct nfs4_state_owner *sp = calldata->state->owner;
2811         struct super_block *sb = calldata->state->inode->i_sb;
2812
2813         if (calldata->roc)
2814                 pnfs_roc_release(calldata->state->inode);
2815         nfs4_put_open_state(calldata->state);
2816         nfs_free_seqid(calldata->arg.seqid);
2817         nfs4_put_state_owner(sp);
2818         nfs_sb_deactive(sb);
2819         kfree(calldata);
2820 }
2821
2822 static void nfs4_close_done(struct rpc_task *task, void *data)
2823 {
2824         struct nfs4_closedata *calldata = data;
2825         struct nfs4_state *state = calldata->state;
2826         struct nfs_server *server = NFS_SERVER(calldata->inode);
2827         nfs4_stateid *res_stateid = NULL;
2828
2829         dprintk("%s: begin!\n", __func__);
2830         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2831                 return;
2832         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2833         /* hmm. we are done with the inode, and in the process of freeing
2834          * the state_owner. we keep this around to process errors
2835          */
2836         switch (task->tk_status) {
2837                 case 0:
2838                         res_stateid = &calldata->res.stateid;
2839                         if (calldata->roc)
2840                                 pnfs_roc_set_barrier(state->inode,
2841                                                      calldata->roc_barrier);
2842                         renew_lease(server, calldata->timestamp);
2843                         break;
2844                 case -NFS4ERR_ADMIN_REVOKED:
2845                 case -NFS4ERR_STALE_STATEID:
2846                 case -NFS4ERR_OLD_STATEID:
2847                 case -NFS4ERR_BAD_STATEID:
2848                 case -NFS4ERR_EXPIRED:
2849                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2850                                                 &state->open_stateid)) {
2851                                 rpc_restart_call_prepare(task);
2852                                 goto out_release;
2853                         }
2854                         if (calldata->arg.fmode == 0)
2855                                 break;
2856                 default:
2857                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2858                                 rpc_restart_call_prepare(task);
2859                                 goto out_release;
2860                         }
2861         }
2862         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2863                         res_stateid, calldata->arg.fmode);
2864 out_release:
2865         nfs_release_seqid(calldata->arg.seqid);
2866         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2867         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2868 }
2869
2870 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2871 {
2872         struct nfs4_closedata *calldata = data;
2873         struct nfs4_state *state = calldata->state;
2874         struct inode *inode = calldata->inode;
2875         bool is_rdonly, is_wronly, is_rdwr;
2876         int call_close = 0;
2877
2878         dprintk("%s: begin!\n", __func__);
2879         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2880                 goto out_wait;
2881
2882         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2883         spin_lock(&state->owner->so_lock);
2884         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2885         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2886         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2887         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2888         /* Calculate the change in open mode */
2889         calldata->arg.fmode = 0;
2890         if (state->n_rdwr == 0) {
2891                 if (state->n_rdonly == 0)
2892                         call_close |= is_rdonly;
2893                 else if (is_rdonly)
2894                         calldata->arg.fmode |= FMODE_READ;
2895                 if (state->n_wronly == 0)
2896                         call_close |= is_wronly;
2897                 else if (is_wronly)
2898                         calldata->arg.fmode |= FMODE_WRITE;
2899                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
2900                         call_close |= is_rdwr;
2901         } else if (is_rdwr)
2902                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2903
2904         if (!nfs4_valid_open_stateid(state))
2905                 call_close = 0;
2906         spin_unlock(&state->owner->so_lock);
2907
2908         if (!call_close) {
2909                 /* Note: exit _without_ calling nfs4_close_done */
2910                 goto out_no_action;
2911         }
2912
2913         if (nfs4_wait_on_layoutreturn(inode, task)) {
2914                 nfs_release_seqid(calldata->arg.seqid);
2915                 goto out_wait;
2916         }
2917
2918         if (calldata->arg.fmode == 0)
2919                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2920         if (calldata->roc)
2921                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2922
2923         calldata->arg.share_access =
2924                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2925                                 calldata->arg.fmode, 0);
2926
2927         nfs_fattr_init(calldata->res.fattr);
2928         calldata->timestamp = jiffies;
2929         if (nfs4_setup_sequence(NFS_SERVER(inode),
2930                                 &calldata->arg.seq_args,
2931                                 &calldata->res.seq_res,
2932                                 task) != 0)
2933                 nfs_release_seqid(calldata->arg.seqid);
2934         dprintk("%s: done!\n", __func__);
2935         return;
2936 out_no_action:
2937         task->tk_action = NULL;
2938 out_wait:
2939         nfs4_sequence_done(task, &calldata->res.seq_res);
2940 }
2941
2942 static const struct rpc_call_ops nfs4_close_ops = {
2943         .rpc_call_prepare = nfs4_close_prepare,
2944         .rpc_call_done = nfs4_close_done,
2945         .rpc_release = nfs4_free_closedata,
2946 };
2947
2948 static bool nfs4_roc(struct inode *inode)
2949 {
2950         if (!nfs_have_layout(inode))
2951                 return false;
2952         return pnfs_roc(inode);
2953 }
2954
2955 /* 
2956  * It is possible for data to be read/written from a mem-mapped file 
2957  * after the sys_close call (which hits the vfs layer as a flush).
2958  * This means that we can't safely call nfsv4 close on a file until 
2959  * the inode is cleared. This in turn means that we are not good
2960  * NFSv4 citizens - we do not indicate to the server to update the file's 
2961  * share state even when we are done with one of the three share 
2962  * stateid's in the inode.
2963  *
2964  * NOTE: Caller must be holding the sp->so_owner semaphore!
2965  */
2966 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2967 {
2968         struct nfs_server *server = NFS_SERVER(state->inode);
2969         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2970         struct nfs4_closedata *calldata;
2971         struct nfs4_state_owner *sp = state->owner;
2972         struct rpc_task *task;
2973         struct rpc_message msg = {
2974                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2975                 .rpc_cred = state->owner->so_cred,
2976         };
2977         struct rpc_task_setup task_setup_data = {
2978                 .rpc_client = server->client,
2979                 .rpc_message = &msg,
2980                 .callback_ops = &nfs4_close_ops,
2981                 .workqueue = nfsiod_workqueue,
2982                 .flags = RPC_TASK_ASYNC,
2983         };
2984         int status = -ENOMEM;
2985
2986         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2987                 &task_setup_data.rpc_client, &msg);
2988
2989         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2990         if (calldata == NULL)
2991                 goto out;
2992         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2993         calldata->inode = state->inode;
2994         calldata->state = state;
2995         calldata->arg.fh = NFS_FH(state->inode);
2996         /* Serialization for the sequence id */
2997         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2998         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2999         if (IS_ERR(calldata->arg.seqid))
3000                 goto out_free_calldata;
3001         calldata->arg.fmode = 0;
3002         calldata->arg.bitmask = server->cache_consistency_bitmask;
3003         calldata->res.fattr = &calldata->fattr;
3004         calldata->res.seqid = calldata->arg.seqid;
3005         calldata->res.server = server;
3006         calldata->roc = nfs4_roc(state->inode);
3007         nfs_sb_active(calldata->inode->i_sb);
3008
3009         msg.rpc_argp = &calldata->arg;
3010         msg.rpc_resp = &calldata->res;
3011         task_setup_data.callback_data = calldata;
3012         task = rpc_run_task(&task_setup_data);
3013         if (IS_ERR(task))
3014                 return PTR_ERR(task);
3015         status = 0;
3016         if (wait)
3017                 status = rpc_wait_for_completion_task(task);
3018         rpc_put_task(task);
3019         return status;
3020 out_free_calldata:
3021         kfree(calldata);
3022 out:
3023         nfs4_put_open_state(state);
3024         nfs4_put_state_owner(sp);
3025         return status;
3026 }
3027
3028 static struct inode *
3029 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3030                 int open_flags, struct iattr *attr, int *opened)
3031 {
3032         struct nfs4_state *state;
3033         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3034
3035         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3036
3037         /* Protect against concurrent sillydeletes */
3038         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3039
3040         nfs4_label_release_security(label);
3041
3042         if (IS_ERR(state))
3043                 return ERR_CAST(state);
3044         return state->inode;
3045 }
3046
3047 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3048 {
3049         if (ctx->state == NULL)
3050                 return;
3051         if (is_sync)
3052                 nfs4_close_sync(ctx->state, ctx->mode);
3053         else
3054                 nfs4_close_state(ctx->state, ctx->mode);
3055 }
3056
3057 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3058 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3059 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3060
3061 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3062 {
3063         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3064         struct nfs4_server_caps_arg args = {
3065                 .fhandle = fhandle,
3066                 .bitmask = bitmask,
3067         };
3068         struct nfs4_server_caps_res res = {};
3069         struct rpc_message msg = {
3070                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3071                 .rpc_argp = &args,
3072                 .rpc_resp = &res,
3073         };
3074         int status;
3075
3076         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3077                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3078                      FATTR4_WORD0_LINK_SUPPORT |
3079                      FATTR4_WORD0_SYMLINK_SUPPORT |
3080                      FATTR4_WORD0_ACLSUPPORT;
3081         if (minorversion)
3082                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3083
3084         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3085         if (status == 0) {
3086                 /* Sanity check the server answers */
3087                 switch (minorversion) {
3088                 case 0:
3089                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3090                         res.attr_bitmask[2] = 0;
3091                         break;
3092                 case 1:
3093                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3094                         break;
3095                 case 2:
3096                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3097                 }
3098                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3099                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3100                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3101                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3102                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3103                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3104                                 NFS_CAP_SECURITY_LABEL);
3105                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3106                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3107                         server->caps |= NFS_CAP_ACLS;
3108                 if (res.has_links != 0)
3109                         server->caps |= NFS_CAP_HARDLINKS;
3110                 if (res.has_symlinks != 0)
3111                         server->caps |= NFS_CAP_SYMLINKS;
3112                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3113                         server->caps |= NFS_CAP_FILEID;
3114                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3115                         server->caps |= NFS_CAP_MODE;
3116                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3117                         server->caps |= NFS_CAP_NLINK;
3118                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3119                         server->caps |= NFS_CAP_OWNER;
3120                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3121                         server->caps |= NFS_CAP_OWNER_GROUP;
3122                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3123                         server->caps |= NFS_CAP_ATIME;
3124                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3125                         server->caps |= NFS_CAP_CTIME;
3126                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3127                         server->caps |= NFS_CAP_MTIME;
3128 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3129                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3130                         server->caps |= NFS_CAP_SECURITY_LABEL;
3131 #endif
3132                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3133                                 sizeof(server->attr_bitmask));
3134                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3135
3136                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3137                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3138                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3139                 server->cache_consistency_bitmask[2] = 0;
3140                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3141                         sizeof(server->exclcreat_bitmask));
3142                 server->acl_bitmask = res.acl_bitmask;
3143                 server->fh_expire_type = res.fh_expire_type;
3144         }
3145
3146         return status;
3147 }
3148
3149 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3150 {
3151         struct nfs4_exception exception = { };
3152         int err;
3153         do {
3154                 err = nfs4_handle_exception(server,
3155                                 _nfs4_server_capabilities(server, fhandle),
3156                                 &exception);
3157         } while (exception.retry);
3158         return err;
3159 }
3160
3161 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3162                 struct nfs_fsinfo *info)
3163 {
3164         u32 bitmask[3];
3165         struct nfs4_lookup_root_arg args = {
3166                 .bitmask = bitmask,
3167         };
3168         struct nfs4_lookup_res res = {
3169                 .server = server,
3170                 .fattr = info->fattr,
3171                 .fh = fhandle,
3172         };
3173         struct rpc_message msg = {
3174                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3175                 .rpc_argp = &args,
3176                 .rpc_resp = &res,
3177         };
3178
3179         bitmask[0] = nfs4_fattr_bitmap[0];
3180         bitmask[1] = nfs4_fattr_bitmap[1];
3181         /*
3182          * Process the label in the upcoming getfattr
3183          */
3184         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3185
3186         nfs_fattr_init(info->fattr);
3187         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3188 }
3189
3190 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3191                 struct nfs_fsinfo *info)
3192 {
3193         struct nfs4_exception exception = { };
3194         int err;
3195         do {
3196                 err = _nfs4_lookup_root(server, fhandle, info);
3197                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3198                 switch (err) {
3199                 case 0:
3200                 case -NFS4ERR_WRONGSEC:
3201                         goto out;
3202                 default:
3203                         err = nfs4_handle_exception(server, err, &exception);
3204                 }
3205         } while (exception.retry);
3206 out:
3207         return err;
3208 }
3209
3210 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3211                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3212 {
3213         struct rpc_auth_create_args auth_args = {
3214                 .pseudoflavor = flavor,
3215         };
3216         struct rpc_auth *auth;
3217         int ret;
3218
3219         auth = rpcauth_create(&auth_args, server->client);
3220         if (IS_ERR(auth)) {
3221                 ret = -EACCES;
3222                 goto out;
3223         }
3224         ret = nfs4_lookup_root(server, fhandle, info);
3225 out:
3226         return ret;
3227 }
3228
3229 /*
3230  * Retry pseudoroot lookup with various security flavors.  We do this when:
3231  *
3232  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3233  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3234  *
3235  * Returns zero on success, or a negative NFS4ERR value, or a
3236  * negative errno value.
3237  */
3238 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3239                               struct nfs_fsinfo *info)
3240 {
3241         /* Per 3530bis 15.33.5 */
3242         static const rpc_authflavor_t flav_array[] = {
3243                 RPC_AUTH_GSS_KRB5P,
3244                 RPC_AUTH_GSS_KRB5I,
3245                 RPC_AUTH_GSS_KRB5,
3246                 RPC_AUTH_UNIX,                  /* courtesy */
3247                 RPC_AUTH_NULL,
3248         };
3249         int status = -EPERM;
3250         size_t i;
3251
3252         if (server->auth_info.flavor_len > 0) {
3253                 /* try each flavor specified by user */
3254                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3255                         status = nfs4_lookup_root_sec(server, fhandle, info,
3256                                                 server->auth_info.flavors[i]);
3257                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3258                                 continue;
3259                         break;
3260                 }
3261         } else {
3262                 /* no flavors specified by user, try default list */
3263                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3264                         status = nfs4_lookup_root_sec(server, fhandle, info,
3265                                                       flav_array[i]);
3266                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3267                                 continue;
3268                         break;
3269                 }
3270         }
3271
3272         /*
3273          * -EACCESS could mean that the user doesn't have correct permissions
3274          * to access the mount.  It could also mean that we tried to mount
3275          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3276          * existing mount programs don't handle -EACCES very well so it should
3277          * be mapped to -EPERM instead.
3278          */
3279         if (status == -EACCES)
3280                 status = -EPERM;
3281         return status;
3282 }
3283
3284 /**
3285  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3286  * @server: initialized nfs_server handle
3287  * @fhandle: we fill in the pseudo-fs root file handle
3288  * @info: we fill in an FSINFO struct
3289  * @auth_probe: probe the auth flavours
3290  *
3291  * Returns zero on success, or a negative errno.
3292  */
3293 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3294                          struct nfs_fsinfo *info,
3295                          bool auth_probe)
3296 {
3297         int status = 0;
3298
3299         if (!auth_probe)
3300                 status = nfs4_lookup_root(server, fhandle, info);
3301
3302         if (auth_probe || status == NFS4ERR_WRONGSEC)
3303                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3304                                 fhandle, info);
3305
3306         if (status == 0)
3307                 status = nfs4_server_capabilities(server, fhandle);
3308         if (status == 0)
3309                 status = nfs4_do_fsinfo(server, fhandle, info);
3310
3311         return nfs4_map_errors(status);
3312 }
3313
3314 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3315                               struct nfs_fsinfo *info)
3316 {
3317         int error;
3318         struct nfs_fattr *fattr = info->fattr;
3319         struct nfs4_label *label = NULL;
3320
3321         error = nfs4_server_capabilities(server, mntfh);
3322         if (error < 0) {
3323                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3324                 return error;
3325         }
3326
3327         label = nfs4_label_alloc(server, GFP_KERNEL);
3328         if (IS_ERR(label))
3329                 return PTR_ERR(label);
3330
3331         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3332         if (error < 0) {
3333                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3334                 goto err_free_label;
3335         }
3336
3337         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3338             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3339                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3340
3341 err_free_label:
3342         nfs4_label_free(label);
3343
3344         return error;
3345 }
3346
3347 /*
3348  * Get locations and (maybe) other attributes of a referral.
3349  * Note that we'll actually follow the referral later when
3350  * we detect fsid mismatch in inode revalidation
3351  */
3352 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3353                              const struct qstr *name, struct nfs_fattr *fattr,
3354                              struct nfs_fh *fhandle)
3355 {
3356         int status = -ENOMEM;
3357         struct page *page = NULL;
3358         struct nfs4_fs_locations *locations = NULL;
3359
3360         page = alloc_page(GFP_KERNEL);
3361         if (page == NULL)
3362                 goto out;
3363         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3364         if (locations == NULL)
3365                 goto out;
3366
3367         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3368         if (status != 0)
3369                 goto out;
3370
3371         /*
3372          * If the fsid didn't change, this is a migration event, not a
3373          * referral.  Cause us to drop into the exception handler, which
3374          * will kick off migration recovery.
3375          */
3376         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3377                 dprintk("%s: server did not return a different fsid for"
3378                         " a referral at %s\n", __func__, name->name);
3379                 status = -NFS4ERR_MOVED;
3380                 goto out;
3381         }
3382         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3383         nfs_fixup_referral_attributes(&locations->fattr);
3384
3385         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3386         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3387         memset(fhandle, 0, sizeof(struct nfs_fh));
3388 out:
3389         if (page)
3390                 __free_page(page);
3391         kfree(locations);
3392         return status;
3393 }
3394
3395 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3396                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3397 {
3398         struct nfs4_getattr_arg args = {
3399                 .fh = fhandle,
3400                 .bitmask = server->attr_bitmask,
3401         };
3402         struct nfs4_getattr_res res = {
3403                 .fattr = fattr,
3404                 .label = label,
3405                 .server = server,
3406         };
3407         struct rpc_message msg = {
3408                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3409                 .rpc_argp = &args,
3410                 .rpc_resp = &res,
3411         };
3412
3413         args.bitmask = nfs4_bitmask(server, label);
3414
3415         nfs_fattr_init(fattr);
3416         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3417 }
3418
3419 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3420                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3421 {
3422         struct nfs4_exception exception = { };
3423         int err;
3424         do {
3425                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3426                 trace_nfs4_getattr(server, fhandle, fattr, err);
3427                 err = nfs4_handle_exception(server, err,
3428                                 &exception);
3429         } while (exception.retry);
3430         return err;
3431 }
3432
3433 /* 
3434  * The file is not closed if it is opened due to the a request to change
3435  * the size of the file. The open call will not be needed once the
3436  * VFS layer lookup-intents are implemented.
3437  *
3438  * Close is called when the inode is destroyed.
3439  * If we haven't opened the file for O_WRONLY, we
3440  * need to in the size_change case to obtain a stateid.
3441  *
3442  * Got race?
3443  * Because OPEN is always done by name in nfsv4, it is
3444  * possible that we opened a different file by the same
3445  * name.  We can recognize this race condition, but we
3446  * can't do anything about it besides returning an error.
3447  *
3448  * This will be fixed with VFS changes (lookup-intent).
3449  */
3450 static int
3451 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3452                   struct iattr *sattr)
3453 {
3454         struct inode *inode = d_inode(dentry);
3455         struct rpc_cred *cred = NULL;
3456         struct nfs4_state *state = NULL;
3457         struct nfs4_label *label = NULL;
3458         int status;
3459
3460         if (pnfs_ld_layoutret_on_setattr(inode) &&
3461             sattr->ia_valid & ATTR_SIZE &&
3462             sattr->ia_size < i_size_read(inode))
3463                 pnfs_commit_and_return_layout(inode);
3464
3465         nfs_fattr_init(fattr);
3466         
3467         /* Deal with open(O_TRUNC) */
3468         if (sattr->ia_valid & ATTR_OPEN)
3469                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3470
3471         /* Optimization: if the end result is no change, don't RPC */
3472         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3473                 return 0;
3474
3475         /* Search for an existing open(O_WRITE) file */
3476         if (sattr->ia_valid & ATTR_FILE) {
3477                 struct nfs_open_context *ctx;
3478
3479                 ctx = nfs_file_open_context(sattr->ia_file);
3480                 if (ctx) {
3481                         cred = ctx->cred;
3482                         state = ctx->state;
3483                 }
3484         }
3485
3486         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3487         if (IS_ERR(label))
3488                 return PTR_ERR(label);
3489
3490         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3491         if (status == 0) {
3492                 nfs_setattr_update_inode(inode, sattr, fattr);
3493                 nfs_setsecurity(inode, fattr, label);
3494         }
3495         nfs4_label_free(label);
3496         return status;
3497 }
3498
3499 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3500                 const struct qstr *name, struct nfs_fh *fhandle,
3501                 struct nfs_fattr *fattr, struct nfs4_label *label)
3502 {
3503         struct nfs_server *server = NFS_SERVER(dir);
3504         int                    status;
3505         struct nfs4_lookup_arg args = {
3506                 .bitmask = server->attr_bitmask,
3507                 .dir_fh = NFS_FH(dir),
3508                 .name = name,
3509         };
3510         struct nfs4_lookup_res res = {
3511                 .server = server,
3512                 .fattr = fattr,
3513                 .label = label,
3514                 .fh = fhandle,
3515         };
3516         struct rpc_message msg = {
3517                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3518                 .rpc_argp = &args,
3519                 .rpc_resp = &res,
3520         };
3521
3522         args.bitmask = nfs4_bitmask(server, label);
3523
3524         nfs_fattr_init(fattr);
3525
3526         dprintk("NFS call  lookup %s\n", name->name);
3527         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3528         dprintk("NFS reply lookup: %d\n", status);
3529         return status;
3530 }
3531
3532 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3533 {
3534         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3535                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3536         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3537         fattr->nlink = 2;
3538 }
3539
3540 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3541                                    const struct qstr *name, struct nfs_fh *fhandle,
3542                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3543 {
3544         struct nfs4_exception exception = { };
3545         struct rpc_clnt *client = *clnt;
3546         int err;
3547         do {
3548                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3549                 trace_nfs4_lookup(dir, name, err);
3550                 switch (err) {
3551                 case -NFS4ERR_BADNAME:
3552                         err = -ENOENT;
3553                         goto out;
3554                 case -NFS4ERR_MOVED:
3555                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3556                         if (err == -NFS4ERR_MOVED)
3557                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3558                         goto out;
3559                 case -NFS4ERR_WRONGSEC:
3560                         err = -EPERM;
3561                         if (client != *clnt)
3562                                 goto out;
3563                         client = nfs4_negotiate_security(client, dir, name);
3564                         if (IS_ERR(client))
3565                                 return PTR_ERR(client);
3566
3567                         exception.retry = 1;
3568                         break;
3569                 default:
3570                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3571                 }
3572         } while (exception.retry);
3573
3574 out:
3575         if (err == 0)
3576                 *clnt = client;
3577         else if (client != *clnt)
3578                 rpc_shutdown_client(client);
3579
3580         return err;
3581 }
3582
3583 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3584                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3585                             struct nfs4_label *label)
3586 {
3587         int status;
3588         struct rpc_clnt *client = NFS_CLIENT(dir);
3589
3590         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3591         if (client != NFS_CLIENT(dir)) {
3592                 rpc_shutdown_client(client);
3593                 nfs_fixup_secinfo_attributes(fattr);
3594         }
3595         return status;
3596 }
3597
3598 struct rpc_clnt *
3599 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3600                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3601 {
3602         struct rpc_clnt *client = NFS_CLIENT(dir);
3603         int status;
3604
3605         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3606         if (status < 0)
3607                 return ERR_PTR(status);
3608         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3609 }
3610
3611 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3612 {
3613         struct nfs_server *server = NFS_SERVER(inode);
3614         struct nfs4_accessargs args = {
3615                 .fh = NFS_FH(inode),
3616                 .bitmask = server->cache_consistency_bitmask,
3617         };
3618         struct nfs4_accessres res = {
3619                 .server = server,
3620         };
3621         struct rpc_message msg = {
3622                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3623                 .rpc_argp = &args,
3624                 .rpc_resp = &res,
3625                 .rpc_cred = entry->cred,
3626         };
3627         int mode = entry->mask;
3628         int status = 0;
3629
3630         /*
3631          * Determine which access bits we want to ask for...
