Merge tag 'samsung-fixes-4.7' of git://git.kernel.org/pub/scm/linux/kernel/git/krzk...
[cascardo/linux.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/file.h>
42 #include <linux/string.h>
43 #include <linux/ratelimit.h>
44 #include <linux/printk.h>
45 #include <linux/slab.h>
46 #include <linux/sunrpc/clnt.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69
70 #include "nfs4trace.h"
71
72 #define NFSDBG_FACILITY         NFSDBG_PROC
73
74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
76
77 /* file attributes which can be mapped to nfs attributes */
78 #define NFS4_VALID_ATTRS (ATTR_MODE \
79         | ATTR_UID \
80         | ATTR_GID \
81         | ATTR_SIZE \
82         | ATTR_ATIME \
83         | ATTR_MTIME \
84         | ATTR_CTIME \
85         | ATTR_ATIME_SET \
86         | ATTR_MTIME_SET)
87
88 struct nfs4_opendata;
89 static int _nfs4_proc_open(struct nfs4_opendata *data);
90 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
91 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
92 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
93 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
95 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
96                             struct nfs_fattr *fattr, struct iattr *sattr,
97                             struct nfs4_state *state, struct nfs4_label *ilabel,
98                             struct nfs4_label *olabel);
99 #ifdef CONFIG_NFS_V4_1
100 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
101                 struct rpc_cred *);
102 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
103                 struct rpc_cred *);
104 #endif
105
106 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
107 static inline struct nfs4_label *
108 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
109         struct iattr *sattr, struct nfs4_label *label)
110 {
111         int err;
112
113         if (label == NULL)
114                 return NULL;
115
116         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
117                 return NULL;
118
119         err = security_dentry_init_security(dentry, sattr->ia_mode,
120                                 &dentry->d_name, (void **)&label->label, &label->len);
121         if (err == 0)
122                 return label;
123
124         return NULL;
125 }
126 static inline void
127 nfs4_label_release_security(struct nfs4_label *label)
128 {
129         if (label)
130                 security_release_secctx(label->label, label->len);
131 }
132 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
133 {
134         if (label)
135                 return server->attr_bitmask;
136
137         return server->attr_bitmask_nl;
138 }
139 #else
140 static inline struct nfs4_label *
141 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
142         struct iattr *sattr, struct nfs4_label *l)
143 { return NULL; }
144 static inline void
145 nfs4_label_release_security(struct nfs4_label *label)
146 { return; }
147 static inline u32 *
148 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
149 { return server->attr_bitmask; }
150 #endif
151
152 /* Prevent leaks of NFSv4 errors into userland */
153 static int nfs4_map_errors(int err)
154 {
155         if (err >= -1000)
156                 return err;
157         switch (err) {
158         case -NFS4ERR_RESOURCE:
159         case -NFS4ERR_LAYOUTTRYLATER:
160         case -NFS4ERR_RECALLCONFLICT:
161                 return -EREMOTEIO;
162         case -NFS4ERR_WRONGSEC:
163         case -NFS4ERR_WRONG_CRED:
164                 return -EPERM;
165         case -NFS4ERR_BADOWNER:
166         case -NFS4ERR_BADNAME:
167                 return -EINVAL;
168         case -NFS4ERR_SHARE_DENIED:
169                 return -EACCES;
170         case -NFS4ERR_MINOR_VERS_MISMATCH:
171                 return -EPROTONOSUPPORT;
172         case -NFS4ERR_FILE_OPEN:
173                 return -EBUSY;
174         default:
175                 dprintk("%s could not handle NFSv4 error %d\n",
176                                 __func__, -err);
177                 break;
178         }
179         return -EIO;
180 }
181
182 /*
183  * This is our standard bitmap for GETATTR requests.
