pnfs: only tear down lsegs that precede seqid in LAYOUTRETURN args
[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         struct inode *inode = exception->inode;
367         int ret = errorcode;
368
369         exception->delay = 0;
370         exception->recovering = 0;
371         exception->retry = 0;
372         switch(errorcode) {
373                 case 0:
374                         return 0;
375                 case -NFS4ERR_OPENMODE:
376                 case -NFS4ERR_DELEG_REVOKED:
377                 case -NFS4ERR_ADMIN_REVOKED:
378                 case -NFS4ERR_BAD_STATEID:
379                         if (inode && nfs_async_inode_return_delegation(inode,
380                                                 NULL) == 0)
381                                 goto wait_on_recovery;
382                         if (state == NULL)
383                                 break;
384                         ret = nfs4_schedule_stateid_recovery(server, state);
385                         if (ret < 0)
386                                 break;
387                         goto wait_on_recovery;
388                 case -NFS4ERR_EXPIRED:
389                         if (state != NULL) {
390                                 ret = nfs4_schedule_stateid_recovery(server, state);
391                                 if (ret < 0)
392                                         break;
393                         }
394                 case -NFS4ERR_STALE_STATEID:
395                 case -NFS4ERR_STALE_CLIENTID:
396                         nfs4_schedule_lease_recovery(clp);
397                         goto wait_on_recovery;
398                 case -NFS4ERR_MOVED:
399                         ret = nfs4_schedule_migration_recovery(server);
400                         if (ret < 0)
401                                 break;
402                         goto wait_on_recovery;
403                 case -NFS4ERR_LEASE_MOVED:
404                         nfs4_schedule_lease_moved_recovery(clp);
405                         goto wait_on_recovery;
406 #if defined(CONFIG_NFS_V4_1)
407                 case -NFS4ERR_BADSESSION:
408                 case -NFS4ERR_BADSLOT:
409                 case -NFS4ERR_BAD_HIGH_SLOT:
410                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
411                 case -NFS4ERR_DEADSESSION:
412                 case -NFS4ERR_SEQ_FALSE_RETRY:
413                 case -NFS4ERR_SEQ_MISORDERED:
414                         dprintk("%s ERROR: %d Reset session\n", __func__,
415                                 errorcode);
416                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
417                         goto wait_on_recovery;
418 #endif /* defined(CONFIG_NFS_V4_1) */
419                 case -NFS4ERR_FILE_OPEN:
420                         if (exception->timeout > HZ) {
421                                 /* We have retried a decent amount, time to
422                                  * fail
423                                  */
424                                 ret = -EBUSY;
425                                 break;
426                         }
427                 case -NFS4ERR_DELAY:
428                         nfs_inc_server_stats(server, NFSIOS_DELAY);
429                 case -NFS4ERR_GRACE:
430                         exception->delay = 1;
431                         return 0;
432
433                 case -NFS4ERR_RETRY_UNCACHED_REP:
434                 case -NFS4ERR_OLD_STATEID:
435                         exception->retry = 1;
436                         break;
437                 case -NFS4ERR_BADOWNER:
438                         /* The following works around a Linux server bug! */
439                 case -NFS4ERR_BADNAME:
440                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
441                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
442                                 exception->retry = 1;
443                                 printk(KERN_WARNING "NFS: v4 server %s "
444                                                 "does not accept raw "
445                                                 "uid/gids. "
446                                                 "Reenabling the idmapper.\n",
447                                                 server->nfs_client->cl_hostname);
448                         }
449         }
450         /* We failed to handle the error */
451         return nfs4_map_errors(ret);
452 wait_on_recovery:
453         exception->recovering = 1;
454         return 0;
455 }
456
457 /* This is the error handling routine for processes that are allowed
458  * to sleep.
459  */
460 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
461 {
462         struct nfs_client *clp = server->nfs_client;
463         int ret;
464
465         ret = nfs4_do_handle_exception(server, errorcode, exception);
466         if (exception->delay) {
467                 ret = nfs4_delay(server->client, &exception->timeout);
468                 goto out_retry;
469         }
470         if (exception->recovering) {
471                 ret = nfs4_wait_clnt_recover(clp);
472                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
473                         return -EIO;
474                 goto out_retry;
475         }
476         return ret;
477 out_retry:
478         if (ret == 0)
479                 exception->retry = 1;
480         return ret;
481 }
482
483 static int
484 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
485                 int errorcode, struct nfs4_exception *exception)
486 {
487         struct nfs_client *clp = server->nfs_client;
488         int ret;
489
490         ret = nfs4_do_handle_exception(server, errorcode, exception);
491         if (exception->delay) {
492                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
493                 goto out_retry;
494         }
495         if (exception->recovering) {
496                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
497                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
498                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
499                 goto out_retry;
500         }
501         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
502                 ret = -EIO;
503         return ret;
504 out_retry:
505         if (ret == 0)
506                 exception->retry = 1;
507         return ret;
508 }
509
510 static int
511 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
512                         struct nfs4_state *state, long *timeout)
513 {
514         struct nfs4_exception exception = {
515                 .state = state,
516         };
517
518         if (task->tk_status >= 0)
519                 return 0;
520         if (timeout)
521                 exception.timeout = *timeout;
522         task->tk_status = nfs4_async_handle_exception(task, server,
523                         task->tk_status,
524                         &exception);
525         if (exception.delay && timeout)
526                 *timeout = exception.timeout;
527         if (exception.retry)
528                 return -EAGAIN;
529         return 0;
530 }
531
532 /*
533  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
534  * or 'false' otherwise.
535  */
536 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
537 {
538         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
539
540         if (flavor == RPC_AUTH_GSS_KRB5I ||
541             flavor == RPC_AUTH_GSS_KRB5P)
542                 return true;
543
544         return false;
545 }
546
547 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
548 {
549         spin_lock(&clp->cl_lock);
550         if (time_before(clp->cl_last_renewal,timestamp))
551                 clp->cl_last_renewal = timestamp;
552         spin_unlock(&clp->cl_lock);
553 }
554
555 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
556 {
557         struct nfs_client *clp = server->nfs_client;
558
559         if (!nfs4_has_session(clp))
560                 do_renew_lease(clp, timestamp);
561 }
562
563 struct nfs4_call_sync_data {
564         const struct nfs_server *seq_server;
565         struct nfs4_sequence_args *seq_args;
566         struct nfs4_sequence_res *seq_res;
567 };
568
569 void nfs4_init_sequence(struct nfs4_sequence_args *args,
570                         struct nfs4_sequence_res *res, int cache_reply)
571 {
572         args->sa_slot = NULL;
573         args->sa_cache_this = cache_reply;
574         args->sa_privileged = 0;
575
576         res->sr_slot = NULL;
577 }
578
579 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
580 {
581         args->sa_privileged = 1;
582 }
583
584 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
585                          struct nfs4_sequence_args *args,
586                          struct nfs4_sequence_res *res,
587                          struct rpc_task *task)
588 {
589         struct nfs4_slot *slot;
590
591         /* slot already allocated? */
592         if (res->sr_slot != NULL)
593                 goto out_start;
594
595         spin_lock(&tbl->slot_tbl_lock);
596         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
597                 goto out_sleep;
598
599         slot = nfs4_alloc_slot(tbl);
600         if (IS_ERR(slot)) {
601                 if (slot == ERR_PTR(-ENOMEM))
602                         task->tk_timeout = HZ >> 2;
603                 goto out_sleep;
604         }
605         spin_unlock(&tbl->slot_tbl_lock);
606
607         args->sa_slot = slot;
608         res->sr_slot = slot;
609
610 out_start:
611         rpc_call_start(task);
612         return 0;
613
614 out_sleep:
615         if (args->sa_privileged)
616                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
617                                 NULL, RPC_PRIORITY_PRIVILEGED);
618         else
619                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
620         spin_unlock(&tbl->slot_tbl_lock);
621         return -EAGAIN;
622 }
623 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
624
625 static int nfs40_sequence_done(struct rpc_task *task,
626                                struct nfs4_sequence_res *res)
627 {
628         struct nfs4_slot *slot = res->sr_slot;
629         struct nfs4_slot_table *tbl;
630
631         if (slot == NULL)
632                 goto out;
633
634         tbl = slot->table;
635         spin_lock(&tbl->slot_tbl_lock);
636         if (!nfs41_wake_and_assign_slot(tbl, slot))
637                 nfs4_free_slot(tbl, slot);
638         spin_unlock(&tbl->slot_tbl_lock);
639
640         res->sr_slot = NULL;
641 out:
642         return 1;
643 }
644
645 #if defined(CONFIG_NFS_V4_1)
646
647 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
648 {
649         struct nfs4_session *session;
650         struct nfs4_slot_table *tbl;
651         struct nfs4_slot *slot = res->sr_slot;
652         bool send_new_highest_used_slotid = false;
653
654         tbl = slot->table;
655         session = tbl->session;
656
657         spin_lock(&tbl->slot_tbl_lock);
658         /* Be nice to the server: try to ensure that the last transmitted
659          * value for highest_user_slotid <= target_highest_slotid
660          */
661         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
662                 send_new_highest_used_slotid = true;
663
664         if (nfs41_wake_and_assign_slot(tbl, slot)) {
665                 send_new_highest_used_slotid = false;
666                 goto out_unlock;
667         }
668         nfs4_free_slot(tbl, slot);
669
670         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
671                 send_new_highest_used_slotid = false;
672 out_unlock:
673         spin_unlock(&tbl->slot_tbl_lock);
674         res->sr_slot = NULL;
675         if (send_new_highest_used_slotid)
676                 nfs41_notify_server(session->clp);
677 }
678
679 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
680 {
681         struct nfs4_session *session;
682         struct nfs4_slot *slot = res->sr_slot;
683         struct nfs_client *clp;
684         bool interrupted = false;
685         int ret = 1;
686
687         if (slot == NULL)
688                 goto out_noaction;
689         /* don't increment the sequence number if the task wasn't sent */
690         if (!RPC_WAS_SENT(task))
691                 goto out;
692
693         session = slot->table->session;
694
695         if (slot->interrupted) {
696                 slot->interrupted = 0;
697                 interrupted = true;
698         }
699
700         trace_nfs4_sequence_done(session, res);
701         /* Check the SEQUENCE operation status */
702         switch (res->sr_status) {
703         case 0:
704                 /* Update the slot's sequence and clientid lease timer */
705                 ++slot->seq_nr;
706                 clp = session->clp;
707                 do_renew_lease(clp, res->sr_timestamp);
708                 /* Check sequence flags */
709                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
710                 nfs41_update_target_slotid(slot->table, slot, res);
711                 break;
712         case 1:
713                 /*
714                  * sr_status remains 1 if an RPC level error occurred.
715                  * The server may or may not have processed the sequence
716                  * operation..
717                  * Mark the slot as having hosted an interrupted RPC call.
718                  */
719                 slot->interrupted = 1;
720                 goto out;
721         case -NFS4ERR_DELAY:
722                 /* The server detected a resend of the RPC call and
723                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
724                  * of RFC5661.
725                  */
726                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
727                         __func__,
728                         slot->slot_nr,
729                         slot->seq_nr);
730                 goto out_retry;
731         case -NFS4ERR_BADSLOT:
732                 /*
733                  * The slot id we used was probably retired. Try again
734                  * using a different slot id.
735                  */
736                 goto retry_nowait;
737         case -NFS4ERR_SEQ_MISORDERED:
738                 /*
739                  * Was the last operation on this sequence interrupted?
740                  * If so, retry after bumping the sequence number.
741                  */
742                 if (interrupted) {
743                         ++slot->seq_nr;
744                         goto retry_nowait;
745                 }
746                 /*
747                  * Could this slot have been previously retired?
748                  * If so, then the server may be expecting seq_nr = 1!
749                  */
750                 if (slot->seq_nr != 1) {
751                         slot->seq_nr = 1;
752                         goto retry_nowait;
753                 }
754                 break;
755         case -NFS4ERR_SEQ_FALSE_RETRY:
756                 ++slot->seq_nr;
757                 goto retry_nowait;
758         default:
759                 /* Just update the slot sequence no. */
760                 ++slot->seq_nr;
761         }
762 out:
763         /* The session may be reset by one of the error handlers. */
764         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
765         nfs41_sequence_free_slot(res);
766 out_noaction:
767         return ret;
768 retry_nowait:
769         if (rpc_restart_call_prepare(task)) {
770                 task->tk_status = 0;
771                 ret = 0;
772         }
773         goto out;
774 out_retry:
775         if (!rpc_restart_call(task))
776                 goto out;
777         rpc_delay(task, NFS4_POLL_RETRY_MAX);
778         return 0;
779 }
780 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
781
782 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
783 {
784         if (res->sr_slot == NULL)
785                 return 1;
786         if (!res->sr_slot->table->session)
787                 return nfs40_sequence_done(task, res);
788         return nfs41_sequence_done(task, res);
789 }
790 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
791
792 int nfs41_setup_sequence(struct nfs4_session *session,
793                                 struct nfs4_sequence_args *args,
794                                 struct nfs4_sequence_res *res,
795                                 struct rpc_task *task)
796 {
797         struct nfs4_slot *slot;
798         struct nfs4_slot_table *tbl;
799
800         dprintk("--> %s\n", __func__);
801         /* slot already allocated? */
802         if (res->sr_slot != NULL)
803                 goto out_success;
804
805         tbl = &session->fc_slot_table;
806
807         task->tk_timeout = 0;
808
809         spin_lock(&tbl->slot_tbl_lock);
810         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
811             !args->sa_privileged) {
812                 /* The state manager will wait until the slot table is empty */
813                 dprintk("%s session is draining\n", __func__);
814                 goto out_sleep;
815         }
816
817         slot = nfs4_alloc_slot(tbl);
818         if (IS_ERR(slot)) {
819                 /* If out of memory, try again in 1/4 second */
820                 if (slot == ERR_PTR(-ENOMEM))
821                         task->tk_timeout = HZ >> 2;
822                 dprintk("<-- %s: no free slots\n", __func__);
823                 goto out_sleep;
824         }
825         spin_unlock(&tbl->slot_tbl_lock);
826
827         args->sa_slot = slot;
828
829         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
830                         slot->slot_nr, slot->seq_nr);
831
832         res->sr_slot = slot;
833         res->sr_timestamp = jiffies;
834         res->sr_status_flags = 0;
835         /*
836          * sr_status is only set in decode_sequence, and so will remain
837          * set to 1 if an rpc level failure occurs.
838          */
839         res->sr_status = 1;
840         trace_nfs4_setup_sequence(session, args);
841 out_success:
842         rpc_call_start(task);
843         return 0;
844 out_sleep:
845         /* Privileged tasks are queued with top priority */
846         if (args->sa_privileged)
847                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
848                                 NULL, RPC_PRIORITY_PRIVILEGED);
849         else
850                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
851         spin_unlock(&tbl->slot_tbl_lock);
852         return -EAGAIN;
853 }
854 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
855
856 static int nfs4_setup_sequence(const struct nfs_server *server,
857                                struct nfs4_sequence_args *args,
858                                struct nfs4_sequence_res *res,
859                                struct rpc_task *task)
860 {
861         struct nfs4_session *session = nfs4_get_session(server);
862         int ret = 0;
863
864         if (!session)
865                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
866                                             args, res, task);
867
868         dprintk("--> %s clp %p session %p sr_slot %u\n",
869                 __func__, session->clp, session, res->sr_slot ?
870                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
871
872         ret = nfs41_setup_sequence(session, args, res, task);
873
874         dprintk("<-- %s status=%d\n", __func__, ret);
875         return ret;
876 }
877
878 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
879 {
880         struct nfs4_call_sync_data *data = calldata;
881         struct nfs4_session *session = nfs4_get_session(data->seq_server);
882
883         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
884
885         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
886 }
887
888 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
889 {
890         struct nfs4_call_sync_data *data = calldata;
891
892         nfs41_sequence_done(task, data->seq_res);
893 }
894
895 static const struct rpc_call_ops nfs41_call_sync_ops = {
896         .rpc_call_prepare = nfs41_call_sync_prepare,
897         .rpc_call_done = nfs41_call_sync_done,
898 };
899
900 #else   /* !CONFIG_NFS_V4_1 */
901
902 static int nfs4_setup_sequence(const struct nfs_server *server,
903                                struct nfs4_sequence_args *args,
904                                struct nfs4_sequence_res *res,
905                                struct rpc_task *task)
906 {
907         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
908                                     args, res, task);
909 }
910
911 int nfs4_sequence_done(struct rpc_task *task,
912                        struct nfs4_sequence_res *res)
913 {
914         return nfs40_sequence_done(task, res);
915 }
916 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
917
918 #endif  /* !CONFIG_NFS_V4_1 */
919
920 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
921 {
922         struct nfs4_call_sync_data *data = calldata;
923         nfs4_setup_sequence(data->seq_server,
924                                 data->seq_args, data->seq_res, task);
925 }
926
927 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
928 {
929         struct nfs4_call_sync_data *data = calldata;
930         nfs4_sequence_done(task, data->seq_res);
931 }
932
933 static const struct rpc_call_ops nfs40_call_sync_ops = {
934         .rpc_call_prepare = nfs40_call_sync_prepare,
935         .rpc_call_done = nfs40_call_sync_done,
936 };
937
938 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
939                                    struct nfs_server *server,
940                                    struct rpc_message *msg,
941                                    struct nfs4_sequence_args *args,
942                                    struct nfs4_sequence_res *res)
943 {
944         int ret;
945         struct rpc_task *task;
946         struct nfs_client *clp = server->nfs_client;
947         struct nfs4_call_sync_data data = {
948                 .seq_server = server,
949                 .seq_args = args,
950                 .seq_res = res,
951         };
952         struct rpc_task_setup task_setup = {
953                 .rpc_client = clnt,
954                 .rpc_message = msg,
955                 .callback_ops = clp->cl_mvops->call_sync_ops,
956                 .callback_data = &data
957         };
958
959         task = rpc_run_task(&task_setup);
960         if (IS_ERR(task))
961                 ret = PTR_ERR(task);
962         else {
963                 ret = task->tk_status;
964                 rpc_put_task(task);
965         }
966         return ret;
967 }
968
969 int nfs4_call_sync(struct rpc_clnt *clnt,
970                    struct nfs_server *server,
971                    struct rpc_message *msg,
972                    struct nfs4_sequence_args *args,
973                    struct nfs4_sequence_res *res,
974                    int cache_reply)
975 {
976         nfs4_init_sequence(args, res, cache_reply);
977         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
978 }
979
980 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
981 {
982         struct nfs_inode *nfsi = NFS_I(dir);
983
984         spin_lock(&dir->i_lock);
985         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
986         if (!cinfo->atomic || cinfo->before != dir->i_version)
987                 nfs_force_lookup_revalidate(dir);
988         dir->i_version = cinfo->after;
989         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
990         nfs_fscache_invalidate(dir);
991         spin_unlock(&dir->i_lock);
992 }
993
994 struct nfs4_opendata {
995         struct kref kref;
996         struct nfs_openargs o_arg;
997         struct nfs_openres o_res;
998         struct nfs_open_confirmargs c_arg;
999         struct nfs_open_confirmres c_res;
1000         struct nfs4_string owner_name;
1001         struct nfs4_string group_name;
1002         struct nfs4_label *a_label;
1003         struct nfs_fattr f_attr;
1004         struct nfs4_label *f_label;
1005         struct dentry *dir;
1006         struct dentry *dentry;
1007         struct nfs4_state_owner *owner;
1008         struct nfs4_state *state;
1009         struct iattr attrs;
1010         unsigned long timestamp;
1011         unsigned int rpc_done : 1;
1012         unsigned int file_created : 1;
1013         unsigned int is_recover : 1;
1014         int rpc_status;
1015         int cancelled;
1016 };
1017
1018 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1019                 int err, struct nfs4_exception *exception)
1020 {
1021         if (err != -EINVAL)
1022                 return false;
1023         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1024                 return false;
1025         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1026         exception->retry = 1;
1027         return true;
1028 }
1029
1030 static u32
1031 nfs4_map_atomic_open_share(struct nfs_server *server,
1032                 fmode_t fmode, int openflags)
1033 {
1034         u32 res = 0;
1035
1036         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1037         case FMODE_READ:
1038                 res = NFS4_SHARE_ACCESS_READ;
1039                 break;
1040         case FMODE_WRITE:
1041                 res = NFS4_SHARE_ACCESS_WRITE;
1042                 break;
1043         case FMODE_READ|FMODE_WRITE:
1044                 res = NFS4_SHARE_ACCESS_BOTH;
1045         }
1046         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1047                 goto out;
1048         /* Want no delegation if we're using O_DIRECT */
1049         if (openflags & O_DIRECT)
1050                 res |= NFS4_SHARE_WANT_NO_DELEG;
1051 out:
1052         return res;
1053 }
1054
1055 static enum open_claim_type4
1056 nfs4_map_atomic_open_claim(struct nfs_server *server,
1057                 enum open_claim_type4 claim)
1058 {
1059         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1060                 return claim;
1061         switch (claim) {
1062         default:
1063                 return claim;
1064         case NFS4_OPEN_CLAIM_FH:
1065                 return NFS4_OPEN_CLAIM_NULL;
1066         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1067                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1068         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1069                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1070         }
1071 }
1072
1073 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1074 {
1075         p->o_res.f_attr = &p->f_attr;
1076         p->o_res.f_label = p->f_label;
1077         p->o_res.seqid = p->o_arg.seqid;
1078         p->c_res.seqid = p->c_arg.seqid;
1079         p->o_res.server = p->o_arg.server;
1080         p->o_res.access_request = p->o_arg.access;
1081         nfs_fattr_init(&p->f_attr);
1082         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1083 }
1084
1085 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1086                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1087                 const struct iattr *attrs,
1088                 struct nfs4_label *label,
1089                 enum open_claim_type4 claim,
1090                 gfp_t gfp_mask)
1091 {
1092         struct dentry *parent = dget_parent(dentry);
1093         struct inode *dir = d_inode(parent);
1094         struct nfs_server *server = NFS_SERVER(dir);
1095         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1096         struct nfs4_opendata *p;
1097
1098         p = kzalloc(sizeof(*p), gfp_mask);
1099         if (p == NULL)
1100                 goto err;
1101
1102         p->f_label = nfs4_label_alloc(server, gfp_mask);
1103         if (IS_ERR(p->f_label))
1104                 goto err_free_p;
1105
1106         p->a_label = nfs4_label_alloc(server, gfp_mask);
1107         if (IS_ERR(p->a_label))
1108                 goto err_free_f;
1109
1110         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1111         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1112         if (IS_ERR(p->o_arg.seqid))
1113                 goto err_free_label;
1114         nfs_sb_active(dentry->d_sb);
1115         p->dentry = dget(dentry);
1116         p->dir = parent;
1117         p->owner = sp;
1118         atomic_inc(&sp->so_count);
1119         p->o_arg.open_flags = flags;
1120         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1121         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1122                         fmode, flags);
1123         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1124          * will return permission denied for all bits until close */
1125         if (!(flags & O_EXCL)) {
1126                 /* ask server to check for all possible rights as results
1127                  * are cached */
1128                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1129                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1130         }
1131         p->o_arg.clientid = server->nfs_client->cl_clientid;
1132         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1133         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1134         p->o_arg.name = &dentry->d_name;
1135         p->o_arg.server = server;
1136         p->o_arg.bitmask = nfs4_bitmask(server, label);
1137         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1138         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1139         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1140         switch (p->o_arg.claim) {
1141         case NFS4_OPEN_CLAIM_NULL:
1142         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1143         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1144                 p->o_arg.fh = NFS_FH(dir);
1145                 break;
1146         case NFS4_OPEN_CLAIM_PREVIOUS:
1147         case NFS4_OPEN_CLAIM_FH:
1148         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1149         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1150                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1151         }
1152         if (attrs != NULL && attrs->ia_valid != 0) {
1153                 __u32 verf[2];
1154
1155                 p->o_arg.u.attrs = &p->attrs;
1156                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1157
1158                 verf[0] = jiffies;
1159                 verf[1] = current->pid;
1160                 memcpy(p->o_arg.u.verifier.data, verf,
1161                                 sizeof(p->o_arg.u.verifier.data));
1162         }
1163         p->c_arg.fh = &p->o_res.fh;
1164         p->c_arg.stateid = &p->o_res.stateid;
1165         p->c_arg.seqid = p->o_arg.seqid;
1166         nfs4_init_opendata_res(p);
1167         kref_init(&p->kref);
1168         return p;
1169
1170 err_free_label:
1171         nfs4_label_free(p->a_label);
1172 err_free_f:
1173         nfs4_label_free(p->f_label);
1174 err_free_p:
1175         kfree(p);
1176 err:
1177         dput(parent);
1178         return NULL;
1179 }
1180
1181 static void nfs4_opendata_free(struct kref *kref)
1182 {
1183         struct nfs4_opendata *p = container_of(kref,
1184                         struct nfs4_opendata, kref);
1185         struct super_block *sb = p->dentry->d_sb;
1186
1187         nfs_free_seqid(p->o_arg.seqid);
1188         if (p->state != NULL)
1189                 nfs4_put_open_state(p->state);
1190         nfs4_put_state_owner(p->owner);
1191
1192         nfs4_label_free(p->a_label);
1193         nfs4_label_free(p->f_label);
1194
1195         dput(p->dir);
1196         dput(p->dentry);
1197         nfs_sb_deactive(sb);
1198         nfs_fattr_free_names(&p->f_attr);
1199         kfree(p->f_attr.mdsthreshold);
1200         kfree(p);
1201 }
1202
1203 static void nfs4_opendata_put(struct nfs4_opendata *p)
1204 {
1205         if (p != NULL)
1206                 kref_put(&p->kref, nfs4_opendata_free);
1207 }
1208
1209 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1210 {
1211         int ret;
1212
1213         ret = rpc_wait_for_completion_task(task);
1214         return ret;
1215 }
1216
1217 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1218                 fmode_t fmode)
1219 {
1220         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1221         case FMODE_READ|FMODE_WRITE:
1222                 return state->n_rdwr != 0;
1223         case FMODE_WRITE:
1224                 return state->n_wronly != 0;
1225         case FMODE_READ:
1226                 return state->n_rdonly != 0;
1227         }
1228         WARN_ON_ONCE(1);
1229         return false;
1230 }
1231
1232 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1233 {
1234         int ret = 0;
1235
1236         if (open_mode & (O_EXCL|O_TRUNC))
1237                 goto out;
1238         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1239                 case FMODE_READ:
1240                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1241                                 && state->n_rdonly != 0;
1242                         break;
1243                 case FMODE_WRITE:
1244                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1245                                 && state->n_wronly != 0;
1246                         break;
1247                 case FMODE_READ|FMODE_WRITE:
1248                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1249                                 && state->n_rdwr != 0;
1250         }
1251 out:
1252         return ret;
1253 }
1254
1255 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1256                 enum open_claim_type4 claim)
1257 {
1258         if (delegation == NULL)
1259                 return 0;
1260         if ((delegation->type & fmode) != fmode)
1261                 return 0;
1262         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1263                 return 0;
1264         switch (claim) {
1265         case NFS4_OPEN_CLAIM_NULL:
1266         case NFS4_OPEN_CLAIM_FH:
1267                 break;
1268         case NFS4_OPEN_CLAIM_PREVIOUS:
1269                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1270                         break;
1271         default:
1272                 return 0;
1273         }
1274         nfs_mark_delegation_referenced(delegation);
1275         return 1;
1276 }
1277
1278 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1279 {
1280         switch (fmode) {
1281                 case FMODE_WRITE:
1282                         state->n_wronly++;
1283                         break;
1284                 case FMODE_READ:
1285                         state->n_rdonly++;
1286                         break;
1287                 case FMODE_READ|FMODE_WRITE:
1288                         state->n_rdwr++;
1289         }
1290         nfs4_state_set_mode_locked(state, state->state | fmode);
1291 }
1292
1293 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1294 {
1295         struct nfs_client *clp = state->owner->so_server->nfs_client;
1296         bool need_recover = false;
1297
1298         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1299                 need_recover = true;
1300         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1301                 need_recover = true;
1302         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1303                 need_recover = true;
1304         if (need_recover)
1305                 nfs4_state_mark_reclaim_nograce(clp, state);
1306 }
1307
1308 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1309                 nfs4_stateid *stateid)
1310 {
1311         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1312                 return true;
1313         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1314                 nfs_test_and_clear_all_open_stateid(state);
1315                 return true;
1316         }
1317         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1318                 return true;
1319         return false;
1320 }
1321
1322 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1323 {
1324         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1325                 return;
1326         if (state->n_wronly)
1327                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1328         if (state->n_rdonly)
1329                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1330         if (state->n_rdwr)
1331                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1332         set_bit(NFS_OPEN_STATE, &state->flags);
1333 }
1334
1335 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1336                 nfs4_stateid *arg_stateid,
1337                 nfs4_stateid *stateid, fmode_t fmode)
1338 {
1339         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1340         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1341         case FMODE_WRITE:
1342                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1343                 break;
1344         case FMODE_READ:
1345                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1346                 break;
1347         case 0:
1348                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1349                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1350                 clear_bit(NFS_OPEN_STATE, &state->flags);
1351         }
1352         if (stateid == NULL)
1353                 return;
1354         /* Handle races with OPEN */
1355         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1356             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1357             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1358                 nfs_resync_open_stateid_locked(state);
1359                 return;
1360         }
1361         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1362                 nfs4_stateid_copy(&state->stateid, stateid);
1363         nfs4_stateid_copy(&state->open_stateid, stateid);
1364 }
1365
1366 static void nfs_clear_open_stateid(struct nfs4_state *state,
1367         nfs4_stateid *arg_stateid,
1368         nfs4_stateid *stateid, fmode_t fmode)
1369 {
1370         write_seqlock(&state->seqlock);
1371         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1372         write_sequnlock(&state->seqlock);
1373         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1374                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1375 }
1376
1377 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1378 {
1379         switch (fmode) {
1380                 case FMODE_READ:
1381                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1382                         break;
1383                 case FMODE_WRITE:
1384                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1385                         break;
1386                 case FMODE_READ|FMODE_WRITE:
1387                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1388         }
1389         if (!nfs_need_update_open_stateid(state, stateid))
1390                 return;
1391         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1392                 nfs4_stateid_copy(&state->stateid, stateid);
1393         nfs4_stateid_copy(&state->open_stateid, stateid);
1394 }
1395
1396 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1397 {
1398         /*
1399          * Protect the call to nfs4_state_set_mode_locked and
1400          * serialise the stateid update
1401          */
1402         spin_lock(&state->owner->so_lock);
1403         write_seqlock(&state->seqlock);
1404         if (deleg_stateid != NULL) {
1405                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1406                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1407         }
1408         if (open_stateid != NULL)
1409                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1410         write_sequnlock(&state->seqlock);
1411         update_open_stateflags(state, fmode);
1412         spin_unlock(&state->owner->so_lock);
1413 }
1414
1415 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1416 {
1417         struct nfs_inode *nfsi = NFS_I(state->inode);
1418         struct nfs_delegation *deleg_cur;
1419         int ret = 0;
1420
1421         fmode &= (FMODE_READ|FMODE_WRITE);
1422
1423         rcu_read_lock();
1424         deleg_cur = rcu_dereference(nfsi->delegation);
1425         if (deleg_cur == NULL)
1426                 goto no_delegation;
1427
1428         spin_lock(&deleg_cur->lock);
1429         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1430            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1431             (deleg_cur->type & fmode) != fmode)
1432                 goto no_delegation_unlock;
1433
1434         if (delegation == NULL)
1435                 delegation = &deleg_cur->stateid;
1436         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1437                 goto no_delegation_unlock;
1438
1439         nfs_mark_delegation_referenced(deleg_cur);
1440         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1441         ret = 1;
1442 no_delegation_unlock:
1443         spin_unlock(&deleg_cur->lock);
1444 no_delegation:
1445         rcu_read_unlock();
1446
1447         if (!ret && open_stateid != NULL) {
1448                 __update_open_stateid(state, open_stateid, NULL, fmode);
1449                 ret = 1;
1450         }
1451         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1452                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1453
1454         return ret;
1455 }
1456
1457 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1458                 const nfs4_stateid *stateid)
1459 {
1460         struct nfs4_state *state = lsp->ls_state;
1461         bool ret = false;
1462
1463         spin_lock(&state->state_lock);
1464         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1465                 goto out_noupdate;
1466         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1467                 goto out_noupdate;
1468         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1469         ret = true;
1470 out_noupdate:
1471         spin_unlock(&state->state_lock);
1472         return ret;
1473 }
1474
1475 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1476 {
1477         struct nfs_delegation *delegation;
1478
1479         rcu_read_lock();
1480         delegation = rcu_dereference(NFS_I(inode)->delegation);
1481         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1482                 rcu_read_unlock();
1483                 return;
1484         }
1485         rcu_read_unlock();
1486         nfs4_inode_return_delegation(inode);
1487 }
1488
1489 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1490 {
1491         struct nfs4_state *state = opendata->state;
1492         struct nfs_inode *nfsi = NFS_I(state->inode);
1493         struct nfs_delegation *delegation;
1494         int open_mode = opendata->o_arg.open_flags;
1495         fmode_t fmode = opendata->o_arg.fmode;
1496         enum open_claim_type4 claim = opendata->o_arg.claim;
1497         nfs4_stateid stateid;
1498         int ret = -EAGAIN;
1499
1500         for (;;) {
1501                 spin_lock(&state->owner->so_lock);
1502                 if (can_open_cached(state, fmode, open_mode)) {
1503                         update_open_stateflags(state, fmode);
1504                         spin_unlock(&state->owner->so_lock);
1505                         goto out_return_state;
1506                 }
1507                 spin_unlock(&state->owner->so_lock);
1508                 rcu_read_lock();
1509                 delegation = rcu_dereference(nfsi->delegation);
1510                 if (!can_open_delegated(delegation, fmode, claim)) {
1511                         rcu_read_unlock();
1512                         break;
1513                 }
1514                 /* Save the delegation */
1515                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1516                 rcu_read_unlock();
1517                 nfs_release_seqid(opendata->o_arg.seqid);
1518                 if (!opendata->is_recover) {
1519                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1520                         if (ret != 0)
1521                                 goto out;
1522                 }
1523                 ret = -EAGAIN;
1524
1525                 /* Try to update the stateid using the delegation */
1526                 if (update_open_stateid(state, NULL, &stateid, fmode))
1527                         goto out_return_state;
1528         }
1529 out:
1530         return ERR_PTR(ret);
1531 out_return_state:
1532         atomic_inc(&state->count);
1533         return state;
1534 }
1535
1536 static void
1537 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1538 {
1539         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1540         struct nfs_delegation *delegation;
1541         int delegation_flags = 0;
1542
1543         rcu_read_lock();
1544         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1545         if (delegation)
1546                 delegation_flags = delegation->flags;
1547         rcu_read_unlock();
1548         switch (data->o_arg.claim) {
1549         default:
1550                 break;
1551         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1552         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1553                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1554                                    "returning a delegation for "
1555                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1556                                    clp->cl_hostname);
1557                 return;
1558         }
1559         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1560                 nfs_inode_set_delegation(state->inode,
1561                                          data->owner->so_cred,
1562                                          &data->o_res);
1563         else
1564                 nfs_inode_reclaim_delegation(state->inode,
1565                                              data->owner->so_cred,
1566                                              &data->o_res);
1567 }
1568
1569 /*
1570  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1571  * and update the nfs4_state.
