nfs: add support for the umask attribute
[sfrench/cifs-2.6.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/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 nfs_open_context *ctx, 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 *, const nfs4_stateid *,
103                 struct rpc_cred *, bool);
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_FILEID,
230 };
231
232 const u32 nfs4_statfs_bitmap[3] = {
233         FATTR4_WORD0_FILES_AVAIL
234         | FATTR4_WORD0_FILES_FREE
235         | FATTR4_WORD0_FILES_TOTAL,
236         FATTR4_WORD1_SPACE_AVAIL
237         | FATTR4_WORD1_SPACE_FREE
238         | FATTR4_WORD1_SPACE_TOTAL
239 };
240
241 const u32 nfs4_pathconf_bitmap[3] = {
242         FATTR4_WORD0_MAXLINK
243         | FATTR4_WORD0_MAXNAME,
244         0
245 };
246
247 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
248                         | FATTR4_WORD0_MAXREAD
249                         | FATTR4_WORD0_MAXWRITE
250                         | FATTR4_WORD0_LEASE_TIME,
251                         FATTR4_WORD1_TIME_DELTA
252                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
253                         FATTR4_WORD2_LAYOUT_BLKSIZE
254                         | FATTR4_WORD2_CLONE_BLKSIZE
255 };
256
257 const u32 nfs4_fs_locations_bitmap[3] = {
258         FATTR4_WORD0_TYPE
259         | FATTR4_WORD0_CHANGE
260         | FATTR4_WORD0_SIZE
261         | FATTR4_WORD0_FSID
262         | FATTR4_WORD0_FILEID
263         | FATTR4_WORD0_FS_LOCATIONS,
264         FATTR4_WORD1_MODE
265         | FATTR4_WORD1_NUMLINKS
266         | FATTR4_WORD1_OWNER
267         | FATTR4_WORD1_OWNER_GROUP
268         | FATTR4_WORD1_RAWDEV
269         | FATTR4_WORD1_SPACE_USED
270         | FATTR4_WORD1_TIME_ACCESS
271         | FATTR4_WORD1_TIME_METADATA
272         | FATTR4_WORD1_TIME_MODIFY
273         | FATTR4_WORD1_MOUNTED_ON_FILEID,
274 };
275
276 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
277                 struct nfs4_readdir_arg *readdir)
278 {
279         __be32 *start, *p;
280
281         if (cookie > 2) {
282                 readdir->cookie = cookie;
283                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
284                 return;
285         }
286
287         readdir->cookie = 0;
288         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
289         if (cookie == 2)
290                 return;
291         
292         /*
293          * NFSv4 servers do not return entries for '.' and '..'
294          * Therefore, we fake these entries here.  We let '.'
295          * have cookie 0 and '..' have cookie 1.  Note that
296          * when talking to the server, we always send cookie 0
297          * instead of 1 or 2.
298          */
299         start = p = kmap_atomic(*readdir->pages);
300         
301         if (cookie == 0) {
302                 *p++ = xdr_one;                                  /* next */
303                 *p++ = xdr_zero;                   /* cookie, first word */
304                 *p++ = xdr_one;                   /* cookie, second word */
305                 *p++ = xdr_one;                             /* entry len */
306                 memcpy(p, ".\0\0\0", 4);                        /* entry */
307                 p++;
308                 *p++ = xdr_one;                         /* bitmap length */
309                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
310                 *p++ = htonl(8);              /* attribute buffer length */
311                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
312         }
313         
314         *p++ = xdr_one;                                  /* next */
315         *p++ = xdr_zero;                   /* cookie, first word */
316         *p++ = xdr_two;                   /* cookie, second word */
317         *p++ = xdr_two;                             /* entry len */
318         memcpy(p, "..\0\0", 4);                         /* entry */
319         p++;
320         *p++ = xdr_one;                         /* bitmap length */
321         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
322         *p++ = htonl(8);              /* attribute buffer length */
323         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
324
325         readdir->pgbase = (char *)p - (char *)start;
326         readdir->count -= readdir->pgbase;
327         kunmap_atomic(start);
328 }
329
330 static void nfs4_test_and_free_stateid(struct nfs_server *server,
331                 nfs4_stateid *stateid,
332                 struct rpc_cred *cred)
333 {
334         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
335
336         ops->test_and_free_expired(server, stateid, cred);
337 }
338
339 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
340                 nfs4_stateid *stateid,
341                 struct rpc_cred *cred)
342 {
343         stateid->type = NFS4_REVOKED_STATEID_TYPE;
344         nfs4_test_and_free_stateid(server, stateid, cred);
345 }
346
347 static void nfs4_free_revoked_stateid(struct nfs_server *server,
348                 const nfs4_stateid *stateid,
349                 struct rpc_cred *cred)
350 {
351         nfs4_stateid tmp;
352
353         nfs4_stateid_copy(&tmp, stateid);
354         __nfs4_free_revoked_stateid(server, &tmp, cred);
355 }
356
357 static long nfs4_update_delay(long *timeout)
358 {
359         long ret;
360         if (!timeout)
361                 return NFS4_POLL_RETRY_MAX;
362         if (*timeout <= 0)
363                 *timeout = NFS4_POLL_RETRY_MIN;
364         if (*timeout > NFS4_POLL_RETRY_MAX)
365                 *timeout = NFS4_POLL_RETRY_MAX;
366         ret = *timeout;
367         *timeout <<= 1;
368         return ret;
369 }
370
371 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
372 {
373         int res = 0;
374
375         might_sleep();
376
377         freezable_schedule_timeout_killable_unsafe(
378                 nfs4_update_delay(timeout));
379         if (fatal_signal_pending(current))
380                 res = -ERESTARTSYS;
381         return res;
382 }
383
384 /* This is the error handling routine for processes that are allowed
385  * to sleep.
386  */
387 static int nfs4_do_handle_exception(struct nfs_server *server,
388                 int errorcode, struct nfs4_exception *exception)
389 {
390         struct nfs_client *clp = server->nfs_client;
391         struct nfs4_state *state = exception->state;
392         const nfs4_stateid *stateid = exception->stateid;
393         struct inode *inode = exception->inode;
394         int ret = errorcode;
395
396         exception->delay = 0;
397         exception->recovering = 0;
398         exception->retry = 0;
399
400         if (stateid == NULL && state != NULL)
401                 stateid = &state->stateid;
402
403         switch(errorcode) {
404                 case 0:
405                         return 0;
406                 case -NFS4ERR_DELEG_REVOKED:
407                 case -NFS4ERR_ADMIN_REVOKED:
408                 case -NFS4ERR_EXPIRED:
409                 case -NFS4ERR_BAD_STATEID:
410                         if (inode != NULL && stateid != NULL) {
411                                 nfs_inode_find_state_and_recover(inode,
412                                                 stateid);
413                                 goto wait_on_recovery;
414                         }
415                 case -NFS4ERR_OPENMODE:
416                         if (inode) {
417                                 int err;
418
419                                 err = nfs_async_inode_return_delegation(inode,
420                                                 stateid);
421                                 if (err == 0)
422                                         goto wait_on_recovery;
423                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
424                                         exception->retry = 1;
425                                         break;
426                                 }
427                         }
428                         if (state == NULL)
429                                 break;
430                         ret = nfs4_schedule_stateid_recovery(server, state);
431                         if (ret < 0)
432                                 break;
433                         goto wait_on_recovery;
434                 case -NFS4ERR_STALE_STATEID:
435                 case -NFS4ERR_STALE_CLIENTID:
436                         nfs4_schedule_lease_recovery(clp);
437                         goto wait_on_recovery;
438                 case -NFS4ERR_MOVED:
439                         ret = nfs4_schedule_migration_recovery(server);
440                         if (ret < 0)
441                                 break;
442                         goto wait_on_recovery;
443                 case -NFS4ERR_LEASE_MOVED:
444                         nfs4_schedule_lease_moved_recovery(clp);
445                         goto wait_on_recovery;
446 #if defined(CONFIG_NFS_V4_1)
447                 case -NFS4ERR_BADSESSION:
448                 case -NFS4ERR_BADSLOT:
449                 case -NFS4ERR_BAD_HIGH_SLOT:
450                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
451                 case -NFS4ERR_DEADSESSION:
452                 case -NFS4ERR_SEQ_FALSE_RETRY:
453                 case -NFS4ERR_SEQ_MISORDERED:
454                         dprintk("%s ERROR: %d Reset session\n", __func__,
455                                 errorcode);
456                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
457                         goto wait_on_recovery;
458 #endif /* defined(CONFIG_NFS_V4_1) */
459                 case -NFS4ERR_FILE_OPEN:
460                         if (exception->timeout > HZ) {
461                                 /* We have retried a decent amount, time to
462                                  * fail
463                                  */
464                                 ret = -EBUSY;
465                                 break;
466                         }
467                 case -NFS4ERR_DELAY:
468                         nfs_inc_server_stats(server, NFSIOS_DELAY);
469                 case -NFS4ERR_GRACE:
470                 case -NFS4ERR_LAYOUTTRYLATER:
471                 case -NFS4ERR_RECALLCONFLICT:
472                         exception->delay = 1;
473                         return 0;
474
475                 case -NFS4ERR_RETRY_UNCACHED_REP:
476                 case -NFS4ERR_OLD_STATEID:
477                         exception->retry = 1;
478                         break;
479                 case -NFS4ERR_BADOWNER:
480                         /* The following works around a Linux server bug! */
481                 case -NFS4ERR_BADNAME:
482                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
483                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
484                                 exception->retry = 1;
485                                 printk(KERN_WARNING "NFS: v4 server %s "
486                                                 "does not accept raw "
487                                                 "uid/gids. "
488                                                 "Reenabling the idmapper.\n",
489                                                 server->nfs_client->cl_hostname);
490                         }
491         }
492         /* We failed to handle the error */
493         return nfs4_map_errors(ret);
494 wait_on_recovery:
495         exception->recovering = 1;
496         return 0;
497 }
498
499 /* This is the error handling routine for processes that are allowed
500  * to sleep.
501  */
502 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
503 {
504         struct nfs_client *clp = server->nfs_client;
505         int ret;
506
507         ret = nfs4_do_handle_exception(server, errorcode, exception);
508         if (exception->delay) {
509                 ret = nfs4_delay(server->client, &exception->timeout);
510                 goto out_retry;
511         }
512         if (exception->recovering) {
513                 ret = nfs4_wait_clnt_recover(clp);
514                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
515                         return -EIO;
516                 goto out_retry;
517         }
518         return ret;
519 out_retry:
520         if (ret == 0)
521                 exception->retry = 1;
522         return ret;
523 }
524
525 static int
526 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
527                 int errorcode, struct nfs4_exception *exception)
528 {
529         struct nfs_client *clp = server->nfs_client;
530         int ret;
531
532         ret = nfs4_do_handle_exception(server, errorcode, exception);
533         if (exception->delay) {
534                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
535                 goto out_retry;
536         }
537         if (exception->recovering) {
538                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
539                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
540                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
541                 goto out_retry;
542         }
543         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
544                 ret = -EIO;
545         return ret;
546 out_retry:
547         if (ret == 0)
548                 exception->retry = 1;
549         return ret;
550 }
551
552 static int
553 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
554                         struct nfs4_state *state, long *timeout)
555 {
556         struct nfs4_exception exception = {
557                 .state = state,
558         };
559
560         if (task->tk_status >= 0)
561                 return 0;
562         if (timeout)
563                 exception.timeout = *timeout;
564         task->tk_status = nfs4_async_handle_exception(task, server,
565                         task->tk_status,
566                         &exception);
567         if (exception.delay && timeout)
568                 *timeout = exception.timeout;
569         if (exception.retry)
570                 return -EAGAIN;
571         return 0;
572 }
573
574 /*
575  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
576  * or 'false' otherwise.
577  */
578 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
579 {
580         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
581
582         if (flavor == RPC_AUTH_GSS_KRB5I ||
583             flavor == RPC_AUTH_GSS_KRB5P)
584                 return true;
585
586         return false;
587 }
588
589 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
590 {
591         spin_lock(&clp->cl_lock);
592         if (time_before(clp->cl_last_renewal,timestamp))
593                 clp->cl_last_renewal = timestamp;
594         spin_unlock(&clp->cl_lock);
595 }
596
597 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
598 {
599         struct nfs_client *clp = server->nfs_client;
600
601         if (!nfs4_has_session(clp))
602                 do_renew_lease(clp, timestamp);
603 }
604
605 struct nfs4_call_sync_data {
606         const struct nfs_server *seq_server;
607         struct nfs4_sequence_args *seq_args;
608         struct nfs4_sequence_res *seq_res;
609 };
610
611 void nfs4_init_sequence(struct nfs4_sequence_args *args,
612                         struct nfs4_sequence_res *res, int cache_reply)
613 {
614         args->sa_slot = NULL;
615         args->sa_cache_this = cache_reply;
616         args->sa_privileged = 0;
617
618         res->sr_slot = NULL;
619 }
620
621 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
622 {
623         args->sa_privileged = 1;
624 }
625
626 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
627                          struct nfs4_sequence_args *args,
628                          struct nfs4_sequence_res *res,
629                          struct rpc_task *task)
630 {
631         struct nfs4_slot *slot;
632
633         /* slot already allocated? */
634         if (res->sr_slot != NULL)
635                 goto out_start;
636
637         spin_lock(&tbl->slot_tbl_lock);
638         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
639                 goto out_sleep;
640
641         slot = nfs4_alloc_slot(tbl);
642         if (IS_ERR(slot)) {
643                 if (slot == ERR_PTR(-ENOMEM))
644                         task->tk_timeout = HZ >> 2;
645                 goto out_sleep;
646         }
647         spin_unlock(&tbl->slot_tbl_lock);
648
649         slot->privileged = args->sa_privileged ? 1 : 0;
650         args->sa_slot = slot;
651         res->sr_slot = slot;
652
653 out_start:
654         rpc_call_start(task);
655         return 0;
656
657 out_sleep:
658         if (args->sa_privileged)
659                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
660                                 NULL, RPC_PRIORITY_PRIVILEGED);
661         else
662                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
663         spin_unlock(&tbl->slot_tbl_lock);
664         return -EAGAIN;
665 }
666 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
667
668 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
669 {
670         struct nfs4_slot *slot = res->sr_slot;
671         struct nfs4_slot_table *tbl;
672
673         tbl = slot->table;
674         spin_lock(&tbl->slot_tbl_lock);
675         if (!nfs41_wake_and_assign_slot(tbl, slot))
676                 nfs4_free_slot(tbl, slot);
677         spin_unlock(&tbl->slot_tbl_lock);
678
679         res->sr_slot = NULL;
680 }
681
682 static int nfs40_sequence_done(struct rpc_task *task,
683                                struct nfs4_sequence_res *res)
684 {
685         if (res->sr_slot != NULL)
686                 nfs40_sequence_free_slot(res);
687         return 1;
688 }
689
690 #if defined(CONFIG_NFS_V4_1)
691
692 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
693 {
694         struct nfs4_session *session;
695         struct nfs4_slot_table *tbl;
696         struct nfs4_slot *slot = res->sr_slot;
697         bool send_new_highest_used_slotid = false;
698
699         tbl = slot->table;
700         session = tbl->session;
701
702         /* Bump the slot sequence number */
703         if (slot->seq_done)
704                 slot->seq_nr++;
705         slot->seq_done = 0;
706
707         spin_lock(&tbl->slot_tbl_lock);
708         /* Be nice to the server: try to ensure that the last transmitted
709          * value for highest_user_slotid <= target_highest_slotid
710          */
711         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
712                 send_new_highest_used_slotid = true;
713
714         if (nfs41_wake_and_assign_slot(tbl, slot)) {
715                 send_new_highest_used_slotid = false;
716                 goto out_unlock;
717         }
718         nfs4_free_slot(tbl, slot);
719
720         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
721                 send_new_highest_used_slotid = false;
722 out_unlock:
723         spin_unlock(&tbl->slot_tbl_lock);
724         res->sr_slot = NULL;
725         if (send_new_highest_used_slotid)
726                 nfs41_notify_server(session->clp);
727         if (waitqueue_active(&tbl->slot_waitq))
728                 wake_up_all(&tbl->slot_waitq);
729 }
730
731 static int nfs41_sequence_process(struct rpc_task *task,
732                 struct nfs4_sequence_res *res)
733 {
734         struct nfs4_session *session;
735         struct nfs4_slot *slot = res->sr_slot;
736         struct nfs_client *clp;
737         bool interrupted = false;
738         int ret = 1;
739
740         if (slot == NULL)
741                 goto out_noaction;
742         /* don't increment the sequence number if the task wasn't sent */
743         if (!RPC_WAS_SENT(task))
744                 goto out;
745
746         session = slot->table->session;
747
748         if (slot->interrupted) {
749                 slot->interrupted = 0;
750                 interrupted = true;
751         }
752
753         trace_nfs4_sequence_done(session, res);
754         /* Check the SEQUENCE operation status */
755         switch (res->sr_status) {
756         case 0:
757                 /* If previous op on slot was interrupted and we reused
758                  * the seq# and got a reply from the cache, then retry
759                  */
760                 if (task->tk_status == -EREMOTEIO && interrupted) {
761                         ++slot->seq_nr;
762                         goto retry_nowait;
763                 }
764                 /* Update the slot's sequence and clientid lease timer */
765                 slot->seq_done = 1;
766                 clp = session->clp;
767                 do_renew_lease(clp, res->sr_timestamp);
768                 /* Check sequence flags */
769                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
770                                 !!slot->privileged);
771                 nfs41_update_target_slotid(slot->table, slot, res);
772                 break;
773         case 1:
774                 /*
775                  * sr_status remains 1 if an RPC level error occurred.
776                  * The server may or may not have processed the sequence
777                  * operation..
778                  * Mark the slot as having hosted an interrupted RPC call.
779                  */
780                 slot->interrupted = 1;
781                 goto out;
782         case -NFS4ERR_DELAY:
783                 /* The server detected a resend of the RPC call and
784                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
785                  * of RFC5661.
786                  */
787                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
788                         __func__,
789                         slot->slot_nr,
790                         slot->seq_nr);
791                 goto out_retry;
792         case -NFS4ERR_BADSLOT:
793                 /*
794                  * The slot id we used was probably retired. Try again
795                  * using a different slot id.
796                  */
797                 goto retry_nowait;
798         case -NFS4ERR_SEQ_MISORDERED:
799                 /*
800                  * Was the last operation on this sequence interrupted?
801                  * If so, retry after bumping the sequence number.
802                  */
803                 if (interrupted) {
804                         ++slot->seq_nr;
805                         goto retry_nowait;
806                 }
807                 /*
808                  * Could this slot have been previously retired?
809                  * If so, then the server may be expecting seq_nr = 1!
810                  */
811                 if (slot->seq_nr != 1) {
812                         slot->seq_nr = 1;
813                         goto retry_nowait;
814                 }
815                 break;
816         case -NFS4ERR_SEQ_FALSE_RETRY:
817                 ++slot->seq_nr;
818                 goto retry_nowait;
819         case -NFS4ERR_DEADSESSION:
820         case -NFS4ERR_BADSESSION:
821                 nfs4_schedule_session_recovery(session, res->sr_status);
822                 goto retry_nowait;
823         default:
824                 /* Just update the slot sequence no. */
825                 slot->seq_done = 1;
826         }
827 out:
828         /* The session may be reset by one of the error handlers. */
829         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
830 out_noaction:
831         return ret;
832 retry_nowait:
833         if (rpc_restart_call_prepare(task)) {
834                 nfs41_sequence_free_slot(res);
835                 task->tk_status = 0;
836                 ret = 0;
837         }
838         goto out;
839 out_retry:
840         if (!rpc_restart_call(task))
841                 goto out;
842         rpc_delay(task, NFS4_POLL_RETRY_MAX);
843         return 0;
844 }
845
846 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
847 {
848         if (!nfs41_sequence_process(task, res))
849                 return 0;
850         if (res->sr_slot != NULL)
851                 nfs41_sequence_free_slot(res);
852         return 1;
853
854 }
855 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
856
857 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
858 {
859         if (res->sr_slot == NULL)
860                 return 1;
861         if (res->sr_slot->table->session != NULL)
862                 return nfs41_sequence_process(task, res);
863         return nfs40_sequence_done(task, res);
864 }
865
866 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
867 {
868         if (res->sr_slot != NULL) {
869                 if (res->sr_slot->table->session != NULL)
870                         nfs41_sequence_free_slot(res);
871                 else
872                         nfs40_sequence_free_slot(res);
873         }
874 }
875
876 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
877 {
878         if (res->sr_slot == NULL)
879                 return 1;
880         if (!res->sr_slot->table->session)
881                 return nfs40_sequence_done(task, res);
882         return nfs41_sequence_done(task, res);
883 }
884 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
885
886 int nfs41_setup_sequence(struct nfs4_session *session,
887                                 struct nfs4_sequence_args *args,
888                                 struct nfs4_sequence_res *res,
889                                 struct rpc_task *task)
890 {
891         struct nfs4_slot *slot;
892         struct nfs4_slot_table *tbl;
893
894         dprintk("--> %s\n", __func__);
895         /* slot already allocated? */
896         if (res->sr_slot != NULL)
897                 goto out_success;
898
899         tbl = &session->fc_slot_table;
900
901         task->tk_timeout = 0;
902
903         spin_lock(&tbl->slot_tbl_lock);
904         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
905             !args->sa_privileged) {
906                 /* The state manager will wait until the slot table is empty */
907                 dprintk("%s session is draining\n", __func__);
908                 goto out_sleep;
909         }
910
911         slot = nfs4_alloc_slot(tbl);
912         if (IS_ERR(slot)) {
913                 /* If out of memory, try again in 1/4 second */
914                 if (slot == ERR_PTR(-ENOMEM))
915                         task->tk_timeout = HZ >> 2;
916                 dprintk("<-- %s: no free slots\n", __func__);
917                 goto out_sleep;
918         }
919         spin_unlock(&tbl->slot_tbl_lock);
920
921         slot->privileged = args->sa_privileged ? 1 : 0;
922         args->sa_slot = slot;
923
924         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
925                         slot->slot_nr, slot->seq_nr);
926
927         res->sr_slot = slot;
928         res->sr_timestamp = jiffies;
929         res->sr_status_flags = 0;
930         /*
931          * sr_status is only set in decode_sequence, and so will remain
932          * set to 1 if an rpc level failure occurs.
933          */
934         res->sr_status = 1;
935         trace_nfs4_setup_sequence(session, args);
936 out_success:
937         rpc_call_start(task);
938         return 0;
939 out_sleep:
940         /* Privileged tasks are queued with top priority */
941         if (args->sa_privileged)
942                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
943                                 NULL, RPC_PRIORITY_PRIVILEGED);
944         else
945                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
946         spin_unlock(&tbl->slot_tbl_lock);
947         return -EAGAIN;
948 }
949 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
950
951 static int nfs4_setup_sequence(const struct nfs_server *server,
952                                struct nfs4_sequence_args *args,
953                                struct nfs4_sequence_res *res,
954                                struct rpc_task *task)
955 {
956         struct nfs4_session *session = nfs4_get_session(server);
957         int ret = 0;
958
959         if (!session)
960                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
961                                             args, res, task);
962
963         dprintk("--> %s clp %p session %p sr_slot %u\n",
964                 __func__, session->clp, session, res->sr_slot ?
965                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
966
967         ret = nfs41_setup_sequence(session, args, res, task);
968
969         dprintk("<-- %s status=%d\n", __func__, ret);
970         return ret;
971 }
972
973 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
974 {
975         struct nfs4_call_sync_data *data = calldata;
976         struct nfs4_session *session = nfs4_get_session(data->seq_server);
977
978         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
979
980         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
981 }
982
983 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
984 {
985         struct nfs4_call_sync_data *data = calldata;
986
987         nfs41_sequence_done(task, data->seq_res);
988 }
989
990 static const struct rpc_call_ops nfs41_call_sync_ops = {
991         .rpc_call_prepare = nfs41_call_sync_prepare,
992         .rpc_call_done = nfs41_call_sync_done,
993 };
994
995 #else   /* !CONFIG_NFS_V4_1 */
996
997 static int nfs4_setup_sequence(const struct nfs_server *server,
998                                struct nfs4_sequence_args *args,
999                                struct nfs4_sequence_res *res,
1000                                struct rpc_task *task)
1001 {
1002         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
1003                                     args, res, task);
1004 }
1005
1006 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1007 {
1008         return nfs40_sequence_done(task, res);
1009 }
1010
1011 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1012 {
1013         if (res->sr_slot != NULL)
1014                 nfs40_sequence_free_slot(res);
1015 }
1016
1017 int nfs4_sequence_done(struct rpc_task *task,
1018                        struct nfs4_sequence_res *res)
1019 {
1020         return nfs40_sequence_done(task, res);
1021 }
1022 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1023
1024 #endif  /* !CONFIG_NFS_V4_1 */
1025
1026 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1027 {
1028         struct nfs4_call_sync_data *data = calldata;
1029         nfs4_setup_sequence(data->seq_server,
1030                                 data->seq_args, data->seq_res, task);
1031 }
1032
1033 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1034 {
1035         struct nfs4_call_sync_data *data = calldata;
1036         nfs4_sequence_done(task, data->seq_res);
1037 }
1038
1039 static const struct rpc_call_ops nfs40_call_sync_ops = {
1040         .rpc_call_prepare = nfs40_call_sync_prepare,
1041         .rpc_call_done = nfs40_call_sync_done,
1042 };
1043
1044 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1045                                    struct nfs_server *server,
1046                                    struct rpc_message *msg,
1047                                    struct nfs4_sequence_args *args,
1048                                    struct nfs4_sequence_res *res)
1049 {
1050         int ret;
1051         struct rpc_task *task;
1052         struct nfs_client *clp = server->nfs_client;
1053         struct nfs4_call_sync_data data = {
1054                 .seq_server = server,
1055                 .seq_args = args,
1056                 .seq_res = res,
1057         };
1058         struct rpc_task_setup task_setup = {
1059                 .rpc_client = clnt,
1060                 .rpc_message = msg,
1061                 .callback_ops = clp->cl_mvops->call_sync_ops,
1062                 .callback_data = &data
1063         };
1064
1065         task = rpc_run_task(&task_setup);
1066         if (IS_ERR(task))
1067                 ret = PTR_ERR(task);
1068         else {
1069                 ret = task->tk_status;
1070                 rpc_put_task(task);
1071         }
1072         return ret;
1073 }
1074
1075 int nfs4_call_sync(struct rpc_clnt *clnt,
1076                    struct nfs_server *server,
1077                    struct rpc_message *msg,
1078                    struct nfs4_sequence_args *args,
1079                    struct nfs4_sequence_res *res,
1080                    int cache_reply)
1081 {
1082         nfs4_init_sequence(args, res, cache_reply);
1083         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1084 }
1085
1086 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
1087 {
1088         struct nfs_inode *nfsi = NFS_I(dir);
1089
1090         spin_lock(&dir->i_lock);
1091         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1092         if (!cinfo->atomic || cinfo->before != dir->i_version)
1093                 nfs_force_lookup_revalidate(dir);
1094         dir->i_version = cinfo->after;
1095         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1096         nfs_fscache_invalidate(dir);
1097         spin_unlock(&dir->i_lock);
1098 }
1099
1100 struct nfs4_opendata {
1101         struct kref kref;
1102         struct nfs_openargs o_arg;
1103         struct nfs_openres o_res;
1104         struct nfs_open_confirmargs c_arg;
1105         struct nfs_open_confirmres c_res;
1106         struct nfs4_string owner_name;
1107         struct nfs4_string group_name;
1108         struct nfs4_label *a_label;
1109         struct nfs_fattr f_attr;
1110         struct nfs4_label *f_label;
1111         struct dentry *dir;
1112         struct dentry *dentry;
1113         struct nfs4_state_owner *owner;
1114         struct nfs4_state *state;
1115         struct iattr attrs;
1116         unsigned long timestamp;
1117         unsigned int rpc_done : 1;
1118         unsigned int file_created : 1;
1119         unsigned int is_recover : 1;
1120         int rpc_status;
1121         int cancelled;
1122 };
1123
1124 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1125                 int err, struct nfs4_exception *exception)
1126 {
1127         if (err != -EINVAL)
1128                 return false;
1129         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1130                 return false;
1131         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1132         exception->retry = 1;
1133         return true;
1134 }
1135
1136 static u32
1137 nfs4_map_atomic_open_share(struct nfs_server *server,
1138                 fmode_t fmode, int openflags)
1139 {
1140         u32 res = 0;
1141
1142         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1143         case FMODE_READ:
1144                 res = NFS4_SHARE_ACCESS_READ;
1145                 break;
1146         case FMODE_WRITE:
1147                 res = NFS4_SHARE_ACCESS_WRITE;
1148                 break;
1149         case FMODE_READ|FMODE_WRITE:
1150                 res = NFS4_SHARE_ACCESS_BOTH;
1151         }
1152         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1153                 goto out;
1154         /* Want no delegation if we're using O_DIRECT */
1155         if (openflags & O_DIRECT)
1156                 res |= NFS4_SHARE_WANT_NO_DELEG;
1157 out:
1158         return res;
1159 }
1160
1161 static enum open_claim_type4
1162 nfs4_map_atomic_open_claim(struct nfs_server *server,
1163                 enum open_claim_type4 claim)
1164 {
1165         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1166                 return claim;
1167         switch (claim) {
1168         default:
1169                 return claim;
1170         case NFS4_OPEN_CLAIM_FH:
1171                 return NFS4_OPEN_CLAIM_NULL;
1172         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1173                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1174         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1175                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1176         }
1177 }
1178
1179 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1180 {
1181         p->o_res.f_attr = &p->f_attr;
1182         p->o_res.f_label = p->f_label;
1183         p->o_res.seqid = p->o_arg.seqid;
1184         p->c_res.seqid = p->c_arg.seqid;
1185         p->o_res.server = p->o_arg.server;
1186         p->o_res.access_request = p->o_arg.access;
1187         nfs_fattr_init(&p->f_attr);
1188         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1189 }
1190
1191 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1192                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1193                 const struct iattr *attrs,
1194                 struct nfs4_label *label,
1195                 enum open_claim_type4 claim,
1196                 gfp_t gfp_mask)
1197 {
1198         struct dentry *parent = dget_parent(dentry);
1199         struct inode *dir = d_inode(parent);
1200         struct nfs_server *server = NFS_SERVER(dir);
1201         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1202         struct nfs4_opendata *p;
1203
1204         p = kzalloc(sizeof(*p), gfp_mask);
1205         if (p == NULL)
1206                 goto err;
1207
1208         p->f_label = nfs4_label_alloc(server, gfp_mask);
1209         if (IS_ERR(p->f_label))
1210                 goto err_free_p;
1211
1212         p->a_label = nfs4_label_alloc(server, gfp_mask);
1213         if (IS_ERR(p->a_label))
1214                 goto err_free_f;
1215
1216         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1217         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1218         if (IS_ERR(p->o_arg.seqid))
1219                 goto err_free_label;
1220         nfs_sb_active(dentry->d_sb);
1221         p->dentry = dget(dentry);
1222         p->dir = parent;
1223         p->owner = sp;
1224         atomic_inc(&sp->so_count);
1225         p->o_arg.open_flags = flags;
1226         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1227         p->o_arg.umask = current_umask();
1228         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1229         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1230                         fmode, flags);
1231         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1232          * will return permission denied for all bits until close */
1233         if (!(flags & O_EXCL)) {
1234                 /* ask server to check for all possible rights as results
1235                  * are cached */
1236                 switch (p->o_arg.claim) {
1237                 default:
1238                         break;
1239                 case NFS4_OPEN_CLAIM_NULL:
1240                 case NFS4_OPEN_CLAIM_FH:
1241                         p->o_arg.access = NFS4_ACCESS_READ |
1242                                 NFS4_ACCESS_MODIFY |
1243                                 NFS4_ACCESS_EXTEND |
1244                                 NFS4_ACCESS_EXECUTE;
1245                 }
1246         }
1247         p->o_arg.clientid = server->nfs_client->cl_clientid;
1248         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1249         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1250         p->o_arg.name = &dentry->d_name;
1251         p->o_arg.server = server;
1252         p->o_arg.bitmask = nfs4_bitmask(server, label);
1253         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1254         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1255         switch (p->o_arg.claim) {
1256         case NFS4_OPEN_CLAIM_NULL:
1257         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1258         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1259                 p->o_arg.fh = NFS_FH(dir);
1260                 break;
1261         case NFS4_OPEN_CLAIM_PREVIOUS:
1262         case NFS4_OPEN_CLAIM_FH:
1263         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1264         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1265                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1266         }
1267         if (attrs != NULL && attrs->ia_valid != 0) {
1268                 __u32 verf[2];
1269
1270                 p->o_arg.u.attrs = &p->attrs;
1271                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1272
1273                 verf[0] = jiffies;
1274                 verf[1] = current->pid;
1275                 memcpy(p->o_arg.u.verifier.data, verf,
1276                                 sizeof(p->o_arg.u.verifier.data));
1277         }
1278         p->c_arg.fh = &p->o_res.fh;
1279         p->c_arg.stateid = &p->o_res.stateid;
1280         p->c_arg.seqid = p->o_arg.seqid;
1281         nfs4_init_opendata_res(p);
1282         kref_init(&p->kref);
1283         return p;
1284
1285 err_free_label:
1286         nfs4_label_free(p->a_label);
1287 err_free_f:
1288         nfs4_label_free(p->f_label);
1289 err_free_p:
1290         kfree(p);
1291 err:
1292         dput(parent);
1293         return NULL;
1294 }
1295
1296 static void nfs4_opendata_free(struct kref *kref)
1297 {
1298         struct nfs4_opendata *p = container_of(kref,
1299                         struct nfs4_opendata, kref);
1300         struct super_block *sb = p->dentry->d_sb;
1301
1302         nfs_free_seqid(p->o_arg.seqid);
1303         nfs4_sequence_free_slot(&p->o_res.seq_res);
1304         if (p->state != NULL)
1305                 nfs4_put_open_state(p->state);
1306         nfs4_put_state_owner(p->owner);
1307
1308         nfs4_label_free(p->a_label);
1309         nfs4_label_free(p->f_label);
1310
1311         dput(p->dir);
1312         dput(p->dentry);
1313         nfs_sb_deactive(sb);
1314         nfs_fattr_free_names(&p->f_attr);
1315         kfree(p->f_attr.mdsthreshold);
1316         kfree(p);
1317 }
1318
1319 static void nfs4_opendata_put(struct nfs4_opendata *p)
1320 {
1321         if (p != NULL)
1322                 kref_put(&p->kref, nfs4_opendata_free);
1323 }
1324
1325 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1326 {
1327         int ret;
1328
1329         ret = rpc_wait_for_completion_task(task);
1330         return ret;
1331 }
1332
1333 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1334                 fmode_t fmode)
1335 {
1336         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1337         case FMODE_READ|FMODE_WRITE:
1338                 return state->n_rdwr != 0;
1339         case FMODE_WRITE:
1340                 return state->n_wronly != 0;
1341         case FMODE_READ:
1342                 return state->n_rdonly != 0;
1343         }
1344         WARN_ON_ONCE(1);
1345         return false;
1346 }
1347
1348 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1349 {
1350         int ret = 0;
1351
1352         if (open_mode & (O_EXCL|O_TRUNC))
1353                 goto out;
1354         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1355                 case FMODE_READ:
1356                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1357                                 && state->n_rdonly != 0;
1358                         break;
1359                 case FMODE_WRITE:
1360                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1361                                 && state->n_wronly != 0;
1362                         break;
1363                 case FMODE_READ|FMODE_WRITE:
1364                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1365                                 && state->n_rdwr != 0;
1366         }
1367 out:
1368         return ret;
1369 }
1370
1371 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1372                 enum open_claim_type4 claim)
1373 {
1374         if (delegation == NULL)
1375                 return 0;
1376         if ((delegation->type & fmode) != fmode)
1377                 return 0;
1378         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1379                 return 0;
1380         switch (claim) {
1381         case NFS4_OPEN_CLAIM_NULL:
1382         case NFS4_OPEN_CLAIM_FH:
1383                 break;
1384         case NFS4_OPEN_CLAIM_PREVIOUS:
1385                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1386                         break;
1387         default:
1388                 return 0;
1389         }
1390         nfs_mark_delegation_referenced(delegation);
1391         return 1;
1392 }
1393
1394 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1395 {
1396         switch (fmode) {
1397                 case FMODE_WRITE:
1398                         state->n_wronly++;
1399                         break;
1400                 case FMODE_READ:
1401                         state->n_rdonly++;
1402                         break;
1403                 case FMODE_READ|FMODE_WRITE:
1404                         state->n_rdwr++;
1405         }
1406         nfs4_state_set_mode_locked(state, state->state | fmode);
1407 }
1408
1409 #ifdef CONFIG_NFS_V4_1
1410 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1411 {
1412         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1413                 return true;
1414         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1415                 return true;
1416         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1417                 return true;
1418         return false;
1419 }
1420 #endif /* CONFIG_NFS_V4_1 */
1421
1422 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1423 {
1424         struct nfs_client *clp = state->owner->so_server->nfs_client;