184  */
185 const u32 nfs4_fattr_bitmap[3] = {
186         FATTR4_WORD0_TYPE
187         | FATTR4_WORD0_CHANGE
188         | FATTR4_WORD0_SIZE
189         | FATTR4_WORD0_FSID
190         | FATTR4_WORD0_FILEID,
191         FATTR4_WORD1_MODE
192         | FATTR4_WORD1_NUMLINKS
193         | FATTR4_WORD1_OWNER
194         | FATTR4_WORD1_OWNER_GROUP
195         | FATTR4_WORD1_RAWDEV
196         | FATTR4_WORD1_SPACE_USED
197         | FATTR4_WORD1_TIME_ACCESS
198         | FATTR4_WORD1_TIME_METADATA
199         | FATTR4_WORD1_TIME_MODIFY
200         | FATTR4_WORD1_MOUNTED_ON_FILEID,
201 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
202         FATTR4_WORD2_SECURITY_LABEL
203 #endif
204 };
205
206 static const u32 nfs4_pnfs_open_bitmap[3] = {
207         FATTR4_WORD0_TYPE
208         | FATTR4_WORD0_CHANGE
209         | FATTR4_WORD0_SIZE
210         | FATTR4_WORD0_FSID
211         | FATTR4_WORD0_FILEID,
212         FATTR4_WORD1_MODE
213         | FATTR4_WORD1_NUMLINKS
214         | FATTR4_WORD1_OWNER
215         | FATTR4_WORD1_OWNER_GROUP
216         | FATTR4_WORD1_RAWDEV
217         | FATTR4_WORD1_SPACE_USED
218         | FATTR4_WORD1_TIME_ACCESS
219         | FATTR4_WORD1_TIME_METADATA
220         | FATTR4_WORD1_TIME_MODIFY,
221         FATTR4_WORD2_MDSTHRESHOLD
222 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
223         | FATTR4_WORD2_SECURITY_LABEL
224 #endif
225 };
226
227 static const u32 nfs4_open_noattr_bitmap[3] = {
228         FATTR4_WORD0_TYPE
229         | FATTR4_WORD0_CHANGE
230         | FATTR4_WORD0_FILEID,
231 };
232
233 const u32 nfs4_statfs_bitmap[3] = {
234         FATTR4_WORD0_FILES_AVAIL
235         | FATTR4_WORD0_FILES_FREE
236         | FATTR4_WORD0_FILES_TOTAL,
237         FATTR4_WORD1_SPACE_AVAIL
238         | FATTR4_WORD1_SPACE_FREE
239         | FATTR4_WORD1_SPACE_TOTAL
240 };
241
242 const u32 nfs4_pathconf_bitmap[3] = {
243         FATTR4_WORD0_MAXLINK
244         | FATTR4_WORD0_MAXNAME,
245         0
246 };
247
248 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
249                         | FATTR4_WORD0_MAXREAD
250                         | FATTR4_WORD0_MAXWRITE
251                         | FATTR4_WORD0_LEASE_TIME,
252                         FATTR4_WORD1_TIME_DELTA
253                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
254                         FATTR4_WORD2_LAYOUT_BLKSIZE
255                         | FATTR4_WORD2_CLONE_BLKSIZE
256 };
257
258 const u32 nfs4_fs_locations_bitmap[3] = {
259         FATTR4_WORD0_TYPE
260         | FATTR4_WORD0_CHANGE
261         | FATTR4_WORD0_SIZE
262         | FATTR4_WORD0_FSID
263         | FATTR4_WORD0_FILEID
264         | FATTR4_WORD0_FS_LOCATIONS,
265         FATTR4_WORD1_MODE
266         | FATTR4_WORD1_NUMLINKS
267         | FATTR4_WORD1_OWNER
268         | FATTR4_WORD1_OWNER_GROUP
269         | FATTR4_WORD1_RAWDEV
270         | FATTR4_WORD1_SPACE_USED
271         | FATTR4_WORD1_TIME_ACCESS
272         | FATTR4_WORD1_TIME_METADATA
273         | FATTR4_WORD1_TIME_MODIFY
274         | FATTR4_WORD1_MOUNTED_ON_FILEID,
275 };
276
277 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
278                 struct nfs4_readdir_arg *readdir)
279 {
280         __be32 *start, *p;
281
282         if (cookie > 2) {
283                 readdir->cookie = cookie;
284                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
285                 return;
286         }
287
288         readdir->cookie = 0;
289         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
290         if (cookie == 2)
291                 return;
292         
293         /*
294          * NFSv4 servers do not return entries for '.' and '..'