1572  */
1573 static struct nfs4_state *
1574 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1575 {
1576         struct inode *inode = data->state->inode;
1577         struct nfs4_state *state = data->state;
1578         int ret;
1579
1580         if (!data->rpc_done) {
1581                 if (data->rpc_status) {
1582                         ret = data->rpc_status;
1583                         goto err;
1584                 }
1585                 /* cached opens have already been processed */
1586                 goto update;
1587         }
1588
1589         ret = nfs_refresh_inode(inode, &data->f_attr);
1590         if (ret)
1591                 goto err;
1592
1593         if (data->o_res.delegation_type != 0)
1594                 nfs4_opendata_check_deleg(data, state);
1595 update:
1596         update_open_stateid(state, &data->o_res.stateid, NULL,
1597                             data->o_arg.fmode);
1598         atomic_inc(&state->count);
1599
1600         return state;
1601 err:
1602         return ERR_PTR(ret);
1603
1604 }
1605
1606 static struct nfs4_state *
1607 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1608 {
1609         struct inode *inode;
1610         struct nfs4_state *state = NULL;
1611         int ret;
1612
1613         if (!data->rpc_done) {
1614                 state = nfs4_try_open_cached(data);
1615                 trace_nfs4_cached_open(data->state);
1616                 goto out;
1617         }
1618
1619         ret = -EAGAIN;
1620         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1621                 goto err;
1622         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1623         ret = PTR_ERR(inode);
1624         if (IS_ERR(inode))
1625                 goto err;
1626         ret = -ENOMEM;
1627         state = nfs4_get_open_state(inode, data->owner);
1628         if (state == NULL)
1629                 goto err_put_inode;
1630         if (data->o_res.delegation_type != 0)
1631                 nfs4_opendata_check_deleg(data, state);
1632         update_open_stateid(state, &data->o_res.stateid, NULL,
1633                         data->o_arg.fmode);
1634         iput(inode);
1635 out:
1636         nfs_release_seqid(data->o_arg.seqid);
1637         return state;
1638 err_put_inode:
1639         iput(inode);
1640 err:
1641         return ERR_PTR(ret);
1642 }
1643
1644 static struct nfs4_state *
1645 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1646 {
1647         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1648                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1649         return _nfs4_opendata_to_nfs4_state(data);
1650 }
1651
1652 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1653 {
1654         struct nfs_inode *nfsi = NFS_I(state->inode);
1655         struct nfs_open_context *ctx;
1656
1657         spin_lock(&state->inode->i_lock);
1658         list_for_each_entry(ctx, &nfsi->open_files, list) {
1659                 if (ctx->state != state)
1660                         continue;
1661                 get_nfs_open_context(ctx);
1662                 spin_unlock(&state->inode->i_lock);
1663                 return ctx;
1664         }
1665         spin_unlock(&state->inode->i_lock);
1666         return ERR_PTR(-ENOENT);
1667 }
1668
1669 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1670                 struct nfs4_state *state, enum open_claim_type4 claim)
1671 {
1672         struct nfs4_opendata *opendata;
1673
1674         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1675                         NULL, NULL, claim, GFP_NOFS);
1676         if (opendata == NULL)
1677                 return ERR_PTR(-ENOMEM);
1678         opendata->state = state;
1679         atomic_inc(&state->count);
1680         return opendata;
1681 }
1682
1683 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1684                 fmode_t fmode)
1685 {
1686         struct nfs4_state *newstate;
1687         int ret;
1688
1689         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1690                 return 0;
1691         opendata->o_arg.open_flags = 0;
1692         opendata->o_arg.fmode = fmode;
1693         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1694                         NFS_SB(opendata->dentry->d_sb),
1695                         fmode, 0);
1696         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1697         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1698         nfs4_init_opendata_res(opendata);
1699         ret = _nfs4_recover_proc_open(opendata);
1700         if (ret != 0)
1701                 return ret; 
1702         newstate = nfs4_opendata_to_nfs4_state(opendata);
1703         if (IS_ERR(newstate))
1704                 return PTR_ERR(newstate);
1705         if (newstate != opendata->state)
1706                 ret = -ESTALE;
1707         nfs4_close_state(newstate, fmode);
1708         return ret;
1709 }
1710
1711 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1712 {
1713         int ret;
1714
1715         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1716         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1717         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1718         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1719         /* memory barrier prior to reading state->n_* */
1720         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1721         clear_bit(NFS_OPEN_STATE, &state->flags);
1722         smp_rmb();
1723         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1724         if (ret != 0)
1725                 return ret;
1726         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1727         if (ret != 0)
1728                 return ret;
1729         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1730         if (ret != 0)
1731                 return ret;
1732         /*
1733          * We may have performed cached opens for all three recoveries.
1734          * Check if we need to update the current stateid.
1735          */
1736         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1737             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1738                 write_seqlock(&state->seqlock);
1739                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1740                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1741                 write_sequnlock(&state->seqlock);
1742         }
1743         return 0;
1744 }
1745
1746 /*
1747  * OPEN_RECLAIM:
1748  *      reclaim state on the server after a reboot.
1749  */
1750 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1751 {
1752         struct nfs_delegation *delegation;
1753         struct nfs4_opendata *opendata;
1754         fmode_t delegation_type = 0;
1755         int status;
1756
1757         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1758                         NFS4_OPEN_CLAIM_PREVIOUS);
1759         if (IS_ERR(opendata))
1760                 return PTR_ERR(opendata);
1761         rcu_read_lock();
1762         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1763         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1764                 delegation_type = delegation->type;
1765         rcu_read_unlock();
1766         opendata->o_arg.u.delegation_type = delegation_type;
1767         status = nfs4_open_recover(opendata, state);
1768         nfs4_opendata_put(opendata);
1769         return status;
1770 }
1771
1772 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1773 {
1774         struct nfs_server *server = NFS_SERVER(state->inode);
1775         struct nfs4_exception exception = { };
1776         int err;
1777         do {
1778                 err = _nfs4_do_open_reclaim(ctx, state);
1779                 trace_nfs4_open_reclaim(ctx, 0, err);
1780                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1781                         continue;
1782                 if (err != -NFS4ERR_DELAY)
1783                         break;
1784                 nfs4_handle_exception(server, err, &exception);
1785         } while (exception.retry);
1786         return err;
1787 }
1788
1789 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1790 {
1791         struct nfs_open_context *ctx;
1792         int ret;
1793
1794         ctx = nfs4_state_find_open_context(state);
1795         if (IS_ERR(ctx))
1796                 return -EAGAIN;
1797         ret = nfs4_do_open_reclaim(ctx, state);
1798         put_nfs_open_context(ctx);
1799         return ret;
1800 }
1801
1802 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1803 {
1804         switch (err) {
1805                 default:
1806                         printk(KERN_ERR "NFS: %s: unhandled error "
1807                                         "%d.\n", __func__, err);
1808                 case 0:
1809                 case -ENOENT:
1810                 case -EAGAIN:
1811                 case -ESTALE:
1812                         break;
1813                 case -NFS4ERR_BADSESSION:
1814                 case -NFS4ERR_BADSLOT:
1815                 case -NFS4ERR_BAD_HIGH_SLOT:
1816                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1817                 case -NFS4ERR_DEADSESSION:
1818                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1819                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1820                         return -EAGAIN;
1821                 case -NFS4ERR_STALE_CLIENTID:
1822                 case -NFS4ERR_STALE_STATEID:
1823                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1824                 case -NFS4ERR_EXPIRED:
1825                         /* Don't recall a delegation if it was lost */
1826                         nfs4_schedule_lease_recovery(server->nfs_client);
1827                         return -EAGAIN;
1828                 case -NFS4ERR_MOVED:
1829                         nfs4_schedule_migration_recovery(server);
1830                         return -EAGAIN;
1831                 case -NFS4ERR_LEASE_MOVED:
1832                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1833                         return -EAGAIN;
1834                 case -NFS4ERR_DELEG_REVOKED:
1835                 case -NFS4ERR_ADMIN_REVOKED:
1836                 case -NFS4ERR_BAD_STATEID:
1837                 case -NFS4ERR_OPENMODE:
1838                         nfs_inode_find_state_and_recover(state->inode,
1839                                         stateid);
1840                         nfs4_schedule_stateid_recovery(server, state);
1841                         return -EAGAIN;
1842                 case -NFS4ERR_DELAY:
1843                 case -NFS4ERR_GRACE:
1844                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1845                         ssleep(1);
1846                         return -EAGAIN;
1847                 case -ENOMEM:
1848                 case -NFS4ERR_DENIED:
1849                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1850                         return 0;
1851         }
1852         return err;
1853 }
1854
1855 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1856                 struct nfs4_state *state, const nfs4_stateid *stateid,
1857                 fmode_t type)
1858 {
1859         struct nfs_server *server = NFS_SERVER(state->inode);
1860         struct nfs4_opendata *opendata;
1861         int err = 0;
1862
1863         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1864                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1865         if (IS_ERR(opendata))
1866                 return PTR_ERR(opendata);
1867         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1868         write_seqlock(&state->seqlock);
1869         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1870         write_sequnlock(&state->seqlock);
1871         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1872         switch (type & (FMODE_READ|FMODE_WRITE)) {
1873         case FMODE_READ|FMODE_WRITE:
1874         case FMODE_WRITE:
1875                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1876                 if (err)
1877                         break;
1878                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1879                 if (err)
1880                         break;
1881         case FMODE_READ:
1882                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1883         }
1884         nfs4_opendata_put(opendata);
1885         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1886 }
1887
1888 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1889 {
1890         struct nfs4_opendata *data = calldata;
1891
1892         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1893                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1894 }
1895
1896 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1897 {
1898         struct nfs4_opendata *data = calldata;
1899
1900         nfs40_sequence_done(task, &data->c_res.seq_res);
1901
1902         data->rpc_status = task->tk_status;
1903         if (data->rpc_status == 0) {
1904                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1905                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1906                 renew_lease(data->o_res.server, data->timestamp);
1907                 data->rpc_done = 1;
1908         }
1909 }
1910
1911 static void nfs4_open_confirm_release(void *calldata)
1912 {
1913         struct nfs4_opendata *data = calldata;
1914         struct nfs4_state *state = NULL;
1915
1916         /* If this request hasn't been cancelled, do nothing */
1917         if (data->cancelled == 0)
1918                 goto out_free;
1919         /* In case of error, no cleanup! */
1920         if (!data->rpc_done)
1921                 goto out_free;
1922         state = nfs4_opendata_to_nfs4_state(data);
1923         if (!IS_ERR(state))
1924                 nfs4_close_state(state, data->o_arg.fmode);
1925 out_free:
1926         nfs4_opendata_put(data);
1927 }
1928
1929 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1930         .rpc_call_prepare = nfs4_open_confirm_prepare,
1931         .rpc_call_done = nfs4_open_confirm_done,
1932         .rpc_release = nfs4_open_confirm_release,
1933 };
1934
1935 /*
1936  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1937  */
1938 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1939 {
1940         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1941         struct rpc_task *task;
1942         struct  rpc_message msg = {
1943                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1944                 .rpc_argp = &data->c_arg,
1945                 .rpc_resp = &data->c_res,
1946                 .rpc_cred = data->owner->so_cred,
1947         };
1948         struct rpc_task_setup task_setup_data = {
1949                 .rpc_client = server->client,
1950                 .rpc_message = &msg,
1951                 .callback_ops = &nfs4_open_confirm_ops,
1952                 .callback_data = data,
1953                 .workqueue = nfsiod_workqueue,
1954                 .flags = RPC_TASK_ASYNC,
1955         };
1956         int status;
1957
1958         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1959         kref_get(&data->kref);
1960         data->rpc_done = 0;
1961         data->rpc_status = 0;
1962         data->timestamp = jiffies;
1963         if (data->is_recover)
1964                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1965         task = rpc_run_task(&task_setup_data);
1966         if (IS_ERR(task))
1967                 return PTR_ERR(task);
1968         status = nfs4_wait_for_completion_rpc_task(task);
1969         if (status != 0) {
1970                 data->cancelled = 1;
1971                 smp_wmb();
1972         } else
1973                 status = data->rpc_status;
1974         rpc_put_task(task);
1975         return status;
1976 }
1977
1978 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1979 {
1980         struct nfs4_opendata *data = calldata;
1981         struct nfs4_state_owner *sp = data->owner;
1982         struct nfs_client *clp = sp->so_server->nfs_client;
1983         enum open_claim_type4 claim = data->o_arg.claim;
1984
1985         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1986                 goto out_wait;
1987         /*
1988          * Check if we still need to send an OPEN call, or if we can use
1989          * a delegation instead.
1990          */
1991         if (data->state != NULL) {
1992                 struct nfs_delegation *delegation;
1993
1994                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1995                         goto out_no_action;
1996                 rcu_read_lock();
1997                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1998                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1999                         goto unlock_no_action;
2000                 rcu_read_unlock();
2001         }
2002         /* Update client id. */
2003         data->o_arg.clientid = clp->cl_clientid;
2004         switch (claim) {
2005         default:
2006                 break;
2007         case NFS4_OPEN_CLAIM_PREVIOUS:
2008         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2009         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2010                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2011         case NFS4_OPEN_CLAIM_FH:
2012                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2013                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2014         }
2015         data->timestamp = jiffies;
2016         if (nfs4_setup_sequence(data->o_arg.server,
2017                                 &data->o_arg.seq_args,
2018                                 &data->o_res.seq_res,
2019                                 task) != 0)
2020                 nfs_release_seqid(data->o_arg.seqid);
2021
2022         /* Set the create mode (note dependency on the session type) */
2023         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2024         if (data->o_arg.open_flags & O_EXCL) {
2025                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2026                 if (nfs4_has_persistent_session(clp))
2027                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2028                 else if (clp->cl_mvops->minor_version > 0)
2029                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2030         }
2031         return;
2032 unlock_no_action:
2033         trace_nfs4_cached_open(data->state);
2034         rcu_read_unlock();
2035 out_no_action:
2036         task->tk_action = NULL;
2037 out_wait:
2038         nfs4_sequence_done(task, &data->o_res.seq_res);
2039 }
2040
2041 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2042 {
2043         struct nfs4_opendata *data = calldata;
2044
2045         data->rpc_status = task->tk_status;
2046
2047         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2048                 return;
2049
2050         if (task->tk_status == 0) {
2051                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2052                         switch (data->o_res.f_attr->mode & S_IFMT) {
2053                         case S_IFREG:
2054                                 break;
2055                         case S_IFLNK:
2056                                 data->rpc_status = -ELOOP;
2057                                 break;
2058                         case S_IFDIR:
2059                                 data->rpc_status = -EISDIR;
2060                                 break;
2061                         default:
2062                                 data->rpc_status = -ENOTDIR;
2063                         }
2064                 }
2065                 renew_lease(data->o_res.server, data->timestamp);
2066                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2067                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2068         }
2069         data->rpc_done = 1;
2070 }
2071
2072 static void nfs4_open_release(void *calldata)
2073 {
2074         struct nfs4_opendata *data = calldata;
2075         struct nfs4_state *state = NULL;
2076
2077         /* If this request hasn't been cancelled, do nothing */
2078         if (data->cancelled == 0)
2079                 goto out_free;
2080         /* In case of error, no cleanup! */
2081         if (data->rpc_status != 0 || !data->rpc_done)
2082                 goto out_free;
2083         /* In case we need an open_confirm, no cleanup! */
2084         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2085                 goto out_free;
2086         state = nfs4_opendata_to_nfs4_state(data);
2087         if (!IS_ERR(state))
2088                 nfs4_close_state(state, data->o_arg.fmode);
2089 out_free:
2090         nfs4_opendata_put(data);
2091 }
2092
2093 static const struct rpc_call_ops nfs4_open_ops = {
2094         .rpc_call_prepare = nfs4_open_prepare,
2095         .rpc_call_done = nfs4_open_done,
2096         .rpc_release = nfs4_open_release,
2097 };
2098
2099 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2100 {
2101         struct inode *dir = d_inode(data->dir);
2102         struct nfs_server *server = NFS_SERVER(dir);
2103         struct nfs_openargs *o_arg = &data->o_arg;
2104         struct nfs_openres *o_res = &data->o_res;
2105         struct rpc_task *task;
2106         struct rpc_message msg = {
2107                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2108                 .rpc_argp = o_arg,
2109                 .rpc_resp = o_res,
2110                 .rpc_cred = data->owner->so_cred,
2111         };
2112         struct rpc_task_setup task_setup_data = {
2113                 .rpc_client = server->client,
2114                 .rpc_message = &msg,
2115                 .callback_ops = &nfs4_open_ops,
2116                 .callback_data = data,
2117                 .workqueue = nfsiod_workqueue,
2118                 .flags = RPC_TASK_ASYNC,
2119         };
2120         int status;
2121
2122         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2123         kref_get(&data->kref);
2124         data->rpc_done = 0;
2125         data->rpc_status = 0;
2126         data->cancelled = 0;
2127         data->is_recover = 0;
2128         if (isrecover) {
2129                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2130                 data->is_recover = 1;
2131         }
2132         task = rpc_run_task(&task_setup_data);
2133         if (IS_ERR(task))
2134                 return PTR_ERR(task);
2135         status = nfs4_wait_for_completion_rpc_task(task);
2136         if (status != 0) {
2137                 data->cancelled = 1;
2138                 smp_wmb();
2139         } else
2140                 status = data->rpc_status;
2141         rpc_put_task(task);
2142
2143         return status;
2144 }
2145
2146 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2147 {
2148         struct inode *dir = d_inode(data->dir);
2149         struct nfs_openres *o_res = &data->o_res;
2150         int status;
2151
2152         status = nfs4_run_open_task(data, 1);
2153         if (status != 0 || !data->rpc_done)
2154                 return status;
2155
2156         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2157
2158         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2159                 status = _nfs4_proc_open_confirm(data);
2160                 if (status != 0)
2161                         return status;
2162         }
2163
2164         return status;
2165 }
2166
2167 /*
2168  * Additional permission checks in order to distinguish between an
2169  * open for read, and an open for execute. This works around the
2170  * fact that NFSv4 OPEN treats read and execute permissions as being
2171  * the same.
2172  * Note that in the non-execute case, we want to turn off permission
2173  * checking if we just created a new file (POSIX open() semantics).
2174  */
2175 static int nfs4_opendata_access(struct rpc_cred *cred,
2176                                 struct nfs4_opendata *opendata,
2177                                 struct nfs4_state *state, fmode_t fmode,
2178                                 int openflags)
2179 {
2180         struct nfs_access_entry cache;
2181         u32 mask;
2182
2183         /* access call failed or for some reason the server doesn't
2184          * support any access modes -- defer access call until later */
2185         if (opendata->o_res.access_supported == 0)
2186                 return 0;
2187
2188         mask = 0;
2189         /*
2190          * Use openflags to check for exec, because fmode won't
2191          * always have FMODE_EXEC set when file open for exec.
2192          */
2193         if (openflags & __FMODE_EXEC) {
2194                 /* ONLY check for exec rights */
2195                 mask = MAY_EXEC;
2196         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2197                 mask = MAY_READ;
2198
2199         cache.cred = cred;
2200         cache.jiffies = jiffies;
2201         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2202         nfs_access_add_cache(state->inode, &cache);
2203
2204         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2205                 return 0;
2206
2207         /* even though OPEN succeeded, access is denied. Close the file */
2208         nfs4_close_state(state, fmode);
2209         return -EACCES;
2210 }
2211
2212 /*
2213  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2214  */
2215 static int _nfs4_proc_open(struct nfs4_opendata *data)
2216 {
2217         struct inode *dir = d_inode(data->dir);
2218         struct nfs_server *server = NFS_SERVER(dir);
2219         struct nfs_openargs *o_arg = &data->o_arg;
2220         struct nfs_openres *o_res = &data->o_res;
2221         int status;
2222
2223         status = nfs4_run_open_task(data, 0);
2224         if (!data->rpc_done)
2225                 return status;
2226         if (status != 0) {
2227                 if (status == -NFS4ERR_BADNAME &&
2228                                 !(o_arg->open_flags & O_CREAT))
2229                         return -ENOENT;
2230                 return status;
2231         }
2232
2233         nfs_fattr_map_and_free_names(server, &data->f_attr);
2234
2235         if (o_arg->open_flags & O_CREAT) {
2236                 update_changeattr(dir, &o_res->cinfo);
2237                 if (o_arg->open_flags & O_EXCL)
2238                         data->file_created = 1;
2239                 else if (o_res->cinfo.before != o_res->cinfo.after)
2240                         data->file_created = 1;
2241         }
2242         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2243                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2244         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2245                 status = _nfs4_proc_open_confirm(data);
2246                 if (status != 0)
2247                         return status;
2248         }
2249         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2250                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2251         return 0;
2252 }
2253
2254 static int nfs4_recover_expired_lease(struct nfs_server *server)
2255 {
2256         return nfs4_client_recover_expired_lease(server->nfs_client);
2257 }
2258
2259 /*
2260  * OPEN_EXPIRED:
2261  *      reclaim state on the server after a network partition.
2262  *      Assumes caller holds the appropriate lock
2263  */
2264 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2265 {
2266         struct nfs4_opendata *opendata;
2267         int ret;
2268
2269         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2270                         NFS4_OPEN_CLAIM_FH);
2271         if (IS_ERR(opendata))
2272                 return PTR_ERR(opendata);
2273         ret = nfs4_open_recover(opendata, state);
2274         if (ret == -ESTALE)
2275                 d_drop(ctx->dentry);
2276         nfs4_opendata_put(opendata);
2277         return ret;
2278 }
2279
2280 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2281 {
2282         struct nfs_server *server = NFS_SERVER(state->inode);
2283         struct nfs4_exception exception = { };
2284         int err;
2285
2286         do {
2287                 err = _nfs4_open_expired(ctx, state);
2288                 trace_nfs4_open_expired(ctx, 0, err);
2289                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2290                         continue;
2291                 switch (err) {
2292                 default:
2293                         goto out;
2294                 case -NFS4ERR_GRACE:
2295                 case -NFS4ERR_DELAY:
2296                         nfs4_handle_exception(server, err, &exception);
2297                         err = 0;
2298                 }
2299         } while (exception.retry);
2300 out:
2301         return err;
2302 }
2303
2304 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2305 {
2306         struct nfs_open_context *ctx;
2307         int ret;
2308
2309         ctx = nfs4_state_find_open_context(state);
2310         if (IS_ERR(ctx))
2311                 return -EAGAIN;
2312         ret = nfs4_do_open_expired(ctx, state);
2313         put_nfs_open_context(ctx);
2314         return ret;
2315 }
2316
2317 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2318 {
2319         nfs_remove_bad_delegation(state->inode);
2320         write_seqlock(&state->seqlock);
2321         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2322         write_sequnlock(&state->seqlock);
2323         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2324 }
2325
2326 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2327 {
2328         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2329                 nfs_finish_clear_delegation_stateid(state);
2330 }
2331
2332 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2333 {
2334         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2335         nfs40_clear_delegation_stateid(state);
2336         return nfs4_open_expired(sp, state);
2337 }
2338
2339 #if defined(CONFIG_NFS_V4_1)
2340 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2341 {
2342         struct nfs_server *server = NFS_SERVER(state->inode);
2343         nfs4_stateid stateid;
2344         struct nfs_delegation *delegation;
2345         struct rpc_cred *cred;
2346         int status;
2347
2348         /* Get the delegation credential for use by test/free_stateid */
2349         rcu_read_lock();
2350         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2351         if (delegation == NULL) {
2352                 rcu_read_unlock();
2353                 return;
2354         }
2355
2356         nfs4_stateid_copy(&stateid, &delegation->stateid);
2357         cred = get_rpccred(delegation->cred);
2358         rcu_read_unlock();
2359         status = nfs41_test_stateid(server, &stateid, cred);
2360         trace_nfs4_test_delegation_stateid(state, NULL, status);
2361
2362         if (status != NFS_OK) {
2363                 /* Free the stateid unless the server explicitly
2364                  * informs us the stateid is unrecognized. */
2365                 if (status != -NFS4ERR_BAD_STATEID)
2366                         nfs41_free_stateid(server, &stateid, cred);
2367                 nfs_finish_clear_delegation_stateid(state);
2368         }
2369
2370         put_rpccred(cred);
2371 }
2372
2373 /**
2374  * nfs41_check_open_stateid - possibly free an open stateid
2375  *
2376  * @state: NFSv4 state for an inode
2377  *
2378  * Returns NFS_OK if recovery for this stateid is now finished.
2379  * Otherwise a negative NFS4ERR value is returned.
2380  */
2381 static int nfs41_check_open_stateid(struct nfs4_state *state)
2382 {
2383         struct nfs_server *server = NFS_SERVER(state->inode);
2384         nfs4_stateid *stateid = &state->open_stateid;
2385         struct rpc_cred *cred = state->owner->so_cred;
2386         int status;
2387
2388         /* If a state reset has been done, test_stateid is unneeded */
2389         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2390             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2391             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2392                 return -NFS4ERR_BAD_STATEID;
2393
2394         status = nfs41_test_stateid(server, stateid, cred);
2395         trace_nfs4_test_open_stateid(state, NULL, status);
2396         if (status != NFS_OK) {
2397                 /* Free the stateid unless the server explicitly
2398                  * informs us the stateid is unrecognized. */
2399                 if (status != -NFS4ERR_BAD_STATEID)
2400                         nfs41_free_stateid(server, stateid, cred);
2401
2402                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2403                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2404                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2405                 clear_bit(NFS_OPEN_STATE, &state->flags);
2406         }
2407         return status;
2408 }
2409
2410 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2411 {
2412         int status;
2413
2414         nfs41_check_delegation_stateid(state);
2415         status = nfs41_check_open_stateid(state);
2416         if (status != NFS_OK)
2417                 status = nfs4_open_expired(sp, state);
2418         return status;
2419 }
2420 #endif
2421
2422 /*
2423  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2424  * fields corresponding to attributes that were used to store the verifier.
2425  * Make sure we clobber those fields in the later setattr call
2426  */
2427 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2428                                 struct iattr *sattr, struct nfs4_label **label)
2429 {
2430         const u32 *attrset = opendata->o_res.attrset;
2431
2432         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2433             !(sattr->ia_valid & ATTR_ATIME_SET))
2434                 sattr->ia_valid |= ATTR_ATIME;
2435
2436         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2437             !(sattr->ia_valid & ATTR_MTIME_SET))
2438                 sattr->ia_valid |= ATTR_MTIME;
2439
2440         /* Except MODE, it seems harmless of setting twice. */
2441         if ((attrset[1] & FATTR4_WORD1_MODE))
2442                 sattr->ia_valid &= ~ATTR_MODE;
2443
2444         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2445                 *label = NULL;
2446 }
2447
2448 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2449                 fmode_t fmode,
2450                 int flags,
2451                 struct nfs_open_context *ctx)
2452 {
2453         struct nfs4_state_owner *sp = opendata->owner;
2454         struct nfs_server *server = sp->so_server;
2455         struct dentry *dentry;
2456         struct nfs4_state *state;
2457         unsigned int seq;
2458         int ret;
2459
2460         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2461
2462         ret = _nfs4_proc_open(opendata);
2463         if (ret != 0)
2464                 goto out;
2465
2466         state = nfs4_opendata_to_nfs4_state(opendata);
2467         ret = PTR_ERR(state);
2468         if (IS_ERR(state))
2469                 goto out;
2470         if (server->caps & NFS_CAP_POSIX_LOCK)
2471                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2472
2473         dentry = opendata->dentry;
2474         if (d_really_is_negative(dentry)) {
2475                 struct dentry *alias;
2476                 d_drop(dentry);
2477                 alias = d_exact_alias(dentry, state->inode);
2478                 if (!alias)
2479                         alias = d_splice_alias(igrab(state->inode), dentry);
2480                 /* d_splice_alias() can't fail here - it's a non-directory */
2481                 if (alias) {
2482                         dput(ctx->dentry);
2483                         ctx->dentry = dentry = alias;
2484                 }
2485                 nfs_set_verifier(dentry,
2486                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2487         }
2488
2489         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2490         if (ret != 0)
2491                 goto out;
2492
2493         ctx->state = state;
2494         if (d_inode(dentry) == state->inode) {
2495                 nfs_inode_attach_open_context(ctx);
2496                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2497                         nfs4_schedule_stateid_recovery(server, state);
2498         }
2499 out:
2500         return ret;
2501 }
2502
2503 /*
2504  * Returns a referenced nfs4_state
2505  */
2506 static int _nfs4_do_open(struct inode *dir,
2507                         struct nfs_open_context *ctx,
2508                         int flags,
2509                         struct iattr *sattr,
2510                         struct nfs4_label *label,
2511                         int *opened)
2512 {
2513         struct nfs4_state_owner  *sp;
2514         struct nfs4_state     *state = NULL;
2515         struct nfs_server       *server = NFS_SERVER(dir);
2516         struct nfs4_opendata *opendata;
2517         struct dentry *dentry = ctx->dentry;
2518         struct rpc_cred *cred = ctx->cred;
2519         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2520         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2521         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2522         struct nfs4_label *olabel = NULL;
2523         int status;
2524
2525         /* Protect against reboot recovery conflicts */
2526         status = -ENOMEM;
2527         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2528         if (sp == NULL) {
2529                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2530                 goto out_err;
2531         }
2532         status = nfs4_recover_expired_lease(server);
2533         if (status != 0)
2534                 goto err_put_state_owner;
2535         if (d_really_is_positive(dentry))
2536                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2537         status = -ENOMEM;
2538         if (d_really_is_positive(dentry))
2539                 claim = NFS4_OPEN_CLAIM_FH;
2540         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2541                         label, claim, GFP_KERNEL);
2542         if (opendata == NULL)
2543                 goto err_put_state_owner;
2544
2545         if (label) {
2546                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2547                 if (IS_ERR(olabel)) {
2548                         status = PTR_ERR(olabel);
2549                         goto err_opendata_put;
2550                 }
2551         }
2552
2553         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2554                 if (!opendata->f_attr.mdsthreshold) {
2555                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2556                         if (!opendata->f_attr.mdsthreshold)
2557                                 goto err_free_label;
2558                 }
2559                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2560         }
2561         if (d_really_is_positive(dentry))
2562                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2563
2564         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2565         if (status != 0)
2566                 goto err_free_label;
2567         state = ctx->state;
2568
2569         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2570             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2571                 nfs4_exclusive_attrset(opendata, sattr, &label);
2572                 /*
2573                  * send create attributes which was not set by open
2574                  * with an extra setattr.