3632          */
3633         if (mode & MAY_READ)
3634                 args.access |= NFS4_ACCESS_READ;
3635         if (S_ISDIR(inode->i_mode)) {
3636                 if (mode & MAY_WRITE)
3637                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3638                 if (mode & MAY_EXEC)
3639                         args.access |= NFS4_ACCESS_LOOKUP;
3640         } else {
3641                 if (mode & MAY_WRITE)
3642                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3643                 if (mode & MAY_EXEC)
3644                         args.access |= NFS4_ACCESS_EXECUTE;
3645         }
3646
3647         res.fattr = nfs_alloc_fattr();
3648         if (res.fattr == NULL)
3649                 return -ENOMEM;
3650
3651         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3652         if (!status) {
3653                 nfs_access_set_mask(entry, res.access);
3654                 nfs_refresh_inode(inode, res.fattr);
3655         }
3656         nfs_free_fattr(res.fattr);
3657         return status;
3658 }
3659
3660 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3661 {
3662         struct nfs4_exception exception = { };
3663         int err;
3664         do {
3665                 err = _nfs4_proc_access(inode, entry);
3666                 trace_nfs4_access(inode, err);
3667                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3668                                 &exception);
3669         } while (exception.retry);
3670         return err;
3671 }
3672
3673 /*
3674  * TODO: For the time being, we don't try to get any attributes
3675  * along with any of the zero-copy operations READ, READDIR,
3676  * READLINK, WRITE.
3677  *
3678  * In the case of the first three, we want to put the GETATTR
3679  * after the read-type operation -- this is because it is hard
3680  * to predict the length of a GETATTR response in v4, and thus
3681  * align the READ data correctly.  This means that the GETATTR
3682  * may end up partially falling into the page cache, and we should
3683  * shift it into the 'tail' of the xdr_buf before processing.
3684  * To do this efficiently, we need to know the total length
3685  * of data received, which doesn't seem to be available outside
3686  * of the RPC layer.
3687  *
3688  * In the case of WRITE, we also want to put the GETATTR after
3689  * the operation -- in this case because we want to make sure
3690  * we get the post-operation mtime and size.
3691  *
3692  * Both of these changes to the XDR layer would in fact be quite
3693  * minor, but I decided to leave them for a subsequent patch.
3694  */
3695 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3696                 unsigned int pgbase, unsigned int pglen)
3697 {
3698         struct nfs4_readlink args = {
3699                 .fh       = NFS_FH(inode),
3700                 .pgbase   = pgbase,
3701                 .pglen    = pglen,
3702                 .pages    = &page,
3703         };
3704         struct nfs4_readlink_res res;
3705         struct rpc_message msg = {
3706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3707                 .rpc_argp = &args,
3708                 .rpc_resp = &res,
3709         };
3710
3711         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3712 }
3713
3714 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3715                 unsigned int pgbase, unsigned int pglen)
3716 {
3717         struct nfs4_exception exception = { };
3718         int err;
3719         do {
3720                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3721                 trace_nfs4_readlink(inode, err);
3722                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3723                                 &exception);
3724         } while (exception.retry);
3725         return err;
3726 }
3727
3728 /*
3729  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3730  */
3731 static int
3732 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3733                  int flags)
3734 {
3735         struct nfs4_label l, *ilabel = NULL;
3736         struct nfs_open_context *ctx;
3737         struct nfs4_state *state;
3738         int status = 0;
3739
3740         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3741         if (IS_ERR(ctx))
3742                 return PTR_ERR(ctx);
3743
3744         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3745
3746         sattr->ia_mode &= ~current_umask();
3747         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3748         if (IS_ERR(state)) {
3749                 status = PTR_ERR(state);
3750                 goto out;
3751         }
3752 out:
3753         nfs4_label_release_security(ilabel);
3754         put_nfs_open_context(ctx);
3755         return status;
3756 }
3757
3758 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3759 {
3760         struct nfs_server *server = NFS_SERVER(dir);
3761         struct nfs_removeargs args = {
3762                 .fh = NFS_FH(dir),
3763                 .name = *name,
3764         };
3765         struct nfs_removeres res = {
3766                 .server = server,
3767         };
3768         struct rpc_message msg = {
3769                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3770                 .rpc_argp = &args,
3771                 .rpc_resp = &res,
3772         };
3773         int status;
3774
3775         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3776         if (status == 0)
3777                 update_changeattr(dir, &res.cinfo);
3778         return status;
3779 }
3780
3781 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3782 {
3783         struct nfs4_exception exception = { };
3784         int err;
3785         do {
3786                 err = _nfs4_proc_remove(dir, name);
3787                 trace_nfs4_remove(dir, name, err);
3788                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3789                                 &exception);
3790         } while (exception.retry);
3791         return err;
3792 }
3793
3794 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3795 {
3796         struct nfs_server *server = NFS_SERVER(dir);
3797         struct nfs_removeargs *args = msg->rpc_argp;
3798         struct nfs_removeres *res = msg->rpc_resp;
3799
3800         res->server = server;
3801         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3802         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3803
3804         nfs_fattr_init(res->dir_attr);
3805 }
3806
3807 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3808 {
3809         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
3810                         &data->args.seq_args,
3811                         &data->res.seq_res,
3812                         task);
3813 }
3814
3815 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3816 {
3817         struct nfs_unlinkdata *data = task->tk_calldata;
3818         struct nfs_removeres *res = &data->res;
3819
3820         if (!nfs4_sequence_done(task, &res->seq_res))
3821                 return 0;
3822         if (nfs4_async_handle_error(task, res->server, NULL,
3823                                     &data->timeout) == -EAGAIN)
3824                 return 0;
3825         update_changeattr(dir, &res->cinfo);
3826         return 1;
3827 }
3828
3829 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3830 {
3831         struct nfs_server *server = NFS_SERVER(dir);
3832         struct nfs_renameargs *arg = msg->rpc_argp;
3833         struct nfs_renameres *res = msg->rpc_resp;
3834
3835         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3836         res->server = server;
3837         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3838 }
3839
3840 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3841 {
3842         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3843                         &data->args.seq_args,
3844                         &data->res.seq_res,
3845                         task);
3846 }
3847
3848 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3849                                  struct inode *new_dir)
3850 {
3851         struct nfs_renamedata *data = task->tk_calldata;
3852         struct nfs_renameres *res = &data->res;
3853
3854         if (!nfs4_sequence_done(task, &res->seq_res))
3855                 return 0;
3856         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3857                 return 0;
3858
3859         update_changeattr(old_dir, &res->old_cinfo);
3860         update_changeattr(new_dir, &res->new_cinfo);
3861         return 1;
3862 }
3863
3864 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
3865 {
3866         struct nfs_server *server = NFS_SERVER(inode);
3867         struct nfs4_link_arg arg = {
3868                 .fh     = NFS_FH(inode),
3869                 .dir_fh = NFS_FH(dir),
3870                 .name   = name,
3871                 .bitmask = server->attr_bitmask,
3872         };
3873         struct nfs4_link_res res = {
3874                 .server = server,
3875                 .label = NULL,
3876         };
3877         struct rpc_message msg = {
3878                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3879                 .rpc_argp = &arg,
3880                 .rpc_resp = &res,
3881         };
3882         int status = -ENOMEM;
3883
3884         res.fattr = nfs_alloc_fattr();
3885         if (res.fattr == NULL)
3886                 goto out;
3887
3888         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3889         if (IS_ERR(res.label)) {
3890                 status = PTR_ERR(res.label);
3891                 goto out;
3892         }
3893         arg.bitmask = nfs4_bitmask(server, res.label);
3894
3895         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3896         if (!status) {
3897                 update_changeattr(dir, &res.cinfo);
3898                 status = nfs_post_op_update_inode(inode, res.fattr);
3899                 if (!status)
3900                         nfs_setsecurity(inode, res.fattr, res.label);
3901         }
3902
3903
3904         nfs4_label_free(res.label);
3905
3906 out:
3907         nfs_free_fattr(res.fattr);
3908         return status;
3909 }
3910
3911 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
3912 {
3913         struct nfs4_exception exception = { };
3914         int err;
3915         do {
3916                 err = nfs4_handle_exception(NFS_SERVER(inode),
3917                                 _nfs4_proc_link(inode, dir, name),
3918                                 &exception);
3919         } while (exception.retry);
3920         return err;
3921 }
3922
3923 struct nfs4_createdata {
3924         struct rpc_message msg;
3925         struct nfs4_create_arg arg;
3926         struct nfs4_create_res res;
3927         struct nfs_fh fh;
3928         struct nfs_fattr fattr;
3929         struct nfs4_label *label;
3930 };
3931
3932 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3933                 const struct qstr *name, struct iattr *sattr, u32 ftype)
3934 {
3935         struct nfs4_createdata *data;
3936
3937         data = kzalloc(sizeof(*data), GFP_KERNEL);
3938         if (data != NULL) {
3939                 struct nfs_server *server = NFS_SERVER(dir);
3940
3941                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3942                 if (IS_ERR(data->label))
3943                         goto out_free;
3944
3945                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3946                 data->msg.rpc_argp = &data->arg;
3947                 data->msg.rpc_resp = &data->res;
3948                 data->arg.dir_fh = NFS_FH(dir);
3949                 data->arg.server = server;
3950                 data->arg.name = name;
3951                 data->arg.attrs = sattr;
3952                 data->arg.ftype = ftype;
3953                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3954                 data->res.server = server;
3955                 data->res.fh = &data->fh;
3956                 data->res.fattr = &data->fattr;
3957                 data->res.label = data->label;
3958                 nfs_fattr_init(data->res.fattr);
3959         }
3960         return data;
3961 out_free:
3962         kfree(data);
3963         return NULL;
3964 }
3965
3966 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3967 {
3968         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3969                                     &data->arg.seq_args, &data->res.seq_res, 1);
3970         if (status == 0) {
3971                 update_changeattr(dir, &data->res.dir_cinfo);
3972                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3973         }
3974         return status;
3975 }
3976
3977 static void nfs4_free_createdata(struct nfs4_createdata *data)
3978 {
3979         nfs4_label_free(data->label);
3980         kfree(data);
3981 }
3982
3983 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3984                 struct page *page, unsigned int len, struct iattr *sattr,
3985                 struct nfs4_label *label)
3986 {
3987         struct nfs4_createdata *data;
3988         int status = -ENAMETOOLONG;
3989
3990         if (len > NFS4_MAXPATHLEN)
3991                 goto out;
3992
3993         status = -ENOMEM;
3994         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3995         if (data == NULL)
3996                 goto out;
3997
3998         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3999         data->arg.u.symlink.pages = &page;
4000         data->arg.u.symlink.len = len;
4001         data->arg.label = label;
4002         
4003         status = nfs4_do_create(dir, dentry, data);
4004
4005         nfs4_free_createdata(data);
4006 out:
4007         return status;
4008 }
4009
4010 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4011                 struct page *page, unsigned int len, struct iattr *sattr)
4012 {
4013         struct nfs4_exception exception = { };
4014         struct nfs4_label l, *label = NULL;
4015         int err;
4016
4017         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4018
4019         do {
4020                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4021                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4022                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4023                                 &exception);
4024         } while (exception.retry);
4025
4026         nfs4_label_release_security(label);
4027         return err;
4028 }
4029
4030 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4031                 struct iattr *sattr, struct nfs4_label *label)
4032 {
4033         struct nfs4_createdata *data;
4034         int status = -ENOMEM;
4035
4036         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4037         if (data == NULL)
4038                 goto out;
4039
4040         data->arg.label = label;
4041         status = nfs4_do_create(dir, dentry, data);
4042
4043         nfs4_free_createdata(data);
4044 out:
4045         return status;
4046 }
4047
4048 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4049                 struct iattr *sattr)
4050 {
4051         struct nfs4_exception exception = { };
4052         struct nfs4_label l, *label = NULL;
4053         int err;
4054
4055         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4056
4057         sattr->ia_mode &= ~current_umask();
4058         do {
4059                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4060                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4061                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4062                                 &exception);
4063         } while (exception.retry);
4064         nfs4_label_release_security(label);
4065
4066         return err;
4067 }
4068
4069 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4070                 u64 cookie, struct page **pages, unsigned int count, int plus)
4071 {
4072         struct inode            *dir = d_inode(dentry);
4073         struct nfs4_readdir_arg args = {
4074                 .fh = NFS_FH(dir),
4075                 .pages = pages,
4076                 .pgbase = 0,
4077                 .count = count,
4078                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4079                 .plus = plus,
4080         };
4081         struct nfs4_readdir_res res;
4082         struct rpc_message msg = {
4083                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4084                 .rpc_argp = &args,
4085                 .rpc_resp = &res,
4086                 .rpc_cred = cred,
4087         };
4088         int                     status;
4089
4090         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4091                         dentry,
4092                         (unsigned long long)cookie);
4093         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4094         res.pgbase = args.pgbase;
4095         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4096         if (status >= 0) {
4097                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4098                 status += args.pgbase;
4099         }
4100
4101         nfs_invalidate_atime(dir);
4102
4103         dprintk("%s: returns %d\n", __func__, status);
4104         return status;
4105 }
4106
4107 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4108                 u64 cookie, struct page **pages, unsigned int count, int plus)
4109 {
4110         struct nfs4_exception exception = { };
4111         int err;
4112         do {
4113                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4114                                 pages, count, plus);
4115                 trace_nfs4_readdir(d_inode(dentry), err);
4116                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4117                                 &exception);
4118         } while (exception.retry);
4119         return err;
4120 }
4121
4122 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4123                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4124 {
4125         struct nfs4_createdata *data;
4126         int mode = sattr->ia_mode;
4127         int status = -ENOMEM;
4128
4129         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4130         if (data == NULL)
4131                 goto out;
4132
4133         if (S_ISFIFO(mode))
4134                 data->arg.ftype = NF4FIFO;
4135         else if (S_ISBLK(mode)) {
4136                 data->arg.ftype = NF4BLK;
4137                 data->arg.u.device.specdata1 = MAJOR(rdev);
4138                 data->arg.u.device.specdata2 = MINOR(rdev);
4139         }
4140         else if (S_ISCHR(mode)) {
4141                 data->arg.ftype = NF4CHR;
4142                 data->arg.u.device.specdata1 = MAJOR(rdev);
4143                 data->arg.u.device.specdata2 = MINOR(rdev);
4144         } else if (!S_ISSOCK(mode)) {
4145                 status = -EINVAL;
4146                 goto out_free;
4147         }
4148
4149         data->arg.label = label;
4150         status = nfs4_do_create(dir, dentry, data);
4151 out_free:
4152         nfs4_free_createdata(data);
4153 out:
4154         return status;
4155 }
4156
4157 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4158                 struct iattr *sattr, dev_t rdev)
4159 {
4160         struct nfs4_exception exception = { };
4161         struct nfs4_label l, *label = NULL;
4162         int err;
4163
4164         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4165
4166         sattr->ia_mode &= ~current_umask();
4167         do {
4168                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4169                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4170                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4171                                 &exception);
4172         } while (exception.retry);
4173
4174         nfs4_label_release_security(label);
4175
4176         return err;
4177 }
4178
4179 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4180                  struct nfs_fsstat *fsstat)
4181 {
4182         struct nfs4_statfs_arg args = {
4183                 .fh = fhandle,
4184                 .bitmask = server->attr_bitmask,
4185         };
4186         struct nfs4_statfs_res res = {
4187                 .fsstat = fsstat,
4188         };
4189         struct rpc_message msg = {
4190                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4191                 .rpc_argp = &args,
4192                 .rpc_resp = &res,
4193         };
4194
4195         nfs_fattr_init(fsstat->fattr);
4196         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4197 }
4198
4199 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4200 {
4201         struct nfs4_exception exception = { };
4202         int err;
4203         do {
4204                 err = nfs4_handle_exception(server,
4205                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4206                                 &exception);
4207         } while (exception.retry);
4208         return err;
4209 }
4210
4211 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4212                 struct nfs_fsinfo *fsinfo)
4213 {
4214         struct nfs4_fsinfo_arg args = {
4215                 .fh = fhandle,
4216                 .bitmask = server->attr_bitmask,
4217         };
4218         struct nfs4_fsinfo_res res = {
4219                 .fsinfo = fsinfo,
4220         };
4221         struct rpc_message msg = {
4222                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4223                 .rpc_argp = &args,
4224                 .rpc_resp = &res,
4225         };
4226
4227         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4228 }
4229
4230 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4231 {
4232         struct nfs4_exception exception = { };
4233         unsigned long now = jiffies;
4234         int err;
4235
4236         do {
4237                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4238                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4239                 if (err == 0) {
4240                         nfs4_set_lease_period(server->nfs_client,
4241                                         fsinfo->lease_time * HZ,
4242                                         now);
4243                         break;
4244                 }
4245                 err = nfs4_handle_exception(server, err, &exception);
4246         } while (exception.retry);
4247         return err;
4248 }
4249
4250 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4251 {
4252         int error;
4253
4254         nfs_fattr_init(fsinfo->fattr);
4255         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4256         if (error == 0) {
4257                 /* block layout checks this! */
4258                 server->pnfs_blksize = fsinfo->blksize;
4259                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4260         }
4261
4262         return error;
4263 }
4264
4265 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4266                 struct nfs_pathconf *pathconf)
4267 {
4268         struct nfs4_pathconf_arg args = {
4269                 .fh = fhandle,
4270                 .bitmask = server->attr_bitmask,
4271         };
4272         struct nfs4_pathconf_res res = {
4273                 .pathconf = pathconf,
4274         };
4275         struct rpc_message msg = {
4276                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4277                 .rpc_argp = &args,
4278                 .rpc_resp = &res,
4279         };
4280
4281         /* None of the pathconf attributes are mandatory to implement */
4282         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4283                 memset(pathconf, 0, sizeof(*pathconf));
4284                 return 0;
4285         }
4286
4287         nfs_fattr_init(pathconf->fattr);
4288         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4289 }
4290
4291 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4292                 struct nfs_pathconf *pathconf)
4293 {
4294         struct nfs4_exception exception = { };
4295         int err;
4296
4297         do {
4298                 err = nfs4_handle_exception(server,
4299                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4300                                 &exception);
4301         } while (exception.retry);
4302         return err;
4303 }
4304
4305 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4306                 const struct nfs_open_context *ctx,
4307                 const struct nfs_lock_context *l_ctx,
4308                 fmode_t fmode)
4309 {
4310         const struct nfs_lockowner *lockowner = NULL;
4311
4312         if (l_ctx != NULL)
4313                 lockowner = &l_ctx->lockowner;
4314         return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL);
4315 }
4316 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4317
4318 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4319                 const struct nfs_open_context *ctx,
4320                 const struct nfs_lock_context *l_ctx,
4321                 fmode_t fmode)
4322 {
4323         nfs4_stateid current_stateid;
4324
4325         /* If the current stateid represents a lost lock, then exit */
4326         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4327                 return true;
4328         return nfs4_stateid_match(stateid, &current_stateid);
4329 }
4330
4331 static bool nfs4_error_stateid_expired(int err)
4332 {
4333         switch (err) {
4334         case -NFS4ERR_DELEG_REVOKED:
4335         case -NFS4ERR_ADMIN_REVOKED:
4336         case -NFS4ERR_BAD_STATEID:
4337         case -NFS4ERR_STALE_STATEID:
4338         case -NFS4ERR_OLD_STATEID:
4339         case -NFS4ERR_OPENMODE:
4340         case -NFS4ERR_EXPIRED:
4341                 return true;
4342         }
4343         return false;
4344 }
4345
4346 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4347 {
4348         nfs_invalidate_atime(hdr->inode);
4349 }
4350
4351 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4352 {
4353         struct nfs_server *server = NFS_SERVER(hdr->inode);
4354
4355         trace_nfs4_read(hdr, task->tk_status);
4356         if (nfs4_async_handle_error(task, server,
4357                                     hdr->args.context->state,
4358                                     NULL) == -EAGAIN) {
4359                 rpc_restart_call_prepare(task);
4360                 return -EAGAIN;
4361         }
4362
4363         __nfs4_read_done_cb(hdr);
4364         if (task->tk_status > 0)
4365                 renew_lease(server, hdr->timestamp);
4366         return 0;
4367 }
4368
4369 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4370                 struct nfs_pgio_args *args)
4371 {
4372
4373         if (!nfs4_error_stateid_expired(task->tk_status) ||
4374                 nfs4_stateid_is_current(&args->stateid,
4375                                 args->context,
4376                                 args->lock_context,
4377                                 FMODE_READ))
4378                 return false;
4379         rpc_restart_call_prepare(task);
4380         return true;
4381 }
4382
4383 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4384 {
4385
4386         dprintk("--> %s\n", __func__);
4387
4388         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4389                 return -EAGAIN;
4390         if (nfs4_read_stateid_changed(task, &hdr->args))
4391                 return -EAGAIN;
4392         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4393                                     nfs4_read_done_cb(task, hdr);
4394 }
4395
4396 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4397                                  struct rpc_message *msg)
4398 {
4399         hdr->timestamp   = jiffies;
4400         if (!hdr->pgio_done_cb)
4401                 hdr->pgio_done_cb = nfs4_read_done_cb;
4402         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4403         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4404 }
4405
4406 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4407                                       struct nfs_pgio_header *hdr)
4408 {
4409         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4410                         &hdr->args.seq_args,
4411                         &hdr->res.seq_res,
4412                         task))
4413                 return 0;
4414         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4415                                 hdr->args.lock_context,
4416                                 hdr->rw_ops->rw_mode) == -EIO)
4417                 return -EIO;
4418         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4419                 return -EIO;
4420         return 0;
4421 }
4422
4423 static int nfs4_write_done_cb(struct rpc_task *task,
4424                               struct nfs_pgio_header *hdr)
4425 {
4426         struct inode *inode = hdr->inode;
4427
4428         trace_nfs4_write(hdr, task->tk_status);
4429         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4430                                     hdr->args.context->state,
4431                                     NULL) == -EAGAIN) {
4432                 rpc_restart_call_prepare(task);
4433                 return -EAGAIN;
4434         }
4435         if (task->tk_status >= 0) {
4436                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4437                 nfs_writeback_update_inode(hdr);
4438         }
4439         return 0;
4440 }
4441
4442 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4443                 struct nfs_pgio_args *args)
4444 {
4445
4446         if (!nfs4_error_stateid_expired(task->tk_status) ||
4447                 nfs4_stateid_is_current(&args->stateid,
4448                                 args->context,
4449                                 args->lock_context,
4450                                 FMODE_WRITE))
4451                 return false;
4452         rpc_restart_call_prepare(task);
4453         return true;
4454 }
4455
4456 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4457 {
4458         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4459                 return -EAGAIN;
4460         if (nfs4_write_stateid_changed(task, &hdr->args))
4461                 return -EAGAIN;
4462         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4463                 nfs4_write_done_cb(task, hdr);
4464 }
4465
4466 static
4467 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4468 {
4469         /* Don't request attributes for pNFS or O_DIRECT writes */
4470         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4471                 return false;
4472         /* Otherwise, request attributes if and only if we don't hold
4473          * a delegation
4474          */
4475         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4476 }
4477
4478 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4479                                   struct rpc_message *msg)
4480 {
4481         struct nfs_server *server = NFS_SERVER(hdr->inode);
4482
4483         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4484                 hdr->args.bitmask = NULL;
4485                 hdr->res.fattr = NULL;
4486         } else
4487                 hdr->args.bitmask = server->cache_consistency_bitmask;
4488
4489         if (!hdr->pgio_done_cb)
4490                 hdr->pgio_done_cb = nfs4_write_done_cb;
4491         hdr->res.server = server;
4492         hdr->timestamp   = jiffies;
4493
4494         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4495         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4496 }
4497
4498 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4499 {
4500         nfs4_setup_sequence(NFS_SERVER(data->inode),
4501                         &data->args.seq_args,
4502                         &data->res.seq_res,
4503                         task);
4504 }
4505
4506 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4507 {
4508         struct inode *inode = data->inode;
4509
4510         trace_nfs4_commit(data, task->tk_status);
4511         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4512                                     NULL, NULL) == -EAGAIN) {
4513                 rpc_restart_call_prepare(task);
4514                 return -EAGAIN;
4515         }
4516         return 0;
4517 }
4518
4519 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4520 {
4521         if (!nfs4_sequence_done(task, &data->res.seq_res))
4522                 return -EAGAIN;
4523         return data->commit_done_cb(task, data);
4524 }
4525
4526 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4527 {
4528         struct nfs_server *server = NFS_SERVER(data->inode);
4529
4530         if (data->commit_done_cb == NULL)
4531                 data->commit_done_cb = nfs4_commit_done_cb;
4532         data->res.server = server;
4533         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4534         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4535 }
4536
4537 struct nfs4_renewdata {
4538         struct nfs_client       *client;
4539         unsigned long           timestamp;
4540 };
4541
4542 /*
4543  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4544  * standalone procedure for queueing an asynchronous RENEW.