295          * Therefore, we fake these entries here.  We let '.'
296          * have cookie 0 and '..' have cookie 1.  Note that
297          * when talking to the server, we always send cookie 0
298          * instead of 1 or 2.
299          */
300         start = p = kmap_atomic(*readdir->pages);
301         
302         if (cookie == 0) {
303                 *p++ = xdr_one;                                  /* next */
304                 *p++ = xdr_zero;                   /* cookie, first word */
305                 *p++ = xdr_one;                   /* cookie, second word */
306                 *p++ = xdr_one;                             /* entry len */
307                 memcpy(p, ".\0\0\0", 4);                        /* entry */
308                 p++;
309                 *p++ = xdr_one;                         /* bitmap length */
310                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
311                 *p++ = htonl(8);              /* attribute buffer length */
312                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
313         }
314         
315         *p++ = xdr_one;                                  /* next */
316         *p++ = xdr_zero;                   /* cookie, first word */
317         *p++ = xdr_two;                   /* cookie, second word */
318         *p++ = xdr_two;                             /* entry len */
319         memcpy(p, "..\0\0", 4);                         /* entry */
320         p++;
321         *p++ = xdr_one;                         /* bitmap length */
322         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
323         *p++ = htonl(8);              /* attribute buffer length */
324         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
325
326         readdir->pgbase = (char *)p - (char *)start;
327         readdir->count -= readdir->pgbase;
328         kunmap_atomic(start);
329 }
330
331 static long nfs4_update_delay(long *timeout)
332 {
333         long ret;
334         if (!timeout)
335                 return NFS4_POLL_RETRY_MAX;
336         if (*timeout <= 0)
337                 *timeout = NFS4_POLL_RETRY_MIN;
338         if (*timeout > NFS4_POLL_RETRY_MAX)
339                 *timeout = NFS4_POLL_RETRY_MAX;
340         ret = *timeout;
341         *timeout <<= 1;
342         return ret;
343 }
344
345 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
346 {
347         int res = 0;
348
349         might_sleep();
350
351         freezable_schedule_timeout_killable_unsafe(
352                 nfs4_update_delay(timeout));
353         if (fatal_signal_pending(current))
354                 res = -ERESTARTSYS;
355         return res;
356 }
357
358 /* This is the error handling routine for processes that are allowed
359  * to sleep.