2575                  */
2576                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2577                         nfs_fattr_init(opendata->o_res.f_attr);
2578                         status = nfs4_do_setattr(state->inode, cred,
2579                                         opendata->o_res.f_attr, sattr,
2580                                         state, label, olabel);
2581                         if (status == 0) {
2582                                 nfs_setattr_update_inode(state->inode, sattr,
2583                                                 opendata->o_res.f_attr);
2584                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2585                         }
2586                 }
2587         }
2588         if (opened && opendata->file_created)
2589                 *opened |= FILE_CREATED;
2590
2591         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2592                 *ctx_th = opendata->f_attr.mdsthreshold;
2593                 opendata->f_attr.mdsthreshold = NULL;
2594         }
2595
2596         nfs4_label_free(olabel);
2597
2598         nfs4_opendata_put(opendata);
2599         nfs4_put_state_owner(sp);
2600         return 0;
2601 err_free_label:
2602         nfs4_label_free(olabel);
2603 err_opendata_put:
2604         nfs4_opendata_put(opendata);
2605 err_put_state_owner:
2606         nfs4_put_state_owner(sp);
2607 out_err:
2608         return status;
2609 }
2610
2611
2612 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2613                                         struct nfs_open_context *ctx,
2614                                         int flags,
2615                                         struct iattr *sattr,
2616                                         struct nfs4_label *label,
2617                                         int *opened)
2618 {
2619         struct nfs_server *server = NFS_SERVER(dir);
2620         struct nfs4_exception exception = { };
2621         struct nfs4_state *res;
2622         int status;
2623
2624         do {
2625                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2626                 res = ctx->state;
2627                 trace_nfs4_open_file(ctx, flags, status);
2628                 if (status == 0)
2629                         break;
2630                 /* NOTE: BAD_SEQID means the server and client disagree about the
2631                  * book-keeping w.r.t. state-changing operations
2632                  * (OPEN/CLOSE/LOCK/LOCKU...)
2633                  * It is actually a sign of a bug on the client or on the server.
2634                  *
2635                  * If we receive a BAD_SEQID error in the particular case of
2636                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2637                  * have unhashed the old state_owner for us, and that we can
2638                  * therefore safely retry using a new one. We should still warn
2639                  * the user though...
2640                  */
2641                 if (status == -NFS4ERR_BAD_SEQID) {
2642                         pr_warn_ratelimited("NFS: v4 server %s "
2643                                         " returned a bad sequence-id error!\n",
2644                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2645                         exception.retry = 1;
2646                         continue;
2647                 }
2648                 /*
2649                  * BAD_STATEID on OPEN means that the server cancelled our
2650                  * state before it received the OPEN_CONFIRM.
2651                  * Recover by retrying the request as per the discussion
2652                  * on Page 181 of RFC3530.
2653                  */
2654                 if (status == -NFS4ERR_BAD_STATEID) {
2655                         exception.retry = 1;
2656                         continue;
2657                 }
2658                 if (status == -EAGAIN) {
2659                         /* We must have found a delegation */
2660                         exception.retry = 1;
2661                         continue;
2662                 }
2663                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2664                         continue;
2665                 res = ERR_PTR(nfs4_handle_exception(server,
2666                                         status, &exception));
2667         } while (exception.retry);
2668         return res;
2669 }
2670
2671 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2672                             struct nfs_fattr *fattr, struct iattr *sattr,
2673                             struct nfs4_state *state, struct nfs4_label *ilabel,
2674                             struct nfs4_label *olabel)
2675 {
2676         struct nfs_server *server = NFS_SERVER(inode);
2677         struct nfs_setattrargs  arg = {
2678                 .fh             = NFS_FH(inode),
2679                 .iap            = sattr,
2680                 .server         = server,
2681                 .bitmask = server->attr_bitmask,
2682                 .label          = ilabel,
2683         };
2684         struct nfs_setattrres  res = {
2685                 .fattr          = fattr,
2686                 .label          = olabel,
2687                 .server         = server,
2688         };
2689         struct rpc_message msg = {
2690                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2691                 .rpc_argp       = &arg,
2692                 .rpc_resp       = &res,
2693                 .rpc_cred       = cred,
2694         };
2695         struct rpc_cred *delegation_cred = NULL;
2696         unsigned long timestamp = jiffies;
2697         fmode_t fmode;
2698         bool truncate;
2699         int status;
2700
2701         arg.bitmask = nfs4_bitmask(server, ilabel);
2702         if (ilabel)
2703                 arg.bitmask = nfs4_bitmask(server, olabel);
2704
2705         nfs_fattr_init(fattr);
2706
2707         /* Servers should only apply open mode checks for file size changes */
2708         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2709         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2710
2711         if (nfs4_copy_delegation_stateid(inode, fmode, &arg.stateid, &delegation_cred)) {
2712                 /* Use that stateid */
2713         } else if (truncate && state != NULL) {
2714                 struct nfs_lockowner lockowner = {
2715                         .l_owner = current->files,
2716                         .l_pid = current->tgid,
2717                 };
2718                 if (!nfs4_valid_open_stateid(state))
2719                         return -EBADF;
2720                 if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner,
2721                                 &arg.stateid, &delegation_cred) == -EIO)
2722                         return -EBADF;
2723         } else
2724                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2725         if (delegation_cred)
2726                 msg.rpc_cred = delegation_cred;
2727
2728         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2729
2730         put_rpccred(delegation_cred);
2731         if (status == 0 && state != NULL)
2732                 renew_lease(server, timestamp);
2733         trace_nfs4_setattr(inode, &arg.stateid, status);
2734         return status;
2735 }
2736
2737 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2738                            struct nfs_fattr *fattr, struct iattr *sattr,
2739                            struct nfs4_state *state, struct nfs4_label *ilabel,
2740                            struct nfs4_label *olabel)
2741 {
2742         struct nfs_server *server = NFS_SERVER(inode);
2743         struct nfs4_exception exception = {
2744                 .state = state,
2745                 .inode = inode,
2746         };
2747         int err;
2748         do {
2749                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2750                 switch (err) {
2751                 case -NFS4ERR_OPENMODE:
2752                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2753                                 pr_warn_once("NFSv4: server %s is incorrectly "
2754                                                 "applying open mode checks to "
2755                                                 "a SETATTR that is not "
2756                                                 "changing file size.\n",
2757                                                 server->nfs_client->cl_hostname);
2758                         }
2759                         if (state && !(state->state & FMODE_WRITE)) {
2760                                 err = -EBADF;
2761                                 if (sattr->ia_valid & ATTR_OPEN)
2762                                         err = -EACCES;
2763                                 goto out;
2764                         }
2765                 }
2766                 err = nfs4_handle_exception(server, err, &exception);
2767         } while (exception.retry);
2768 out:
2769         return err;
2770 }
2771
2772 static bool
2773 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2774 {
2775         if (inode == NULL || !nfs_have_layout(inode))
2776                 return false;
2777
2778         return pnfs_wait_on_layoutreturn(inode, task);
2779 }
2780
2781 struct nfs4_closedata {
2782         struct inode *inode;
2783         struct nfs4_state *state;
2784         struct nfs_closeargs arg;
2785         struct nfs_closeres res;
2786         struct nfs_fattr fattr;
2787         unsigned long timestamp;
2788         bool roc;
2789         u32 roc_barrier;
2790 };
2791
2792 static void nfs4_free_closedata(void *data)
2793 {
2794         struct nfs4_closedata *calldata = data;
2795         struct nfs4_state_owner *sp = calldata->state->owner;
2796         struct super_block *sb = calldata->state->inode->i_sb;
2797
2798         if (calldata->roc)
2799                 pnfs_roc_release(calldata->state->inode);
2800         nfs4_put_open_state(calldata->state);
2801         nfs_free_seqid(calldata->arg.seqid);
2802         nfs4_put_state_owner(sp);
2803         nfs_sb_deactive(sb);
2804         kfree(calldata);
2805 }
2806
2807 static void nfs4_close_done(struct rpc_task *task, void *data)
2808 {
2809         struct nfs4_closedata *calldata = data;
2810         struct nfs4_state *state = calldata->state;
2811         struct nfs_server *server = NFS_SERVER(calldata->inode);
2812         nfs4_stateid *res_stateid = NULL;
2813
2814         dprintk("%s: begin!\n", __func__);
2815         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2816                 return;
2817         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2818         /* hmm. we are done with the inode, and in the process of freeing
2819          * the state_owner. we keep this around to process errors
2820          */
2821         switch (task->tk_status) {
2822                 case 0:
2823                         res_stateid = &calldata->res.stateid;
2824                         if (calldata->roc)
2825                                 pnfs_roc_set_barrier(state->inode,
2826                                                      calldata->roc_barrier);
2827                         renew_lease(server, calldata->timestamp);
2828                         break;
2829                 case -NFS4ERR_ADMIN_REVOKED:
2830                 case -NFS4ERR_STALE_STATEID:
2831                 case -NFS4ERR_OLD_STATEID:
2832                 case -NFS4ERR_BAD_STATEID:
2833                 case -NFS4ERR_EXPIRED:
2834                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2835                                                 &state->open_stateid)) {
2836                                 rpc_restart_call_prepare(task);
2837                                 goto out_release;
2838                         }
2839                         if (calldata->arg.fmode == 0)
2840                                 break;
2841                 default:
2842                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2843                                 rpc_restart_call_prepare(task);
2844                                 goto out_release;
2845                         }
2846         }
2847         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2848                         res_stateid, calldata->arg.fmode);
2849 out_release:
2850         nfs_release_seqid(calldata->arg.seqid);
2851         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2852         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2853 }
2854
2855 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2856 {
2857         struct nfs4_closedata *calldata = data;
2858         struct nfs4_state *state = calldata->state;
2859         struct inode *inode = calldata->inode;
2860         bool is_rdonly, is_wronly, is_rdwr;
2861         int call_close = 0;
2862
2863         dprintk("%s: begin!\n", __func__);
2864         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2865                 goto out_wait;
2866
2867         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2868         spin_lock(&state->owner->so_lock);
2869         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2870         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2871         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2872         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2873         /* Calculate the change in open mode */
2874         calldata->arg.fmode = 0;
2875         if (state->n_rdwr == 0) {
2876                 if (state->n_rdonly == 0)
2877                         call_close |= is_rdonly;
2878                 else if (is_rdonly)
2879                         calldata->arg.fmode |= FMODE_READ;
2880                 if (state->n_wronly == 0)
2881                         call_close |= is_wronly;
2882                 else if (is_wronly)
2883                         calldata->arg.fmode |= FMODE_WRITE;
2884         } else if (is_rdwr)
2885                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2886
2887         if (calldata->arg.fmode == 0)
2888                 call_close |= is_rdwr;
2889
2890         if (!nfs4_valid_open_stateid(state))
2891                 call_close = 0;
2892         spin_unlock(&state->owner->so_lock);
2893
2894         if (!call_close) {
2895                 /* Note: exit _without_ calling nfs4_close_done */
2896                 goto out_no_action;
2897         }
2898
2899         if (nfs4_wait_on_layoutreturn(inode, task)) {
2900                 nfs_release_seqid(calldata->arg.seqid);
2901                 goto out_wait;
2902         }
2903
2904         if (calldata->arg.fmode == 0)
2905                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2906         if (calldata->roc)
2907                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2908
2909         calldata->arg.share_access =
2910                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2911                                 calldata->arg.fmode, 0);
2912
2913         nfs_fattr_init(calldata->res.fattr);
2914         calldata->timestamp = jiffies;
2915         if (nfs4_setup_sequence(NFS_SERVER(inode),
2916                                 &calldata->arg.seq_args,
2917                                 &calldata->res.seq_res,
2918                                 task) != 0)
2919                 nfs_release_seqid(calldata->arg.seqid);
2920         dprintk("%s: done!\n", __func__);
2921         return;
2922 out_no_action:
2923         task->tk_action = NULL;
2924 out_wait:
2925         nfs4_sequence_done(task, &calldata->res.seq_res);
2926 }
2927
2928 static const struct rpc_call_ops nfs4_close_ops = {
2929         .rpc_call_prepare = nfs4_close_prepare,
2930         .rpc_call_done = nfs4_close_done,
2931         .rpc_release = nfs4_free_closedata,
2932 };
2933
2934 static bool nfs4_roc(struct inode *inode)
2935 {
2936         if (!nfs_have_layout(inode))
2937                 return false;
2938         return pnfs_roc(inode);
2939 }
2940
2941 /* 
2942  * It is possible for data to be read/written from a mem-mapped file 
2943  * after the sys_close call (which hits the vfs layer as a flush).
2944  * This means that we can't safely call nfsv4 close on a file until 
2945  * the inode is cleared. This in turn means that we are not good
2946  * NFSv4 citizens - we do not indicate to the server to update the file's 
2947  * share state even when we are done with one of the three share 
2948  * stateid's in the inode.
2949  *
2950  * NOTE: Caller must be holding the sp->so_owner semaphore!
2951  */
2952 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2953 {
2954         struct nfs_server *server = NFS_SERVER(state->inode);
2955         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2956         struct nfs4_closedata *calldata;
2957         struct nfs4_state_owner *sp = state->owner;
2958         struct rpc_task *task;
2959         struct rpc_message msg = {
2960                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2961                 .rpc_cred = state->owner->so_cred,
2962         };
2963         struct rpc_task_setup task_setup_data = {
2964                 .rpc_client = server->client,
2965                 .rpc_message = &msg,
2966                 .callback_ops = &nfs4_close_ops,
2967                 .workqueue = nfsiod_workqueue,
2968                 .flags = RPC_TASK_ASYNC,
2969         };
2970         int status = -ENOMEM;
2971
2972         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2973                 &task_setup_data.rpc_client, &msg);
2974
2975         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2976         if (calldata == NULL)
2977                 goto out;
2978         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2979         calldata->inode = state->inode;
2980         calldata->state = state;
2981         calldata->arg.fh = NFS_FH(state->inode);
2982         /* Serialization for the sequence id */
2983         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2984         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2985         if (IS_ERR(calldata->arg.seqid))
2986                 goto out_free_calldata;
2987         calldata->arg.fmode = 0;
2988         calldata->arg.bitmask = server->cache_consistency_bitmask;
2989         calldata->res.fattr = &calldata->fattr;
2990         calldata->res.seqid = calldata->arg.seqid;
2991         calldata->res.server = server;
2992         calldata->roc = nfs4_roc(state->inode);
2993         nfs_sb_active(calldata->inode->i_sb);
2994
2995         msg.rpc_argp = &calldata->arg;
2996         msg.rpc_resp = &calldata->res;
2997         task_setup_data.callback_data = calldata;
2998         task = rpc_run_task(&task_setup_data);
2999         if (IS_ERR(task))
3000                 return PTR_ERR(task);
3001         status = 0;
3002         if (wait)
3003                 status = rpc_wait_for_completion_task(task);
3004         rpc_put_task(task);
3005         return status;
3006 out_free_calldata:
3007         kfree(calldata);
3008 out:
3009         nfs4_put_open_state(state);
3010         nfs4_put_state_owner(sp);
3011         return status;
3012 }
3013
3014 static struct inode *
3015 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3016                 int open_flags, struct iattr *attr, int *opened)
3017 {
3018         struct nfs4_state *state;
3019         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3020
3021         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3022
3023         /* Protect against concurrent sillydeletes */
3024         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3025
3026         nfs4_label_release_security(label);
3027
3028         if (IS_ERR(state))
3029                 return ERR_CAST(state);
3030         return state->inode;
3031 }
3032
3033 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3034 {
3035         if (ctx->state == NULL)
3036                 return;
3037         if (is_sync)
3038                 nfs4_close_sync(ctx->state, ctx->mode);
3039         else
3040                 nfs4_close_state(ctx->state, ctx->mode);
3041 }
3042
3043 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3044 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3045 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3046
3047 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3048 {
3049         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3050         struct nfs4_server_caps_arg args = {
3051                 .fhandle = fhandle,
3052                 .bitmask = bitmask,
3053         };
3054         struct nfs4_server_caps_res res = {};
3055         struct rpc_message msg = {
3056                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3057                 .rpc_argp = &args,
3058                 .rpc_resp = &res,
3059         };
3060         int status;
3061
3062         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3063                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3064                      FATTR4_WORD0_LINK_SUPPORT |
3065                      FATTR4_WORD0_SYMLINK_SUPPORT |
3066                      FATTR4_WORD0_ACLSUPPORT;
3067         if (minorversion)
3068                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3069
3070         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3071         if (status == 0) {
3072                 /* Sanity check the server answers */
3073                 switch (minorversion) {
3074                 case 0:
3075                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3076                         res.attr_bitmask[2] = 0;
3077                         break;
3078                 case 1:
3079                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3080                         break;
3081                 case 2:
3082                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3083                 }
3084                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3085                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3086                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3087                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3088                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3089                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3090                                 NFS_CAP_SECURITY_LABEL);
3091                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3092                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3093                         server->caps |= NFS_CAP_ACLS;
3094                 if (res.has_links != 0)
3095                         server->caps |= NFS_CAP_HARDLINKS;
3096                 if (res.has_symlinks != 0)
3097                         server->caps |= NFS_CAP_SYMLINKS;
3098                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3099                         server->caps |= NFS_CAP_FILEID;
3100                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3101                         server->caps |= NFS_CAP_MODE;
3102                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3103                         server->caps |= NFS_CAP_NLINK;
3104                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3105                         server->caps |= NFS_CAP_OWNER;
3106                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3107                         server->caps |= NFS_CAP_OWNER_GROUP;
3108                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3109                         server->caps |= NFS_CAP_ATIME;
3110                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3111                         server->caps |= NFS_CAP_CTIME;
3112                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3113                         server->caps |= NFS_CAP_MTIME;
3114 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3115                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3116                         server->caps |= NFS_CAP_SECURITY_LABEL;
3117 #endif
3118                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3119                                 sizeof(server->attr_bitmask));
3120                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3121
3122                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3123                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3124                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3125                 server->cache_consistency_bitmask[2] = 0;
3126                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3127                         sizeof(server->exclcreat_bitmask));
3128                 server->acl_bitmask = res.acl_bitmask;
3129                 server->fh_expire_type = res.fh_expire_type;
3130         }
3131
3132         return status;
3133 }
3134
3135 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3136 {
3137         struct nfs4_exception exception = { };
3138         int err;
3139         do {
3140                 err = nfs4_handle_exception(server,
3141                                 _nfs4_server_capabilities(server, fhandle),
3142                                 &exception);
3143         } while (exception.retry);
3144         return err;
3145 }
3146
3147 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3148                 struct nfs_fsinfo *info)
3149 {
3150         u32 bitmask[3];
3151         struct nfs4_lookup_root_arg args = {
3152                 .bitmask = bitmask,
3153         };
3154         struct nfs4_lookup_res res = {
3155                 .server = server,
3156                 .fattr = info->fattr,
3157                 .fh = fhandle,
3158         };
3159         struct rpc_message msg = {
3160                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3161                 .rpc_argp = &args,
3162                 .rpc_resp = &res,
3163         };
3164
3165         bitmask[0] = nfs4_fattr_bitmap[0];
3166         bitmask[1] = nfs4_fattr_bitmap[1];
3167         /*
3168          * Process the label in the upcoming getfattr
3169          */
3170         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3171
3172         nfs_fattr_init(info->fattr);
3173         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3174 }
3175
3176 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3177                 struct nfs_fsinfo *info)
3178 {
3179         struct nfs4_exception exception = { };
3180         int err;
3181         do {
3182                 err = _nfs4_lookup_root(server, fhandle, info);
3183                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3184                 switch (err) {
3185                 case 0:
3186                 case -NFS4ERR_WRONGSEC:
3187                         goto out;
3188                 default:
3189                         err = nfs4_handle_exception(server, err, &exception);
3190                 }
3191         } while (exception.retry);
3192 out:
3193         return err;
3194 }
3195
3196 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3197                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3198 {
3199         struct rpc_auth_create_args auth_args = {
3200                 .pseudoflavor = flavor,
3201         };
3202         struct rpc_auth *auth;
3203         int ret;
3204
3205         auth = rpcauth_create(&auth_args, server->client);
3206         if (IS_ERR(auth)) {
3207                 ret = -EACCES;
3208                 goto out;
3209         }
3210         ret = nfs4_lookup_root(server, fhandle, info);
3211 out:
3212         return ret;
3213 }
3214
3215 /*
3216  * Retry pseudoroot lookup with various security flavors.  We do this when:
3217  *
3218  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3219  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3220  *
3221  * Returns zero on success, or a negative NFS4ERR value, or a
3222  * negative errno value.
3223  */
3224 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3225                               struct nfs_fsinfo *info)
3226 {
3227         /* Per 3530bis 15.33.5 */
3228         static const rpc_authflavor_t flav_array[] = {
3229                 RPC_AUTH_GSS_KRB5P,
3230                 RPC_AUTH_GSS_KRB5I,
3231                 RPC_AUTH_GSS_KRB5,
3232                 RPC_AUTH_UNIX,                  /* courtesy */
3233                 RPC_AUTH_NULL,
3234         };
3235         int status = -EPERM;
3236         size_t i;
3237
3238         if (server->auth_info.flavor_len > 0) {
3239                 /* try each flavor specified by user */
3240                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3241                         status = nfs4_lookup_root_sec(server, fhandle, info,
3242                                                 server->auth_info.flavors[i]);
3243                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3244                                 continue;
3245                         break;
3246                 }
3247         } else {
3248                 /* no flavors specified by user, try default list */
3249                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3250                         status = nfs4_lookup_root_sec(server, fhandle, info,
3251                                                       flav_array[i]);
3252                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3253                                 continue;
3254                         break;
3255                 }
3256         }
3257
3258         /*
3259          * -EACCESS could mean that the user doesn't have correct permissions
3260          * to access the mount.  It could also mean that we tried to mount
3261          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3262          * existing mount programs don't handle -EACCES very well so it should
3263          * be mapped to -EPERM instead.
3264          */
3265         if (status == -EACCES)
3266                 status = -EPERM;
3267         return status;
3268 }
3269
3270 static int nfs4_do_find_root_sec(struct nfs_server *server,
3271                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3272 {
3273         int mv = server->nfs_client->cl_minorversion;
3274         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3275 }
3276
3277 /**
3278  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3279  * @server: initialized nfs_server handle
3280  * @fhandle: we fill in the pseudo-fs root file handle
3281  * @info: we fill in an FSINFO struct
3282  * @auth_probe: probe the auth flavours
3283  *
3284  * Returns zero on success, or a negative errno.
3285  */
3286 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3287                          struct nfs_fsinfo *info,
3288                          bool auth_probe)
3289 {
3290         int status = 0;
3291
3292         if (!auth_probe)
3293                 status = nfs4_lookup_root(server, fhandle, info);
3294
3295         if (auth_probe || status == NFS4ERR_WRONGSEC)
3296                 status = nfs4_do_find_root_sec(server, fhandle, info);
3297
3298         if (status == 0)
3299                 status = nfs4_server_capabilities(server, fhandle);
3300         if (status == 0)
3301                 status = nfs4_do_fsinfo(server, fhandle, info);
3302
3303         return nfs4_map_errors(status);
3304 }
3305
3306 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3307                               struct nfs_fsinfo *info)
3308 {
3309         int error;
3310         struct nfs_fattr *fattr = info->fattr;
3311         struct nfs4_label *label = NULL;
3312
3313         error = nfs4_server_capabilities(server, mntfh);
3314         if (error < 0) {
3315                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3316                 return error;
3317         }
3318
3319         label = nfs4_label_alloc(server, GFP_KERNEL);
3320         if (IS_ERR(label))
3321                 return PTR_ERR(label);
3322
3323         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3324         if (error < 0) {
3325                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3326                 goto err_free_label;
3327         }
3328
3329         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3330             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3331                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3332
3333 err_free_label:
3334         nfs4_label_free(label);
3335
3336         return error;
3337 }
3338
3339 /*
3340  * Get locations and (maybe) other attributes of a referral.
3341  * Note that we'll actually follow the referral later when
3342  * we detect fsid mismatch in inode revalidation
3343  */
3344 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3345                              const struct qstr *name, struct nfs_fattr *fattr,
3346                              struct nfs_fh *fhandle)
3347 {
3348         int status = -ENOMEM;
3349         struct page *page = NULL;
3350         struct nfs4_fs_locations *locations = NULL;
3351
3352         page = alloc_page(GFP_KERNEL);
3353         if (page == NULL)
3354                 goto out;
3355         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3356         if (locations == NULL)
3357                 goto out;
3358
3359         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3360         if (status != 0)
3361                 goto out;
3362
3363         /*
3364          * If the fsid didn't change, this is a migration event, not a
3365          * referral.  Cause us to drop into the exception handler, which
3366          * will kick off migration recovery.
3367          */
3368         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3369                 dprintk("%s: server did not return a different fsid for"
3370                         " a referral at %s\n", __func__, name->name);
3371                 status = -NFS4ERR_MOVED;
3372                 goto out;
3373         }
3374         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3375         nfs_fixup_referral_attributes(&locations->fattr);
3376
3377         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3378         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3379         memset(fhandle, 0, sizeof(struct nfs_fh));
3380 out:
3381         if (page)
3382                 __free_page(page);
3383         kfree(locations);
3384         return status;
3385 }
3386
3387 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3388                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3389 {
3390         struct nfs4_getattr_arg args = {
3391                 .fh = fhandle,
3392                 .bitmask = server->attr_bitmask,
3393         };
3394         struct nfs4_getattr_res res = {
3395                 .fattr = fattr,
3396                 .label = label,
3397                 .server = server,
3398         };
3399         struct rpc_message msg = {
3400                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3401                 .rpc_argp = &args,
3402                 .rpc_resp = &res,
3403         };
3404
3405         args.bitmask = nfs4_bitmask(server, label);
3406
3407         nfs_fattr_init(fattr);
3408         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3409 }
3410
3411 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3412                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3413 {
3414         struct nfs4_exception exception = { };
3415         int err;
3416         do {
3417                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3418                 trace_nfs4_getattr(server, fhandle, fattr, err);
3419                 err = nfs4_handle_exception(server, err,
3420                                 &exception);
3421         } while (exception.retry);
3422         return err;
3423 }
3424
3425 /* 
3426  * The file is not closed if it is opened due to the a request to change
3427  * the size of the file. The open call will not be needed once the
3428  * VFS layer lookup-intents are implemented.
3429  *
3430  * Close is called when the inode is destroyed.
3431  * If we haven't opened the file for O_WRONLY, we
3432  * need to in the size_change case to obtain a stateid.
3433  *
3434  * Got race?
3435  * Because OPEN is always done by name in nfsv4, it is
3436  * possible that we opened a different file by the same
3437  * name.  We can recognize this race condition, but we
3438  * can't do anything about it besides returning an error.
3439  *
3440  * This will be fixed with VFS changes (lookup-intent).
3441  */
3442 static int
3443 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3444                   struct iattr *sattr)
3445 {
3446         struct inode *inode = d_inode(dentry);
3447         struct rpc_cred *cred = NULL;
3448         struct nfs4_state *state = NULL;
3449         struct nfs4_label *label = NULL;
3450         int status;
3451
3452         if (pnfs_ld_layoutret_on_setattr(inode) &&
3453             sattr->ia_valid & ATTR_SIZE &&
3454             sattr->ia_size < i_size_read(inode))
3455                 pnfs_commit_and_return_layout(inode);
3456
3457         nfs_fattr_init(fattr);
3458         
3459         /* Deal with open(O_TRUNC) */
3460         if (sattr->ia_valid & ATTR_OPEN)
3461                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3462
3463         /* Optimization: if the end result is no change, don't RPC */
3464         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3465                 return 0;
3466
3467         /* Search for an existing open(O_WRITE) file */
3468         if (sattr->ia_valid & ATTR_FILE) {
3469                 struct nfs_open_context *ctx;
3470
3471                 ctx = nfs_file_open_context(sattr->ia_file);
3472                 if (ctx) {
3473                         cred = ctx->cred;
3474                         state = ctx->state;
3475                 }
3476         }
3477
3478         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3479         if (IS_ERR(label))
3480                 return PTR_ERR(label);
3481
3482         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3483         if (status == 0) {
3484                 nfs_setattr_update_inode(inode, sattr, fattr);
3485                 nfs_setsecurity(inode, fattr, label);
3486         }
3487         nfs4_label_free(label);
3488         return status;
3489 }
3490
3491 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3492                 const struct qstr *name, struct nfs_fh *fhandle,
3493                 struct nfs_fattr *fattr, struct nfs4_label *label)
3494 {
3495         struct nfs_server *server = NFS_SERVER(dir);
3496         int                    status;
3497         struct nfs4_lookup_arg args = {
3498                 .bitmask = server->attr_bitmask,
3499                 .dir_fh = NFS_FH(dir),
3500                 .name = name,
3501         };
3502         struct nfs4_lookup_res res = {
3503                 .server = server,
3504                 .fattr = fattr,
3505                 .label = label,
3506                 .fh = fhandle,
3507         };
3508         struct rpc_message msg = {
3509                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3510                 .rpc_argp = &args,
3511                 .rpc_resp = &res,
3512         };
3513
3514         args.bitmask = nfs4_bitmask(server, label);
3515
3516         nfs_fattr_init(fattr);
3517
3518         dprintk("NFS call  lookup %s\n", name->name);
3519         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3520         dprintk("NFS reply lookup: %d\n", status);
3521         return status;
3522 }
3523
3524 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3525 {
3526         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3527                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3528         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3529         fattr->nlink = 2;
3530 }
3531
3532 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3533                                    struct qstr *name, struct nfs_fh *fhandle,
3534                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3535 {
3536         struct nfs4_exception exception = { };
3537         struct rpc_clnt *client = *clnt;
3538         int err;
3539         do {
3540                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3541                 trace_nfs4_lookup(dir, name, err);
3542                 switch (err) {
3543                 case -NFS4ERR_BADNAME:
3544                         err = -ENOENT;
3545                         goto out;
3546                 case -NFS4ERR_MOVED:
3547                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3548                         if (err == -NFS4ERR_MOVED)
3549                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3550                         goto out;
3551                 case -NFS4ERR_WRONGSEC:
3552                         err = -EPERM;
3553                         if (client != *clnt)
3554                                 goto out;
3555                         client = nfs4_negotiate_security(client, dir, name);
3556                         if (IS_ERR(client))
3557                                 return PTR_ERR(client);
3558
3559                         exception.retry = 1;
3560                         break;
3561                 default:
3562                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3563                 }
3564         } while (exception.retry);
3565
3566 out:
3567         if (err == 0)
3568                 *clnt = client;
3569         else if (client != *clnt)
3570                 rpc_shutdown_client(client);
3571
3572         return err;
3573 }
3574
3575 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3576                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3577                             struct nfs4_label *label)
3578 {
3579         int status;
3580         struct rpc_clnt *client = NFS_CLIENT(dir);
3581
3582         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3583         if (client != NFS_CLIENT(dir)) {
3584                 rpc_shutdown_client(client);
3585                 nfs_fixup_secinfo_attributes(fattr);
3586         }
3587         return status;
3588 }
3589
3590 struct rpc_clnt *
3591 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3592                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3593 {
3594         struct rpc_clnt *client = NFS_CLIENT(dir);
3595         int status;
3596
3597         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3598         if (status < 0)
3599                 return ERR_PTR(status);
3600         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3601 }
3602
3603 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3604 {
3605         struct nfs_server *server = NFS_SERVER(inode);
3606         struct nfs4_accessargs args = {
3607                 .fh = NFS_FH(inode),
3608                 .bitmask = server->cache_consistency_bitmask,
3609         };
3610         struct nfs4_accessres res = {
3611                 .server = server,
3612         };
3613         struct rpc_message msg = {
3614                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3615                 .rpc_argp = &args,
3616                 .rpc_resp = &res,
3617                 .rpc_cred = entry->cred,
3618         };
3619         int mode = entry->mask;
3620         int status = 0;
3621
3622         /*
3623          * Determine which access bits we want to ask for...