1425         bool need_recover = false;
1426
1427         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1428                 need_recover = true;
1429         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1430                 need_recover = true;
1431         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1432                 need_recover = true;
1433         if (need_recover)
1434                 nfs4_state_mark_reclaim_nograce(clp, state);
1435 }
1436
1437 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1438                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1439 {
1440         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1441                 return true;
1442         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1443                 nfs4_stateid_copy(freeme, &state->open_stateid);
1444                 nfs_test_and_clear_all_open_stateid(state);
1445                 return true;
1446         }
1447         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1448                 return true;
1449         return false;
1450 }
1451
1452 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1453 {
1454         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1455                 return;
1456         if (state->n_wronly)
1457                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1458         if (state->n_rdonly)
1459                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1460         if (state->n_rdwr)
1461                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1462         set_bit(NFS_OPEN_STATE, &state->flags);
1463 }
1464
1465 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1466                 nfs4_stateid *stateid, fmode_t fmode)
1467 {
1468         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1469         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1470         case FMODE_WRITE:
1471                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1472                 break;
1473         case FMODE_READ:
1474                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1475                 break;
1476         case 0:
1477                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1478                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1479                 clear_bit(NFS_OPEN_STATE, &state->flags);
1480         }
1481         if (stateid == NULL)
1482                 return;
1483         /* Handle OPEN+OPEN_DOWNGRADE races */
1484         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1485             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1486                 nfs_resync_open_stateid_locked(state);
1487                 return;
1488         }
1489         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1490                 nfs4_stateid_copy(&state->stateid, stateid);
1491         nfs4_stateid_copy(&state->open_stateid, stateid);
1492 }
1493
1494 static void nfs_clear_open_stateid(struct nfs4_state *state,
1495         nfs4_stateid *arg_stateid,
1496         nfs4_stateid *stateid, fmode_t fmode)
1497 {
1498         write_seqlock(&state->seqlock);
1499         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1500         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1501                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1502         write_sequnlock(&state->seqlock);
1503         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1504                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1505 }
1506
1507 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1508                 const nfs4_stateid *stateid, fmode_t fmode,
1509                 nfs4_stateid *freeme)
1510 {
1511         switch (fmode) {
1512                 case FMODE_READ:
1513                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1514                         break;
1515                 case FMODE_WRITE:
1516                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1517                         break;
1518                 case FMODE_READ|FMODE_WRITE:
1519                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1520         }
1521         if (!nfs_need_update_open_stateid(state, stateid, freeme))
1522                 return;
1523         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1524                 nfs4_stateid_copy(&state->stateid, stateid);
1525         nfs4_stateid_copy(&state->open_stateid, stateid);
1526 }
1527
1528 static void __update_open_stateid(struct nfs4_state *state,
1529                 const nfs4_stateid *open_stateid,
1530                 const nfs4_stateid *deleg_stateid,
1531                 fmode_t fmode,
1532                 nfs4_stateid *freeme)
1533 {
1534         /*
1535          * Protect the call to nfs4_state_set_mode_locked and
1536          * serialise the stateid update
1537          */
1538         spin_lock(&state->owner->so_lock);
1539         write_seqlock(&state->seqlock);
1540         if (deleg_stateid != NULL) {
1541                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1542                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1543         }
1544         if (open_stateid != NULL)
1545                 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1546         write_sequnlock(&state->seqlock);
1547         update_open_stateflags(state, fmode);
1548         spin_unlock(&state->owner->so_lock);
1549 }
1550
1551 static int update_open_stateid(struct nfs4_state *state,
1552                 const nfs4_stateid *open_stateid,
1553                 const nfs4_stateid *delegation,
1554                 fmode_t fmode)
1555 {
1556         struct nfs_server *server = NFS_SERVER(state->inode);
1557         struct nfs_client *clp = server->nfs_client;
1558         struct nfs_inode *nfsi = NFS_I(state->inode);
1559         struct nfs_delegation *deleg_cur;
1560         nfs4_stateid freeme = { };
1561         int ret = 0;
1562
1563         fmode &= (FMODE_READ|FMODE_WRITE);
1564
1565         rcu_read_lock();
1566         deleg_cur = rcu_dereference(nfsi->delegation);
1567         if (deleg_cur == NULL)
1568                 goto no_delegation;
1569
1570         spin_lock(&deleg_cur->lock);
1571         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1572            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1573             (deleg_cur->type & fmode) != fmode)
1574                 goto no_delegation_unlock;
1575
1576         if (delegation == NULL)
1577                 delegation = &deleg_cur->stateid;
1578         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1579                 goto no_delegation_unlock;
1580
1581         nfs_mark_delegation_referenced(deleg_cur);
1582         __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1583                         fmode, &freeme);
1584         ret = 1;
1585 no_delegation_unlock:
1586         spin_unlock(&deleg_cur->lock);
1587 no_delegation:
1588         rcu_read_unlock();
1589
1590         if (!ret && open_stateid != NULL) {
1591                 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1592                 ret = 1;
1593         }
1594         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1595                 nfs4_schedule_state_manager(clp);
1596         if (freeme.type != 0)
1597                 nfs4_test_and_free_stateid(server, &freeme,
1598                                 state->owner->so_cred);
1599
1600         return ret;
1601 }
1602
1603 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1604                 const nfs4_stateid *stateid)
1605 {
1606         struct nfs4_state *state = lsp->ls_state;
1607         bool ret = false;
1608
1609         spin_lock(&state->state_lock);
1610         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1611                 goto out_noupdate;
1612         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1613                 goto out_noupdate;
1614         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1615         ret = true;
1616 out_noupdate:
1617         spin_unlock(&state->state_lock);
1618         return ret;
1619 }
1620
1621 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1622 {
1623         struct nfs_delegation *delegation;
1624
1625         rcu_read_lock();
1626         delegation = rcu_dereference(NFS_I(inode)->delegation);
1627         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1628                 rcu_read_unlock();
1629                 return;
1630         }
1631         rcu_read_unlock();
1632         nfs4_inode_return_delegation(inode);
1633 }
1634
1635 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1636 {
1637         struct nfs4_state *state = opendata->state;
1638         struct nfs_inode *nfsi = NFS_I(state->inode);
1639         struct nfs_delegation *delegation;
1640         int open_mode = opendata->o_arg.open_flags;
1641         fmode_t fmode = opendata->o_arg.fmode;
1642         enum open_claim_type4 claim = opendata->o_arg.claim;
1643         nfs4_stateid stateid;
1644         int ret = -EAGAIN;
1645
1646         for (;;) {
1647                 spin_lock(&state->owner->so_lock);
1648                 if (can_open_cached(state, fmode, open_mode)) {
1649                         update_open_stateflags(state, fmode);
1650                         spin_unlock(&state->owner->so_lock);
1651                         goto out_return_state;
1652                 }
1653                 spin_unlock(&state->owner->so_lock);
1654                 rcu_read_lock();
1655                 delegation = rcu_dereference(nfsi->delegation);
1656                 if (!can_open_delegated(delegation, fmode, claim)) {
1657                         rcu_read_unlock();
1658                         break;
1659                 }
1660                 /* Save the delegation */
1661                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1662                 rcu_read_unlock();
1663                 nfs_release_seqid(opendata->o_arg.seqid);
1664                 if (!opendata->is_recover) {
1665                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1666                         if (ret != 0)
1667                                 goto out;
1668                 }
1669                 ret = -EAGAIN;
1670
1671                 /* Try to update the stateid using the delegation */
1672                 if (update_open_stateid(state, NULL, &stateid, fmode))
1673                         goto out_return_state;
1674         }
1675 out:
1676         return ERR_PTR(ret);
1677 out_return_state:
1678         atomic_inc(&state->count);
1679         return state;
1680 }
1681
1682 static void
1683 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1684 {
1685         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1686         struct nfs_delegation *delegation;
1687         int delegation_flags = 0;
1688
1689         rcu_read_lock();
1690         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1691         if (delegation)
1692                 delegation_flags = delegation->flags;
1693         rcu_read_unlock();
1694         switch (data->o_arg.claim) {
1695         default:
1696                 break;
1697         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1698         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1699                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1700                                    "returning a delegation for "
1701                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1702                                    clp->cl_hostname);
1703                 return;
1704         }
1705         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1706                 nfs_inode_set_delegation(state->inode,
1707                                          data->owner->so_cred,
1708                                          &data->o_res);
1709         else
1710                 nfs_inode_reclaim_delegation(state->inode,
1711                                              data->owner->so_cred,
1712                                              &data->o_res);
1713 }
1714
1715 /*
1716  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1717  * and update the nfs4_state.
1718  */
1719 static struct nfs4_state *
1720 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1721 {
1722         struct inode *inode = data->state->inode;
1723         struct nfs4_state *state = data->state;
1724         int ret;
1725
1726         if (!data->rpc_done) {
1727                 if (data->rpc_status) {
1728                         ret = data->rpc_status;
1729                         goto err;
1730                 }
1731                 /* cached opens have already been processed */
1732                 goto update;
1733         }
1734
1735         ret = nfs_refresh_inode(inode, &data->f_attr);
1736         if (ret)
1737                 goto err;
1738
1739         if (data->o_res.delegation_type != 0)
1740                 nfs4_opendata_check_deleg(data, state);
1741 update:
1742         update_open_stateid(state, &data->o_res.stateid, NULL,
1743                             data->o_arg.fmode);
1744         atomic_inc(&state->count);
1745
1746         return state;
1747 err:
1748         return ERR_PTR(ret);
1749
1750 }
1751
1752 static struct nfs4_state *
1753 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1754 {
1755         struct inode *inode;
1756         struct nfs4_state *state = NULL;
1757         int ret;
1758
1759         if (!data->rpc_done) {
1760                 state = nfs4_try_open_cached(data);
1761                 trace_nfs4_cached_open(data->state);
1762                 goto out;
1763         }
1764
1765         ret = -EAGAIN;
1766         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1767                 goto err;
1768         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1769         ret = PTR_ERR(inode);
1770         if (IS_ERR(inode))
1771                 goto err;
1772         ret = -ENOMEM;
1773         state = nfs4_get_open_state(inode, data->owner);
1774         if (state == NULL)
1775                 goto err_put_inode;
1776         if (data->o_res.delegation_type != 0)
1777                 nfs4_opendata_check_deleg(data, state);
1778         update_open_stateid(state, &data->o_res.stateid, NULL,
1779                         data->o_arg.fmode);
1780         iput(inode);
1781 out:
1782         nfs_release_seqid(data->o_arg.seqid);
1783         return state;
1784 err_put_inode:
1785         iput(inode);
1786 err:
1787         return ERR_PTR(ret);
1788 }
1789
1790 static struct nfs4_state *
1791 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1792 {
1793         struct nfs4_state *ret;
1794
1795         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1796                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1797         else
1798                 ret = _nfs4_opendata_to_nfs4_state(data);
1799         nfs4_sequence_free_slot(&data->o_res.seq_res);
1800         return ret;
1801 }
1802
1803 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1804 {
1805         struct nfs_inode *nfsi = NFS_I(state->inode);
1806         struct nfs_open_context *ctx;
1807
1808         spin_lock(&state->inode->i_lock);
1809         list_for_each_entry(ctx, &nfsi->open_files, list) {
1810                 if (ctx->state != state)
1811                         continue;
1812                 get_nfs_open_context(ctx);
1813                 spin_unlock(&state->inode->i_lock);
1814                 return ctx;
1815         }
1816         spin_unlock(&state->inode->i_lock);
1817         return ERR_PTR(-ENOENT);
1818 }
1819
1820 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1821                 struct nfs4_state *state, enum open_claim_type4 claim)
1822 {
1823         struct nfs4_opendata *opendata;
1824
1825         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1826                         NULL, NULL, claim, GFP_NOFS);
1827         if (opendata == NULL)
1828                 return ERR_PTR(-ENOMEM);
1829         opendata->state = state;
1830         atomic_inc(&state->count);
1831         return opendata;
1832 }
1833
1834 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1835                 fmode_t fmode)
1836 {
1837         struct nfs4_state *newstate;
1838         int ret;
1839
1840         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1841                 return 0;
1842         opendata->o_arg.open_flags = 0;
1843         opendata->o_arg.fmode = fmode;
1844         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1845                         NFS_SB(opendata->dentry->d_sb),
1846                         fmode, 0);
1847         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1848         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1849         nfs4_init_opendata_res(opendata);
1850         ret = _nfs4_recover_proc_open(opendata);
1851         if (ret != 0)
1852                 return ret; 
1853         newstate = nfs4_opendata_to_nfs4_state(opendata);
1854         if (IS_ERR(newstate))
1855                 return PTR_ERR(newstate);
1856         if (newstate != opendata->state)
1857                 ret = -ESTALE;
1858         nfs4_close_state(newstate, fmode);
1859         return ret;
1860 }
1861
1862 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1863 {
1864         int ret;
1865
1866         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1867         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1868         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1869         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1870         /* memory barrier prior to reading state->n_* */
1871         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1872         clear_bit(NFS_OPEN_STATE, &state->flags);
1873         smp_rmb();
1874         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1875         if (ret != 0)
1876                 return ret;
1877         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1878         if (ret != 0)
1879                 return ret;
1880         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1881         if (ret != 0)
1882                 return ret;
1883         /*
1884          * We may have performed cached opens for all three recoveries.
1885          * Check if we need to update the current stateid.
1886          */
1887         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1888             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1889                 write_seqlock(&state->seqlock);
1890                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1891                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1892                 write_sequnlock(&state->seqlock);
1893         }
1894         return 0;
1895 }
1896
1897 /*
1898  * OPEN_RECLAIM:
1899  *      reclaim state on the server after a reboot.
1900  */
1901 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1902 {
1903         struct nfs_delegation *delegation;
1904         struct nfs4_opendata *opendata;
1905         fmode_t delegation_type = 0;
1906         int status;
1907
1908         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1909                         NFS4_OPEN_CLAIM_PREVIOUS);
1910         if (IS_ERR(opendata))
1911                 return PTR_ERR(opendata);
1912         rcu_read_lock();
1913         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1914         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1915                 delegation_type = delegation->type;
1916         rcu_read_unlock();
1917         opendata->o_arg.u.delegation_type = delegation_type;
1918         status = nfs4_open_recover(opendata, state);
1919         nfs4_opendata_put(opendata);
1920         return status;
1921 }
1922
1923 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1924 {
1925         struct nfs_server *server = NFS_SERVER(state->inode);
1926         struct nfs4_exception exception = { };
1927         int err;
1928         do {
1929                 err = _nfs4_do_open_reclaim(ctx, state);
1930                 trace_nfs4_open_reclaim(ctx, 0, err);
1931                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1932                         continue;
1933                 if (err != -NFS4ERR_DELAY)
1934                         break;
1935                 nfs4_handle_exception(server, err, &exception);
1936         } while (exception.retry);
1937         return err;
1938 }
1939
1940 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1941 {
1942         struct nfs_open_context *ctx;
1943         int ret;
1944
1945         ctx = nfs4_state_find_open_context(state);
1946         if (IS_ERR(ctx))
1947                 return -EAGAIN;
1948         ret = nfs4_do_open_reclaim(ctx, state);
1949         put_nfs_open_context(ctx);
1950         return ret;
1951 }
1952
1953 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1954 {
1955         switch (err) {
1956                 default:
1957                         printk(KERN_ERR "NFS: %s: unhandled error "
1958                                         "%d.\n", __func__, err);
1959                 case 0:
1960                 case -ENOENT:
1961                 case -EAGAIN:
1962                 case -ESTALE:
1963                         break;
1964                 case -NFS4ERR_BADSESSION:
1965                 case -NFS4ERR_BADSLOT:
1966                 case -NFS4ERR_BAD_HIGH_SLOT:
1967                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1968                 case -NFS4ERR_DEADSESSION:
1969                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1970                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1971                         return -EAGAIN;
1972                 case -NFS4ERR_STALE_CLIENTID:
1973                 case -NFS4ERR_STALE_STATEID:
1974                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1975                         /* Don't recall a delegation if it was lost */
1976                         nfs4_schedule_lease_recovery(server->nfs_client);
1977                         return -EAGAIN;
1978                 case -NFS4ERR_MOVED:
1979                         nfs4_schedule_migration_recovery(server);
1980                         return -EAGAIN;
1981                 case -NFS4ERR_LEASE_MOVED:
1982                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1983                         return -EAGAIN;
1984                 case -NFS4ERR_DELEG_REVOKED:
1985                 case -NFS4ERR_ADMIN_REVOKED:
1986                 case -NFS4ERR_EXPIRED:
1987                 case -NFS4ERR_BAD_STATEID:
1988                 case -NFS4ERR_OPENMODE:
1989                         nfs_inode_find_state_and_recover(state->inode,
1990                                         stateid);
1991                         nfs4_schedule_stateid_recovery(server, state);
1992                         return -EAGAIN;
1993                 case -NFS4ERR_DELAY:
1994                 case -NFS4ERR_GRACE:
1995                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1996                         ssleep(1);
1997                         return -EAGAIN;
1998                 case -ENOMEM:
1999                 case -NFS4ERR_DENIED:
2000                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
2001                         return 0;
2002         }
2003         return err;
2004 }
2005
2006 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2007                 struct nfs4_state *state, const nfs4_stateid *stateid,
2008                 fmode_t type)
2009 {
2010         struct nfs_server *server = NFS_SERVER(state->inode);
2011         struct nfs4_opendata *opendata;
2012         int err = 0;
2013
2014         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2015                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2016         if (IS_ERR(opendata))
2017                 return PTR_ERR(opendata);
2018         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2019         write_seqlock(&state->seqlock);
2020         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2021         write_sequnlock(&state->seqlock);
2022         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2023         switch (type & (FMODE_READ|FMODE_WRITE)) {
2024         case FMODE_READ|FMODE_WRITE:
2025         case FMODE_WRITE:
2026                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2027                 if (err)
2028                         break;
2029                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2030                 if (err)
2031                         break;
2032         case FMODE_READ:
2033                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2034         }
2035         nfs4_opendata_put(opendata);
2036         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
2037 }
2038
2039 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2040 {
2041         struct nfs4_opendata *data = calldata;
2042
2043         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
2044                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
2045 }
2046
2047 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2048 {
2049         struct nfs4_opendata *data = calldata;
2050
2051         nfs40_sequence_done(task, &data->c_res.seq_res);
2052
2053         data->rpc_status = task->tk_status;
2054         if (data->rpc_status == 0) {
2055                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2056                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2057                 renew_lease(data->o_res.server, data->timestamp);
2058                 data->rpc_done = 1;
2059         }
2060 }
2061
2062 static void nfs4_open_confirm_release(void *calldata)
2063 {
2064         struct nfs4_opendata *data = calldata;
2065         struct nfs4_state *state = NULL;
2066
2067         /* If this request hasn't been cancelled, do nothing */
2068         if (data->cancelled == 0)
2069                 goto out_free;
2070         /* In case of error, no cleanup! */
2071         if (!data->rpc_done)
2072                 goto out_free;
2073         state = nfs4_opendata_to_nfs4_state(data);
2074         if (!IS_ERR(state))
2075                 nfs4_close_state(state, data->o_arg.fmode);
2076 out_free:
2077         nfs4_opendata_put(data);
2078 }
2079
2080 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2081         .rpc_call_prepare = nfs4_open_confirm_prepare,
2082         .rpc_call_done = nfs4_open_confirm_done,
2083         .rpc_release = nfs4_open_confirm_release,
2084 };
2085
2086 /*
2087  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2088  */
2089 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2090 {
2091         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2092         struct rpc_task *task;
2093         struct  rpc_message msg = {
2094                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2095                 .rpc_argp = &data->c_arg,
2096                 .rpc_resp = &data->c_res,
2097                 .rpc_cred = data->owner->so_cred,
2098         };
2099         struct rpc_task_setup task_setup_data = {
2100                 .rpc_client = server->client,
2101                 .rpc_message = &msg,
2102                 .callback_ops = &nfs4_open_confirm_ops,
2103                 .callback_data = data,
2104                 .workqueue = nfsiod_workqueue,
2105                 .flags = RPC_TASK_ASYNC,
2106         };
2107         int status;
2108
2109         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2110         kref_get(&data->kref);
2111         data->rpc_done = 0;
2112         data->rpc_status = 0;
2113         data->timestamp = jiffies;
2114         if (data->is_recover)
2115                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2116         task = rpc_run_task(&task_setup_data);
2117         if (IS_ERR(task))
2118                 return PTR_ERR(task);
2119         status = nfs4_wait_for_completion_rpc_task(task);
2120         if (status != 0) {
2121                 data->cancelled = 1;
2122                 smp_wmb();
2123         } else
2124                 status = data->rpc_status;
2125         rpc_put_task(task);
2126         return status;
2127 }
2128
2129 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2130 {
2131         struct nfs4_opendata *data = calldata;
2132         struct nfs4_state_owner *sp = data->owner;
2133         struct nfs_client *clp = sp->so_server->nfs_client;
2134         enum open_claim_type4 claim = data->o_arg.claim;
2135
2136         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2137                 goto out_wait;
2138         /*
2139          * Check if we still need to send an OPEN call, or if we can use
2140          * a delegation instead.
2141          */
2142         if (data->state != NULL) {
2143                 struct nfs_delegation *delegation;
2144
2145                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2146                         goto out_no_action;
2147                 rcu_read_lock();
2148                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2149                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2150                         goto unlock_no_action;
2151                 rcu_read_unlock();
2152         }
2153         /* Update client id. */
2154         data->o_arg.clientid = clp->cl_clientid;
2155         switch (claim) {
2156         default:
2157                 break;
2158         case NFS4_OPEN_CLAIM_PREVIOUS:
2159         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2160         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2161                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2162         case NFS4_OPEN_CLAIM_FH:
2163                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2164                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2165         }
2166         data->timestamp = jiffies;
2167         if (nfs4_setup_sequence(data->o_arg.server,
2168                                 &data->o_arg.seq_args,
2169                                 &data->o_res.seq_res,
2170                                 task) != 0)
2171                 nfs_release_seqid(data->o_arg.seqid);
2172
2173         /* Set the create mode (note dependency on the session type) */
2174         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2175         if (data->o_arg.open_flags & O_EXCL) {
2176                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2177                 if (nfs4_has_persistent_session(clp))
2178                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2179                 else if (clp->cl_mvops->minor_version > 0)
2180                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2181         }
2182         return;
2183 unlock_no_action:
2184         trace_nfs4_cached_open(data->state);
2185         rcu_read_unlock();
2186 out_no_action:
2187         task->tk_action = NULL;
2188 out_wait:
2189         nfs4_sequence_done(task, &data->o_res.seq_res);
2190 }
2191
2192 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2193 {
2194         struct nfs4_opendata *data = calldata;
2195
2196         data->rpc_status = task->tk_status;
2197
2198         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2199                 return;
2200
2201         if (task->tk_status == 0) {
2202                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2203                         switch (data->o_res.f_attr->mode & S_IFMT) {
2204                         case S_IFREG:
2205                                 break;
2206                         case S_IFLNK:
2207                                 data->rpc_status = -ELOOP;
2208                                 break;
2209                         case S_IFDIR:
2210                                 data->rpc_status = -EISDIR;
2211                                 break;
2212                         default:
2213                                 data->rpc_status = -ENOTDIR;
2214                         }
2215                 }
2216                 renew_lease(data->o_res.server, data->timestamp);
2217                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2218                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2219         }
2220         data->rpc_done = 1;
2221 }
2222
2223 static void nfs4_open_release(void *calldata)
2224 {
2225         struct nfs4_opendata *data = calldata;
2226         struct nfs4_state *state = NULL;
2227
2228         /* If this request hasn't been cancelled, do nothing */
2229         if (data->cancelled == 0)
2230                 goto out_free;
2231         /* In case of error, no cleanup! */
2232         if (data->rpc_status != 0 || !data->rpc_done)
2233                 goto out_free;
2234         /* In case we need an open_confirm, no cleanup! */
2235         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2236                 goto out_free;
2237         state = nfs4_opendata_to_nfs4_state(data);
2238         if (!IS_ERR(state))
2239                 nfs4_close_state(state, data->o_arg.fmode);
2240 out_free:
2241         nfs4_opendata_put(data);
2242 }
2243
2244 static const struct rpc_call_ops nfs4_open_ops = {
2245         .rpc_call_prepare = nfs4_open_prepare,
2246         .rpc_call_done = nfs4_open_done,
2247         .rpc_release = nfs4_open_release,
2248 };
2249
2250 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2251 {
2252         struct inode *dir = d_inode(data->dir);
2253         struct nfs_server *server = NFS_SERVER(dir);
2254         struct nfs_openargs *o_arg = &data->o_arg;
2255         struct nfs_openres *o_res = &data->o_res;
2256         struct rpc_task *task;
2257         struct rpc_message msg = {
2258                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2259                 .rpc_argp = o_arg,
2260                 .rpc_resp = o_res,
2261                 .rpc_cred = data->owner->so_cred,
2262         };
2263         struct rpc_task_setup task_setup_data = {
2264                 .rpc_client = server->client,
2265                 .rpc_message = &msg,
2266                 .callback_ops = &nfs4_open_ops,
2267                 .callback_data = data,
2268                 .workqueue = nfsiod_workqueue,
2269                 .flags = RPC_TASK_ASYNC,
2270         };
2271         int status;
2272
2273         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2274         kref_get(&data->kref);
2275         data->rpc_done = 0;
2276         data->rpc_status = 0;
2277         data->cancelled = 0;
2278         data->is_recover = 0;
2279         if (isrecover) {
2280                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2281                 data->is_recover = 1;
2282         }
2283         task = rpc_run_task(&task_setup_data);
2284         if (IS_ERR(task))
2285                 return PTR_ERR(task);
2286         status = nfs4_wait_for_completion_rpc_task(task);
2287         if (status != 0) {
2288                 data->cancelled = 1;
2289                 smp_wmb();
2290         } else
2291                 status = data->rpc_status;
2292         rpc_put_task(task);
2293
2294         return status;
2295 }
2296
2297 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2298 {
2299         struct inode *dir = d_inode(data->dir);
2300         struct nfs_openres *o_res = &data->o_res;
2301         int status;
2302
2303         status = nfs4_run_open_task(data, 1);
2304         if (status != 0 || !data->rpc_done)
2305                 return status;
2306
2307         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2308
2309         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2310                 status = _nfs4_proc_open_confirm(data);
2311                 if (status != 0)
2312                         return status;
2313         }
2314
2315         return status;
2316 }
2317
2318 /*
2319  * Additional permission checks in order to distinguish between an
2320  * open for read, and an open for execute. This works around the
2321  * fact that NFSv4 OPEN treats read and execute permissions as being
2322  * the same.
2323  * Note that in the non-execute case, we want to turn off permission
2324  * checking if we just created a new file (POSIX open() semantics).
2325  */
2326 static int nfs4_opendata_access(struct rpc_cred *cred,
2327                                 struct nfs4_opendata *opendata,
2328                                 struct nfs4_state *state, fmode_t fmode,
2329                                 int openflags)
2330 {
2331         struct nfs_access_entry cache;
2332         u32 mask;
2333
2334         /* access call failed or for some reason the server doesn't
2335          * support any access modes -- defer access call until later */
2336         if (opendata->o_res.access_supported == 0)
2337                 return 0;
2338
2339         mask = 0;
2340         /*
2341          * Use openflags to check for exec, because fmode won't
2342          * always have FMODE_EXEC set when file open for exec.
2343          */
2344         if (openflags & __FMODE_EXEC) {
2345                 /* ONLY check for exec rights */
2346                 mask = MAY_EXEC;
2347         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2348                 mask = MAY_READ;
2349
2350         cache.cred = cred;
2351         cache.jiffies = jiffies;
2352         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2353         nfs_access_add_cache(state->inode, &cache);
2354
2355         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2356                 return 0;
2357
2358         /* even though OPEN succeeded, access is denied. Close the file */
2359         nfs4_close_state(state, fmode);
2360         return -EACCES;
2361 }
2362
2363 /*
2364  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2365  */
2366 static int _nfs4_proc_open(struct nfs4_opendata *data)
2367 {
2368         struct inode *dir = d_inode(data->dir);
2369         struct nfs_server *server = NFS_SERVER(dir);
2370         struct nfs_openargs *o_arg = &data->o_arg;
2371         struct nfs_openres *o_res = &data->o_res;
2372         int status;
2373
2374         status = nfs4_run_open_task(data, 0);
2375         if (!data->rpc_done)
2376                 return status;
2377         if (status != 0) {
2378                 if (status == -NFS4ERR_BADNAME &&
2379                                 !(o_arg->open_flags & O_CREAT))
2380                         return -ENOENT;
2381                 return status;
2382         }
2383
2384         nfs_fattr_map_and_free_names(server, &data->f_attr);
2385
2386         if (o_arg->open_flags & O_CREAT) {
2387                 update_changeattr(dir, &o_res->cinfo);
2388                 if (o_arg->open_flags & O_EXCL)
2389                         data->file_created = 1;
2390                 else if (o_res->cinfo.before != o_res->cinfo.after)
2391                         data->file_created = 1;
2392         }
2393         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2394                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2395         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2396                 status = _nfs4_proc_open_confirm(data);
2397                 if (status != 0)
2398                         return status;
2399         }
2400         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2401                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2402         return 0;
2403 }
2404
2405 static int nfs4_recover_expired_lease(struct nfs_server *server)
2406 {
2407         return nfs4_client_recover_expired_lease(server->nfs_client);
2408 }
2409
2410 /*
2411  * OPEN_EXPIRED:
2412  *      reclaim state on the server after a network partition.
2413  *      Assumes caller holds the appropriate lock
2414  */
2415 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2416 {
2417         struct nfs4_opendata *opendata;
2418         int ret;
2419
2420         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2421                         NFS4_OPEN_CLAIM_FH);
2422         if (IS_ERR(opendata))
2423                 return PTR_ERR(opendata);
2424         ret = nfs4_open_recover(opendata, state);
2425         if (ret == -ESTALE)
2426                 d_drop(ctx->dentry);
2427         nfs4_opendata_put(opendata);
2428         return ret;
2429 }
2430
2431 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2432 {
2433         struct nfs_server *server = NFS_SERVER(state->inode);
2434         struct nfs4_exception exception = { };
2435         int err;
2436
2437         do {
2438                 err = _nfs4_open_expired(ctx, state);
2439                 trace_nfs4_open_expired(ctx, 0, err);
2440                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2441                         continue;
2442                 switch (err) {
2443                 default:
2444                         goto out;
2445                 case -NFS4ERR_GRACE:
2446                 case -NFS4ERR_DELAY:
2447                         nfs4_handle_exception(server, err, &exception);
2448                         err = 0;
2449                 }
2450         } while (exception.retry);
2451 out:
2452         return err;
2453 }
2454
2455 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2456 {
2457         struct nfs_open_context *ctx;
2458         int ret;
2459
2460         ctx = nfs4_state_find_open_context(state);
2461         if (IS_ERR(ctx))
2462                 return -EAGAIN;
2463         ret = nfs4_do_open_expired(ctx, state);
2464         put_nfs_open_context(ctx);
2465         return ret;
2466 }
2467
2468 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2469                 const nfs4_stateid *stateid)
2470 {
2471         nfs_remove_bad_delegation(state->inode, stateid);
2472         write_seqlock(&state->seqlock);
2473         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2474         write_sequnlock(&state->seqlock);
2475         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2476 }
2477
2478 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2479 {
2480         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2481                 nfs_finish_clear_delegation_stateid(state, NULL);
2482 }
2483
2484 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2485 {
2486         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2487         nfs40_clear_delegation_stateid(state);
2488         return nfs4_open_expired(sp, state);
2489 }
2490
2491 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2492                 nfs4_stateid *stateid,
2493                 struct rpc_cred *cred)
2494 {
2495         return -NFS4ERR_BAD_STATEID;
2496 }
2497
2498 #if defined(CONFIG_NFS_V4_1)
2499 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2500                 nfs4_stateid *stateid,
2501                 struct rpc_cred *cred)
2502 {
2503         int status;
2504
2505         switch (stateid->type) {
2506         default:
2507                 break;
2508         case NFS4_INVALID_STATEID_TYPE:
2509         case NFS4_SPECIAL_STATEID_TYPE:
2510                 return -NFS4ERR_BAD_STATEID;
2511         case NFS4_REVOKED_STATEID_TYPE:
2512                 goto out_free;
2513         }
2514
2515         status = nfs41_test_stateid(server, stateid, cred);
2516         switch (status) {
2517         case -NFS4ERR_EXPIRED:
2518         case -NFS4ERR_ADMIN_REVOKED:
2519         case -NFS4ERR_DELEG_REVOKED:
2520                 break;
2521         default:
2522                 return status;
2523         }
2524 out_free:
2525         /* Ack the revoked state to the server */
2526         nfs41_free_stateid(server, stateid, cred, true);
2527         return -NFS4ERR_EXPIRED;
2528 }
2529
2530 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2531 {
2532         struct nfs_server *server = NFS_SERVER(state->inode);
2533         nfs4_stateid stateid;
2534         struct nfs_delegation *delegation;
2535         struct rpc_cred *cred;
2536         int status;
2537
2538         /* Get the delegation credential for use by test/free_stateid */
2539         rcu_read_lock();
2540         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2541         if (delegation == NULL) {
2542                 rcu_read_unlock();
2543                 return;
2544         }
2545
2546         nfs4_stateid_copy(&stateid, &delegation->stateid);
2547         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2548                 rcu_read_unlock();
2549                 nfs_finish_clear_delegation_stateid(state, &stateid);
2550                 return;
2551         }
2552
2553         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED, &delegation->flags)) {
2554                 rcu_read_unlock();
2555                 return;
2556         }
2557
2558         cred = get_rpccred(delegation->cred);
2559         rcu_read_unlock();
2560         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2561         trace_nfs4_test_delegation_stateid(state, NULL, status);
2562         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2563                 nfs_finish_clear_delegation_stateid(state, &stateid);
2564
2565         put_rpccred(cred);
2566 }
2567
2568 /**
2569  * nfs41_check_expired_locks - possibly free a lock stateid
2570  *
2571  * @state: NFSv4 state for an inode
2572  *
2573  * Returns NFS_OK if recovery for this stateid is now finished.