360  */
361 static int nfs4_do_handle_exception(struct nfs_server *server,
362                 int errorcode, struct nfs4_exception *exception)
363 {
364         struct nfs_client *clp = server->nfs_client;
365         struct nfs4_state *state = exception->state;
366         struct inode *inode = exception->inode;
367         int ret = errorcode;
368
369         exception->delay = 0;
370         exception->recovering = 0;
371         exception->retry = 0;
372         switch(errorcode) {
373                 case 0:
374                         return 0;
375                 case -NFS4ERR_OPENMODE:
376                 case -NFS4ERR_DELEG_REVOKED:
377                 case -NFS4ERR_ADMIN_REVOKED:
378                 case -NFS4ERR_BAD_STATEID:
379                         if (inode && nfs_async_inode_return_delegation(inode,
380                                                 NULL) == 0)
381                                 goto wait_on_recovery;
382                         if (state == NULL)
383                                 break;
384                         ret = nfs4_schedule_stateid_recovery(server, state);
385                         if (ret < 0)
386                                 break;
387                         goto wait_on_recovery;
388                 case -NFS4ERR_EXPIRED:
389                         if (state != NULL) {
390                                 ret = nfs4_schedule_stateid_recovery(server, state);
391                                 if (ret < 0)
392                                         break;
393                         }
394                 case -NFS4ERR_STALE_STATEID:
395                 case -NFS4ERR_STALE_CLIENTID:
396                         nfs4_schedule_lease_recovery(clp);
397                         goto wait_on_recovery;
398                 case -NFS4ERR_MOVED:
399                         ret = nfs4_schedule_migration_recovery(server);
400                         if (ret < 0)
401                                 break;
402                         goto wait_on_recovery;
403                 case -NFS4ERR_LEASE_MOVED:
404                         nfs4_schedule_lease_moved_recovery(clp);
405                         goto wait_on_recovery;
406 #if defined(CONFIG_NFS_V4_1)
407                 case -NFS4ERR_BADSESSION:
408                 case -NFS4ERR_BADSLOT:
409                 case -NFS4ERR_BAD_HIGH_SLOT:
410                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
411                 case -NFS4ERR_DEADSESSION:
412                 case -NFS4ERR_SEQ_FALSE_RETRY:
413                 case -NFS4ERR_SEQ_MISORDERED:
414                         dprintk("%s ERROR: %d Reset session\n", __func__,
415                                 errorcode);
416                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
417                         goto wait_on_recovery;
418 #endif /* defined(CONFIG_NFS_V4_1) */
419                 case -NFS4ERR_FILE_OPEN:
420                         if (exception->timeout > HZ) {
421                                 /* We have retried a decent amount, time to
422                                  * fail
423                                  */
424                                 ret = -EBUSY;
425                                 break;
426                         }
427                 case -NFS4ERR_DELAY:
428                         nfs_inc_server_stats(server, NFSIOS_DELAY);
429                 case -NFS4ERR_GRACE:
430                 case -NFS4ERR_RECALLCONFLICT:
431                         exception->delay = 1;
432                         return 0;
433
434                 case -NFS4ERR_RETRY_UNCACHED_REP:
435                 case -NFS4ERR_OLD_STATEID:
436                         exception->retry = 1;
437                         break;
438                 case -NFS4ERR_BADOWNER:
439                         /* The following works around a Linux server bug! */
440                 case -NFS4ERR_BADNAME:
441                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
442                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
443                                 exception->retry = 1;
444                                 printk(KERN_WARNING "NFS: v4 server %s "
445                                                 "does not accept raw "
446                                                 "uid/gids. "
447                                                 "Reenabling the idmapper.\n",
448                                                 server->nfs_client->cl_hostname);
449                         }
450         }
451         /* We failed to handle the error */
452         return nfs4_map_errors(ret);
453 wait_on_recovery:
454         exception->recovering = 1;
455         return 0;
456 }
457
458 /* This is the error handling routine for processes that are allowed
459  * to sleep.
460  */
461 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
462 {
463         struct nfs_client *clp = server->nfs_client;
464         int ret;
465
466         ret = nfs4_do_handle_exception(server, errorcode, exception);
467         if (exception->delay) {
468                 ret = nfs4_delay(server->client, &exception->timeout);
469                 goto out_retry;
470         }
471         if (exception->recovering) {
472                 ret = nfs4_wait_clnt_recover(clp);
473                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
474                         return -EIO;
475                 goto out_retry;
476         }
477         return ret;
478 out_retry:
479         if (ret == 0)
480                 exception->retry = 1;
481         return ret;
482 }
483
484 static int
485 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
486                 int errorcode, struct nfs4_exception *exception)
487 {
488         struct nfs_client *clp = server->nfs_client;
489         int ret;
490
491         ret = nfs4_do_handle_exception(server, errorcode, exception);
492         if (exception->delay) {
493                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
494                 goto out_retry;
495         }
496         if (exception->recovering) {
497                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
498                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
499                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
500                 goto out_retry;
501         }
502         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
503                 ret = -EIO;
504         return ret;
505 out_retry:
506         if (ret == 0)
507                 exception->retry = 1;
508         return ret;
509 }
510
511 static int
512 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
513                         struct nfs4_state *state, long *timeout)
514 {
515         struct nfs4_exception exception = {
516                 .state = state,
517         };
518
519         if (task->tk_status >= 0)
520                 return 0;
521         if (timeout)
522                 exception.timeout = *timeout;
523         task->tk_status = nfs4_async_handle_exception(task, server,
524                         task->tk_status,
525                         &exception);
526         if (exception.delay && timeout)
527                 *timeout = exception.timeout;
528         if (exception.retry)
529                 return -EAGAIN;
530         return 0;
531 }
532
533 /*
534  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
535  * or 'false' otherwise.