3624          */
3625         if (mode & MAY_READ)
3626                 args.access |= NFS4_ACCESS_READ;
3627         if (S_ISDIR(inode->i_mode)) {
3628                 if (mode & MAY_WRITE)
3629                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3630                 if (mode & MAY_EXEC)
3631                         args.access |= NFS4_ACCESS_LOOKUP;
3632         } else {
3633                 if (mode & MAY_WRITE)
3634                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3635                 if (mode & MAY_EXEC)
3636                         args.access |= NFS4_ACCESS_EXECUTE;
3637         }
3638
3639         res.fattr = nfs_alloc_fattr();
3640         if (res.fattr == NULL)
3641                 return -ENOMEM;
3642
3643         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3644         if (!status) {
3645                 nfs_access_set_mask(entry, res.access);
3646                 nfs_refresh_inode(inode, res.fattr);
3647         }
3648         nfs_free_fattr(res.fattr);
3649         return status;
3650 }
3651
3652 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3653 {
3654         struct nfs4_exception exception = { };
3655         int err;
3656         do {
3657                 err = _nfs4_proc_access(inode, entry);
3658                 trace_nfs4_access(inode, err);
3659                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3660                                 &exception);
3661         } while (exception.retry);
3662         return err;
3663 }
3664
3665 /*
3666  * TODO: For the time being, we don't try to get any attributes
3667  * along with any of the zero-copy operations READ, READDIR,
3668  * READLINK, WRITE.
3669  *
3670  * In the case of the first three, we want to put the GETATTR
3671  * after the read-type operation -- this is because it is hard
3672  * to predict the length of a GETATTR response in v4, and thus
3673  * align the READ data correctly.  This means that the GETATTR
3674  * may end up partially falling into the page cache, and we should
3675  * shift it into the 'tail' of the xdr_buf before processing.
3676  * To do this efficiently, we need to know the total length
3677  * of data received, which doesn't seem to be available outside
3678  * of the RPC layer.
3679  *
3680  * In the case of WRITE, we also want to put the GETATTR after
3681  * the operation -- in this case because we want to make sure
3682  * we get the post-operation mtime and size.
3683  *
3684  * Both of these changes to the XDR layer would in fact be quite
3685  * minor, but I decided to leave them for a subsequent patch.
3686  */
3687 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3688                 unsigned int pgbase, unsigned int pglen)
3689 {
3690         struct nfs4_readlink args = {
3691                 .fh       = NFS_FH(inode),
3692                 .pgbase   = pgbase,
3693                 .pglen    = pglen,
3694                 .pages    = &page,
3695         };
3696         struct nfs4_readlink_res res;
3697         struct rpc_message msg = {
3698                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3699                 .rpc_argp = &args,
3700                 .rpc_resp = &res,
3701         };
3702
3703         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3704 }
3705
3706 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3707                 unsigned int pgbase, unsigned int pglen)
3708 {
3709         struct nfs4_exception exception = { };
3710         int err;
3711         do {
3712                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3713                 trace_nfs4_readlink(inode, err);
3714                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3715                                 &exception);
3716         } while (exception.retry);
3717         return err;
3718 }
3719
3720 /*
3721  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3722  */
3723 static int
3724 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3725                  int flags)
3726 {
3727         struct nfs4_label l, *ilabel = NULL;
3728         struct nfs_open_context *ctx;
3729         struct nfs4_state *state;
3730         int status = 0;
3731
3732         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3733         if (IS_ERR(ctx))
3734                 return PTR_ERR(ctx);
3735
3736         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3737
3738         sattr->ia_mode &= ~current_umask();
3739         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3740         if (IS_ERR(state)) {
3741                 status = PTR_ERR(state);
3742                 goto out;
3743         }
3744 out:
3745         nfs4_label_release_security(ilabel);
3746         put_nfs_open_context(ctx);
3747         return status;
3748 }
3749
3750 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3751 {
3752         struct nfs_server *server = NFS_SERVER(dir);
3753         struct nfs_removeargs args = {
3754                 .fh = NFS_FH(dir),
3755                 .name = *name,
3756         };
3757         struct nfs_removeres res = {
3758                 .server = server,
3759         };
3760         struct rpc_message msg = {
3761                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3762                 .rpc_argp = &args,
3763                 .rpc_resp = &res,
3764         };
3765         int status;
3766
3767         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3768         if (status == 0)
3769                 update_changeattr(dir, &res.cinfo);
3770         return status;
3771 }
3772
3773 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3774 {
3775         struct nfs4_exception exception = { };
3776         int err;
3777         do {
3778                 err = _nfs4_proc_remove(dir, name);
3779                 trace_nfs4_remove(dir, name, err);
3780                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3781                                 &exception);
3782         } while (exception.retry);
3783         return err;
3784 }
3785
3786 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3787 {
3788         struct nfs_server *server = NFS_SERVER(dir);
3789         struct nfs_removeargs *args = msg->rpc_argp;
3790         struct nfs_removeres *res = msg->rpc_resp;
3791
3792         res->server = server;
3793         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3794         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3795
3796         nfs_fattr_init(res->dir_attr);
3797 }
3798
3799 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3800 {
3801         nfs4_setup_sequence(NFS_SERVER(data->dir),
3802                         &data->args.seq_args,
3803                         &data->res.seq_res,
3804                         task);
3805 }
3806
3807 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3808 {
3809         struct nfs_unlinkdata *data = task->tk_calldata;
3810         struct nfs_removeres *res = &data->res;
3811
3812         if (!nfs4_sequence_done(task, &res->seq_res))
3813                 return 0;
3814         if (nfs4_async_handle_error(task, res->server, NULL,
3815                                     &data->timeout) == -EAGAIN)
3816                 return 0;
3817         update_changeattr(dir, &res->cinfo);
3818         return 1;
3819 }
3820
3821 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3822 {
3823         struct nfs_server *server = NFS_SERVER(dir);
3824         struct nfs_renameargs *arg = msg->rpc_argp;
3825         struct nfs_renameres *res = msg->rpc_resp;
3826
3827         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3828         res->server = server;
3829         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3830 }
3831
3832 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3833 {
3834         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3835                         &data->args.seq_args,
3836                         &data->res.seq_res,
3837                         task);
3838 }
3839
3840 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3841                                  struct inode *new_dir)
3842 {
3843         struct nfs_renamedata *data = task->tk_calldata;
3844         struct nfs_renameres *res = &data->res;
3845
3846         if (!nfs4_sequence_done(task, &res->seq_res))
3847                 return 0;
3848         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3849                 return 0;
3850
3851         update_changeattr(old_dir, &res->old_cinfo);
3852         update_changeattr(new_dir, &res->new_cinfo);
3853         return 1;
3854 }
3855
3856 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3857 {
3858         struct nfs_server *server = NFS_SERVER(inode);
3859         struct nfs4_link_arg arg = {
3860                 .fh     = NFS_FH(inode),
3861                 .dir_fh = NFS_FH(dir),
3862                 .name   = name,
3863                 .bitmask = server->attr_bitmask,
3864         };
3865         struct nfs4_link_res res = {
3866                 .server = server,
3867                 .label = NULL,
3868         };
3869         struct rpc_message msg = {
3870                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3871                 .rpc_argp = &arg,
3872                 .rpc_resp = &res,
3873         };
3874         int status = -ENOMEM;
3875
3876         res.fattr = nfs_alloc_fattr();
3877         if (res.fattr == NULL)
3878                 goto out;
3879
3880         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3881         if (IS_ERR(res.label)) {
3882                 status = PTR_ERR(res.label);
3883                 goto out;
3884         }
3885         arg.bitmask = nfs4_bitmask(server, res.label);
3886
3887         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3888         if (!status) {
3889                 update_changeattr(dir, &res.cinfo);
3890                 status = nfs_post_op_update_inode(inode, res.fattr);
3891                 if (!status)
3892                         nfs_setsecurity(inode, res.fattr, res.label);
3893         }
3894
3895
3896         nfs4_label_free(res.label);
3897
3898 out:
3899         nfs_free_fattr(res.fattr);
3900         return status;
3901 }
3902
3903 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3904 {
3905         struct nfs4_exception exception = { };
3906         int err;
3907         do {
3908                 err = nfs4_handle_exception(NFS_SERVER(inode),
3909                                 _nfs4_proc_link(inode, dir, name),
3910                                 &exception);
3911         } while (exception.retry);
3912         return err;
3913 }
3914
3915 struct nfs4_createdata {
3916         struct rpc_message msg;
3917         struct nfs4_create_arg arg;
3918         struct nfs4_create_res res;
3919         struct nfs_fh fh;
3920         struct nfs_fattr fattr;
3921         struct nfs4_label *label;
3922 };
3923
3924 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3925                 struct qstr *name, struct iattr *sattr, u32 ftype)
3926 {
3927         struct nfs4_createdata *data;
3928
3929         data = kzalloc(sizeof(*data), GFP_KERNEL);
3930         if (data != NULL) {
3931                 struct nfs_server *server = NFS_SERVER(dir);
3932
3933                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3934                 if (IS_ERR(data->label))
3935                         goto out_free;
3936
3937                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3938                 data->msg.rpc_argp = &data->arg;
3939                 data->msg.rpc_resp = &data->res;
3940                 data->arg.dir_fh = NFS_FH(dir);
3941                 data->arg.server = server;
3942                 data->arg.name = name;
3943                 data->arg.attrs = sattr;
3944                 data->arg.ftype = ftype;
3945                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3946                 data->res.server = server;
3947                 data->res.fh = &data->fh;
3948                 data->res.fattr = &data->fattr;
3949                 data->res.label = data->label;
3950                 nfs_fattr_init(data->res.fattr);
3951         }
3952         return data;
3953 out_free:
3954         kfree(data);
3955         return NULL;
3956 }
3957
3958 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3959 {
3960         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3961                                     &data->arg.seq_args, &data->res.seq_res, 1);
3962         if (status == 0) {
3963                 update_changeattr(dir, &data->res.dir_cinfo);
3964                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3965         }
3966         return status;
3967 }
3968
3969 static void nfs4_free_createdata(struct nfs4_createdata *data)
3970 {
3971         nfs4_label_free(data->label);
3972         kfree(data);
3973 }
3974
3975 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3976                 struct page *page, unsigned int len, struct iattr *sattr,
3977                 struct nfs4_label *label)
3978 {
3979         struct nfs4_createdata *data;
3980         int status = -ENAMETOOLONG;
3981
3982         if (len > NFS4_MAXPATHLEN)
3983                 goto out;
3984
3985         status = -ENOMEM;
3986         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3987         if (data == NULL)
3988                 goto out;
3989
3990         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3991         data->arg.u.symlink.pages = &page;
3992         data->arg.u.symlink.len = len;
3993         data->arg.label = label;
3994         
3995         status = nfs4_do_create(dir, dentry, data);
3996
3997         nfs4_free_createdata(data);
3998 out:
3999         return status;
4000 }
4001
4002 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4003                 struct page *page, unsigned int len, struct iattr *sattr)
4004 {
4005         struct nfs4_exception exception = { };
4006         struct nfs4_label l, *label = NULL;
4007         int err;
4008
4009         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4010
4011         do {
4012                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4013                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4014                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4015                                 &exception);
4016         } while (exception.retry);
4017
4018         nfs4_label_release_security(label);
4019         return err;
4020 }
4021
4022 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4023                 struct iattr *sattr, struct nfs4_label *label)
4024 {
4025         struct nfs4_createdata *data;
4026         int status = -ENOMEM;
4027
4028         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4029         if (data == NULL)
4030                 goto out;
4031
4032         data->arg.label = label;
4033         status = nfs4_do_create(dir, dentry, data);
4034
4035         nfs4_free_createdata(data);
4036 out:
4037         return status;
4038 }
4039
4040 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4041                 struct iattr *sattr)
4042 {
4043         struct nfs4_exception exception = { };
4044         struct nfs4_label l, *label = NULL;
4045         int err;
4046
4047         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4048
4049         sattr->ia_mode &= ~current_umask();
4050         do {
4051                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4052                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4053                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4054                                 &exception);
4055         } while (exception.retry);
4056         nfs4_label_release_security(label);
4057
4058         return err;
4059 }
4060
4061 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4062                 u64 cookie, struct page **pages, unsigned int count, int plus)
4063 {
4064         struct inode            *dir = d_inode(dentry);
4065         struct nfs4_readdir_arg args = {
4066                 .fh = NFS_FH(dir),
4067                 .pages = pages,
4068                 .pgbase = 0,
4069                 .count = count,
4070                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4071                 .plus = plus,
4072         };
4073         struct nfs4_readdir_res res;
4074         struct rpc_message msg = {
4075                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4076                 .rpc_argp = &args,
4077                 .rpc_resp = &res,
4078                 .rpc_cred = cred,
4079         };
4080         int                     status;
4081
4082         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4083                         dentry,
4084                         (unsigned long long)cookie);
4085         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4086         res.pgbase = args.pgbase;
4087         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4088         if (status >= 0) {
4089                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4090                 status += args.pgbase;
4091         }
4092
4093         nfs_invalidate_atime(dir);
4094
4095         dprintk("%s: returns %d\n", __func__, status);
4096         return status;
4097 }
4098
4099 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4100                 u64 cookie, struct page **pages, unsigned int count, int plus)
4101 {
4102         struct nfs4_exception exception = { };
4103         int err;
4104         do {
4105                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4106                                 pages, count, plus);
4107                 trace_nfs4_readdir(d_inode(dentry), err);
4108                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4109                                 &exception);
4110         } while (exception.retry);
4111         return err;
4112 }
4113
4114 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4115                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4116 {
4117         struct nfs4_createdata *data;
4118         int mode = sattr->ia_mode;
4119         int status = -ENOMEM;
4120
4121         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4122         if (data == NULL)
4123                 goto out;
4124
4125         if (S_ISFIFO(mode))
4126                 data->arg.ftype = NF4FIFO;
4127         else if (S_ISBLK(mode)) {
4128                 data->arg.ftype = NF4BLK;
4129                 data->arg.u.device.specdata1 = MAJOR(rdev);
4130                 data->arg.u.device.specdata2 = MINOR(rdev);
4131         }
4132         else if (S_ISCHR(mode)) {
4133                 data->arg.ftype = NF4CHR;
4134                 data->arg.u.device.specdata1 = MAJOR(rdev);
4135                 data->arg.u.device.specdata2 = MINOR(rdev);
4136         } else if (!S_ISSOCK(mode)) {
4137                 status = -EINVAL;
4138                 goto out_free;
4139         }
4140
4141         data->arg.label = label;
4142         status = nfs4_do_create(dir, dentry, data);
4143 out_free:
4144         nfs4_free_createdata(data);
4145 out:
4146         return status;
4147 }
4148
4149 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4150                 struct iattr *sattr, dev_t rdev)
4151 {
4152         struct nfs4_exception exception = { };
4153         struct nfs4_label l, *label = NULL;
4154         int err;
4155
4156         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4157
4158         sattr->ia_mode &= ~current_umask();
4159         do {
4160                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4161                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4162                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4163                                 &exception);
4164         } while (exception.retry);
4165
4166         nfs4_label_release_security(label);
4167
4168         return err;
4169 }
4170
4171 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4172                  struct nfs_fsstat *fsstat)
4173 {
4174         struct nfs4_statfs_arg args = {
4175                 .fh = fhandle,
4176                 .bitmask = server->attr_bitmask,
4177         };
4178         struct nfs4_statfs_res res = {
4179                 .fsstat = fsstat,
4180         };
4181         struct rpc_message msg = {
4182                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4183                 .rpc_argp = &args,
4184                 .rpc_resp = &res,
4185         };
4186
4187         nfs_fattr_init(fsstat->fattr);
4188         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4189 }
4190
4191 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4192 {
4193         struct nfs4_exception exception = { };
4194         int err;
4195         do {
4196                 err = nfs4_handle_exception(server,
4197                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4198                                 &exception);
4199         } while (exception.retry);
4200         return err;
4201 }
4202
4203 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4204                 struct nfs_fsinfo *fsinfo)
4205 {
4206         struct nfs4_fsinfo_arg args = {
4207                 .fh = fhandle,
4208                 .bitmask = server->attr_bitmask,
4209         };
4210         struct nfs4_fsinfo_res res = {
4211                 .fsinfo = fsinfo,
4212         };
4213         struct rpc_message msg = {
4214                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4215                 .rpc_argp = &args,
4216                 .rpc_resp = &res,
4217         };
4218
4219         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4220 }
4221
4222 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4223 {
4224         struct nfs4_exception exception = { };
4225         unsigned long now = jiffies;
4226         int err;
4227
4228         do {
4229                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4230                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4231                 if (err == 0) {
4232                         struct nfs_client *clp = server->nfs_client;
4233
4234                         spin_lock(&clp->cl_lock);
4235                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4236                         clp->cl_last_renewal = now;
4237                         spin_unlock(&clp->cl_lock);
4238                         break;
4239                 }
4240                 err = nfs4_handle_exception(server, err, &exception);
4241         } while (exception.retry);
4242         return err;
4243 }
4244
4245 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4246 {
4247         int error;
4248
4249         nfs_fattr_init(fsinfo->fattr);
4250         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4251         if (error == 0) {
4252                 /* block layout checks this! */
4253                 server->pnfs_blksize = fsinfo->blksize;
4254                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4255         }
4256
4257         return error;
4258 }
4259
4260 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4261                 struct nfs_pathconf *pathconf)
4262 {
4263         struct nfs4_pathconf_arg args = {
4264                 .fh = fhandle,
4265                 .bitmask = server->attr_bitmask,
4266         };
4267         struct nfs4_pathconf_res res = {
4268                 .pathconf = pathconf,
4269         };
4270         struct rpc_message msg = {
4271                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4272                 .rpc_argp = &args,
4273                 .rpc_resp = &res,
4274         };
4275
4276         /* None of the pathconf attributes are mandatory to implement */
4277         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4278                 memset(pathconf, 0, sizeof(*pathconf));
4279                 return 0;
4280         }
4281
4282         nfs_fattr_init(pathconf->fattr);
4283         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4284 }
4285
4286 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4287                 struct nfs_pathconf *pathconf)
4288 {
4289         struct nfs4_exception exception = { };
4290         int err;
4291
4292         do {
4293                 err = nfs4_handle_exception(server,
4294                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4295                                 &exception);
4296         } while (exception.retry);
4297         return err;
4298 }
4299
4300 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4301                 const struct nfs_open_context *ctx,
4302                 const struct nfs_lock_context *l_ctx,
4303                 fmode_t fmode)
4304 {
4305         const struct nfs_lockowner *lockowner = NULL;
4306
4307         if (l_ctx != NULL)
4308                 lockowner = &l_ctx->lockowner;
4309         return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL);
4310 }
4311 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4312
4313 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4314                 const struct nfs_open_context *ctx,
4315                 const struct nfs_lock_context *l_ctx,
4316                 fmode_t fmode)
4317 {
4318         nfs4_stateid current_stateid;
4319
4320         /* If the current stateid represents a lost lock, then exit */
4321         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4322                 return true;
4323         return nfs4_stateid_match(stateid, &current_stateid);
4324 }
4325
4326 static bool nfs4_error_stateid_expired(int err)
4327 {
4328         switch (err) {
4329         case -NFS4ERR_DELEG_REVOKED:
4330         case -NFS4ERR_ADMIN_REVOKED:
4331         case -NFS4ERR_BAD_STATEID:
4332         case -NFS4ERR_STALE_STATEID:
4333         case -NFS4ERR_OLD_STATEID:
4334         case -NFS4ERR_OPENMODE:
4335         case -NFS4ERR_EXPIRED:
4336                 return true;
4337         }
4338         return false;
4339 }
4340
4341 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4342 {
4343         nfs_invalidate_atime(hdr->inode);
4344 }
4345
4346 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4347 {
4348         struct nfs_server *server = NFS_SERVER(hdr->inode);
4349
4350         trace_nfs4_read(hdr, task->tk_status);
4351         if (nfs4_async_handle_error(task, server,
4352                                     hdr->args.context->state,
4353                                     NULL) == -EAGAIN) {
4354                 rpc_restart_call_prepare(task);
4355                 return -EAGAIN;
4356         }
4357
4358         __nfs4_read_done_cb(hdr);
4359         if (task->tk_status > 0)
4360                 renew_lease(server, hdr->timestamp);
4361         return 0;
4362 }
4363
4364 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4365                 struct nfs_pgio_args *args)
4366 {
4367
4368         if (!nfs4_error_stateid_expired(task->tk_status) ||
4369                 nfs4_stateid_is_current(&args->stateid,
4370                                 args->context,
4371                                 args->lock_context,
4372                                 FMODE_READ))
4373                 return false;
4374         rpc_restart_call_prepare(task);
4375         return true;
4376 }
4377
4378 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4379 {
4380
4381         dprintk("--> %s\n", __func__);
4382
4383         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4384                 return -EAGAIN;
4385         if (nfs4_read_stateid_changed(task, &hdr->args))
4386                 return -EAGAIN;
4387         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4388                                     nfs4_read_done_cb(task, hdr);
4389 }
4390
4391 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4392                                  struct rpc_message *msg)
4393 {
4394         hdr->timestamp   = jiffies;
4395         hdr->pgio_done_cb = nfs4_read_done_cb;
4396         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4397         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4398 }
4399
4400 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4401                                       struct nfs_pgio_header *hdr)
4402 {
4403         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4404                         &hdr->args.seq_args,
4405                         &hdr->res.seq_res,
4406                         task))
4407                 return 0;
4408         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4409                                 hdr->args.lock_context,
4410                                 hdr->rw_ops->rw_mode) == -EIO)
4411                 return -EIO;
4412         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4413                 return -EIO;
4414         return 0;
4415 }
4416
4417 static int nfs4_write_done_cb(struct rpc_task *task,
4418                               struct nfs_pgio_header *hdr)
4419 {
4420         struct inode *inode = hdr->inode;
4421
4422         trace_nfs4_write(hdr, task->tk_status);
4423         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4424                                     hdr->args.context->state,
4425                                     NULL) == -EAGAIN) {
4426                 rpc_restart_call_prepare(task);
4427                 return -EAGAIN;
4428         }
4429         if (task->tk_status >= 0) {
4430                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4431                 nfs_writeback_update_inode(hdr);
4432         }
4433         return 0;
4434 }
4435
4436 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4437                 struct nfs_pgio_args *args)
4438 {
4439
4440         if (!nfs4_error_stateid_expired(task->tk_status) ||
4441                 nfs4_stateid_is_current(&args->stateid,
4442                                 args->context,
4443                                 args->lock_context,
4444                                 FMODE_WRITE))
4445                 return false;
4446         rpc_restart_call_prepare(task);
4447         return true;
4448 }
4449
4450 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4451 {
4452         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4453                 return -EAGAIN;
4454         if (nfs4_write_stateid_changed(task, &hdr->args))
4455                 return -EAGAIN;
4456         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4457                 nfs4_write_done_cb(task, hdr);
4458 }
4459
4460 static
4461 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4462 {
4463         /* Don't request attributes for pNFS or O_DIRECT writes */
4464         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4465                 return false;
4466         /* Otherwise, request attributes if and only if we don't hold
4467          * a delegation
4468          */
4469         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4470 }
4471
4472 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4473                                   struct rpc_message *msg)
4474 {
4475         struct nfs_server *server = NFS_SERVER(hdr->inode);
4476
4477         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4478                 hdr->args.bitmask = NULL;
4479                 hdr->res.fattr = NULL;
4480         } else
4481                 hdr->args.bitmask = server->cache_consistency_bitmask;
4482
4483         if (!hdr->pgio_done_cb)
4484                 hdr->pgio_done_cb = nfs4_write_done_cb;
4485         hdr->res.server = server;
4486         hdr->timestamp   = jiffies;
4487
4488         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4489         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4490 }
4491
4492 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4493 {
4494         nfs4_setup_sequence(NFS_SERVER(data->inode),
4495                         &data->args.seq_args,
4496                         &data->res.seq_res,
4497                         task);
4498 }
4499
4500 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4501 {
4502         struct inode *inode = data->inode;
4503
4504         trace_nfs4_commit(data, task->tk_status);
4505         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4506                                     NULL, NULL) == -EAGAIN) {
4507                 rpc_restart_call_prepare(task);
4508                 return -EAGAIN;
4509         }
4510         return 0;
4511 }
4512
4513 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4514 {
4515         if (!nfs4_sequence_done(task, &data->res.seq_res))
4516                 return -EAGAIN;
4517         return data->commit_done_cb(task, data);
4518 }
4519
4520 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4521 {
4522         struct nfs_server *server = NFS_SERVER(data->inode);
4523
4524         if (data->commit_done_cb == NULL)
4525                 data->commit_done_cb = nfs4_commit_done_cb;
4526         data->res.server = server;
4527         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4528         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4529 }
4530
4531 struct nfs4_renewdata {
4532         struct nfs_client       *client;
4533         unsigned long           timestamp;
4534 };
4535
4536 /*
4537  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4538  * standalone procedure for queueing an asynchronous RENEW.
4539  */
4540 static void nfs4_renew_release(void *calldata)
4541 {
4542         struct nfs4_renewdata *data = calldata;
4543         struct nfs_client *clp = data->client;
4544
4545         if (atomic_read(&clp->cl_count) > 1)
4546                 nfs4_schedule_state_renewal(clp);
4547         nfs_put_client(clp);
4548         kfree(data);
4549 }
4550
4551 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4552 {
4553         struct nfs4_renewdata *data = calldata;
4554         struct nfs_client *clp = data->client;
4555         unsigned long timestamp = data->timestamp;
4556
4557         trace_nfs4_renew_async(clp, task->tk_status);
4558         switch (task->tk_status) {
4559         case 0:
4560                 break;
4561         case -NFS4ERR_LEASE_MOVED:
4562                 nfs4_schedule_lease_moved_recovery(clp);
4563                 break;
4564         default:
4565                 /* Unless we're shutting down, schedule state recovery! */
4566                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4567                         return;
4568                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4569                         nfs4_schedule_lease_recovery(clp);
4570                         return;
4571                 }
4572                 nfs4_schedule_path_down_recovery(clp);
4573         }
4574         do_renew_lease(clp, timestamp);
4575 }
4576
4577 static const struct rpc_call_ops nfs4_renew_ops = {
4578         .rpc_call_done = nfs4_renew_done,
4579         .rpc_release = nfs4_renew_release,
4580 };
4581
4582 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4583 {
4584         struct rpc_message msg = {
4585                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4586                 .rpc_argp       = clp,
4587                 .rpc_cred       = cred,
4588         };
4589         struct nfs4_renewdata *data;
4590
4591         if (renew_flags == 0)
4592                 return 0;
4593         if (!atomic_inc_not_zero(&clp->cl_count))
4594                 return -EIO;
4595         data = kmalloc(sizeof(*data), GFP_NOFS);
4596         if (data == NULL)
4597                 return -ENOMEM;
4598         data->client = clp;
4599         data->timestamp = jiffies;
4600         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4601                         &nfs4_renew_ops, data);
4602 }
4603
4604 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4605 {
4606         struct rpc_message msg = {
4607                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4608                 .rpc_argp       = clp,
4609                 .rpc_cred       = cred,
4610         };
4611         unsigned long now = jiffies;
4612         int status;
4613
4614         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4615         if (status < 0)
4616                 return status;
4617         do_renew_lease(clp, now);
4618         return 0;
4619 }
4620
4621 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4622 {
4623         return server->caps & NFS_CAP_ACLS;
4624 }
4625
4626 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4627  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4628  * the stack.
4629  */
4630 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4631
4632 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4633                 struct page **pages)
4634 {
4635         struct page *newpage, **spages;
4636         int rc = 0;
4637         size_t len;
4638         spages = pages;
4639
4640         do {
4641                 len = min_t(size_t, PAGE_SIZE, buflen);
4642                 newpage = alloc_page(GFP_KERNEL);
4643
4644                 if (newpage == NULL)
4645                         goto unwind;
4646                 memcpy(page_address(newpage), buf, len);
4647                 buf += len;
4648                 buflen -= len;
4649                 *pages++ = newpage;
4650                 rc++;
4651         } while (buflen != 0);
4652
4653         return rc;
4654
4655 unwind:
4656         for(; rc > 0; rc--)
4657                 __free_page(spages[rc-1]);
4658         return -ENOMEM;
4659 }
4660
4661 struct nfs4_cached_acl {
4662         int cached;
4663         size_t len;
4664         char data[0];
4665 };
4666
4667 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4668 {
4669         struct nfs_inode *nfsi = NFS_I(inode);
4670
4671         spin_lock(&inode->i_lock);
4672         kfree(nfsi->nfs4_acl);
4673         nfsi->nfs4_acl = acl;
4674         spin_unlock(&inode->i_lock);
4675 }
4676
4677 static void nfs4_zap_acl_attr(struct inode *inode)
4678 {
4679         nfs4_set_cached_acl(inode, NULL);
4680 }
4681
4682 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4683 {
4684         struct nfs_inode *nfsi = NFS_I(inode);
4685         struct nfs4_cached_acl *acl;
4686         int ret = -ENOENT;
4687
4688         spin_lock(&inode->i_lock);
4689         acl = nfsi->nfs4_acl;
4690         if (acl == NULL)
4691                 goto out;
4692         if (buf == NULL) /* user is just asking for length */
4693                 goto out_len;
4694         if (acl->cached == 0)
4695                 goto out;
4696         ret = -ERANGE; /* see getxattr(2) man page */
4697         if (acl->len > buflen)
4698                 goto out;
4699         memcpy(buf, acl->data, acl->len);
4700 out_len:
4701         ret = acl->len;
4702 out:
4703         spin_unlock(&inode->i_lock);
4704         return ret;
4705 }
4706
4707 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4708 {
4709         struct nfs4_cached_acl *acl;
4710         size_t buflen = sizeof(*acl) + acl_len;
4711
4712         if (buflen <= PAGE_SIZE) {
4713                 acl = kmalloc(buflen, GFP_KERNEL);
4714                 if (acl == NULL)
4715                         goto out;
4716                 acl->cached = 1;
4717                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4718         } else {
4719                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4720                 if (acl == NULL)
4721                         goto out;
4722                 acl->cached = 0;
4723         }
4724         acl->len = acl_len;
4725 out:
4726         nfs4_set_cached_acl(inode, acl);
4727 }
4728
4729 /*
4730  * The getxattr API returns the required buffer length when called with a
4731  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4732  * the required buf.  On a NULL buf, we send a page of data to the server
4733  * guessing that the ACL request can be serviced by a page. If so, we cache
4734  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4735  * the cache. If not so, we throw away the page, and cache the required
4736  * length. The next getxattr call will then produce another round trip to
4737  * the server, this time with the input buf of the required size.