4545  */
4546 static void nfs4_renew_release(void *calldata)
4547 {
4548         struct nfs4_renewdata *data = calldata;
4549         struct nfs_client *clp = data->client;
4550
4551         if (atomic_read(&clp->cl_count) > 1)
4552                 nfs4_schedule_state_renewal(clp);
4553         nfs_put_client(clp);
4554         kfree(data);
4555 }
4556
4557 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4558 {
4559         struct nfs4_renewdata *data = calldata;
4560         struct nfs_client *clp = data->client;
4561         unsigned long timestamp = data->timestamp;
4562
4563         trace_nfs4_renew_async(clp, task->tk_status);
4564         switch (task->tk_status) {
4565         case 0:
4566                 break;
4567         case -NFS4ERR_LEASE_MOVED:
4568                 nfs4_schedule_lease_moved_recovery(clp);
4569                 break;
4570         default:
4571                 /* Unless we're shutting down, schedule state recovery! */
4572                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4573                         return;
4574                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4575                         nfs4_schedule_lease_recovery(clp);
4576                         return;
4577                 }
4578                 nfs4_schedule_path_down_recovery(clp);
4579         }
4580         do_renew_lease(clp, timestamp);
4581 }
4582
4583 static const struct rpc_call_ops nfs4_renew_ops = {
4584         .rpc_call_done = nfs4_renew_done,
4585         .rpc_release = nfs4_renew_release,
4586 };
4587
4588 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4589 {
4590         struct rpc_message msg = {
4591                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4592                 .rpc_argp       = clp,
4593                 .rpc_cred       = cred,
4594         };
4595         struct nfs4_renewdata *data;
4596
4597         if (renew_flags == 0)
4598                 return 0;
4599         if (!atomic_inc_not_zero(&clp->cl_count))
4600                 return -EIO;
4601         data = kmalloc(sizeof(*data), GFP_NOFS);
4602         if (data == NULL)
4603                 return -ENOMEM;
4604         data->client = clp;
4605         data->timestamp = jiffies;
4606         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4607                         &nfs4_renew_ops, data);
4608 }
4609
4610 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4611 {
4612         struct rpc_message msg = {
4613                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4614                 .rpc_argp       = clp,
4615                 .rpc_cred       = cred,
4616         };
4617         unsigned long now = jiffies;
4618         int status;
4619
4620         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4621         if (status < 0)
4622                 return status;
4623         do_renew_lease(clp, now);
4624         return 0;
4625 }
4626
4627 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4628 {
4629         return server->caps & NFS_CAP_ACLS;
4630 }
4631
4632 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4633  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4634  * the stack.
4635  */
4636 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4637
4638 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4639                 struct page **pages)
4640 {
4641         struct page *newpage, **spages;
4642         int rc = 0;
4643         size_t len;
4644         spages = pages;
4645
4646         do {
4647                 len = min_t(size_t, PAGE_SIZE, buflen);
4648                 newpage = alloc_page(GFP_KERNEL);
4649
4650                 if (newpage == NULL)
4651                         goto unwind;
4652                 memcpy(page_address(newpage), buf, len);
4653                 buf += len;
4654                 buflen -= len;
4655                 *pages++ = newpage;
4656                 rc++;
4657         } while (buflen != 0);
4658
4659         return rc;
4660
4661 unwind:
4662         for(; rc > 0; rc--)
4663                 __free_page(spages[rc-1]);
4664         return -ENOMEM;
4665 }
4666
4667 struct nfs4_cached_acl {
4668         int cached;
4669         size_t len;
4670         char data[0];
4671 };
4672
4673 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4674 {
4675         struct nfs_inode *nfsi = NFS_I(inode);
4676
4677         spin_lock(&inode->i_lock);
4678         kfree(nfsi->nfs4_acl);
4679         nfsi->nfs4_acl = acl;
4680         spin_unlock(&inode->i_lock);
4681 }
4682
4683 static void nfs4_zap_acl_attr(struct inode *inode)
4684 {
4685         nfs4_set_cached_acl(inode, NULL);
4686 }
4687
4688 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4689 {
4690         struct nfs_inode *nfsi = NFS_I(inode);
4691         struct nfs4_cached_acl *acl;
4692         int ret = -ENOENT;
4693
4694         spin_lock(&inode->i_lock);
4695         acl = nfsi->nfs4_acl;
4696         if (acl == NULL)
4697                 goto out;
4698         if (buf == NULL) /* user is just asking for length */
4699                 goto out_len;
4700         if (acl->cached == 0)
4701                 goto out;
4702         ret = -ERANGE; /* see getxattr(2) man page */
4703         if (acl->len > buflen)
4704                 goto out;
4705         memcpy(buf, acl->data, acl->len);
4706 out_len:
4707         ret = acl->len;
4708 out:
4709         spin_unlock(&inode->i_lock);
4710         return ret;
4711 }
4712
4713 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4714 {
4715         struct nfs4_cached_acl *acl;
4716         size_t buflen = sizeof(*acl) + acl_len;
4717
4718         if (buflen <= PAGE_SIZE) {
4719                 acl = kmalloc(buflen, GFP_KERNEL);
4720                 if (acl == NULL)
4721                         goto out;
4722                 acl->cached = 1;
4723                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4724         } else {
4725                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4726                 if (acl == NULL)
4727                         goto out;
4728                 acl->cached = 0;
4729         }
4730         acl->len = acl_len;
4731 out:
4732         nfs4_set_cached_acl(inode, acl);
4733 }
4734
4735 /*
4736  * The getxattr API returns the required buffer length when called with a
4737  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4738  * the required buf.  On a NULL buf, we send a page of data to the server
4739  * guessing that the ACL request can be serviced by a page. If so, we cache
4740  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4741  * the cache. If not so, we throw away the page, and cache the required
4742  * length. The next getxattr call will then produce another round trip to
4743  * the server, this time with the input buf of the required size.
4744  */
4745 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4746 {
4747         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4748         struct nfs_getaclargs args = {
4749                 .fh = NFS_FH(inode),
4750                 .acl_pages = pages,
4751                 .acl_len = buflen,
4752         };
4753         struct nfs_getaclres res = {
4754                 .acl_len = buflen,
4755         };
4756         struct rpc_message msg = {
4757                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4758                 .rpc_argp = &args,
4759                 .rpc_resp = &res,
4760         };
4761         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4762         int ret = -ENOMEM, i;
4763
4764         /* As long as we're doing a round trip to the server anyway,
4765          * let's be prepared for a page of acl data. */
4766         if (npages == 0)
4767                 npages = 1;
4768         if (npages > ARRAY_SIZE(pages))
4769                 return -ERANGE;
4770
4771         for (i = 0; i < npages; i++) {
4772                 pages[i] = alloc_page(GFP_KERNEL);
4773                 if (!pages[i])
4774                         goto out_free;
4775         }
4776
4777         /* for decoding across pages */
4778         res.acl_scratch = alloc_page(GFP_KERNEL);
4779         if (!res.acl_scratch)
4780                 goto out_free;
4781
4782         args.acl_len = npages * PAGE_SIZE;
4783
4784         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4785                 __func__, buf, buflen, npages, args.acl_len);
4786         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4787                              &msg, &args.seq_args, &res.seq_res, 0);
4788         if (ret)
4789                 goto out_free;
4790
4791         /* Handle the case where the passed-in buffer is too short */
4792         if (res.acl_flags & NFS4_ACL_TRUNC) {
4793                 /* Did the user only issue a request for the acl length? */
4794                 if (buf == NULL)
4795                         goto out_ok;
4796                 ret = -ERANGE;
4797                 goto out_free;
4798         }
4799         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4800         if (buf) {
4801                 if (res.acl_len > buflen) {
4802                         ret = -ERANGE;
4803                         goto out_free;
4804                 }
4805                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4806         }
4807 out_ok:
4808         ret = res.acl_len;
4809 out_free:
4810         for (i = 0; i < npages; i++)
4811                 if (pages[i])
4812                         __free_page(pages[i]);
4813         if (res.acl_scratch)
4814                 __free_page(res.acl_scratch);
4815         return ret;
4816 }
4817
4818 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4819 {
4820         struct nfs4_exception exception = { };
4821         ssize_t ret;
4822         do {
4823                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4824                 trace_nfs4_get_acl(inode, ret);
4825                 if (ret >= 0)
4826                         break;
4827                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4828         } while (exception.retry);
4829         return ret;
4830 }
4831
4832 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4833 {
4834         struct nfs_server *server = NFS_SERVER(inode);
4835         int ret;
4836
4837         if (!nfs4_server_supports_acls(server))
4838                 return -EOPNOTSUPP;
4839         ret = nfs_revalidate_inode(server, inode);
4840         if (ret < 0)
4841                 return ret;
4842         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4843                 nfs_zap_acl_cache(inode);
4844         ret = nfs4_read_cached_acl(inode, buf, buflen);
4845         if (ret != -ENOENT)
4846                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4847                  * but no cached acl data, just the acl length */
4848                 return ret;
4849         return nfs4_get_acl_uncached(inode, buf, buflen);
4850 }
4851
4852 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4853 {
4854         struct nfs_server *server = NFS_SERVER(inode);
4855         struct page *pages[NFS4ACL_MAXPAGES];
4856         struct nfs_setaclargs arg = {
4857                 .fh             = NFS_FH(inode),
4858                 .acl_pages      = pages,
4859                 .acl_len        = buflen,
4860         };
4861         struct nfs_setaclres res;
4862         struct rpc_message msg = {
4863                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4864                 .rpc_argp       = &arg,
4865                 .rpc_resp       = &res,
4866         };
4867         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4868         int ret, i;
4869
4870         if (!nfs4_server_supports_acls(server))
4871                 return -EOPNOTSUPP;
4872         if (npages > ARRAY_SIZE(pages))
4873                 return -ERANGE;
4874         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4875         if (i < 0)
4876                 return i;
4877         nfs4_inode_return_delegation(inode);
4878         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4879
4880         /*
4881          * Free each page after tx, so the only ref left is
4882          * held by the network stack
4883          */
4884         for (; i > 0; i--)
4885                 put_page(pages[i-1]);
4886
4887         /*
4888          * Acl update can result in inode attribute update.
4889          * so mark the attribute cache invalid.
4890          */
4891         spin_lock(&inode->i_lock);
4892         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4893         spin_unlock(&inode->i_lock);
4894         nfs_access_zap_cache(inode);
4895         nfs_zap_acl_cache(inode);
4896         return ret;
4897 }
4898
4899 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4900 {
4901         struct nfs4_exception exception = { };
4902         int err;
4903         do {
4904                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4905                 trace_nfs4_set_acl(inode, err);
4906                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4907                                 &exception);
4908         } while (exception.retry);
4909         return err;
4910 }
4911
4912 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4913 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4914                                         size_t buflen)
4915 {
4916         struct nfs_server *server = NFS_SERVER(inode);
4917         struct nfs_fattr fattr;
4918         struct nfs4_label label = {0, 0, buflen, buf};
4919
4920         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4921         struct nfs4_getattr_arg arg = {
4922                 .fh             = NFS_FH(inode),
4923                 .bitmask        = bitmask,
4924         };
4925         struct nfs4_getattr_res res = {
4926                 .fattr          = &fattr,
4927                 .label          = &label,
4928                 .server         = server,
4929         };
4930         struct rpc_message msg = {
4931                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4932                 .rpc_argp       = &arg,
4933                 .rpc_resp       = &res,
4934         };
4935         int ret;
4936
4937         nfs_fattr_init(&fattr);
4938
4939         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4940         if (ret)
4941                 return ret;
4942         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4943                 return -ENOENT;
4944         if (buflen < label.len)
4945                 return -ERANGE;
4946         return 0;
4947 }
4948
4949 static int nfs4_get_security_label(struct inode *inode, void *buf,
4950                                         size_t buflen)
4951 {
4952         struct nfs4_exception exception = { };
4953         int err;
4954
4955         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4956                 return -EOPNOTSUPP;
4957
4958         do {
4959                 err = _nfs4_get_security_label(inode, buf, buflen);
4960                 trace_nfs4_get_security_label(inode, err);
4961                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4962                                 &exception);
4963         } while (exception.retry);
4964         return err;
4965 }
4966
4967 static int _nfs4_do_set_security_label(struct inode *inode,
4968                 struct nfs4_label *ilabel,
4969                 struct nfs_fattr *fattr,
4970                 struct nfs4_label *olabel)
4971 {
4972
4973         struct iattr sattr = {0};
4974         struct nfs_server *server = NFS_SERVER(inode);
4975         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4976         struct nfs_setattrargs arg = {
4977                 .fh             = NFS_FH(inode),
4978                 .iap            = &sattr,
4979                 .server         = server,
4980                 .bitmask        = bitmask,
4981                 .label          = ilabel,
4982         };
4983         struct nfs_setattrres res = {
4984                 .fattr          = fattr,
4985                 .label          = olabel,
4986                 .server         = server,
4987         };
4988         struct rpc_message msg = {
4989                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4990                 .rpc_argp       = &arg,
4991                 .rpc_resp       = &res,
4992         };
4993         int status;
4994
4995         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4996
4997         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4998         if (status)
4999                 dprintk("%s failed: %d\n", __func__, status);
5000
5001         return status;
5002 }
5003
5004 static int nfs4_do_set_security_label(struct inode *inode,
5005                 struct nfs4_label *ilabel,
5006                 struct nfs_fattr *fattr,
5007                 struct nfs4_label *olabel)
5008 {
5009         struct nfs4_exception exception = { };
5010         int err;
5011
5012         do {
5013                 err = _nfs4_do_set_security_label(inode, ilabel,
5014                                 fattr, olabel);
5015                 trace_nfs4_set_security_label(inode, err);
5016                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5017                                 &exception);
5018         } while (exception.retry);
5019         return err;
5020 }
5021
5022 static int
5023 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5024 {
5025         struct nfs4_label ilabel, *olabel = NULL;
5026         struct nfs_fattr fattr;
5027         struct rpc_cred *cred;
5028         int status;
5029
5030         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5031                 return -EOPNOTSUPP;
5032
5033         nfs_fattr_init(&fattr);
5034
5035         ilabel.pi = 0;
5036         ilabel.lfs = 0;
5037         ilabel.label = (char *)buf;
5038         ilabel.len = buflen;
5039
5040         cred = rpc_lookup_cred();
5041         if (IS_ERR(cred))
5042                 return PTR_ERR(cred);
5043
5044         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5045         if (IS_ERR(olabel)) {
5046                 status = -PTR_ERR(olabel);
5047                 goto out;
5048         }
5049
5050         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5051         if (status == 0)
5052                 nfs_setsecurity(inode, &fattr, olabel);
5053
5054         nfs4_label_free(olabel);
5055 out:
5056         put_rpccred(cred);
5057         return status;
5058 }
5059 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5060
5061
5062 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5063                                     nfs4_verifier *bootverf)
5064 {
5065         __be32 verf[2];
5066
5067         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5068                 /* An impossible timestamp guarantees this value
5069                  * will never match a generated boot time. */
5070                 verf[0] = 0;
5071                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5072         } else {
5073                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5074                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5075                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5076         }
5077         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5078 }
5079
5080 static int
5081 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5082 {
5083         size_t len;
5084         char *str;
5085
5086         if (clp->cl_owner_id != NULL)
5087                 return 0;
5088
5089         rcu_read_lock();
5090         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5091                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5092                 1 +
5093                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5094                 1;
5095         rcu_read_unlock();
5096
5097         if (len > NFS4_OPAQUE_LIMIT + 1)
5098                 return -EINVAL;
5099
5100         /*
5101          * Since this string is allocated at mount time, and held until the
5102          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5103          * about a memory-reclaim deadlock.
5104          */
5105         str = kmalloc(len, GFP_KERNEL);
5106         if (!str)
5107                 return -ENOMEM;
5108
5109         rcu_read_lock();
5110         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5111                         clp->cl_ipaddr,
5112                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5113                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5114         rcu_read_unlock();
5115
5116         clp->cl_owner_id = str;
5117         return 0;
5118 }
5119
5120 static int
5121 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5122 {
5123         size_t len;
5124         char *str;
5125
5126         len = 10 + 10 + 1 + 10 + 1 +
5127                 strlen(nfs4_client_id_uniquifier) + 1 +
5128                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5129
5130         if (len > NFS4_OPAQUE_LIMIT + 1)
5131                 return -EINVAL;
5132
5133         /*
5134          * Since this string is allocated at mount time, and held until the
5135          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5136          * about a memory-reclaim deadlock.
5137          */
5138         str = kmalloc(len, GFP_KERNEL);
5139         if (!str)
5140                 return -ENOMEM;
5141
5142         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5143                         clp->rpc_ops->version, clp->cl_minorversion,
5144                         nfs4_client_id_uniquifier,
5145                         clp->cl_rpcclient->cl_nodename);
5146         clp->cl_owner_id = str;
5147         return 0;
5148 }
5149
5150 static int
5151 nfs4_init_uniform_client_string(struct nfs_client *clp)
5152 {
5153         size_t len;
5154         char *str;
5155
5156         if (clp->cl_owner_id != NULL)
5157                 return 0;
5158
5159         if (nfs4_client_id_uniquifier[0] != '\0')
5160                 return nfs4_init_uniquifier_client_string(clp);
5161
5162         len = 10 + 10 + 1 + 10 + 1 +
5163                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5164
5165         if (len > NFS4_OPAQUE_LIMIT + 1)
5166                 return -EINVAL;
5167
5168         /*
5169          * Since this string is allocated at mount time, and held until the
5170          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5171          * about a memory-reclaim deadlock.
5172          */
5173         str = kmalloc(len, GFP_KERNEL);
5174         if (!str)
5175                 return -ENOMEM;
5176
5177         scnprintf(str, len, "Linux NFSv%u.%u %s",
5178                         clp->rpc_ops->version, clp->cl_minorversion,
5179                         clp->cl_rpcclient->cl_nodename);
5180         clp->cl_owner_id = str;
5181         return 0;
5182 }
5183
5184 /*
5185  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5186  * services.  Advertise one based on the address family of the
5187  * clientaddr.