2574  * Otherwise a negative NFS4ERR value is returned.
2575  */
2576 static int nfs41_check_expired_locks(struct nfs4_state *state)
2577 {
2578         int status, ret = NFS_OK;
2579         struct nfs4_lock_state *lsp, *prev = NULL;
2580         struct nfs_server *server = NFS_SERVER(state->inode);
2581
2582         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2583                 goto out;
2584
2585         spin_lock(&state->state_lock);
2586         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2587                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2588                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2589
2590                         atomic_inc(&lsp->ls_count);
2591                         spin_unlock(&state->state_lock);
2592
2593                         nfs4_put_lock_state(prev);
2594                         prev = lsp;
2595
2596                         status = nfs41_test_and_free_expired_stateid(server,
2597                                         &lsp->ls_stateid,
2598                                         cred);
2599                         trace_nfs4_test_lock_stateid(state, lsp, status);
2600                         if (status == -NFS4ERR_EXPIRED ||
2601                             status == -NFS4ERR_BAD_STATEID) {
2602                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2603                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2604                                 if (!recover_lost_locks)
2605                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2606                         } else if (status != NFS_OK) {
2607                                 ret = status;
2608                                 nfs4_put_lock_state(prev);
2609                                 goto out;
2610                         }
2611                         spin_lock(&state->state_lock);
2612                 }
2613         }
2614         spin_unlock(&state->state_lock);
2615         nfs4_put_lock_state(prev);
2616 out:
2617         return ret;
2618 }
2619
2620 /**
2621  * nfs41_check_open_stateid - possibly free an open stateid
2622  *
2623  * @state: NFSv4 state for an inode
2624  *
2625  * Returns NFS_OK if recovery for this stateid is now finished.
2626  * Otherwise a negative NFS4ERR value is returned.
2627  */
2628 static int nfs41_check_open_stateid(struct nfs4_state *state)
2629 {
2630         struct nfs_server *server = NFS_SERVER(state->inode);
2631         nfs4_stateid *stateid = &state->open_stateid;
2632         struct rpc_cred *cred = state->owner->so_cred;
2633         int status;
2634
2635         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2636                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2637                         if (nfs4_have_delegation(state->inode, state->state))
2638                                 return NFS_OK;
2639                         return -NFS4ERR_OPENMODE;
2640                 }
2641                 return -NFS4ERR_BAD_STATEID;
2642         }
2643         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2644         trace_nfs4_test_open_stateid(state, NULL, status);
2645         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2646                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2647                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2648                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2649                 clear_bit(NFS_OPEN_STATE, &state->flags);
2650                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2651         }
2652         if (status != NFS_OK)
2653                 return status;
2654         if (nfs_open_stateid_recover_openmode(state))
2655                 return -NFS4ERR_OPENMODE;
2656         return NFS_OK;
2657 }
2658
2659 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2660 {
2661         int status;
2662
2663         nfs41_check_delegation_stateid(state);
2664         status = nfs41_check_expired_locks(state);
2665         if (status != NFS_OK)
2666                 return status;
2667         status = nfs41_check_open_stateid(state);
2668         if (status != NFS_OK)
2669                 status = nfs4_open_expired(sp, state);
2670         return status;
2671 }
2672 #endif
2673
2674 /*
2675  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2676  * fields corresponding to attributes that were used to store the verifier.
2677  * Make sure we clobber those fields in the later setattr call
2678  */
2679 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2680                                 struct iattr *sattr, struct nfs4_label **label)
2681 {
2682         const u32 *attrset = opendata->o_res.attrset;
2683
2684         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2685             !(sattr->ia_valid & ATTR_ATIME_SET))
2686                 sattr->ia_valid |= ATTR_ATIME;
2687
2688         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2689             !(sattr->ia_valid & ATTR_MTIME_SET))
2690                 sattr->ia_valid |= ATTR_MTIME;
2691
2692         /* Except MODE, it seems harmless of setting twice. */
2693         if ((attrset[1] & FATTR4_WORD1_MODE))
2694                 sattr->ia_valid &= ~ATTR_MODE;
2695
2696         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2697                 *label = NULL;
2698 }
2699
2700 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2701                 fmode_t fmode,
2702                 int flags,
2703                 struct nfs_open_context *ctx)
2704 {
2705         struct nfs4_state_owner *sp = opendata->owner;
2706         struct nfs_server *server = sp->so_server;
2707         struct dentry *dentry;
2708         struct nfs4_state *state;
2709         unsigned int seq;
2710         int ret;
2711
2712         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2713
2714         ret = _nfs4_proc_open(opendata);
2715         if (ret != 0)
2716                 goto out;
2717
2718         state = nfs4_opendata_to_nfs4_state(opendata);
2719         ret = PTR_ERR(state);
2720         if (IS_ERR(state))
2721                 goto out;
2722         if (server->caps & NFS_CAP_POSIX_LOCK)
2723                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2724         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2725                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2726
2727         dentry = opendata->dentry;
2728         if (d_really_is_negative(dentry)) {
2729                 struct dentry *alias;
2730                 d_drop(dentry);
2731                 alias = d_exact_alias(dentry, state->inode);
2732                 if (!alias)
2733                         alias = d_splice_alias(igrab(state->inode), dentry);
2734                 /* d_splice_alias() can't fail here - it's a non-directory */
2735                 if (alias) {
2736                         dput(ctx->dentry);
2737                         ctx->dentry = dentry = alias;
2738                 }
2739                 nfs_set_verifier(dentry,
2740                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2741         }
2742
2743         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2744         if (ret != 0)
2745                 goto out;
2746
2747         ctx->state = state;
2748         if (d_inode(dentry) == state->inode) {
2749                 nfs_inode_attach_open_context(ctx);
2750                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2751                         nfs4_schedule_stateid_recovery(server, state);
2752         }
2753 out:
2754         return ret;
2755 }
2756
2757 /*
2758  * Returns a referenced nfs4_state
2759  */
2760 static int _nfs4_do_open(struct inode *dir,
2761                         struct nfs_open_context *ctx,
2762                         int flags,
2763                         struct iattr *sattr,
2764                         struct nfs4_label *label,
2765                         int *opened)
2766 {
2767         struct nfs4_state_owner  *sp;
2768         struct nfs4_state     *state = NULL;
2769         struct nfs_server       *server = NFS_SERVER(dir);
2770         struct nfs4_opendata *opendata;
2771         struct dentry *dentry = ctx->dentry;
2772         struct rpc_cred *cred = ctx->cred;
2773         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2774         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2775         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2776         struct nfs4_label *olabel = NULL;
2777         int status;
2778
2779         /* Protect against reboot recovery conflicts */
2780         status = -ENOMEM;
2781         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2782         if (sp == NULL) {
2783                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2784                 goto out_err;
2785         }
2786         status = nfs4_recover_expired_lease(server);
2787         if (status != 0)
2788                 goto err_put_state_owner;
2789         if (d_really_is_positive(dentry))
2790                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2791         status = -ENOMEM;
2792         if (d_really_is_positive(dentry))
2793                 claim = NFS4_OPEN_CLAIM_FH;
2794         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2795                         label, claim, GFP_KERNEL);
2796         if (opendata == NULL)
2797                 goto err_put_state_owner;
2798
2799         if (label) {
2800                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2801                 if (IS_ERR(olabel)) {
2802                         status = PTR_ERR(olabel);
2803                         goto err_opendata_put;
2804                 }
2805         }
2806
2807         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2808                 if (!opendata->f_attr.mdsthreshold) {
2809                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2810                         if (!opendata->f_attr.mdsthreshold)
2811                                 goto err_free_label;
2812                 }
2813                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2814         }
2815         if (d_really_is_positive(dentry))
2816                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2817
2818         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2819         if (status != 0)
2820                 goto err_free_label;
2821         state = ctx->state;
2822
2823         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2824             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2825                 nfs4_exclusive_attrset(opendata, sattr, &label);
2826                 /*
2827                  * send create attributes which was not set by open
2828                  * with an extra setattr.
2829                  */
2830                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2831                         nfs_fattr_init(opendata->o_res.f_attr);
2832                         status = nfs4_do_setattr(state->inode, cred,
2833                                         opendata->o_res.f_attr, sattr,
2834                                         ctx, label, olabel);
2835                         if (status == 0) {
2836                                 nfs_setattr_update_inode(state->inode, sattr,
2837                                                 opendata->o_res.f_attr);
2838                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2839                         }
2840                 }
2841         }
2842         if (opened && opendata->file_created)
2843                 *opened |= FILE_CREATED;
2844
2845         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2846                 *ctx_th = opendata->f_attr.mdsthreshold;
2847                 opendata->f_attr.mdsthreshold = NULL;
2848         }
2849
2850         nfs4_label_free(olabel);
2851
2852         nfs4_opendata_put(opendata);
2853         nfs4_put_state_owner(sp);
2854         return 0;
2855 err_free_label:
2856         nfs4_label_free(olabel);
2857 err_opendata_put:
2858         nfs4_opendata_put(opendata);
2859 err_put_state_owner:
2860         nfs4_put_state_owner(sp);
2861 out_err:
2862         return status;
2863 }
2864
2865
2866 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2867                                         struct nfs_open_context *ctx,
2868                                         int flags,
2869                                         struct iattr *sattr,
2870                                         struct nfs4_label *label,
2871                                         int *opened)
2872 {
2873         struct nfs_server *server = NFS_SERVER(dir);
2874         struct nfs4_exception exception = { };
2875         struct nfs4_state *res;
2876         int status;
2877
2878         do {
2879                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2880                 res = ctx->state;
2881                 trace_nfs4_open_file(ctx, flags, status);
2882                 if (status == 0)
2883                         break;
2884                 /* NOTE: BAD_SEQID means the server and client disagree about the
2885                  * book-keeping w.r.t. state-changing operations
2886                  * (OPEN/CLOSE/LOCK/LOCKU...)
2887                  * It is actually a sign of a bug on the client or on the server.
2888                  *
2889                  * If we receive a BAD_SEQID error in the particular case of
2890                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2891                  * have unhashed the old state_owner for us, and that we can
2892                  * therefore safely retry using a new one. We should still warn
2893                  * the user though...
2894                  */
2895                 if (status == -NFS4ERR_BAD_SEQID) {
2896                         pr_warn_ratelimited("NFS: v4 server %s "
2897                                         " returned a bad sequence-id error!\n",
2898                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2899                         exception.retry = 1;
2900                         continue;
2901                 }
2902                 /*
2903                  * BAD_STATEID on OPEN means that the server cancelled our
2904                  * state before it received the OPEN_CONFIRM.
2905                  * Recover by retrying the request as per the discussion
2906                  * on Page 181 of RFC3530.
2907                  */
2908                 if (status == -NFS4ERR_BAD_STATEID) {
2909                         exception.retry = 1;
2910                         continue;
2911                 }
2912                 if (status == -EAGAIN) {
2913                         /* We must have found a delegation */
2914                         exception.retry = 1;
2915                         continue;
2916                 }
2917                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2918                         continue;
2919                 res = ERR_PTR(nfs4_handle_exception(server,
2920                                         status, &exception));
2921         } while (exception.retry);
2922         return res;
2923 }
2924
2925 static int _nfs4_do_setattr(struct inode *inode,
2926                             struct nfs_setattrargs *arg,
2927                             struct nfs_setattrres *res,
2928                             struct rpc_cred *cred,
2929                             struct nfs_open_context *ctx)
2930 {
2931         struct nfs_server *server = NFS_SERVER(inode);
2932         struct rpc_message msg = {
2933                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2934                 .rpc_argp       = arg,
2935                 .rpc_resp       = res,
2936                 .rpc_cred       = cred,
2937         };
2938         struct rpc_cred *delegation_cred = NULL;
2939         unsigned long timestamp = jiffies;
2940         fmode_t fmode;
2941         bool truncate;
2942         int status;
2943
2944         nfs_fattr_init(res->fattr);
2945
2946         /* Servers should only apply open mode checks for file size changes */
2947         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2948         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2949
2950         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2951                 /* Use that stateid */
2952         } else if (truncate && ctx != NULL) {
2953                 struct nfs_lock_context *l_ctx;
2954                 if (!nfs4_valid_open_stateid(ctx->state))
2955                         return -EBADF;
2956                 l_ctx = nfs_get_lock_context(ctx);
2957                 if (IS_ERR(l_ctx))
2958                         return PTR_ERR(l_ctx);
2959                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2960                                                 &arg->stateid, &delegation_cred);
2961                 nfs_put_lock_context(l_ctx);
2962                 if (status == -EIO)
2963                         return -EBADF;
2964         } else
2965                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2966         if (delegation_cred)
2967                 msg.rpc_cred = delegation_cred;
2968
2969         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2970
2971         put_rpccred(delegation_cred);
2972         if (status == 0 && ctx != NULL)
2973                 renew_lease(server, timestamp);
2974         trace_nfs4_setattr(inode, &arg->stateid, status);
2975         return status;
2976 }
2977
2978 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2979                            struct nfs_fattr *fattr, struct iattr *sattr,
2980                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2981                            struct nfs4_label *olabel)
2982 {
2983         struct nfs_server *server = NFS_SERVER(inode);
2984         struct nfs4_state *state = ctx ? ctx->state : NULL;
2985         struct nfs_setattrargs  arg = {
2986                 .fh             = NFS_FH(inode),
2987                 .iap            = sattr,
2988                 .server         = server,
2989                 .bitmask = server->attr_bitmask,
2990                 .label          = ilabel,
2991         };
2992         struct nfs_setattrres  res = {
2993                 .fattr          = fattr,
2994                 .label          = olabel,
2995                 .server         = server,
2996         };
2997         struct nfs4_exception exception = {
2998                 .state = state,
2999                 .inode = inode,
3000                 .stateid = &arg.stateid,
3001         };
3002         int err;
3003
3004         arg.bitmask = nfs4_bitmask(server, ilabel);
3005         if (ilabel)
3006                 arg.bitmask = nfs4_bitmask(server, olabel);
3007
3008         do {
3009                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3010                 switch (err) {
3011                 case -NFS4ERR_OPENMODE:
3012                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3013                                 pr_warn_once("NFSv4: server %s is incorrectly "
3014                                                 "applying open mode checks to "
3015                                                 "a SETATTR that is not "
3016                                                 "changing file size.\n",
3017                                                 server->nfs_client->cl_hostname);
3018                         }
3019                         if (state && !(state->state & FMODE_WRITE)) {
3020                                 err = -EBADF;
3021                                 if (sattr->ia_valid & ATTR_OPEN)
3022                                         err = -EACCES;
3023                                 goto out;
3024                         }
3025                 }
3026                 err = nfs4_handle_exception(server, err, &exception);
3027         } while (exception.retry);
3028 out:
3029         return err;
3030 }
3031
3032 static bool
3033 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3034 {
3035         if (inode == NULL || !nfs_have_layout(inode))
3036                 return false;
3037
3038         return pnfs_wait_on_layoutreturn(inode, task);
3039 }
3040
3041 struct nfs4_closedata {
3042         struct inode *inode;
3043         struct nfs4_state *state;
3044         struct nfs_closeargs arg;
3045         struct nfs_closeres res;
3046         struct {
3047                 struct nfs4_layoutreturn_args arg;
3048                 struct nfs4_layoutreturn_res res;
3049                 struct nfs4_xdr_opaque_data ld_private;
3050                 u32 roc_barrier;
3051                 bool roc;
3052         } lr;
3053         struct nfs_fattr fattr;
3054         unsigned long timestamp;
3055 };
3056
3057 static void nfs4_free_closedata(void *data)
3058 {
3059         struct nfs4_closedata *calldata = data;
3060         struct nfs4_state_owner *sp = calldata->state->owner;
3061         struct super_block *sb = calldata->state->inode->i_sb;
3062
3063         if (calldata->lr.roc)
3064                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3065                                 calldata->res.lr_ret);
3066         nfs4_put_open_state(calldata->state);
3067         nfs_free_seqid(calldata->arg.seqid);
3068         nfs4_put_state_owner(sp);
3069         nfs_sb_deactive(sb);
3070         kfree(calldata);
3071 }
3072
3073 static void nfs4_close_done(struct rpc_task *task, void *data)
3074 {
3075         struct nfs4_closedata *calldata = data;
3076         struct nfs4_state *state = calldata->state;
3077         struct nfs_server *server = NFS_SERVER(calldata->inode);
3078         nfs4_stateid *res_stateid = NULL;
3079
3080         dprintk("%s: begin!\n", __func__);
3081         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3082                 return;
3083         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3084
3085         /* Handle Layoutreturn errors */
3086         if (calldata->arg.lr_args && task->tk_status != 0) {
3087                 switch (calldata->res.lr_ret) {
3088                 default:
3089                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3090                         break;
3091                 case 0:
3092                         calldata->arg.lr_args = NULL;
3093                         calldata->res.lr_res = NULL;
3094                         break;
3095                 case -NFS4ERR_ADMIN_REVOKED:
3096                 case -NFS4ERR_DELEG_REVOKED:
3097                 case -NFS4ERR_EXPIRED:
3098                 case -NFS4ERR_BAD_STATEID:
3099                 case -NFS4ERR_OLD_STATEID:
3100                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3101                 case -NFS4ERR_WRONG_CRED:
3102                         calldata->arg.lr_args = NULL;
3103                         calldata->res.lr_res = NULL;
3104                         calldata->res.lr_ret = 0;
3105                         rpc_restart_call_prepare(task);
3106                         return;
3107                 }
3108         }
3109
3110         /* hmm. we are done with the inode, and in the process of freeing
3111          * the state_owner. we keep this around to process errors
3112          */
3113         switch (task->tk_status) {
3114                 case 0:
3115                         res_stateid = &calldata->res.stateid;
3116                         renew_lease(server, calldata->timestamp);
3117                         break;
3118                 case -NFS4ERR_ADMIN_REVOKED:
3119                 case -NFS4ERR_STALE_STATEID:
3120                 case -NFS4ERR_EXPIRED:
3121                         nfs4_free_revoked_stateid(server,
3122                                         &calldata->arg.stateid,
3123                                         task->tk_msg.rpc_cred);
3124                 case -NFS4ERR_OLD_STATEID:
3125                 case -NFS4ERR_BAD_STATEID:
3126                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3127                                                 &state->open_stateid)) {
3128                                 rpc_restart_call_prepare(task);
3129                                 goto out_release;
3130                         }
3131                         if (calldata->arg.fmode == 0)
3132                                 break;
3133                 default:
3134                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3135                                 rpc_restart_call_prepare(task);
3136                                 goto out_release;
3137                         }
3138         }
3139         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3140                         res_stateid, calldata->arg.fmode);
3141 out_release:
3142         nfs_release_seqid(calldata->arg.seqid);
3143         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
3144         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3145 }
3146
3147 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3148 {
3149         struct nfs4_closedata *calldata = data;
3150         struct nfs4_state *state = calldata->state;
3151         struct inode *inode = calldata->inode;
3152         bool is_rdonly, is_wronly, is_rdwr;
3153         int call_close = 0;
3154
3155         dprintk("%s: begin!\n", __func__);
3156         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3157                 goto out_wait;
3158
3159         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3160         spin_lock(&state->owner->so_lock);
3161         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3162         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3163         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3164         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3165         /* Calculate the change in open mode */
3166         calldata->arg.fmode = 0;
3167         if (state->n_rdwr == 0) {
3168                 if (state->n_rdonly == 0)
3169                         call_close |= is_rdonly;
3170                 else if (is_rdonly)
3171                         calldata->arg.fmode |= FMODE_READ;
3172                 if (state->n_wronly == 0)
3173                         call_close |= is_wronly;
3174                 else if (is_wronly)
3175                         calldata->arg.fmode |= FMODE_WRITE;
3176                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3177                         call_close |= is_rdwr;
3178         } else if (is_rdwr)
3179                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3180
3181         if (!nfs4_valid_open_stateid(state) ||
3182             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3183                 call_close = 0;
3184         spin_unlock(&state->owner->so_lock);
3185
3186         if (!call_close) {
3187                 /* Note: exit _without_ calling nfs4_close_done */
3188                 goto out_no_action;
3189         }
3190
3191         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3192                 nfs_release_seqid(calldata->arg.seqid);
3193                 goto out_wait;
3194         }
3195
3196         if (calldata->arg.fmode == 0) {
3197                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3198
3199                 /* Close-to-open cache consistency revalidation */
3200                 if (!nfs4_have_delegation(inode, FMODE_READ))
3201                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3202                 else
3203                         calldata->arg.bitmask = NULL;
3204         }
3205
3206         calldata->arg.share_access =
3207                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3208                                 calldata->arg.fmode, 0);
3209
3210         nfs_fattr_init(calldata->res.fattr);
3211         calldata->timestamp = jiffies;
3212         if (nfs4_setup_sequence(NFS_SERVER(inode),
3213                                 &calldata->arg.seq_args,
3214                                 &calldata->res.seq_res,
3215                                 task) != 0)
3216                 nfs_release_seqid(calldata->arg.seqid);
3217         dprintk("%s: done!\n", __func__);
3218         return;
3219 out_no_action:
3220         task->tk_action = NULL;
3221 out_wait:
3222         nfs4_sequence_done(task, &calldata->res.seq_res);
3223 }
3224
3225 static const struct rpc_call_ops nfs4_close_ops = {
3226         .rpc_call_prepare = nfs4_close_prepare,
3227         .rpc_call_done = nfs4_close_done,
3228         .rpc_release = nfs4_free_closedata,
3229 };
3230
3231 /* 
3232  * It is possible for data to be read/written from a mem-mapped file 
3233  * after the sys_close call (which hits the vfs layer as a flush).
3234  * This means that we can't safely call nfsv4 close on a file until 
3235  * the inode is cleared. This in turn means that we are not good
3236  * NFSv4 citizens - we do not indicate to the server to update the file's 
3237  * share state even when we are done with one of the three share 
3238  * stateid's in the inode.
3239  *
3240  * NOTE: Caller must be holding the sp->so_owner semaphore!
3241  */
3242 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3243 {
3244         struct nfs_server *server = NFS_SERVER(state->inode);
3245         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3246         struct nfs4_closedata *calldata;
3247         struct nfs4_state_owner *sp = state->owner;
3248         struct rpc_task *task;
3249         struct rpc_message msg = {
3250                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3251                 .rpc_cred = state->owner->so_cred,
3252         };
3253         struct rpc_task_setup task_setup_data = {
3254                 .rpc_client = server->client,
3255                 .rpc_message = &msg,
3256                 .callback_ops = &nfs4_close_ops,
3257                 .workqueue = nfsiod_workqueue,
3258                 .flags = RPC_TASK_ASYNC,
3259         };
3260         int status = -ENOMEM;
3261
3262         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3263                 &task_setup_data.rpc_client, &msg);
3264
3265         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3266         if (calldata == NULL)
3267                 goto out;
3268         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3269         calldata->inode = state->inode;
3270         calldata->state = state;
3271         calldata->arg.fh = NFS_FH(state->inode);
3272         /* Serialization for the sequence id */
3273         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3274         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3275         if (IS_ERR(calldata->arg.seqid))
3276                 goto out_free_calldata;
3277         calldata->arg.fmode = 0;
3278         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3279         calldata->res.fattr = &calldata->fattr;
3280         calldata->res.seqid = calldata->arg.seqid;
3281         calldata->res.server = server;
3282         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3283         calldata->lr.roc = pnfs_roc(state->inode,
3284                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3285         if (calldata->lr.roc) {
3286                 calldata->arg.lr_args = &calldata->lr.arg;
3287                 calldata->res.lr_res = &calldata->lr.res;
3288         }
3289         nfs_sb_active(calldata->inode->i_sb);
3290
3291         msg.rpc_argp = &calldata->arg;
3292         msg.rpc_resp = &calldata->res;
3293         task_setup_data.callback_data = calldata;
3294         task = rpc_run_task(&task_setup_data);
3295         if (IS_ERR(task))
3296                 return PTR_ERR(task);
3297         status = 0;
3298         if (wait)
3299                 status = rpc_wait_for_completion_task(task);
3300         rpc_put_task(task);
3301         return status;
3302 out_free_calldata:
3303         kfree(calldata);
3304 out:
3305         nfs4_put_open_state(state);
3306         nfs4_put_state_owner(sp);
3307         return status;
3308 }
3309
3310 static struct inode *
3311 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3312                 int open_flags, struct iattr *attr, int *opened)
3313 {
3314         struct nfs4_state *state;
3315         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3316
3317         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3318
3319         /* Protect against concurrent sillydeletes */
3320         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3321
3322         nfs4_label_release_security(label);
3323
3324         if (IS_ERR(state))
3325                 return ERR_CAST(state);
3326         return state->inode;
3327 }
3328
3329 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3330 {
3331         if (ctx->state == NULL)
3332                 return;
3333         if (is_sync)
3334                 nfs4_close_sync(ctx->state, ctx->mode);
3335         else
3336                 nfs4_close_state(ctx->state, ctx->mode);
3337 }
3338
3339 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3340 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3341 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3342
3343 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3344 {
3345         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3346         struct nfs4_server_caps_arg args = {
3347                 .fhandle = fhandle,
3348                 .bitmask = bitmask,
3349         };
3350         struct nfs4_server_caps_res res = {};
3351         struct rpc_message msg = {
3352                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3353                 .rpc_argp = &args,
3354                 .rpc_resp = &res,
3355         };
3356         int status;
3357
3358         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3359                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3360                      FATTR4_WORD0_LINK_SUPPORT |
3361                      FATTR4_WORD0_SYMLINK_SUPPORT |
3362                      FATTR4_WORD0_ACLSUPPORT;
3363         if (minorversion)
3364                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3365
3366         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3367         if (status == 0) {
3368                 /* Sanity check the server answers */
3369                 switch (minorversion) {
3370                 case 0:
3371                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3372                         res.attr_bitmask[2] = 0;
3373                         break;
3374                 case 1:
3375                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3376                         break;
3377                 case 2:
3378                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3379                 }
3380                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3381                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3382                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3383                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3384                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3385                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3386                                 NFS_CAP_SECURITY_LABEL);
3387                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3388                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3389                         server->caps |= NFS_CAP_ACLS;
3390                 if (res.has_links != 0)
3391                         server->caps |= NFS_CAP_HARDLINKS;
3392                 if (res.has_symlinks != 0)
3393                         server->caps |= NFS_CAP_SYMLINKS;
3394                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3395                         server->caps |= NFS_CAP_FILEID;
3396                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3397                         server->caps |= NFS_CAP_MODE;
3398                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3399                         server->caps |= NFS_CAP_NLINK;
3400                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3401                         server->caps |= NFS_CAP_OWNER;
3402                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3403                         server->caps |= NFS_CAP_OWNER_GROUP;
3404                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3405                         server->caps |= NFS_CAP_ATIME;
3406                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3407                         server->caps |= NFS_CAP_CTIME;
3408                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3409                         server->caps |= NFS_CAP_MTIME;
3410 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3411                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3412                         server->caps |= NFS_CAP_SECURITY_LABEL;
3413 #endif
3414                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3415                                 sizeof(server->attr_bitmask));
3416                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3417
3418                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3419                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3420                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3421                 server->cache_consistency_bitmask[2] = 0;
3422                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3423                         sizeof(server->exclcreat_bitmask));
3424                 server->acl_bitmask = res.acl_bitmask;
3425                 server->fh_expire_type = res.fh_expire_type;
3426         }
3427
3428         return status;
3429 }
3430
3431 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3432 {
3433         struct nfs4_exception exception = { };
3434         int err;
3435         do {
3436                 err = nfs4_handle_exception(server,
3437                                 _nfs4_server_capabilities(server, fhandle),
3438                                 &exception);
3439         } while (exception.retry);
3440         return err;
3441 }
3442
3443 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3444                 struct nfs_fsinfo *info)
3445 {
3446         u32 bitmask[3];
3447         struct nfs4_lookup_root_arg args = {
3448                 .bitmask = bitmask,
3449         };
3450         struct nfs4_lookup_res res = {
3451                 .server = server,
3452                 .fattr = info->fattr,
3453                 .fh = fhandle,
3454         };
3455         struct rpc_message msg = {
3456                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3457                 .rpc_argp = &args,
3458                 .rpc_resp = &res,
3459         };
3460
3461         bitmask[0] = nfs4_fattr_bitmap[0];
3462         bitmask[1] = nfs4_fattr_bitmap[1];
3463         /*
3464          * Process the label in the upcoming getfattr
3465          */
3466         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3467
3468         nfs_fattr_init(info->fattr);
3469         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3470 }
3471
3472 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3473                 struct nfs_fsinfo *info)
3474 {
3475         struct nfs4_exception exception = { };
3476         int err;
3477         do {
3478                 err = _nfs4_lookup_root(server, fhandle, info);
3479                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3480                 switch (err) {
3481                 case 0:
3482                 case -NFS4ERR_WRONGSEC:
3483                         goto out;
3484                 default:
3485                         err = nfs4_handle_exception(server, err, &exception);
3486                 }
3487         } while (exception.retry);
3488 out:
3489         return err;
3490 }
3491
3492 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3493                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3494 {
3495         struct rpc_auth_create_args auth_args = {
3496                 .pseudoflavor = flavor,
3497         };
3498         struct rpc_auth *auth;
3499         int ret;
3500
3501         auth = rpcauth_create(&auth_args, server->client);
3502         if (IS_ERR(auth)) {
3503                 ret = -EACCES;
3504                 goto out;
3505         }
3506         ret = nfs4_lookup_root(server, fhandle, info);
3507 out:
3508         return ret;
3509 }
3510
3511 /*
3512  * Retry pseudoroot lookup with various security flavors.  We do this when:
3513  *
3514  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3515  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3516  *
3517  * Returns zero on success, or a negative NFS4ERR value, or a
3518  * negative errno value.
3519  */
3520 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3521                               struct nfs_fsinfo *info)
3522 {
3523         /* Per 3530bis 15.33.5 */
3524         static const rpc_authflavor_t flav_array[] = {
3525                 RPC_AUTH_GSS_KRB5P,
3526                 RPC_AUTH_GSS_KRB5I,
3527                 RPC_AUTH_GSS_KRB5,
3528                 RPC_AUTH_UNIX,                  /* courtesy */
3529                 RPC_AUTH_NULL,
3530         };
3531         int status = -EPERM;
3532         size_t i;
3533
3534         if (server->auth_info.flavor_len > 0) {
3535                 /* try each flavor specified by user */
3536                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3537                         status = nfs4_lookup_root_sec(server, fhandle, info,
3538                                                 server->auth_info.flavors[i]);
3539                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3540                                 continue;
3541                         break;
3542                 }
3543         } else {
3544                 /* no flavors specified by user, try default list */
3545                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3546                         status = nfs4_lookup_root_sec(server, fhandle, info,
3547                                                       flav_array[i]);
3548                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3549                                 continue;
3550                         break;
3551                 }
3552         }
3553
3554         /*
3555          * -EACCESS could mean that the user doesn't have correct permissions
3556          * to access the mount.  It could also mean that we tried to mount
3557          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3558          * existing mount programs don't handle -EACCES very well so it should
3559          * be mapped to -EPERM instead.
3560          */
3561         if (status == -EACCES)
3562                 status = -EPERM;
3563         return status;
3564 }
3565
3566 /**
3567  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3568  * @server: initialized nfs_server handle
3569  * @fhandle: we fill in the pseudo-fs root file handle
3570  * @info: we fill in an FSINFO struct
3571  * @auth_probe: probe the auth flavours
3572  *
3573  * Returns zero on success, or a negative errno.
3574  */
3575 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3576                          struct nfs_fsinfo *info,
3577                          bool auth_probe)
3578 {
3579         int status = 0;
3580
3581         if (!auth_probe)
3582                 status = nfs4_lookup_root(server, fhandle, info);
3583
3584         if (auth_probe || status == NFS4ERR_WRONGSEC)
3585                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3586                                 fhandle, info);
3587
3588         if (status == 0)
3589                 status = nfs4_server_capabilities(server, fhandle);
3590         if (status == 0)
3591                 status = nfs4_do_fsinfo(server, fhandle, info);
3592
3593         return nfs4_map_errors(status);
3594 }
3595
3596 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3597                               struct nfs_fsinfo *info)
3598 {
3599         int error;
3600         struct nfs_fattr *fattr = info->fattr;
3601         struct nfs4_label *label = NULL;
3602
3603         error = nfs4_server_capabilities(server, mntfh);
3604         if (error < 0) {
3605                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3606                 return error;
3607         }
3608
3609         label = nfs4_label_alloc(server, GFP_KERNEL);
3610         if (IS_ERR(label))
3611                 return PTR_ERR(label);
3612
3613         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3614         if (error < 0) {
3615                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3616                 goto err_free_label;
3617         }
3618
3619         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3620             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3621                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3622
3623 err_free_label:
3624         nfs4_label_free(label);
3625
3626         return error;
3627 }
3628
3629 /*
3630  * Get locations and (maybe) other attributes of a referral.