4738  */
4739 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4740 {
4741         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4742         struct nfs_getaclargs args = {
4743                 .fh = NFS_FH(inode),
4744                 .acl_pages = pages,
4745                 .acl_len = buflen,
4746         };
4747         struct nfs_getaclres res = {
4748                 .acl_len = buflen,
4749         };
4750         struct rpc_message msg = {
4751                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4752                 .rpc_argp = &args,
4753                 .rpc_resp = &res,
4754         };
4755         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4756         int ret = -ENOMEM, i;
4757
4758         /* As long as we're doing a round trip to the server anyway,
4759          * let's be prepared for a page of acl data. */
4760         if (npages == 0)
4761                 npages = 1;
4762         if (npages > ARRAY_SIZE(pages))
4763                 return -ERANGE;
4764
4765         for (i = 0; i < npages; i++) {
4766                 pages[i] = alloc_page(GFP_KERNEL);
4767                 if (!pages[i])
4768                         goto out_free;
4769         }
4770
4771         /* for decoding across pages */
4772         res.acl_scratch = alloc_page(GFP_KERNEL);
4773         if (!res.acl_scratch)
4774                 goto out_free;
4775
4776         args.acl_len = npages * PAGE_SIZE;
4777
4778         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4779                 __func__, buf, buflen, npages, args.acl_len);
4780         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4781                              &msg, &args.seq_args, &res.seq_res, 0);
4782         if (ret)
4783                 goto out_free;
4784
4785         /* Handle the case where the passed-in buffer is too short */
4786         if (res.acl_flags & NFS4_ACL_TRUNC) {
4787                 /* Did the user only issue a request for the acl length? */
4788                 if (buf == NULL)
4789                         goto out_ok;
4790                 ret = -ERANGE;
4791                 goto out_free;
4792         }
4793         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4794         if (buf) {
4795                 if (res.acl_len > buflen) {
4796                         ret = -ERANGE;
4797                         goto out_free;
4798                 }
4799                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4800         }
4801 out_ok:
4802         ret = res.acl_len;
4803 out_free:
4804         for (i = 0; i < npages; i++)
4805                 if (pages[i])
4806                         __free_page(pages[i]);
4807         if (res.acl_scratch)
4808                 __free_page(res.acl_scratch);
4809         return ret;
4810 }
4811
4812 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4813 {
4814         struct nfs4_exception exception = { };
4815         ssize_t ret;
4816         do {
4817                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4818                 trace_nfs4_get_acl(inode, ret);
4819                 if (ret >= 0)
4820                         break;
4821                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4822         } while (exception.retry);
4823         return ret;
4824 }
4825
4826 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4827 {
4828         struct nfs_server *server = NFS_SERVER(inode);
4829         int ret;
4830
4831         if (!nfs4_server_supports_acls(server))
4832                 return -EOPNOTSUPP;
4833         ret = nfs_revalidate_inode(server, inode);
4834         if (ret < 0)
4835                 return ret;
4836         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4837                 nfs_zap_acl_cache(inode);
4838         ret = nfs4_read_cached_acl(inode, buf, buflen);
4839         if (ret != -ENOENT)
4840                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4841                  * but no cached acl data, just the acl length */
4842                 return ret;
4843         return nfs4_get_acl_uncached(inode, buf, buflen);
4844 }
4845
4846 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4847 {
4848         struct nfs_server *server = NFS_SERVER(inode);
4849         struct page *pages[NFS4ACL_MAXPAGES];
4850         struct nfs_setaclargs arg = {
4851                 .fh             = NFS_FH(inode),
4852                 .acl_pages      = pages,
4853                 .acl_len        = buflen,
4854         };
4855         struct nfs_setaclres res;
4856         struct rpc_message msg = {
4857                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4858                 .rpc_argp       = &arg,
4859                 .rpc_resp       = &res,
4860         };
4861         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4862         int ret, i;
4863
4864         if (!nfs4_server_supports_acls(server))
4865                 return -EOPNOTSUPP;
4866         if (npages > ARRAY_SIZE(pages))
4867                 return -ERANGE;
4868         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4869         if (i < 0)
4870                 return i;
4871         nfs4_inode_return_delegation(inode);
4872         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4873
4874         /*
4875          * Free each page after tx, so the only ref left is
4876          * held by the network stack
4877          */
4878         for (; i > 0; i--)
4879                 put_page(pages[i-1]);
4880
4881         /*
4882          * Acl update can result in inode attribute update.
4883          * so mark the attribute cache invalid.
4884          */
4885         spin_lock(&inode->i_lock);
4886         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4887         spin_unlock(&inode->i_lock);
4888         nfs_access_zap_cache(inode);
4889         nfs_zap_acl_cache(inode);
4890         return ret;
4891 }
4892
4893 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4894 {
4895         struct nfs4_exception exception = { };
4896         int err;
4897         do {
4898                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4899                 trace_nfs4_set_acl(inode, err);
4900                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4901                                 &exception);
4902         } while (exception.retry);
4903         return err;
4904 }
4905
4906 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4907 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4908                                         size_t buflen)
4909 {
4910         struct nfs_server *server = NFS_SERVER(inode);
4911         struct nfs_fattr fattr;
4912         struct nfs4_label label = {0, 0, buflen, buf};
4913
4914         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4915         struct nfs4_getattr_arg arg = {
4916                 .fh             = NFS_FH(inode),
4917                 .bitmask        = bitmask,
4918         };
4919         struct nfs4_getattr_res res = {
4920                 .fattr          = &fattr,
4921                 .label          = &label,
4922                 .server         = server,
4923         };
4924         struct rpc_message msg = {
4925                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4926                 .rpc_argp       = &arg,
4927                 .rpc_resp       = &res,
4928         };
4929         int ret;
4930
4931         nfs_fattr_init(&fattr);
4932
4933         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4934         if (ret)
4935                 return ret;
4936         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4937                 return -ENOENT;
4938         if (buflen < label.len)
4939                 return -ERANGE;
4940         return 0;
4941 }
4942
4943 static int nfs4_get_security_label(struct inode *inode, void *buf,
4944                                         size_t buflen)
4945 {
4946         struct nfs4_exception exception = { };
4947         int err;
4948
4949         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4950                 return -EOPNOTSUPP;
4951
4952         do {
4953                 err = _nfs4_get_security_label(inode, buf, buflen);
4954                 trace_nfs4_get_security_label(inode, err);
4955                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4956                                 &exception);
4957         } while (exception.retry);
4958         return err;
4959 }
4960
4961 static int _nfs4_do_set_security_label(struct inode *inode,
4962                 struct nfs4_label *ilabel,
4963                 struct nfs_fattr *fattr,
4964                 struct nfs4_label *olabel)
4965 {
4966
4967         struct iattr sattr = {0};
4968         struct nfs_server *server = NFS_SERVER(inode);
4969         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4970         struct nfs_setattrargs arg = {
4971                 .fh             = NFS_FH(inode),
4972                 .iap            = &sattr,
4973                 .server         = server,
4974                 .bitmask        = bitmask,
4975                 .label          = ilabel,
4976         };
4977         struct nfs_setattrres res = {
4978                 .fattr          = fattr,
4979                 .label          = olabel,
4980                 .server         = server,
4981         };
4982         struct rpc_message msg = {
4983                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4984                 .rpc_argp       = &arg,
4985                 .rpc_resp       = &res,
4986         };
4987         int status;
4988
4989         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4990
4991         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4992         if (status)
4993                 dprintk("%s failed: %d\n", __func__, status);
4994
4995         return status;
4996 }
4997
4998 static int nfs4_do_set_security_label(struct inode *inode,
4999                 struct nfs4_label *ilabel,
5000                 struct nfs_fattr *fattr,
5001                 struct nfs4_label *olabel)
5002 {
5003         struct nfs4_exception exception = { };
5004         int err;
5005
5006         do {
5007                 err = _nfs4_do_set_security_label(inode, ilabel,
5008                                 fattr, olabel);
5009                 trace_nfs4_set_security_label(inode, err);
5010                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5011                                 &exception);
5012         } while (exception.retry);
5013         return err;
5014 }
5015
5016 static int
5017 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
5018 {
5019         struct nfs4_label ilabel, *olabel = NULL;
5020         struct nfs_fattr fattr;
5021         struct rpc_cred *cred;
5022         struct inode *inode = d_inode(dentry);
5023         int status;
5024
5025         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5026                 return -EOPNOTSUPP;
5027
5028         nfs_fattr_init(&fattr);
5029
5030         ilabel.pi = 0;
5031         ilabel.lfs = 0;
5032         ilabel.label = (char *)buf;
5033         ilabel.len = buflen;
5034
5035         cred = rpc_lookup_cred();
5036         if (IS_ERR(cred))
5037                 return PTR_ERR(cred);
5038
5039         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5040         if (IS_ERR(olabel)) {
5041                 status = -PTR_ERR(olabel);
5042                 goto out;
5043         }
5044
5045         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5046         if (status == 0)
5047                 nfs_setsecurity(inode, &fattr, olabel);
5048
5049         nfs4_label_free(olabel);
5050 out:
5051         put_rpccred(cred);
5052         return status;
5053 }
5054 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5055
5056
5057 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5058                                     nfs4_verifier *bootverf)
5059 {
5060         __be32 verf[2];
5061
5062         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5063                 /* An impossible timestamp guarantees this value
5064                  * will never match a generated boot time. */
5065                 verf[0] = 0;
5066                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5067         } else {
5068                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5069                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5070                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5071         }
5072         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5073 }
5074
5075 static int
5076 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5077 {
5078         size_t len;
5079         char *str;
5080
5081         if (clp->cl_owner_id != NULL)
5082                 return 0;
5083
5084         rcu_read_lock();
5085         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5086                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5087                 1 +
5088                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5089                 1;
5090         rcu_read_unlock();
5091
5092         if (len > NFS4_OPAQUE_LIMIT + 1)
5093                 return -EINVAL;
5094
5095         /*
5096          * Since this string is allocated at mount time, and held until the
5097          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5098          * about a memory-reclaim deadlock.
5099          */
5100         str = kmalloc(len, GFP_KERNEL);
5101         if (!str)
5102                 return -ENOMEM;
5103
5104         rcu_read_lock();
5105         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5106                         clp->cl_ipaddr,
5107                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5108                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5109         rcu_read_unlock();
5110
5111         clp->cl_owner_id = str;
5112         return 0;
5113 }
5114
5115 static int
5116 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5117 {
5118         size_t len;
5119         char *str;
5120
5121         len = 10 + 10 + 1 + 10 + 1 +
5122                 strlen(nfs4_client_id_uniquifier) + 1 +
5123                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5124
5125         if (len > NFS4_OPAQUE_LIMIT + 1)
5126                 return -EINVAL;
5127
5128         /*
5129          * Since this string is allocated at mount time, and held until the
5130          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5131          * about a memory-reclaim deadlock.
5132          */
5133         str = kmalloc(len, GFP_KERNEL);
5134         if (!str)
5135                 return -ENOMEM;
5136
5137         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5138                         clp->rpc_ops->version, clp->cl_minorversion,
5139                         nfs4_client_id_uniquifier,
5140                         clp->cl_rpcclient->cl_nodename);
5141         clp->cl_owner_id = str;
5142         return 0;
5143 }
5144
5145 static int
5146 nfs4_init_uniform_client_string(struct nfs_client *clp)
5147 {
5148         size_t len;
5149         char *str;
5150
5151         if (clp->cl_owner_id != NULL)
5152                 return 0;
5153
5154         if (nfs4_client_id_uniquifier[0] != '\0')
5155                 return nfs4_init_uniquifier_client_string(clp);
5156
5157         len = 10 + 10 + 1 + 10 + 1 +
5158                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5159
5160         if (len > NFS4_OPAQUE_LIMIT + 1)
5161                 return -EINVAL;
5162
5163         /*
5164          * Since this string is allocated at mount time, and held until the
5165          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5166          * about a memory-reclaim deadlock.
5167          */
5168         str = kmalloc(len, GFP_KERNEL);
5169         if (!str)
5170                 return -ENOMEM;
5171
5172         scnprintf(str, len, "Linux NFSv%u.%u %s",
5173                         clp->rpc_ops->version, clp->cl_minorversion,
5174                         clp->cl_rpcclient->cl_nodename);
5175         clp->cl_owner_id = str;
5176         return 0;
5177 }
5178
5179 /*
5180  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5181  * services.  Advertise one based on the address family of the
5182  * clientaddr.
5183  */
5184 static unsigned int
5185 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5186 {
5187         if (strchr(clp->cl_ipaddr, ':') != NULL)
5188                 return scnprintf(buf, len, "tcp6");
5189         else
5190                 return scnprintf(buf, len, "tcp");
5191 }
5192
5193 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5194 {
5195         struct nfs4_setclientid *sc = calldata;
5196
5197         if (task->tk_status == 0)
5198                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5199 }
5200
5201 static const struct rpc_call_ops nfs4_setclientid_ops = {
5202         .rpc_call_done = nfs4_setclientid_done,
5203 };
5204
5205 /**
5206  * nfs4_proc_setclientid - Negotiate client ID
5207  * @clp: state data structure
5208  * @program: RPC program for NFSv4 callback service
5209  * @port: IP port number for NFS4 callback service
5210  * @cred: RPC credential to use for this call
5211  * @res: where to place the result
5212  *
5213  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5214  */
5215 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5216                 unsigned short port, struct rpc_cred *cred,
5217                 struct nfs4_setclientid_res *res)
5218 {
5219         nfs4_verifier sc_verifier;
5220         struct nfs4_setclientid setclientid = {
5221                 .sc_verifier = &sc_verifier,
5222                 .sc_prog = program,
5223                 .sc_clnt = clp,
5224         };
5225         struct rpc_message msg = {
5226                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5227                 .rpc_argp = &setclientid,
5228                 .rpc_resp = res,
5229                 .rpc_cred = cred,
5230         };
5231         struct rpc_task *task;
5232         struct rpc_task_setup task_setup_data = {
5233                 .rpc_client = clp->cl_rpcclient,
5234                 .rpc_message = &msg,
5235                 .callback_ops = &nfs4_setclientid_ops,
5236                 .callback_data = &setclientid,
5237                 .flags = RPC_TASK_TIMEOUT,
5238         };
5239         int status;
5240
5241         /* nfs_client_id4 */
5242         nfs4_init_boot_verifier(clp, &sc_verifier);
5243
5244         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5245                 status = nfs4_init_uniform_client_string(clp);
5246         else
5247                 status = nfs4_init_nonuniform_client_string(clp);
5248
5249         if (status)
5250                 goto out;
5251
5252         /* cb_client4 */
5253         setclientid.sc_netid_len =
5254                                 nfs4_init_callback_netid(clp,
5255                                                 setclientid.sc_netid,
5256                                                 sizeof(setclientid.sc_netid));
5257         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5258                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5259                                 clp->cl_ipaddr, port >> 8, port & 255);
5260
5261         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5262                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5263                 clp->cl_owner_id);
5264         task = rpc_run_task(&task_setup_data);
5265         if (IS_ERR(task)) {
5266                 status = PTR_ERR(task);
5267                 goto out;
5268         }
5269         status = task->tk_status;
5270         if (setclientid.sc_cred) {
5271                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5272                 put_rpccred(setclientid.sc_cred);
5273         }
5274         rpc_put_task(task);
5275 out:
5276         trace_nfs4_setclientid(clp, status);
5277         dprintk("NFS reply setclientid: %d\n", status);
5278         return status;
5279 }
5280
5281 /**
5282  * nfs4_proc_setclientid_confirm - Confirm client ID
5283  * @clp: state data structure
5284  * @res: result of a previous SETCLIENTID
5285  * @cred: RPC credential to use for this call
5286  *
5287  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5288  */
5289 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5290                 struct nfs4_setclientid_res *arg,
5291                 struct rpc_cred *cred)
5292 {
5293         struct rpc_message msg = {
5294                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5295                 .rpc_argp = arg,
5296                 .rpc_cred = cred,
5297         };
5298         int status;
5299
5300         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5301                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5302                 clp->cl_clientid);
5303         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5304         trace_nfs4_setclientid_confirm(clp, status);
5305         dprintk("NFS reply setclientid_confirm: %d\n", status);
5306         return status;
5307 }
5308
5309 struct nfs4_delegreturndata {
5310         struct nfs4_delegreturnargs args;
5311         struct nfs4_delegreturnres res;
5312         struct nfs_fh fh;
5313         nfs4_stateid stateid;
5314         unsigned long timestamp;
5315         struct nfs_fattr fattr;
5316         int rpc_status;
5317         struct inode *inode;
5318         bool roc;
5319         u32 roc_barrier;
5320 };
5321
5322 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5323 {
5324         struct nfs4_delegreturndata *data = calldata;
5325
5326         if (!nfs4_sequence_done(task, &data->res.seq_res))
5327                 return;
5328
5329         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5330         switch (task->tk_status) {
5331         case 0:
5332                 renew_lease(data->res.server, data->timestamp);
5333         case -NFS4ERR_ADMIN_REVOKED:
5334         case -NFS4ERR_DELEG_REVOKED:
5335         case -NFS4ERR_BAD_STATEID:
5336         case -NFS4ERR_OLD_STATEID:
5337         case -NFS4ERR_STALE_STATEID:
5338         case -NFS4ERR_EXPIRED:
5339                 task->tk_status = 0;
5340                 if (data->roc)
5341                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5342                 break;
5343         default:
5344                 if (nfs4_async_handle_error(task, data->res.server,
5345                                             NULL, NULL) == -EAGAIN) {
5346                         rpc_restart_call_prepare(task);
5347                         return;
5348                 }
5349         }
5350         data->rpc_status = task->tk_status;
5351 }
5352
5353 static void nfs4_delegreturn_release(void *calldata)
5354 {
5355         struct nfs4_delegreturndata *data = calldata;
5356         struct inode *inode = data->inode;
5357
5358         if (inode) {
5359                 if (data->roc)
5360                         pnfs_roc_release(inode);
5361                 nfs_iput_and_deactive(inode);
5362         }
5363         kfree(calldata);
5364 }
5365
5366 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5367 {
5368         struct nfs4_delegreturndata *d_data;
5369
5370         d_data = (struct nfs4_delegreturndata *)data;
5371
5372         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5373                 return;
5374
5375         if (d_data->roc)
5376                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5377
5378         nfs4_setup_sequence(d_data->res.server,
5379                         &d_data->args.seq_args,
5380                         &d_data->res.seq_res,
5381                         task);
5382 }
5383
5384 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5385         .rpc_call_prepare = nfs4_delegreturn_prepare,
5386         .rpc_call_done = nfs4_delegreturn_done,
5387         .rpc_release = nfs4_delegreturn_release,
5388 };
5389
5390 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5391 {
5392         struct nfs4_delegreturndata *data;
5393         struct nfs_server *server = NFS_SERVER(inode);
5394         struct rpc_task *task;
5395         struct rpc_message msg = {
5396                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5397                 .rpc_cred = cred,
5398         };
5399         struct rpc_task_setup task_setup_data = {
5400                 .rpc_client = server->client,
5401                 .rpc_message = &msg,
5402                 .callback_ops = &nfs4_delegreturn_ops,
5403                 .flags = RPC_TASK_ASYNC,
5404         };
5405         int status = 0;
5406
5407         data = kzalloc(sizeof(*data), GFP_NOFS);
5408         if (data == NULL)
5409                 return -ENOMEM;
5410         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5411
5412         nfs4_state_protect(server->nfs_client,
5413                         NFS_SP4_MACH_CRED_CLEANUP,
5414                         &task_setup_data.rpc_client, &msg);
5415
5416         data->args.fhandle = &data->fh;
5417         data->args.stateid = &data->stateid;
5418         data->args.bitmask = server->cache_consistency_bitmask;
5419         nfs_copy_fh(&data->fh, NFS_FH(inode));
5420         nfs4_stateid_copy(&data->stateid, stateid);
5421         data->res.fattr = &data->fattr;
5422         data->res.server = server;
5423         nfs_fattr_init(data->res.fattr);
5424         data->timestamp = jiffies;
5425         data->rpc_status = 0;
5426         data->inode = nfs_igrab_and_active(inode);
5427         if (data->inode)
5428                 data->roc = nfs4_roc(inode);
5429
5430         task_setup_data.callback_data = data;
5431         msg.rpc_argp = &data->args;
5432         msg.rpc_resp = &data->res;
5433         task = rpc_run_task(&task_setup_data);
5434         if (IS_ERR(task))
5435                 return PTR_ERR(task);
5436         if (!issync)
5437                 goto out;
5438         status = nfs4_wait_for_completion_rpc_task(task);
5439         if (status != 0)
5440                 goto out;
5441         status = data->rpc_status;
5442         if (status == 0)
5443                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5444         else
5445                 nfs_refresh_inode(inode, &data->fattr);
5446 out:
5447         rpc_put_task(task);
5448         return status;
5449 }
5450
5451 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5452 {
5453         struct nfs_server *server = NFS_SERVER(inode);
5454         struct nfs4_exception exception = { };
5455         int err;
5456         do {
5457                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5458                 trace_nfs4_delegreturn(inode, stateid, err);
5459                 switch (err) {
5460                         case -NFS4ERR_STALE_STATEID:
5461                         case -NFS4ERR_EXPIRED:
5462                         case 0:
5463                                 return 0;
5464                 }
5465                 err = nfs4_handle_exception(server, err, &exception);
5466         } while (exception.retry);
5467         return err;
5468 }
5469
5470 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5471 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5472
5473 /* 
5474  * sleep, with exponential backoff, and retry the LOCK operation. 
5475  */
5476 static unsigned long
5477 nfs4_set_lock_task_retry(unsigned long timeout)
5478 {
5479         freezable_schedule_timeout_killable_unsafe(timeout);
5480         timeout <<= 1;
5481         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5482                 return NFS4_LOCK_MAXTIMEOUT;
5483         return timeout;
5484 }
5485
5486 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5487 {
5488         struct inode *inode = state->inode;
5489         struct nfs_server *server = NFS_SERVER(inode);
5490         struct nfs_client *clp = server->nfs_client;
5491         struct nfs_lockt_args arg = {
5492                 .fh = NFS_FH(inode),
5493                 .fl = request,
5494         };
5495         struct nfs_lockt_res res = {
5496                 .denied = request,
5497         };
5498         struct rpc_message msg = {
5499                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5500                 .rpc_argp       = &arg,
5501                 .rpc_resp       = &res,
5502                 .rpc_cred       = state->owner->so_cred,
5503         };
5504         struct nfs4_lock_state *lsp;
5505         int status;
5506
5507         arg.lock_owner.clientid = clp->cl_clientid;
5508         status = nfs4_set_lock_state(state, request);
5509         if (status != 0)
5510                 goto out;
5511         lsp = request->fl_u.nfs4_fl.owner;
5512         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5513         arg.lock_owner.s_dev = server->s_dev;
5514         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5515         switch (status) {
5516                 case 0:
5517                         request->fl_type = F_UNLCK;
5518                         break;
5519                 case -NFS4ERR_DENIED:
5520                         status = 0;
5521         }
5522         request->fl_ops->fl_release_private(request);
5523         request->fl_ops = NULL;
5524 out:
5525         return status;
5526 }
5527
5528 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5529 {
5530         struct nfs4_exception exception = { };
5531         int err;
5532
5533         do {
5534                 err = _nfs4_proc_getlk(state, cmd, request);
5535                 trace_nfs4_get_lock(request, state, cmd, err);
5536                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5537                                 &exception);
5538         } while (exception.retry);
5539         return err;
5540 }
5541
5542 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5543 {
5544         return locks_lock_inode_wait(inode, fl);
5545 }
5546
5547 struct nfs4_unlockdata {
5548         struct nfs_locku_args arg;
5549         struct nfs_locku_res res;
5550         struct nfs4_lock_state *lsp;
5551         struct nfs_open_context *ctx;
5552         struct file_lock fl;
5553         struct nfs_server *server;
5554         unsigned long timestamp;
5555 };
5556
5557 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5558                 struct nfs_open_context *ctx,
5559                 struct nfs4_lock_state *lsp,
5560                 struct nfs_seqid *seqid)
5561 {
5562         struct nfs4_unlockdata *p;
5563         struct inode *inode = lsp->ls_state->inode;
5564
5565         p = kzalloc(sizeof(*p), GFP_NOFS);
5566         if (p == NULL)
5567                 return NULL;
5568         p->arg.fh = NFS_FH(inode);
5569         p->arg.fl = &p->fl;
5570         p->arg.seqid = seqid;
5571         p->res.seqid = seqid;
5572         p->lsp = lsp;
5573         atomic_inc(&lsp->ls_count);
5574         /* Ensure we don't close file until we're done freeing locks! */
5575         p->ctx = get_nfs_open_context(ctx);
5576         memcpy(&p->fl, fl, sizeof(p->fl));
5577         p->server = NFS_SERVER(inode);
5578         return p;
5579 }
5580
5581 static void nfs4_locku_release_calldata(void *data)
5582 {
5583         struct nfs4_unlockdata *calldata = data;
5584         nfs_free_seqid(calldata->arg.seqid);
5585         nfs4_put_lock_state(calldata->lsp);
5586         put_nfs_open_context(calldata->ctx);
5587         kfree(calldata);
5588 }
5589
5590 static void nfs4_locku_done(struct rpc_task *task, void *data)
5591 {
5592         struct nfs4_unlockdata *calldata = data;
5593
5594         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5595                 return;
5596         switch (task->tk_status) {
5597                 case 0:
5598                         renew_lease(calldata->server, calldata->timestamp);
5599                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5600                         if (nfs4_update_lock_stateid(calldata->lsp,
5601                                         &calldata->res.stateid))
5602                                 break;
5603                 case -NFS4ERR_BAD_STATEID:
5604                 case -NFS4ERR_OLD_STATEID:
5605                 case -NFS4ERR_STALE_STATEID:
5606                 case -NFS4ERR_EXPIRED:
5607                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5608                                                 &calldata->lsp->ls_stateid))
5609                                 rpc_restart_call_prepare(task);
5610                         break;
5611                 default:
5612                         if (nfs4_async_handle_error(task, calldata->server,
5613                                                     NULL, NULL) == -EAGAIN)
5614                                 rpc_restart_call_prepare(task);
5615         }
5616         nfs_release_seqid(calldata->arg.seqid);
5617 }
5618
5619 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5620 {
5621         struct nfs4_unlockdata *calldata = data;
5622
5623         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5624                 goto out_wait;
5625         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5626         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5627                 /* Note: exit _without_ running nfs4_locku_done */
5628                 goto out_no_action;
5629         }
5630         calldata->timestamp = jiffies;
5631         if (nfs4_setup_sequence(calldata->server,
5632                                 &calldata->arg.seq_args,
5633                                 &calldata->res.seq_res,
5634                                 task) != 0)
5635                 nfs_release_seqid(calldata->arg.seqid);
5636         return;
5637 out_no_action:
5638         task->tk_action = NULL;
5639 out_wait:
5640         nfs4_sequence_done(task, &calldata->res.seq_res);
5641 }
5642
5643 static const struct rpc_call_ops nfs4_locku_ops = {
5644         .rpc_call_prepare = nfs4_locku_prepare,
5645         .rpc_call_done = nfs4_locku_done,
5646         .rpc_release = nfs4_locku_release_calldata,
5647 };
5648
5649 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5650                 struct nfs_open_context *ctx,
5651                 struct nfs4_lock_state *lsp,
5652                 struct nfs_seqid *seqid)
5653 {
5654         struct nfs4_unlockdata *data;
5655         struct rpc_message msg = {
5656                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5657                 .rpc_cred = ctx->cred,
5658         };
5659         struct rpc_task_setup task_setup_data = {
5660                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5661                 .rpc_message = &msg,
5662                 .callback_ops = &nfs4_locku_ops,
5663                 .workqueue = nfsiod_workqueue,
5664                 .flags = RPC_TASK_ASYNC,
5665         };
5666
5667         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5668                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5669
5670         /* Ensure this is an unlock - when canceling a lock, the
5671          * canceled lock is passed in, and it won't be an unlock.