5188  */
5189 static unsigned int
5190 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5191 {
5192         if (strchr(clp->cl_ipaddr, ':') != NULL)
5193                 return scnprintf(buf, len, "tcp6");
5194         else
5195                 return scnprintf(buf, len, "tcp");
5196 }
5197
5198 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5199 {
5200         struct nfs4_setclientid *sc = calldata;
5201
5202         if (task->tk_status == 0)
5203                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5204 }
5205
5206 static const struct rpc_call_ops nfs4_setclientid_ops = {
5207         .rpc_call_done = nfs4_setclientid_done,
5208 };
5209
5210 /**
5211  * nfs4_proc_setclientid - Negotiate client ID
5212  * @clp: state data structure
5213  * @program: RPC program for NFSv4 callback service
5214  * @port: IP port number for NFS4 callback service
5215  * @cred: RPC credential to use for this call
5216  * @res: where to place the result
5217  *
5218  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5219  */
5220 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5221                 unsigned short port, struct rpc_cred *cred,
5222                 struct nfs4_setclientid_res *res)
5223 {
5224         nfs4_verifier sc_verifier;
5225         struct nfs4_setclientid setclientid = {
5226                 .sc_verifier = &sc_verifier,
5227                 .sc_prog = program,
5228                 .sc_clnt = clp,
5229         };
5230         struct rpc_message msg = {
5231                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5232                 .rpc_argp = &setclientid,
5233                 .rpc_resp = res,
5234                 .rpc_cred = cred,
5235         };
5236         struct rpc_task *task;
5237         struct rpc_task_setup task_setup_data = {
5238                 .rpc_client = clp->cl_rpcclient,
5239                 .rpc_message = &msg,
5240                 .callback_ops = &nfs4_setclientid_ops,
5241                 .callback_data = &setclientid,
5242                 .flags = RPC_TASK_TIMEOUT,
5243         };
5244         int status;
5245
5246         /* nfs_client_id4 */
5247         nfs4_init_boot_verifier(clp, &sc_verifier);
5248
5249         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5250                 status = nfs4_init_uniform_client_string(clp);
5251         else
5252                 status = nfs4_init_nonuniform_client_string(clp);
5253
5254         if (status)
5255                 goto out;
5256
5257         /* cb_client4 */
5258         setclientid.sc_netid_len =
5259                                 nfs4_init_callback_netid(clp,
5260                                                 setclientid.sc_netid,
5261                                                 sizeof(setclientid.sc_netid));
5262         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5263                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5264                                 clp->cl_ipaddr, port >> 8, port & 255);
5265
5266         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5267                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5268                 clp->cl_owner_id);
5269         task = rpc_run_task(&task_setup_data);
5270         if (IS_ERR(task)) {
5271                 status = PTR_ERR(task);
5272                 goto out;
5273         }
5274         status = task->tk_status;
5275         if (setclientid.sc_cred) {
5276                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5277                 put_rpccred(setclientid.sc_cred);
5278         }
5279         rpc_put_task(task);
5280 out:
5281         trace_nfs4_setclientid(clp, status);
5282         dprintk("NFS reply setclientid: %d\n", status);
5283         return status;
5284 }
5285
5286 /**
5287  * nfs4_proc_setclientid_confirm - Confirm client ID
5288  * @clp: state data structure
5289  * @res: result of a previous SETCLIENTID
5290  * @cred: RPC credential to use for this call
5291  *
5292  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5293  */
5294 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5295                 struct nfs4_setclientid_res *arg,
5296                 struct rpc_cred *cred)
5297 {
5298         struct rpc_message msg = {
5299                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5300                 .rpc_argp = arg,
5301                 .rpc_cred = cred,
5302         };
5303         int status;
5304
5305         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5306                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5307                 clp->cl_clientid);
5308         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5309         trace_nfs4_setclientid_confirm(clp, status);
5310         dprintk("NFS reply setclientid_confirm: %d\n", status);
5311         return status;
5312 }
5313
5314 struct nfs4_delegreturndata {
5315         struct nfs4_delegreturnargs args;
5316         struct nfs4_delegreturnres res;
5317         struct nfs_fh fh;
5318         nfs4_stateid stateid;
5319         unsigned long timestamp;
5320         struct nfs_fattr fattr;
5321         int rpc_status;
5322         struct inode *inode;
5323         bool roc;
5324         u32 roc_barrier;
5325 };
5326
5327 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5328 {
5329         struct nfs4_delegreturndata *data = calldata;
5330
5331         if (!nfs4_sequence_done(task, &data->res.seq_res))
5332                 return;
5333
5334         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5335         switch (task->tk_status) {
5336         case 0:
5337                 renew_lease(data->res.server, data->timestamp);
5338         case -NFS4ERR_ADMIN_REVOKED:
5339         case -NFS4ERR_DELEG_REVOKED:
5340         case -NFS4ERR_BAD_STATEID:
5341         case -NFS4ERR_OLD_STATEID:
5342         case -NFS4ERR_STALE_STATEID:
5343         case -NFS4ERR_EXPIRED:
5344                 task->tk_status = 0;
5345                 if (data->roc)
5346                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5347                 break;
5348         default:
5349                 if (nfs4_async_handle_error(task, data->res.server,
5350                                             NULL, NULL) == -EAGAIN) {
5351                         rpc_restart_call_prepare(task);
5352                         return;
5353                 }
5354         }
5355         data->rpc_status = task->tk_status;
5356 }
5357
5358 static void nfs4_delegreturn_release(void *calldata)
5359 {
5360         struct nfs4_delegreturndata *data = calldata;
5361         struct inode *inode = data->inode;
5362
5363         if (inode) {
5364                 if (data->roc)
5365                         pnfs_roc_release(inode);
5366                 nfs_iput_and_deactive(inode);
5367         }
5368         kfree(calldata);
5369 }
5370
5371 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5372 {
5373         struct nfs4_delegreturndata *d_data;
5374
5375         d_data = (struct nfs4_delegreturndata *)data;
5376
5377         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5378                 return;
5379
5380         if (d_data->roc)
5381                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5382
5383         nfs4_setup_sequence(d_data->res.server,
5384                         &d_data->args.seq_args,
5385                         &d_data->res.seq_res,
5386                         task);
5387 }
5388
5389 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5390         .rpc_call_prepare = nfs4_delegreturn_prepare,
5391         .rpc_call_done = nfs4_delegreturn_done,
5392         .rpc_release = nfs4_delegreturn_release,
5393 };
5394
5395 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5396 {
5397         struct nfs4_delegreturndata *data;
5398         struct nfs_server *server = NFS_SERVER(inode);
5399         struct rpc_task *task;
5400         struct rpc_message msg = {
5401                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5402                 .rpc_cred = cred,
5403         };
5404         struct rpc_task_setup task_setup_data = {
5405                 .rpc_client = server->client,
5406                 .rpc_message = &msg,
5407                 .callback_ops = &nfs4_delegreturn_ops,
5408                 .flags = RPC_TASK_ASYNC,
5409         };
5410         int status = 0;
5411
5412         data = kzalloc(sizeof(*data), GFP_NOFS);
5413         if (data == NULL)
5414                 return -ENOMEM;
5415         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5416
5417         nfs4_state_protect(server->nfs_client,
5418                         NFS_SP4_MACH_CRED_CLEANUP,
5419                         &task_setup_data.rpc_client, &msg);
5420
5421         data->args.fhandle = &data->fh;
5422         data->args.stateid = &data->stateid;
5423         data->args.bitmask = server->cache_consistency_bitmask;
5424         nfs_copy_fh(&data->fh, NFS_FH(inode));
5425         nfs4_stateid_copy(&data->stateid, stateid);
5426         data->res.fattr = &data->fattr;
5427         data->res.server = server;
5428         nfs_fattr_init(data->res.fattr);
5429         data->timestamp = jiffies;
5430         data->rpc_status = 0;
5431         data->inode = nfs_igrab_and_active(inode);
5432         if (data->inode)
5433                 data->roc = nfs4_roc(inode);
5434
5435         task_setup_data.callback_data = data;
5436         msg.rpc_argp = &data->args;
5437         msg.rpc_resp = &data->res;
5438         task = rpc_run_task(&task_setup_data);
5439         if (IS_ERR(task))
5440                 return PTR_ERR(task);
5441         if (!issync)
5442                 goto out;
5443         status = nfs4_wait_for_completion_rpc_task(task);
5444         if (status != 0)
5445                 goto out;
5446         status = data->rpc_status;
5447         if (status == 0)
5448                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5449         else
5450                 nfs_refresh_inode(inode, &data->fattr);
5451 out:
5452         rpc_put_task(task);
5453         return status;
5454 }
5455
5456 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5457 {
5458         struct nfs_server *server = NFS_SERVER(inode);
5459         struct nfs4_exception exception = { };
5460         int err;
5461         do {
5462                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5463                 trace_nfs4_delegreturn(inode, stateid, err);
5464                 switch (err) {
5465                         case -NFS4ERR_STALE_STATEID:
5466                         case -NFS4ERR_EXPIRED:
5467                         case 0:
5468                                 return 0;
5469                 }
5470                 err = nfs4_handle_exception(server, err, &exception);
5471         } while (exception.retry);
5472         return err;
5473 }
5474
5475 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5476 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5477
5478 /* 
5479  * sleep, with exponential backoff, and retry the LOCK operation. 
5480  */
5481 static unsigned long
5482 nfs4_set_lock_task_retry(unsigned long timeout)
5483 {
5484         freezable_schedule_timeout_killable_unsafe(timeout);
5485         timeout <<= 1;
5486         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5487                 return NFS4_LOCK_MAXTIMEOUT;
5488         return timeout;
5489 }
5490
5491 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5492 {
5493         struct inode *inode = state->inode;
5494         struct nfs_server *server = NFS_SERVER(inode);
5495         struct nfs_client *clp = server->nfs_client;
5496         struct nfs_lockt_args arg = {
5497                 .fh = NFS_FH(inode),
5498                 .fl = request,
5499         };
5500         struct nfs_lockt_res res = {
5501                 .denied = request,
5502         };
5503         struct rpc_message msg = {
5504                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5505                 .rpc_argp       = &arg,
5506                 .rpc_resp       = &res,
5507                 .rpc_cred       = state->owner->so_cred,
5508         };
5509         struct nfs4_lock_state *lsp;
5510         int status;
5511
5512         arg.lock_owner.clientid = clp->cl_clientid;
5513         status = nfs4_set_lock_state(state, request);
5514         if (status != 0)
5515                 goto out;
5516         lsp = request->fl_u.nfs4_fl.owner;
5517         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5518         arg.lock_owner.s_dev = server->s_dev;
5519         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5520         switch (status) {
5521                 case 0:
5522                         request->fl_type = F_UNLCK;
5523                         break;
5524                 case -NFS4ERR_DENIED:
5525                         status = 0;
5526         }
5527         request->fl_ops->fl_release_private(request);
5528         request->fl_ops = NULL;
5529 out:
5530         return status;
5531 }
5532
5533 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5534 {
5535         struct nfs4_exception exception = { };
5536         int err;
5537
5538         do {
5539                 err = _nfs4_proc_getlk(state, cmd, request);
5540                 trace_nfs4_get_lock(request, state, cmd, err);
5541                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5542                                 &exception);
5543         } while (exception.retry);
5544         return err;
5545 }
5546
5547 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5548 {
5549         return locks_lock_inode_wait(inode, fl);
5550 }
5551
5552 struct nfs4_unlockdata {
5553         struct nfs_locku_args arg;
5554         struct nfs_locku_res res;
5555         struct nfs4_lock_state *lsp;
5556         struct nfs_open_context *ctx;
5557         struct file_lock fl;
5558         struct nfs_server *server;
5559         unsigned long timestamp;
5560 };
5561
5562 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5563                 struct nfs_open_context *ctx,
5564                 struct nfs4_lock_state *lsp,
5565                 struct nfs_seqid *seqid)
5566 {
5567         struct nfs4_unlockdata *p;
5568         struct inode *inode = lsp->ls_state->inode;
5569
5570         p = kzalloc(sizeof(*p), GFP_NOFS);
5571         if (p == NULL)
5572                 return NULL;
5573         p->arg.fh = NFS_FH(inode);
5574         p->arg.fl = &p->fl;
5575         p->arg.seqid = seqid;
5576         p->res.seqid = seqid;
5577         p->lsp = lsp;
5578         atomic_inc(&lsp->ls_count);
5579         /* Ensure we don't close file until we're done freeing locks! */
5580         p->ctx = get_nfs_open_context(ctx);
5581         memcpy(&p->fl, fl, sizeof(p->fl));
5582         p->server = NFS_SERVER(inode);
5583         return p;
5584 }
5585
5586 static void nfs4_locku_release_calldata(void *data)
5587 {
5588         struct nfs4_unlockdata *calldata = data;
5589         nfs_free_seqid(calldata->arg.seqid);
5590         nfs4_put_lock_state(calldata->lsp);
5591         put_nfs_open_context(calldata->ctx);
5592         kfree(calldata);
5593 }
5594
5595 static void nfs4_locku_done(struct rpc_task *task, void *data)
5596 {
5597         struct nfs4_unlockdata *calldata = data;
5598
5599         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5600                 return;
5601         switch (task->tk_status) {
5602                 case 0:
5603                         renew_lease(calldata->server, calldata->timestamp);
5604                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5605                         if (nfs4_update_lock_stateid(calldata->lsp,
5606                                         &calldata->res.stateid))
5607                                 break;
5608                 case -NFS4ERR_BAD_STATEID:
5609                 case -NFS4ERR_OLD_STATEID:
5610                 case -NFS4ERR_STALE_STATEID:
5611                 case -NFS4ERR_EXPIRED:
5612                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5613                                                 &calldata->lsp->ls_stateid))
5614                                 rpc_restart_call_prepare(task);
5615                         break;
5616                 default:
5617                         if (nfs4_async_handle_error(task, calldata->server,
5618                                                     NULL, NULL) == -EAGAIN)
5619                                 rpc_restart_call_prepare(task);
5620         }
5621         nfs_release_seqid(calldata->arg.seqid);
5622 }
5623
5624 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5625 {
5626         struct nfs4_unlockdata *calldata = data;
5627
5628         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5629                 goto out_wait;
5630         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5631         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5632                 /* Note: exit _without_ running nfs4_locku_done */
5633                 goto out_no_action;
5634         }
5635         calldata->timestamp = jiffies;
5636         if (nfs4_setup_sequence(calldata->server,
5637                                 &calldata->arg.seq_args,
5638                                 &calldata->res.seq_res,
5639                                 task) != 0)
5640                 nfs_release_seqid(calldata->arg.seqid);
5641         return;
5642 out_no_action:
5643         task->tk_action = NULL;
5644 out_wait:
5645         nfs4_sequence_done(task, &calldata->res.seq_res);
5646 }
5647
5648 static const struct rpc_call_ops nfs4_locku_ops = {
5649         .rpc_call_prepare = nfs4_locku_prepare,
5650         .rpc_call_done = nfs4_locku_done,
5651         .rpc_release = nfs4_locku_release_calldata,
5652 };
5653
5654 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5655                 struct nfs_open_context *ctx,
5656                 struct nfs4_lock_state *lsp,
5657                 struct nfs_seqid *seqid)
5658 {
5659         struct nfs4_unlockdata *data;
5660         struct rpc_message msg = {
5661                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5662                 .rpc_cred = ctx->cred,
5663         };
5664         struct rpc_task_setup task_setup_data = {
5665                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5666                 .rpc_message = &msg,
5667                 .callback_ops = &nfs4_locku_ops,
5668                 .workqueue = nfsiod_workqueue,
5669                 .flags = RPC_TASK_ASYNC,
5670         };
5671
5672         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5673                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5674
5675         /* Ensure this is an unlock - when canceling a lock, the
5676          * canceled lock is passed in, and it won't be an unlock.
5677          */
5678         fl->fl_type = F_UNLCK;
5679
5680         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5681         if (data == NULL) {
5682                 nfs_free_seqid(seqid);
5683                 return ERR_PTR(-ENOMEM);
5684         }
5685
5686         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5687         msg.rpc_argp = &data->arg;
5688         msg.rpc_resp = &data->res;
5689         task_setup_data.callback_data = data;
5690         return rpc_run_task(&task_setup_data);
5691 }
5692
5693 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5694 {
5695         struct inode *inode = state->inode;
5696         struct nfs4_state_owner *sp = state->owner;
5697         struct nfs_inode *nfsi = NFS_I(inode);
5698         struct nfs_seqid *seqid;
5699         struct nfs4_lock_state *lsp;
5700         struct rpc_task *task;
5701         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5702         int status = 0;
5703         unsigned char fl_flags = request->fl_flags;
5704
5705         status = nfs4_set_lock_state(state, request);
5706         /* Unlock _before_ we do the RPC call */
5707         request->fl_flags |= FL_EXISTS;
5708         /* Exclude nfs_delegation_claim_locks() */
5709         mutex_lock(&sp->so_delegreturn_mutex);
5710         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5711         down_read(&nfsi->rwsem);
5712         if (do_vfs_lock(inode, request) == -ENOENT) {
5713                 up_read(&nfsi->rwsem);
5714                 mutex_unlock(&sp->so_delegreturn_mutex);
5715                 goto out;
5716         }
5717         up_read(&nfsi->rwsem);
5718         mutex_unlock(&sp->so_delegreturn_mutex);
5719         if (status != 0)
5720                 goto out;
5721         /* Is this a delegated lock? */
5722         lsp = request->fl_u.nfs4_fl.owner;
5723         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5724                 goto out;
5725         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5726         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5727         status = -ENOMEM;
5728         if (IS_ERR(seqid))
5729                 goto out;
5730         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5731         status = PTR_ERR(task);
5732         if (IS_ERR(task))
5733                 goto out;
5734         status = nfs4_wait_for_completion_rpc_task(task);
5735         rpc_put_task(task);
5736 out:
5737         request->fl_flags = fl_flags;
5738         trace_nfs4_unlock(request, state, F_SETLK, status);
5739         return status;
5740 }
5741
5742 struct nfs4_lockdata {
5743         struct nfs_lock_args arg;
5744         struct nfs_lock_res res;
5745         struct nfs4_lock_state *lsp;
5746         struct nfs_open_context *ctx;
5747         struct file_lock fl;
5748         unsigned long timestamp;
5749         int rpc_status;
5750         int cancelled;
5751         struct nfs_server *server;
5752 };
5753
5754 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5755                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5756                 gfp_t gfp_mask)
5757 {
5758         struct nfs4_lockdata *p;
5759         struct inode *inode = lsp->ls_state->inode;
5760         struct nfs_server *server = NFS_SERVER(inode);
5761         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5762
5763         p = kzalloc(sizeof(*p), gfp_mask);
5764         if (p == NULL)
5765                 return NULL;
5766
5767         p->arg.fh = NFS_FH(inode);
5768         p->arg.fl = &p->fl;
5769         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5770         if (IS_ERR(p->arg.open_seqid))
5771                 goto out_free;
5772         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5773         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5774         if (IS_ERR(p->arg.lock_seqid))
5775                 goto out_free_seqid;
5776         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5777         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5778         p->arg.lock_owner.s_dev = server->s_dev;
5779         p->res.lock_seqid = p->arg.lock_seqid;
5780         p->lsp = lsp;
5781         p->server = server;
5782         atomic_inc(&lsp->ls_count);
5783         p->ctx = get_nfs_open_context(ctx);
5784         get_file(fl->fl_file);
5785         memcpy(&p->fl, fl, sizeof(p->fl));
5786         return p;
5787 out_free_seqid:
5788         nfs_free_seqid(p->arg.open_seqid);
5789 out_free:
5790         kfree(p);
5791         return NULL;
5792 }
5793
5794 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5795 {
5796         struct nfs4_lockdata *data = calldata;
5797         struct nfs4_state *state = data->lsp->ls_state;
5798
5799         dprintk("%s: begin!\n", __func__);
5800         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5801                 goto out_wait;
5802         /* Do we need to do an open_to_lock_owner? */
5803         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5804                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5805                         goto out_release_lock_seqid;
5806                 }
5807                 nfs4_stateid_copy(&data->arg.open_stateid,
5808                                 &state->open_stateid);
5809                 data->arg.new_lock_owner = 1;
5810                 data->res.open_seqid = data->arg.open_seqid;
5811         } else {
5812                 data->arg.new_lock_owner = 0;
5813                 nfs4_stateid_copy(&data->arg.lock_stateid,
5814                                 &data->lsp->ls_stateid);
5815         }
5816         if (!nfs4_valid_open_stateid(state)) {
5817                 data->rpc_status = -EBADF;
5818                 task->tk_action = NULL;
5819                 goto out_release_open_seqid;
5820         }
5821         data->timestamp = jiffies;
5822         if (nfs4_setup_sequence(data->server,
5823                                 &data->arg.seq_args,
5824                                 &data->res.seq_res,
5825                                 task) == 0)
5826                 return;
5827 out_release_open_seqid:
5828         nfs_release_seqid(data->arg.open_seqid);
5829 out_release_lock_seqid:
5830         nfs_release_seqid(data->arg.lock_seqid);
5831 out_wait:
5832         nfs4_sequence_done(task, &data->res.seq_res);
5833         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5834 }
5835
5836 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5837 {
5838         struct nfs4_lockdata *data = calldata;
5839         struct nfs4_lock_state *lsp = data->lsp;
5840
5841         dprintk("%s: begin!\n", __func__);
5842
5843         if (!nfs4_sequence_done(task, &data->res.seq_res))
5844                 return;
5845
5846         data->rpc_status = task->tk_status;
5847         switch (task->tk_status) {
5848         case 0:
5849                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5850                                 data->timestamp);
5851                 if (data->arg.new_lock) {
5852                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5853                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5854                                 rpc_restart_call_prepare(task);
5855                                 break;
5856                         }
5857                 }
5858                 if (data->arg.new_lock_owner != 0) {
5859                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5860                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5861                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5862                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5863                         rpc_restart_call_prepare(task);
5864                 break;
5865         case -NFS4ERR_BAD_STATEID:
5866         case -NFS4ERR_OLD_STATEID:
5867         case -NFS4ERR_STALE_STATEID:
5868         case -NFS4ERR_EXPIRED:
5869                 if (data->arg.new_lock_owner != 0) {
5870                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5871                                                 &lsp->ls_state->open_stateid))
5872                                 rpc_restart_call_prepare(task);
5873                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5874                                                 &lsp->ls_stateid))
5875                                 rpc_restart_call_prepare(task);
5876         }
5877         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5878 }
5879
5880 static void nfs4_lock_release(void *calldata)
5881 {
5882         struct nfs4_lockdata *data = calldata;
5883
5884         dprintk("%s: begin!\n", __func__);
5885         nfs_free_seqid(data->arg.open_seqid);
5886         if (data->cancelled != 0) {
5887                 struct rpc_task *task;
5888                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5889                                 data->arg.lock_seqid);
5890                 if (!IS_ERR(task))
5891                         rpc_put_task_async(task);
5892                 dprintk("%s: cancelling lock!\n", __func__);
5893         } else
5894                 nfs_free_seqid(data->arg.lock_seqid);
5895         nfs4_put_lock_state(data->lsp);
5896         put_nfs_open_context(data->ctx);
5897         fput(data->fl.fl_file);
5898         kfree(data);
5899         dprintk("%s: done!\n", __func__);
5900 }
5901
5902 static const struct rpc_call_ops nfs4_lock_ops = {
5903         .rpc_call_prepare = nfs4_lock_prepare,
5904         .rpc_call_done = nfs4_lock_done,
5905         .rpc_release = nfs4_lock_release,
5906 };
5907
5908 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5909 {
5910         switch (error) {
5911         case -NFS4ERR_ADMIN_REVOKED:
5912         case -NFS4ERR_BAD_STATEID:
5913                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5914                 if (new_lock_owner != 0 ||
5915                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5916                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5917                 break;
5918         case -NFS4ERR_STALE_STATEID:
5919                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5920         case -NFS4ERR_EXPIRED:
5921                 nfs4_schedule_lease_recovery(server->nfs_client);
5922         };
5923 }
5924
5925 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5926 {
5927         struct nfs4_lockdata *data;
5928         struct rpc_task *task;
5929         struct rpc_message msg = {
5930                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5931                 .rpc_cred = state->owner->so_cred,
5932         };
5933         struct rpc_task_setup task_setup_data = {
5934                 .rpc_client = NFS_CLIENT(state->inode),
5935                 .rpc_message = &msg,
5936                 .callback_ops = &nfs4_lock_ops,
5937                 .workqueue = nfsiod_workqueue,
5938                 .flags = RPC_TASK_ASYNC,
5939         };
5940         int ret;
5941
5942         dprintk("%s: begin!\n", __func__);
5943         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5944                         fl->fl_u.nfs4_fl.owner,
5945                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5946         if (data == NULL)
5947                 return -ENOMEM;
5948         if (IS_SETLKW(cmd))
5949                 data->arg.block = 1;
5950         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5951         msg.rpc_argp = &data->arg;
5952         msg.rpc_resp = &data->res;
5953         task_setup_data.callback_data = data;
5954         if (recovery_type > NFS_LOCK_NEW) {
5955                 if (recovery_type == NFS_LOCK_RECLAIM)
5956                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5957                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5958         } else
5959                 data->arg.new_lock = 1;
5960         task = rpc_run_task(&task_setup_data);
5961         if (IS_ERR(task))
5962                 return PTR_ERR(task);
5963         ret = nfs4_wait_for_completion_rpc_task(task);
5964         if (ret == 0) {
5965                 ret = data->rpc_status;
5966                 if (ret)
5967                         nfs4_handle_setlk_error(data->server, data->lsp,
5968                                         data->arg.new_lock_owner, ret);
5969         } else
5970                 data->cancelled = 1;
5971         rpc_put_task(task);
5972         dprintk("%s: done, ret = %d!\n", __func__, ret);
5973         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
5974         return ret;
5975 }
5976
5977 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5978 {
5979         struct nfs_server *server = NFS_SERVER(state->inode);
5980         struct nfs4_exception exception = {
5981                 .inode = state->inode,
5982         };
5983         int err;
5984
5985         do {
5986                 /* Cache the lock if possible... */
5987                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5988                         return 0;
5989                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5990                 if (err != -NFS4ERR_DELAY)
5991                         break;
5992                 nfs4_handle_exception(server, err, &exception);
5993         } while (exception.retry);
5994         return err;
5995 }
5996
5997 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5998 {
5999         struct nfs_server *server = NFS_SERVER(state->inode);
6000         struct nfs4_exception exception = {
6001                 .inode = state->inode,
6002         };
6003         int err;
6004
6005         err = nfs4_set_lock_state(state, request);
6006         if (err != 0)
6007                 return err;
6008         if (!recover_lost_locks) {
6009                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6010                 return 0;
6011         }
6012         do {
6013                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6014                         return 0;
6015                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6016                 switch (err) {
6017                 default:
6018                         goto out;
6019                 case -NFS4ERR_GRACE:
6020                 case -NFS4ERR_DELAY:
6021                         nfs4_handle_exception(server, err, &exception);
6022                         err = 0;
6023                 }
6024         } while (exception.retry);
6025 out:
6026         return err;
6027 }
6028
6029 #if defined(CONFIG_NFS_V4_1)
6030 /**
6031  * nfs41_check_expired_locks - possibly free a lock stateid
6032  *
6033  * @state: NFSv4 state for an inode
6034  *
6035  * Returns NFS_OK if recovery for this stateid is now finished.