3631  * Note that we'll actually follow the referral later when
3632  * we detect fsid mismatch in inode revalidation
3633  */
3634 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3635                              const struct qstr *name, struct nfs_fattr *fattr,
3636                              struct nfs_fh *fhandle)
3637 {
3638         int status = -ENOMEM;
3639         struct page *page = NULL;
3640         struct nfs4_fs_locations *locations = NULL;
3641
3642         page = alloc_page(GFP_KERNEL);
3643         if (page == NULL)
3644                 goto out;
3645         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3646         if (locations == NULL)
3647                 goto out;
3648
3649         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3650         if (status != 0)
3651                 goto out;
3652
3653         /*
3654          * If the fsid didn't change, this is a migration event, not a
3655          * referral.  Cause us to drop into the exception handler, which
3656          * will kick off migration recovery.
3657          */
3658         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3659                 dprintk("%s: server did not return a different fsid for"
3660                         " a referral at %s\n", __func__, name->name);
3661                 status = -NFS4ERR_MOVED;
3662                 goto out;
3663         }
3664         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3665         nfs_fixup_referral_attributes(&locations->fattr);
3666
3667         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3668         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3669         memset(fhandle, 0, sizeof(struct nfs_fh));
3670 out:
3671         if (page)
3672                 __free_page(page);
3673         kfree(locations);
3674         return status;
3675 }
3676
3677 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3678                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3679 {
3680         struct nfs4_getattr_arg args = {
3681                 .fh = fhandle,
3682                 .bitmask = server->attr_bitmask,
3683         };
3684         struct nfs4_getattr_res res = {
3685                 .fattr = fattr,
3686                 .label = label,
3687                 .server = server,
3688         };
3689         struct rpc_message msg = {
3690                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3691                 .rpc_argp = &args,
3692                 .rpc_resp = &res,
3693         };
3694
3695         args.bitmask = nfs4_bitmask(server, label);
3696
3697         nfs_fattr_init(fattr);
3698         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3699 }
3700
3701 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3702                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3703 {
3704         struct nfs4_exception exception = { };
3705         int err;
3706         do {
3707                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3708                 trace_nfs4_getattr(server, fhandle, fattr, err);
3709                 err = nfs4_handle_exception(server, err,
3710                                 &exception);
3711         } while (exception.retry);
3712         return err;
3713 }
3714
3715 /* 
3716  * The file is not closed if it is opened due to the a request to change
3717  * the size of the file. The open call will not be needed once the
3718  * VFS layer lookup-intents are implemented.
3719  *
3720  * Close is called when the inode is destroyed.
3721  * If we haven't opened the file for O_WRONLY, we
3722  * need to in the size_change case to obtain a stateid.
3723  *
3724  * Got race?
3725  * Because OPEN is always done by name in nfsv4, it is
3726  * possible that we opened a different file by the same
3727  * name.  We can recognize this race condition, but we
3728  * can't do anything about it besides returning an error.
3729  *
3730  * This will be fixed with VFS changes (lookup-intent).
3731  */
3732 static int
3733 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3734                   struct iattr *sattr)
3735 {
3736         struct inode *inode = d_inode(dentry);
3737         struct rpc_cred *cred = NULL;
3738         struct nfs_open_context *ctx = NULL;
3739         struct nfs4_label *label = NULL;
3740         int status;
3741
3742         if (pnfs_ld_layoutret_on_setattr(inode) &&
3743             sattr->ia_valid & ATTR_SIZE &&
3744             sattr->ia_size < i_size_read(inode))
3745                 pnfs_commit_and_return_layout(inode);
3746
3747         nfs_fattr_init(fattr);
3748         
3749         /* Deal with open(O_TRUNC) */
3750         if (sattr->ia_valid & ATTR_OPEN)
3751                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3752
3753         /* Optimization: if the end result is no change, don't RPC */
3754         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3755                 return 0;
3756
3757         /* Search for an existing open(O_WRITE) file */
3758         if (sattr->ia_valid & ATTR_FILE) {
3759
3760                 ctx = nfs_file_open_context(sattr->ia_file);
3761                 if (ctx)
3762                         cred = ctx->cred;
3763         }
3764
3765         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3766         if (IS_ERR(label))
3767                 return PTR_ERR(label);
3768
3769         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3770         if (status == 0) {
3771                 nfs_setattr_update_inode(inode, sattr, fattr);
3772                 nfs_setsecurity(inode, fattr, label);
3773         }
3774         nfs4_label_free(label);
3775         return status;
3776 }
3777
3778 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3779                 const struct qstr *name, struct nfs_fh *fhandle,
3780                 struct nfs_fattr *fattr, struct nfs4_label *label)
3781 {
3782         struct nfs_server *server = NFS_SERVER(dir);
3783         int                    status;
3784         struct nfs4_lookup_arg args = {
3785                 .bitmask = server->attr_bitmask,
3786                 .dir_fh = NFS_FH(dir),
3787                 .name = name,
3788         };
3789         struct nfs4_lookup_res res = {
3790                 .server = server,
3791                 .fattr = fattr,
3792                 .label = label,
3793                 .fh = fhandle,
3794         };
3795         struct rpc_message msg = {
3796                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3797                 .rpc_argp = &args,
3798                 .rpc_resp = &res,
3799         };
3800
3801         args.bitmask = nfs4_bitmask(server, label);
3802
3803         nfs_fattr_init(fattr);
3804
3805         dprintk("NFS call  lookup %s\n", name->name);
3806         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3807         dprintk("NFS reply lookup: %d\n", status);
3808         return status;
3809 }
3810
3811 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3812 {
3813         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3814                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3815         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3816         fattr->nlink = 2;
3817 }
3818
3819 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3820                                    const struct qstr *name, struct nfs_fh *fhandle,
3821                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3822 {
3823         struct nfs4_exception exception = { };
3824         struct rpc_clnt *client = *clnt;
3825         int err;
3826         do {
3827                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3828                 trace_nfs4_lookup(dir, name, err);
3829                 switch (err) {
3830                 case -NFS4ERR_BADNAME:
3831                         err = -ENOENT;
3832                         goto out;
3833                 case -NFS4ERR_MOVED:
3834                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3835                         if (err == -NFS4ERR_MOVED)
3836                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3837                         goto out;
3838                 case -NFS4ERR_WRONGSEC:
3839                         err = -EPERM;
3840                         if (client != *clnt)
3841                                 goto out;
3842                         client = nfs4_negotiate_security(client, dir, name);
3843                         if (IS_ERR(client))
3844                                 return PTR_ERR(client);
3845
3846                         exception.retry = 1;
3847                         break;
3848                 default:
3849                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3850                 }
3851         } while (exception.retry);
3852
3853 out:
3854         if (err == 0)
3855                 *clnt = client;
3856         else if (client != *clnt)
3857                 rpc_shutdown_client(client);
3858
3859         return err;
3860 }
3861
3862 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3863                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3864                             struct nfs4_label *label)
3865 {
3866         int status;
3867         struct rpc_clnt *client = NFS_CLIENT(dir);
3868
3869         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3870         if (client != NFS_CLIENT(dir)) {
3871                 rpc_shutdown_client(client);
3872                 nfs_fixup_secinfo_attributes(fattr);
3873         }
3874         return status;
3875 }
3876
3877 struct rpc_clnt *
3878 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3879                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3880 {
3881         struct rpc_clnt *client = NFS_CLIENT(dir);
3882         int status;
3883
3884         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3885         if (status < 0)
3886                 return ERR_PTR(status);
3887         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3888 }
3889
3890 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3891 {
3892         struct nfs_server *server = NFS_SERVER(inode);
3893         struct nfs4_accessargs args = {
3894                 .fh = NFS_FH(inode),
3895                 .bitmask = server->cache_consistency_bitmask,
3896         };
3897         struct nfs4_accessres res = {
3898                 .server = server,
3899         };
3900         struct rpc_message msg = {
3901                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3902                 .rpc_argp = &args,
3903                 .rpc_resp = &res,
3904                 .rpc_cred = entry->cred,
3905         };
3906         int mode = entry->mask;
3907         int status = 0;
3908
3909         /*
3910          * Determine which access bits we want to ask for...
3911          */
3912         if (mode & MAY_READ)
3913                 args.access |= NFS4_ACCESS_READ;
3914         if (S_ISDIR(inode->i_mode)) {
3915                 if (mode & MAY_WRITE)
3916                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3917                 if (mode & MAY_EXEC)
3918                         args.access |= NFS4_ACCESS_LOOKUP;
3919         } else {
3920                 if (mode & MAY_WRITE)
3921                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3922                 if (mode & MAY_EXEC)
3923                         args.access |= NFS4_ACCESS_EXECUTE;
3924         }
3925
3926         res.fattr = nfs_alloc_fattr();
3927         if (res.fattr == NULL)
3928                 return -ENOMEM;
3929
3930         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3931         if (!status) {
3932                 nfs_access_set_mask(entry, res.access);
3933                 nfs_refresh_inode(inode, res.fattr);
3934         }
3935         nfs_free_fattr(res.fattr);
3936         return status;
3937 }
3938
3939 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3940 {
3941         struct nfs4_exception exception = { };
3942         int err;
3943         do {
3944                 err = _nfs4_proc_access(inode, entry);
3945                 trace_nfs4_access(inode, err);
3946                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3947                                 &exception);
3948         } while (exception.retry);
3949         return err;
3950 }
3951
3952 /*
3953  * TODO: For the time being, we don't try to get any attributes
3954  * along with any of the zero-copy operations READ, READDIR,
3955  * READLINK, WRITE.
3956  *
3957  * In the case of the first three, we want to put the GETATTR
3958  * after the read-type operation -- this is because it is hard
3959  * to predict the length of a GETATTR response in v4, and thus
3960  * align the READ data correctly.  This means that the GETATTR
3961  * may end up partially falling into the page cache, and we should
3962  * shift it into the 'tail' of the xdr_buf before processing.
3963  * To do this efficiently, we need to know the total length
3964  * of data received, which doesn't seem to be available outside
3965  * of the RPC layer.
3966  *
3967  * In the case of WRITE, we also want to put the GETATTR after
3968  * the operation -- in this case because we want to make sure
3969  * we get the post-operation mtime and size.
3970  *
3971  * Both of these changes to the XDR layer would in fact be quite
3972  * minor, but I decided to leave them for a subsequent patch.
3973  */
3974 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3975                 unsigned int pgbase, unsigned int pglen)
3976 {
3977         struct nfs4_readlink args = {
3978                 .fh       = NFS_FH(inode),
3979                 .pgbase   = pgbase,
3980                 .pglen    = pglen,
3981                 .pages    = &page,
3982         };
3983         struct nfs4_readlink_res res;
3984         struct rpc_message msg = {
3985                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3986                 .rpc_argp = &args,
3987                 .rpc_resp = &res,
3988         };
3989
3990         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3991 }
3992
3993 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3994                 unsigned int pgbase, unsigned int pglen)
3995 {
3996         struct nfs4_exception exception = { };
3997         int err;
3998         do {
3999                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4000                 trace_nfs4_readlink(inode, err);
4001                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4002                                 &exception);
4003         } while (exception.retry);
4004         return err;
4005 }
4006
4007 /*
4008  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4009  */
4010 static int
4011 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4012                  int flags)
4013 {
4014         struct nfs_server *server = NFS_SERVER(dir);
4015         struct nfs4_label l, *ilabel = NULL;
4016         struct nfs_open_context *ctx;
4017         struct nfs4_state *state;
4018         int status = 0;
4019
4020         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4021         if (IS_ERR(ctx))
4022                 return PTR_ERR(ctx);
4023
4024         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4025
4026         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4027                 sattr->ia_mode &= ~current_umask();
4028         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4029         if (IS_ERR(state)) {
4030                 status = PTR_ERR(state);
4031                 goto out;
4032         }
4033 out:
4034         nfs4_label_release_security(ilabel);
4035         put_nfs_open_context(ctx);
4036         return status;
4037 }
4038
4039 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4040 {
4041         struct nfs_server *server = NFS_SERVER(dir);
4042         struct nfs_removeargs args = {
4043                 .fh = NFS_FH(dir),
4044                 .name = *name,
4045         };
4046         struct nfs_removeres res = {
4047                 .server = server,
4048         };
4049         struct rpc_message msg = {
4050                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4051                 .rpc_argp = &args,
4052                 .rpc_resp = &res,
4053         };
4054         int status;
4055
4056         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4057         if (status == 0)
4058                 update_changeattr(dir, &res.cinfo);
4059         return status;
4060 }
4061
4062 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4063 {
4064         struct nfs4_exception exception = { };
4065         int err;
4066         do {
4067                 err = _nfs4_proc_remove(dir, name);
4068                 trace_nfs4_remove(dir, name, err);
4069                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4070                                 &exception);
4071         } while (exception.retry);
4072         return err;
4073 }
4074
4075 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4076 {
4077         struct nfs_server *server = NFS_SERVER(dir);
4078         struct nfs_removeargs *args = msg->rpc_argp;
4079         struct nfs_removeres *res = msg->rpc_resp;
4080
4081         res->server = server;
4082         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4083         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4084
4085         nfs_fattr_init(res->dir_attr);
4086 }
4087
4088 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4089 {
4090         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
4091                         &data->args.seq_args,
4092                         &data->res.seq_res,
4093                         task);
4094 }
4095
4096 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4097 {
4098         struct nfs_unlinkdata *data = task->tk_calldata;
4099         struct nfs_removeres *res = &data->res;
4100
4101         if (!nfs4_sequence_done(task, &res->seq_res))
4102                 return 0;
4103         if (nfs4_async_handle_error(task, res->server, NULL,
4104                                     &data->timeout) == -EAGAIN)
4105                 return 0;
4106         update_changeattr(dir, &res->cinfo);
4107         return 1;
4108 }
4109
4110 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4111 {
4112         struct nfs_server *server = NFS_SERVER(dir);
4113         struct nfs_renameargs *arg = msg->rpc_argp;
4114         struct nfs_renameres *res = msg->rpc_resp;
4115
4116         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4117         res->server = server;
4118         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4119 }
4120
4121 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4122 {
4123         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
4124                         &data->args.seq_args,
4125                         &data->res.seq_res,
4126                         task);
4127 }
4128
4129 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4130                                  struct inode *new_dir)
4131 {
4132         struct nfs_renamedata *data = task->tk_calldata;
4133         struct nfs_renameres *res = &data->res;
4134
4135         if (!nfs4_sequence_done(task, &res->seq_res))
4136                 return 0;
4137         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4138                 return 0;
4139
4140         update_changeattr(old_dir, &res->old_cinfo);
4141         update_changeattr(new_dir, &res->new_cinfo);
4142         return 1;
4143 }
4144
4145 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4146 {
4147         struct nfs_server *server = NFS_SERVER(inode);
4148         struct nfs4_link_arg arg = {
4149                 .fh     = NFS_FH(inode),
4150                 .dir_fh = NFS_FH(dir),
4151                 .name   = name,
4152                 .bitmask = server->attr_bitmask,
4153         };
4154         struct nfs4_link_res res = {
4155                 .server = server,
4156                 .label = NULL,
4157         };
4158         struct rpc_message msg = {
4159                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4160                 .rpc_argp = &arg,
4161                 .rpc_resp = &res,
4162         };
4163         int status = -ENOMEM;
4164
4165         res.fattr = nfs_alloc_fattr();
4166         if (res.fattr == NULL)
4167                 goto out;
4168
4169         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4170         if (IS_ERR(res.label)) {
4171                 status = PTR_ERR(res.label);
4172                 goto out;
4173         }
4174         arg.bitmask = nfs4_bitmask(server, res.label);
4175
4176         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4177         if (!status) {
4178                 update_changeattr(dir, &res.cinfo);
4179                 status = nfs_post_op_update_inode(inode, res.fattr);
4180                 if (!status)
4181                         nfs_setsecurity(inode, res.fattr, res.label);
4182         }
4183
4184
4185         nfs4_label_free(res.label);
4186
4187 out:
4188         nfs_free_fattr(res.fattr);
4189         return status;
4190 }
4191
4192 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4193 {
4194         struct nfs4_exception exception = { };
4195         int err;
4196         do {
4197                 err = nfs4_handle_exception(NFS_SERVER(inode),
4198                                 _nfs4_proc_link(inode, dir, name),
4199                                 &exception);
4200         } while (exception.retry);
4201         return err;
4202 }
4203
4204 struct nfs4_createdata {
4205         struct rpc_message msg;
4206         struct nfs4_create_arg arg;
4207         struct nfs4_create_res res;
4208         struct nfs_fh fh;
4209         struct nfs_fattr fattr;
4210         struct nfs4_label *label;
4211 };
4212
4213 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4214                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4215 {
4216         struct nfs4_createdata *data;
4217
4218         data = kzalloc(sizeof(*data), GFP_KERNEL);
4219         if (data != NULL) {
4220                 struct nfs_server *server = NFS_SERVER(dir);
4221
4222                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4223                 if (IS_ERR(data->label))
4224                         goto out_free;
4225
4226                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4227                 data->msg.rpc_argp = &data->arg;
4228                 data->msg.rpc_resp = &data->res;
4229                 data->arg.dir_fh = NFS_FH(dir);
4230                 data->arg.server = server;
4231                 data->arg.name = name;
4232                 data->arg.attrs = sattr;
4233                 data->arg.ftype = ftype;
4234                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4235                 data->arg.umask = current_umask();
4236                 data->res.server = server;
4237                 data->res.fh = &data->fh;
4238                 data->res.fattr = &data->fattr;
4239                 data->res.label = data->label;
4240                 nfs_fattr_init(data->res.fattr);
4241         }
4242         return data;
4243 out_free:
4244         kfree(data);
4245         return NULL;
4246 }
4247
4248 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4249 {
4250         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4251                                     &data->arg.seq_args, &data->res.seq_res, 1);
4252         if (status == 0) {
4253                 update_changeattr(dir, &data->res.dir_cinfo);
4254                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4255         }
4256         return status;
4257 }
4258
4259 static void nfs4_free_createdata(struct nfs4_createdata *data)
4260 {
4261         nfs4_label_free(data->label);
4262         kfree(data);
4263 }
4264
4265 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4266                 struct page *page, unsigned int len, struct iattr *sattr,
4267                 struct nfs4_label *label)
4268 {
4269         struct nfs4_createdata *data;
4270         int status = -ENAMETOOLONG;
4271
4272         if (len > NFS4_MAXPATHLEN)
4273                 goto out;
4274
4275         status = -ENOMEM;
4276         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4277         if (data == NULL)
4278                 goto out;
4279
4280         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4281         data->arg.u.symlink.pages = &page;
4282         data->arg.u.symlink.len = len;
4283         data->arg.label = label;
4284         
4285         status = nfs4_do_create(dir, dentry, data);
4286
4287         nfs4_free_createdata(data);
4288 out:
4289         return status;
4290 }
4291
4292 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4293                 struct page *page, unsigned int len, struct iattr *sattr)
4294 {
4295         struct nfs4_exception exception = { };
4296         struct nfs4_label l, *label = NULL;
4297         int err;
4298
4299         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4300
4301         do {
4302                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4303                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4304                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4305                                 &exception);
4306         } while (exception.retry);
4307
4308         nfs4_label_release_security(label);
4309         return err;
4310 }
4311
4312 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4313                 struct iattr *sattr, struct nfs4_label *label)
4314 {
4315         struct nfs4_createdata *data;
4316         int status = -ENOMEM;
4317
4318         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4319         if (data == NULL)
4320                 goto out;
4321
4322         data->arg.label = label;
4323         status = nfs4_do_create(dir, dentry, data);
4324
4325         nfs4_free_createdata(data);
4326 out:
4327         return status;
4328 }
4329
4330 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4331                 struct iattr *sattr)
4332 {
4333         struct nfs_server *server = NFS_SERVER(dir);
4334         struct nfs4_exception exception = { };
4335         struct nfs4_label l, *label = NULL;
4336         int err;
4337
4338         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4339
4340         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4341                 sattr->ia_mode &= ~current_umask();
4342         do {
4343                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4344                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4345                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4346                                 &exception);
4347         } while (exception.retry);
4348         nfs4_label_release_security(label);
4349
4350         return err;
4351 }
4352
4353 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4354                 u64 cookie, struct page **pages, unsigned int count, int plus)
4355 {
4356         struct inode            *dir = d_inode(dentry);
4357         struct nfs4_readdir_arg args = {
4358                 .fh = NFS_FH(dir),
4359                 .pages = pages,
4360                 .pgbase = 0,
4361                 .count = count,
4362                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4363                 .plus = plus,
4364         };
4365         struct nfs4_readdir_res res;
4366         struct rpc_message msg = {
4367                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4368                 .rpc_argp = &args,
4369                 .rpc_resp = &res,
4370                 .rpc_cred = cred,
4371         };
4372         int                     status;
4373
4374         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4375                         dentry,
4376                         (unsigned long long)cookie);
4377         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4378         res.pgbase = args.pgbase;
4379         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4380         if (status >= 0) {
4381                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4382                 status += args.pgbase;
4383         }
4384
4385         nfs_invalidate_atime(dir);
4386
4387         dprintk("%s: returns %d\n", __func__, status);
4388         return status;
4389 }
4390
4391 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4392                 u64 cookie, struct page **pages, unsigned int count, int plus)
4393 {
4394         struct nfs4_exception exception = { };
4395         int err;
4396         do {
4397                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4398                                 pages, count, plus);
4399                 trace_nfs4_readdir(d_inode(dentry), err);
4400                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4401                                 &exception);
4402         } while (exception.retry);
4403         return err;
4404 }
4405
4406 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4407                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4408 {
4409         struct nfs4_createdata *data;
4410         int mode = sattr->ia_mode;
4411         int status = -ENOMEM;
4412
4413         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4414         if (data == NULL)
4415                 goto out;
4416
4417         if (S_ISFIFO(mode))
4418                 data->arg.ftype = NF4FIFO;
4419         else if (S_ISBLK(mode)) {
4420                 data->arg.ftype = NF4BLK;
4421                 data->arg.u.device.specdata1 = MAJOR(rdev);
4422                 data->arg.u.device.specdata2 = MINOR(rdev);
4423         }
4424         else if (S_ISCHR(mode)) {
4425                 data->arg.ftype = NF4CHR;
4426                 data->arg.u.device.specdata1 = MAJOR(rdev);
4427                 data->arg.u.device.specdata2 = MINOR(rdev);
4428         } else if (!S_ISSOCK(mode)) {
4429                 status = -EINVAL;
4430                 goto out_free;
4431         }
4432
4433         data->arg.label = label;
4434         status = nfs4_do_create(dir, dentry, data);
4435 out_free:
4436         nfs4_free_createdata(data);
4437 out:
4438         return status;
4439 }
4440
4441 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4442                 struct iattr *sattr, dev_t rdev)
4443 {
4444         struct nfs_server *server = NFS_SERVER(dir);
4445         struct nfs4_exception exception = { };
4446         struct nfs4_label l, *label = NULL;
4447         int err;
4448
4449         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4450
4451         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4452                 sattr->ia_mode &= ~current_umask();
4453         do {
4454                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4455                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4456                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4457                                 &exception);
4458         } while (exception.retry);
4459
4460         nfs4_label_release_security(label);
4461
4462         return err;
4463 }
4464
4465 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4466                  struct nfs_fsstat *fsstat)
4467 {
4468         struct nfs4_statfs_arg args = {
4469                 .fh = fhandle,
4470                 .bitmask = server->attr_bitmask,
4471         };
4472         struct nfs4_statfs_res res = {
4473                 .fsstat = fsstat,
4474         };
4475         struct rpc_message msg = {
4476                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4477                 .rpc_argp = &args,
4478                 .rpc_resp = &res,
4479         };
4480
4481         nfs_fattr_init(fsstat->fattr);
4482         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4483 }
4484
4485 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4486 {
4487         struct nfs4_exception exception = { };
4488         int err;
4489         do {
4490                 err = nfs4_handle_exception(server,
4491                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4492                                 &exception);
4493         } while (exception.retry);
4494         return err;
4495 }
4496
4497 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4498                 struct nfs_fsinfo *fsinfo)
4499 {
4500         struct nfs4_fsinfo_arg args = {
4501                 .fh = fhandle,
4502                 .bitmask = server->attr_bitmask,
4503         };
4504         struct nfs4_fsinfo_res res = {
4505                 .fsinfo = fsinfo,
4506         };
4507         struct rpc_message msg = {
4508                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4509                 .rpc_argp = &args,
4510                 .rpc_resp = &res,
4511         };
4512
4513         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4514 }
4515
4516 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4517 {
4518         struct nfs4_exception exception = { };
4519         unsigned long now = jiffies;
4520         int err;
4521
4522         do {
4523                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4524                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4525                 if (err == 0) {
4526                         nfs4_set_lease_period(server->nfs_client,
4527                                         fsinfo->lease_time * HZ,
4528                                         now);
4529                         break;
4530                 }
4531                 err = nfs4_handle_exception(server, err, &exception);
4532         } while (exception.retry);
4533         return err;
4534 }
4535
4536 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4537 {
4538         int error;
4539
4540         nfs_fattr_init(fsinfo->fattr);
4541         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4542         if (error == 0) {
4543                 /* block layout checks this! */
4544                 server->pnfs_blksize = fsinfo->blksize;
4545                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4546         }
4547
4548         return error;
4549 }
4550
4551 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4552                 struct nfs_pathconf *pathconf)
4553 {
4554         struct nfs4_pathconf_arg args = {
4555                 .fh = fhandle,
4556                 .bitmask = server->attr_bitmask,
4557         };
4558         struct nfs4_pathconf_res res = {
4559                 .pathconf = pathconf,
4560         };
4561         struct rpc_message msg = {
4562                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4563                 .rpc_argp = &args,
4564                 .rpc_resp = &res,
4565         };
4566
4567         /* None of the pathconf attributes are mandatory to implement */
4568         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4569                 memset(pathconf, 0, sizeof(*pathconf));
4570                 return 0;
4571         }
4572
4573         nfs_fattr_init(pathconf->fattr);
4574         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4575 }
4576
4577 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4578                 struct nfs_pathconf *pathconf)
4579 {
4580         struct nfs4_exception exception = { };
4581         int err;
4582
4583         do {
4584                 err = nfs4_handle_exception(server,
4585                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4586                                 &exception);
4587         } while (exception.retry);
4588         return err;
4589 }
4590
4591 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4592                 const struct nfs_open_context *ctx,
4593                 const struct nfs_lock_context *l_ctx,
4594                 fmode_t fmode)
4595 {
4596         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4597 }
4598 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4599
4600 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4601                 const struct nfs_open_context *ctx,
4602                 const struct nfs_lock_context *l_ctx,
4603                 fmode_t fmode)
4604 {
4605         nfs4_stateid current_stateid;
4606
4607         /* If the current stateid represents a lost lock, then exit */
4608         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4609                 return true;
4610         return nfs4_stateid_match(stateid, &current_stateid);
4611 }
4612
4613 static bool nfs4_error_stateid_expired(int err)
4614 {
4615         switch (err) {
4616         case -NFS4ERR_DELEG_REVOKED:
4617         case -NFS4ERR_ADMIN_REVOKED:
4618         case -NFS4ERR_BAD_STATEID:
4619         case -NFS4ERR_STALE_STATEID:
4620         case -NFS4ERR_OLD_STATEID:
4621         case -NFS4ERR_OPENMODE:
4622         case -NFS4ERR_EXPIRED:
4623                 return true;
4624         }
4625         return false;
4626 }
4627
4628 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4629 {
4630         struct nfs_server *server = NFS_SERVER(hdr->inode);
4631
4632         trace_nfs4_read(hdr, task->tk_status);
4633         if (task->tk_status < 0) {
4634                 struct nfs4_exception exception = {
4635                         .inode = hdr->inode,
4636                         .state = hdr->args.context->state,
4637                         .stateid = &hdr->args.stateid,
4638                 };
4639                 task->tk_status = nfs4_async_handle_exception(task,
4640                                 server, task->tk_status, &exception);
4641                 if (exception.retry) {
4642                         rpc_restart_call_prepare(task);
4643                         return -EAGAIN;
4644                 }
4645         }
4646
4647         if (task->tk_status > 0)
4648                 renew_lease(server, hdr->timestamp);
4649         return 0;
4650 }
4651
4652 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4653                 struct nfs_pgio_args *args)
4654 {
4655
4656         if (!nfs4_error_stateid_expired(task->tk_status) ||
4657                 nfs4_stateid_is_current(&args->stateid,
4658                                 args->context,
4659                                 args->lock_context,
4660                                 FMODE_READ))
4661                 return false;
4662         rpc_restart_call_prepare(task);
4663         return true;
4664 }
4665
4666 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4667 {
4668
4669         dprintk("--> %s\n", __func__);
4670
4671         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4672                 return -EAGAIN;
4673         if (nfs4_read_stateid_changed(task, &hdr->args))
4674                 return -EAGAIN;
4675         if (task->tk_status > 0)
4676                 nfs_invalidate_atime(hdr->inode);
4677         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4678                                     nfs4_read_done_cb(task, hdr);
4679 }
4680
4681 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4682                                  struct rpc_message *msg)
4683 {
4684         hdr->timestamp   = jiffies;
4685         if (!hdr->pgio_done_cb)
4686                 hdr->pgio_done_cb = nfs4_read_done_cb;
4687         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4688         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4689 }
4690
4691 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4692                                       struct nfs_pgio_header *hdr)
4693 {
4694         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4695                         &hdr->args.seq_args,
4696                         &hdr->res.seq_res,
4697                         task))
4698                 return 0;
4699         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4700                                 hdr->args.lock_context,
4701                                 hdr->rw_ops->rw_mode) == -EIO)
4702                 return -EIO;
4703         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4704                 return -EIO;
4705         return 0;
4706 }
4707
4708 static int nfs4_write_done_cb(struct rpc_task *task,
4709                               struct nfs_pgio_header *hdr)
4710 {
4711         struct inode *inode = hdr->inode;
4712
4713         trace_nfs4_write(hdr, task->tk_status);
4714         if (task->tk_status < 0) {
4715                 struct nfs4_exception exception = {
4716                         .inode = hdr->inode,
4717                         .state = hdr->args.context->state,
4718                         .stateid = &hdr->args.stateid,
4719                 };
4720                 task->tk_status = nfs4_async_handle_exception(task,
4721                                 NFS_SERVER(inode), task->tk_status,
4722                                 &exception);
4723                 if (exception.retry) {
4724                         rpc_restart_call_prepare(task);
4725                         return -EAGAIN;
4726                 }
4727         }
4728         if (task->tk_status >= 0) {
4729                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4730                 nfs_writeback_update_inode(hdr);
4731         }
4732         return 0;
4733 }
4734
4735 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4736                 struct nfs_pgio_args *args)
4737 {
4738
4739         if (!nfs4_error_stateid_expired(task->tk_status) ||
4740                 nfs4_stateid_is_current(&args->stateid,
4741                                 args->context,
4742                                 args->lock_context,
4743                                 FMODE_WRITE))
4744                 return false;
4745         rpc_restart_call_prepare(task);
4746         return true;
4747 }
4748
4749 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4750 {
4751         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4752                 return -EAGAIN;
4753         if (nfs4_write_stateid_changed(task, &hdr->args))
4754                 return -EAGAIN;
4755         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4756                 nfs4_write_done_cb(task, hdr);
4757 }
4758
4759 static
4760 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4761 {
4762         /* Don't request attributes for pNFS or O_DIRECT writes */
4763         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4764                 return false;
4765         /* Otherwise, request attributes if and only if we don't hold
4766          * a delegation
4767          */
4768         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4769 }
4770
4771 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4772                                   struct rpc_message *msg)
4773 {
4774         struct nfs_server *server = NFS_SERVER(hdr->inode);
4775
4776         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4777                 hdr->args.bitmask = NULL;
4778                 hdr->res.fattr = NULL;
4779         } else
4780                 hdr->args.bitmask = server->cache_consistency_bitmask;
4781
4782         if (!hdr->pgio_done_cb)
4783                 hdr->pgio_done_cb = nfs4_write_done_cb;
4784         hdr->res.server = server;
4785         hdr->timestamp   = jiffies;
4786
4787         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4788         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4789 }
4790
4791 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4792 {
4793         nfs4_setup_sequence(NFS_SERVER(data->inode),
4794                         &data->args.seq_args,
4795                         &data->res.seq_res,
4796                         task);
4797 }
4798
4799 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4800 {
4801         struct inode *inode = data->inode;
4802
4803         trace_nfs4_commit(data, task->tk_status);
4804         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4805                                     NULL, NULL) == -EAGAIN) {
4806                 rpc_restart_call_prepare(task);
4807                 return -EAGAIN;
4808         }
4809         return 0;
4810 }
4811
4812 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4813 {
4814         if (!nfs4_sequence_done(task, &data->res.seq_res))
4815                 return -EAGAIN;
4816         return data->commit_done_cb(task, data);
4817 }
4818
4819 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4820 {
4821         struct nfs_server *server = NFS_SERVER(data->inode);
4822
4823         if (data->commit_done_cb == NULL)
4824                 data->commit_done_cb = nfs4_commit_done_cb;
4825         data->res.server = server;
4826         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4827         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4828 }
4829
4830 struct nfs4_renewdata {
4831         struct nfs_client       *client;
4832         unsigned long           timestamp;
4833 };
4834
4835 /*
4836  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4837  * standalone procedure for queueing an asynchronous RENEW.
4838  */
4839 static void nfs4_renew_release(void *calldata)
4840 {
4841         struct nfs4_renewdata *data = calldata;
4842         struct nfs_client *clp = data->client;
4843
4844         if (atomic_read(&clp->cl_count) > 1)
4845                 nfs4_schedule_state_renewal(clp);
4846         nfs_put_client(clp);
4847         kfree(data);
4848 }
4849
4850 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4851 {
4852         struct nfs4_renewdata *data = calldata;
4853         struct nfs_client *clp = data->client;
4854         unsigned long timestamp = data->timestamp;
4855
4856         trace_nfs4_renew_async(clp, task->tk_status);
4857         switch (task->tk_status) {
4858         case 0:
4859                 break;
4860         case -NFS4ERR_LEASE_MOVED:
4861                 nfs4_schedule_lease_moved_recovery(clp);
4862                 break;
4863         default:
4864                 /* Unless we're shutting down, schedule state recovery! */
4865                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4866                         return;
4867                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4868                         nfs4_schedule_lease_recovery(clp);
4869                         return;
4870                 }
4871                 nfs4_schedule_path_down_recovery(clp);
4872         }
4873         do_renew_lease(clp, timestamp);
4874 }
4875
4876 static const struct rpc_call_ops nfs4_renew_ops = {
4877         .rpc_call_done = nfs4_renew_done,
4878         .rpc_release = nfs4_renew_release,
4879 };
4880
4881 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4882 {
4883         struct rpc_message msg = {
4884                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4885                 .rpc_argp       = clp,
4886                 .rpc_cred       = cred,
4887         };
4888         struct nfs4_renewdata *data;
4889
4890         if (renew_flags == 0)
4891                 return 0;
4892         if (!atomic_inc_not_zero(&clp->cl_count))
4893                 return -EIO;
4894         data = kmalloc(sizeof(*data), GFP_NOFS);
4895         if (data == NULL)
4896                 return -ENOMEM;
4897         data->client = clp;
4898         data->timestamp = jiffies;
4899         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4900                         &nfs4_renew_ops, data);
4901 }
4902
4903 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4904 {
4905         struct rpc_message msg = {
4906                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4907                 .rpc_argp       = clp,
4908                 .rpc_cred       = cred,
4909         };
4910         unsigned long now = jiffies;
4911         int status;
4912
4913         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4914         if (status < 0)
4915                 return status;
4916         do_renew_lease(clp, now);
4917         return 0;
4918 }
4919
4920 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4921 {
4922         return server->caps & NFS_CAP_ACLS;
4923 }
4924
4925 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4926  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4927  * the stack.