5672          */
5673         fl->fl_type = F_UNLCK;
5674
5675         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5676         if (data == NULL) {
5677                 nfs_free_seqid(seqid);
5678                 return ERR_PTR(-ENOMEM);
5679         }
5680
5681         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5682         msg.rpc_argp = &data->arg;
5683         msg.rpc_resp = &data->res;
5684         task_setup_data.callback_data = data;
5685         return rpc_run_task(&task_setup_data);
5686 }
5687
5688 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5689 {
5690         struct inode *inode = state->inode;
5691         struct nfs4_state_owner *sp = state->owner;
5692         struct nfs_inode *nfsi = NFS_I(inode);
5693         struct nfs_seqid *seqid;
5694         struct nfs4_lock_state *lsp;
5695         struct rpc_task *task;
5696         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5697         int status = 0;
5698         unsigned char fl_flags = request->fl_flags;
5699
5700         status = nfs4_set_lock_state(state, request);
5701         /* Unlock _before_ we do the RPC call */
5702         request->fl_flags |= FL_EXISTS;
5703         /* Exclude nfs_delegation_claim_locks() */
5704         mutex_lock(&sp->so_delegreturn_mutex);
5705         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5706         down_read(&nfsi->rwsem);
5707         if (do_vfs_lock(inode, request) == -ENOENT) {
5708                 up_read(&nfsi->rwsem);
5709                 mutex_unlock(&sp->so_delegreturn_mutex);
5710                 goto out;
5711         }
5712         up_read(&nfsi->rwsem);
5713         mutex_unlock(&sp->so_delegreturn_mutex);
5714         if (status != 0)
5715                 goto out;
5716         /* Is this a delegated lock? */
5717         lsp = request->fl_u.nfs4_fl.owner;
5718         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5719                 goto out;
5720         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5721         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5722         status = -ENOMEM;
5723         if (IS_ERR(seqid))
5724                 goto out;
5725         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5726         status = PTR_ERR(task);
5727         if (IS_ERR(task))
5728                 goto out;
5729         status = nfs4_wait_for_completion_rpc_task(task);
5730         rpc_put_task(task);
5731 out:
5732         request->fl_flags = fl_flags;
5733         trace_nfs4_unlock(request, state, F_SETLK, status);
5734         return status;
5735 }
5736
5737 struct nfs4_lockdata {
5738         struct nfs_lock_args arg;
5739         struct nfs_lock_res res;
5740         struct nfs4_lock_state *lsp;
5741         struct nfs_open_context *ctx;
5742         struct file_lock fl;
5743         unsigned long timestamp;
5744         int rpc_status;
5745         int cancelled;
5746         struct nfs_server *server;
5747 };
5748
5749 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5750                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5751                 gfp_t gfp_mask)
5752 {
5753         struct nfs4_lockdata *p;
5754         struct inode *inode = lsp->ls_state->inode;
5755         struct nfs_server *server = NFS_SERVER(inode);
5756         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5757
5758         p = kzalloc(sizeof(*p), gfp_mask);
5759         if (p == NULL)
5760                 return NULL;
5761
5762         p->arg.fh = NFS_FH(inode);
5763         p->arg.fl = &p->fl;
5764         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5765         if (IS_ERR(p->arg.open_seqid))
5766                 goto out_free;
5767         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5768         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5769         if (IS_ERR(p->arg.lock_seqid))
5770                 goto out_free_seqid;
5771         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5772         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5773         p->arg.lock_owner.s_dev = server->s_dev;
5774         p->res.lock_seqid = p->arg.lock_seqid;
5775         p->lsp = lsp;
5776         p->server = server;
5777         atomic_inc(&lsp->ls_count);
5778         p->ctx = get_nfs_open_context(ctx);
5779         get_file(fl->fl_file);
5780         memcpy(&p->fl, fl, sizeof(p->fl));
5781         return p;
5782 out_free_seqid:
5783         nfs_free_seqid(p->arg.open_seqid);
5784 out_free:
5785         kfree(p);
5786         return NULL;
5787 }
5788
5789 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5790 {
5791         struct nfs4_lockdata *data = calldata;
5792         struct nfs4_state *state = data->lsp->ls_state;
5793
5794         dprintk("%s: begin!\n", __func__);
5795         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5796                 goto out_wait;
5797         /* Do we need to do an open_to_lock_owner? */
5798         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5799                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5800                         goto out_release_lock_seqid;
5801                 }
5802                 nfs4_stateid_copy(&data->arg.open_stateid,
5803                                 &state->open_stateid);
5804                 data->arg.new_lock_owner = 1;
5805                 data->res.open_seqid = data->arg.open_seqid;
5806         } else {
5807                 data->arg.new_lock_owner = 0;
5808                 nfs4_stateid_copy(&data->arg.lock_stateid,
5809                                 &data->lsp->ls_stateid);
5810         }
5811         if (!nfs4_valid_open_stateid(state)) {
5812                 data->rpc_status = -EBADF;
5813                 task->tk_action = NULL;
5814                 goto out_release_open_seqid;
5815         }
5816         data->timestamp = jiffies;
5817         if (nfs4_setup_sequence(data->server,
5818                                 &data->arg.seq_args,
5819                                 &data->res.seq_res,
5820                                 task) == 0)
5821                 return;
5822 out_release_open_seqid:
5823         nfs_release_seqid(data->arg.open_seqid);
5824 out_release_lock_seqid:
5825         nfs_release_seqid(data->arg.lock_seqid);
5826 out_wait:
5827         nfs4_sequence_done(task, &data->res.seq_res);
5828         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5829 }
5830
5831 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5832 {
5833         struct nfs4_lockdata *data = calldata;
5834         struct nfs4_lock_state *lsp = data->lsp;
5835
5836         dprintk("%s: begin!\n", __func__);
5837
5838         if (!nfs4_sequence_done(task, &data->res.seq_res))
5839                 return;
5840
5841         data->rpc_status = task->tk_status;
5842         switch (task->tk_status) {
5843         case 0:
5844                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5845                                 data->timestamp);
5846                 if (data->arg.new_lock) {
5847                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5848                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5849                                 rpc_restart_call_prepare(task);
5850                                 break;
5851                         }
5852                 }
5853                 if (data->arg.new_lock_owner != 0) {
5854                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5855                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5856                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5857                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5858                         rpc_restart_call_prepare(task);
5859                 break;
5860         case -NFS4ERR_BAD_STATEID:
5861         case -NFS4ERR_OLD_STATEID:
5862         case -NFS4ERR_STALE_STATEID:
5863         case -NFS4ERR_EXPIRED:
5864                 if (data->arg.new_lock_owner != 0) {
5865                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5866                                                 &lsp->ls_state->open_stateid))
5867                                 rpc_restart_call_prepare(task);
5868                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5869                                                 &lsp->ls_stateid))
5870                                 rpc_restart_call_prepare(task);
5871         }
5872         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5873 }
5874
5875 static void nfs4_lock_release(void *calldata)
5876 {
5877         struct nfs4_lockdata *data = calldata;
5878
5879         dprintk("%s: begin!\n", __func__);
5880         nfs_free_seqid(data->arg.open_seqid);
5881         if (data->cancelled != 0) {
5882                 struct rpc_task *task;
5883                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5884                                 data->arg.lock_seqid);
5885                 if (!IS_ERR(task))
5886                         rpc_put_task_async(task);
5887                 dprintk("%s: cancelling lock!\n", __func__);
5888         } else
5889                 nfs_free_seqid(data->arg.lock_seqid);
5890         nfs4_put_lock_state(data->lsp);
5891         put_nfs_open_context(data->ctx);
5892         fput(data->fl.fl_file);
5893         kfree(data);
5894         dprintk("%s: done!\n", __func__);
5895 }
5896
5897 static const struct rpc_call_ops nfs4_lock_ops = {
5898         .rpc_call_prepare = nfs4_lock_prepare,
5899         .rpc_call_done = nfs4_lock_done,
5900         .rpc_release = nfs4_lock_release,
5901 };
5902
5903 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5904 {
5905         switch (error) {
5906         case -NFS4ERR_ADMIN_REVOKED:
5907         case -NFS4ERR_BAD_STATEID:
5908                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5909                 if (new_lock_owner != 0 ||
5910                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5911                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5912                 break;
5913         case -NFS4ERR_STALE_STATEID:
5914                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5915         case -NFS4ERR_EXPIRED:
5916                 nfs4_schedule_lease_recovery(server->nfs_client);
5917         };
5918 }
5919
5920 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5921 {
5922         struct nfs4_lockdata *data;
5923         struct rpc_task *task;
5924         struct rpc_message msg = {
5925                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5926                 .rpc_cred = state->owner->so_cred,
5927         };
5928         struct rpc_task_setup task_setup_data = {
5929                 .rpc_client = NFS_CLIENT(state->inode),
5930                 .rpc_message = &msg,
5931                 .callback_ops = &nfs4_lock_ops,
5932                 .workqueue = nfsiod_workqueue,
5933                 .flags = RPC_TASK_ASYNC,
5934         };
5935         int ret;
5936
5937         dprintk("%s: begin!\n", __func__);
5938         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5939                         fl->fl_u.nfs4_fl.owner,
5940                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5941         if (data == NULL)
5942                 return -ENOMEM;
5943         if (IS_SETLKW(cmd))
5944                 data->arg.block = 1;
5945         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5946         msg.rpc_argp = &data->arg;
5947         msg.rpc_resp = &data->res;
5948         task_setup_data.callback_data = data;
5949         if (recovery_type > NFS_LOCK_NEW) {
5950                 if (recovery_type == NFS_LOCK_RECLAIM)
5951                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5952                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5953         } else
5954                 data->arg.new_lock = 1;
5955         task = rpc_run_task(&task_setup_data);
5956         if (IS_ERR(task))
5957                 return PTR_ERR(task);
5958         ret = nfs4_wait_for_completion_rpc_task(task);
5959         if (ret == 0) {
5960                 ret = data->rpc_status;
5961                 if (ret)
5962                         nfs4_handle_setlk_error(data->server, data->lsp,
5963                                         data->arg.new_lock_owner, ret);
5964         } else
5965                 data->cancelled = 1;
5966         rpc_put_task(task);
5967         dprintk("%s: done, ret = %d!\n", __func__, ret);
5968         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
5969         return ret;
5970 }
5971
5972 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5973 {
5974         struct nfs_server *server = NFS_SERVER(state->inode);
5975         struct nfs4_exception exception = {
5976                 .inode = state->inode,
5977         };
5978         int err;
5979
5980         do {
5981                 /* Cache the lock if possible... */
5982                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5983                         return 0;
5984                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5985                 if (err != -NFS4ERR_DELAY)
5986                         break;
5987                 nfs4_handle_exception(server, err, &exception);
5988         } while (exception.retry);
5989         return err;
5990 }
5991
5992 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5993 {
5994         struct nfs_server *server = NFS_SERVER(state->inode);
5995         struct nfs4_exception exception = {
5996                 .inode = state->inode,
5997         };
5998         int err;
5999
6000         err = nfs4_set_lock_state(state, request);
6001         if (err != 0)
6002                 return err;
6003         if (!recover_lost_locks) {
6004                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6005                 return 0;
6006         }
6007         do {
6008                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6009                         return 0;
6010                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6011                 switch (err) {
6012                 default:
6013                         goto out;
6014                 case -NFS4ERR_GRACE:
6015                 case -NFS4ERR_DELAY:
6016                         nfs4_handle_exception(server, err, &exception);
6017                         err = 0;
6018                 }
6019         } while (exception.retry);
6020 out:
6021         return err;
6022 }
6023
6024 #if defined(CONFIG_NFS_V4_1)
6025 /**
6026  * nfs41_check_expired_locks - possibly free a lock stateid
6027  *
6028  * @state: NFSv4 state for an inode
6029  *
6030  * Returns NFS_OK if recovery for this stateid is now finished.
6031  * Otherwise a negative NFS4ERR value is returned.
6032  */
6033 static int nfs41_check_expired_locks(struct nfs4_state *state)
6034 {
6035         int status, ret = -NFS4ERR_BAD_STATEID;
6036         struct nfs4_lock_state *lsp;
6037         struct nfs_server *server = NFS_SERVER(state->inode);
6038
6039         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6040                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6041                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6042
6043                         status = nfs41_test_stateid(server,
6044                                         &lsp->ls_stateid,
6045                                         cred);
6046                         trace_nfs4_test_lock_stateid(state, lsp, status);
6047                         if (status != NFS_OK) {
6048                                 /* Free the stateid unless the server
6049                                  * informs us the stateid is unrecognized. */
6050                                 if (status != -NFS4ERR_BAD_STATEID)
6051                                         nfs41_free_stateid(server,
6052                                                         &lsp->ls_stateid,
6053                                                         cred);
6054                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6055                                 ret = status;
6056                         }
6057                 }
6058         };
6059
6060         return ret;
6061 }
6062
6063 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6064 {
6065         int status = NFS_OK;
6066
6067         if (test_bit(LK_STATE_IN_USE, &state->flags))
6068                 status = nfs41_check_expired_locks(state);
6069         if (status != NFS_OK)
6070                 status = nfs4_lock_expired(state, request);
6071         return status;
6072 }
6073 #endif
6074
6075 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6076 {
6077         struct nfs_inode *nfsi = NFS_I(state->inode);
6078         struct nfs4_state_owner *sp = state->owner;
6079         unsigned char fl_flags = request->fl_flags;
6080         int status = -ENOLCK;
6081
6082         if ((fl_flags & FL_POSIX) &&
6083                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6084                 goto out;
6085         /* Is this a delegated open? */
6086         status = nfs4_set_lock_state(state, request);
6087         if (status != 0)
6088                 goto out;
6089         request->fl_flags |= FL_ACCESS;
6090         status = do_vfs_lock(state->inode, request);
6091         if (status < 0)
6092                 goto out;
6093         mutex_lock(&sp->so_delegreturn_mutex);
6094         down_read(&nfsi->rwsem);
6095         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6096                 /* Yes: cache locks! */
6097                 /* ...but avoid races with delegation recall... */
6098                 request->fl_flags = fl_flags & ~FL_SLEEP;
6099                 status = do_vfs_lock(state->inode, request);
6100                 up_read(&nfsi->rwsem);
6101                 mutex_unlock(&sp->so_delegreturn_mutex);
6102                 goto out;
6103         }
6104         up_read(&nfsi->rwsem);
6105         mutex_unlock(&sp->so_delegreturn_mutex);
6106         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6107 out:
6108         request->fl_flags = fl_flags;
6109         return status;
6110 }
6111
6112 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6113 {
6114         struct nfs4_exception exception = {
6115                 .state = state,
6116                 .inode = state->inode,
6117         };
6118         int err;
6119
6120         do {
6121                 err = _nfs4_proc_setlk(state, cmd, request);
6122                 if (err == -NFS4ERR_DENIED)
6123                         err = -EAGAIN;
6124                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6125                                 err, &exception);
6126         } while (exception.retry);
6127         return err;
6128 }
6129
6130 static int
6131 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6132 {
6133         struct nfs_open_context *ctx;
6134         struct nfs4_state *state;
6135         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6136         int status;
6137
6138         /* verify open state */
6139         ctx = nfs_file_open_context(filp);
6140         state = ctx->state;
6141
6142         if (request->fl_start < 0 || request->fl_end < 0)
6143                 return -EINVAL;
6144
6145         if (IS_GETLK(cmd)) {
6146                 if (state != NULL)
6147                         return nfs4_proc_getlk(state, F_GETLK, request);
6148                 return 0;
6149         }
6150
6151         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6152                 return -EINVAL;
6153
6154         if (request->fl_type == F_UNLCK) {
6155                 if (state != NULL)
6156                         return nfs4_proc_unlck(state, cmd, request);
6157                 return 0;
6158         }
6159
6160         if (state == NULL)
6161                 return -ENOLCK;
6162         /*
6163          * Don't rely on the VFS having checked the file open mode,
6164          * since it won't do this for flock() locks.
6165          */
6166         switch (request->fl_type) {
6167         case F_RDLCK:
6168                 if (!(filp->f_mode & FMODE_READ))
6169                         return -EBADF;
6170                 break;
6171         case F_WRLCK:
6172                 if (!(filp->f_mode & FMODE_WRITE))
6173                         return -EBADF;
6174         }
6175
6176         do {
6177                 status = nfs4_proc_setlk(state, cmd, request);
6178                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6179                         break;
6180                 timeout = nfs4_set_lock_task_retry(timeout);
6181                 status = -ERESTARTSYS;
6182                 if (signalled())
6183                         break;
6184         } while(status < 0);
6185         return status;
6186 }
6187
6188 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6189 {
6190         struct nfs_server *server = NFS_SERVER(state->inode);
6191         int err;
6192
6193         err = nfs4_set_lock_state(state, fl);
6194         if (err != 0)
6195                 return err;
6196         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6197         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6198 }
6199
6200 struct nfs_release_lockowner_data {
6201         struct nfs4_lock_state *lsp;
6202         struct nfs_server *server;
6203         struct nfs_release_lockowner_args args;
6204         struct nfs_release_lockowner_res res;
6205         unsigned long timestamp;
6206 };
6207
6208 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6209 {
6210         struct nfs_release_lockowner_data *data = calldata;
6211         struct nfs_server *server = data->server;
6212         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6213                              &data->args.seq_args, &data->res.seq_res, task);
6214         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6215         data->timestamp = jiffies;
6216 }
6217
6218 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6219 {
6220         struct nfs_release_lockowner_data *data = calldata;
6221         struct nfs_server *server = data->server;
6222
6223         nfs40_sequence_done(task, &data->res.seq_res);
6224
6225         switch (task->tk_status) {
6226         case 0:
6227                 renew_lease(server, data->timestamp);
6228                 break;
6229         case -NFS4ERR_STALE_CLIENTID:
6230         case -NFS4ERR_EXPIRED:
6231                 nfs4_schedule_lease_recovery(server->nfs_client);
6232                 break;
6233         case -NFS4ERR_LEASE_MOVED:
6234         case -NFS4ERR_DELAY:
6235                 if (nfs4_async_handle_error(task, server,
6236                                             NULL, NULL) == -EAGAIN)
6237                         rpc_restart_call_prepare(task);
6238         }
6239 }
6240
6241 static void nfs4_release_lockowner_release(void *calldata)
6242 {
6243         struct nfs_release_lockowner_data *data = calldata;
6244         nfs4_free_lock_state(data->server, data->lsp);
6245         kfree(calldata);
6246 }
6247
6248 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6249         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6250         .rpc_call_done = nfs4_release_lockowner_done,
6251         .rpc_release = nfs4_release_lockowner_release,
6252 };
6253
6254 static void
6255 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6256 {
6257         struct nfs_release_lockowner_data *data;
6258         struct rpc_message msg = {
6259                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6260         };
6261
6262         if (server->nfs_client->cl_mvops->minor_version != 0)
6263                 return;
6264
6265         data = kmalloc(sizeof(*data), GFP_NOFS);
6266         if (!data)
6267                 return;
6268         data->lsp = lsp;
6269         data->server = server;
6270         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6271         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6272         data->args.lock_owner.s_dev = server->s_dev;
6273
6274         msg.rpc_argp = &data->args;
6275         msg.rpc_resp = &data->res;
6276         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6277         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6278 }
6279
6280 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6281
6282 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6283                                    struct dentry *dentry, const char *key,
6284                                    const void *buf, size_t buflen,
6285                                    int flags)
6286 {
6287         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6288 }
6289
6290 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6291                                    struct dentry *dentry, const char *key,
6292                                    void *buf, size_t buflen)
6293 {
6294         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6295 }
6296
6297 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6298 {
6299         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6300 }
6301
6302 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6303
6304 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6305                                      struct dentry *dentry, const char *key,
6306                                      const void *buf, size_t buflen,
6307                                      int flags)
6308 {
6309         if (security_ismaclabel(key))
6310                 return nfs4_set_security_label(dentry, buf, buflen);
6311
6312         return -EOPNOTSUPP;
6313 }
6314
6315 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6316                                      struct dentry *dentry, const char *key,
6317                                      void *buf, size_t buflen)
6318 {
6319         if (security_ismaclabel(key))
6320                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6321         return -EOPNOTSUPP;
6322 }
6323
6324 static ssize_t
6325 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6326 {
6327         int len = 0;
6328
6329         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6330                 len = security_inode_listsecurity(inode, list, list_len);
6331                 if (list_len && len > list_len)
6332                         return -ERANGE;
6333         }
6334         return len;
6335 }
6336
6337 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6338         .prefix = XATTR_SECURITY_PREFIX,
6339         .get    = nfs4_xattr_get_nfs4_label,
6340         .set    = nfs4_xattr_set_nfs4_label,
6341 };
6342
6343 #else
6344
6345 static ssize_t
6346 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6347 {
6348         return 0;
6349 }
6350
6351 #endif
6352
6353 /*
6354  * nfs_fhget will use either the mounted_on_fileid or the fileid
6355  */
6356 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6357 {
6358         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6359                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6360               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6361               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6362                 return;
6363
6364         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6365                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6366         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6367         fattr->nlink = 2;
6368 }
6369
6370 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6371                                    const struct qstr *name,
6372                                    struct nfs4_fs_locations *fs_locations,
6373                                    struct page *page)
6374 {
6375         struct nfs_server *server = NFS_SERVER(dir);
6376         u32 bitmask[3] = {
6377                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6378         };
6379         struct nfs4_fs_locations_arg args = {
6380                 .dir_fh = NFS_FH(dir),
6381                 .name = name,
6382                 .page = page,
6383                 .bitmask = bitmask,
6384         };
6385         struct nfs4_fs_locations_res res = {
6386                 .fs_locations = fs_locations,
6387         };
6388         struct rpc_message msg = {
6389                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6390                 .rpc_argp = &args,
6391                 .rpc_resp = &res,
6392         };
6393         int status;
6394
6395         dprintk("%s: start\n", __func__);
6396
6397         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6398          * is not supported */
6399         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6400                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6401         else
6402                 bitmask[0] |= FATTR4_WORD0_FILEID;
6403
6404         nfs_fattr_init(&fs_locations->fattr);
6405         fs_locations->server = server;
6406         fs_locations->nlocations = 0;
6407         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6408         dprintk("%s: returned status = %d\n", __func__, status);
6409         return status;
6410 }
6411
6412 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6413                            const struct qstr *name,
6414                            struct nfs4_fs_locations *fs_locations,
6415                            struct page *page)
6416 {
6417         struct nfs4_exception exception = { };
6418         int err;
6419         do {
6420                 err = _nfs4_proc_fs_locations(client, dir, name,
6421                                 fs_locations, page);
6422                 trace_nfs4_get_fs_locations(dir, name, err);
6423                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6424                                 &exception);
6425         } while (exception.retry);
6426         return err;
6427 }
6428
6429 /*
6430  * This operation also signals the server that this client is
6431  * performing migration recovery.  The server can stop returning
6432  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6433  * appended to this compound to identify the client ID which is
6434  * performing recovery.
6435  */
6436 static int _nfs40_proc_get_locations(struct inode *inode,
6437                                      struct nfs4_fs_locations *locations,
6438                                      struct page *page, struct rpc_cred *cred)
6439 {
6440         struct nfs_server *server = NFS_SERVER(inode);
6441         struct rpc_clnt *clnt = server->client;
6442         u32 bitmask[2] = {
6443                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6444         };
6445         struct nfs4_fs_locations_arg args = {
6446                 .clientid       = server->nfs_client->cl_clientid,
6447                 .fh             = NFS_FH(inode),
6448                 .page           = page,
6449                 .bitmask        = bitmask,
6450                 .migration      = 1,            /* skip LOOKUP */
6451                 .renew          = 1,            /* append RENEW */
6452         };
6453         struct nfs4_fs_locations_res res = {
6454                 .fs_locations   = locations,
6455                 .migration      = 1,
6456                 .renew          = 1,
6457         };
6458         struct rpc_message msg = {
6459                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6460                 .rpc_argp       = &args,
6461                 .rpc_resp       = &res,
6462                 .rpc_cred       = cred,
6463         };
6464         unsigned long now = jiffies;
6465         int status;
6466
6467         nfs_fattr_init(&locations->fattr);
6468         locations->server = server;
6469         locations->nlocations = 0;
6470
6471         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6472         nfs4_set_sequence_privileged(&args.seq_args);
6473         status = nfs4_call_sync_sequence(clnt, server, &msg,
6474                                         &args.seq_args, &res.seq_res);
6475         if (status)
6476                 return status;
6477
6478         renew_lease(server, now);
6479         return 0;
6480 }
6481
6482 #ifdef CONFIG_NFS_V4_1
6483
6484 /*
6485  * This operation also signals the server that this client is
6486  * performing migration recovery.  The server can stop asserting
6487  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6488  * performing this operation is identified in the SEQUENCE
6489  * operation in this compound.
6490  *
6491  * When the client supports GETATTR(fs_locations_info), it can
6492  * be plumbed in here.
6493  */
6494 static int _nfs41_proc_get_locations(struct inode *inode,
6495                                      struct nfs4_fs_locations *locations,
6496                                      struct page *page, struct rpc_cred *cred)
6497 {
6498         struct nfs_server *server = NFS_SERVER(inode);
6499         struct rpc_clnt *clnt = server->client;
6500         u32 bitmask[2] = {
6501                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6502         };
6503         struct nfs4_fs_locations_arg args = {
6504                 .fh             = NFS_FH(inode),
6505                 .page           = page,
6506                 .bitmask        = bitmask,
6507                 .migration      = 1,            /* skip LOOKUP */
6508         };
6509         struct nfs4_fs_locations_res res = {
6510                 .fs_locations   = locations,
6511                 .migration      = 1,
6512         };
6513         struct rpc_message msg = {
6514                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6515                 .rpc_argp       = &args,
6516                 .rpc_resp       = &res,
6517                 .rpc_cred       = cred,
6518         };
6519         int status;
6520
6521         nfs_fattr_init(&locations->fattr);
6522         locations->server = server;
6523         locations->nlocations = 0;
6524
6525         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6526         nfs4_set_sequence_privileged(&args.seq_args);
6527         status = nfs4_call_sync_sequence(clnt, server, &msg,
6528                                         &args.seq_args, &res.seq_res);
6529         if (status == NFS4_OK &&
6530             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6531                 status = -NFS4ERR_LEASE_MOVED;
6532         return status;
6533 }
6534
6535 #endif  /* CONFIG_NFS_V4_1 */
6536
6537 /**
6538  * nfs4_proc_get_locations - discover locations for a migrated FSID
6539  * @inode: inode on FSID that is migrating
6540  * @locations: result of query
6541  * @page: buffer
6542  * @cred: credential to use for this operation
6543  *
6544  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6545  * operation failed, or a negative errno if a local error occurred.
6546  *
6547  * On success, "locations" is filled in, but if the server has
6548  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6549  * asserted.
6550  *
6551  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6552  * from this client that require migration recovery.
6553  */
6554 int nfs4_proc_get_locations(struct inode *inode,
6555                             struct nfs4_fs_locations *locations,
6556                             struct page *page, struct rpc_cred *cred)
6557 {
6558         struct nfs_server *server = NFS_SERVER(inode);
6559         struct nfs_client *clp = server->nfs_client;
6560         const struct nfs4_mig_recovery_ops *ops =
6561                                         clp->cl_mvops->mig_recovery_ops;
6562         struct nfs4_exception exception = { };
6563         int status;
6564
6565         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6566                 (unsigned long long)server->fsid.major,
6567                 (unsigned long long)server->fsid.minor,
6568                 clp->cl_hostname);
6569         nfs_display_fhandle(NFS_FH(inode), __func__);
6570
6571         do {
6572                 status = ops->get_locations(inode, locations, page, cred);
6573                 if (status != -NFS4ERR_DELAY)
6574                         break;
6575                 nfs4_handle_exception(server, status, &exception);
6576         } while (exception.retry);
6577         return status;
6578 }
6579
6580 /*
6581  * This operation also signals the server that this client is
6582  * performing "lease moved" recovery.  The server can stop
6583  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6584  * is appended to this compound to identify the client ID which is
6585  * performing recovery.
6586  */
6587 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6588 {
6589         struct nfs_server *server = NFS_SERVER(inode);
6590         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6591         struct rpc_clnt *clnt = server->client;
6592         struct nfs4_fsid_present_arg args = {
6593                 .fh             = NFS_FH(inode),
6594                 .clientid       = clp->cl_clientid,
6595                 .renew          = 1,            /* append RENEW */
6596         };
6597         struct nfs4_fsid_present_res res = {
6598                 .renew          = 1,
6599         };
6600         struct rpc_message msg = {
6601                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6602                 .rpc_argp       = &args,
6603                 .rpc_resp       = &res,
6604                 .rpc_cred       = cred,
6605         };
6606         unsigned long now = jiffies;
6607         int status;
6608
6609         res.fh = nfs_alloc_fhandle();
6610         if (res.fh == NULL)
6611                 return -ENOMEM;
6612
6613         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6614         nfs4_set_sequence_privileged(&args.seq_args);
6615         status = nfs4_call_sync_sequence(clnt, server, &msg,
6616                                                 &args.seq_args, &res.seq_res);
6617         nfs_free_fhandle(res.fh);
6618         if (status)
6619                 return status;
6620
6621         do_renew_lease(clp, now);
6622         return 0;
6623 }
6624
6625 #ifdef CONFIG_NFS_V4_1
6626
6627 /*
6628  * This operation also signals the server that this client is
6629  * performing "lease moved" recovery.  The server can stop asserting
6630  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6631  * this operation is identified in the SEQUENCE operation in this
6632  * compound.
6633  */
6634 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6635 {
6636         struct nfs_server *server = NFS_SERVER(inode);
6637         struct rpc_clnt *clnt = server->client;
6638         struct nfs4_fsid_present_arg args = {
6639                 .fh             = NFS_FH(inode),
6640         };
6641         struct nfs4_fsid_present_res res = {
6642         };
6643         struct rpc_message msg = {
6644                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6645                 .rpc_argp       = &args,
6646                 .rpc_resp       = &res,
6647                 .rpc_cred       = cred,
6648         };
6649         int status;
6650
6651         res.fh = nfs_alloc_fhandle();
6652         if (res.fh == NULL)
6653                 return -ENOMEM;
6654
6655         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6656         nfs4_set_sequence_privileged(&args.seq_args);
6657         status = nfs4_call_sync_sequence(clnt, server, &msg,
6658                                                 &args.seq_args, &res.seq_res);
6659         nfs_free_fhandle(res.fh);
6660         if (status == NFS4_OK &&
6661             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6662                 status = -NFS4ERR_LEASE_MOVED;
6663         return status;
6664 }
6665
6666 #endif  /* CONFIG_NFS_V4_1 */
6667
6668 /**
6669  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6670  * @inode: inode on FSID to check
6671  * @cred: credential to use for this operation
6672  *
6673  * Server indicates whether the FSID is present, moved, or not
6674  * recognized.  This operation is necessary to clear a LEASE_MOVED
6675  * condition for this client ID.
6676  *
6677  * Returns NFS4_OK if the FSID is present on this server,
6678  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6679  *  NFS4ERR code if some error occurred on the server, or a
6680  *  negative errno if a local failure occurred.
6681  */
6682 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6683 {
6684         struct nfs_server *server = NFS_SERVER(inode);
6685         struct nfs_client *clp = server->nfs_client;
6686         const struct nfs4_mig_recovery_ops *ops =
6687                                         clp->cl_mvops->mig_recovery_ops;
6688         struct nfs4_exception exception = { };
6689         int status;
6690
6691         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6692                 (unsigned long long)server->fsid.major,
6693                 (unsigned long long)server->fsid.minor,
6694                 clp->cl_hostname);
6695         nfs_display_fhandle(NFS_FH(inode), __func__);
6696
6697         do {
6698                 status = ops->fsid_present(inode, cred);
6699                 if (status != -NFS4ERR_DELAY)
6700                         break;
6701                 nfs4_handle_exception(server, status, &exception);
6702         } while (exception.retry);
6703         return status;
6704 }
6705
6706 /**
6707  * If 'use_integrity' is true and the state managment nfs_client
6708  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6709  * and the machine credential as per RFC3530bis and RFC5661 Security
6710  * Considerations sections. Otherwise, just use the user cred with the
6711  * filesystem's rpc_client.
6712  */
6713 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6714 {
6715         int status;
6716         struct nfs4_secinfo_arg args = {
6717                 .dir_fh = NFS_FH(dir),
6718                 .name   = name,
6719         };
6720         struct nfs4_secinfo_res res = {
6721                 .flavors     = flavors,
6722         };
6723         struct rpc_message msg = {
6724                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6725                 .rpc_argp = &args,
6726                 .rpc_resp = &res,
6727         };
6728         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6729         struct rpc_cred *cred = NULL;
6730
6731         if (use_integrity) {
6732                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6733                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6734                 msg.rpc_cred = cred;
6735         }
6736
6737         dprintk("NFS call  secinfo %s\n", name->name);
6738
6739         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6740                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6741
6742         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6743                                 &res.seq_res, 0);
6744         dprintk("NFS reply  secinfo: %d\n", status);
6745
6746         if (cred)
6747                 put_rpccred(cred);
6748
6749         return status;
6750 }
6751
6752 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6753                       struct nfs4_secinfo_flavors *flavors)
6754 {
6755         struct nfs4_exception exception = { };
6756         int err;
6757         do {
6758                 err = -NFS4ERR_WRONGSEC;
6759
6760                 /* try to use integrity protection with machine cred */
6761                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6762                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6763
6764                 /*
6765                  * if unable to use integrity protection, or SECINFO with
6766                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6767                  * disallowed by spec, but exists in deployed servers) use
6768                  * the current filesystem's rpc_client and the user cred.
6769                  */
6770                 if (err == -NFS4ERR_WRONGSEC)
6771                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6772
6773                 trace_nfs4_secinfo(dir, name, err);
6774                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6775                                 &exception);
6776         } while (exception.retry);
6777         return err;
6778 }
6779
6780 #ifdef CONFIG_NFS_V4_1
6781 /*
6782  * Check the exchange flags returned by the server for invalid flags, having
6783  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6784  * DS flags set.
6785  */
6786 static int nfs4_check_cl_exchange_flags(u32 flags)
6787 {
6788         if (flags & ~EXCHGID4_FLAG_MASK_R)
6789                 goto out_inval;
6790         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6791             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6792                 goto out_inval;
6793         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6794                 goto out_inval;
6795         return NFS_OK;
6796 out_inval:
6797         return -NFS4ERR_INVAL;
6798 }
6799
6800 static bool
6801 nfs41_same_server_scope(struct nfs41_server_scope *a,
6802                         struct nfs41_server_scope *b)
6803 {
6804         if (a->server_scope_sz == b->server_scope_sz &&
6805             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6806                 return true;
6807
6808         return false;
6809 }
6810
6811 static void
6812 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
6813 {
6814 }
6815
6816 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
6817         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
6818 };
6819
6820 /*
6821  * nfs4_proc_bind_one_conn_to_session()
6822  *
6823  * The 4.1 client currently uses the same TCP connection for the
6824  * fore and backchannel.