6036  * Otherwise a negative NFS4ERR value is returned.
6037  */
6038 static int nfs41_check_expired_locks(struct nfs4_state *state)
6039 {
6040         int status, ret = -NFS4ERR_BAD_STATEID;
6041         struct nfs4_lock_state *lsp;
6042         struct nfs_server *server = NFS_SERVER(state->inode);
6043
6044         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6045                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6046                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6047
6048                         status = nfs41_test_stateid(server,
6049                                         &lsp->ls_stateid,
6050                                         cred);
6051                         trace_nfs4_test_lock_stateid(state, lsp, status);
6052                         if (status != NFS_OK) {
6053                                 /* Free the stateid unless the server
6054                                  * informs us the stateid is unrecognized. */
6055                                 if (status != -NFS4ERR_BAD_STATEID)
6056                                         nfs41_free_stateid(server,
6057                                                         &lsp->ls_stateid,
6058                                                         cred);
6059                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6060                                 ret = status;
6061                         }
6062                 }
6063         };
6064
6065         return ret;
6066 }
6067
6068 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6069 {
6070         int status = NFS_OK;
6071
6072         if (test_bit(LK_STATE_IN_USE, &state->flags))
6073                 status = nfs41_check_expired_locks(state);
6074         if (status != NFS_OK)
6075                 status = nfs4_lock_expired(state, request);
6076         return status;
6077 }
6078 #endif
6079
6080 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6081 {
6082         struct nfs_inode *nfsi = NFS_I(state->inode);
6083         struct nfs4_state_owner *sp = state->owner;
6084         unsigned char fl_flags = request->fl_flags;
6085         int status = -ENOLCK;
6086
6087         if ((fl_flags & FL_POSIX) &&
6088                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6089                 goto out;
6090         /* Is this a delegated open? */
6091         status = nfs4_set_lock_state(state, request);
6092         if (status != 0)
6093                 goto out;
6094         request->fl_flags |= FL_ACCESS;
6095         status = do_vfs_lock(state->inode, request);
6096         if (status < 0)
6097                 goto out;
6098         mutex_lock(&sp->so_delegreturn_mutex);
6099         down_read(&nfsi->rwsem);
6100         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6101                 /* Yes: cache locks! */
6102                 /* ...but avoid races with delegation recall... */
6103                 request->fl_flags = fl_flags & ~FL_SLEEP;
6104                 status = do_vfs_lock(state->inode, request);
6105                 up_read(&nfsi->rwsem);
6106                 mutex_unlock(&sp->so_delegreturn_mutex);
6107                 goto out;
6108         }
6109         up_read(&nfsi->rwsem);
6110         mutex_unlock(&sp->so_delegreturn_mutex);
6111         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6112 out:
6113         request->fl_flags = fl_flags;
6114         return status;
6115 }
6116
6117 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6118 {
6119         struct nfs4_exception exception = {
6120                 .state = state,
6121                 .inode = state->inode,
6122         };
6123         int err;
6124
6125         do {
6126                 err = _nfs4_proc_setlk(state, cmd, request);
6127                 if (err == -NFS4ERR_DENIED)
6128                         err = -EAGAIN;
6129                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6130                                 err, &exception);
6131         } while (exception.retry);
6132         return err;
6133 }
6134
6135 static int
6136 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6137 {
6138         struct nfs_open_context *ctx;
6139         struct nfs4_state *state;
6140         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6141         int status;
6142
6143         /* verify open state */
6144         ctx = nfs_file_open_context(filp);
6145         state = ctx->state;
6146
6147         if (request->fl_start < 0 || request->fl_end < 0)
6148                 return -EINVAL;
6149
6150         if (IS_GETLK(cmd)) {
6151                 if (state != NULL)
6152                         return nfs4_proc_getlk(state, F_GETLK, request);
6153                 return 0;
6154         }
6155
6156         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6157                 return -EINVAL;
6158
6159         if (request->fl_type == F_UNLCK) {
6160                 if (state != NULL)
6161                         return nfs4_proc_unlck(state, cmd, request);
6162                 return 0;
6163         }
6164
6165         if (state == NULL)
6166                 return -ENOLCK;
6167         /*
6168          * Don't rely on the VFS having checked the file open mode,
6169          * since it won't do this for flock() locks.
6170          */
6171         switch (request->fl_type) {
6172         case F_RDLCK:
6173                 if (!(filp->f_mode & FMODE_READ))
6174                         return -EBADF;
6175                 break;
6176         case F_WRLCK:
6177                 if (!(filp->f_mode & FMODE_WRITE))
6178                         return -EBADF;
6179         }
6180
6181         do {
6182                 status = nfs4_proc_setlk(state, cmd, request);
6183                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6184                         break;
6185                 timeout = nfs4_set_lock_task_retry(timeout);
6186                 status = -ERESTARTSYS;
6187                 if (signalled())
6188                         break;
6189         } while(status < 0);
6190         return status;
6191 }
6192
6193 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6194 {
6195         struct nfs_server *server = NFS_SERVER(state->inode);
6196         int err;
6197
6198         err = nfs4_set_lock_state(state, fl);
6199         if (err != 0)
6200                 return err;
6201         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6202         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6203 }
6204
6205 struct nfs_release_lockowner_data {
6206         struct nfs4_lock_state *lsp;
6207         struct nfs_server *server;
6208         struct nfs_release_lockowner_args args;
6209         struct nfs_release_lockowner_res res;
6210         unsigned long timestamp;
6211 };
6212
6213 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6214 {
6215         struct nfs_release_lockowner_data *data = calldata;
6216         struct nfs_server *server = data->server;
6217         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6218                              &data->args.seq_args, &data->res.seq_res, task);
6219         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6220         data->timestamp = jiffies;
6221 }
6222
6223 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6224 {
6225         struct nfs_release_lockowner_data *data = calldata;
6226         struct nfs_server *server = data->server;
6227
6228         nfs40_sequence_done(task, &data->res.seq_res);
6229
6230         switch (task->tk_status) {
6231         case 0:
6232                 renew_lease(server, data->timestamp);
6233                 break;
6234         case -NFS4ERR_STALE_CLIENTID:
6235         case -NFS4ERR_EXPIRED:
6236                 nfs4_schedule_lease_recovery(server->nfs_client);
6237                 break;
6238         case -NFS4ERR_LEASE_MOVED:
6239         case -NFS4ERR_DELAY:
6240                 if (nfs4_async_handle_error(task, server,
6241                                             NULL, NULL) == -EAGAIN)
6242                         rpc_restart_call_prepare(task);
6243         }
6244 }
6245
6246 static void nfs4_release_lockowner_release(void *calldata)
6247 {
6248         struct nfs_release_lockowner_data *data = calldata;
6249         nfs4_free_lock_state(data->server, data->lsp);
6250         kfree(calldata);
6251 }
6252
6253 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6254         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6255         .rpc_call_done = nfs4_release_lockowner_done,
6256         .rpc_release = nfs4_release_lockowner_release,
6257 };
6258
6259 static void
6260 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6261 {
6262         struct nfs_release_lockowner_data *data;
6263         struct rpc_message msg = {
6264                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6265         };
6266
6267         if (server->nfs_client->cl_mvops->minor_version != 0)
6268                 return;
6269
6270         data = kmalloc(sizeof(*data), GFP_NOFS);
6271         if (!data)
6272                 return;
6273         data->lsp = lsp;
6274         data->server = server;
6275         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6276         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6277         data->args.lock_owner.s_dev = server->s_dev;
6278
6279         msg.rpc_argp = &data->args;
6280         msg.rpc_resp = &data->res;
6281         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6282         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6283 }
6284
6285 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6286
6287 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6288                                    struct dentry *unused, struct inode *inode,
6289                                    const char *key, const void *buf,
6290                                    size_t buflen, int flags)
6291 {
6292         return nfs4_proc_set_acl(inode, buf, buflen);
6293 }
6294
6295 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6296                                    struct dentry *unused, struct inode *inode,
6297                                    const char *key, void *buf, size_t buflen)
6298 {
6299         return nfs4_proc_get_acl(inode, buf, buflen);
6300 }
6301
6302 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6303 {
6304         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6305 }
6306
6307 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6308
6309 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6310                                      struct dentry *unused, struct inode *inode,
6311                                      const char *key, const void *buf,
6312                                      size_t buflen, int flags)
6313 {
6314         if (security_ismaclabel(key))
6315                 return nfs4_set_security_label(inode, buf, buflen);
6316
6317         return -EOPNOTSUPP;
6318 }
6319
6320 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6321                                      struct dentry *unused, struct inode *inode,
6322                                      const char *key, void *buf, size_t buflen)
6323 {
6324         if (security_ismaclabel(key))
6325                 return nfs4_get_security_label(inode, buf, buflen);
6326         return -EOPNOTSUPP;
6327 }
6328
6329 static ssize_t
6330 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6331 {
6332         int len = 0;
6333
6334         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6335                 len = security_inode_listsecurity(inode, list, list_len);
6336                 if (list_len && len > list_len)
6337                         return -ERANGE;
6338         }
6339         return len;
6340 }
6341
6342 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6343         .prefix = XATTR_SECURITY_PREFIX,
6344         .get    = nfs4_xattr_get_nfs4_label,
6345         .set    = nfs4_xattr_set_nfs4_label,
6346 };
6347
6348 #else
6349
6350 static ssize_t
6351 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6352 {
6353         return 0;
6354 }
6355
6356 #endif
6357
6358 /*
6359  * nfs_fhget will use either the mounted_on_fileid or the fileid
6360  */
6361 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6362 {
6363         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6364                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6365               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6366               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6367                 return;
6368
6369         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6370                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6371         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6372         fattr->nlink = 2;
6373 }
6374
6375 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6376                                    const struct qstr *name,
6377                                    struct nfs4_fs_locations *fs_locations,
6378                                    struct page *page)
6379 {
6380         struct nfs_server *server = NFS_SERVER(dir);
6381         u32 bitmask[3] = {
6382                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6383         };
6384         struct nfs4_fs_locations_arg args = {
6385                 .dir_fh = NFS_FH(dir),
6386                 .name = name,
6387                 .page = page,
6388                 .bitmask = bitmask,
6389         };
6390         struct nfs4_fs_locations_res res = {
6391                 .fs_locations = fs_locations,
6392         };
6393         struct rpc_message msg = {
6394                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6395                 .rpc_argp = &args,
6396                 .rpc_resp = &res,
6397         };
6398         int status;
6399
6400         dprintk("%s: start\n", __func__);
6401
6402         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6403          * is not supported */
6404         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6405                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6406         else
6407                 bitmask[0] |= FATTR4_WORD0_FILEID;
6408
6409         nfs_fattr_init(&fs_locations->fattr);
6410         fs_locations->server = server;
6411         fs_locations->nlocations = 0;
6412         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6413         dprintk("%s: returned status = %d\n", __func__, status);
6414         return status;
6415 }
6416
6417 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6418                            const struct qstr *name,
6419                            struct nfs4_fs_locations *fs_locations,
6420                            struct page *page)
6421 {
6422         struct nfs4_exception exception = { };
6423         int err;
6424         do {
6425                 err = _nfs4_proc_fs_locations(client, dir, name,
6426                                 fs_locations, page);
6427                 trace_nfs4_get_fs_locations(dir, name, err);
6428                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6429                                 &exception);
6430         } while (exception.retry);
6431         return err;
6432 }
6433
6434 /*
6435  * This operation also signals the server that this client is
6436  * performing migration recovery.  The server can stop returning
6437  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6438  * appended to this compound to identify the client ID which is
6439  * performing recovery.
6440  */
6441 static int _nfs40_proc_get_locations(struct inode *inode,
6442                                      struct nfs4_fs_locations *locations,
6443                                      struct page *page, struct rpc_cred *cred)
6444 {
6445         struct nfs_server *server = NFS_SERVER(inode);
6446         struct rpc_clnt *clnt = server->client;
6447         u32 bitmask[2] = {
6448                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6449         };
6450         struct nfs4_fs_locations_arg args = {
6451                 .clientid       = server->nfs_client->cl_clientid,
6452                 .fh             = NFS_FH(inode),
6453                 .page           = page,
6454                 .bitmask        = bitmask,
6455                 .migration      = 1,            /* skip LOOKUP */
6456                 .renew          = 1,            /* append RENEW */
6457         };
6458         struct nfs4_fs_locations_res res = {
6459                 .fs_locations   = locations,
6460                 .migration      = 1,
6461                 .renew          = 1,
6462         };
6463         struct rpc_message msg = {
6464                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6465                 .rpc_argp       = &args,
6466                 .rpc_resp       = &res,
6467                 .rpc_cred       = cred,
6468         };
6469         unsigned long now = jiffies;
6470         int status;
6471
6472         nfs_fattr_init(&locations->fattr);
6473         locations->server = server;
6474         locations->nlocations = 0;
6475
6476         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6477         nfs4_set_sequence_privileged(&args.seq_args);
6478         status = nfs4_call_sync_sequence(clnt, server, &msg,
6479                                         &args.seq_args, &res.seq_res);
6480         if (status)
6481                 return status;
6482
6483         renew_lease(server, now);
6484         return 0;
6485 }
6486
6487 #ifdef CONFIG_NFS_V4_1
6488
6489 /*
6490  * This operation also signals the server that this client is
6491  * performing migration recovery.  The server can stop asserting
6492  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6493  * performing this operation is identified in the SEQUENCE
6494  * operation in this compound.
6495  *
6496  * When the client supports GETATTR(fs_locations_info), it can
6497  * be plumbed in here.
6498  */
6499 static int _nfs41_proc_get_locations(struct inode *inode,
6500                                      struct nfs4_fs_locations *locations,
6501                                      struct page *page, struct rpc_cred *cred)
6502 {
6503         struct nfs_server *server = NFS_SERVER(inode);
6504         struct rpc_clnt *clnt = server->client;
6505         u32 bitmask[2] = {
6506                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6507         };
6508         struct nfs4_fs_locations_arg args = {
6509                 .fh             = NFS_FH(inode),
6510                 .page           = page,
6511                 .bitmask        = bitmask,
6512                 .migration      = 1,            /* skip LOOKUP */
6513         };
6514         struct nfs4_fs_locations_res res = {
6515                 .fs_locations   = locations,
6516                 .migration      = 1,
6517         };
6518         struct rpc_message msg = {
6519                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6520                 .rpc_argp       = &args,
6521                 .rpc_resp       = &res,
6522                 .rpc_cred       = cred,
6523         };
6524         int status;
6525
6526         nfs_fattr_init(&locations->fattr);
6527         locations->server = server;
6528         locations->nlocations = 0;
6529
6530         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6531         nfs4_set_sequence_privileged(&args.seq_args);
6532         status = nfs4_call_sync_sequence(clnt, server, &msg,
6533                                         &args.seq_args, &res.seq_res);
6534         if (status == NFS4_OK &&
6535             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6536                 status = -NFS4ERR_LEASE_MOVED;
6537         return status;
6538 }
6539
6540 #endif  /* CONFIG_NFS_V4_1 */
6541
6542 /**
6543  * nfs4_proc_get_locations - discover locations for a migrated FSID
6544  * @inode: inode on FSID that is migrating
6545  * @locations: result of query
6546  * @page: buffer
6547  * @cred: credential to use for this operation
6548  *
6549  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6550  * operation failed, or a negative errno if a local error occurred.
6551  *
6552  * On success, "locations" is filled in, but if the server has
6553  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6554  * asserted.
6555  *
6556  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6557  * from this client that require migration recovery.
6558  */
6559 int nfs4_proc_get_locations(struct inode *inode,
6560                             struct nfs4_fs_locations *locations,
6561                             struct page *page, struct rpc_cred *cred)
6562 {
6563         struct nfs_server *server = NFS_SERVER(inode);
6564         struct nfs_client *clp = server->nfs_client;
6565         const struct nfs4_mig_recovery_ops *ops =
6566                                         clp->cl_mvops->mig_recovery_ops;
6567         struct nfs4_exception exception = { };
6568         int status;
6569
6570         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6571                 (unsigned long long)server->fsid.major,
6572                 (unsigned long long)server->fsid.minor,
6573                 clp->cl_hostname);
6574         nfs_display_fhandle(NFS_FH(inode), __func__);
6575
6576         do {
6577                 status = ops->get_locations(inode, locations, page, cred);
6578                 if (status != -NFS4ERR_DELAY)
6579                         break;
6580                 nfs4_handle_exception(server, status, &exception);
6581         } while (exception.retry);
6582         return status;
6583 }
6584
6585 /*
6586  * This operation also signals the server that this client is
6587  * performing "lease moved" recovery.  The server can stop
6588  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6589  * is appended to this compound to identify the client ID which is
6590  * performing recovery.
6591  */
6592 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6593 {
6594         struct nfs_server *server = NFS_SERVER(inode);
6595         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6596         struct rpc_clnt *clnt = server->client;
6597         struct nfs4_fsid_present_arg args = {
6598                 .fh             = NFS_FH(inode),
6599                 .clientid       = clp->cl_clientid,
6600                 .renew          = 1,            /* append RENEW */
6601         };
6602         struct nfs4_fsid_present_res res = {
6603                 .renew          = 1,
6604         };
6605         struct rpc_message msg = {
6606                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6607                 .rpc_argp       = &args,
6608                 .rpc_resp       = &res,
6609                 .rpc_cred       = cred,
6610         };
6611         unsigned long now = jiffies;
6612         int status;
6613
6614         res.fh = nfs_alloc_fhandle();
6615         if (res.fh == NULL)
6616                 return -ENOMEM;
6617
6618         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6619         nfs4_set_sequence_privileged(&args.seq_args);
6620         status = nfs4_call_sync_sequence(clnt, server, &msg,
6621                                                 &args.seq_args, &res.seq_res);
6622         nfs_free_fhandle(res.fh);
6623         if (status)
6624                 return status;
6625
6626         do_renew_lease(clp, now);
6627         return 0;
6628 }
6629
6630 #ifdef CONFIG_NFS_V4_1
6631
6632 /*
6633  * This operation also signals the server that this client is
6634  * performing "lease moved" recovery.  The server can stop asserting
6635  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6636  * this operation is identified in the SEQUENCE operation in this
6637  * compound.
6638  */
6639 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6640 {
6641         struct nfs_server *server = NFS_SERVER(inode);
6642         struct rpc_clnt *clnt = server->client;
6643         struct nfs4_fsid_present_arg args = {
6644                 .fh             = NFS_FH(inode),
6645         };
6646         struct nfs4_fsid_present_res res = {
6647         };
6648         struct rpc_message msg = {
6649                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6650                 .rpc_argp       = &args,
6651                 .rpc_resp       = &res,
6652                 .rpc_cred       = cred,
6653         };
6654         int status;
6655
6656         res.fh = nfs_alloc_fhandle();
6657         if (res.fh == NULL)
6658                 return -ENOMEM;
6659
6660         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6661         nfs4_set_sequence_privileged(&args.seq_args);
6662         status = nfs4_call_sync_sequence(clnt, server, &msg,
6663                                                 &args.seq_args, &res.seq_res);
6664         nfs_free_fhandle(res.fh);
6665         if (status == NFS4_OK &&
6666             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6667                 status = -NFS4ERR_LEASE_MOVED;
6668         return status;
6669 }
6670
6671 #endif  /* CONFIG_NFS_V4_1 */
6672
6673 /**
6674  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6675  * @inode: inode on FSID to check
6676  * @cred: credential to use for this operation
6677  *
6678  * Server indicates whether the FSID is present, moved, or not
6679  * recognized.  This operation is necessary to clear a LEASE_MOVED
6680  * condition for this client ID.
6681  *
6682  * Returns NFS4_OK if the FSID is present on this server,
6683  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6684  *  NFS4ERR code if some error occurred on the server, or a
6685  *  negative errno if a local failure occurred.