536  */
537 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
538 {
539         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
540
541         if (flavor == RPC_AUTH_GSS_KRB5I ||
542             flavor == RPC_AUTH_GSS_KRB5P)
543                 return true;
544
545         return false;
546 }
547
548 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
549 {
550         spin_lock(&clp->cl_lock);
551         if (time_before(clp->cl_last_renewal,timestamp))
552                 clp->cl_last_renewal = timestamp;
553         spin_unlock(&clp->cl_lock);
554 }
555
556 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
557 {
558         struct nfs_client *clp = server->nfs_client;
559
560         if (!nfs4_has_session(clp))
561                 do_renew_lease(clp, timestamp);
562 }
563
564 struct nfs4_call_sync_data {
565         const struct nfs_server *seq_server;
566         struct nfs4_sequence_args *seq_args;
567         struct nfs4_sequence_res *seq_res;
568 };
569
570 void nfs4_init_sequence(struct nfs4_sequence_args *args,
571                         struct nfs4_sequence_res *res, int cache_reply)
572 {
573         args->sa_slot = NULL;
574         args->sa_cache_this = cache_reply;
575         args->sa_privileged = 0;
576
577         res->sr_slot = NULL;
578 }
579
580 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
581 {
582         args->sa_privileged = 1;
583 }
584
585 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
586                          struct nfs4_sequence_args *args,
587                          struct nfs4_sequence_res *res,
588                          struct rpc_task *task)
589 {
590         struct nfs4_slot *slot;
591
592         /* slot already allocated? */
593         if (res->sr_slot != NULL)
594                 goto out_start;
595
596         spin_lock(&tbl->slot_tbl_lock);
597         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
598                 goto out_sleep;
599
600         slot = nfs4_alloc_slot(tbl);
601         if (IS_ERR(slot)) {
602                 if (slot == ERR_PTR(-ENOMEM))
603                         task->tk_timeout = HZ >> 2;
604                 goto out_sleep;
605         }
606         spin_unlock(&tbl->slot_tbl_lock);
607
608         args->sa_slot = slot;
609         res->sr_slot = slot;
610
611 out_start:
612         rpc_call_start(task);
613         return 0;
614
615 out_sleep:
616         if (args->sa_privileged)
617                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
618                                 NULL, RPC_PRIORITY_PRIVILEGED);
619         else
620                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
621         spin_unlock(&tbl->slot_tbl_lock);
622         return -EAGAIN;
623 }
624 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
625
626 static int nfs40_sequence_done(struct rpc_task *task,
627                                struct nfs4_sequence_res *res)
628 {
629         struct nfs4_slot *slot = res->sr_slot;
630         struct nfs4_slot_table *tbl;
631
632         if (slot == NULL)
633                 goto out;
634
635         tbl = slot->table;
636         spin_lock(&tbl->slot_tbl_lock);
637         if (!nfs41_wake_and_assign_slot(tbl, slot))
638                 nfs4_free_slot(tbl, slot);
639         spin_unlock(&tbl->slot_tbl_lock);
640
641         res->sr_slot = NULL;
642 out:
643         return 1;
644 }
645
646 #if defined(CONFIG_NFS_V4_1)
647
648 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
649 {
650         struct nfs4_session *session;
651         struct nfs4_slot_table *tbl;
652         struct nfs4_slot *slot = res->sr_slot;
653         bool send_new_highest_used_slotid = false;
654
655         tbl = slot->table;
656         session = tbl->session;
657
658         spin_lock(&tbl->slot_tbl_lock);