6825  */
6826 static
6827 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
6828                 struct rpc_xprt *xprt,
6829                 struct nfs_client *clp,
6830                 struct rpc_cred *cred)
6831 {
6832         int status;
6833         struct nfs41_bind_conn_to_session_args args = {
6834                 .client = clp,
6835                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6836         };
6837         struct nfs41_bind_conn_to_session_res res;
6838         struct rpc_message msg = {
6839                 .rpc_proc =
6840                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6841                 .rpc_argp = &args,
6842                 .rpc_resp = &res,
6843                 .rpc_cred = cred,
6844         };
6845         struct rpc_task_setup task_setup_data = {
6846                 .rpc_client = clnt,
6847                 .rpc_xprt = xprt,
6848                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
6849                 .rpc_message = &msg,
6850                 .flags = RPC_TASK_TIMEOUT,
6851         };
6852         struct rpc_task *task;
6853
6854         dprintk("--> %s\n", __func__);
6855
6856         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6857         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6858                 args.dir = NFS4_CDFC4_FORE;
6859
6860         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
6861         if (xprt != rcu_access_pointer(clnt->cl_xprt))
6862                 args.dir = NFS4_CDFC4_FORE;
6863
6864         task = rpc_run_task(&task_setup_data);
6865         if (!IS_ERR(task)) {
6866                 status = task->tk_status;
6867                 rpc_put_task(task);
6868         } else
6869                 status = PTR_ERR(task);
6870         trace_nfs4_bind_conn_to_session(clp, status);
6871         if (status == 0) {
6872                 if (memcmp(res.sessionid.data,
6873                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6874                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6875                         status = -EIO;
6876                         goto out;
6877                 }
6878                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6879                         dprintk("NFS: %s: Unexpected direction from server\n",
6880                                 __func__);
6881                         status = -EIO;
6882                         goto out;
6883                 }
6884                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6885                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6886                                 __func__);
6887                         status = -EIO;
6888                         goto out;
6889                 }
6890         }
6891 out:
6892         dprintk("<-- %s status= %d\n", __func__, status);
6893         return status;
6894 }
6895
6896 struct rpc_bind_conn_calldata {
6897         struct nfs_client *clp;
6898         struct rpc_cred *cred;
6899 };
6900
6901 static int
6902 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
6903                 struct rpc_xprt *xprt,
6904                 void *calldata)
6905 {
6906         struct rpc_bind_conn_calldata *p = calldata;
6907
6908         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
6909 }
6910
6911 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6912 {
6913         struct rpc_bind_conn_calldata data = {
6914                 .clp = clp,
6915                 .cred = cred,
6916         };
6917         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
6918                         nfs4_proc_bind_conn_to_session_callback, &data);
6919 }
6920
6921 /*
6922  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6923  * and operations we'd like to see to enable certain features in the allow map
6924  */
6925 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6926         .how = SP4_MACH_CRED,
6927         .enforce.u.words = {
6928                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6929                       1 << (OP_EXCHANGE_ID - 32) |
6930                       1 << (OP_CREATE_SESSION - 32) |
6931                       1 << (OP_DESTROY_SESSION - 32) |
6932                       1 << (OP_DESTROY_CLIENTID - 32)
6933         },
6934         .allow.u.words = {
6935                 [0] = 1 << (OP_CLOSE) |
6936                       1 << (OP_OPEN_DOWNGRADE) |
6937                       1 << (OP_LOCKU) |
6938                       1 << (OP_DELEGRETURN) |
6939                       1 << (OP_COMMIT),
6940                 [1] = 1 << (OP_SECINFO - 32) |
6941                       1 << (OP_SECINFO_NO_NAME - 32) |
6942                       1 << (OP_LAYOUTRETURN - 32) |
6943                       1 << (OP_TEST_STATEID - 32) |
6944                       1 << (OP_FREE_STATEID - 32) |
6945                       1 << (OP_WRITE - 32)
6946         }
6947 };
6948
6949 /*
6950  * Select the state protection mode for client `clp' given the server results
6951  * from exchange_id in `sp'.
6952  *
6953  * Returns 0 on success, negative errno otherwise.
6954  */
6955 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6956                                  struct nfs41_state_protection *sp)
6957 {
6958         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6959                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6960                       1 << (OP_EXCHANGE_ID - 32) |
6961                       1 << (OP_CREATE_SESSION - 32) |
6962                       1 << (OP_DESTROY_SESSION - 32) |
6963                       1 << (OP_DESTROY_CLIENTID - 32)
6964         };
6965         unsigned int i;
6966
6967         if (sp->how == SP4_MACH_CRED) {
6968                 /* Print state protect result */
6969                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6970                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6971                         if (test_bit(i, sp->enforce.u.longs))
6972                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6973                         if (test_bit(i, sp->allow.u.longs))
6974                                 dfprintk(MOUNT, "  allow op %d\n", i);
6975                 }
6976
6977                 /* make sure nothing is on enforce list that isn't supported */
6978                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6979                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6980                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6981                                 return -EINVAL;
6982                         }
6983                 }
6984
6985                 /*
6986                  * Minimal mode - state operations are allowed to use machine
6987                  * credential.  Note this already happens by default, so the
6988                  * client doesn't have to do anything more than the negotiation.
6989                  *
6990                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6991                  *       we're already using the machine cred for exchange_id
6992                  *       and will never use a different cred.
6993                  */
6994                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6995                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6996                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6997                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6998                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6999                         dfprintk(MOUNT, "  minimal mode enabled\n");
7000                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7001                 } else {
7002                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7003                         return -EINVAL;
7004                 }
7005
7006                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7007                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7008                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7009                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7010                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7011                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7012                 }
7013
7014                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7015                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7016                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7017                                 &clp->cl_sp4_flags);
7018                 }
7019
7020                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7021                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7022                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7023                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7024                 }
7025
7026                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7027                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7028                         dfprintk(MOUNT, "  stateid mode enabled\n");
7029                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7030                 }
7031
7032                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7033                         dfprintk(MOUNT, "  write mode enabled\n");
7034                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7035                 }
7036
7037                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7038                         dfprintk(MOUNT, "  commit mode enabled\n");
7039                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7040                 }
7041         }
7042
7043         return 0;
7044 }
7045
7046 /*
7047  * _nfs4_proc_exchange_id()
7048  *
7049  * Wrapper for EXCHANGE_ID operation.
7050  */
7051 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7052         u32 sp4_how)
7053 {
7054         nfs4_verifier verifier;
7055         struct nfs41_exchange_id_args args = {
7056                 .verifier = &verifier,
7057                 .client = clp,
7058 #ifdef CONFIG_NFS_V4_1_MIGRATION
7059                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7060                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
7061                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7062 #else
7063                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7064                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
7065 #endif
7066         };
7067         struct nfs41_exchange_id_res res = {
7068                 0
7069         };
7070         int status;
7071         struct rpc_message msg = {
7072                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7073                 .rpc_argp = &args,
7074                 .rpc_resp = &res,
7075                 .rpc_cred = cred,
7076         };
7077
7078         nfs4_init_boot_verifier(clp, &verifier);
7079
7080         status = nfs4_init_uniform_client_string(clp);
7081         if (status)
7082                 goto out;
7083
7084         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7085                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7086                 clp->cl_owner_id);
7087
7088         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7089                                         GFP_NOFS);
7090         if (unlikely(res.server_owner == NULL)) {
7091                 status = -ENOMEM;
7092                 goto out;
7093         }
7094
7095         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7096                                         GFP_NOFS);
7097         if (unlikely(res.server_scope == NULL)) {
7098                 status = -ENOMEM;
7099                 goto out_server_owner;
7100         }
7101
7102         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7103         if (unlikely(res.impl_id == NULL)) {
7104                 status = -ENOMEM;
7105                 goto out_server_scope;
7106         }
7107
7108         switch (sp4_how) {
7109         case SP4_NONE:
7110                 args.state_protect.how = SP4_NONE;
7111                 break;
7112
7113         case SP4_MACH_CRED:
7114                 args.state_protect = nfs4_sp4_mach_cred_request;
7115                 break;
7116
7117         default:
7118                 /* unsupported! */
7119                 WARN_ON_ONCE(1);
7120                 status = -EINVAL;
7121                 goto out_impl_id;
7122         }
7123
7124         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7125         trace_nfs4_exchange_id(clp, status);
7126         if (status == 0)
7127                 status = nfs4_check_cl_exchange_flags(res.flags);
7128
7129         if (status == 0)
7130                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7131
7132         if (status == 0) {
7133                 clp->cl_clientid = res.clientid;
7134                 clp->cl_exchange_flags = res.flags;
7135                 /* Client ID is not confirmed */
7136                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7137                         clear_bit(NFS4_SESSION_ESTABLISHED,
7138                                         &clp->cl_session->session_state);
7139                         clp->cl_seqid = res.seqid;
7140                 }
7141
7142                 kfree(clp->cl_serverowner);
7143                 clp->cl_serverowner = res.server_owner;
7144                 res.server_owner = NULL;
7145
7146                 /* use the most recent implementation id */
7147                 kfree(clp->cl_implid);
7148                 clp->cl_implid = res.impl_id;
7149                 res.impl_id = NULL;
7150
7151                 if (clp->cl_serverscope != NULL &&
7152                     !nfs41_same_server_scope(clp->cl_serverscope,
7153                                              res.server_scope)) {
7154                         dprintk("%s: server_scope mismatch detected\n",
7155                                 __func__);
7156                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7157                         kfree(clp->cl_serverscope);
7158                         clp->cl_serverscope = NULL;
7159                 }
7160
7161                 if (clp->cl_serverscope == NULL) {
7162                         clp->cl_serverscope = res.server_scope;
7163                         res.server_scope = NULL;
7164                 }
7165         }
7166
7167 out_impl_id:
7168         kfree(res.impl_id);
7169 out_server_scope:
7170         kfree(res.server_scope);
7171 out_server_owner:
7172         kfree(res.server_owner);
7173 out:
7174         if (clp->cl_implid != NULL)
7175                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7176                         "domain: %s, name: %s, date: %llu,%u\n",
7177                         clp->cl_implid->domain, clp->cl_implid->name,
7178                         clp->cl_implid->date.seconds,
7179                         clp->cl_implid->date.nseconds);
7180         dprintk("NFS reply exchange_id: %d\n", status);
7181         return status;
7182 }
7183
7184 /*
7185  * nfs4_proc_exchange_id()
7186  *
7187  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7188  *
7189  * Since the clientid has expired, all compounds using sessions
7190  * associated with the stale clientid will be returning
7191  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7192  * be in some phase of session reset.
7193  *
7194  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7195  */
7196 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7197 {
7198         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7199         int status;
7200
7201         /* try SP4_MACH_CRED if krb5i/p */
7202         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7203             authflavor == RPC_AUTH_GSS_KRB5P) {
7204                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7205                 if (!status)
7206                         return 0;
7207         }
7208
7209         /* try SP4_NONE */
7210         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7211 }
7212
7213 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7214                 struct rpc_cred *cred)
7215 {
7216         struct rpc_message msg = {
7217                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7218                 .rpc_argp = clp,
7219                 .rpc_cred = cred,
7220         };
7221         int status;
7222
7223         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7224         trace_nfs4_destroy_clientid(clp, status);
7225         if (status)
7226                 dprintk("NFS: Got error %d from the server %s on "
7227                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7228         return status;
7229 }
7230
7231 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7232                 struct rpc_cred *cred)
7233 {
7234         unsigned int loop;
7235         int ret;
7236
7237         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7238                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7239                 switch (ret) {
7240                 case -NFS4ERR_DELAY:
7241                 case -NFS4ERR_CLIENTID_BUSY:
7242                         ssleep(1);
7243                         break;
7244                 default:
7245                         return ret;
7246                 }
7247         }
7248         return 0;
7249 }
7250
7251 int nfs4_destroy_clientid(struct nfs_client *clp)
7252 {
7253         struct rpc_cred *cred;
7254         int ret = 0;
7255
7256         if (clp->cl_mvops->minor_version < 1)
7257                 goto out;
7258         if (clp->cl_exchange_flags == 0)
7259                 goto out;
7260         if (clp->cl_preserve_clid)
7261                 goto out;
7262         cred = nfs4_get_clid_cred(clp);
7263         ret = nfs4_proc_destroy_clientid(clp, cred);
7264         if (cred)
7265                 put_rpccred(cred);
7266         switch (ret) {
7267         case 0:
7268         case -NFS4ERR_STALE_CLIENTID:
7269                 clp->cl_exchange_flags = 0;
7270         }
7271 out:
7272         return ret;
7273 }
7274
7275 struct nfs4_get_lease_time_data {
7276         struct nfs4_get_lease_time_args *args;
7277         struct nfs4_get_lease_time_res *res;
7278         struct nfs_client *clp;
7279 };
7280
7281 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7282                                         void *calldata)
7283 {
7284         struct nfs4_get_lease_time_data *data =
7285                         (struct nfs4_get_lease_time_data *)calldata;
7286
7287         dprintk("--> %s\n", __func__);
7288         /* just setup sequence, do not trigger session recovery
7289            since we're invoked within one */
7290         nfs41_setup_sequence(data->clp->cl_session,
7291                         &data->args->la_seq_args,
7292                         &data->res->lr_seq_res,
7293                         task);
7294         dprintk("<-- %s\n", __func__);
7295 }
7296
7297 /*
7298  * Called from nfs4_state_manager thread for session setup, so don't recover
7299  * from sequence operation or clientid errors.
7300  */
7301 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7302 {
7303         struct nfs4_get_lease_time_data *data =
7304                         (struct nfs4_get_lease_time_data *)calldata;
7305
7306         dprintk("--> %s\n", __func__);
7307         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7308                 return;
7309         switch (task->tk_status) {
7310         case -NFS4ERR_DELAY:
7311         case -NFS4ERR_GRACE:
7312                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7313                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7314                 task->tk_status = 0;
7315                 /* fall through */
7316         case -NFS4ERR_RETRY_UNCACHED_REP:
7317                 rpc_restart_call_prepare(task);
7318                 return;
7319         }
7320         dprintk("<-- %s\n", __func__);
7321 }
7322
7323 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7324         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7325         .rpc_call_done = nfs4_get_lease_time_done,
7326 };
7327
7328 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7329 {
7330         struct rpc_task *task;
7331         struct nfs4_get_lease_time_args args;
7332         struct nfs4_get_lease_time_res res = {
7333                 .lr_fsinfo = fsinfo,
7334         };
7335         struct nfs4_get_lease_time_data data = {
7336                 .args = &args,
7337                 .res = &res,
7338                 .clp = clp,
7339         };
7340         struct rpc_message msg = {
7341                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7342                 .rpc_argp = &args,
7343                 .rpc_resp = &res,
7344         };
7345         struct rpc_task_setup task_setup = {
7346                 .rpc_client = clp->cl_rpcclient,
7347                 .rpc_message = &msg,
7348                 .callback_ops = &nfs4_get_lease_time_ops,
7349                 .callback_data = &data,
7350                 .flags = RPC_TASK_TIMEOUT,
7351         };
7352         int status;
7353
7354         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7355         nfs4_set_sequence_privileged(&args.la_seq_args);
7356         dprintk("--> %s\n", __func__);
7357         task = rpc_run_task(&task_setup);
7358
7359         if (IS_ERR(task))
7360                 status = PTR_ERR(task);
7361         else {
7362                 status = task->tk_status;
7363                 rpc_put_task(task);
7364         }
7365         dprintk("<-- %s return %d\n", __func__, status);
7366
7367         return status;
7368 }
7369
7370 /*
7371  * Initialize the values to be used by the client in CREATE_SESSION
7372  * If nfs4_init_session set the fore channel request and response sizes,
7373  * use them.
7374  *
7375  * Set the back channel max_resp_sz_cached to zero to force the client to
7376  * always set csa_cachethis to FALSE because the current implementation
7377  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7378  */
7379 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7380                                     struct rpc_clnt *clnt)
7381 {
7382         unsigned int max_rqst_sz, max_resp_sz;
7383         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7384
7385         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7386         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7387
7388         /* Fore channel attributes */
7389         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7390         args->fc_attrs.max_resp_sz = max_resp_sz;
7391         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7392         args->fc_attrs.max_reqs = max_session_slots;
7393
7394         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7395                 "max_ops=%u max_reqs=%u\n",
7396                 __func__,
7397                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7398                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7399
7400         /* Back channel attributes */
7401         args->bc_attrs.max_rqst_sz = max_bc_payload;
7402         args->bc_attrs.max_resp_sz = max_bc_payload;
7403         args->bc_attrs.max_resp_sz_cached = 0;
7404         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7405         args->bc_attrs.max_reqs = NFS41_BC_MAX_CALLBACKS;
7406
7407         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7408                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7409                 __func__,
7410                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7411                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7412                 args->bc_attrs.max_reqs);
7413 }
7414
7415 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7416                 struct nfs41_create_session_res *res)
7417 {
7418         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7419         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7420
7421         if (rcvd->max_resp_sz > sent->max_resp_sz)
7422                 return -EINVAL;
7423         /*
7424          * Our requested max_ops is the minimum we need; we're not
7425          * prepared to break up compounds into smaller pieces than that.
7426          * So, no point even trying to continue if the server won't
7427          * cooperate:
7428          */
7429         if (rcvd->max_ops < sent->max_ops)
7430                 return -EINVAL;
7431         if (rcvd->max_reqs == 0)
7432                 return -EINVAL;
7433         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7434                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7435         return 0;
7436 }
7437
7438 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7439                 struct nfs41_create_session_res *res)
7440 {
7441         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7442         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7443
7444         if (!(res->flags & SESSION4_BACK_CHAN))
7445                 goto out;
7446         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7447                 return -EINVAL;
7448         if (rcvd->max_resp_sz < sent->max_resp_sz)
7449                 return -EINVAL;
7450         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7451                 return -EINVAL;
7452         /* These would render the backchannel useless: */
7453         if (rcvd->max_ops != sent->max_ops)
7454                 return -EINVAL;
7455         if (rcvd->max_reqs != sent->max_reqs)
7456                 return -EINVAL;
7457 out:
7458         return 0;
7459 }
7460
7461 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7462                                      struct nfs41_create_session_res *res)
7463 {
7464         int ret;
7465
7466         ret = nfs4_verify_fore_channel_attrs(args, res);
7467         if (ret)
7468                 return ret;
7469         return nfs4_verify_back_channel_attrs(args, res);
7470 }
7471
7472 static void nfs4_update_session(struct nfs4_session *session,
7473                 struct nfs41_create_session_res *res)
7474 {
7475         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7476         /* Mark client id and session as being confirmed */
7477         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7478         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7479         session->flags = res->flags;
7480         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7481         if (res->flags & SESSION4_BACK_CHAN)
7482                 memcpy(&session->bc_attrs, &res->bc_attrs,
7483                                 sizeof(session->bc_attrs));
7484 }
7485
7486 static int _nfs4_proc_create_session(struct nfs_client *clp,
7487                 struct rpc_cred *cred)
7488 {
7489         struct nfs4_session *session = clp->cl_session;
7490         struct nfs41_create_session_args args = {
7491                 .client = clp,
7492                 .clientid = clp->cl_clientid,
7493                 .seqid = clp->cl_seqid,
7494                 .cb_program = NFS4_CALLBACK,
7495         };
7496         struct nfs41_create_session_res res;
7497
7498         struct rpc_message msg = {
7499                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7500                 .rpc_argp = &args,
7501                 .rpc_resp = &res,
7502                 .rpc_cred = cred,
7503         };
7504         int status;
7505
7506         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7507         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7508
7509         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7510         trace_nfs4_create_session(clp, status);
7511
7512         if (!status) {
7513                 /* Verify the session's negotiated channel_attrs values */
7514                 status = nfs4_verify_channel_attrs(&args, &res);
7515                 /* Increment the clientid slot sequence id */
7516                 if (clp->cl_seqid == res.seqid)
7517                         clp->cl_seqid++;
7518                 if (status)
7519                         goto out;
7520                 nfs4_update_session(session, &res);
7521         }
7522 out:
7523         return status;
7524 }
7525
7526 /*
7527  * Issues a CREATE_SESSION operation to the server.
7528  * It is the responsibility of the caller to verify the session is
7529  * expired before calling this routine.
7530  */
7531 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7532 {
7533         int status;
7534         unsigned *ptr;
7535         struct nfs4_session *session = clp->cl_session;
7536
7537         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7538
7539         status = _nfs4_proc_create_session(clp, cred);
7540         if (status)
7541                 goto out;
7542
7543         /* Init or reset the session slot tables */
7544         status = nfs4_setup_session_slot_tables(session);
7545         dprintk("slot table setup returned %d\n", status);
7546         if (status)
7547                 goto out;
7548
7549         ptr = (unsigned *)&session->sess_id.data[0];
7550         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7551                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7552 out:
7553         dprintk("<-- %s\n", __func__);
7554         return status;
7555 }
7556
7557 /*
7558  * Issue the over-the-wire RPC DESTROY_SESSION.
7559  * The caller must serialize access to this routine.
7560  */
7561 int nfs4_proc_destroy_session(struct nfs4_session *session,
7562                 struct rpc_cred *cred)
7563 {
7564         struct rpc_message msg = {
7565                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7566                 .rpc_argp = session,
7567                 .rpc_cred = cred,
7568         };
7569         int status = 0;
7570
7571         dprintk("--> nfs4_proc_destroy_session\n");
7572
7573         /* session is still being setup */
7574         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7575                 return 0;
7576
7577         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7578         trace_nfs4_destroy_session(session->clp, status);
7579
7580         if (status)
7581                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7582                         "Session has been destroyed regardless...\n", status);
7583
7584         dprintk("<-- nfs4_proc_destroy_session\n");
7585         return status;
7586 }
7587
7588 /*
7589  * Renew the cl_session lease.
7590  */
7591 struct nfs4_sequence_data {
7592         struct nfs_client *clp;
7593         struct nfs4_sequence_args args;
7594         struct nfs4_sequence_res res;
7595 };
7596
7597 static void nfs41_sequence_release(void *data)
7598 {
7599         struct nfs4_sequence_data *calldata = data;
7600         struct nfs_client *clp = calldata->clp;
7601
7602         if (atomic_read(&clp->cl_count) > 1)
7603                 nfs4_schedule_state_renewal(clp);
7604         nfs_put_client(clp);
7605         kfree(calldata);
7606 }
7607
7608 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7609 {
7610         switch(task->tk_status) {
7611         case -NFS4ERR_DELAY:
7612                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7613                 return -EAGAIN;
7614         default:
7615                 nfs4_schedule_lease_recovery(clp);
7616         }
7617         return 0;
7618 }
7619
7620 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7621 {
7622         struct nfs4_sequence_data *calldata = data;
7623         struct nfs_client *clp = calldata->clp;
7624
7625         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7626                 return;
7627
7628         trace_nfs4_sequence(clp, task->tk_status);
7629         if (task->tk_status < 0) {
7630                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7631                 if (atomic_read(&clp->cl_count) == 1)
7632                         goto out;
7633
7634                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7635                         rpc_restart_call_prepare(task);
7636                         return;
7637                 }
7638         }
7639         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7640 out:
7641         dprintk("<-- %s\n", __func__);
7642 }
7643
7644 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7645 {
7646         struct nfs4_sequence_data *calldata = data;
7647         struct nfs_client *clp = calldata->clp;
7648         struct nfs4_sequence_args *args;
7649         struct nfs4_sequence_res *res;
7650
7651         args = task->tk_msg.rpc_argp;
7652         res = task->tk_msg.rpc_resp;
7653
7654         nfs41_setup_sequence(clp->cl_session, args, res, task);
7655 }
7656
7657 static const struct rpc_call_ops nfs41_sequence_ops = {
7658         .rpc_call_done = nfs41_sequence_call_done,
7659         .rpc_call_prepare = nfs41_sequence_prepare,
7660         .rpc_release = nfs41_sequence_release,
7661 };
7662
7663 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7664                 struct rpc_cred *cred,
7665                 bool is_privileged)
7666 {
7667         struct nfs4_sequence_data *calldata;
7668         struct rpc_message msg = {
7669                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7670                 .rpc_cred = cred,
7671         };
7672         struct rpc_task_setup task_setup_data = {
7673                 .rpc_client = clp->cl_rpcclient,
7674                 .rpc_message = &msg,
7675                 .callback_ops = &nfs41_sequence_ops,
7676                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7677         };
7678
7679         if (!atomic_inc_not_zero(&clp->cl_count))
7680                 return ERR_PTR(-EIO);
7681         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7682         if (calldata == NULL) {
7683                 nfs_put_client(clp);
7684                 return ERR_PTR(-ENOMEM);
7685         }
7686         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7687         if (is_privileged)
7688                 nfs4_set_sequence_privileged(&calldata->args);
7689         msg.rpc_argp = &calldata->args;
7690         msg.rpc_resp = &calldata->res;
7691         calldata->clp = clp;
7692         task_setup_data.callback_data = calldata;
7693
7694         return rpc_run_task(&task_setup_data);
7695 }
7696
7697 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7698 {
7699         struct rpc_task *task;
7700         int ret = 0;
7701
7702         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7703                 return -EAGAIN;
7704         task = _nfs41_proc_sequence(clp, cred, false);
7705         if (IS_ERR(task))
7706                 ret = PTR_ERR(task);
7707         else
7708                 rpc_put_task_async(task);
7709         dprintk("<-- %s status=%d\n", __func__, ret);
7710         return ret;
7711 }
7712
7713 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7714 {
7715         struct rpc_task *task;
7716         int ret;
7717
7718         task = _nfs41_proc_sequence(clp, cred, true);
7719         if (IS_ERR(task)) {
7720                 ret = PTR_ERR(task);
7721                 goto out;
7722         }
7723         ret = rpc_wait_for_completion_task(task);
7724         if (!ret)
7725                 ret = task->tk_status;
7726         rpc_put_task(task);
7727 out:
7728         dprintk("<-- %s status=%d\n", __func__, ret);
7729         return ret;
7730 }
7731
7732 struct nfs4_reclaim_complete_data {
7733         struct nfs_client *clp;
7734         struct nfs41_reclaim_complete_args arg;
7735         struct nfs41_reclaim_complete_res res;
7736 };
7737
7738 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7739 {
7740         struct nfs4_reclaim_complete_data *calldata = data;
7741
7742         nfs41_setup_sequence(calldata->clp->cl_session,
7743                         &calldata->arg.seq_args,
7744                         &calldata->res.seq_res,
7745                         task);
7746 }
7747
7748 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7749 {
7750         switch(task->tk_status) {
7751         case 0:
7752         case -NFS4ERR_COMPLETE_ALREADY:
7753         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7754                 break;
7755         case -NFS4ERR_DELAY:
7756                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7757                 /* fall through */
7758         case -NFS4ERR_RETRY_UNCACHED_REP:
7759                 return -EAGAIN;
7760         default:
7761                 nfs4_schedule_lease_recovery(clp);
7762         }
7763         return 0;
7764 }
7765
7766 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7767 {
7768         struct nfs4_reclaim_complete_data *calldata = data;
7769         struct nfs_client *clp = calldata->clp;
7770         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7771
7772         dprintk("--> %s\n", __func__);
7773         if (!nfs41_sequence_done(task, res))
7774                 return;
7775
7776         trace_nfs4_reclaim_complete(clp, task->tk_status);
7777         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7778                 rpc_restart_call_prepare(task);
7779                 return;
7780         }
7781         dprintk("<-- %s\n", __func__);
7782 }
7783
7784 static void nfs4_free_reclaim_complete_data(void *data)
7785 {
7786         struct nfs4_reclaim_complete_data *calldata = data;
7787
7788         kfree(calldata);
7789 }
7790
7791 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7792         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7793         .rpc_call_done = nfs4_reclaim_complete_done,
7794         .rpc_release = nfs4_free_reclaim_complete_data,
7795 };
7796
7797 /*
7798  * Issue a global reclaim complete.
7799  */
7800 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7801                 struct rpc_cred *cred)
7802 {
7803         struct nfs4_reclaim_complete_data *calldata;
7804         struct rpc_task *task;
7805         struct rpc_message msg = {
7806                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7807                 .rpc_cred = cred,
7808         };
7809         struct rpc_task_setup task_setup_data = {
7810                 .rpc_client = clp->cl_rpcclient,
7811                 .rpc_message = &msg,
7812                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7813                 .flags = RPC_TASK_ASYNC,
7814         };
7815         int status = -ENOMEM;
7816
7817         dprintk("--> %s\n", __func__);
7818         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7819         if (calldata == NULL)
7820                 goto out;
7821         calldata->clp = clp;
7822         calldata->arg.one_fs = 0;
7823
7824         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7825         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7826         msg.rpc_argp = &calldata->arg;
7827         msg.rpc_resp = &calldata->res;
7828         task_setup_data.callback_data = calldata;
7829         task = rpc_run_task(&task_setup_data);
7830         if (IS_ERR(task)) {
7831                 status = PTR_ERR(task);
7832                 goto out;
7833         }
7834         status = nfs4_wait_for_completion_rpc_task(task);
7835         if (status == 0)
7836                 status = task->tk_status;
7837         rpc_put_task(task);
7838         return 0;
7839 out:
7840         dprintk("<-- %s status=%d\n", __func__, status);
7841         return status;
7842 }
7843
7844 static void
7845 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7846 {
7847         struct nfs4_layoutget *lgp = calldata;
7848         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7849         struct nfs4_session *session = nfs4_get_session(server);
7850         int ret;
7851
7852         dprintk("--> %s\n", __func__);
7853         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7854          * right now covering the LAYOUTGET we are about to send.
7855          * However, that is not so catastrophic, and there seems
7856          * to be no way to prevent it completely.
7857          */
7858         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7859                                 &lgp->res.seq_res, task))
7860                 return;
7861         ret = pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7862                                           NFS_I(lgp->args.inode)->layout,
7863                                           &lgp->args.range,
7864                                           lgp->args.ctx->state);
7865         if (ret < 0)
7866                 rpc_exit(task, ret);
7867 }
7868
7869 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7870 {
7871         struct nfs4_layoutget *lgp = calldata;
7872         struct inode *inode = lgp->args.inode;
7873         struct nfs_server *server = NFS_SERVER(inode);
7874         struct pnfs_layout_hdr *lo;
7875         struct nfs4_state *state = NULL;
7876         unsigned long timeo, now, giveup;
7877
7878         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7879
7880         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7881                 goto out;
7882
7883         switch (task->tk_status) {
7884         case 0:
7885                 goto out;
7886
7887         /*
7888          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
7889          * on the file. set tk_status to -ENODATA to tell upper layer to
7890          * retry go inband.
7891          */
7892         case -NFS4ERR_LAYOUTUNAVAILABLE:
7893                 task->tk_status = -ENODATA;
7894                 goto out;
7895         /*
7896          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7897          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7898          */
7899         case -NFS4ERR_BADLAYOUT:
7900                 goto out_overflow;
7901         /*
7902          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7903          * (or clients) writing to the same RAID stripe except when
7904          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7905          */
7906         case -NFS4ERR_LAYOUTTRYLATER:
7907                 if (lgp->args.minlength == 0)
7908                         goto out_overflow;
7909         /*
7910          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7911          * existing layout before getting a new one).
7912          */
7913         case -NFS4ERR_RECALLCONFLICT:
7914                 timeo = rpc_get_timeout(task->tk_client);
7915                 giveup = lgp->args.timestamp + timeo;
7916                 now = jiffies;
7917                 if (time_after(giveup, now)) {
7918                         unsigned long delay;
7919
7920                         /* Delay for:
7921                          * - Not less then NFS4_POLL_RETRY_MIN.
7922                          * - One last time a jiffie before we give up
7923                          * - exponential backoff (time_now minus start_attempt)
7924                          */
7925                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7926                                     min((giveup - now - 1),
7927                                         now - lgp->args.timestamp));
7928
7929                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7930                                 __func__, delay);
7931                         rpc_delay(task, delay);
7932                         /* Do not call nfs4_async_handle_error() */
7933                         goto out_restart;
7934                 }
7935                 break;
7936         case -NFS4ERR_EXPIRED:
7937         case -NFS4ERR_BAD_STATEID:
7938                 spin_lock(&inode->i_lock);
7939                 if (nfs4_stateid_match(&lgp->args.stateid,
7940                                         &lgp->args.ctx->state->stateid)) {
7941                         spin_unlock(&inode->i_lock);
7942                         /* If the open stateid was bad, then recover it. */
7943                         state = lgp->args.ctx->state;
7944                         break;
7945                 }
7946                 lo = NFS_I(inode)->layout;
7947                 if (lo && nfs4_stateid_match(&lgp->args.stateid,
7948                                         &lo->plh_stateid)) {
7949                         LIST_HEAD(head);
7950
7951                         /*
7952                          * Mark the bad layout state as invalid, then retry
7953                          * with the current stateid.