6686  */
6687 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6688 {
6689         struct nfs_server *server = NFS_SERVER(inode);
6690         struct nfs_client *clp = server->nfs_client;
6691         const struct nfs4_mig_recovery_ops *ops =
6692                                         clp->cl_mvops->mig_recovery_ops;
6693         struct nfs4_exception exception = { };
6694         int status;
6695
6696         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6697                 (unsigned long long)server->fsid.major,
6698                 (unsigned long long)server->fsid.minor,
6699                 clp->cl_hostname);
6700         nfs_display_fhandle(NFS_FH(inode), __func__);
6701
6702         do {
6703                 status = ops->fsid_present(inode, cred);
6704                 if (status != -NFS4ERR_DELAY)
6705                         break;
6706                 nfs4_handle_exception(server, status, &exception);
6707         } while (exception.retry);
6708         return status;
6709 }
6710
6711 /**
6712  * If 'use_integrity' is true and the state managment nfs_client
6713  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6714  * and the machine credential as per RFC3530bis and RFC5661 Security
6715  * Considerations sections. Otherwise, just use the user cred with the
6716  * filesystem's rpc_client.
6717  */
6718 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6719 {
6720         int status;
6721         struct nfs4_secinfo_arg args = {
6722                 .dir_fh = NFS_FH(dir),
6723                 .name   = name,
6724         };
6725         struct nfs4_secinfo_res res = {
6726                 .flavors     = flavors,
6727         };
6728         struct rpc_message msg = {
6729                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6730                 .rpc_argp = &args,
6731                 .rpc_resp = &res,
6732         };
6733         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6734         struct rpc_cred *cred = NULL;
6735
6736         if (use_integrity) {
6737                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6738                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6739                 msg.rpc_cred = cred;
6740         }
6741
6742         dprintk("NFS call  secinfo %s\n", name->name);
6743
6744         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6745                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6746
6747         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6748                                 &res.seq_res, 0);
6749         dprintk("NFS reply  secinfo: %d\n", status);
6750
6751         if (cred)
6752                 put_rpccred(cred);
6753
6754         return status;
6755 }
6756
6757 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6758                       struct nfs4_secinfo_flavors *flavors)
6759 {
6760         struct nfs4_exception exception = { };
6761         int err;
6762         do {
6763                 err = -NFS4ERR_WRONGSEC;
6764
6765                 /* try to use integrity protection with machine cred */
6766                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6767                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6768
6769                 /*
6770                  * if unable to use integrity protection, or SECINFO with
6771                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6772                  * disallowed by spec, but exists in deployed servers) use
6773                  * the current filesystem's rpc_client and the user cred.
6774                  */
6775                 if (err == -NFS4ERR_WRONGSEC)
6776                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6777
6778                 trace_nfs4_secinfo(dir, name, err);
6779                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6780                                 &exception);
6781         } while (exception.retry);
6782         return err;
6783 }
6784
6785 #ifdef CONFIG_NFS_V4_1
6786 /*
6787  * Check the exchange flags returned by the server for invalid flags, having
6788  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6789  * DS flags set.
6790  */
6791 static int nfs4_check_cl_exchange_flags(u32 flags)
6792 {
6793         if (flags & ~EXCHGID4_FLAG_MASK_R)
6794                 goto out_inval;
6795         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6796             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6797                 goto out_inval;
6798         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6799                 goto out_inval;
6800         return NFS_OK;
6801 out_inval:
6802         return -NFS4ERR_INVAL;
6803 }
6804
6805 static bool
6806 nfs41_same_server_scope(struct nfs41_server_scope *a,
6807                         struct nfs41_server_scope *b)
6808 {
6809         if (a->server_scope_sz == b->server_scope_sz &&
6810             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6811                 return true;
6812
6813         return false;
6814 }
6815
6816 static void
6817 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
6818 {
6819 }
6820
6821 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
6822         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
6823 };
6824
6825 /*
6826  * nfs4_proc_bind_one_conn_to_session()
6827  *
6828  * The 4.1 client currently uses the same TCP connection for the
6829  * fore and backchannel.
6830  */
6831 static
6832 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
6833                 struct rpc_xprt *xprt,
6834                 struct nfs_client *clp,
6835                 struct rpc_cred *cred)
6836 {
6837         int status;
6838         struct nfs41_bind_conn_to_session_args args = {
6839                 .client = clp,
6840                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6841         };
6842         struct nfs41_bind_conn_to_session_res res;
6843         struct rpc_message msg = {
6844                 .rpc_proc =
6845                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6846                 .rpc_argp = &args,
6847                 .rpc_resp = &res,
6848                 .rpc_cred = cred,
6849         };
6850         struct rpc_task_setup task_setup_data = {
6851                 .rpc_client = clnt,
6852                 .rpc_xprt = xprt,
6853                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
6854                 .rpc_message = &msg,
6855                 .flags = RPC_TASK_TIMEOUT,
6856         };
6857         struct rpc_task *task;
6858
6859         dprintk("--> %s\n", __func__);
6860
6861         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6862         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6863                 args.dir = NFS4_CDFC4_FORE;
6864
6865         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
6866         if (xprt != rcu_access_pointer(clnt->cl_xprt))
6867                 args.dir = NFS4_CDFC4_FORE;
6868
6869         task = rpc_run_task(&task_setup_data);
6870         if (!IS_ERR(task)) {
6871                 status = task->tk_status;
6872                 rpc_put_task(task);
6873         } else
6874                 status = PTR_ERR(task);
6875         trace_nfs4_bind_conn_to_session(clp, status);
6876         if (status == 0) {
6877                 if (memcmp(res.sessionid.data,
6878                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6879                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6880                         status = -EIO;
6881                         goto out;
6882                 }
6883                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6884                         dprintk("NFS: %s: Unexpected direction from server\n",
6885                                 __func__);
6886                         status = -EIO;
6887                         goto out;
6888                 }
6889                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6890                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6891                                 __func__);
6892                         status = -EIO;
6893                         goto out;
6894                 }
6895         }
6896 out:
6897         dprintk("<-- %s status= %d\n", __func__, status);
6898         return status;
6899 }
6900
6901 struct rpc_bind_conn_calldata {
6902         struct nfs_client *clp;
6903         struct rpc_cred *cred;
6904 };
6905
6906 static int
6907 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
6908                 struct rpc_xprt *xprt,
6909                 void *calldata)
6910 {
6911         struct rpc_bind_conn_calldata *p = calldata;
6912
6913         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
6914 }
6915
6916 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6917 {
6918         struct rpc_bind_conn_calldata data = {
6919                 .clp = clp,
6920                 .cred = cred,
6921         };
6922         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
6923                         nfs4_proc_bind_conn_to_session_callback, &data);
6924 }
6925
6926 /*
6927  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6928  * and operations we'd like to see to enable certain features in the allow map
6929  */
6930 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6931         .how = SP4_MACH_CRED,
6932         .enforce.u.words = {
6933                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6934                       1 << (OP_EXCHANGE_ID - 32) |
6935                       1 << (OP_CREATE_SESSION - 32) |
6936                       1 << (OP_DESTROY_SESSION - 32) |
6937                       1 << (OP_DESTROY_CLIENTID - 32)
6938         },
6939         .allow.u.words = {
6940                 [0] = 1 << (OP_CLOSE) |
6941                       1 << (OP_OPEN_DOWNGRADE) |
6942                       1 << (OP_LOCKU) |
6943                       1 << (OP_DELEGRETURN) |
6944                       1 << (OP_COMMIT),
6945                 [1] = 1 << (OP_SECINFO - 32) |
6946                       1 << (OP_SECINFO_NO_NAME - 32) |
6947                       1 << (OP_LAYOUTRETURN - 32) |
6948                       1 << (OP_TEST_STATEID - 32) |
6949                       1 << (OP_FREE_STATEID - 32) |
6950                       1 << (OP_WRITE - 32)
6951         }
6952 };
6953
6954 /*
6955  * Select the state protection mode for client `clp' given the server results
6956  * from exchange_id in `sp'.
6957  *
6958  * Returns 0 on success, negative errno otherwise.
6959  */
6960 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6961                                  struct nfs41_state_protection *sp)
6962 {
6963         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6964                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6965                       1 << (OP_EXCHANGE_ID - 32) |
6966                       1 << (OP_CREATE_SESSION - 32) |
6967                       1 << (OP_DESTROY_SESSION - 32) |
6968                       1 << (OP_DESTROY_CLIENTID - 32)
6969         };
6970         unsigned int i;
6971
6972         if (sp->how == SP4_MACH_CRED) {
6973                 /* Print state protect result */
6974                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6975                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6976                         if (test_bit(i, sp->enforce.u.longs))
6977                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6978                         if (test_bit(i, sp->allow.u.longs))
6979                                 dfprintk(MOUNT, "  allow op %d\n", i);
6980                 }
6981
6982                 /* make sure nothing is on enforce list that isn't supported */
6983                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6984                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6985                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6986                                 return -EINVAL;
6987                         }
6988                 }
6989
6990                 /*
6991                  * Minimal mode - state operations are allowed to use machine
6992                  * credential.  Note this already happens by default, so the
6993                  * client doesn't have to do anything more than the negotiation.
6994                  *
6995                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6996                  *       we're already using the machine cred for exchange_id
6997                  *       and will never use a different cred.
6998                  */
6999                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7000                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7001                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7002                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7003                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7004                         dfprintk(MOUNT, "  minimal mode enabled\n");
7005                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7006                 } else {
7007                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7008                         return -EINVAL;
7009                 }
7010
7011                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7012                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7013                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7014                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7015                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7016                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7017                 }
7018
7019                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7020                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7021                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7022                                 &clp->cl_sp4_flags);
7023                 }
7024
7025                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7026                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7027                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7028                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7029                 }
7030
7031                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7032                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7033                         dfprintk(MOUNT, "  stateid mode enabled\n");
7034                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7035                 }
7036
7037                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7038                         dfprintk(MOUNT, "  write mode enabled\n");
7039                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7040                 }
7041
7042                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7043                         dfprintk(MOUNT, "  commit mode enabled\n");
7044                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7045                 }
7046         }
7047
7048         return 0;
7049 }
7050
7051 /*
7052  * _nfs4_proc_exchange_id()
7053  *
7054  * Wrapper for EXCHANGE_ID operation.
7055  */
7056 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7057         u32 sp4_how)
7058 {
7059         nfs4_verifier verifier;
7060         struct nfs41_exchange_id_args args = {
7061                 .verifier = &verifier,
7062                 .client = clp,
7063 #ifdef CONFIG_NFS_V4_1_MIGRATION
7064                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7065                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
7066                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7067 #else
7068                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7069                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
7070 #endif
7071         };
7072         struct nfs41_exchange_id_res res = {
7073                 0
7074         };
7075         int status;
7076         struct rpc_message msg = {
7077                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7078                 .rpc_argp = &args,
7079                 .rpc_resp = &res,
7080                 .rpc_cred = cred,
7081         };
7082
7083         nfs4_init_boot_verifier(clp, &verifier);
7084
7085         status = nfs4_init_uniform_client_string(clp);
7086         if (status)
7087                 goto out;
7088
7089         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7090                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7091                 clp->cl_owner_id);
7092
7093         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7094                                         GFP_NOFS);
7095         if (unlikely(res.server_owner == NULL)) {
7096                 status = -ENOMEM;
7097                 goto out;
7098         }
7099
7100         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7101                                         GFP_NOFS);
7102         if (unlikely(res.server_scope == NULL)) {
7103                 status = -ENOMEM;
7104                 goto out_server_owner;
7105         }
7106
7107         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7108         if (unlikely(res.impl_id == NULL)) {
7109                 status = -ENOMEM;
7110                 goto out_server_scope;
7111         }
7112
7113         switch (sp4_how) {
7114         case SP4_NONE:
7115                 args.state_protect.how = SP4_NONE;
7116                 break;
7117
7118         case SP4_MACH_CRED:
7119                 args.state_protect = nfs4_sp4_mach_cred_request;
7120                 break;
7121
7122         default:
7123                 /* unsupported! */
7124                 WARN_ON_ONCE(1);
7125                 status = -EINVAL;
7126                 goto out_impl_id;
7127         }
7128
7129         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7130         trace_nfs4_exchange_id(clp, status);
7131         if (status == 0)
7132                 status = nfs4_check_cl_exchange_flags(res.flags);
7133
7134         if (status == 0)
7135                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7136
7137         if (status == 0) {
7138                 clp->cl_clientid = res.clientid;
7139                 clp->cl_exchange_flags = res.flags;
7140                 /* Client ID is not confirmed */
7141                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7142                         clear_bit(NFS4_SESSION_ESTABLISHED,
7143                                         &clp->cl_session->session_state);
7144                         clp->cl_seqid = res.seqid;
7145                 }
7146
7147                 kfree(clp->cl_serverowner);
7148                 clp->cl_serverowner = res.server_owner;
7149                 res.server_owner = NULL;
7150
7151                 /* use the most recent implementation id */
7152                 kfree(clp->cl_implid);
7153                 clp->cl_implid = res.impl_id;
7154                 res.impl_id = NULL;
7155
7156                 if (clp->cl_serverscope != NULL &&
7157                     !nfs41_same_server_scope(clp->cl_serverscope,
7158                                              res.server_scope)) {
7159                         dprintk("%s: server_scope mismatch detected\n",
7160                                 __func__);
7161                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7162                         kfree(clp->cl_serverscope);
7163                         clp->cl_serverscope = NULL;
7164                 }
7165
7166                 if (clp->cl_serverscope == NULL) {
7167                         clp->cl_serverscope = res.server_scope;
7168                         res.server_scope = NULL;
7169                 }
7170         }
7171
7172 out_impl_id:
7173         kfree(res.impl_id);
7174 out_server_scope:
7175         kfree(res.server_scope);
7176 out_server_owner:
7177         kfree(res.server_owner);
7178 out:
7179         if (clp->cl_implid != NULL)
7180                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7181                         "domain: %s, name: %s, date: %llu,%u\n",
7182                         clp->cl_implid->domain, clp->cl_implid->name,
7183                         clp->cl_implid->date.seconds,
7184                         clp->cl_implid->date.nseconds);
7185         dprintk("NFS reply exchange_id: %d\n", status);
7186         return status;
7187 }
7188
7189 /*
7190  * nfs4_proc_exchange_id()
7191  *
7192  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7193  *
7194  * Since the clientid has expired, all compounds using sessions
7195  * associated with the stale clientid will be returning
7196  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7197  * be in some phase of session reset.
7198  *
7199  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7200  */
7201 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7202 {
7203         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7204         int status;
7205
7206         /* try SP4_MACH_CRED if krb5i/p */
7207         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7208             authflavor == RPC_AUTH_GSS_KRB5P) {
7209                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7210                 if (!status)
7211                         return 0;
7212         }
7213
7214         /* try SP4_NONE */
7215         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7216 }
7217
7218 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7219                 struct rpc_cred *cred)
7220 {
7221         struct rpc_message msg = {
7222                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7223                 .rpc_argp = clp,
7224                 .rpc_cred = cred,
7225         };
7226         int status;
7227
7228         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7229         trace_nfs4_destroy_clientid(clp, status);
7230         if (status)
7231                 dprintk("NFS: Got error %d from the server %s on "
7232                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7233         return status;
7234 }
7235
7236 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7237                 struct rpc_cred *cred)
7238 {
7239         unsigned int loop;
7240         int ret;
7241
7242         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7243                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7244                 switch (ret) {
7245                 case -NFS4ERR_DELAY:
7246                 case -NFS4ERR_CLIENTID_BUSY:
7247                         ssleep(1);
7248                         break;
7249                 default:
7250                         return ret;
7251                 }
7252         }
7253         return 0;
7254 }
7255
7256 int nfs4_destroy_clientid(struct nfs_client *clp)
7257 {
7258         struct rpc_cred *cred;
7259         int ret = 0;
7260
7261         if (clp->cl_mvops->minor_version < 1)
7262                 goto out;
7263         if (clp->cl_exchange_flags == 0)
7264                 goto out;
7265         if (clp->cl_preserve_clid)
7266                 goto out;
7267         cred = nfs4_get_clid_cred(clp);
7268         ret = nfs4_proc_destroy_clientid(clp, cred);
7269         if (cred)
7270                 put_rpccred(cred);
7271         switch (ret) {
7272         case 0:
7273         case -NFS4ERR_STALE_CLIENTID:
7274                 clp->cl_exchange_flags = 0;
7275         }
7276 out:
7277         return ret;
7278 }
7279
7280 struct nfs4_get_lease_time_data {
7281         struct nfs4_get_lease_time_args *args;
7282         struct nfs4_get_lease_time_res *res;
7283         struct nfs_client *clp;
7284 };
7285
7286 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7287                                         void *calldata)
7288 {
7289         struct nfs4_get_lease_time_data *data =
7290                         (struct nfs4_get_lease_time_data *)calldata;
7291
7292         dprintk("--> %s\n", __func__);
7293         /* just setup sequence, do not trigger session recovery
7294            since we're invoked within one */
7295         nfs41_setup_sequence(data->clp->cl_session,
7296                         &data->args->la_seq_args,
7297                         &data->res->lr_seq_res,
7298                         task);
7299         dprintk("<-- %s\n", __func__);
7300 }
7301
7302 /*
7303  * Called from nfs4_state_manager thread for session setup, so don't recover
7304  * from sequence operation or clientid errors.
7305  */
7306 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7307 {
7308         struct nfs4_get_lease_time_data *data =
7309                         (struct nfs4_get_lease_time_data *)calldata;
7310
7311         dprintk("--> %s\n", __func__);
7312         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7313                 return;
7314         switch (task->tk_status) {
7315         case -NFS4ERR_DELAY:
7316         case -NFS4ERR_GRACE:
7317                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7318                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7319                 task->tk_status = 0;
7320                 /* fall through */
7321         case -NFS4ERR_RETRY_UNCACHED_REP:
7322                 rpc_restart_call_prepare(task);
7323                 return;
7324         }
7325         dprintk("<-- %s\n", __func__);
7326 }
7327
7328 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7329         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7330         .rpc_call_done = nfs4_get_lease_time_done,
7331 };
7332
7333 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7334 {
7335         struct rpc_task *task;
7336         struct nfs4_get_lease_time_args args;
7337         struct nfs4_get_lease_time_res res = {
7338                 .lr_fsinfo = fsinfo,
7339         };
7340         struct nfs4_get_lease_time_data data = {
7341                 .args = &args,
7342                 .res = &res,
7343                 .clp = clp,
7344         };
7345         struct rpc_message msg = {
7346                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7347                 .rpc_argp = &args,
7348                 .rpc_resp = &res,
7349         };
7350         struct rpc_task_setup task_setup = {
7351                 .rpc_client = clp->cl_rpcclient,
7352                 .rpc_message = &msg,
7353                 .callback_ops = &nfs4_get_lease_time_ops,
7354                 .callback_data = &data,
7355                 .flags = RPC_TASK_TIMEOUT,
7356         };
7357         int status;
7358
7359         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7360         nfs4_set_sequence_privileged(&args.la_seq_args);
7361         dprintk("--> %s\n", __func__);
7362         task = rpc_run_task(&task_setup);
7363
7364         if (IS_ERR(task))
7365                 status = PTR_ERR(task);
7366         else {
7367                 status = task->tk_status;
7368                 rpc_put_task(task);
7369         }
7370         dprintk("<-- %s return %d\n", __func__, status);
7371
7372         return status;
7373 }
7374
7375 /*
7376  * Initialize the values to be used by the client in CREATE_SESSION
7377  * If nfs4_init_session set the fore channel request and response sizes,
7378  * use them.
7379  *
7380  * Set the back channel max_resp_sz_cached to zero to force the client to
7381  * always set csa_cachethis to FALSE because the current implementation
7382  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7383  */
7384 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7385                                     struct rpc_clnt *clnt)
7386 {
7387         unsigned int max_rqst_sz, max_resp_sz;
7388         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7389
7390         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7391         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7392
7393         /* Fore channel attributes */
7394         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7395         args->fc_attrs.max_resp_sz = max_resp_sz;
7396         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7397         args->fc_attrs.max_reqs = max_session_slots;
7398
7399         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7400                 "max_ops=%u max_reqs=%u\n",
7401                 __func__,
7402                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7403                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7404
7405         /* Back channel attributes */
7406         args->bc_attrs.max_rqst_sz = max_bc_payload;
7407         args->bc_attrs.max_resp_sz = max_bc_payload;
7408         args->bc_attrs.max_resp_sz_cached = 0;
7409         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7410         args->bc_attrs.max_reqs = NFS41_BC_MAX_CALLBACKS;
7411
7412         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7413                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7414                 __func__,
7415                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7416                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7417                 args->bc_attrs.max_reqs);
7418 }
7419
7420 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7421                 struct nfs41_create_session_res *res)
7422 {
7423         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7424         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7425
7426         if (rcvd->max_resp_sz > sent->max_resp_sz)
7427                 return -EINVAL;
7428         /*
7429          * Our requested max_ops is the minimum we need; we're not
7430          * prepared to break up compounds into smaller pieces than that.
7431          * So, no point even trying to continue if the server won't
7432          * cooperate:
7433          */
7434         if (rcvd->max_ops < sent->max_ops)
7435                 return -EINVAL;
7436         if (rcvd->max_reqs == 0)
7437                 return -EINVAL;
7438         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7439                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7440         return 0;
7441 }
7442
7443 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7444                 struct nfs41_create_session_res *res)
7445 {
7446         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7447         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7448
7449         if (!(res->flags & SESSION4_BACK_CHAN))
7450                 goto out;
7451         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7452                 return -EINVAL;
7453         if (rcvd->max_resp_sz < sent->max_resp_sz)
7454                 return -EINVAL;
7455         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7456                 return -EINVAL;
7457         /* These would render the backchannel useless: */
7458         if (rcvd->max_ops != sent->max_ops)
7459                 return -EINVAL;
7460         if (rcvd->max_reqs != sent->max_reqs)
7461                 return -EINVAL;
7462 out:
7463         return 0;
7464 }
7465
7466 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7467                                      struct nfs41_create_session_res *res)
7468 {
7469         int ret;
7470
7471         ret = nfs4_verify_fore_channel_attrs(args, res);
7472         if (ret)
7473                 return ret;
7474         return nfs4_verify_back_channel_attrs(args, res);
7475 }
7476
7477 static void nfs4_update_session(struct nfs4_session *session,
7478                 struct nfs41_create_session_res *res)
7479 {
7480         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7481         /* Mark client id and session as being confirmed */
7482         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7483         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7484         session->flags = res->flags;
7485         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7486         if (res->flags & SESSION4_BACK_CHAN)
7487                 memcpy(&session->bc_attrs, &res->bc_attrs,
7488                                 sizeof(session->bc_attrs));
7489 }
7490
7491 static int _nfs4_proc_create_session(struct nfs_client *clp,
7492                 struct rpc_cred *cred)
7493 {
7494         struct nfs4_session *session = clp->cl_session;
7495         struct nfs41_create_session_args args = {
7496                 .client = clp,
7497                 .clientid = clp->cl_clientid,
7498                 .seqid = clp->cl_seqid,
7499                 .cb_program = NFS4_CALLBACK,
7500         };
7501         struct nfs41_create_session_res res;
7502
7503         struct rpc_message msg = {
7504                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7505                 .rpc_argp = &args,
7506                 .rpc_resp = &res,
7507                 .rpc_cred = cred,
7508         };
7509         int status;
7510
7511         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7512         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7513
7514         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7515         trace_nfs4_create_session(clp, status);
7516
7517         if (!status) {
7518                 /* Verify the session's negotiated channel_attrs values */
7519                 status = nfs4_verify_channel_attrs(&args, &res);
7520                 /* Increment the clientid slot sequence id */
7521                 if (clp->cl_seqid == res.seqid)
7522                         clp->cl_seqid++;
7523                 if (status)
7524                         goto out;
7525                 nfs4_update_session(session, &res);
7526         }
7527 out:
7528         return status;
7529 }
7530
7531 /*
7532  * Issues a CREATE_SESSION operation to the server.