4928  */
4929 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4930
4931 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4932                 struct page **pages)
4933 {
4934         struct page *newpage, **spages;
4935         int rc = 0;
4936         size_t len;
4937         spages = pages;
4938
4939         do {
4940                 len = min_t(size_t, PAGE_SIZE, buflen);
4941                 newpage = alloc_page(GFP_KERNEL);
4942
4943                 if (newpage == NULL)
4944                         goto unwind;
4945                 memcpy(page_address(newpage), buf, len);
4946                 buf += len;
4947                 buflen -= len;
4948                 *pages++ = newpage;
4949                 rc++;
4950         } while (buflen != 0);
4951
4952         return rc;
4953
4954 unwind:
4955         for(; rc > 0; rc--)
4956                 __free_page(spages[rc-1]);
4957         return -ENOMEM;
4958 }
4959
4960 struct nfs4_cached_acl {
4961         int cached;
4962         size_t len;
4963         char data[0];
4964 };
4965
4966 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4967 {
4968         struct nfs_inode *nfsi = NFS_I(inode);
4969
4970         spin_lock(&inode->i_lock);
4971         kfree(nfsi->nfs4_acl);
4972         nfsi->nfs4_acl = acl;
4973         spin_unlock(&inode->i_lock);
4974 }
4975
4976 static void nfs4_zap_acl_attr(struct inode *inode)
4977 {
4978         nfs4_set_cached_acl(inode, NULL);
4979 }
4980
4981 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4982 {
4983         struct nfs_inode *nfsi = NFS_I(inode);
4984         struct nfs4_cached_acl *acl;
4985         int ret = -ENOENT;
4986
4987         spin_lock(&inode->i_lock);
4988         acl = nfsi->nfs4_acl;
4989         if (acl == NULL)
4990                 goto out;
4991         if (buf == NULL) /* user is just asking for length */
4992                 goto out_len;
4993         if (acl->cached == 0)
4994                 goto out;
4995         ret = -ERANGE; /* see getxattr(2) man page */
4996         if (acl->len > buflen)
4997                 goto out;
4998         memcpy(buf, acl->data, acl->len);
4999 out_len:
5000         ret = acl->len;
5001 out:
5002         spin_unlock(&inode->i_lock);
5003         return ret;
5004 }
5005
5006 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5007 {
5008         struct nfs4_cached_acl *acl;
5009         size_t buflen = sizeof(*acl) + acl_len;
5010
5011         if (buflen <= PAGE_SIZE) {
5012                 acl = kmalloc(buflen, GFP_KERNEL);
5013                 if (acl == NULL)
5014                         goto out;
5015                 acl->cached = 1;
5016                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5017         } else {
5018                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5019                 if (acl == NULL)
5020                         goto out;
5021                 acl->cached = 0;
5022         }
5023         acl->len = acl_len;
5024 out:
5025         nfs4_set_cached_acl(inode, acl);
5026 }
5027
5028 /*
5029  * The getxattr API returns the required buffer length when called with a
5030  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5031  * the required buf.  On a NULL buf, we send a page of data to the server
5032  * guessing that the ACL request can be serviced by a page. If so, we cache
5033  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5034  * the cache. If not so, we throw away the page, and cache the required
5035  * length. The next getxattr call will then produce another round trip to
5036  * the server, this time with the input buf of the required size.
5037  */
5038 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5039 {
5040         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
5041         struct nfs_getaclargs args = {
5042                 .fh = NFS_FH(inode),
5043                 .acl_pages = pages,
5044                 .acl_len = buflen,
5045         };
5046         struct nfs_getaclres res = {
5047                 .acl_len = buflen,
5048         };
5049         struct rpc_message msg = {
5050                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5051                 .rpc_argp = &args,
5052                 .rpc_resp = &res,
5053         };
5054         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5055         int ret = -ENOMEM, i;
5056
5057         /* As long as we're doing a round trip to the server anyway,
5058          * let's be prepared for a page of acl data. */
5059         if (npages == 0)
5060                 npages = 1;
5061         if (npages > ARRAY_SIZE(pages))
5062                 return -ERANGE;
5063
5064         for (i = 0; i < npages; i++) {
5065                 pages[i] = alloc_page(GFP_KERNEL);
5066                 if (!pages[i])
5067                         goto out_free;
5068         }
5069
5070         /* for decoding across pages */
5071         res.acl_scratch = alloc_page(GFP_KERNEL);
5072         if (!res.acl_scratch)
5073                 goto out_free;
5074
5075         args.acl_len = npages * PAGE_SIZE;
5076
5077         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5078                 __func__, buf, buflen, npages, args.acl_len);
5079         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5080                              &msg, &args.seq_args, &res.seq_res, 0);
5081         if (ret)
5082                 goto out_free;
5083
5084         /* Handle the case where the passed-in buffer is too short */
5085         if (res.acl_flags & NFS4_ACL_TRUNC) {
5086                 /* Did the user only issue a request for the acl length? */
5087                 if (buf == NULL)
5088                         goto out_ok;
5089                 ret = -ERANGE;
5090                 goto out_free;
5091         }
5092         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5093         if (buf) {
5094                 if (res.acl_len > buflen) {
5095                         ret = -ERANGE;
5096                         goto out_free;
5097                 }
5098                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5099         }
5100 out_ok:
5101         ret = res.acl_len;
5102 out_free:
5103         for (i = 0; i < npages; i++)
5104                 if (pages[i])
5105                         __free_page(pages[i]);
5106         if (res.acl_scratch)
5107                 __free_page(res.acl_scratch);
5108         return ret;
5109 }
5110
5111 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5112 {
5113         struct nfs4_exception exception = { };
5114         ssize_t ret;
5115         do {
5116                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5117                 trace_nfs4_get_acl(inode, ret);
5118                 if (ret >= 0)
5119                         break;
5120                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5121         } while (exception.retry);
5122         return ret;
5123 }
5124
5125 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5126 {
5127         struct nfs_server *server = NFS_SERVER(inode);
5128         int ret;
5129
5130         if (!nfs4_server_supports_acls(server))
5131                 return -EOPNOTSUPP;
5132         ret = nfs_revalidate_inode(server, inode);
5133         if (ret < 0)
5134                 return ret;
5135         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5136                 nfs_zap_acl_cache(inode);
5137         ret = nfs4_read_cached_acl(inode, buf, buflen);
5138         if (ret != -ENOENT)
5139                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5140                  * but no cached acl data, just the acl length */
5141                 return ret;
5142         return nfs4_get_acl_uncached(inode, buf, buflen);
5143 }
5144
5145 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5146 {
5147         struct nfs_server *server = NFS_SERVER(inode);
5148         struct page *pages[NFS4ACL_MAXPAGES];
5149         struct nfs_setaclargs arg = {
5150                 .fh             = NFS_FH(inode),
5151                 .acl_pages      = pages,
5152                 .acl_len        = buflen,
5153         };
5154         struct nfs_setaclres res;
5155         struct rpc_message msg = {
5156                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5157                 .rpc_argp       = &arg,
5158                 .rpc_resp       = &res,
5159         };
5160         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5161         int ret, i;
5162
5163         if (!nfs4_server_supports_acls(server))
5164                 return -EOPNOTSUPP;
5165         if (npages > ARRAY_SIZE(pages))
5166                 return -ERANGE;
5167         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5168         if (i < 0)
5169                 return i;
5170         nfs4_inode_return_delegation(inode);
5171         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5172
5173         /*
5174          * Free each page after tx, so the only ref left is
5175          * held by the network stack
5176          */
5177         for (; i > 0; i--)
5178                 put_page(pages[i-1]);
5179
5180         /*
5181          * Acl update can result in inode attribute update.
5182          * so mark the attribute cache invalid.
5183          */
5184         spin_lock(&inode->i_lock);
5185         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5186         spin_unlock(&inode->i_lock);
5187         nfs_access_zap_cache(inode);
5188         nfs_zap_acl_cache(inode);
5189         return ret;
5190 }
5191
5192 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5193 {
5194         struct nfs4_exception exception = { };
5195         int err;
5196         do {
5197                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5198                 trace_nfs4_set_acl(inode, err);
5199                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5200                                 &exception);
5201         } while (exception.retry);
5202         return err;
5203 }
5204
5205 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5206 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5207                                         size_t buflen)
5208 {
5209         struct nfs_server *server = NFS_SERVER(inode);
5210         struct nfs_fattr fattr;
5211         struct nfs4_label label = {0, 0, buflen, buf};
5212
5213         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5214         struct nfs4_getattr_arg arg = {
5215                 .fh             = NFS_FH(inode),
5216                 .bitmask        = bitmask,
5217         };
5218         struct nfs4_getattr_res res = {
5219                 .fattr          = &fattr,
5220                 .label          = &label,
5221                 .server         = server,
5222         };
5223         struct rpc_message msg = {
5224                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5225                 .rpc_argp       = &arg,
5226                 .rpc_resp       = &res,
5227         };
5228         int ret;
5229
5230         nfs_fattr_init(&fattr);
5231
5232         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5233         if (ret)
5234                 return ret;
5235         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5236                 return -ENOENT;
5237         if (buflen < label.len)
5238                 return -ERANGE;
5239         return 0;
5240 }
5241
5242 static int nfs4_get_security_label(struct inode *inode, void *buf,
5243                                         size_t buflen)
5244 {
5245         struct nfs4_exception exception = { };
5246         int err;
5247
5248         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5249                 return -EOPNOTSUPP;
5250
5251         do {
5252                 err = _nfs4_get_security_label(inode, buf, buflen);
5253                 trace_nfs4_get_security_label(inode, err);
5254                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5255                                 &exception);
5256         } while (exception.retry);
5257         return err;
5258 }
5259
5260 static int _nfs4_do_set_security_label(struct inode *inode,
5261                 struct nfs4_label *ilabel,
5262                 struct nfs_fattr *fattr,
5263                 struct nfs4_label *olabel)
5264 {
5265
5266         struct iattr sattr = {0};
5267         struct nfs_server *server = NFS_SERVER(inode);
5268         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5269         struct nfs_setattrargs arg = {
5270                 .fh             = NFS_FH(inode),
5271                 .iap            = &sattr,
5272                 .server         = server,
5273                 .bitmask        = bitmask,
5274                 .label          = ilabel,
5275         };
5276         struct nfs_setattrres res = {
5277                 .fattr          = fattr,
5278                 .label          = olabel,
5279                 .server         = server,
5280         };
5281         struct rpc_message msg = {
5282                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5283                 .rpc_argp       = &arg,
5284                 .rpc_resp       = &res,
5285         };
5286         int status;
5287
5288         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5289
5290         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5291         if (status)
5292                 dprintk("%s failed: %d\n", __func__, status);
5293
5294         return status;
5295 }
5296
5297 static int nfs4_do_set_security_label(struct inode *inode,
5298                 struct nfs4_label *ilabel,
5299                 struct nfs_fattr *fattr,
5300                 struct nfs4_label *olabel)
5301 {
5302         struct nfs4_exception exception = { };
5303         int err;
5304
5305         do {
5306                 err = _nfs4_do_set_security_label(inode, ilabel,
5307                                 fattr, olabel);
5308                 trace_nfs4_set_security_label(inode, err);
5309                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5310                                 &exception);
5311         } while (exception.retry);
5312         return err;
5313 }
5314
5315 static int
5316 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5317 {
5318         struct nfs4_label ilabel, *olabel = NULL;
5319         struct nfs_fattr fattr;
5320         struct rpc_cred *cred;
5321         int status;
5322
5323         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5324                 return -EOPNOTSUPP;
5325
5326         nfs_fattr_init(&fattr);
5327
5328         ilabel.pi = 0;
5329         ilabel.lfs = 0;
5330         ilabel.label = (char *)buf;
5331         ilabel.len = buflen;
5332
5333         cred = rpc_lookup_cred();
5334         if (IS_ERR(cred))
5335                 return PTR_ERR(cred);
5336
5337         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5338         if (IS_ERR(olabel)) {
5339                 status = -PTR_ERR(olabel);
5340                 goto out;
5341         }
5342
5343         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5344         if (status == 0)
5345                 nfs_setsecurity(inode, &fattr, olabel);
5346
5347         nfs4_label_free(olabel);
5348 out:
5349         put_rpccred(cred);
5350         return status;
5351 }
5352 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5353
5354
5355 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5356                                     nfs4_verifier *bootverf)
5357 {
5358         __be32 verf[2];
5359
5360         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5361                 /* An impossible timestamp guarantees this value
5362                  * will never match a generated boot time. */
5363                 verf[0] = cpu_to_be32(U32_MAX);
5364                 verf[1] = cpu_to_be32(U32_MAX);
5365         } else {
5366                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5367                 u64 ns = ktime_to_ns(nn->boot_time);
5368
5369                 verf[0] = cpu_to_be32(ns >> 32);
5370                 verf[1] = cpu_to_be32(ns);
5371         }
5372         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5373 }
5374
5375 static int
5376 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5377 {
5378         size_t len;
5379         char *str;
5380
5381         if (clp->cl_owner_id != NULL)
5382                 return 0;
5383
5384         rcu_read_lock();
5385         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5386                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5387                 1 +
5388                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5389                 1;
5390         rcu_read_unlock();
5391
5392         if (len > NFS4_OPAQUE_LIMIT + 1)
5393                 return -EINVAL;
5394
5395         /*
5396          * Since this string is allocated at mount time, and held until the
5397          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5398          * about a memory-reclaim deadlock.
5399          */
5400         str = kmalloc(len, GFP_KERNEL);
5401         if (!str)
5402                 return -ENOMEM;
5403
5404         rcu_read_lock();
5405         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5406                         clp->cl_ipaddr,
5407                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5408                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5409         rcu_read_unlock();
5410
5411         clp->cl_owner_id = str;
5412         return 0;
5413 }
5414
5415 static int
5416 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5417 {
5418         size_t len;
5419         char *str;
5420
5421         len = 10 + 10 + 1 + 10 + 1 +
5422                 strlen(nfs4_client_id_uniquifier) + 1 +
5423                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5424
5425         if (len > NFS4_OPAQUE_LIMIT + 1)
5426                 return -EINVAL;
5427
5428         /*
5429          * Since this string is allocated at mount time, and held until the
5430          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5431          * about a memory-reclaim deadlock.
5432          */
5433         str = kmalloc(len, GFP_KERNEL);
5434         if (!str)
5435                 return -ENOMEM;
5436
5437         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5438                         clp->rpc_ops->version, clp->cl_minorversion,
5439                         nfs4_client_id_uniquifier,
5440                         clp->cl_rpcclient->cl_nodename);
5441         clp->cl_owner_id = str;
5442         return 0;
5443 }
5444
5445 static int
5446 nfs4_init_uniform_client_string(struct nfs_client *clp)
5447 {
5448         size_t len;
5449         char *str;
5450
5451         if (clp->cl_owner_id != NULL)
5452                 return 0;
5453
5454         if (nfs4_client_id_uniquifier[0] != '\0')
5455                 return nfs4_init_uniquifier_client_string(clp);
5456
5457         len = 10 + 10 + 1 + 10 + 1 +
5458                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5459
5460         if (len > NFS4_OPAQUE_LIMIT + 1)
5461                 return -EINVAL;
5462
5463         /*
5464          * Since this string is allocated at mount time, and held until the
5465          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5466          * about a memory-reclaim deadlock.
5467          */
5468         str = kmalloc(len, GFP_KERNEL);
5469         if (!str)
5470                 return -ENOMEM;
5471
5472         scnprintf(str, len, "Linux NFSv%u.%u %s",
5473                         clp->rpc_ops->version, clp->cl_minorversion,
5474                         clp->cl_rpcclient->cl_nodename);
5475         clp->cl_owner_id = str;
5476         return 0;
5477 }
5478
5479 /*
5480  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5481  * services.  Advertise one based on the address family of the
5482  * clientaddr.
5483  */
5484 static unsigned int
5485 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5486 {
5487         if (strchr(clp->cl_ipaddr, ':') != NULL)
5488                 return scnprintf(buf, len, "tcp6");
5489         else
5490                 return scnprintf(buf, len, "tcp");
5491 }
5492
5493 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5494 {
5495         struct nfs4_setclientid *sc = calldata;
5496
5497         if (task->tk_status == 0)
5498                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5499 }
5500
5501 static const struct rpc_call_ops nfs4_setclientid_ops = {
5502         .rpc_call_done = nfs4_setclientid_done,
5503 };
5504
5505 /**
5506  * nfs4_proc_setclientid - Negotiate client ID
5507  * @clp: state data structure
5508  * @program: RPC program for NFSv4 callback service
5509  * @port: IP port number for NFS4 callback service
5510  * @cred: RPC credential to use for this call
5511  * @res: where to place the result
5512  *
5513  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5514  */
5515 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5516                 unsigned short port, struct rpc_cred *cred,
5517                 struct nfs4_setclientid_res *res)
5518 {
5519         nfs4_verifier sc_verifier;
5520         struct nfs4_setclientid setclientid = {
5521                 .sc_verifier = &sc_verifier,
5522                 .sc_prog = program,
5523                 .sc_clnt = clp,
5524         };
5525         struct rpc_message msg = {
5526                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5527                 .rpc_argp = &setclientid,
5528                 .rpc_resp = res,
5529                 .rpc_cred = cred,
5530         };
5531         struct rpc_task *task;
5532         struct rpc_task_setup task_setup_data = {
5533                 .rpc_client = clp->cl_rpcclient,
5534                 .rpc_message = &msg,
5535                 .callback_ops = &nfs4_setclientid_ops,
5536                 .callback_data = &setclientid,
5537                 .flags = RPC_TASK_TIMEOUT,
5538         };
5539         int status;
5540
5541         /* nfs_client_id4 */
5542         nfs4_init_boot_verifier(clp, &sc_verifier);
5543
5544         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5545                 status = nfs4_init_uniform_client_string(clp);
5546         else
5547                 status = nfs4_init_nonuniform_client_string(clp);
5548
5549         if (status)
5550                 goto out;
5551
5552         /* cb_client4 */
5553         setclientid.sc_netid_len =
5554                                 nfs4_init_callback_netid(clp,
5555                                                 setclientid.sc_netid,
5556                                                 sizeof(setclientid.sc_netid));
5557         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5558                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5559                                 clp->cl_ipaddr, port >> 8, port & 255);
5560
5561         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5562                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5563                 clp->cl_owner_id);
5564         task = rpc_run_task(&task_setup_data);
5565         if (IS_ERR(task)) {
5566                 status = PTR_ERR(task);
5567                 goto out;
5568         }
5569         status = task->tk_status;
5570         if (setclientid.sc_cred) {
5571                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5572                 put_rpccred(setclientid.sc_cred);
5573         }
5574         rpc_put_task(task);
5575 out:
5576         trace_nfs4_setclientid(clp, status);
5577         dprintk("NFS reply setclientid: %d\n", status);
5578         return status;
5579 }
5580
5581 /**
5582  * nfs4_proc_setclientid_confirm - Confirm client ID
5583  * @clp: state data structure
5584  * @res: result of a previous SETCLIENTID
5585  * @cred: RPC credential to use for this call
5586  *
5587  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5588  */
5589 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5590                 struct nfs4_setclientid_res *arg,
5591                 struct rpc_cred *cred)
5592 {
5593         struct rpc_message msg = {
5594                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5595                 .rpc_argp = arg,
5596                 .rpc_cred = cred,
5597         };
5598         int status;
5599
5600         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5601                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5602                 clp->cl_clientid);
5603         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5604         trace_nfs4_setclientid_confirm(clp, status);
5605         dprintk("NFS reply setclientid_confirm: %d\n", status);
5606         return status;
5607 }
5608
5609 struct nfs4_delegreturndata {
5610         struct nfs4_delegreturnargs args;
5611         struct nfs4_delegreturnres res;
5612         struct nfs_fh fh;
5613         nfs4_stateid stateid;
5614         unsigned long timestamp;
5615         struct {
5616                 struct nfs4_layoutreturn_args arg;
5617                 struct nfs4_layoutreturn_res res;
5618                 struct nfs4_xdr_opaque_data ld_private;
5619                 u32 roc_barrier;
5620                 bool roc;
5621         } lr;
5622         struct nfs_fattr fattr;
5623         int rpc_status;
5624         struct inode *inode;
5625 };
5626
5627 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5628 {
5629         struct nfs4_delegreturndata *data = calldata;
5630
5631         if (!nfs4_sequence_done(task, &data->res.seq_res))
5632                 return;
5633
5634         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5635
5636         /* Handle Layoutreturn errors */
5637         if (data->args.lr_args && task->tk_status != 0) {
5638                 switch(data->res.lr_ret) {
5639                 default:
5640                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5641                         break;
5642                 case 0:
5643                         data->args.lr_args = NULL;
5644                         data->res.lr_res = NULL;
5645                         break;
5646                 case -NFS4ERR_ADMIN_REVOKED:
5647                 case -NFS4ERR_DELEG_REVOKED:
5648                 case -NFS4ERR_EXPIRED:
5649                 case -NFS4ERR_BAD_STATEID:
5650                 case -NFS4ERR_OLD_STATEID:
5651                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5652                 case -NFS4ERR_WRONG_CRED:
5653                         data->args.lr_args = NULL;
5654                         data->res.lr_res = NULL;
5655                         data->res.lr_ret = 0;
5656                         rpc_restart_call_prepare(task);
5657                         return;
5658                 }
5659         }
5660
5661         switch (task->tk_status) {
5662         case 0:
5663                 renew_lease(data->res.server, data->timestamp);
5664                 break;
5665         case -NFS4ERR_ADMIN_REVOKED:
5666         case -NFS4ERR_DELEG_REVOKED:
5667         case -NFS4ERR_EXPIRED:
5668                 nfs4_free_revoked_stateid(data->res.server,
5669                                 data->args.stateid,
5670                                 task->tk_msg.rpc_cred);
5671         case -NFS4ERR_BAD_STATEID:
5672         case -NFS4ERR_OLD_STATEID:
5673         case -NFS4ERR_STALE_STATEID:
5674                 task->tk_status = 0;
5675                 break;
5676         default:
5677                 if (nfs4_async_handle_error(task, data->res.server,
5678                                             NULL, NULL) == -EAGAIN) {
5679                         rpc_restart_call_prepare(task);
5680                         return;
5681                 }
5682         }
5683         data->rpc_status = task->tk_status;
5684 }
5685
5686 static void nfs4_delegreturn_release(void *calldata)
5687 {
5688         struct nfs4_delegreturndata *data = calldata;
5689         struct inode *inode = data->inode;
5690
5691         if (inode) {
5692                 if (data->lr.roc)
5693                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
5694                                         data->res.lr_ret);
5695                 nfs_iput_and_deactive(inode);
5696         }
5697         kfree(calldata);
5698 }
5699
5700 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5701 {
5702         struct nfs4_delegreturndata *d_data;
5703
5704         d_data = (struct nfs4_delegreturndata *)data;
5705
5706         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5707                 return;
5708
5709         nfs4_setup_sequence(d_data->res.server,
5710                         &d_data->args.seq_args,
5711                         &d_data->res.seq_res,
5712                         task);
5713 }
5714
5715 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5716         .rpc_call_prepare = nfs4_delegreturn_prepare,
5717         .rpc_call_done = nfs4_delegreturn_done,
5718         .rpc_release = nfs4_delegreturn_release,
5719 };
5720
5721 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5722 {
5723         struct nfs4_delegreturndata *data;
5724         struct nfs_server *server = NFS_SERVER(inode);
5725         struct rpc_task *task;
5726         struct rpc_message msg = {
5727                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5728                 .rpc_cred = cred,
5729         };
5730         struct rpc_task_setup task_setup_data = {
5731                 .rpc_client = server->client,
5732                 .rpc_message = &msg,
5733                 .callback_ops = &nfs4_delegreturn_ops,
5734                 .flags = RPC_TASK_ASYNC,
5735         };
5736         int status = 0;
5737
5738         data = kzalloc(sizeof(*data), GFP_NOFS);
5739         if (data == NULL)
5740                 return -ENOMEM;
5741         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5742
5743         nfs4_state_protect(server->nfs_client,
5744                         NFS_SP4_MACH_CRED_CLEANUP,
5745                         &task_setup_data.rpc_client, &msg);
5746
5747         data->args.fhandle = &data->fh;
5748         data->args.stateid = &data->stateid;
5749         data->args.bitmask = server->cache_consistency_bitmask;
5750         nfs_copy_fh(&data->fh, NFS_FH(inode));
5751         nfs4_stateid_copy(&data->stateid, stateid);
5752         data->res.fattr = &data->fattr;
5753         data->res.server = server;
5754         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5755         data->lr.arg.ld_private = &data->lr.ld_private;
5756         nfs_fattr_init(data->res.fattr);
5757         data->timestamp = jiffies;
5758         data->rpc_status = 0;
5759         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5760         data->inode = nfs_igrab_and_active(inode);
5761         if (data->inode) {
5762                 if (data->lr.roc) {
5763                         data->args.lr_args = &data->lr.arg;
5764                         data->res.lr_res = &data->lr.res;
5765                 }
5766         } else if (data->lr.roc) {
5767                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5768                 data->lr.roc = false;
5769         }
5770
5771         task_setup_data.callback_data = data;
5772         msg.rpc_argp = &data->args;
5773         msg.rpc_resp = &data->res;
5774         task = rpc_run_task(&task_setup_data);
5775         if (IS_ERR(task))
5776                 return PTR_ERR(task);
5777         if (!issync)
5778                 goto out;
5779         status = nfs4_wait_for_completion_rpc_task(task);
5780         if (status != 0)
5781                 goto out;
5782         status = data->rpc_status;
5783         if (status == 0)
5784                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5785         else
5786                 nfs_refresh_inode(inode, &data->fattr);
5787 out:
5788         rpc_put_task(task);
5789         return status;
5790 }
5791
5792 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5793 {
5794         struct nfs_server *server = NFS_SERVER(inode);
5795         struct nfs4_exception exception = { };
5796         int err;
5797         do {
5798                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5799                 trace_nfs4_delegreturn(inode, stateid, err);
5800                 switch (err) {
5801                         case -NFS4ERR_STALE_STATEID:
5802                         case -NFS4ERR_EXPIRED:
5803                         case 0:
5804                                 return 0;
5805                 }
5806                 err = nfs4_handle_exception(server, err, &exception);
5807         } while (exception.retry);
5808         return err;
5809 }
5810
5811 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5812 {
5813         struct inode *inode = state->inode;
5814         struct nfs_server *server = NFS_SERVER(inode);
5815         struct nfs_client *clp = server->nfs_client;
5816         struct nfs_lockt_args arg = {
5817                 .fh = NFS_FH(inode),
5818                 .fl = request,
5819         };
5820         struct nfs_lockt_res res = {
5821                 .denied = request,
5822         };
5823         struct rpc_message msg = {
5824                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5825                 .rpc_argp       = &arg,
5826                 .rpc_resp       = &res,
5827                 .rpc_cred       = state->owner->so_cred,
5828         };
5829         struct nfs4_lock_state *lsp;
5830         int status;
5831
5832         arg.lock_owner.clientid = clp->cl_clientid;
5833         status = nfs4_set_lock_state(state, request);
5834         if (status != 0)
5835                 goto out;
5836         lsp = request->fl_u.nfs4_fl.owner;
5837         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5838         arg.lock_owner.s_dev = server->s_dev;
5839         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5840         switch (status) {
5841                 case 0:
5842                         request->fl_type = F_UNLCK;
5843                         break;
5844                 case -NFS4ERR_DENIED:
5845                         status = 0;
5846         }
5847         request->fl_ops->fl_release_private(request);
5848         request->fl_ops = NULL;
5849 out:
5850         return status;
5851 }
5852
5853 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5854 {
5855         struct nfs4_exception exception = { };
5856         int err;
5857
5858         do {
5859                 err = _nfs4_proc_getlk(state, cmd, request);
5860                 trace_nfs4_get_lock(request, state, cmd, err);
5861                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5862                                 &exception);
5863         } while (exception.retry);
5864         return err;
5865 }
5866
5867 struct nfs4_unlockdata {
5868         struct nfs_locku_args arg;
5869         struct nfs_locku_res res;
5870         struct nfs4_lock_state *lsp;
5871         struct nfs_open_context *ctx;
5872         struct file_lock fl;
5873         struct nfs_server *server;
5874         unsigned long timestamp;
5875 };
5876
5877 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5878                 struct nfs_open_context *ctx,
5879                 struct nfs4_lock_state *lsp,
5880                 struct nfs_seqid *seqid)
5881 {
5882         struct nfs4_unlockdata *p;
5883         struct inode *inode = lsp->ls_state->inode;
5884
5885         p = kzalloc(sizeof(*p), GFP_NOFS);
5886         if (p == NULL)
5887                 return NULL;
5888         p->arg.fh = NFS_FH(inode);
5889         p->arg.fl = &p->fl;
5890         p->arg.seqid = seqid;
5891         p->res.seqid = seqid;
5892         p->lsp = lsp;
5893         atomic_inc(&lsp->ls_count);
5894         /* Ensure we don't close file until we're done freeing locks! */
5895         p->ctx = get_nfs_open_context(ctx);
5896         memcpy(&p->fl, fl, sizeof(p->fl));
5897         p->server = NFS_SERVER(inode);
5898         return p;
5899 }
5900
5901 static void nfs4_locku_release_calldata(void *data)
5902 {
5903         struct nfs4_unlockdata *calldata = data;
5904         nfs_free_seqid(calldata->arg.seqid);
5905         nfs4_put_lock_state(calldata->lsp);
5906         put_nfs_open_context(calldata->ctx);
5907         kfree(calldata);
5908 }
5909
5910 static void nfs4_locku_done(struct rpc_task *task, void *data)
5911 {
5912         struct nfs4_unlockdata *calldata = data;
5913
5914         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5915                 return;
5916         switch (task->tk_status) {
5917                 case 0:
5918                         renew_lease(calldata->server, calldata->timestamp);
5919                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5920                         if (nfs4_update_lock_stateid(calldata->lsp,
5921                                         &calldata->res.stateid))
5922                                 break;
5923                 case -NFS4ERR_ADMIN_REVOKED:
5924                 case -NFS4ERR_EXPIRED:
5925                         nfs4_free_revoked_stateid(calldata->server,
5926                                         &calldata->arg.stateid,
5927                                         task->tk_msg.rpc_cred);
5928                 case -NFS4ERR_BAD_STATEID:
5929                 case -NFS4ERR_OLD_STATEID:
5930                 case -NFS4ERR_STALE_STATEID:
5931                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5932                                                 &calldata->lsp->ls_stateid))
5933                                 rpc_restart_call_prepare(task);
5934                         break;
5935                 default:
5936                         if (nfs4_async_handle_error(task, calldata->server,
5937                                                     NULL, NULL) == -EAGAIN)
5938                                 rpc_restart_call_prepare(task);
5939         }
5940         nfs_release_seqid(calldata->arg.seqid);
5941 }
5942
5943 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5944 {
5945         struct nfs4_unlockdata *calldata = data;
5946
5947         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5948                 goto out_wait;
5949         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5950         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5951                 /* Note: exit _without_ running nfs4_locku_done */
5952                 goto out_no_action;
5953         }
5954         calldata->timestamp = jiffies;
5955         if (nfs4_setup_sequence(calldata->server,
5956                                 &calldata->arg.seq_args,
5957                                 &calldata->res.seq_res,
5958                                 task) != 0)
5959                 nfs_release_seqid(calldata->arg.seqid);
5960         return;
5961 out_no_action:
5962         task->tk_action = NULL;
5963 out_wait:
5964         nfs4_sequence_done(task, &calldata->res.seq_res);
5965 }
5966
5967 static const struct rpc_call_ops nfs4_locku_ops = {
5968         .rpc_call_prepare = nfs4_locku_prepare,
5969         .rpc_call_done = nfs4_locku_done,
5970         .rpc_release = nfs4_locku_release_calldata,
5971 };
5972
5973 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5974                 struct nfs_open_context *ctx,
5975                 struct nfs4_lock_state *lsp,
5976                 struct nfs_seqid *seqid)
5977 {
5978         struct nfs4_unlockdata *data;
5979         struct rpc_message msg = {
5980                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5981                 .rpc_cred = ctx->cred,
5982         };
5983         struct rpc_task_setup task_setup_data = {
5984                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5985                 .rpc_message = &msg,
5986                 .callback_ops = &nfs4_locku_ops,
5987                 .workqueue = nfsiod_workqueue,
5988                 .flags = RPC_TASK_ASYNC,
5989         };
5990
5991         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5992                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5993
5994         /* Ensure this is an unlock - when canceling a lock, the
5995          * canceled lock is passed in, and it won't be an unlock.