7954                          */
7955                         set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7956                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL, 0);
7957                         spin_unlock(&inode->i_lock);
7958                         pnfs_free_lseg_list(&head);
7959                 } else
7960                         spin_unlock(&inode->i_lock);
7961                 goto out_restart;
7962         }
7963         if (nfs4_async_handle_error(task, server, state, &lgp->timeout) == -EAGAIN)
7964                 goto out_restart;
7965 out:
7966         dprintk("<-- %s\n", __func__);
7967         return;
7968 out_restart:
7969         task->tk_status = 0;
7970         rpc_restart_call_prepare(task);
7971         return;
7972 out_overflow:
7973         task->tk_status = -EOVERFLOW;
7974         goto out;
7975 }
7976
7977 static size_t max_response_pages(struct nfs_server *server)
7978 {
7979         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7980         return nfs_page_array_len(0, max_resp_sz);
7981 }
7982
7983 static void nfs4_free_pages(struct page **pages, size_t size)
7984 {
7985         int i;
7986
7987         if (!pages)
7988                 return;
7989
7990         for (i = 0; i < size; i++) {
7991                 if (!pages[i])
7992                         break;
7993                 __free_page(pages[i]);
7994         }
7995         kfree(pages);
7996 }
7997
7998 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7999 {
8000         struct page **pages;
8001         int i;
8002
8003         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8004         if (!pages) {
8005                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8006                 return NULL;
8007         }
8008
8009         for (i = 0; i < size; i++) {
8010                 pages[i] = alloc_page(gfp_flags);
8011                 if (!pages[i]) {
8012                         dprintk("%s: failed to allocate page\n", __func__);
8013                         nfs4_free_pages(pages, size);
8014                         return NULL;
8015                 }
8016         }
8017
8018         return pages;
8019 }
8020
8021 static void nfs4_layoutget_release(void *calldata)
8022 {
8023         struct nfs4_layoutget *lgp = calldata;
8024         struct inode *inode = lgp->args.inode;
8025         struct nfs_server *server = NFS_SERVER(inode);
8026         size_t max_pages = max_response_pages(server);
8027
8028         dprintk("--> %s\n", __func__);
8029         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8030         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8031         put_nfs_open_context(lgp->args.ctx);
8032         kfree(calldata);
8033         dprintk("<-- %s\n", __func__);
8034 }
8035
8036 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8037         .rpc_call_prepare = nfs4_layoutget_prepare,
8038         .rpc_call_done = nfs4_layoutget_done,
8039         .rpc_release = nfs4_layoutget_release,
8040 };
8041
8042 struct pnfs_layout_segment *
8043 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
8044 {
8045         struct inode *inode = lgp->args.inode;
8046         struct nfs_server *server = NFS_SERVER(inode);
8047         size_t max_pages = max_response_pages(server);
8048         struct rpc_task *task;
8049         struct rpc_message msg = {
8050                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8051                 .rpc_argp = &lgp->args,
8052                 .rpc_resp = &lgp->res,
8053                 .rpc_cred = lgp->cred,
8054         };
8055         struct rpc_task_setup task_setup_data = {
8056                 .rpc_client = server->client,
8057                 .rpc_message = &msg,
8058                 .callback_ops = &nfs4_layoutget_call_ops,
8059                 .callback_data = lgp,
8060                 .flags = RPC_TASK_ASYNC,
8061         };
8062         struct pnfs_layout_segment *lseg = NULL;
8063         int status = 0;
8064
8065         dprintk("--> %s\n", __func__);
8066
8067         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8068         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8069
8070         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8071         if (!lgp->args.layout.pages) {
8072                 nfs4_layoutget_release(lgp);
8073                 return ERR_PTR(-ENOMEM);
8074         }
8075         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8076         lgp->args.timestamp = jiffies;
8077
8078         lgp->res.layoutp = &lgp->args.layout;
8079         lgp->res.seq_res.sr_slot = NULL;
8080         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8081
8082         task = rpc_run_task(&task_setup_data);
8083         if (IS_ERR(task))
8084                 return ERR_CAST(task);
8085         status = nfs4_wait_for_completion_rpc_task(task);
8086         if (status == 0)
8087                 status = task->tk_status;
8088         trace_nfs4_layoutget(lgp->args.ctx,
8089                         &lgp->args.range,
8090                         &lgp->res.range,
8091                         &lgp->res.stateid,
8092                         status);
8093         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8094         if (status == 0 && lgp->res.layoutp->len)
8095                 lseg = pnfs_layout_process(lgp);
8096         rpc_put_task(task);
8097         dprintk("<-- %s status=%d\n", __func__, status);
8098         if (status)
8099                 return ERR_PTR(status);
8100         return lseg;
8101 }
8102
8103 static void
8104 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8105 {
8106         struct nfs4_layoutreturn *lrp = calldata;
8107
8108         dprintk("--> %s\n", __func__);
8109         nfs41_setup_sequence(lrp->clp->cl_session,
8110                         &lrp->args.seq_args,
8111                         &lrp->res.seq_res,
8112                         task);
8113 }
8114
8115 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8116 {
8117         struct nfs4_layoutreturn *lrp = calldata;
8118         struct nfs_server *server;
8119
8120         dprintk("--> %s\n", __func__);
8121
8122         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8123                 return;
8124
8125         server = NFS_SERVER(lrp->args.inode);
8126         switch (task->tk_status) {
8127         default:
8128                 task->tk_status = 0;
8129         case 0:
8130                 break;
8131         case -NFS4ERR_DELAY:
8132                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8133                         break;
8134                 rpc_restart_call_prepare(task);
8135                 return;
8136         }
8137         dprintk("<-- %s\n", __func__);
8138 }
8139
8140 static void nfs4_layoutreturn_release(void *calldata)
8141 {
8142         struct nfs4_layoutreturn *lrp = calldata;
8143         struct pnfs_layout_hdr *lo = lrp->args.layout;
8144         LIST_HEAD(freeme);
8145
8146         dprintk("--> %s\n", __func__);
8147         spin_lock(&lo->plh_inode->i_lock);
8148         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range,
8149                         be32_to_cpu(lrp->args.stateid.seqid));
8150         pnfs_mark_layout_returned_if_empty(lo);
8151         if (lrp->res.lrs_present)
8152                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8153         pnfs_clear_layoutreturn_waitbit(lo);
8154         spin_unlock(&lo->plh_inode->i_lock);
8155         pnfs_free_lseg_list(&freeme);
8156         pnfs_put_layout_hdr(lrp->args.layout);
8157         nfs_iput_and_deactive(lrp->inode);
8158         kfree(calldata);
8159         dprintk("<-- %s\n", __func__);
8160 }
8161
8162 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8163         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8164         .rpc_call_done = nfs4_layoutreturn_done,
8165         .rpc_release = nfs4_layoutreturn_release,
8166 };
8167
8168 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8169 {
8170         struct rpc_task *task;
8171         struct rpc_message msg = {
8172                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8173                 .rpc_argp = &lrp->args,
8174                 .rpc_resp = &lrp->res,
8175                 .rpc_cred = lrp->cred,
8176         };
8177         struct rpc_task_setup task_setup_data = {
8178                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8179                 .rpc_message = &msg,
8180                 .callback_ops = &nfs4_layoutreturn_call_ops,
8181                 .callback_data = lrp,
8182         };
8183         int status = 0;
8184
8185         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8186                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8187                         &task_setup_data.rpc_client, &msg);
8188
8189         dprintk("--> %s\n", __func__);
8190         if (!sync) {
8191                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8192                 if (!lrp->inode) {
8193                         nfs4_layoutreturn_release(lrp);
8194                         return -EAGAIN;
8195                 }
8196                 task_setup_data.flags |= RPC_TASK_ASYNC;
8197         }
8198         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8199         task = rpc_run_task(&task_setup_data);
8200         if (IS_ERR(task))
8201                 return PTR_ERR(task);
8202         if (sync)
8203                 status = task->tk_status;
8204         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8205         dprintk("<-- %s status=%d\n", __func__, status);
8206         rpc_put_task(task);
8207         return status;
8208 }
8209
8210 static int
8211 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8212                 struct pnfs_device *pdev,
8213                 struct rpc_cred *cred)
8214 {
8215         struct nfs4_getdeviceinfo_args args = {
8216                 .pdev = pdev,
8217                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8218                         NOTIFY_DEVICEID4_DELETE,
8219         };
8220         struct nfs4_getdeviceinfo_res res = {
8221                 .pdev = pdev,
8222         };
8223         struct rpc_message msg = {
8224                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8225                 .rpc_argp = &args,
8226                 .rpc_resp = &res,
8227                 .rpc_cred = cred,
8228         };
8229         int status;
8230
8231         dprintk("--> %s\n", __func__);
8232         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8233         if (res.notification & ~args.notify_types)
8234                 dprintk("%s: unsupported notification\n", __func__);
8235         if (res.notification != args.notify_types)
8236                 pdev->nocache = 1;
8237
8238         dprintk("<-- %s status=%d\n", __func__, status);
8239
8240         return status;
8241 }
8242
8243 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8244                 struct pnfs_device *pdev,
8245                 struct rpc_cred *cred)
8246 {
8247         struct nfs4_exception exception = { };
8248         int err;
8249
8250         do {
8251                 err = nfs4_handle_exception(server,
8252                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8253                                         &exception);
8254         } while (exception.retry);
8255         return err;
8256 }
8257 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8258
8259 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8260 {
8261         struct nfs4_layoutcommit_data *data = calldata;
8262         struct nfs_server *server = NFS_SERVER(data->args.inode);
8263         struct nfs4_session *session = nfs4_get_session(server);
8264
8265         nfs41_setup_sequence(session,
8266                         &data->args.seq_args,
8267                         &data->res.seq_res,
8268                         task);
8269 }
8270
8271 static void
8272 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8273 {
8274         struct nfs4_layoutcommit_data *data = calldata;
8275         struct nfs_server *server = NFS_SERVER(data->args.inode);
8276
8277         if (!nfs41_sequence_done(task, &data->res.seq_res))
8278                 return;
8279
8280         switch (task->tk_status) { /* Just ignore these failures */
8281         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8282         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8283         case -NFS4ERR_BADLAYOUT:     /* no layout */
8284         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8285                 task->tk_status = 0;
8286         case 0:
8287                 break;
8288         default:
8289                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8290                         rpc_restart_call_prepare(task);
8291                         return;
8292                 }
8293         }
8294 }
8295
8296 static void nfs4_layoutcommit_release(void *calldata)
8297 {
8298         struct nfs4_layoutcommit_data *data = calldata;
8299
8300         pnfs_cleanup_layoutcommit(data);
8301         nfs_post_op_update_inode_force_wcc(data->args.inode,
8302                                            data->res.fattr);
8303         put_rpccred(data->cred);
8304         nfs_iput_and_deactive(data->inode);
8305         kfree(data);
8306 }
8307
8308 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8309         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8310         .rpc_call_done = nfs4_layoutcommit_done,
8311         .rpc_release = nfs4_layoutcommit_release,
8312 };
8313
8314 int
8315 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8316 {
8317         struct rpc_message msg = {
8318                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8319                 .rpc_argp = &data->args,
8320                 .rpc_resp = &data->res,
8321                 .rpc_cred = data->cred,
8322         };
8323         struct rpc_task_setup task_setup_data = {
8324                 .task = &data->task,
8325                 .rpc_client = NFS_CLIENT(data->args.inode),
8326                 .rpc_message = &msg,
8327                 .callback_ops = &nfs4_layoutcommit_ops,
8328                 .callback_data = data,
8329         };
8330         struct rpc_task *task;
8331         int status = 0;
8332
8333         dprintk("NFS: initiating layoutcommit call. sync %d "
8334                 "lbw: %llu inode %lu\n", sync,
8335                 data->args.lastbytewritten,
8336                 data->args.inode->i_ino);
8337
8338         if (!sync) {
8339                 data->inode = nfs_igrab_and_active(data->args.inode);
8340                 if (data->inode == NULL) {
8341                         nfs4_layoutcommit_release(data);
8342                         return -EAGAIN;
8343                 }
8344                 task_setup_data.flags = RPC_TASK_ASYNC;
8345         }
8346         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8347         task = rpc_run_task(&task_setup_data);
8348         if (IS_ERR(task))
8349                 return PTR_ERR(task);
8350         if (sync)
8351                 status = task->tk_status;
8352         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8353         dprintk("%s: status %d\n", __func__, status);
8354         rpc_put_task(task);
8355         return status;
8356 }
8357
8358 /**
8359  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8360  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8361  */
8362 static int
8363 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8364                     struct nfs_fsinfo *info,
8365                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8366 {
8367         struct nfs41_secinfo_no_name_args args = {
8368                 .style = SECINFO_STYLE_CURRENT_FH,
8369         };
8370         struct nfs4_secinfo_res res = {
8371                 .flavors = flavors,
8372         };
8373         struct rpc_message msg = {
8374                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8375                 .rpc_argp = &args,
8376                 .rpc_resp = &res,
8377         };
8378         struct rpc_clnt *clnt = server->client;
8379         struct rpc_cred *cred = NULL;
8380         int status;
8381
8382         if (use_integrity) {
8383                 clnt = server->nfs_client->cl_rpcclient;
8384                 cred = nfs4_get_clid_cred(server->nfs_client);
8385                 msg.rpc_cred = cred;
8386         }
8387
8388         dprintk("--> %s\n", __func__);
8389         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8390                                 &res.seq_res, 0);
8391         dprintk("<-- %s status=%d\n", __func__, status);
8392
8393         if (cred)
8394                 put_rpccred(cred);
8395
8396         return status;
8397 }
8398
8399 static int
8400 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8401                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8402 {
8403         struct nfs4_exception exception = { };
8404         int err;
8405         do {
8406                 /* first try using integrity protection */
8407                 err = -NFS4ERR_WRONGSEC;
8408
8409                 /* try to use integrity protection with machine cred */
8410                 if (_nfs4_is_integrity_protected(server->nfs_client))
8411                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8412                                                           flavors, true);
8413
8414                 /*
8415                  * if unable to use integrity protection, or SECINFO with
8416                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8417                  * disallowed by spec, but exists in deployed servers) use
8418                  * the current filesystem's rpc_client and the user cred.
8419                  */
8420                 if (err == -NFS4ERR_WRONGSEC)
8421                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8422                                                           flavors, false);
8423
8424                 switch (err) {
8425                 case 0:
8426                 case -NFS4ERR_WRONGSEC:
8427                 case -ENOTSUPP:
8428                         goto out;
8429                 default:
8430                         err = nfs4_handle_exception(server, err, &exception);
8431                 }
8432         } while (exception.retry);
8433 out:
8434         return err;
8435 }
8436
8437 static int
8438 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8439                     struct nfs_fsinfo *info)
8440 {
8441         int err;
8442         struct page *page;
8443         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8444         struct nfs4_secinfo_flavors *flavors;
8445         struct nfs4_secinfo4 *secinfo;
8446         int i;
8447
8448         page = alloc_page(GFP_KERNEL);
8449         if (!page) {
8450                 err = -ENOMEM;
8451                 goto out;
8452         }
8453
8454         flavors = page_address(page);
8455         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8456
8457         /*
8458          * Fall back on "guess and check" method if
8459          * the server doesn't support SECINFO_NO_NAME
8460          */
8461         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8462                 err = nfs4_find_root_sec(server, fhandle, info);
8463                 goto out_freepage;
8464         }
8465         if (err)
8466                 goto out_freepage;
8467
8468         for (i = 0; i < flavors->num_flavors; i++) {
8469                 secinfo = &flavors->flavors[i];
8470
8471                 switch (secinfo->flavor) {
8472                 case RPC_AUTH_NULL:
8473                 case RPC_AUTH_UNIX:
8474                 case RPC_AUTH_GSS:
8475                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8476                                         &secinfo->flavor_info);
8477                         break;
8478                 default:
8479                         flavor = RPC_AUTH_MAXFLAVOR;
8480                         break;
8481                 }
8482
8483                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8484                         flavor = RPC_AUTH_MAXFLAVOR;
8485
8486                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8487                         err = nfs4_lookup_root_sec(server, fhandle,
8488                                                    info, flavor);
8489                         if (!err)
8490                                 break;
8491                 }
8492         }
8493
8494         if (flavor == RPC_AUTH_MAXFLAVOR)
8495                 err = -EPERM;
8496
8497 out_freepage:
8498         put_page(page);
8499         if (err == -EACCES)
8500                 return -EPERM;
8501 out:
8502         return err;
8503 }
8504
8505 static int _nfs41_test_stateid(struct nfs_server *server,
8506                 nfs4_stateid *stateid,
8507                 struct rpc_cred *cred)
8508 {
8509         int status;
8510         struct nfs41_test_stateid_args args = {
8511                 .stateid = stateid,
8512         };
8513         struct nfs41_test_stateid_res res;
8514         struct rpc_message msg = {
8515                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8516                 .rpc_argp = &args,
8517                 .rpc_resp = &res,
8518                 .rpc_cred = cred,
8519         };
8520         struct rpc_clnt *rpc_client = server->client;
8521
8522         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8523                 &rpc_client, &msg);
8524
8525         dprintk("NFS call  test_stateid %p\n", stateid);
8526         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8527         nfs4_set_sequence_privileged(&args.seq_args);
8528         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8529                         &args.seq_args, &res.seq_res);
8530         if (status != NFS_OK) {
8531                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8532                 return status;
8533         }
8534         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8535         return -res.status;
8536 }
8537
8538 /**
8539  * nfs41_test_stateid - perform a TEST_STATEID operation
8540  *
8541  * @server: server / transport on which to perform the operation
8542  * @stateid: state ID to test
8543  * @cred: credential
8544  *
8545  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8546  * Otherwise a negative NFS4ERR value is returned if the operation
8547  * failed or the state ID is not currently valid.
8548  */
8549 static int nfs41_test_stateid(struct nfs_server *server,
8550                 nfs4_stateid *stateid,
8551                 struct rpc_cred *cred)
8552 {
8553         struct nfs4_exception exception = { };
8554         int err;
8555         do {
8556                 err = _nfs41_test_stateid(server, stateid, cred);
8557                 if (err != -NFS4ERR_DELAY)
8558                         break;
8559                 nfs4_handle_exception(server, err, &exception);
8560         } while (exception.retry);
8561         return err;
8562 }
8563
8564 struct nfs_free_stateid_data {
8565         struct nfs_server *server;
8566         struct nfs41_free_stateid_args args;
8567         struct nfs41_free_stateid_res res;
8568 };
8569
8570 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8571 {
8572         struct nfs_free_stateid_data *data = calldata;
8573         nfs41_setup_sequence(nfs4_get_session(data->server),
8574                         &data->args.seq_args,
8575                         &data->res.seq_res,
8576                         task);
8577 }
8578
8579 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8580 {
8581         struct nfs_free_stateid_data *data = calldata;
8582
8583         nfs41_sequence_done(task, &data->res.seq_res);
8584
8585         switch (task->tk_status) {
8586         case -NFS4ERR_DELAY:
8587                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8588                         rpc_restart_call_prepare(task);
8589         }
8590 }
8591
8592 static void nfs41_free_stateid_release(void *calldata)
8593 {
8594         kfree(calldata);
8595 }
8596
8597 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8598         .rpc_call_prepare = nfs41_free_stateid_prepare,
8599         .rpc_call_done = nfs41_free_stateid_done,
8600         .rpc_release = nfs41_free_stateid_release,
8601 };
8602
8603 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8604                 nfs4_stateid *stateid,
8605                 struct rpc_cred *cred,
8606                 bool privileged)
8607 {
8608         struct rpc_message msg = {
8609                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8610                 .rpc_cred = cred,
8611         };
8612         struct rpc_task_setup task_setup = {
8613                 .rpc_client = server->client,
8614                 .rpc_message = &msg,
8615                 .callback_ops = &nfs41_free_stateid_ops,
8616                 .flags = RPC_TASK_ASYNC,
8617         };
8618         struct nfs_free_stateid_data *data;
8619
8620         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8621                 &task_setup.rpc_client, &msg);
8622
8623         dprintk("NFS call  free_stateid %p\n", stateid);
8624         data = kmalloc(sizeof(*data), GFP_NOFS);
8625         if (!data)
8626                 return ERR_PTR(-ENOMEM);
8627         data->server = server;
8628         nfs4_stateid_copy(&data->args.stateid, stateid);
8629
8630         task_setup.callback_data = data;
8631
8632         msg.rpc_argp = &data->args;
8633         msg.rpc_resp = &data->res;
8634         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8635         if (privileged)
8636                 nfs4_set_sequence_privileged(&data->args.seq_args);
8637
8638         return rpc_run_task(&task_setup);
8639 }
8640
8641 /**
8642  * nfs41_free_stateid - perform a FREE_STATEID operation
8643  *
8644  * @server: server / transport on which to perform the operation
8645  * @stateid: state ID to release
8646  * @cred: credential
8647  *
8648  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8649  * negative NFS4ERR value is returned.
8650  */
8651 static int nfs41_free_stateid(struct nfs_server *server,
8652                 nfs4_stateid *stateid,
8653                 struct rpc_cred *cred)
8654 {
8655         struct rpc_task *task;
8656         int ret;
8657
8658         task = _nfs41_free_stateid(server, stateid, cred, true);
8659         if (IS_ERR(task))
8660                 return PTR_ERR(task);
8661         ret = rpc_wait_for_completion_task(task);
8662         if (!ret)
8663                 ret = task->tk_status;
8664         rpc_put_task(task);
8665         return ret;
8666 }
8667
8668 static void
8669 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8670 {
8671         struct rpc_task *task;
8672         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8673
8674         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8675         nfs4_free_lock_state(server, lsp);
8676         if (IS_ERR(task))
8677                 return;
8678         rpc_put_task(task);
8679 }
8680
8681 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8682                 const nfs4_stateid *s2)
8683 {
8684         if (s1->type != s2->type)
8685                 return false;
8686
8687         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8688                 return false;
8689
8690         if (s1->seqid == s2->seqid)
8691                 return true;
8692         if (s1->seqid == 0 || s2->seqid == 0)
8693                 return true;
8694
8695         return false;
8696 }
8697
8698 #endif /* CONFIG_NFS_V4_1 */
8699
8700 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8701                 const nfs4_stateid *s2)
8702 {
8703         return nfs4_stateid_match(s1, s2);
8704 }
8705
8706
8707 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8708         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8709         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8710         .recover_open   = nfs4_open_reclaim,
8711         .recover_lock   = nfs4_lock_reclaim,
8712         .establish_clid = nfs4_init_clientid,
8713         .detect_trunking = nfs40_discover_server_trunking,
8714 };
8715
8716 #if defined(CONFIG_NFS_V4_1)
8717 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8718         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8719         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8720         .recover_open   = nfs4_open_reclaim,
8721         .recover_lock   = nfs4_lock_reclaim,
8722         .establish_clid = nfs41_init_clientid,
8723         .reclaim_complete = nfs41_proc_reclaim_complete,
8724         .detect_trunking = nfs41_discover_server_trunking,
8725 };
8726 #endif /* CONFIG_NFS_V4_1 */
8727
8728 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8729         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8730         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8731         .recover_open   = nfs40_open_expired,
8732         .recover_lock   = nfs4_lock_expired,
8733         .establish_clid = nfs4_init_clientid,
8734 };
8735
8736 #if defined(CONFIG_NFS_V4_1)
8737 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8738         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8739         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8740         .recover_open   = nfs41_open_expired,
8741         .recover_lock   = nfs41_lock_expired,
8742         .establish_clid = nfs41_init_clientid,
8743 };
8744 #endif /* CONFIG_NFS_V4_1 */
8745
8746 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8747         .sched_state_renewal = nfs4_proc_async_renew,
8748         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8749         .renew_lease = nfs4_proc_renew,
8750 };
8751
8752 #if defined(CONFIG_NFS_V4_1)
8753 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8754         .sched_state_renewal = nfs41_proc_async_sequence,
8755         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8756         .renew_lease = nfs4_proc_sequence,
8757 };
8758 #endif
8759
8760 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8761         .get_locations = _nfs40_proc_get_locations,
8762         .fsid_present = _nfs40_proc_fsid_present,
8763 };
8764
8765 #if defined(CONFIG_NFS_V4_1)
8766 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8767         .get_locations = _nfs41_proc_get_locations,
8768         .fsid_present = _nfs41_proc_fsid_present,
8769 };
8770 #endif  /* CONFIG_NFS_V4_1 */
8771
8772 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8773         .minor_version = 0,
8774         .init_caps = NFS_CAP_READDIRPLUS
8775                 | NFS_CAP_ATOMIC_OPEN
8776                 | NFS_CAP_POSIX_LOCK,
8777         .init_client = nfs40_init_client,
8778         .shutdown_client = nfs40_shutdown_client,
8779         .match_stateid = nfs4_match_stateid,
8780         .find_root_sec = nfs4_find_root_sec,
8781         .free_lock_state = nfs4_release_lockowner,
8782         .alloc_seqid = nfs_alloc_seqid,
8783         .call_sync_ops = &nfs40_call_sync_ops,
8784         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8785         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8786         .state_renewal_ops = &nfs40_state_renewal_ops,
8787         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8788 };
8789
8790 #if defined(CONFIG_NFS_V4_1)
8791 static struct nfs_seqid *
8792 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8793 {
8794         return NULL;
8795 }
8796
8797 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8798         .minor_version = 1,
8799         .init_caps = NFS_CAP_READDIRPLUS
8800                 | NFS_CAP_ATOMIC_OPEN
8801                 | NFS_CAP_POSIX_LOCK
8802                 | NFS_CAP_STATEID_NFSV41
8803                 | NFS_CAP_ATOMIC_OPEN_V1,
8804         .init_client = nfs41_init_client,
8805         .shutdown_client = nfs41_shutdown_client,
8806         .match_stateid = nfs41_match_stateid,
8807         .find_root_sec = nfs41_find_root_sec,
8808         .free_lock_state = nfs41_free_lock_state,
8809         .alloc_seqid = nfs_alloc_no_seqid,
8810         .call_sync_ops = &nfs41_call_sync_ops,
8811         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8812         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8813         .state_renewal_ops = &nfs41_state_renewal_ops,
8814         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8815 };
8816 #endif
8817
8818 #if defined(CONFIG_NFS_V4_2)
8819 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8820         .minor_version = 2,
8821         .init_caps = NFS_CAP_READDIRPLUS
8822                 | NFS_CAP_ATOMIC_OPEN
8823                 | NFS_CAP_POSIX_LOCK
8824                 | NFS_CAP_STATEID_NFSV41
8825                 | NFS_CAP_ATOMIC_OPEN_V1
8826                 | NFS_CAP_ALLOCATE
8827                 | NFS_CAP_COPY
8828                 | NFS_CAP_DEALLOCATE
8829                 | NFS_CAP_SEEK
8830                 | NFS_CAP_LAYOUTSTATS
8831                 | NFS_CAP_CLONE,
8832         .init_client = nfs41_init_client,
8833         .shutdown_client = nfs41_shutdown_client,
8834         .match_stateid = nfs41_match_stateid,
8835         .find_root_sec = nfs41_find_root_sec,
8836         .free_lock_state = nfs41_free_lock_state,
8837         .call_sync_ops = &nfs41_call_sync_ops,
8838         .alloc_seqid = nfs_alloc_no_seqid,
8839         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8840         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8841         .state_renewal_ops = &nfs41_state_renewal_ops,
8842         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8843 };
8844 #endif
8845
8846 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8847         [0] = &nfs_v4_0_minor_ops,
8848 #if defined(CONFIG_NFS_V4_1)
8849         [1] = &nfs_v4_1_minor_ops,
8850 #endif
8851 #if defined(CONFIG_NFS_V4_2)
8852         [2] = &nfs_v4_2_minor_ops,
8853 #endif
8854 };
8855
8856 ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
8857 {
8858         ssize_t error, error2;
8859
8860         error = generic_listxattr(dentry, list, size);
8861         if (error < 0)
8862                 return error;
8863         if (list) {
8864                 list += error;
8865                 size -= error;
8866         }
8867
8868         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
8869         if (error2 < 0)
8870                 return error2;
8871         return error + error2;
8872 }
8873
8874 static const struct inode_operations nfs4_dir_inode_operations = {
8875         .create         = nfs_create,
8876         .lookup         = nfs_lookup,
8877         .atomic_open    = nfs_atomic_open,
8878         .link           = nfs_link,
8879         .unlink         = nfs_unlink,
8880         .symlink        = nfs_symlink,
8881         .mkdir          = nfs_mkdir,
8882         .rmdir          = nfs_rmdir,
8883         .mknod          = nfs_mknod,
8884         .rename         = nfs_rename,
8885         .permission     = nfs_permission,
8886         .getattr        = nfs_getattr,
8887         .setattr        = nfs_setattr,
8888         .getxattr       = generic_getxattr,
8889         .setxattr       = generic_setxattr,
8890         .listxattr      = nfs4_listxattr,
8891         .removexattr    = generic_removexattr,
8892 };
8893
8894 static const struct inode_operations nfs4_file_inode_operations = {
8895         .permission     = nfs_permission,
8896         .getattr        = nfs_getattr,
8897         .setattr        = nfs_setattr,
8898         .getxattr       = generic_getxattr,
8899         .setxattr       = generic_setxattr,
8900         .listxattr      = nfs4_listxattr,
8901         .removexattr    = generic_removexattr,
8902 };
8903
8904 const struct nfs_rpc_ops nfs_v4_clientops = {
8905         .version        = 4,                    /* protocol version */
8906         .dentry_ops     = &nfs4_dentry_operations,
8907         .dir_inode_ops  = &nfs4_dir_inode_operations,
8908         .file_inode_ops = &nfs4_file_inode_operations,
8909         .file_ops       = &nfs4_file_operations,
8910         .getroot        = nfs4_proc_get_root,
8911         .submount       = nfs4_submount,
8912         .try_mount      = nfs4_try_mount,
8913         .getattr        = nfs4_proc_getattr,
8914         .setattr        = nfs4_proc_setattr,
8915         .lookup         = nfs4_proc_lookup,
8916         .access         = nfs4_proc_access,
8917         .readlink       = nfs4_proc_readlink,
8918         .create         = nfs4_proc_create,
8919         .remove         = nfs4_proc_remove,
8920         .unlink_setup   = nfs4_proc_unlink_setup,
8921         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8922         .unlink_done    = nfs4_proc_unlink_done,
8923         .rename_setup   = nfs4_proc_rename_setup,
8924         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8925         .rename_done    = nfs4_proc_rename_done,
8926         .link           = nfs4_proc_link,
8927         .symlink        = nfs4_proc_symlink,
8928         .mkdir          = nfs4_proc_mkdir,
8929         .rmdir          = nfs4_proc_remove,
8930         .readdir        = nfs4_proc_readdir,
8931         .mknod          = nfs4_proc_mknod,
8932         .statfs         = nfs4_proc_statfs,
8933         .fsinfo         = nfs4_proc_fsinfo,
8934         .pathconf       = nfs4_proc_pathconf,
8935         .set_capabilities = nfs4_server_capabilities,
8936         .decode_dirent  = nfs4_decode_dirent,
8937         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8938         .read_setup     = nfs4_proc_read_setup,
8939         .read_done      = nfs4_read_done,
8940         .write_setup    = nfs4_proc_write_setup,
8941         .write_done     = nfs4_write_done,
8942         .commit_setup   = nfs4_proc_commit_setup,
8943         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8944         .commit_done    = nfs4_commit_done,
8945         .lock           = nfs4_proc_lock,
8946         .clear_acl_cache = nfs4_zap_acl_attr,
8947         .close_context  = nfs4_close_context,
8948         .open_context   = nfs4_atomic_open,
8949         .have_delegation = nfs4_have_delegation,
8950         .return_delegation = nfs4_inode_return_delegation,
8951         .alloc_client   = nfs4_alloc_client,
8952         .init_client    = nfs4_init_client,
8953         .free_client    = nfs4_free_client,
8954         .create_server  = nfs4_create_server,
8955         .clone_server   = nfs_clone_server,
8956 };
8957
8958 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8959         .name   = XATTR_NAME_NFSV4_ACL,
8960         .list   = nfs4_xattr_list_nfs4_acl,
8961         .get    = nfs4_xattr_get_nfs4_acl,
8962         .set    = nfs4_xattr_set_nfs4_acl,
8963 };
8964
8965 const struct xattr_handler *nfs4_xattr_handlers[] = {
8966         &nfs4_xattr_nfs4_acl_handler,
8967 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8968         &nfs4_xattr_nfs4_label_handler,
8969 #endif
8970         NULL
8971 };
8972
8973 /*
8974  * Local variables:
8975  *  c-basic-offset: 8
8976  * End:
8977  */