659         /* Be nice to the server: try to ensure that the last transmitted
660          * value for highest_user_slotid <= target_highest_slotid
661          */
662         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
663                 send_new_highest_used_slotid = true;
664
665         if (nfs41_wake_and_assign_slot(tbl, slot)) {
666                 send_new_highest_used_slotid = false;
667                 goto out_unlock;
668         }
669         nfs4_free_slot(tbl, slot);
670
671         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
672                 send_new_highest_used_slotid = false;
673 out_unlock:
674         spin_unlock(&tbl->slot_tbl_lock);
675         res->sr_slot = NULL;
676         if (send_new_highest_used_slotid)
677                 nfs41_notify_server(session->clp);
678 }
679
680 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
681 {
682         struct nfs4_session *session;
683         struct nfs4_slot *slot = res->sr_slot;
684         struct nfs_client *clp;
685         bool interrupted = false;
686         int ret = 1;
687
688         if (slot == NULL)
689                 goto out_noaction;
690         /* don't increment the sequence number if the task wasn't sent */
691         if (!RPC_WAS_SENT(task))
692                 goto out;
693
694         session = slot->table->session;
695
696         if (slot->interrupted) {
697                 slot->interrupted = 0;
698                 interrupted = true;
699         }
700
701         trace_nfs4_sequence_done(session, res);
702         /* Check the SEQUENCE operation status */
703         switch (res->sr_status) {
704         case 0:
705                 /* Update the slot's sequence and clientid lease timer */
706                 ++slot->seq_nr;
707                 clp = session->clp;
708                 do_renew_lease(clp, res->sr_timestamp);
709                 /* Check sequence flags */
710                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
711                 nfs41_update_target_slotid(slot->table, slot, res);
712                 break;
713         case 1:
714                 /*
715                  * sr_status remains 1 if an RPC level error occurred.
716                  * The server may or may not have processed the sequence
717                  * operation..
718                  * Mark the slot as having hosted an interrupted RPC call.
719                  */
720                 slot->interrupted = 1;
721                 goto out;
722         case -NFS4ERR_DELAY:
723                 /* The server detected a resend of the RPC call and
724                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
725                  * of RFC5661.
726                  */
727                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
728                         __func__,
729                         slot->slot_nr,
730                         slot->seq_nr);
731                 goto out_retry;
732         case -NFS4ERR_BADSLOT:
733                 /*
734                  * The slot id we used was probably retired. Try again
735                  * using a different slot id.
736                  */
737                 goto retry_nowait;
738         case -NFS4ERR_SEQ_MISORDERED:
739                 /*
740                  * Was the last operation on this sequence interrupted?
741                  * If so, retry after bumping the sequence number.
742                  */
743                 if (interrupted) {
744                         ++slot->seq_nr;
745                         goto retry_nowait;
746                 }
747                 /*
748                  * Could this slot have been previously retired?
749                  * If so, then the server may be expecting seq_nr = 1!