5996          */
5997         fl->fl_type = F_UNLCK;
5998
5999         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6000         if (data == NULL) {
6001                 nfs_free_seqid(seqid);
6002                 return ERR_PTR(-ENOMEM);
6003         }
6004
6005         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6006         msg.rpc_argp = &data->arg;
6007         msg.rpc_resp = &data->res;
6008         task_setup_data.callback_data = data;
6009         return rpc_run_task(&task_setup_data);
6010 }
6011
6012 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6013 {
6014         struct inode *inode = state->inode;
6015         struct nfs4_state_owner *sp = state->owner;
6016         struct nfs_inode *nfsi = NFS_I(inode);
6017         struct nfs_seqid *seqid;
6018         struct nfs4_lock_state *lsp;
6019         struct rpc_task *task;
6020         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6021         int status = 0;
6022         unsigned char fl_flags = request->fl_flags;
6023
6024         status = nfs4_set_lock_state(state, request);
6025         /* Unlock _before_ we do the RPC call */
6026         request->fl_flags |= FL_EXISTS;
6027         /* Exclude nfs_delegation_claim_locks() */
6028         mutex_lock(&sp->so_delegreturn_mutex);
6029         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6030         down_read(&nfsi->rwsem);
6031         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6032                 up_read(&nfsi->rwsem);
6033                 mutex_unlock(&sp->so_delegreturn_mutex);
6034                 goto out;
6035         }
6036         up_read(&nfsi->rwsem);
6037         mutex_unlock(&sp->so_delegreturn_mutex);
6038         if (status != 0)
6039                 goto out;
6040         /* Is this a delegated lock? */
6041         lsp = request->fl_u.nfs4_fl.owner;
6042         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6043                 goto out;
6044         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6045         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6046         status = -ENOMEM;
6047         if (IS_ERR(seqid))
6048                 goto out;
6049         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6050         status = PTR_ERR(task);
6051         if (IS_ERR(task))
6052                 goto out;
6053         status = nfs4_wait_for_completion_rpc_task(task);
6054         rpc_put_task(task);
6055 out:
6056         request->fl_flags = fl_flags;
6057         trace_nfs4_unlock(request, state, F_SETLK, status);
6058         return status;
6059 }
6060
6061 struct nfs4_lockdata {
6062         struct nfs_lock_args arg;
6063         struct nfs_lock_res res;
6064         struct nfs4_lock_state *lsp;
6065         struct nfs_open_context *ctx;
6066         struct file_lock fl;
6067         unsigned long timestamp;
6068         int rpc_status;
6069         int cancelled;
6070         struct nfs_server *server;
6071 };
6072
6073 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6074                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6075                 gfp_t gfp_mask)
6076 {
6077         struct nfs4_lockdata *p;
6078         struct inode *inode = lsp->ls_state->inode;
6079         struct nfs_server *server = NFS_SERVER(inode);
6080         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6081
6082         p = kzalloc(sizeof(*p), gfp_mask);
6083         if (p == NULL)
6084                 return NULL;
6085
6086         p->arg.fh = NFS_FH(inode);
6087         p->arg.fl = &p->fl;
6088         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6089         if (IS_ERR(p->arg.open_seqid))
6090                 goto out_free;
6091         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6092         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6093         if (IS_ERR(p->arg.lock_seqid))
6094                 goto out_free_seqid;
6095         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6096         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6097         p->arg.lock_owner.s_dev = server->s_dev;
6098         p->res.lock_seqid = p->arg.lock_seqid;
6099         p->lsp = lsp;
6100         p->server = server;
6101         atomic_inc(&lsp->ls_count);
6102         p->ctx = get_nfs_open_context(ctx);
6103         get_file(fl->fl_file);
6104         memcpy(&p->fl, fl, sizeof(p->fl));
6105         return p;
6106 out_free_seqid:
6107         nfs_free_seqid(p->arg.open_seqid);
6108 out_free:
6109         kfree(p);
6110         return NULL;
6111 }
6112
6113 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6114 {
6115         struct nfs4_lockdata *data = calldata;
6116         struct nfs4_state *state = data->lsp->ls_state;
6117
6118         dprintk("%s: begin!\n", __func__);
6119         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6120                 goto out_wait;
6121         /* Do we need to do an open_to_lock_owner? */
6122         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6123                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6124                         goto out_release_lock_seqid;
6125                 }
6126                 nfs4_stateid_copy(&data->arg.open_stateid,
6127                                 &state->open_stateid);
6128                 data->arg.new_lock_owner = 1;
6129                 data->res.open_seqid = data->arg.open_seqid;
6130         } else {
6131                 data->arg.new_lock_owner = 0;
6132                 nfs4_stateid_copy(&data->arg.lock_stateid,
6133                                 &data->lsp->ls_stateid);
6134         }
6135         if (!nfs4_valid_open_stateid(state)) {
6136                 data->rpc_status = -EBADF;
6137                 task->tk_action = NULL;
6138                 goto out_release_open_seqid;
6139         }
6140         data->timestamp = jiffies;
6141         if (nfs4_setup_sequence(data->server,
6142                                 &data->arg.seq_args,
6143                                 &data->res.seq_res,
6144                                 task) == 0)
6145                 return;
6146 out_release_open_seqid:
6147         nfs_release_seqid(data->arg.open_seqid);
6148 out_release_lock_seqid:
6149         nfs_release_seqid(data->arg.lock_seqid);
6150 out_wait:
6151         nfs4_sequence_done(task, &data->res.seq_res);
6152         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6153 }
6154
6155 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6156 {
6157         struct nfs4_lockdata *data = calldata;
6158         struct nfs4_lock_state *lsp = data->lsp;
6159
6160         dprintk("%s: begin!\n", __func__);
6161
6162         if (!nfs4_sequence_done(task, &data->res.seq_res))
6163                 return;
6164
6165         data->rpc_status = task->tk_status;
6166         switch (task->tk_status) {
6167         case 0:
6168                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6169                                 data->timestamp);
6170                 if (data->arg.new_lock) {
6171                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6172                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6173                                 rpc_restart_call_prepare(task);
6174                                 break;
6175                         }
6176                 }
6177                 if (data->arg.new_lock_owner != 0) {
6178                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6179                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6180                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6181                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6182                         rpc_restart_call_prepare(task);
6183                 break;
6184         case -NFS4ERR_BAD_STATEID:
6185         case -NFS4ERR_OLD_STATEID:
6186         case -NFS4ERR_STALE_STATEID:
6187         case -NFS4ERR_EXPIRED:
6188                 if (data->arg.new_lock_owner != 0) {
6189                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6190                                                 &lsp->ls_state->open_stateid))
6191                                 rpc_restart_call_prepare(task);
6192                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6193                                                 &lsp->ls_stateid))
6194                                 rpc_restart_call_prepare(task);
6195         }
6196         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6197 }
6198
6199 static void nfs4_lock_release(void *calldata)
6200 {
6201         struct nfs4_lockdata *data = calldata;
6202
6203         dprintk("%s: begin!\n", __func__);
6204         nfs_free_seqid(data->arg.open_seqid);
6205         if (data->cancelled != 0) {
6206                 struct rpc_task *task;
6207                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6208                                 data->arg.lock_seqid);
6209                 if (!IS_ERR(task))
6210                         rpc_put_task_async(task);
6211                 dprintk("%s: cancelling lock!\n", __func__);
6212         } else
6213                 nfs_free_seqid(data->arg.lock_seqid);
6214         nfs4_put_lock_state(data->lsp);
6215         put_nfs_open_context(data->ctx);
6216         fput(data->fl.fl_file);
6217         kfree(data);
6218         dprintk("%s: done!\n", __func__);
6219 }
6220
6221 static const struct rpc_call_ops nfs4_lock_ops = {
6222         .rpc_call_prepare = nfs4_lock_prepare,
6223         .rpc_call_done = nfs4_lock_done,
6224         .rpc_release = nfs4_lock_release,
6225 };
6226
6227 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6228 {
6229         switch (error) {
6230         case -NFS4ERR_ADMIN_REVOKED:
6231         case -NFS4ERR_EXPIRED:
6232         case -NFS4ERR_BAD_STATEID:
6233                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6234                 if (new_lock_owner != 0 ||
6235                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6236                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6237                 break;
6238         case -NFS4ERR_STALE_STATEID:
6239                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6240                 nfs4_schedule_lease_recovery(server->nfs_client);
6241         };
6242 }
6243
6244 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6245 {
6246         struct nfs4_lockdata *data;
6247         struct rpc_task *task;
6248         struct rpc_message msg = {
6249                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6250                 .rpc_cred = state->owner->so_cred,
6251         };
6252         struct rpc_task_setup task_setup_data = {
6253                 .rpc_client = NFS_CLIENT(state->inode),
6254                 .rpc_message = &msg,
6255                 .callback_ops = &nfs4_lock_ops,
6256                 .workqueue = nfsiod_workqueue,
6257                 .flags = RPC_TASK_ASYNC,
6258         };
6259         int ret;
6260
6261         dprintk("%s: begin!\n", __func__);
6262         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6263                         fl->fl_u.nfs4_fl.owner,
6264                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6265         if (data == NULL)
6266                 return -ENOMEM;
6267         if (IS_SETLKW(cmd))
6268                 data->arg.block = 1;
6269         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6270         msg.rpc_argp = &data->arg;
6271         msg.rpc_resp = &data->res;
6272         task_setup_data.callback_data = data;
6273         if (recovery_type > NFS_LOCK_NEW) {
6274                 if (recovery_type == NFS_LOCK_RECLAIM)
6275                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6276                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6277         } else
6278                 data->arg.new_lock = 1;
6279         task = rpc_run_task(&task_setup_data);
6280         if (IS_ERR(task))
6281                 return PTR_ERR(task);
6282         ret = nfs4_wait_for_completion_rpc_task(task);
6283         if (ret == 0) {
6284                 ret = data->rpc_status;
6285                 if (ret)
6286                         nfs4_handle_setlk_error(data->server, data->lsp,
6287                                         data->arg.new_lock_owner, ret);
6288         } else
6289                 data->cancelled = 1;
6290         rpc_put_task(task);
6291         dprintk("%s: done, ret = %d!\n", __func__, ret);
6292         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6293         return ret;
6294 }
6295
6296 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6297 {
6298         struct nfs_server *server = NFS_SERVER(state->inode);
6299         struct nfs4_exception exception = {
6300                 .inode = state->inode,
6301         };
6302         int err;
6303
6304         do {
6305                 /* Cache the lock if possible... */
6306                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6307                         return 0;
6308                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6309                 if (err != -NFS4ERR_DELAY)
6310                         break;
6311                 nfs4_handle_exception(server, err, &exception);
6312         } while (exception.retry);
6313         return err;
6314 }
6315
6316 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6317 {
6318         struct nfs_server *server = NFS_SERVER(state->inode);
6319         struct nfs4_exception exception = {
6320                 .inode = state->inode,
6321         };
6322         int err;
6323
6324         err = nfs4_set_lock_state(state, request);
6325         if (err != 0)
6326                 return err;
6327         if (!recover_lost_locks) {
6328                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6329                 return 0;
6330         }
6331         do {
6332                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6333                         return 0;
6334                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6335                 switch (err) {
6336                 default:
6337                         goto out;
6338                 case -NFS4ERR_GRACE:
6339                 case -NFS4ERR_DELAY:
6340                         nfs4_handle_exception(server, err, &exception);
6341                         err = 0;
6342                 }
6343         } while (exception.retry);
6344 out:
6345         return err;
6346 }
6347
6348 #if defined(CONFIG_NFS_V4_1)
6349 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6350 {
6351         struct nfs4_lock_state *lsp;
6352         int status;
6353
6354         status = nfs4_set_lock_state(state, request);
6355         if (status != 0)
6356                 return status;
6357         lsp = request->fl_u.nfs4_fl.owner;
6358         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6359             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6360                 return 0;
6361         status = nfs4_lock_expired(state, request);
6362         return status;
6363 }
6364 #endif
6365
6366 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6367 {
6368         struct nfs_inode *nfsi = NFS_I(state->inode);
6369         struct nfs4_state_owner *sp = state->owner;
6370         unsigned char fl_flags = request->fl_flags;
6371         int status;
6372
6373         request->fl_flags |= FL_ACCESS;
6374         status = locks_lock_inode_wait(state->inode, request);
6375         if (status < 0)
6376                 goto out;
6377         mutex_lock(&sp->so_delegreturn_mutex);
6378         down_read(&nfsi->rwsem);
6379         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6380                 /* Yes: cache locks! */
6381                 /* ...but avoid races with delegation recall... */
6382                 request->fl_flags = fl_flags & ~FL_SLEEP;
6383                 status = locks_lock_inode_wait(state->inode, request);
6384                 up_read(&nfsi->rwsem);
6385                 mutex_unlock(&sp->so_delegreturn_mutex);
6386                 goto out;
6387         }
6388         up_read(&nfsi->rwsem);
6389         mutex_unlock(&sp->so_delegreturn_mutex);
6390         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6391 out:
6392         request->fl_flags = fl_flags;
6393         return status;
6394 }
6395
6396 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6397 {
6398         struct nfs4_exception exception = {
6399                 .state = state,
6400                 .inode = state->inode,
6401         };
6402         int err;
6403
6404         do {
6405                 err = _nfs4_proc_setlk(state, cmd, request);
6406                 if (err == -NFS4ERR_DENIED)
6407                         err = -EAGAIN;
6408                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6409                                 err, &exception);
6410         } while (exception.retry);
6411         return err;
6412 }
6413
6414 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6415 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6416
6417 static int
6418 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6419                         struct file_lock *request)
6420 {
6421         int             status = -ERESTARTSYS;
6422         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6423
6424         while(!signalled()) {
6425                 status = nfs4_proc_setlk(state, cmd, request);
6426                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6427                         break;
6428                 freezable_schedule_timeout_interruptible(timeout);
6429                 timeout *= 2;
6430                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6431                 status = -ERESTARTSYS;
6432         }
6433         return status;
6434 }
6435
6436 #ifdef CONFIG_NFS_V4_1
6437 struct nfs4_lock_waiter {
6438         struct task_struct      *task;
6439         struct inode            *inode;
6440         struct nfs_lowner       *owner;
6441         bool                    notified;
6442 };
6443
6444 static int
6445 nfs4_wake_lock_waiter(wait_queue_t *wait, unsigned int mode, int flags, void *key)
6446 {
6447         int ret;
6448         struct cb_notify_lock_args *cbnl = key;
6449         struct nfs4_lock_waiter *waiter = wait->private;
6450         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6451                                 *wowner = waiter->owner;
6452
6453         /* Only wake if the callback was for the same owner */
6454         if (lowner->clientid != wowner->clientid ||
6455             lowner->id != wowner->id             ||
6456             lowner->s_dev != wowner->s_dev)
6457                 return 0;
6458
6459         /* Make sure it's for the right inode */
6460         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6461                 return 0;
6462
6463         waiter->notified = true;
6464
6465         /* override "private" so we can use default_wake_function */
6466         wait->private = waiter->task;
6467         ret = autoremove_wake_function(wait, mode, flags, key);
6468         wait->private = waiter;
6469         return ret;
6470 }
6471
6472 static int
6473 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6474 {
6475         int status = -ERESTARTSYS;
6476         unsigned long flags;
6477         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6478         struct nfs_server *server = NFS_SERVER(state->inode);
6479         struct nfs_client *clp = server->nfs_client;
6480         wait_queue_head_t *q = &clp->cl_lock_waitq;
6481         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6482                                     .id = lsp->ls_seqid.owner_id,
6483                                     .s_dev = server->s_dev };
6484         struct nfs4_lock_waiter waiter = { .task  = current,
6485                                            .inode = state->inode,
6486                                            .owner = &owner,
6487                                            .notified = false };
6488         wait_queue_t wait;
6489
6490         /* Don't bother with waitqueue if we don't expect a callback */
6491         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6492                 return nfs4_retry_setlk_simple(state, cmd, request);
6493
6494         init_wait(&wait);
6495         wait.private = &waiter;
6496         wait.func = nfs4_wake_lock_waiter;
6497         add_wait_queue(q, &wait);
6498
6499         while(!signalled()) {
6500                 status = nfs4_proc_setlk(state, cmd, request);
6501                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6502                         break;
6503
6504                 status = -ERESTARTSYS;
6505                 spin_lock_irqsave(&q->lock, flags);
6506                 if (waiter.notified) {
6507                         spin_unlock_irqrestore(&q->lock, flags);
6508                         continue;
6509                 }
6510                 set_current_state(TASK_INTERRUPTIBLE);
6511                 spin_unlock_irqrestore(&q->lock, flags);
6512
6513                 freezable_schedule_timeout_interruptible(NFS4_LOCK_MAXTIMEOUT);
6514         }
6515
6516         finish_wait(q, &wait);
6517         return status;
6518 }
6519 #else /* !CONFIG_NFS_V4_1 */
6520 static inline int
6521 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6522 {
6523         return nfs4_retry_setlk_simple(state, cmd, request);
6524 }
6525 #endif
6526
6527 static int
6528 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6529 {
6530         struct nfs_open_context *ctx;
6531         struct nfs4_state *state;
6532         int status;
6533
6534         /* verify open state */
6535         ctx = nfs_file_open_context(filp);
6536         state = ctx->state;
6537
6538         if (request->fl_start < 0 || request->fl_end < 0)
6539                 return -EINVAL;
6540
6541         if (IS_GETLK(cmd)) {
6542                 if (state != NULL)
6543                         return nfs4_proc_getlk(state, F_GETLK, request);
6544                 return 0;
6545         }
6546
6547         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6548                 return -EINVAL;
6549
6550         if (request->fl_type == F_UNLCK) {
6551                 if (state != NULL)
6552                         return nfs4_proc_unlck(state, cmd, request);
6553                 return 0;
6554         }
6555
6556         if (state == NULL)
6557                 return -ENOLCK;
6558
6559         if ((request->fl_flags & FL_POSIX) &&
6560             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6561                 return -ENOLCK;
6562
6563         /*
6564          * Don't rely on the VFS having checked the file open mode,
6565          * since it won't do this for flock() locks.
6566          */
6567         switch (request->fl_type) {
6568         case F_RDLCK:
6569                 if (!(filp->f_mode & FMODE_READ))
6570                         return -EBADF;
6571                 break;
6572         case F_WRLCK:
6573                 if (!(filp->f_mode & FMODE_WRITE))
6574                         return -EBADF;
6575         }
6576
6577         status = nfs4_set_lock_state(state, request);
6578         if (status != 0)
6579                 return status;
6580
6581         return nfs4_retry_setlk(state, cmd, request);
6582 }
6583
6584 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6585 {
6586         struct nfs_server *server = NFS_SERVER(state->inode);
6587         int err;
6588
6589         err = nfs4_set_lock_state(state, fl);
6590         if (err != 0)
6591                 return err;
6592         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6593         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6594 }
6595
6596 struct nfs_release_lockowner_data {
6597         struct nfs4_lock_state *lsp;
6598         struct nfs_server *server;
6599         struct nfs_release_lockowner_args args;
6600         struct nfs_release_lockowner_res res;
6601         unsigned long timestamp;
6602 };
6603
6604 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6605 {
6606         struct nfs_release_lockowner_data *data = calldata;
6607         struct nfs_server *server = data->server;
6608         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6609                              &data->args.seq_args, &data->res.seq_res, task);
6610         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6611         data->timestamp = jiffies;
6612 }
6613
6614 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6615 {
6616         struct nfs_release_lockowner_data *data = calldata;
6617         struct nfs_server *server = data->server;
6618
6619         nfs40_sequence_done(task, &data->res.seq_res);
6620
6621         switch (task->tk_status) {
6622         case 0:
6623                 renew_lease(server, data->timestamp);
6624                 break;
6625         case -NFS4ERR_STALE_CLIENTID:
6626         case -NFS4ERR_EXPIRED:
6627                 nfs4_schedule_lease_recovery(server->nfs_client);
6628                 break;
6629         case -NFS4ERR_LEASE_MOVED:
6630         case -NFS4ERR_DELAY:
6631                 if (nfs4_async_handle_error(task, server,
6632                                             NULL, NULL) == -EAGAIN)
6633                         rpc_restart_call_prepare(task);
6634         }
6635 }
6636
6637 static void nfs4_release_lockowner_release(void *calldata)
6638 {
6639         struct nfs_release_lockowner_data *data = calldata;
6640         nfs4_free_lock_state(data->server, data->lsp);
6641         kfree(calldata);
6642 }
6643
6644 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6645         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6646         .rpc_call_done = nfs4_release_lockowner_done,
6647         .rpc_release = nfs4_release_lockowner_release,
6648 };
6649
6650 static void
6651 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6652 {
6653         struct nfs_release_lockowner_data *data;
6654         struct rpc_message msg = {
6655                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6656         };
6657
6658         if (server->nfs_client->cl_mvops->minor_version != 0)
6659                 return;
6660
6661         data = kmalloc(sizeof(*data), GFP_NOFS);
6662         if (!data)
6663                 return;
6664         data->lsp = lsp;
6665         data->server = server;
6666         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6667         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6668         data->args.lock_owner.s_dev = server->s_dev;
6669
6670         msg.rpc_argp = &data->args;
6671         msg.rpc_resp = &data->res;
6672         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6673         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6674 }
6675
6676 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6677
6678 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6679                                    struct dentry *unused, struct inode *inode,
6680                                    const char *key, const void *buf,
6681                                    size_t buflen, int flags)
6682 {
6683         return nfs4_proc_set_acl(inode, buf, buflen);
6684 }
6685
6686 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6687                                    struct dentry *unused, struct inode *inode,
6688                                    const char *key, void *buf, size_t buflen)
6689 {
6690         return nfs4_proc_get_acl(inode, buf, buflen);
6691 }
6692
6693 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6694 {
6695         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6696 }
6697
6698 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6699
6700 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6701                                      struct dentry *unused, struct inode *inode,
6702                                      const char *key, const void *buf,
6703                                      size_t buflen, int flags)
6704 {
6705         if (security_ismaclabel(key))
6706                 return nfs4_set_security_label(inode, buf, buflen);
6707
6708         return -EOPNOTSUPP;
6709 }
6710
6711 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6712                                      struct dentry *unused, struct inode *inode,
6713                                      const char *key, void *buf, size_t buflen)
6714 {
6715         if (security_ismaclabel(key))
6716                 return nfs4_get_security_label(inode, buf, buflen);
6717         return -EOPNOTSUPP;
6718 }
6719
6720 static ssize_t
6721 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6722 {
6723         int len = 0;
6724
6725         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6726                 len = security_inode_listsecurity(inode, list, list_len);
6727                 if (list_len && len > list_len)
6728                         return -ERANGE;
6729         }
6730         return len;
6731 }
6732
6733 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6734         .prefix = XATTR_SECURITY_PREFIX,
6735         .get    = nfs4_xattr_get_nfs4_label,
6736         .set    = nfs4_xattr_set_nfs4_label,
6737 };
6738
6739 #else
6740
6741 static ssize_t
6742 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6743 {
6744         return 0;
6745 }
6746
6747 #endif
6748
6749 /*
6750  * nfs_fhget will use either the mounted_on_fileid or the fileid
6751  */
6752 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6753 {
6754         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6755                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6756               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6757               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6758                 return;
6759
6760         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6761                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6762         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6763         fattr->nlink = 2;
6764 }
6765
6766 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6767                                    const struct qstr *name,
6768                                    struct nfs4_fs_locations *fs_locations,
6769                                    struct page *page)
6770 {
6771         struct nfs_server *server = NFS_SERVER(dir);
6772         u32 bitmask[3] = {
6773                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6774         };
6775         struct nfs4_fs_locations_arg args = {
6776                 .dir_fh = NFS_FH(dir),
6777                 .name = name,
6778                 .page = page,
6779                 .bitmask = bitmask,
6780         };
6781         struct nfs4_fs_locations_res res = {
6782                 .fs_locations = fs_locations,
6783         };
6784         struct rpc_message msg = {
6785                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6786                 .rpc_argp = &args,
6787                 .rpc_resp = &res,
6788         };
6789         int status;
6790
6791         dprintk("%s: start\n", __func__);
6792
6793         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6794          * is not supported */
6795         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6796                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6797         else
6798                 bitmask[0] |= FATTR4_WORD0_FILEID;
6799
6800         nfs_fattr_init(&fs_locations->fattr);
6801         fs_locations->server = server;
6802         fs_locations->nlocations = 0;
6803         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6804         dprintk("%s: returned status = %d\n", __func__, status);
6805         return status;
6806 }
6807
6808 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6809                            const struct qstr *name,
6810                            struct nfs4_fs_locations *fs_locations,
6811                            struct page *page)
6812 {
6813         struct nfs4_exception exception = { };
6814         int err;
6815         do {
6816                 err = _nfs4_proc_fs_locations(client, dir, name,
6817                                 fs_locations, page);
6818                 trace_nfs4_get_fs_locations(dir, name, err);
6819                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6820                                 &exception);
6821         } while (exception.retry);
6822         return err;
6823 }
6824
6825 /*
6826  * This operation also signals the server that this client is
6827  * performing migration recovery.  The server can stop returning
6828  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6829  * appended to this compound to identify the client ID which is
6830  * performing recovery.
6831  */
6832 static int _nfs40_proc_get_locations(struct inode *inode,
6833                                      struct nfs4_fs_locations *locations,
6834                                      struct page *page, struct rpc_cred *cred)
6835 {
6836         struct nfs_server *server = NFS_SERVER(inode);
6837         struct rpc_clnt *clnt = server->client;
6838         u32 bitmask[2] = {
6839                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6840         };
6841         struct nfs4_fs_locations_arg args = {
6842                 .clientid       = server->nfs_client->cl_clientid,
6843                 .fh             = NFS_FH(inode),
6844                 .page           = page,
6845                 .bitmask        = bitmask,
6846                 .migration      = 1,            /* skip LOOKUP */
6847                 .renew          = 1,            /* append RENEW */
6848         };
6849         struct nfs4_fs_locations_res res = {
6850                 .fs_locations   = locations,
6851                 .migration      = 1,
6852                 .renew          = 1,
6853         };
6854         struct rpc_message msg = {
6855                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6856                 .rpc_argp       = &args,
6857                 .rpc_resp       = &res,
6858                 .rpc_cred       = cred,
6859         };
6860         unsigned long now = jiffies;
6861         int status;
6862
6863         nfs_fattr_init(&locations->fattr);
6864         locations->server = server;
6865         locations->nlocations = 0;
6866
6867         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6868         nfs4_set_sequence_privileged(&args.seq_args);
6869         status = nfs4_call_sync_sequence(clnt, server, &msg,
6870                                         &args.seq_args, &res.seq_res);
6871         if (status)
6872                 return status;
6873
6874         renew_lease(server, now);
6875         return 0;
6876 }
6877
6878 #ifdef CONFIG_NFS_V4_1
6879
6880 /*
6881  * This operation also signals the server that this client is
6882  * performing migration recovery.  The server can stop asserting
6883  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6884  * performing this operation is identified in the SEQUENCE
6885  * operation in this compound.
6886  *
6887  * When the client supports GETATTR(fs_locations_info), it can
6888  * be plumbed in here.
6889  */
6890 static int _nfs41_proc_get_locations(struct inode *inode,
6891                                      struct nfs4_fs_locations *locations,
6892                                      struct page *page, struct rpc_cred *cred)
6893 {
6894         struct nfs_server *server = NFS_SERVER(inode);
6895         struct rpc_clnt *clnt = server->client;
6896         u32 bitmask[2] = {
6897                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6898         };
6899         struct nfs4_fs_locations_arg args = {
6900                 .fh             = NFS_FH(inode),
6901                 .page           = page,
6902                 .bitmask        = bitmask,
6903                 .migration      = 1,            /* skip LOOKUP */
6904         };
6905         struct nfs4_fs_locations_res res = {
6906                 .fs_locations   = locations,
6907                 .migration      = 1,
6908         };
6909         struct rpc_message msg = {
6910                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6911                 .rpc_argp       = &args,
6912                 .rpc_resp       = &res,
6913                 .rpc_cred       = cred,
6914         };
6915         int status;
6916
6917         nfs_fattr_init(&locations->fattr);
6918         locations->server = server;
6919         locations->nlocations = 0;
6920
6921         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6922         nfs4_set_sequence_privileged(&args.seq_args);
6923         status = nfs4_call_sync_sequence(clnt, server, &msg,
6924                                         &args.seq_args, &res.seq_res);
6925         if (status == NFS4_OK &&
6926             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6927                 status = -NFS4ERR_LEASE_MOVED;
6928         return status;
6929 }
6930
6931 #endif  /* CONFIG_NFS_V4_1 */
6932
6933 /**
6934  * nfs4_proc_get_locations - discover locations for a migrated FSID
6935  * @inode: inode on FSID that is migrating
6936  * @locations: result of query
6937  * @page: buffer
6938  * @cred: credential to use for this operation
6939  *
6940  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6941  * operation failed, or a negative errno if a local error occurred.
6942  *
6943  * On success, "locations" is filled in, but if the server has
6944  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6945  * asserted.
6946  *
6947  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6948  * from this client that require migration recovery.
6949  */
6950 int nfs4_proc_get_locations(struct inode *inode,
6951                             struct nfs4_fs_locations *locations,
6952                             struct page *page, struct rpc_cred *cred)
6953 {
6954         struct nfs_server *server = NFS_SERVER(inode);
6955         struct nfs_client *clp = server->nfs_client;
6956         const struct nfs4_mig_recovery_ops *ops =
6957                                         clp->cl_mvops->mig_recovery_ops;
6958         struct nfs4_exception exception = { };
6959         int status;
6960
6961         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6962                 (unsigned long long)server->fsid.major,
6963                 (unsigned long long)server->fsid.minor,
6964                 clp->cl_hostname);
6965         nfs_display_fhandle(NFS_FH(inode), __func__);
6966
6967         do {
6968                 status = ops->get_locations(inode, locations, page, cred);
6969                 if (status != -NFS4ERR_DELAY)
6970                         break;
6971                 nfs4_handle_exception(server, status, &exception);
6972         } while (exception.retry);
6973         return status;
6974 }
6975
6976 /*
6977  * This operation also signals the server that this client is
6978  * performing "lease moved" recovery.  The server can stop
6979  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6980  * is appended to this compound to identify the client ID which is
6981  * performing recovery.
6982  */
6983 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6984 {
6985         struct nfs_server *server = NFS_SERVER(inode);
6986         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6987         struct rpc_clnt *clnt = server->client;
6988         struct nfs4_fsid_present_arg args = {
6989                 .fh             = NFS_FH(inode),
6990                 .clientid       = clp->cl_clientid,
6991                 .renew          = 1,            /* append RENEW */
6992         };
6993         struct nfs4_fsid_present_res res = {
6994                 .renew          = 1,
6995         };
6996         struct rpc_message msg = {
6997                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6998                 .rpc_argp       = &args,
6999                 .rpc_resp       = &res,
7000                 .rpc_cred       = cred,
7001         };
7002         unsigned long now = jiffies;
7003         int status;
7004
7005         res.fh = nfs_alloc_fhandle();
7006         if (res.fh == NULL)
7007                 return -ENOMEM;
7008
7009         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7010         nfs4_set_sequence_privileged(&args.seq_args);
7011         status = nfs4_call_sync_sequence(clnt, server, &msg,
7012                                                 &args.seq_args, &res.seq_res);
7013         nfs_free_fhandle(res.fh);
7014         if (status)
7015                 return status;
7016
7017         do_renew_lease(clp, now);
7018         return 0;
7019 }
7020
7021 #ifdef CONFIG_NFS_V4_1
7022
7023 /*
7024  * This operation also signals the server that this client is
7025  * performing "lease moved" recovery.  The server can stop asserting
7026  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7027  * this operation is identified in the SEQUENCE operation in this
7028  * compound.
7029  */
7030 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7031 {
7032         struct nfs_server *server = NFS_SERVER(inode);
7033         struct rpc_clnt *clnt = server->client;
7034         struct nfs4_fsid_present_arg args = {
7035                 .fh             = NFS_FH(inode),
7036         };
7037         struct nfs4_fsid_present_res res = {
7038         };
7039         struct rpc_message msg = {
7040                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7041                 .rpc_argp       = &args,
7042                 .rpc_resp       = &res,
7043                 .rpc_cred       = cred,
7044         };
7045         int status;
7046
7047         res.fh = nfs_alloc_fhandle();
7048         if (res.fh == NULL)
7049                 return -ENOMEM;
7050
7051         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7052         nfs4_set_sequence_privileged(&args.seq_args);
7053         status = nfs4_call_sync_sequence(clnt, server, &msg,
7054                                                 &args.seq_args, &res.seq_res);
7055         nfs_free_fhandle(res.fh);
7056         if (status == NFS4_OK &&
7057             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7058                 status = -NFS4ERR_LEASE_MOVED;
7059         return status;
7060 }
7061
7062 #endif  /* CONFIG_NFS_V4_1 */
7063
7064 /**
7065  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7066  * @inode: inode on FSID to check
7067  * @cred: credential to use for this operation
7068  *
7069  * Server indicates whether the FSID is present, moved, or not
7070  * recognized.  This operation is necessary to clear a LEASE_MOVED
7071  * condition for this client ID.
7072  *
7073  * Returns NFS4_OK if the FSID is present on this server,
7074  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7075  *  NFS4ERR code if some error occurred on the server, or a
7076  *  negative errno if a local failure occurred.
7077  */
7078 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7079 {
7080         struct nfs_server *server = NFS_SERVER(inode);
7081         struct nfs_client *clp = server->nfs_client;
7082         const struct nfs4_mig_recovery_ops *ops =
7083                                         clp->cl_mvops->mig_recovery_ops;
7084         struct nfs4_exception exception = { };
7085         int status;
7086
7087         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7088                 (unsigned long long)server->fsid.major,
7089                 (unsigned long long)server->fsid.minor,
7090                 clp->cl_hostname);
7091         nfs_display_fhandle(NFS_FH(inode), __func__);
7092
7093         do {
7094                 status = ops->fsid_present(inode, cred);
7095                 if (status != -NFS4ERR_DELAY)
7096                         break;
7097                 nfs4_handle_exception(server, status, &exception);
7098         } while (exception.retry);
7099         return status;
7100 }
7101
7102 /**
7103  * If 'use_integrity' is true and the state managment nfs_client
7104  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7105  * and the machine credential as per RFC3530bis and RFC5661 Security
7106  * Considerations sections. Otherwise, just use the user cred with the
7107  * filesystem's rpc_client.
7108  */
7109 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7110 {
7111         int status;
7112         struct nfs4_secinfo_arg args = {
7113                 .dir_fh = NFS_FH(dir),
7114                 .name   = name,
7115         };
7116         struct nfs4_secinfo_res res = {
7117                 .flavors     = flavors,
7118         };
7119         struct rpc_message msg = {
7120                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7121                 .rpc_argp = &args,
7122                 .rpc_resp = &res,
7123         };
7124         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7125         struct rpc_cred *cred = NULL;
7126
7127         if (use_integrity) {
7128                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7129                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7130                 msg.rpc_cred = cred;
7131         }
7132
7133         dprintk("NFS call  secinfo %s\n", name->name);
7134
7135         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7136                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7137
7138         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7139                                 &res.seq_res, 0);
7140         dprintk("NFS reply  secinfo: %d\n", status);
7141
7142         if (cred)
7143                 put_rpccred(cred);
7144
7145         return status;
7146 }
7147
7148 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7149                       struct nfs4_secinfo_flavors *flavors)
7150 {
7151         struct nfs4_exception exception = { };
7152         int err;
7153         do {
7154                 err = -NFS4ERR_WRONGSEC;
7155
7156                 /* try to use integrity protection with machine cred */
7157                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7158                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7159
7160                 /*
7161                  * if unable to use integrity protection, or SECINFO with
7162                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7163                  * disallowed by spec, but exists in deployed servers) use
7164                  * the current filesystem's rpc_client and the user cred.
7165                  */
7166                 if (err == -NFS4ERR_WRONGSEC)
7167                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7168
7169                 trace_nfs4_secinfo(dir, name, err);
7170                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7171                                 &exception);
7172         } while (exception.retry);
7173         return err;
7174 }
7175
7176 #ifdef CONFIG_NFS_V4_1
7177 /*
7178  * Check the exchange flags returned by the server for invalid flags, having
7179  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7180  * DS flags set.