7533  * It is the responsibility of the caller to verify the session is
7534  * expired before calling this routine.
7535  */
7536 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7537 {
7538         int status;
7539         unsigned *ptr;
7540         struct nfs4_session *session = clp->cl_session;
7541
7542         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7543
7544         status = _nfs4_proc_create_session(clp, cred);
7545         if (status)
7546                 goto out;
7547
7548         /* Init or reset the session slot tables */
7549         status = nfs4_setup_session_slot_tables(session);
7550         dprintk("slot table setup returned %d\n", status);
7551         if (status)
7552                 goto out;
7553
7554         ptr = (unsigned *)&session->sess_id.data[0];
7555         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7556                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7557 out:
7558         dprintk("<-- %s\n", __func__);
7559         return status;
7560 }
7561
7562 /*
7563  * Issue the over-the-wire RPC DESTROY_SESSION.
7564  * The caller must serialize access to this routine.
7565  */
7566 int nfs4_proc_destroy_session(struct nfs4_session *session,
7567                 struct rpc_cred *cred)
7568 {
7569         struct rpc_message msg = {
7570                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7571                 .rpc_argp = session,
7572                 .rpc_cred = cred,
7573         };
7574         int status = 0;
7575
7576         dprintk("--> nfs4_proc_destroy_session\n");
7577
7578         /* session is still being setup */
7579         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7580                 return 0;
7581
7582         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7583         trace_nfs4_destroy_session(session->clp, status);
7584
7585         if (status)
7586                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7587                         "Session has been destroyed regardless...\n", status);
7588
7589         dprintk("<-- nfs4_proc_destroy_session\n");
7590         return status;
7591 }
7592
7593 /*
7594  * Renew the cl_session lease.
7595  */
7596 struct nfs4_sequence_data {
7597         struct nfs_client *clp;
7598         struct nfs4_sequence_args args;
7599         struct nfs4_sequence_res res;
7600 };
7601
7602 static void nfs41_sequence_release(void *data)
7603 {
7604         struct nfs4_sequence_data *calldata = data;
7605         struct nfs_client *clp = calldata->clp;
7606
7607         if (atomic_read(&clp->cl_count) > 1)
7608                 nfs4_schedule_state_renewal(clp);
7609         nfs_put_client(clp);
7610         kfree(calldata);
7611 }
7612
7613 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7614 {
7615         switch(task->tk_status) {
7616         case -NFS4ERR_DELAY:
7617                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7618                 return -EAGAIN;
7619         default:
7620                 nfs4_schedule_lease_recovery(clp);
7621         }
7622         return 0;
7623 }
7624
7625 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7626 {
7627         struct nfs4_sequence_data *calldata = data;
7628         struct nfs_client *clp = calldata->clp;
7629
7630         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7631                 return;
7632
7633         trace_nfs4_sequence(clp, task->tk_status);
7634         if (task->tk_status < 0) {
7635                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7636                 if (atomic_read(&clp->cl_count) == 1)
7637                         goto out;
7638
7639                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7640                         rpc_restart_call_prepare(task);
7641                         return;
7642                 }
7643         }
7644         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7645 out:
7646         dprintk("<-- %s\n", __func__);
7647 }
7648
7649 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7650 {
7651         struct nfs4_sequence_data *calldata = data;
7652         struct nfs_client *clp = calldata->clp;
7653         struct nfs4_sequence_args *args;
7654         struct nfs4_sequence_res *res;
7655
7656         args = task->tk_msg.rpc_argp;
7657         res = task->tk_msg.rpc_resp;
7658
7659         nfs41_setup_sequence(clp->cl_session, args, res, task);
7660 }
7661
7662 static const struct rpc_call_ops nfs41_sequence_ops = {
7663         .rpc_call_done = nfs41_sequence_call_done,
7664         .rpc_call_prepare = nfs41_sequence_prepare,
7665         .rpc_release = nfs41_sequence_release,
7666 };
7667
7668 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7669                 struct rpc_cred *cred,
7670                 bool is_privileged)
7671 {
7672         struct nfs4_sequence_data *calldata;
7673         struct rpc_message msg = {
7674                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7675                 .rpc_cred = cred,
7676         };
7677         struct rpc_task_setup task_setup_data = {
7678                 .rpc_client = clp->cl_rpcclient,
7679                 .rpc_message = &msg,
7680                 .callback_ops = &nfs41_sequence_ops,
7681                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7682         };
7683
7684         if (!atomic_inc_not_zero(&clp->cl_count))
7685                 return ERR_PTR(-EIO);
7686         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7687         if (calldata == NULL) {
7688                 nfs_put_client(clp);
7689                 return ERR_PTR(-ENOMEM);
7690         }
7691         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7692         if (is_privileged)
7693                 nfs4_set_sequence_privileged(&calldata->args);
7694         msg.rpc_argp = &calldata->args;
7695         msg.rpc_resp = &calldata->res;
7696         calldata->clp = clp;
7697         task_setup_data.callback_data = calldata;
7698
7699         return rpc_run_task(&task_setup_data);
7700 }
7701
7702 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7703 {
7704         struct rpc_task *task;
7705         int ret = 0;
7706
7707         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7708                 return -EAGAIN;
7709         task = _nfs41_proc_sequence(clp, cred, false);
7710         if (IS_ERR(task))
7711                 ret = PTR_ERR(task);
7712         else
7713                 rpc_put_task_async(task);
7714         dprintk("<-- %s status=%d\n", __func__, ret);
7715         return ret;
7716 }
7717
7718 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7719 {
7720         struct rpc_task *task;
7721         int ret;
7722
7723         task = _nfs41_proc_sequence(clp, cred, true);
7724         if (IS_ERR(task)) {
7725                 ret = PTR_ERR(task);
7726                 goto out;
7727         }
7728         ret = rpc_wait_for_completion_task(task);
7729         if (!ret)
7730                 ret = task->tk_status;
7731         rpc_put_task(task);
7732 out:
7733         dprintk("<-- %s status=%d\n", __func__, ret);
7734         return ret;
7735 }
7736
7737 struct nfs4_reclaim_complete_data {
7738         struct nfs_client *clp;
7739         struct nfs41_reclaim_complete_args arg;
7740         struct nfs41_reclaim_complete_res res;
7741 };
7742
7743 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7744 {
7745         struct nfs4_reclaim_complete_data *calldata = data;
7746
7747         nfs41_setup_sequence(calldata->clp->cl_session,
7748                         &calldata->arg.seq_args,
7749                         &calldata->res.seq_res,
7750                         task);
7751 }
7752
7753 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7754 {
7755         switch(task->tk_status) {
7756         case 0:
7757         case -NFS4ERR_COMPLETE_ALREADY:
7758         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7759                 break;
7760         case -NFS4ERR_DELAY:
7761                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7762                 /* fall through */
7763         case -NFS4ERR_RETRY_UNCACHED_REP:
7764                 return -EAGAIN;
7765         default:
7766                 nfs4_schedule_lease_recovery(clp);
7767         }
7768         return 0;
7769 }
7770
7771 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7772 {
7773         struct nfs4_reclaim_complete_data *calldata = data;
7774         struct nfs_client *clp = calldata->clp;
7775         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7776
7777         dprintk("--> %s\n", __func__);
7778         if (!nfs41_sequence_done(task, res))
7779                 return;
7780
7781         trace_nfs4_reclaim_complete(clp, task->tk_status);
7782         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7783                 rpc_restart_call_prepare(task);
7784                 return;
7785         }
7786         dprintk("<-- %s\n", __func__);
7787 }
7788
7789 static void nfs4_free_reclaim_complete_data(void *data)
7790 {
7791         struct nfs4_reclaim_complete_data *calldata = data;
7792
7793         kfree(calldata);
7794 }
7795
7796 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7797         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7798         .rpc_call_done = nfs4_reclaim_complete_done,
7799         .rpc_release = nfs4_free_reclaim_complete_data,
7800 };
7801
7802 /*
7803  * Issue a global reclaim complete.
7804  */
7805 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7806                 struct rpc_cred *cred)
7807 {
7808         struct nfs4_reclaim_complete_data *calldata;
7809         struct rpc_task *task;
7810         struct rpc_message msg = {
7811                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7812                 .rpc_cred = cred,
7813         };
7814         struct rpc_task_setup task_setup_data = {
7815                 .rpc_client = clp->cl_rpcclient,
7816                 .rpc_message = &msg,
7817                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7818                 .flags = RPC_TASK_ASYNC,
7819         };
7820         int status = -ENOMEM;
7821
7822         dprintk("--> %s\n", __func__);
7823         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7824         if (calldata == NULL)
7825                 goto out;
7826         calldata->clp = clp;
7827         calldata->arg.one_fs = 0;
7828
7829         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7830         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7831         msg.rpc_argp = &calldata->arg;
7832         msg.rpc_resp = &calldata->res;
7833         task_setup_data.callback_data = calldata;
7834         task = rpc_run_task(&task_setup_data);
7835         if (IS_ERR(task)) {
7836                 status = PTR_ERR(task);
7837                 goto out;
7838         }
7839         status = nfs4_wait_for_completion_rpc_task(task);
7840         if (status == 0)
7841                 status = task->tk_status;
7842         rpc_put_task(task);
7843         return 0;
7844 out:
7845         dprintk("<-- %s status=%d\n", __func__, status);
7846         return status;
7847 }
7848
7849 static void
7850 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7851 {
7852         struct nfs4_layoutget *lgp = calldata;
7853         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7854         struct nfs4_session *session = nfs4_get_session(server);
7855
7856         dprintk("--> %s\n", __func__);
7857         nfs41_setup_sequence(session, &lgp->args.seq_args,
7858                                 &lgp->res.seq_res, task);
7859         dprintk("<-- %s\n", __func__);
7860 }
7861
7862 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7863 {
7864         struct nfs4_layoutget *lgp = calldata;
7865
7866         dprintk("--> %s\n", __func__);
7867         nfs41_sequence_done(task, &lgp->res.seq_res);
7868         dprintk("<-- %s\n", __func__);
7869 }
7870
7871 static int
7872 nfs4_layoutget_handle_exception(struct rpc_task *task,
7873                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
7874 {
7875         struct inode *inode = lgp->args.inode;
7876         struct nfs_server *server = NFS_SERVER(inode);
7877         struct pnfs_layout_hdr *lo;
7878         int nfs4err = task->tk_status;
7879         int err, status = 0;
7880         LIST_HEAD(head);
7881
7882         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7883
7884         switch (nfs4err) {
7885         case 0:
7886                 goto out;
7887
7888         /*
7889          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
7890          * on the file. set tk_status to -ENODATA to tell upper layer to
7891          * retry go inband.
7892          */
7893         case -NFS4ERR_LAYOUTUNAVAILABLE:
7894                 status = -ENODATA;
7895                 goto out;
7896         /*
7897          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7898          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7899          */
7900         case -NFS4ERR_BADLAYOUT:
7901                 status = -EOVERFLOW;
7902                 goto out;
7903         /*
7904          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7905          * (or clients) writing to the same RAID stripe except when
7906          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7907          *
7908          * Treat it like we would RECALLCONFLICT -- we retry for a little
7909          * while, and then eventually give up.
7910          */
7911         case -NFS4ERR_LAYOUTTRYLATER:
7912                 if (lgp->args.minlength == 0) {
7913                         status = -EOVERFLOW;
7914                         goto out;
7915                 }
7916                 status = -EBUSY;
7917                 break;
7918         case -NFS4ERR_RECALLCONFLICT:
7919                 status = -ERECALLCONFLICT;
7920                 break;
7921         case -NFS4ERR_EXPIRED:
7922         case -NFS4ERR_BAD_STATEID:
7923                 exception->timeout = 0;
7924                 spin_lock(&inode->i_lock);
7925                 lo = NFS_I(inode)->layout;
7926                 /* If the open stateid was bad, then recover it. */
7927                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
7928                     nfs4_stateid_match_other(&lgp->args.stateid,
7929                                         &lgp->args.ctx->state->stateid)) {
7930                         spin_unlock(&inode->i_lock);
7931                         exception->state = lgp->args.ctx->state;
7932                         break;
7933                 }
7934
7935                 /*
7936                  * Mark the bad layout state as invalid, then retry
7937                  */
7938                 pnfs_mark_layout_stateid_invalid(lo, &head);
7939                 spin_unlock(&inode->i_lock);
7940                 pnfs_free_lseg_list(&head);
7941                 status = -EAGAIN;
7942                 goto out;
7943         }
7944
7945         err = nfs4_handle_exception(server, nfs4err, exception);
7946         if (!status) {
7947                 if (exception->retry)
7948                         status = -EAGAIN;
7949                 else
7950                         status = err;
7951         }
7952 out:
7953         dprintk("<-- %s\n", __func__);
7954         return status;
7955 }
7956
7957 static size_t max_response_pages(struct nfs_server *server)
7958 {
7959         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7960         return nfs_page_array_len(0, max_resp_sz);
7961 }
7962
7963 static void nfs4_free_pages(struct page **pages, size_t size)
7964 {
7965         int i;
7966
7967         if (!pages)
7968                 return;
7969
7970         for (i = 0; i < size; i++) {
7971                 if (!pages[i])
7972                         break;
7973                 __free_page(pages[i]);
7974         }
7975         kfree(pages);
7976 }
7977
7978 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7979 {
7980         struct page **pages;
7981         int i;
7982
7983         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7984         if (!pages) {
7985                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7986                 return NULL;
7987         }
7988
7989         for (i = 0; i < size; i++) {
7990                 pages[i] = alloc_page(gfp_flags);
7991                 if (!pages[i]) {
7992                         dprintk("%s: failed to allocate page\n", __func__);
7993                         nfs4_free_pages(pages, size);
7994                         return NULL;
7995                 }
7996         }
7997
7998         return pages;
7999 }
8000
8001 static void nfs4_layoutget_release(void *calldata)
8002 {
8003         struct nfs4_layoutget *lgp = calldata;
8004         struct inode *inode = lgp->args.inode;
8005         struct nfs_server *server = NFS_SERVER(inode);
8006         size_t max_pages = max_response_pages(server);
8007
8008         dprintk("--> %s\n", __func__);
8009         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8010         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8011         put_nfs_open_context(lgp->args.ctx);
8012         kfree(calldata);
8013         dprintk("<-- %s\n", __func__);
8014 }
8015
8016 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8017         .rpc_call_prepare = nfs4_layoutget_prepare,
8018         .rpc_call_done = nfs4_layoutget_done,
8019         .rpc_release = nfs4_layoutget_release,
8020 };
8021
8022 struct pnfs_layout_segment *
8023 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8024 {
8025         struct inode *inode = lgp->args.inode;
8026         struct nfs_server *server = NFS_SERVER(inode);
8027         size_t max_pages = max_response_pages(server);
8028         struct rpc_task *task;
8029         struct rpc_message msg = {
8030                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8031                 .rpc_argp = &lgp->args,
8032                 .rpc_resp = &lgp->res,
8033                 .rpc_cred = lgp->cred,
8034         };
8035         struct rpc_task_setup task_setup_data = {
8036                 .rpc_client = server->client,
8037                 .rpc_message = &msg,
8038                 .callback_ops = &nfs4_layoutget_call_ops,
8039                 .callback_data = lgp,
8040                 .flags = RPC_TASK_ASYNC,
8041         };
8042         struct pnfs_layout_segment *lseg = NULL;
8043         struct nfs4_exception exception = {
8044                 .inode = inode,
8045                 .timeout = *timeout,
8046         };
8047         int status = 0;
8048
8049         dprintk("--> %s\n", __func__);
8050
8051         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8052         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8053
8054         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8055         if (!lgp->args.layout.pages) {
8056                 nfs4_layoutget_release(lgp);
8057                 return ERR_PTR(-ENOMEM);
8058         }
8059         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8060
8061         lgp->res.layoutp = &lgp->args.layout;
8062         lgp->res.seq_res.sr_slot = NULL;
8063         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8064
8065         task = rpc_run_task(&task_setup_data);
8066         if (IS_ERR(task))
8067                 return ERR_CAST(task);
8068         status = nfs4_wait_for_completion_rpc_task(task);
8069         if (status == 0) {
8070                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8071                 *timeout = exception.timeout;
8072         }
8073
8074         trace_nfs4_layoutget(lgp->args.ctx,
8075                         &lgp->args.range,
8076                         &lgp->res.range,
8077                         &lgp->res.stateid,
8078                         status);
8079
8080         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8081         if (status == 0 && lgp->res.layoutp->len)
8082                 lseg = pnfs_layout_process(lgp);
8083         rpc_put_task(task);
8084         dprintk("<-- %s status=%d\n", __func__, status);
8085         if (status)
8086                 return ERR_PTR(status);
8087         return lseg;
8088 }
8089
8090 static void
8091 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8092 {
8093         struct nfs4_layoutreturn *lrp = calldata;
8094
8095         dprintk("--> %s\n", __func__);
8096         nfs41_setup_sequence(lrp->clp->cl_session,
8097                         &lrp->args.seq_args,
8098                         &lrp->res.seq_res,
8099                         task);
8100 }
8101
8102 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8103 {
8104         struct nfs4_layoutreturn *lrp = calldata;
8105         struct nfs_server *server;
8106
8107         dprintk("--> %s\n", __func__);
8108
8109         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8110                 return;
8111
8112         server = NFS_SERVER(lrp->args.inode);
8113         switch (task->tk_status) {
8114         default:
8115                 task->tk_status = 0;
8116         case 0:
8117                 break;
8118         case -NFS4ERR_DELAY:
8119                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8120                         break;
8121                 rpc_restart_call_prepare(task);
8122                 return;
8123         }
8124         dprintk("<-- %s\n", __func__);
8125 }
8126
8127 static void nfs4_layoutreturn_release(void *calldata)
8128 {
8129         struct nfs4_layoutreturn *lrp = calldata;
8130         struct pnfs_layout_hdr *lo = lrp->args.layout;
8131         LIST_HEAD(freeme);
8132
8133         dprintk("--> %s\n", __func__);
8134         spin_lock(&lo->plh_inode->i_lock);
8135         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range,
8136                         be32_to_cpu(lrp->args.stateid.seqid));
8137         if (lrp->res.lrs_present && pnfs_layout_is_valid(lo))
8138                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8139         pnfs_clear_layoutreturn_waitbit(lo);
8140         spin_unlock(&lo->plh_inode->i_lock);
8141         pnfs_free_lseg_list(&freeme);
8142         pnfs_put_layout_hdr(lrp->args.layout);
8143         nfs_iput_and_deactive(lrp->inode);
8144         kfree(calldata);
8145         dprintk("<-- %s\n", __func__);
8146 }
8147
8148 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8149         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8150         .rpc_call_done = nfs4_layoutreturn_done,
8151         .rpc_release = nfs4_layoutreturn_release,
8152 };
8153
8154 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8155 {
8156         struct rpc_task *task;
8157         struct rpc_message msg = {
8158                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8159                 .rpc_argp = &lrp->args,
8160                 .rpc_resp = &lrp->res,
8161                 .rpc_cred = lrp->cred,
8162         };
8163         struct rpc_task_setup task_setup_data = {
8164                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8165                 .rpc_message = &msg,
8166                 .callback_ops = &nfs4_layoutreturn_call_ops,
8167                 .callback_data = lrp,
8168         };
8169         int status = 0;
8170
8171         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8172                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8173                         &task_setup_data.rpc_client, &msg);
8174
8175         dprintk("--> %s\n", __func__);
8176         if (!sync) {
8177                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8178                 if (!lrp->inode) {
8179                         nfs4_layoutreturn_release(lrp);
8180                         return -EAGAIN;
8181                 }
8182                 task_setup_data.flags |= RPC_TASK_ASYNC;
8183         }
8184         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8185         task = rpc_run_task(&task_setup_data);
8186         if (IS_ERR(task))
8187                 return PTR_ERR(task);
8188         if (sync)
8189                 status = task->tk_status;
8190         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8191         dprintk("<-- %s status=%d\n", __func__, status);
8192         rpc_put_task(task);
8193         return status;
8194 }
8195
8196 static int
8197 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8198                 struct pnfs_device *pdev,
8199                 struct rpc_cred *cred)
8200 {
8201         struct nfs4_getdeviceinfo_args args = {
8202                 .pdev = pdev,
8203                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8204                         NOTIFY_DEVICEID4_DELETE,
8205         };
8206         struct nfs4_getdeviceinfo_res res = {
8207                 .pdev = pdev,
8208         };
8209         struct rpc_message msg = {
8210                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8211                 .rpc_argp = &args,
8212                 .rpc_resp = &res,
8213                 .rpc_cred = cred,
8214         };
8215         int status;
8216
8217         dprintk("--> %s\n", __func__);
8218         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8219         if (res.notification & ~args.notify_types)
8220                 dprintk("%s: unsupported notification\n", __func__);
8221         if (res.notification != args.notify_types)
8222                 pdev->nocache = 1;
8223
8224         dprintk("<-- %s status=%d\n", __func__, status);
8225
8226         return status;
8227 }
8228
8229 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8230                 struct pnfs_device *pdev,
8231                 struct rpc_cred *cred)
8232 {
8233         struct nfs4_exception exception = { };
8234         int err;
8235
8236         do {
8237                 err = nfs4_handle_exception(server,
8238                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8239                                         &exception);
8240         } while (exception.retry);
8241         return err;
8242 }
8243 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8244
8245 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8246 {
8247         struct nfs4_layoutcommit_data *data = calldata;
8248         struct nfs_server *server = NFS_SERVER(data->args.inode);
8249         struct nfs4_session *session = nfs4_get_session(server);
8250
8251         nfs41_setup_sequence(session,
8252                         &data->args.seq_args,
8253                         &data->res.seq_res,
8254                         task);
8255 }
8256
8257 static void
8258 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8259 {
8260         struct nfs4_layoutcommit_data *data = calldata;
8261         struct nfs_server *server = NFS_SERVER(data->args.inode);
8262
8263         if (!nfs41_sequence_done(task, &data->res.seq_res))
8264                 return;
8265
8266         switch (task->tk_status) { /* Just ignore these failures */
8267         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8268         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8269         case -NFS4ERR_BADLAYOUT:     /* no layout */
8270         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8271                 task->tk_status = 0;
8272         case 0:
8273                 break;
8274         default:
8275                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8276                         rpc_restart_call_prepare(task);
8277                         return;
8278                 }
8279         }
8280 }
8281
8282 static void nfs4_layoutcommit_release(void *calldata)
8283 {
8284         struct nfs4_layoutcommit_data *data = calldata;
8285
8286         pnfs_cleanup_layoutcommit(data);
8287         nfs_post_op_update_inode_force_wcc(data->args.inode,
8288                                            data->res.fattr);
8289         put_rpccred(data->cred);
8290         nfs_iput_and_deactive(data->inode);
8291         kfree(data);
8292 }
8293
8294 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8295         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8296         .rpc_call_done = nfs4_layoutcommit_done,
8297         .rpc_release = nfs4_layoutcommit_release,
8298 };
8299
8300 int
8301 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8302 {
8303         struct rpc_message msg = {
8304                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8305                 .rpc_argp = &data->args,
8306                 .rpc_resp = &data->res,
8307                 .rpc_cred = data->cred,
8308         };
8309         struct rpc_task_setup task_setup_data = {
8310                 .task = &data->task,
8311                 .rpc_client = NFS_CLIENT(data->args.inode),
8312                 .rpc_message = &msg,
8313                 .callback_ops = &nfs4_layoutcommit_ops,
8314                 .callback_data = data,
8315         };
8316         struct rpc_task *task;
8317         int status = 0;
8318
8319         dprintk("NFS: initiating layoutcommit call. sync %d "
8320                 "lbw: %llu inode %lu\n", sync,
8321                 data->args.lastbytewritten,
8322                 data->args.inode->i_ino);
8323
8324         if (!sync) {
8325                 data->inode = nfs_igrab_and_active(data->args.inode);
8326                 if (data->inode == NULL) {
8327                         nfs4_layoutcommit_release(data);
8328                         return -EAGAIN;
8329                 }
8330                 task_setup_data.flags = RPC_TASK_ASYNC;
8331         }
8332         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8333         task = rpc_run_task(&task_setup_data);
8334         if (IS_ERR(task))
8335                 return PTR_ERR(task);
8336         if (sync)
8337                 status = task->tk_status;
8338         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8339         dprintk("%s: status %d\n", __func__, status);
8340         rpc_put_task(task);
8341         return status;
8342 }
8343
8344 /**
8345  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8346  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8347  */
8348 static int
8349 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8350                     struct nfs_fsinfo *info,
8351                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8352 {
8353         struct nfs41_secinfo_no_name_args args = {
8354                 .style = SECINFO_STYLE_CURRENT_FH,
8355         };
8356         struct nfs4_secinfo_res res = {
8357                 .flavors = flavors,
8358         };
8359         struct rpc_message msg = {
8360                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8361                 .rpc_argp = &args,
8362                 .rpc_resp = &res,
8363         };
8364         struct rpc_clnt *clnt = server->client;
8365         struct rpc_cred *cred = NULL;
8366         int status;
8367
8368         if (use_integrity) {
8369                 clnt = server->nfs_client->cl_rpcclient;
8370                 cred = nfs4_get_clid_cred(server->nfs_client);
8371                 msg.rpc_cred = cred;
8372         }
8373
8374         dprintk("--> %s\n", __func__);
8375         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8376                                 &res.seq_res, 0);
8377         dprintk("<-- %s status=%d\n", __func__, status);
8378
8379         if (cred)
8380                 put_rpccred(cred);
8381
8382         return status;
8383 }
8384
8385 static int
8386 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8387                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8388 {
8389         struct nfs4_exception exception = { };
8390         int err;
8391         do {
8392                 /* first try using integrity protection */
8393                 err = -NFS4ERR_WRONGSEC;
8394
8395                 /* try to use integrity protection with machine cred */
8396                 if (_nfs4_is_integrity_protected(server->nfs_client))
8397                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8398                                                           flavors, true);
8399
8400                 /*
8401                  * if unable to use integrity protection, or SECINFO with
8402                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8403                  * disallowed by spec, but exists in deployed servers) use
8404                  * the current filesystem's rpc_client and the user cred.