750                  */
751                 if (slot->seq_nr != 1) {
752                         slot->seq_nr = 1;
753                         goto retry_nowait;
754                 }
755                 break;
756         case -NFS4ERR_SEQ_FALSE_RETRY:
757                 ++slot->seq_nr;
758                 goto retry_nowait;
759         default:
760                 /* Just update the slot sequence no. */
761                 ++slot->seq_nr;
762         }
763 out:
764         /* The session may be reset by one of the error handlers. */
765         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
766         nfs41_sequence_free_slot(res);
767 out_noaction:
768         return ret;
769 retry_nowait:
770         if (rpc_restart_call_prepare(task)) {
771                 task->tk_status = 0;
772                 ret = 0;
773         }
774         goto out;
775 out_retry:
776         if (!rpc_restart_call(task))
777                 goto out;
778         rpc_delay(task, NFS4_POLL_RETRY_MAX);
779         return 0;
780 }
781 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
782
783 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
784 {
785         if (res->sr_slot == NULL)
786                 return 1;
787         if (!res->sr_slot->table->session)
788                 return nfs40_sequence_done(task, res);
789         return nfs41_sequence_done(task, res);
790 }
791 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
792
793 int nfs41_setup_sequence(struct nfs4_session *session,
794                                 struct nfs4_sequence_args *args,
795                                 struct nfs4_sequence_res *res,
796                                 struct rpc_task *task)
797 {
798         struct nfs4_slot *slot;
799         struct nfs4_slot_table *tbl;
800
801         dprintk("--> %s\n", __func__);
802         /* slot already allocated? */
803         if (res->sr_slot != NULL)
804                 goto out_success;
805
806         tbl = &session->fc_slot_table;
807
808         task->tk_timeout = 0;
809
810         spin_lock(&tbl->slot_tbl_lock);
811         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
812             !args->sa_privileged) {
813                 /* The state manager will wait until the slot table is empty */
814                 dprintk("%s session is draining\n", __func__);
815                 goto out_sleep;
816         }
817
818         slot = nfs4_alloc_slot(tbl);
819         if (IS_ERR(slot)) {
820                 /* If out of memory, try again in 1/4 second */
821                 if (slot == ERR_PTR(-ENOMEM))
822                         task->tk_timeout = HZ >> 2;
823                 dprintk("<-- %s: no free slots\n", __func__);
824                 goto out_sleep;
825         }
826         spin_unlock(&tbl->slot_tbl_lock);
827
828         args->sa_slot = slot;
829
830         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
831                         slot->slot_nr, slot->seq_nr);
832
833         res->sr_slot = slot;
834         res->sr_timestamp = jiffies;
835         res->sr_status_flags = 0;
836         /*
837          * sr_status is only set in decode_sequence, and so will remain
838          * set to 1 if an rpc level failure occurs.
839          */
840         res->sr_status = 1;
841         trace_nfs4_setup_sequence(session, args);
842 out_success:
843         rpc_call_start(task);
844         return 0;
845 out_sleep:
846         /* Privileged tasks are queued with top priority */
847         if (args->sa_privileged)
848                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
849                                 NULL, RPC_PRIORITY_PRIVILEGED);
850         else
851                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
852         spin_unlock(&tbl->slot_tbl_lock);
853         return -EAGAIN;
854 }
855 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
856
857 static int nfs4_setup_sequence(const struct nfs_server *server,
858                                struct nfs4_sequence_args *args,
859                                struct nfs4_sequence_res *res,
860                                struct rpc_task *task)
861 {
862         struct nfs4_session *session = nfs4_get_session(server);
863         int ret = 0;
864
865         if (!session)
866                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
867                                             args, res, task);
868
869         dprintk("--> %s clp %p session %p sr_slot %u\n",
870                 __func__, session->clp, session, res->sr_slot ?