7181  */
7182 static int nfs4_check_cl_exchange_flags(u32 flags)
7183 {
7184         if (flags & ~EXCHGID4_FLAG_MASK_R)
7185                 goto out_inval;
7186         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7187             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7188                 goto out_inval;
7189         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7190                 goto out_inval;
7191         return NFS_OK;
7192 out_inval:
7193         return -NFS4ERR_INVAL;
7194 }
7195
7196 static bool
7197 nfs41_same_server_scope(struct nfs41_server_scope *a,
7198                         struct nfs41_server_scope *b)
7199 {
7200         if (a->server_scope_sz == b->server_scope_sz &&
7201             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
7202                 return true;
7203
7204         return false;
7205 }
7206
7207 static void
7208 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7209 {
7210 }
7211
7212 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7213         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7214 };
7215
7216 /*
7217  * nfs4_proc_bind_one_conn_to_session()
7218  *
7219  * The 4.1 client currently uses the same TCP connection for the
7220  * fore and backchannel.
7221  */
7222 static
7223 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7224                 struct rpc_xprt *xprt,
7225                 struct nfs_client *clp,
7226                 struct rpc_cred *cred)
7227 {
7228         int status;
7229         struct nfs41_bind_conn_to_session_args args = {
7230                 .client = clp,
7231                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7232         };
7233         struct nfs41_bind_conn_to_session_res res;
7234         struct rpc_message msg = {
7235                 .rpc_proc =
7236                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7237                 .rpc_argp = &args,
7238                 .rpc_resp = &res,
7239                 .rpc_cred = cred,
7240         };
7241         struct rpc_task_setup task_setup_data = {
7242                 .rpc_client = clnt,
7243                 .rpc_xprt = xprt,
7244                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7245                 .rpc_message = &msg,
7246                 .flags = RPC_TASK_TIMEOUT,
7247         };
7248         struct rpc_task *task;
7249
7250         dprintk("--> %s\n", __func__);
7251
7252         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7253         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7254                 args.dir = NFS4_CDFC4_FORE;
7255
7256         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7257         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7258                 args.dir = NFS4_CDFC4_FORE;
7259
7260         task = rpc_run_task(&task_setup_data);
7261         if (!IS_ERR(task)) {
7262                 status = task->tk_status;
7263                 rpc_put_task(task);
7264         } else
7265                 status = PTR_ERR(task);
7266         trace_nfs4_bind_conn_to_session(clp, status);
7267         if (status == 0) {
7268                 if (memcmp(res.sessionid.data,
7269                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7270                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7271                         status = -EIO;
7272                         goto out;
7273                 }
7274                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7275                         dprintk("NFS: %s: Unexpected direction from server\n",
7276                                 __func__);
7277                         status = -EIO;
7278                         goto out;
7279                 }
7280                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7281                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7282                                 __func__);
7283                         status = -EIO;
7284                         goto out;
7285                 }
7286         }
7287 out:
7288         dprintk("<-- %s status= %d\n", __func__, status);
7289         return status;
7290 }
7291
7292 struct rpc_bind_conn_calldata {
7293         struct nfs_client *clp;
7294         struct rpc_cred *cred;
7295 };
7296
7297 static int
7298 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7299                 struct rpc_xprt *xprt,
7300                 void *calldata)
7301 {
7302         struct rpc_bind_conn_calldata *p = calldata;
7303
7304         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7305 }
7306
7307 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7308 {
7309         struct rpc_bind_conn_calldata data = {
7310                 .clp = clp,
7311                 .cred = cred,
7312         };
7313         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7314                         nfs4_proc_bind_conn_to_session_callback, &data);
7315 }
7316
7317 /*
7318  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7319  * and operations we'd like to see to enable certain features in the allow map
7320  */
7321 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7322         .how = SP4_MACH_CRED,
7323         .enforce.u.words = {
7324                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7325                       1 << (OP_EXCHANGE_ID - 32) |
7326                       1 << (OP_CREATE_SESSION - 32) |
7327                       1 << (OP_DESTROY_SESSION - 32) |
7328                       1 << (OP_DESTROY_CLIENTID - 32)
7329         },
7330         .allow.u.words = {
7331                 [0] = 1 << (OP_CLOSE) |
7332                       1 << (OP_OPEN_DOWNGRADE) |
7333                       1 << (OP_LOCKU) |
7334                       1 << (OP_DELEGRETURN) |
7335                       1 << (OP_COMMIT),
7336                 [1] = 1 << (OP_SECINFO - 32) |
7337                       1 << (OP_SECINFO_NO_NAME - 32) |
7338                       1 << (OP_LAYOUTRETURN - 32) |
7339                       1 << (OP_TEST_STATEID - 32) |
7340                       1 << (OP_FREE_STATEID - 32) |
7341                       1 << (OP_WRITE - 32)
7342         }
7343 };
7344
7345 /*
7346  * Select the state protection mode for client `clp' given the server results
7347  * from exchange_id in `sp'.
7348  *
7349  * Returns 0 on success, negative errno otherwise.
7350  */
7351 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7352                                  struct nfs41_state_protection *sp)
7353 {
7354         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7355                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7356                       1 << (OP_EXCHANGE_ID - 32) |
7357                       1 << (OP_CREATE_SESSION - 32) |
7358                       1 << (OP_DESTROY_SESSION - 32) |
7359                       1 << (OP_DESTROY_CLIENTID - 32)
7360         };
7361         unsigned int i;
7362
7363         if (sp->how == SP4_MACH_CRED) {
7364                 /* Print state protect result */
7365                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7366                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7367                         if (test_bit(i, sp->enforce.u.longs))
7368                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7369                         if (test_bit(i, sp->allow.u.longs))
7370                                 dfprintk(MOUNT, "  allow op %d\n", i);
7371                 }
7372
7373                 /* make sure nothing is on enforce list that isn't supported */
7374                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7375                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7376                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7377                                 return -EINVAL;
7378                         }
7379                 }
7380
7381                 /*
7382                  * Minimal mode - state operations are allowed to use machine
7383                  * credential.  Note this already happens by default, so the
7384                  * client doesn't have to do anything more than the negotiation.
7385                  *
7386                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7387                  *       we're already using the machine cred for exchange_id
7388                  *       and will never use a different cred.
7389                  */
7390                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7391                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7392                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7393                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7394                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7395                         dfprintk(MOUNT, "  minimal mode enabled\n");
7396                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7397                 } else {
7398                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7399                         return -EINVAL;
7400                 }
7401
7402                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7403                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7404                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7405                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7406                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7407                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7408                 }
7409
7410                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7411                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7412                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7413                                 &clp->cl_sp4_flags);
7414                 }
7415
7416                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7417                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7418                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7419                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7420                 }
7421
7422                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7423                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7424                         dfprintk(MOUNT, "  stateid mode enabled\n");
7425                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7426                 }
7427
7428                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7429                         dfprintk(MOUNT, "  write mode enabled\n");
7430                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7431                 }
7432
7433                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7434                         dfprintk(MOUNT, "  commit mode enabled\n");
7435                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7436                 }
7437         }
7438
7439         return 0;
7440 }
7441
7442 struct nfs41_exchange_id_data {
7443         struct nfs41_exchange_id_res res;
7444         struct nfs41_exchange_id_args args;
7445         struct rpc_xprt *xprt;
7446         int rpc_status;
7447 };
7448
7449 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7450 {
7451         struct nfs41_exchange_id_data *cdata =
7452                                         (struct nfs41_exchange_id_data *)data;
7453         struct nfs_client *clp = cdata->args.client;
7454         int status = task->tk_status;
7455
7456         trace_nfs4_exchange_id(clp, status);
7457
7458         if (status == 0)
7459                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7460
7461         if (cdata->xprt && status == 0) {
7462                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7463                                                       cdata->xprt);
7464                 goto out;
7465         }
7466
7467         if (status  == 0)
7468                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7469
7470         if (status == 0) {
7471                 clp->cl_clientid = cdata->res.clientid;
7472                 clp->cl_exchange_flags = cdata->res.flags;
7473                 /* Client ID is not confirmed */
7474                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7475                         clear_bit(NFS4_SESSION_ESTABLISHED,
7476                         &clp->cl_session->session_state);
7477                         clp->cl_seqid = cdata->res.seqid;
7478                 }
7479
7480                 kfree(clp->cl_serverowner);
7481                 clp->cl_serverowner = cdata->res.server_owner;
7482                 cdata->res.server_owner = NULL;
7483
7484                 /* use the most recent implementation id */
7485                 kfree(clp->cl_implid);
7486                 clp->cl_implid = cdata->res.impl_id;
7487                 cdata->res.impl_id = NULL;
7488
7489                 if (clp->cl_serverscope != NULL &&
7490                     !nfs41_same_server_scope(clp->cl_serverscope,
7491                                         cdata->res.server_scope)) {
7492                         dprintk("%s: server_scope mismatch detected\n",
7493                                 __func__);
7494                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7495                         kfree(clp->cl_serverscope);
7496                         clp->cl_serverscope = NULL;
7497                 }
7498
7499                 if (clp->cl_serverscope == NULL) {
7500                         clp->cl_serverscope = cdata->res.server_scope;
7501                         cdata->res.server_scope = NULL;
7502                 }
7503                 /* Save the EXCHANGE_ID verifier session trunk tests */
7504                 memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7505                        sizeof(clp->cl_confirm.data));
7506         }
7507 out:
7508         cdata->rpc_status = status;
7509         return;
7510 }
7511
7512 static void nfs4_exchange_id_release(void *data)
7513 {
7514         struct nfs41_exchange_id_data *cdata =
7515                                         (struct nfs41_exchange_id_data *)data;
7516
7517         nfs_put_client(cdata->args.client);
7518         if (cdata->xprt) {
7519                 xprt_put(cdata->xprt);
7520                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7521         }
7522         kfree(cdata->res.impl_id);
7523         kfree(cdata->res.server_scope);
7524         kfree(cdata->res.server_owner);
7525         kfree(cdata);
7526 }
7527
7528 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7529         .rpc_call_done = nfs4_exchange_id_done,
7530         .rpc_release = nfs4_exchange_id_release,
7531 };
7532
7533 /*
7534  * _nfs4_proc_exchange_id()
7535  *
7536  * Wrapper for EXCHANGE_ID operation.
7537  */
7538 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7539                         u32 sp4_how, struct rpc_xprt *xprt)
7540 {
7541         nfs4_verifier verifier;
7542         struct rpc_message msg = {
7543                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7544                 .rpc_cred = cred,
7545         };
7546         struct rpc_task_setup task_setup_data = {
7547                 .rpc_client = clp->cl_rpcclient,
7548                 .callback_ops = &nfs4_exchange_id_call_ops,
7549                 .rpc_message = &msg,
7550                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7551         };
7552         struct nfs41_exchange_id_data *calldata;
7553         struct rpc_task *task;
7554         int status = -EIO;
7555
7556         if (!atomic_inc_not_zero(&clp->cl_count))
7557                 goto out;
7558
7559         status = -ENOMEM;
7560         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7561         if (!calldata)
7562                 goto out;
7563
7564         if (!xprt)
7565                 nfs4_init_boot_verifier(clp, &verifier);
7566
7567         status = nfs4_init_uniform_client_string(clp);
7568         if (status)
7569                 goto out_calldata;
7570
7571         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7572                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7573                 clp->cl_owner_id);
7574
7575         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7576                                                 GFP_NOFS);
7577         status = -ENOMEM;
7578         if (unlikely(calldata->res.server_owner == NULL))
7579                 goto out_calldata;
7580
7581         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7582                                         GFP_NOFS);
7583         if (unlikely(calldata->res.server_scope == NULL))
7584                 goto out_server_owner;
7585
7586         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7587         if (unlikely(calldata->res.impl_id == NULL))
7588                 goto out_server_scope;
7589
7590         switch (sp4_how) {
7591         case SP4_NONE:
7592                 calldata->args.state_protect.how = SP4_NONE;
7593                 break;
7594
7595         case SP4_MACH_CRED:
7596                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7597                 break;
7598
7599         default:
7600                 /* unsupported! */
7601                 WARN_ON_ONCE(1);
7602                 status = -EINVAL;
7603                 goto out_impl_id;
7604         }
7605         if (xprt) {
7606                 calldata->xprt = xprt;
7607                 task_setup_data.rpc_xprt = xprt;
7608                 task_setup_data.flags =
7609                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7610                 calldata->args.verifier = &clp->cl_confirm;
7611         } else {
7612                 calldata->args.verifier = &verifier;
7613         }
7614         calldata->args.client = clp;
7615 #ifdef CONFIG_NFS_V4_1_MIGRATION
7616         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7617         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7618         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7619 #else
7620         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7621         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7622 #endif
7623         msg.rpc_argp = &calldata->args;
7624         msg.rpc_resp = &calldata->res;
7625         task_setup_data.callback_data = calldata;
7626
7627         task = rpc_run_task(&task_setup_data);
7628         if (IS_ERR(task)) {
7629         status = PTR_ERR(task);
7630                 goto out_impl_id;
7631         }
7632
7633         if (!xprt) {
7634                 status = rpc_wait_for_completion_task(task);
7635                 if (!status)
7636                         status = calldata->rpc_status;
7637         } else  /* session trunking test */
7638                 status = calldata->rpc_status;
7639
7640         rpc_put_task(task);
7641 out:
7642         if (clp->cl_implid != NULL)
7643                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7644                         "domain: %s, name: %s, date: %llu,%u\n",
7645                         clp->cl_implid->domain, clp->cl_implid->name,
7646                         clp->cl_implid->date.seconds,
7647                         clp->cl_implid->date.nseconds);
7648         dprintk("NFS reply exchange_id: %d\n", status);
7649         return status;
7650
7651 out_impl_id:
7652         kfree(calldata->res.impl_id);
7653 out_server_scope:
7654         kfree(calldata->res.server_scope);
7655 out_server_owner:
7656         kfree(calldata->res.server_owner);
7657 out_calldata:
7658         kfree(calldata);
7659         goto out;
7660 }
7661
7662 /*
7663  * nfs4_proc_exchange_id()
7664  *
7665  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7666  *
7667  * Since the clientid has expired, all compounds using sessions
7668  * associated with the stale clientid will be returning
7669  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7670  * be in some phase of session reset.
7671  *
7672  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7673  */
7674 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7675 {
7676         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7677         int status;
7678
7679         /* try SP4_MACH_CRED if krb5i/p */
7680         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7681             authflavor == RPC_AUTH_GSS_KRB5P) {
7682                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7683                 if (!status)
7684                         return 0;
7685         }
7686
7687         /* try SP4_NONE */
7688         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7689 }
7690
7691 /**
7692  * nfs4_test_session_trunk
7693  *
7694  * This is an add_xprt_test() test function called from
7695  * rpc_clnt_setup_test_and_add_xprt.
7696  *
7697  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7698  * and is dereferrenced in nfs4_exchange_id_release
7699  *
7700  * Upon success, add the new transport to the rpc_clnt
7701  *
7702  * @clnt: struct rpc_clnt to get new transport
7703  * @xprt: the rpc_xprt to test
7704  * @data: call data for _nfs4_proc_exchange_id.
7705  */
7706 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7707                             void *data)
7708 {
7709         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7710         u32 sp4_how;
7711
7712         dprintk("--> %s try %s\n", __func__,
7713                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7714
7715         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7716
7717         /* Test connection for session trunking. Async exchange_id call */
7718         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7719 }
7720 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7721
7722 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7723                 struct rpc_cred *cred)
7724 {
7725         struct rpc_message msg = {
7726                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7727                 .rpc_argp = clp,
7728                 .rpc_cred = cred,
7729         };
7730         int status;
7731
7732         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7733         trace_nfs4_destroy_clientid(clp, status);
7734         if (status)
7735                 dprintk("NFS: Got error %d from the server %s on "
7736                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7737         return status;
7738 }
7739
7740 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7741                 struct rpc_cred *cred)
7742 {
7743         unsigned int loop;
7744         int ret;
7745
7746         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7747                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7748                 switch (ret) {
7749                 case -NFS4ERR_DELAY:
7750                 case -NFS4ERR_CLIENTID_BUSY:
7751                         ssleep(1);
7752                         break;
7753                 default:
7754                         return ret;
7755                 }
7756         }
7757         return 0;
7758 }
7759
7760 int nfs4_destroy_clientid(struct nfs_client *clp)
7761 {
7762         struct rpc_cred *cred;
7763         int ret = 0;
7764
7765         if (clp->cl_mvops->minor_version < 1)
7766                 goto out;
7767         if (clp->cl_exchange_flags == 0)
7768                 goto out;
7769         if (clp->cl_preserve_clid)
7770                 goto out;
7771         cred = nfs4_get_clid_cred(clp);
7772         ret = nfs4_proc_destroy_clientid(clp, cred);
7773         if (cred)
7774                 put_rpccred(cred);
7775         switch (ret) {
7776         case 0:
7777         case -NFS4ERR_STALE_CLIENTID:
7778                 clp->cl_exchange_flags = 0;
7779         }
7780 out:
7781         return ret;
7782 }
7783
7784 struct nfs4_get_lease_time_data {
7785         struct nfs4_get_lease_time_args *args;
7786         struct nfs4_get_lease_time_res *res;
7787         struct nfs_client *clp;
7788 };
7789
7790 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7791                                         void *calldata)
7792 {
7793         struct nfs4_get_lease_time_data *data =
7794                         (struct nfs4_get_lease_time_data *)calldata;
7795
7796         dprintk("--> %s\n", __func__);
7797         /* just setup sequence, do not trigger session recovery
7798            since we're invoked within one */
7799         nfs41_setup_sequence(data->clp->cl_session,
7800                         &data->args->la_seq_args,
7801                         &data->res->lr_seq_res,
7802                         task);
7803         dprintk("<-- %s\n", __func__);
7804 }
7805
7806 /*
7807  * Called from nfs4_state_manager thread for session setup, so don't recover
7808  * from sequence operation or clientid errors.
7809  */
7810 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7811 {
7812         struct nfs4_get_lease_time_data *data =
7813                         (struct nfs4_get_lease_time_data *)calldata;
7814
7815         dprintk("--> %s\n", __func__);
7816         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7817                 return;
7818         switch (task->tk_status) {
7819         case -NFS4ERR_DELAY:
7820         case -NFS4ERR_GRACE:
7821                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7822                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7823                 task->tk_status = 0;
7824                 /* fall through */
7825         case -NFS4ERR_RETRY_UNCACHED_REP:
7826                 rpc_restart_call_prepare(task);
7827                 return;
7828         }
7829         dprintk("<-- %s\n", __func__);
7830 }
7831
7832 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7833         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7834         .rpc_call_done = nfs4_get_lease_time_done,
7835 };
7836
7837 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7838 {
7839         struct rpc_task *task;
7840         struct nfs4_get_lease_time_args args;
7841         struct nfs4_get_lease_time_res res = {
7842                 .lr_fsinfo = fsinfo,
7843         };
7844         struct nfs4_get_lease_time_data data = {
7845                 .args = &args,
7846                 .res = &res,
7847                 .clp = clp,
7848         };
7849         struct rpc_message msg = {
7850                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7851                 .rpc_argp = &args,
7852                 .rpc_resp = &res,
7853         };
7854         struct rpc_task_setup task_setup = {
7855                 .rpc_client = clp->cl_rpcclient,
7856                 .rpc_message = &msg,
7857                 .callback_ops = &nfs4_get_lease_time_ops,
7858                 .callback_data = &data,
7859                 .flags = RPC_TASK_TIMEOUT,
7860         };
7861         int status;
7862
7863         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7864         nfs4_set_sequence_privileged(&args.la_seq_args);
7865         dprintk("--> %s\n", __func__);
7866         task = rpc_run_task(&task_setup);
7867
7868         if (IS_ERR(task))
7869                 status = PTR_ERR(task);
7870         else {
7871                 status = task->tk_status;
7872                 rpc_put_task(task);
7873         }
7874         dprintk("<-- %s return %d\n", __func__, status);
7875
7876         return status;
7877 }
7878
7879 /*
7880  * Initialize the values to be used by the client in CREATE_SESSION
7881  * If nfs4_init_session set the fore channel request and response sizes,
7882  * use them.
7883  *
7884  * Set the back channel max_resp_sz_cached to zero to force the client to
7885  * always set csa_cachethis to FALSE because the current implementation
7886  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7887  */
7888 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7889                                     struct rpc_clnt *clnt)
7890 {
7891         unsigned int max_rqst_sz, max_resp_sz;
7892         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7893
7894         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7895         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7896
7897         /* Fore channel attributes */
7898         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7899         args->fc_attrs.max_resp_sz = max_resp_sz;
7900         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7901         args->fc_attrs.max_reqs = max_session_slots;
7902
7903         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7904                 "max_ops=%u max_reqs=%u\n",
7905                 __func__,
7906                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7907                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7908
7909         /* Back channel attributes */
7910         args->bc_attrs.max_rqst_sz = max_bc_payload;
7911         args->bc_attrs.max_resp_sz = max_bc_payload;
7912         args->bc_attrs.max_resp_sz_cached = 0;
7913         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7914         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7915
7916         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7917                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7918                 __func__,
7919                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7920                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7921                 args->bc_attrs.max_reqs);
7922 }
7923
7924 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7925                 struct nfs41_create_session_res *res)
7926 {
7927         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7928         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7929
7930         if (rcvd->max_resp_sz > sent->max_resp_sz)
7931                 return -EINVAL;
7932         /*
7933          * Our requested max_ops is the minimum we need; we're not
7934          * prepared to break up compounds into smaller pieces than that.
7935          * So, no point even trying to continue if the server won't
7936          * cooperate:
7937          */
7938         if (rcvd->max_ops < sent->max_ops)
7939                 return -EINVAL;
7940         if (rcvd->max_reqs == 0)
7941                 return -EINVAL;
7942         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7943                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7944         return 0;
7945 }
7946
7947 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7948                 struct nfs41_create_session_res *res)
7949 {
7950         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7951         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7952
7953         if (!(res->flags & SESSION4_BACK_CHAN))
7954                 goto out;
7955         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7956                 return -EINVAL;
7957         if (rcvd->max_resp_sz < sent->max_resp_sz)
7958                 return -EINVAL;
7959         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7960                 return -EINVAL;
7961         if (rcvd->max_ops > sent->max_ops)
7962                 return -EINVAL;
7963         if (rcvd->max_reqs > sent->max_reqs)
7964                 return -EINVAL;
7965 out:
7966         return 0;
7967 }
7968
7969 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7970                                      struct nfs41_create_session_res *res)
7971 {
7972         int ret;
7973
7974         ret = nfs4_verify_fore_channel_attrs(args, res);
7975         if (ret)
7976                 return ret;
7977         return nfs4_verify_back_channel_attrs(args, res);
7978 }
7979
7980 static void nfs4_update_session(struct nfs4_session *session,
7981                 struct nfs41_create_session_res *res)
7982 {
7983         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7984         /* Mark client id and session as being confirmed */
7985         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7986         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7987         session->flags = res->flags;
7988         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7989         if (res->flags & SESSION4_BACK_CHAN)
7990                 memcpy(&session->bc_attrs, &res->bc_attrs,
7991                                 sizeof(session->bc_attrs));
7992 }
7993
7994 static int _nfs4_proc_create_session(struct nfs_client *clp,
7995                 struct rpc_cred *cred)
7996 {
7997         struct nfs4_session *session = clp->cl_session;
7998         struct nfs41_create_session_args args = {
7999                 .client = clp,
8000                 .clientid = clp->cl_clientid,
8001                 .seqid = clp->cl_seqid,
8002                 .cb_program = NFS4_CALLBACK,
8003         };
8004         struct nfs41_create_session_res res;
8005
8006         struct rpc_message msg = {
8007                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8008                 .rpc_argp = &args,
8009                 .rpc_resp = &res,
8010                 .rpc_cred = cred,
8011         };
8012         int status;
8013
8014         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8015         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8016
8017         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8018         trace_nfs4_create_session(clp, status);
8019
8020         switch (status) {
8021         case -NFS4ERR_STALE_CLIENTID:
8022         case -NFS4ERR_DELAY:
8023         case -ETIMEDOUT:
8024         case -EACCES:
8025         case -EAGAIN:
8026                 goto out;
8027         };
8028
8029         clp->cl_seqid++;
8030         if (!status) {
8031                 /* Verify the session's negotiated channel_attrs values */
8032                 status = nfs4_verify_channel_attrs(&args, &res);
8033                 /* Increment the clientid slot sequence id */
8034                 if (status)
8035                         goto out;
8036                 nfs4_update_session(session, &res);
8037         }
8038 out:
8039         return status;
8040 }
8041
8042 /*
8043  * Issues a CREATE_SESSION operation to the server.
8044  * It is the responsibility of the caller to verify the session is
8045  * expired before calling this routine.
8046  */
8047 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8048 {
8049         int status;
8050         unsigned *ptr;
8051         struct nfs4_session *session = clp->cl_session;
8052
8053         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8054
8055         status = _nfs4_proc_create_session(clp, cred);
8056         if (status)
8057                 goto out;
8058
8059         /* Init or reset the session slot tables */
8060         status = nfs4_setup_session_slot_tables(session);
8061         dprintk("slot table setup returned %d\n", status);
8062         if (status)
8063                 goto out;
8064
8065         ptr = (unsigned *)&session->sess_id.data[0];
8066         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8067                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8068 out:
8069         dprintk("<-- %s\n", __func__);
8070         return status;
8071 }
8072
8073 /*
8074  * Issue the over-the-wire RPC DESTROY_SESSION.
8075  * The caller must serialize access to this routine.
8076  */
8077 int nfs4_proc_destroy_session(struct nfs4_session *session,
8078                 struct rpc_cred *cred)
8079 {
8080         struct rpc_message msg = {
8081                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8082                 .rpc_argp = session,
8083                 .rpc_cred = cred,
8084         };
8085         int status = 0;
8086
8087         dprintk("--> nfs4_proc_destroy_session\n");
8088
8089         /* session is still being setup */
8090         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8091                 return 0;
8092
8093         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8094         trace_nfs4_destroy_session(session->clp, status);
8095
8096         if (status)
8097                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8098                         "Session has been destroyed regardless...\n", status);
8099
8100         dprintk("<-- nfs4_proc_destroy_session\n");
8101         return status;
8102 }
8103
8104 /*
8105  * Renew the cl_session lease.
8106  */
8107 struct nfs4_sequence_data {
8108         struct nfs_client *clp;
8109         struct nfs4_sequence_args args;
8110         struct nfs4_sequence_res res;
8111 };
8112
8113 static void nfs41_sequence_release(void *data)
8114 {
8115         struct nfs4_sequence_data *calldata = data;
8116         struct nfs_client *clp = calldata->clp;
8117
8118         if (atomic_read(&clp->cl_count) > 1)
8119                 nfs4_schedule_state_renewal(clp);
8120         nfs_put_client(clp);
8121         kfree(calldata);
8122 }
8123
8124 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8125 {
8126         switch(task->tk_status) {
8127         case -NFS4ERR_DELAY:
8128                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8129                 return -EAGAIN;
8130         default:
8131                 nfs4_schedule_lease_recovery(clp);
8132         }
8133         return 0;
8134 }
8135
8136 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8137 {
8138         struct nfs4_sequence_data *calldata = data;
8139         struct nfs_client *clp = calldata->clp;
8140
8141         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8142                 return;
8143
8144         trace_nfs4_sequence(clp, task->tk_status);
8145         if (task->tk_status < 0) {
8146                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8147                 if (atomic_read(&clp->cl_count) == 1)
8148                         goto out;
8149
8150                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8151                         rpc_restart_call_prepare(task);
8152                         return;
8153                 }
8154         }
8155         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8156 out:
8157         dprintk("<-- %s\n", __func__);
8158 }
8159
8160 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8161 {
8162         struct nfs4_sequence_data *calldata = data;
8163         struct nfs_client *clp = calldata->clp;
8164         struct nfs4_sequence_args *args;
8165         struct nfs4_sequence_res *res;
8166
8167         args = task->tk_msg.rpc_argp;
8168         res = task->tk_msg.rpc_resp;
8169
8170         nfs41_setup_sequence(clp->cl_session, args, res, task);
8171 }
8172
8173 static const struct rpc_call_ops nfs41_sequence_ops = {
8174         .rpc_call_done = nfs41_sequence_call_done,
8175         .rpc_call_prepare = nfs41_sequence_prepare,
8176         .rpc_release = nfs41_sequence_release,
8177 };
8178
8179 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8180                 struct rpc_cred *cred,
8181                 bool is_privileged)
8182 {
8183         struct nfs4_sequence_data *calldata;
8184         struct rpc_message msg = {
8185                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8186                 .rpc_cred = cred,
8187         };
8188         struct rpc_task_setup task_setup_data = {
8189                 .rpc_client = clp->cl_rpcclient,
8190                 .rpc_message = &msg,
8191                 .callback_ops = &nfs41_sequence_ops,
8192                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8193         };
8194
8195         if (!atomic_inc_not_zero(&clp->cl_count))
8196                 return ERR_PTR(-EIO);
8197         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8198         if (calldata == NULL) {
8199                 nfs_put_client(clp);
8200                 return ERR_PTR(-ENOMEM);
8201         }
8202         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8203         if (is_privileged)
8204                 nfs4_set_sequence_privileged(&calldata->args);
8205         msg.rpc_argp = &calldata->args;
8206         msg.rpc_resp = &calldata->res;
8207         calldata->clp = clp;
8208         task_setup_data.callback_data = calldata;
8209
8210         return rpc_run_task(&task_setup_data);
8211 }
8212
8213 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8214 {
8215         struct rpc_task *task;
8216         int ret = 0;
8217
8218         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8219                 return -EAGAIN;
8220         task = _nfs41_proc_sequence(clp, cred, false);
8221         if (IS_ERR(task))
8222                 ret = PTR_ERR(task);
8223         else
8224                 rpc_put_task_async(task);
8225         dprintk("<-- %s status=%d\n", __func__, ret);
8226         return ret;
8227 }
8228
8229 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8230 {
8231         struct rpc_task *task;
8232         int ret;
8233
8234         task = _nfs41_proc_sequence(clp, cred, true);
8235         if (IS_ERR(task)) {
8236                 ret = PTR_ERR(task);
8237                 goto out;
8238         }
8239         ret = rpc_wait_for_completion_task(task);
8240         if (!ret)
8241                 ret = task->tk_status;
8242         rpc_put_task(task);
8243 out:
8244         dprintk("<-- %s status=%d\n", __func__, ret);
8245         return ret;
8246 }
8247
8248 struct nfs4_reclaim_complete_data {
8249         struct nfs_client *clp;
8250         struct nfs41_reclaim_complete_args arg;
8251         struct nfs41_reclaim_complete_res res;
8252 };
8253
8254 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8255 {
8256         struct nfs4_reclaim_complete_data *calldata = data;
8257
8258         nfs41_setup_sequence(calldata->clp->cl_session,
8259                         &calldata->arg.seq_args,
8260                         &calldata->res.seq_res,
8261                         task);
8262 }
8263
8264 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8265 {
8266         switch(task->tk_status) {
8267         case 0:
8268         case -NFS4ERR_COMPLETE_ALREADY:
8269         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8270                 break;
8271         case -NFS4ERR_DELAY:
8272                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8273                 /* fall through */
8274         case -NFS4ERR_RETRY_UNCACHED_REP:
8275                 return -EAGAIN;
8276         default:
8277                 nfs4_schedule_lease_recovery(clp);
8278         }
8279         return 0;
8280 }
8281
8282 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8283 {
8284         struct nfs4_reclaim_complete_data *calldata = data;
8285         struct nfs_client *clp = calldata->clp;
8286         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8287
8288         dprintk("--> %s\n", __func__);
8289         if (!nfs41_sequence_done(task, res))
8290                 return;
8291
8292         trace_nfs4_reclaim_complete(clp, task->tk_status);
8293         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8294                 rpc_restart_call_prepare(task);
8295                 return;
8296         }
8297         dprintk("<-- %s\n", __func__);
8298 }
8299
8300 static void nfs4_free_reclaim_complete_data(void *data)
8301 {
8302         struct nfs4_reclaim_complete_data *calldata = data;
8303
8304         kfree(calldata);
8305 }
8306
8307 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8308         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8309         .rpc_call_done = nfs4_reclaim_complete_done,
8310         .rpc_release = nfs4_free_reclaim_complete_data,
8311 };
8312
8313 /*
8314  * Issue a global reclaim complete.
8315  */
8316 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8317                 struct rpc_cred *cred)
8318 {
8319         struct nfs4_reclaim_complete_data *calldata;
8320         struct rpc_task *task;
8321         struct rpc_message msg = {
8322                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8323                 .rpc_cred = cred,
8324         };
8325         struct rpc_task_setup task_setup_data = {
8326                 .rpc_client = clp->cl_rpcclient,
8327                 .rpc_message = &msg,
8328                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8329                 .flags = RPC_TASK_ASYNC,
8330         };
8331         int status = -ENOMEM;
8332
8333         dprintk("--> %s\n", __func__);
8334         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8335         if (calldata == NULL)
8336                 goto out;
8337         calldata->clp = clp;
8338         calldata->arg.one_fs = 0;
8339
8340         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8341         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8342         msg.rpc_argp = &calldata->arg;
8343         msg.rpc_resp = &calldata->res;
8344         task_setup_data.callback_data = calldata;
8345         task = rpc_run_task(&task_setup_data);
8346         if (IS_ERR(task)) {
8347                 status = PTR_ERR(task);
8348                 goto out;
8349         }
8350         status = nfs4_wait_for_completion_rpc_task(task);
8351         if (status == 0)
8352                 status = task->tk_status;
8353         rpc_put_task(task);
8354         return 0;
8355 out:
8356         dprintk("<-- %s status=%d\n", __func__, status);
8357         return status;
8358 }
8359
8360 static void
8361 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8362 {
8363         struct nfs4_layoutget *lgp = calldata;
8364         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8365         struct nfs4_session *session = nfs4_get_session(server);
8366
8367         dprintk("--> %s\n", __func__);
8368         nfs41_setup_sequence(session, &lgp->args.seq_args,
8369                                 &lgp->res.seq_res, task);
8370         dprintk("<-- %s\n", __func__);
8371 }
8372
8373 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8374 {
8375         struct nfs4_layoutget *lgp = calldata;
8376
8377         dprintk("--> %s\n", __func__);
8378         nfs41_sequence_process(task, &lgp->res.seq_res);
8379         dprintk("<-- %s\n", __func__);
8380 }
8381
8382 static int
8383 nfs4_layoutget_handle_exception(struct rpc_task *task,
8384                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8385 {
8386         struct inode *inode = lgp->args.inode;
8387         struct nfs_server *server = NFS_SERVER(inode);
8388         struct pnfs_layout_hdr *lo;
8389         int nfs4err = task->tk_status;
8390         int err, status = 0;
8391         LIST_HEAD(head);
8392
8393         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8394
8395         switch (nfs4err) {
8396         case 0:
8397                 goto out;
8398
8399         /*
8400          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8401          * on the file. set tk_status to -ENODATA to tell upper layer to
8402          * retry go inband.