8405                  */
8406                 if (err == -NFS4ERR_WRONGSEC)
8407                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8408                                                           flavors, false);
8409
8410                 switch (err) {
8411                 case 0:
8412                 case -NFS4ERR_WRONGSEC:
8413                 case -ENOTSUPP:
8414                         goto out;
8415                 default:
8416                         err = nfs4_handle_exception(server, err, &exception);
8417                 }
8418         } while (exception.retry);
8419 out:
8420         return err;
8421 }
8422
8423 static int
8424 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8425                     struct nfs_fsinfo *info)
8426 {
8427         int err;
8428         struct page *page;
8429         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8430         struct nfs4_secinfo_flavors *flavors;
8431         struct nfs4_secinfo4 *secinfo;
8432         int i;
8433
8434         page = alloc_page(GFP_KERNEL);
8435         if (!page) {
8436                 err = -ENOMEM;
8437                 goto out;
8438         }
8439
8440         flavors = page_address(page);
8441         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8442
8443         /*
8444          * Fall back on "guess and check" method if
8445          * the server doesn't support SECINFO_NO_NAME
8446          */
8447         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8448                 err = nfs4_find_root_sec(server, fhandle, info);
8449                 goto out_freepage;
8450         }
8451         if (err)
8452                 goto out_freepage;
8453
8454         for (i = 0; i < flavors->num_flavors; i++) {
8455                 secinfo = &flavors->flavors[i];
8456
8457                 switch (secinfo->flavor) {
8458                 case RPC_AUTH_NULL:
8459                 case RPC_AUTH_UNIX:
8460                 case RPC_AUTH_GSS:
8461                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8462                                         &secinfo->flavor_info);
8463                         break;
8464                 default:
8465                         flavor = RPC_AUTH_MAXFLAVOR;
8466                         break;
8467                 }
8468
8469                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8470                         flavor = RPC_AUTH_MAXFLAVOR;
8471
8472                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8473                         err = nfs4_lookup_root_sec(server, fhandle,
8474                                                    info, flavor);
8475                         if (!err)
8476                                 break;
8477                 }
8478         }
8479
8480         if (flavor == RPC_AUTH_MAXFLAVOR)
8481                 err = -EPERM;
8482
8483 out_freepage:
8484         put_page(page);
8485         if (err == -EACCES)
8486                 return -EPERM;
8487 out:
8488         return err;
8489 }
8490
8491 static int _nfs41_test_stateid(struct nfs_server *server,
8492                 nfs4_stateid *stateid,
8493                 struct rpc_cred *cred)
8494 {
8495         int status;
8496         struct nfs41_test_stateid_args args = {
8497                 .stateid = stateid,
8498         };
8499         struct nfs41_test_stateid_res res;
8500         struct rpc_message msg = {
8501                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8502                 .rpc_argp = &args,
8503                 .rpc_resp = &res,
8504                 .rpc_cred = cred,
8505         };
8506         struct rpc_clnt *rpc_client = server->client;
8507
8508         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8509                 &rpc_client, &msg);
8510
8511         dprintk("NFS call  test_stateid %p\n", stateid);
8512         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8513         nfs4_set_sequence_privileged(&args.seq_args);
8514         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8515                         &args.seq_args, &res.seq_res);
8516         if (status != NFS_OK) {
8517                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8518                 return status;
8519         }
8520         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8521         return -res.status;
8522 }
8523
8524 /**
8525  * nfs41_test_stateid - perform a TEST_STATEID operation
8526  *
8527  * @server: server / transport on which to perform the operation
8528  * @stateid: state ID to test
8529  * @cred: credential
8530  *
8531  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8532  * Otherwise a negative NFS4ERR value is returned if the operation
8533  * failed or the state ID is not currently valid.
8534  */
8535 static int nfs41_test_stateid(struct nfs_server *server,
8536                 nfs4_stateid *stateid,
8537                 struct rpc_cred *cred)
8538 {
8539         struct nfs4_exception exception = { };
8540         int err;
8541         do {
8542                 err = _nfs41_test_stateid(server, stateid, cred);
8543                 if (err != -NFS4ERR_DELAY)
8544                         break;
8545                 nfs4_handle_exception(server, err, &exception);
8546         } while (exception.retry);
8547         return err;
8548 }
8549
8550 struct nfs_free_stateid_data {
8551         struct nfs_server *server;
8552         struct nfs41_free_stateid_args args;
8553         struct nfs41_free_stateid_res res;
8554 };
8555
8556 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8557 {
8558         struct nfs_free_stateid_data *data = calldata;
8559         nfs41_setup_sequence(nfs4_get_session(data->server),
8560                         &data->args.seq_args,
8561                         &data->res.seq_res,
8562                         task);
8563 }
8564
8565 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8566 {
8567         struct nfs_free_stateid_data *data = calldata;
8568
8569         nfs41_sequence_done(task, &data->res.seq_res);
8570
8571         switch (task->tk_status) {
8572         case -NFS4ERR_DELAY:
8573                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8574                         rpc_restart_call_prepare(task);
8575         }
8576 }
8577
8578 static void nfs41_free_stateid_release(void *calldata)
8579 {
8580         kfree(calldata);
8581 }
8582
8583 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8584         .rpc_call_prepare = nfs41_free_stateid_prepare,
8585         .rpc_call_done = nfs41_free_stateid_done,
8586         .rpc_release = nfs41_free_stateid_release,
8587 };
8588
8589 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8590                 nfs4_stateid *stateid,
8591                 struct rpc_cred *cred,
8592                 bool privileged)
8593 {
8594         struct rpc_message msg = {
8595                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8596                 .rpc_cred = cred,
8597         };
8598         struct rpc_task_setup task_setup = {
8599                 .rpc_client = server->client,
8600                 .rpc_message = &msg,
8601                 .callback_ops = &nfs41_free_stateid_ops,
8602                 .flags = RPC_TASK_ASYNC,
8603         };
8604         struct nfs_free_stateid_data *data;
8605
8606         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8607                 &task_setup.rpc_client, &msg);
8608
8609         dprintk("NFS call  free_stateid %p\n", stateid);
8610         data = kmalloc(sizeof(*data), GFP_NOFS);
8611         if (!data)
8612                 return ERR_PTR(-ENOMEM);
8613         data->server = server;
8614         nfs4_stateid_copy(&data->args.stateid, stateid);
8615
8616         task_setup.callback_data = data;
8617
8618         msg.rpc_argp = &data->args;
8619         msg.rpc_resp = &data->res;
8620         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8621         if (privileged)
8622                 nfs4_set_sequence_privileged(&data->args.seq_args);
8623
8624         return rpc_run_task(&task_setup);
8625 }
8626
8627 /**
8628  * nfs41_free_stateid - perform a FREE_STATEID operation
8629  *
8630  * @server: server / transport on which to perform the operation
8631  * @stateid: state ID to release
8632  * @cred: credential
8633  *
8634  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8635  * negative NFS4ERR value is returned.
8636  */
8637 static int nfs41_free_stateid(struct nfs_server *server,
8638                 nfs4_stateid *stateid,
8639                 struct rpc_cred *cred)
8640 {
8641         struct rpc_task *task;
8642         int ret;
8643
8644         task = _nfs41_free_stateid(server, stateid, cred, true);
8645         if (IS_ERR(task))
8646                 return PTR_ERR(task);
8647         ret = rpc_wait_for_completion_task(task);
8648         if (!ret)
8649                 ret = task->tk_status;
8650         rpc_put_task(task);
8651         return ret;
8652 }
8653
8654 static void
8655 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8656 {
8657         struct rpc_task *task;
8658         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8659
8660         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8661         nfs4_free_lock_state(server, lsp);
8662         if (IS_ERR(task))
8663                 return;
8664         rpc_put_task(task);
8665 }
8666
8667 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8668                 const nfs4_stateid *s2)
8669 {
8670         if (s1->type != s2->type)
8671                 return false;
8672
8673         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8674                 return false;
8675
8676         if (s1->seqid == s2->seqid)
8677                 return true;
8678         if (s1->seqid == 0 || s2->seqid == 0)
8679                 return true;
8680
8681         return false;
8682 }
8683
8684 #endif /* CONFIG_NFS_V4_1 */
8685
8686 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8687                 const nfs4_stateid *s2)
8688 {
8689         return nfs4_stateid_match(s1, s2);
8690 }
8691
8692
8693 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8694         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8695         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8696         .recover_open   = nfs4_open_reclaim,
8697         .recover_lock   = nfs4_lock_reclaim,
8698         .establish_clid = nfs4_init_clientid,
8699         .detect_trunking = nfs40_discover_server_trunking,
8700 };
8701
8702 #if defined(CONFIG_NFS_V4_1)
8703 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8704         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8705         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8706         .recover_open   = nfs4_open_reclaim,
8707         .recover_lock   = nfs4_lock_reclaim,
8708         .establish_clid = nfs41_init_clientid,
8709         .reclaim_complete = nfs41_proc_reclaim_complete,
8710         .detect_trunking = nfs41_discover_server_trunking,
8711 };
8712 #endif /* CONFIG_NFS_V4_1 */
8713
8714 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8715         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8716         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8717         .recover_open   = nfs40_open_expired,
8718         .recover_lock   = nfs4_lock_expired,
8719         .establish_clid = nfs4_init_clientid,
8720 };
8721
8722 #if defined(CONFIG_NFS_V4_1)
8723 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8724         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8725         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8726         .recover_open   = nfs41_open_expired,
8727         .recover_lock   = nfs41_lock_expired,
8728         .establish_clid = nfs41_init_clientid,
8729 };
8730 #endif /* CONFIG_NFS_V4_1 */
8731
8732 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8733         .sched_state_renewal = nfs4_proc_async_renew,
8734         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8735         .renew_lease = nfs4_proc_renew,
8736 };
8737
8738 #if defined(CONFIG_NFS_V4_1)
8739 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8740         .sched_state_renewal = nfs41_proc_async_sequence,
8741         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8742         .renew_lease = nfs4_proc_sequence,
8743 };
8744 #endif
8745
8746 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8747         .get_locations = _nfs40_proc_get_locations,
8748         .fsid_present = _nfs40_proc_fsid_present,
8749 };
8750
8751 #if defined(CONFIG_NFS_V4_1)
8752 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8753         .get_locations = _nfs41_proc_get_locations,
8754         .fsid_present = _nfs41_proc_fsid_present,
8755 };
8756 #endif  /* CONFIG_NFS_V4_1 */
8757
8758 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8759         .minor_version = 0,
8760         .init_caps = NFS_CAP_READDIRPLUS
8761                 | NFS_CAP_ATOMIC_OPEN
8762                 | NFS_CAP_POSIX_LOCK,
8763         .init_client = nfs40_init_client,
8764         .shutdown_client = nfs40_shutdown_client,
8765         .match_stateid = nfs4_match_stateid,
8766         .find_root_sec = nfs4_find_root_sec,
8767         .free_lock_state = nfs4_release_lockowner,
8768         .alloc_seqid = nfs_alloc_seqid,
8769         .call_sync_ops = &nfs40_call_sync_ops,
8770         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8771         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8772         .state_renewal_ops = &nfs40_state_renewal_ops,
8773         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8774 };
8775
8776 #if defined(CONFIG_NFS_V4_1)
8777 static struct nfs_seqid *
8778 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8779 {
8780         return NULL;
8781 }
8782
8783 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8784         .minor_version = 1,
8785         .init_caps = NFS_CAP_READDIRPLUS
8786                 | NFS_CAP_ATOMIC_OPEN
8787                 | NFS_CAP_POSIX_LOCK
8788                 | NFS_CAP_STATEID_NFSV41
8789                 | NFS_CAP_ATOMIC_OPEN_V1,
8790         .init_client = nfs41_init_client,
8791         .shutdown_client = nfs41_shutdown_client,
8792         .match_stateid = nfs41_match_stateid,
8793         .find_root_sec = nfs41_find_root_sec,
8794         .free_lock_state = nfs41_free_lock_state,
8795         .alloc_seqid = nfs_alloc_no_seqid,
8796         .call_sync_ops = &nfs41_call_sync_ops,
8797         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8798         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8799         .state_renewal_ops = &nfs41_state_renewal_ops,
8800         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8801 };
8802 #endif
8803
8804 #if defined(CONFIG_NFS_V4_2)
8805 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8806         .minor_version = 2,
8807         .init_caps = NFS_CAP_READDIRPLUS
8808                 | NFS_CAP_ATOMIC_OPEN
8809                 | NFS_CAP_POSIX_LOCK
8810                 | NFS_CAP_STATEID_NFSV41
8811                 | NFS_CAP_ATOMIC_OPEN_V1
8812                 | NFS_CAP_ALLOCATE
8813                 | NFS_CAP_COPY
8814                 | NFS_CAP_DEALLOCATE
8815                 | NFS_CAP_SEEK
8816                 | NFS_CAP_LAYOUTSTATS
8817                 | NFS_CAP_CLONE,
8818         .init_client = nfs41_init_client,
8819         .shutdown_client = nfs41_shutdown_client,
8820         .match_stateid = nfs41_match_stateid,
8821         .find_root_sec = nfs41_find_root_sec,
8822         .free_lock_state = nfs41_free_lock_state,
8823         .call_sync_ops = &nfs41_call_sync_ops,
8824         .alloc_seqid = nfs_alloc_no_seqid,
8825         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8826         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8827         .state_renewal_ops = &nfs41_state_renewal_ops,
8828         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8829 };
8830 #endif
8831
8832 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8833         [0] = &nfs_v4_0_minor_ops,
8834 #if defined(CONFIG_NFS_V4_1)
8835         [1] = &nfs_v4_1_minor_ops,
8836 #endif
8837 #if defined(CONFIG_NFS_V4_2)
8838         [2] = &nfs_v4_2_minor_ops,
8839 #endif
8840 };
8841
8842 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
8843 {
8844         ssize_t error, error2;
8845
8846         error = generic_listxattr(dentry, list, size);
8847         if (error < 0)
8848                 return error;
8849         if (list) {
8850                 list += error;
8851                 size -= error;
8852         }
8853
8854         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
8855         if (error2 < 0)
8856                 return error2;
8857         return error + error2;
8858 }
8859
8860 static const struct inode_operations nfs4_dir_inode_operations = {
8861         .create         = nfs_create,
8862         .lookup         = nfs_lookup,
8863         .atomic_open    = nfs_atomic_open,
8864         .link           = nfs_link,
8865         .unlink         = nfs_unlink,
8866         .symlink        = nfs_symlink,
8867         .mkdir          = nfs_mkdir,
8868         .rmdir          = nfs_rmdir,
8869         .mknod          = nfs_mknod,
8870         .rename         = nfs_rename,
8871         .permission     = nfs_permission,
8872         .getattr        = nfs_getattr,
8873         .setattr        = nfs_setattr,
8874         .getxattr       = generic_getxattr,
8875         .setxattr       = generic_setxattr,
8876         .listxattr      = nfs4_listxattr,
8877         .removexattr    = generic_removexattr,
8878 };
8879
8880 static const struct inode_operations nfs4_file_inode_operations = {
8881         .permission     = nfs_permission,
8882         .getattr        = nfs_getattr,
8883         .setattr        = nfs_setattr,
8884         .getxattr       = generic_getxattr,
8885         .setxattr       = generic_setxattr,
8886         .listxattr      = nfs4_listxattr,
8887         .removexattr    = generic_removexattr,
8888 };
8889
8890 const struct nfs_rpc_ops nfs_v4_clientops = {
8891         .version        = 4,                    /* protocol version */
8892         .dentry_ops     = &nfs4_dentry_operations,
8893         .dir_inode_ops  = &nfs4_dir_inode_operations,
8894         .file_inode_ops = &nfs4_file_inode_operations,
8895         .file_ops       = &nfs4_file_operations,
8896         .getroot        = nfs4_proc_get_root,
8897         .submount       = nfs4_submount,
8898         .try_mount      = nfs4_try_mount,
8899         .getattr        = nfs4_proc_getattr,
8900         .setattr        = nfs4_proc_setattr,
8901         .lookup         = nfs4_proc_lookup,
8902         .access         = nfs4_proc_access,
8903         .readlink       = nfs4_proc_readlink,
8904         .create         = nfs4_proc_create,
8905         .remove         = nfs4_proc_remove,
8906         .unlink_setup   = nfs4_proc_unlink_setup,
8907         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8908         .unlink_done    = nfs4_proc_unlink_done,
8909         .rename_setup   = nfs4_proc_rename_setup,
8910         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8911         .rename_done    = nfs4_proc_rename_done,
8912         .link           = nfs4_proc_link,
8913         .symlink        = nfs4_proc_symlink,
8914         .mkdir          = nfs4_proc_mkdir,
8915         .rmdir          = nfs4_proc_remove,
8916         .readdir        = nfs4_proc_readdir,
8917         .mknod          = nfs4_proc_mknod,
8918         .statfs         = nfs4_proc_statfs,
8919         .fsinfo         = nfs4_proc_fsinfo,
8920         .pathconf       = nfs4_proc_pathconf,
8921         .set_capabilities = nfs4_server_capabilities,
8922         .decode_dirent  = nfs4_decode_dirent,
8923         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8924         .read_setup     = nfs4_proc_read_setup,
8925         .read_done      = nfs4_read_done,
8926         .write_setup    = nfs4_proc_write_setup,
8927         .write_done     = nfs4_write_done,
8928         .commit_setup   = nfs4_proc_commit_setup,
8929         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8930         .commit_done    = nfs4_commit_done,
8931         .lock           = nfs4_proc_lock,
8932         .clear_acl_cache = nfs4_zap_acl_attr,
8933         .close_context  = nfs4_close_context,
8934         .open_context   = nfs4_atomic_open,
8935         .have_delegation = nfs4_have_delegation,
8936         .return_delegation = nfs4_inode_return_delegation,
8937         .alloc_client   = nfs4_alloc_client,
8938         .init_client    = nfs4_init_client,
8939         .free_client    = nfs4_free_client,
8940         .create_server  = nfs4_create_server,
8941         .clone_server   = nfs_clone_server,
8942 };
8943
8944 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8945         .name   = XATTR_NAME_NFSV4_ACL,
8946         .list   = nfs4_xattr_list_nfs4_acl,
8947         .get    = nfs4_xattr_get_nfs4_acl,
8948         .set    = nfs4_xattr_set_nfs4_acl,
8949 };
8950
8951 const struct xattr_handler *nfs4_xattr_handlers[] = {
8952         &nfs4_xattr_nfs4_acl_handler,
8953 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8954         &nfs4_xattr_nfs4_label_handler,
8955 #endif
8956         NULL
8957 };
8958
8959 /*
8960  * Local variables:
8961  *  c-basic-offset: 8
8962  * End:
8963  */