8403          */
8404         case -NFS4ERR_LAYOUTUNAVAILABLE:
8405                 status = -ENODATA;
8406                 goto out;
8407         /*
8408          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8409          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8410          */
8411         case -NFS4ERR_BADLAYOUT:
8412                 status = -EOVERFLOW;
8413                 goto out;
8414         /*
8415          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8416          * (or clients) writing to the same RAID stripe except when
8417          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8418          *
8419          * Treat it like we would RECALLCONFLICT -- we retry for a little
8420          * while, and then eventually give up.
8421          */
8422         case -NFS4ERR_LAYOUTTRYLATER:
8423                 if (lgp->args.minlength == 0) {
8424                         status = -EOVERFLOW;
8425                         goto out;
8426                 }
8427                 status = -EBUSY;
8428                 break;
8429         case -NFS4ERR_RECALLCONFLICT:
8430                 status = -ERECALLCONFLICT;
8431                 break;
8432         case -NFS4ERR_DELEG_REVOKED:
8433         case -NFS4ERR_ADMIN_REVOKED:
8434         case -NFS4ERR_EXPIRED:
8435         case -NFS4ERR_BAD_STATEID:
8436                 exception->timeout = 0;
8437                 spin_lock(&inode->i_lock);
8438                 lo = NFS_I(inode)->layout;
8439                 /* If the open stateid was bad, then recover it. */
8440                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8441                     nfs4_stateid_match_other(&lgp->args.stateid,
8442                                         &lgp->args.ctx->state->stateid)) {
8443                         spin_unlock(&inode->i_lock);
8444                         exception->state = lgp->args.ctx->state;
8445                         exception->stateid = &lgp->args.stateid;
8446                         break;
8447                 }
8448
8449                 /*
8450                  * Mark the bad layout state as invalid, then retry
8451                  */
8452                 pnfs_mark_layout_stateid_invalid(lo, &head);
8453                 spin_unlock(&inode->i_lock);
8454                 pnfs_free_lseg_list(&head);
8455                 status = -EAGAIN;
8456                 goto out;
8457         }
8458
8459         err = nfs4_handle_exception(server, nfs4err, exception);
8460         if (!status) {
8461                 if (exception->retry)
8462                         status = -EAGAIN;
8463                 else
8464                         status = err;
8465         }
8466 out:
8467         dprintk("<-- %s\n", __func__);
8468         return status;
8469 }
8470
8471 static size_t max_response_pages(struct nfs_server *server)
8472 {
8473         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8474         return nfs_page_array_len(0, max_resp_sz);
8475 }
8476
8477 static void nfs4_free_pages(struct page **pages, size_t size)
8478 {
8479         int i;
8480
8481         if (!pages)
8482                 return;
8483
8484         for (i = 0; i < size; i++) {
8485                 if (!pages[i])
8486                         break;
8487                 __free_page(pages[i]);
8488         }
8489         kfree(pages);
8490 }
8491
8492 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8493 {
8494         struct page **pages;
8495         int i;
8496
8497         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8498         if (!pages) {
8499                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8500                 return NULL;
8501         }
8502
8503         for (i = 0; i < size; i++) {
8504                 pages[i] = alloc_page(gfp_flags);
8505                 if (!pages[i]) {
8506                         dprintk("%s: failed to allocate page\n", __func__);
8507                         nfs4_free_pages(pages, size);
8508                         return NULL;
8509                 }
8510         }
8511
8512         return pages;
8513 }
8514
8515 static void nfs4_layoutget_release(void *calldata)
8516 {
8517         struct nfs4_layoutget *lgp = calldata;
8518         struct inode *inode = lgp->args.inode;
8519         struct nfs_server *server = NFS_SERVER(inode);
8520         size_t max_pages = max_response_pages(server);
8521
8522         dprintk("--> %s\n", __func__);
8523         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8524         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8525         put_nfs_open_context(lgp->args.ctx);
8526         kfree(calldata);
8527         dprintk("<-- %s\n", __func__);
8528 }
8529
8530 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8531         .rpc_call_prepare = nfs4_layoutget_prepare,
8532         .rpc_call_done = nfs4_layoutget_done,
8533         .rpc_release = nfs4_layoutget_release,
8534 };
8535
8536 struct pnfs_layout_segment *
8537 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8538 {
8539         struct inode *inode = lgp->args.inode;
8540         struct nfs_server *server = NFS_SERVER(inode);
8541         size_t max_pages = max_response_pages(server);
8542         struct rpc_task *task;
8543         struct rpc_message msg = {
8544                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8545                 .rpc_argp = &lgp->args,
8546                 .rpc_resp = &lgp->res,
8547                 .rpc_cred = lgp->cred,
8548         };
8549         struct rpc_task_setup task_setup_data = {
8550                 .rpc_client = server->client,
8551                 .rpc_message = &msg,
8552                 .callback_ops = &nfs4_layoutget_call_ops,
8553                 .callback_data = lgp,
8554                 .flags = RPC_TASK_ASYNC,
8555         };
8556         struct pnfs_layout_segment *lseg = NULL;
8557         struct nfs4_exception exception = {
8558                 .inode = inode,
8559                 .timeout = *timeout,
8560         };
8561         int status = 0;
8562
8563         dprintk("--> %s\n", __func__);
8564
8565         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8566         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8567
8568         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8569         if (!lgp->args.layout.pages) {
8570                 nfs4_layoutget_release(lgp);
8571                 return ERR_PTR(-ENOMEM);
8572         }
8573         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8574
8575         lgp->res.layoutp = &lgp->args.layout;
8576         lgp->res.seq_res.sr_slot = NULL;
8577         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8578
8579         task = rpc_run_task(&task_setup_data);
8580         if (IS_ERR(task))
8581                 return ERR_CAST(task);
8582         status = nfs4_wait_for_completion_rpc_task(task);
8583         if (status == 0) {
8584                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8585                 *timeout = exception.timeout;
8586         }
8587
8588         trace_nfs4_layoutget(lgp->args.ctx,
8589                         &lgp->args.range,
8590                         &lgp->res.range,
8591                         &lgp->res.stateid,
8592                         status);
8593
8594         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8595         if (status == 0 && lgp->res.layoutp->len)
8596                 lseg = pnfs_layout_process(lgp);
8597         nfs4_sequence_free_slot(&lgp->res.seq_res);
8598         rpc_put_task(task);
8599         dprintk("<-- %s status=%d\n", __func__, status);
8600         if (status)
8601                 return ERR_PTR(status);
8602         return lseg;
8603 }
8604
8605 static void
8606 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8607 {
8608         struct nfs4_layoutreturn *lrp = calldata;
8609
8610         dprintk("--> %s\n", __func__);
8611         nfs41_setup_sequence(lrp->clp->cl_session,
8612                         &lrp->args.seq_args,
8613                         &lrp->res.seq_res,
8614                         task);
8615 }
8616
8617 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8618 {
8619         struct nfs4_layoutreturn *lrp = calldata;
8620         struct nfs_server *server;
8621
8622         dprintk("--> %s\n", __func__);
8623
8624         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8625                 return;
8626
8627         server = NFS_SERVER(lrp->args.inode);
8628         switch (task->tk_status) {
8629         default:
8630                 task->tk_status = 0;
8631         case 0:
8632                 break;
8633         case -NFS4ERR_DELAY:
8634                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8635                         break;
8636                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8637                 rpc_restart_call_prepare(task);
8638                 return;
8639         }
8640         dprintk("<-- %s\n", __func__);
8641 }
8642
8643 static void nfs4_layoutreturn_release(void *calldata)
8644 {
8645         struct nfs4_layoutreturn *lrp = calldata;
8646         struct pnfs_layout_hdr *lo = lrp->args.layout;
8647
8648         dprintk("--> %s\n", __func__);
8649         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8650                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8651         nfs4_sequence_free_slot(&lrp->res.seq_res);
8652         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8653                 lrp->ld_private.ops->free(&lrp->ld_private);
8654         pnfs_put_layout_hdr(lrp->args.layout);
8655         nfs_iput_and_deactive(lrp->inode);
8656         kfree(calldata);
8657         dprintk("<-- %s\n", __func__);
8658 }
8659
8660 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8661         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8662         .rpc_call_done = nfs4_layoutreturn_done,
8663         .rpc_release = nfs4_layoutreturn_release,
8664 };
8665
8666 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8667 {
8668         struct rpc_task *task;
8669         struct rpc_message msg = {
8670                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8671                 .rpc_argp = &lrp->args,
8672                 .rpc_resp = &lrp->res,
8673                 .rpc_cred = lrp->cred,
8674         };
8675         struct rpc_task_setup task_setup_data = {
8676                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8677                 .rpc_message = &msg,
8678                 .callback_ops = &nfs4_layoutreturn_call_ops,
8679                 .callback_data = lrp,
8680         };
8681         int status = 0;
8682
8683         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8684                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8685                         &task_setup_data.rpc_client, &msg);
8686
8687         dprintk("--> %s\n", __func__);
8688         if (!sync) {
8689                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8690                 if (!lrp->inode) {
8691                         nfs4_layoutreturn_release(lrp);
8692                         return -EAGAIN;
8693                 }
8694                 task_setup_data.flags |= RPC_TASK_ASYNC;
8695         }
8696         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8697         task = rpc_run_task(&task_setup_data);
8698         if (IS_ERR(task))
8699                 return PTR_ERR(task);
8700         if (sync)
8701                 status = task->tk_status;
8702         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8703         dprintk("<-- %s status=%d\n", __func__, status);
8704         rpc_put_task(task);
8705         return status;
8706 }
8707
8708 static int
8709 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8710                 struct pnfs_device *pdev,
8711                 struct rpc_cred *cred)
8712 {
8713         struct nfs4_getdeviceinfo_args args = {
8714                 .pdev = pdev,
8715                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8716                         NOTIFY_DEVICEID4_DELETE,
8717         };
8718         struct nfs4_getdeviceinfo_res res = {
8719                 .pdev = pdev,
8720         };
8721         struct rpc_message msg = {
8722                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8723                 .rpc_argp = &args,
8724                 .rpc_resp = &res,
8725                 .rpc_cred = cred,
8726         };
8727         int status;
8728
8729         dprintk("--> %s\n", __func__);
8730         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8731         if (res.notification & ~args.notify_types)
8732                 dprintk("%s: unsupported notification\n", __func__);
8733         if (res.notification != args.notify_types)
8734                 pdev->nocache = 1;
8735
8736         dprintk("<-- %s status=%d\n", __func__, status);
8737
8738         return status;
8739 }
8740
8741 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8742                 struct pnfs_device *pdev,
8743                 struct rpc_cred *cred)
8744 {
8745         struct nfs4_exception exception = { };
8746         int err;
8747
8748         do {
8749                 err = nfs4_handle_exception(server,
8750                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8751                                         &exception);
8752         } while (exception.retry);
8753         return err;
8754 }
8755 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8756
8757 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8758 {
8759         struct nfs4_layoutcommit_data *data = calldata;
8760         struct nfs_server *server = NFS_SERVER(data->args.inode);
8761         struct nfs4_session *session = nfs4_get_session(server);
8762
8763         nfs41_setup_sequence(session,
8764                         &data->args.seq_args,
8765                         &data->res.seq_res,
8766                         task);
8767 }
8768
8769 static void
8770 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8771 {
8772         struct nfs4_layoutcommit_data *data = calldata;
8773         struct nfs_server *server = NFS_SERVER(data->args.inode);
8774
8775         if (!nfs41_sequence_done(task, &data->res.seq_res))
8776                 return;
8777
8778         switch (task->tk_status) { /* Just ignore these failures */
8779         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8780         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8781         case -NFS4ERR_BADLAYOUT:     /* no layout */
8782         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8783                 task->tk_status = 0;
8784         case 0:
8785                 break;
8786         default:
8787                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8788                         rpc_restart_call_prepare(task);
8789                         return;
8790                 }
8791         }
8792 }
8793
8794 static void nfs4_layoutcommit_release(void *calldata)
8795 {
8796         struct nfs4_layoutcommit_data *data = calldata;
8797
8798         pnfs_cleanup_layoutcommit(data);
8799         nfs_post_op_update_inode_force_wcc(data->args.inode,
8800                                            data->res.fattr);
8801         put_rpccred(data->cred);
8802         nfs_iput_and_deactive(data->inode);
8803         kfree(data);
8804 }
8805
8806 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8807         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8808         .rpc_call_done = nfs4_layoutcommit_done,
8809         .rpc_release = nfs4_layoutcommit_release,
8810 };
8811
8812 int
8813 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8814 {
8815         struct rpc_message msg = {
8816                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8817                 .rpc_argp = &data->args,
8818                 .rpc_resp = &data->res,
8819                 .rpc_cred = data->cred,
8820         };
8821         struct rpc_task_setup task_setup_data = {
8822                 .task = &data->task,
8823                 .rpc_client = NFS_CLIENT(data->args.inode),
8824                 .rpc_message = &msg,
8825                 .callback_ops = &nfs4_layoutcommit_ops,
8826                 .callback_data = data,
8827         };
8828         struct rpc_task *task;
8829         int status = 0;
8830
8831         dprintk("NFS: initiating layoutcommit call. sync %d "
8832                 "lbw: %llu inode %lu\n", sync,
8833                 data->args.lastbytewritten,
8834                 data->args.inode->i_ino);
8835
8836         if (!sync) {
8837                 data->inode = nfs_igrab_and_active(data->args.inode);
8838                 if (data->inode == NULL) {
8839                         nfs4_layoutcommit_release(data);
8840                         return -EAGAIN;
8841                 }
8842                 task_setup_data.flags = RPC_TASK_ASYNC;
8843         }
8844         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8845         task = rpc_run_task(&task_setup_data);
8846         if (IS_ERR(task))
8847                 return PTR_ERR(task);
8848         if (sync)
8849                 status = task->tk_status;
8850         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8851         dprintk("%s: status %d\n", __func__, status);
8852         rpc_put_task(task);
8853         return status;
8854 }
8855
8856 /**
8857  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8858  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8859  */
8860 static int
8861 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8862                     struct nfs_fsinfo *info,
8863                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8864 {
8865         struct nfs41_secinfo_no_name_args args = {
8866                 .style = SECINFO_STYLE_CURRENT_FH,
8867         };
8868         struct nfs4_secinfo_res res = {
8869                 .flavors = flavors,
8870         };
8871         struct rpc_message msg = {
8872                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8873                 .rpc_argp = &args,
8874                 .rpc_resp = &res,
8875         };
8876         struct rpc_clnt *clnt = server->client;
8877         struct rpc_cred *cred = NULL;
8878         int status;
8879
8880         if (use_integrity) {
8881                 clnt = server->nfs_client->cl_rpcclient;
8882                 cred = nfs4_get_clid_cred(server->nfs_client);
8883                 msg.rpc_cred = cred;
8884         }
8885
8886         dprintk("--> %s\n", __func__);
8887         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8888                                 &res.seq_res, 0);
8889         dprintk("<-- %s status=%d\n", __func__, status);
8890
8891         if (cred)
8892                 put_rpccred(cred);
8893
8894         return status;
8895 }
8896
8897 static int
8898 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8899                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8900 {
8901         struct nfs4_exception exception = { };
8902         int err;
8903         do {
8904                 /* first try using integrity protection */
8905                 err = -NFS4ERR_WRONGSEC;
8906
8907                 /* try to use integrity protection with machine cred */
8908                 if (_nfs4_is_integrity_protected(server->nfs_client))
8909                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8910                                                           flavors, true);
8911
8912                 /*
8913                  * if unable to use integrity protection, or SECINFO with
8914                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8915                  * disallowed by spec, but exists in deployed servers) use
8916                  * the current filesystem's rpc_client and the user cred.
8917                  */
8918                 if (err == -NFS4ERR_WRONGSEC)
8919                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8920                                                           flavors, false);
8921
8922                 switch (err) {
8923                 case 0:
8924                 case -NFS4ERR_WRONGSEC:
8925                 case -ENOTSUPP:
8926                         goto out;
8927                 default:
8928                         err = nfs4_handle_exception(server, err, &exception);
8929                 }
8930         } while (exception.retry);
8931 out:
8932         return err;
8933 }
8934
8935 static int
8936 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8937                     struct nfs_fsinfo *info)
8938 {
8939         int err;
8940         struct page *page;
8941         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8942         struct nfs4_secinfo_flavors *flavors;
8943         struct nfs4_secinfo4 *secinfo;
8944         int i;
8945
8946         page = alloc_page(GFP_KERNEL);
8947         if (!page) {
8948                 err = -ENOMEM;
8949                 goto out;
8950         }
8951
8952         flavors = page_address(page);
8953         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8954
8955         /*
8956          * Fall back on "guess and check" method if
8957          * the server doesn't support SECINFO_NO_NAME
8958          */
8959         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8960                 err = nfs4_find_root_sec(server, fhandle, info);
8961                 goto out_freepage;
8962         }
8963         if (err)
8964                 goto out_freepage;
8965
8966         for (i = 0; i < flavors->num_flavors; i++) {
8967                 secinfo = &flavors->flavors[i];
8968
8969                 switch (secinfo->flavor) {
8970                 case RPC_AUTH_NULL:
8971                 case RPC_AUTH_UNIX:
8972                 case RPC_AUTH_GSS:
8973                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8974                                         &secinfo->flavor_info);
8975                         break;
8976                 default:
8977                         flavor = RPC_AUTH_MAXFLAVOR;
8978                         break;
8979                 }
8980
8981                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8982                         flavor = RPC_AUTH_MAXFLAVOR;
8983
8984                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8985                         err = nfs4_lookup_root_sec(server, fhandle,
8986                                                    info, flavor);
8987                         if (!err)
8988                                 break;
8989                 }
8990         }
8991
8992         if (flavor == RPC_AUTH_MAXFLAVOR)
8993                 err = -EPERM;
8994
8995 out_freepage:
8996         put_page(page);
8997         if (err == -EACCES)
8998                 return -EPERM;
8999 out:
9000         return err;
9001 }
9002
9003 static int _nfs41_test_stateid(struct nfs_server *server,
9004                 nfs4_stateid *stateid,
9005                 struct rpc_cred *cred)
9006 {
9007         int status;
9008         struct nfs41_test_stateid_args args = {
9009                 .stateid = stateid,
9010         };
9011         struct nfs41_test_stateid_res res;
9012         struct rpc_message msg = {
9013                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9014                 .rpc_argp = &args,
9015                 .rpc_resp = &res,
9016                 .rpc_cred = cred,
9017         };
9018         struct rpc_clnt *rpc_client = server->client;
9019
9020         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9021                 &rpc_client, &msg);
9022
9023         dprintk("NFS call  test_stateid %p\n", stateid);
9024         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
9025         nfs4_set_sequence_privileged(&args.seq_args);
9026         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9027                         &args.seq_args, &res.seq_res);
9028         if (status != NFS_OK) {
9029                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9030                 return status;
9031         }
9032         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9033         return -res.status;
9034 }
9035
9036 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9037                 int err, struct nfs4_exception *exception)
9038 {
9039         exception->retry = 0;
9040         switch(err) {
9041         case -NFS4ERR_DELAY:
9042         case -NFS4ERR_RETRY_UNCACHED_REP:
9043                 nfs4_handle_exception(server, err, exception);
9044                 break;
9045         case -NFS4ERR_BADSESSION:
9046         case -NFS4ERR_BADSLOT:
9047         case -NFS4ERR_BAD_HIGH_SLOT:
9048         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9049         case -NFS4ERR_DEADSESSION:
9050                 nfs4_do_handle_exception(server, err, exception);
9051         }
9052 }
9053
9054 /**
9055  * nfs41_test_stateid - perform a TEST_STATEID operation
9056  *
9057  * @server: server / transport on which to perform the operation
9058  * @stateid: state ID to test
9059  * @cred: credential
9060  *
9061  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9062  * Otherwise a negative NFS4ERR value is returned if the operation
9063  * failed or the state ID is not currently valid.
9064  */
9065 static int nfs41_test_stateid(struct nfs_server *server,
9066                 nfs4_stateid *stateid,
9067                 struct rpc_cred *cred)
9068 {
9069         struct nfs4_exception exception = { };
9070         int err;
9071         do {
9072                 err = _nfs41_test_stateid(server, stateid, cred);
9073                 nfs4_handle_delay_or_session_error(server, err, &exception);
9074         } while (exception.retry);
9075         return err;
9076 }
9077
9078 struct nfs_free_stateid_data {
9079         struct nfs_server *server;
9080         struct nfs41_free_stateid_args args;
9081         struct nfs41_free_stateid_res res;
9082 };
9083
9084 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9085 {
9086         struct nfs_free_stateid_data *data = calldata;
9087         nfs41_setup_sequence(nfs4_get_session(data->server),
9088                         &data->args.seq_args,
9089                         &data->res.seq_res,
9090                         task);
9091 }
9092
9093 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9094 {
9095         struct nfs_free_stateid_data *data = calldata;
9096
9097         nfs41_sequence_done(task, &data->res.seq_res);
9098
9099         switch (task->tk_status) {
9100         case -NFS4ERR_DELAY:
9101                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9102                         rpc_restart_call_prepare(task);
9103         }
9104 }
9105
9106 static void nfs41_free_stateid_release(void *calldata)
9107 {
9108         kfree(calldata);
9109 }
9110
9111 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9112         .rpc_call_prepare = nfs41_free_stateid_prepare,
9113         .rpc_call_done = nfs41_free_stateid_done,
9114         .rpc_release = nfs41_free_stateid_release,
9115 };
9116
9117 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9118                 const nfs4_stateid *stateid,
9119                 struct rpc_cred *cred,
9120                 bool privileged)
9121 {
9122         struct rpc_message msg = {
9123                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9124                 .rpc_cred = cred,
9125         };
9126         struct rpc_task_setup task_setup = {
9127                 .rpc_client = server->client,
9128                 .rpc_message = &msg,
9129                 .callback_ops = &nfs41_free_stateid_ops,
9130                 .flags = RPC_TASK_ASYNC,
9131         };
9132         struct nfs_free_stateid_data *data;
9133
9134         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9135                 &task_setup.rpc_client, &msg);
9136
9137         dprintk("NFS call  free_stateid %p\n", stateid);
9138         data = kmalloc(sizeof(*data), GFP_NOFS);
9139         if (!data)
9140                 return ERR_PTR(-ENOMEM);
9141         data->server = server;
9142         nfs4_stateid_copy(&data->args.stateid, stateid);
9143
9144         task_setup.callback_data = data;
9145
9146         msg.rpc_argp = &data->args;
9147         msg.rpc_resp = &data->res;
9148         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9149         if (privileged)
9150                 nfs4_set_sequence_privileged(&data->args.seq_args);
9151
9152         return rpc_run_task(&task_setup);
9153 }
9154
9155 /**
9156  * nfs41_free_stateid - perform a FREE_STATEID operation
9157  *
9158  * @server: server / transport on which to perform the operation
9159  * @stateid: state ID to release
9160  * @cred: credential
9161  * @is_recovery: set to true if this call needs to be privileged
9162  *
9163  * Note: this function is always asynchronous.
9164  */
9165 static int nfs41_free_stateid(struct nfs_server *server,
9166                 const nfs4_stateid *stateid,
9167                 struct rpc_cred *cred,
9168                 bool is_recovery)
9169 {
9170         struct rpc_task *task;
9171
9172         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9173         if (IS_ERR(task))
9174                 return PTR_ERR(task);
9175         rpc_put_task(task);
9176         return 0;
9177 }
9178
9179 static void
9180 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9181 {
9182         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9183
9184         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9185         nfs4_free_lock_state(server, lsp);
9186 }
9187
9188 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9189                 const nfs4_stateid *s2)
9190 {
9191         if (s1->type != s2->type)
9192                 return false;
9193
9194         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9195                 return false;
9196
9197         if (s1->seqid == s2->seqid)
9198                 return true;
9199         if (s1->seqid == 0 || s2->seqid == 0)
9200                 return true;
9201
9202         return false;
9203 }
9204
9205 #endif /* CONFIG_NFS_V4_1 */
9206
9207 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9208                 const nfs4_stateid *s2)
9209 {
9210         return nfs4_stateid_match(s1, s2);
9211 }
9212
9213
9214 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9215         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9216         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9217         .recover_open   = nfs4_open_reclaim,
9218         .recover_lock   = nfs4_lock_reclaim,
9219         .establish_clid = nfs4_init_clientid,
9220         .detect_trunking = nfs40_discover_server_trunking,
9221 };
9222
9223 #if defined(CONFIG_NFS_V4_1)
9224 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9225         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9226         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9227         .recover_open   = nfs4_open_reclaim,
9228         .recover_lock   = nfs4_lock_reclaim,
9229         .establish_clid = nfs41_init_clientid,
9230         .reclaim_complete = nfs41_proc_reclaim_complete,
9231         .detect_trunking = nfs41_discover_server_trunking,
9232 };
9233 #endif /* CONFIG_NFS_V4_1 */
9234
9235 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9236         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9237         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9238         .recover_open   = nfs40_open_expired,
9239         .recover_lock   = nfs4_lock_expired,
9240         .establish_clid = nfs4_init_clientid,
9241 };
9242
9243 #if defined(CONFIG_NFS_V4_1)
9244 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9245         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9246         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9247         .recover_open   = nfs41_open_expired,
9248         .recover_lock   = nfs41_lock_expired,
9249         .establish_clid = nfs41_init_clientid,
9250 };
9251 #endif /* CONFIG_NFS_V4_1 */
9252
9253 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9254         .sched_state_renewal = nfs4_proc_async_renew,
9255         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9256         .renew_lease = nfs4_proc_renew,
9257 };
9258
9259 #if defined(CONFIG_NFS_V4_1)
9260 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9261         .sched_state_renewal = nfs41_proc_async_sequence,
9262         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9263         .renew_lease = nfs4_proc_sequence,
9264 };
9265 #endif
9266
9267 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9268         .get_locations = _nfs40_proc_get_locations,
9269         .fsid_present = _nfs40_proc_fsid_present,
9270 };
9271
9272 #if defined(CONFIG_NFS_V4_1)
9273 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9274         .get_locations = _nfs41_proc_get_locations,
9275         .fsid_present = _nfs41_proc_fsid_present,
9276 };
9277 #endif  /* CONFIG_NFS_V4_1 */
9278
9279 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9280         .minor_version = 0,
9281         .init_caps = NFS_CAP_READDIRPLUS
9282                 | NFS_CAP_ATOMIC_OPEN
9283                 | NFS_CAP_POSIX_LOCK,
9284         .init_client = nfs40_init_client,
9285         .shutdown_client = nfs40_shutdown_client,
9286         .match_stateid = nfs4_match_stateid,
9287         .find_root_sec = nfs4_find_root_sec,
9288         .free_lock_state = nfs4_release_lockowner,
9289         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9290         .alloc_seqid = nfs_alloc_seqid,
9291         .call_sync_ops = &nfs40_call_sync_ops,
9292         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9293         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9294         .state_renewal_ops = &nfs40_state_renewal_ops,
9295         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9296 };
9297
9298 #if defined(CONFIG_NFS_V4_1)
9299 static struct nfs_seqid *
9300 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9301 {
9302         return NULL;
9303 }
9304
9305 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9306         .minor_version = 1,
9307         .init_caps = NFS_CAP_READDIRPLUS
9308                 | NFS_CAP_ATOMIC_OPEN
9309                 | NFS_CAP_POSIX_LOCK
9310                 | NFS_CAP_STATEID_NFSV41
9311                 | NFS_CAP_ATOMIC_OPEN_V1,
9312         .init_client = nfs41_init_client,
9313         .shutdown_client = nfs41_shutdown_client,
9314         .match_stateid = nfs41_match_stateid,
9315         .find_root_sec = nfs41_find_root_sec,
9316         .free_lock_state = nfs41_free_lock_state,
9317         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9318         .alloc_seqid = nfs_alloc_no_seqid,
9319         .session_trunk = nfs4_test_session_trunk,
9320         .call_sync_ops = &nfs41_call_sync_ops,
9321         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9322         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9323         .state_renewal_ops = &nfs41_state_renewal_ops,
9324         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9325 };
9326 #endif
9327
9328 #if defined(CONFIG_NFS_V4_2)
9329 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9330         .minor_version = 2,
9331         .init_caps = NFS_CAP_READDIRPLUS
9332                 | NFS_CAP_ATOMIC_OPEN
9333                 | NFS_CAP_POSIX_LOCK
9334                 | NFS_CAP_STATEID_NFSV41
9335                 | NFS_CAP_ATOMIC_OPEN_V1
9336                 | NFS_CAP_ALLOCATE
9337                 | NFS_CAP_COPY
9338                 | NFS_CAP_DEALLOCATE
9339                 | NFS_CAP_SEEK
9340                 | NFS_CAP_LAYOUTSTATS
9341                 | NFS_CAP_CLONE,
9342         .init_client = nfs41_init_client,
9343         .shutdown_client = nfs41_shutdown_client,
9344         .match_stateid = nfs41_match_stateid,
9345         .find_root_sec = nfs41_find_root_sec,
9346         .free_lock_state = nfs41_free_lock_state,
9347         .call_sync_ops = &nfs41_call_sync_ops,
9348         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9349         .alloc_seqid = nfs_alloc_no_seqid,
9350         .session_trunk = nfs4_test_session_trunk,
9351         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9352         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9353         .state_renewal_ops = &nfs41_state_renewal_ops,
9354         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9355 };
9356 #endif
9357
9358 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9359         [0] = &nfs_v4_0_minor_ops,
9360 #if defined(CONFIG_NFS_V4_1)
9361         [1] = &nfs_v4_1_minor_ops,
9362 #endif
9363 #if defined(CONFIG_NFS_V4_2)
9364         [2] = &nfs_v4_2_minor_ops,
9365 #endif
9366 };
9367
9368 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9369 {
9370         ssize_t error, error2;
9371
9372         error = generic_listxattr(dentry, list, size);
9373         if (error < 0)
9374                 return error;
9375         if (list) {
9376                 list += error;
9377                 size -= error;
9378         }
9379
9380         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9381         if (error2 < 0)
9382                 return error2;
9383         return error + error2;
9384 }
9385
9386 static const struct inode_operations nfs4_dir_inode_operations = {
9387         .create         = nfs_create,
9388         .lookup         = nfs_lookup,
9389         .atomic_open    = nfs_atomic_open,
9390         .link           = nfs_link,
9391         .unlink         = nfs_unlink,
9392         .symlink        = nfs_symlink,
9393         .mkdir          = nfs_mkdir,
9394         .rmdir          = nfs_rmdir,
9395         .mknod          = nfs_mknod,
9396         .rename         = nfs_rename,
9397         .permission     = nfs_permission,
9398         .getattr        = nfs_getattr,
9399         .setattr        = nfs_setattr,
9400         .listxattr      = nfs4_listxattr,
9401 };
9402
9403 static const struct inode_operations nfs4_file_inode_operations = {
9404         .permission     = nfs_permission,
9405         .getattr        = nfs_getattr,
9406         .setattr        = nfs_setattr,
9407         .listxattr      = nfs4_listxattr,
9408 };
9409
9410 const struct nfs_rpc_ops nfs_v4_clientops = {
9411         .version        = 4,                    /* protocol version */
9412         .dentry_ops     = &nfs4_dentry_operations,
9413         .dir_inode_ops  = &nfs4_dir_inode_operations,
9414         .file_inode_ops = &nfs4_file_inode_operations,
9415         .file_ops       = &nfs4_file_operations,
9416         .getroot        = nfs4_proc_get_root,
9417         .submount       = nfs4_submount,
9418         .try_mount      = nfs4_try_mount,
9419         .getattr        = nfs4_proc_getattr,
9420         .setattr        = nfs4_proc_setattr,
9421         .lookup         = nfs4_proc_lookup,
9422         .access         = nfs4_proc_access,
9423         .readlink       = nfs4_proc_readlink,
9424         .create         = nfs4_proc_create,
9425         .remove         = nfs4_proc_remove,
9426         .unlink_setup   = nfs4_proc_unlink_setup,
9427         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9428         .unlink_done    = nfs4_proc_unlink_done,
9429         .rename_setup   = nfs4_proc_rename_setup,
9430         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9431         .rename_done    = nfs4_proc_rename_done,
9432         .link           = nfs4_proc_link,
9433         .symlink        = nfs4_proc_symlink,
9434         .mkdir          = nfs4_proc_mkdir,
9435         .rmdir          = nfs4_proc_remove,
9436         .readdir        = nfs4_proc_readdir,
9437         .mknod          = nfs4_proc_mknod,
9438         .statfs         = nfs4_proc_statfs,
9439         .fsinfo         = nfs4_proc_fsinfo,
9440         .pathconf       = nfs4_proc_pathconf,
9441         .set_capabilities = nfs4_server_capabilities,
9442         .decode_dirent  = nfs4_decode_dirent,
9443         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9444         .read_setup     = nfs4_proc_read_setup,
9445         .read_done      = nfs4_read_done,
9446         .write_setup    = nfs4_proc_write_setup,
9447         .write_done     = nfs4_write_done,
9448         .commit_setup   = nfs4_proc_commit_setup,
9449         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9450         .commit_done    = nfs4_commit_done,
9451         .lock           = nfs4_proc_lock,
9452         .clear_acl_cache = nfs4_zap_acl_attr,
9453         .close_context  = nfs4_close_context,
9454         .open_context   = nfs4_atomic_open,
9455         .have_delegation = nfs4_have_delegation,
9456         .return_delegation = nfs4_inode_return_delegation,
9457         .alloc_client   = nfs4_alloc_client,
9458         .init_client    = nfs4_init_client,
9459         .free_client    = nfs4_free_client,
9460         .create_server  = nfs4_create_server,
9461         .clone_server   = nfs_clone_server,
9462 };
9463
9464 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9465         .name   = XATTR_NAME_NFSV4_ACL,
9466         .list   = nfs4_xattr_list_nfs4_acl,
9467         .get    = nfs4_xattr_get_nfs4_acl,
9468         .set    = nfs4_xattr_set_nfs4_acl,
9469 };
9470
9471 const struct xattr_handler *nfs4_xattr_handlers[] = {
9472         &nfs4_xattr_nfs4_acl_handler,
9473 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9474         &nfs4_xattr_nfs4_label_handler,
9475 #endif
9476         NULL
9477 };
9478
9479 /*
9480  * Local variables:
9481  *  c-basic-offset: 8
9482  * End:
9483  */