871                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
872
873         ret = nfs41_setup_sequence(session, args, res, task);
874
875         dprintk("<-- %s status=%d\n", __func__, ret);
876         return ret;
877 }
878
879 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
880 {
881         struct nfs4_call_sync_data *data = calldata;
882         struct nfs4_session *session = nfs4_get_session(data->seq_server);
883
884         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
885
886         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
887 }
888
889 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
890 {
891         struct nfs4_call_sync_data *data = calldata;
892
893         nfs41_sequence_done(task, data->seq_res);
894 }
895
896 static const struct rpc_call_ops nfs41_call_sync_ops = {
897         .rpc_call_prepare = nfs41_call_sync_prepare,
898         .rpc_call_done = nfs41_call_sync_done,
899 };
900
901 #else   /* !CONFIG_NFS_V4_1 */
902
903 static int nfs4_setup_sequence(const struct nfs_server *server,
904                                struct nfs4_sequence_args *args,
905                                struct nfs4_sequence_res *res,
906                                struct rpc_task *task)
907 {
908         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
909                                     args, res, task);
910 }
911
912 int nfs4_sequence_done(struct rpc_task *task,
913                        struct nfs4_sequence_res *res)
914 {
915         return nfs40_sequence_done(task, res);
916 }
917 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
918
919 #endif  /* !CONFIG_NFS_V4_1 */
920
921 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
922 {
923         struct nfs4_call_sync_data *data = calldata;
924         nfs4_setup_sequence(data->seq_server,
925                                 data->seq_args, data->seq_res, task);
926 }
927
928 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
929 {
930         struct nfs4_call_sync_data *data = calldata;
931         nfs4_sequence_done(task, data->seq_res);
932 }
933
934 static const struct rpc_call_ops nfs40_call_sync_ops = {
935         .rpc_call_prepare = nfs40_call_sync_prepare,
936         .rpc_call_done = nfs40_call_sync_done,
937 };
938
939 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
940                                    struct nfs_server *server,
941                                    struct rpc_message *msg,
942                                    struct nfs4_sequence_args *args,
943                                    struct nfs4_sequence_res *res)
944 {
945         int ret;
946         struct rpc_task *task;
947         struct nfs_client *clp = server->nfs_client;
948         struct nfs4_call_sync_data data = {
949                 .seq_server = server,
950                 .seq_args = args,
951                 .seq_res = res,
952         };
953         struct rpc_task_setup task_setup = {
954                 .rpc_client = clnt,
955                 .rpc_message = msg,
956                 .callback_ops = clp->cl_mvops->call_sync_ops,
957                 .callback_data = &data
958         };
959
960         task = rpc_run_task(&task_setup);
961         if (IS_ERR(task))
962                 ret = PTR_ERR(task);
963         else {
964                 ret = task->tk_status;
965                 rpc_put_task(task);
966         }
967         return ret;
968 }
969
970 int nfs4_call_sync(struct rpc_clnt *clnt,
971                    struct nfs_server *server,
972                    struct rpc_message *msg,
973                    struct nfs4_sequence_args *args,
974                    struct nfs4_sequence_res *res,
975                    int cache_reply)
976 {
977         nfs4_init_sequence(args, res, cache_reply);
978         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
979 }
980
981 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
982 {
983         struct nfs_inode *nfsi = NFS_I(dir);
984
985         spin_lock(&dir->i_lock);
986         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
987         if (!cinfo->atomic || cinfo->before != dir->i_version)
988                 nfs_force_lookup_revalidate(dir);
989         dir->i_version = cinfo->after;
990         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
991         nfs_fscache_invalidate(dir);
992         spin_unlock(&dir->i_lock);
993 }
994
995 struct nfs4_opendata {
996         struct kref kref;
997         struct nfs_openargs o_arg;
998         struct nfs_openres o_res;
999         struct nfs_open_confirmargs c_arg;
1000         struct nfs_open_confirmres c_res;
1001         struct nfs4_string owner_name;
1002         struct nfs4_string group_name;
1003         struct nfs4_label *a_label;
1004         struct nfs_fattr f_attr;
1005         struct nfs4_label *f_label;
1006         struct dentry *dir;
1007         struct dentry *dentry;
1008         struct nfs4_state_owner *owner;
1009         struct nfs4_state *state;
1010         struct iattr attrs;
1011         unsigned long timestamp;
1012         unsigned int rpc_done : 1;
1013         unsigned int file_created : 1;
1014         unsigned int is_recover : 1;
1015         int rpc_status;
1016         int cancelled;
1017 };
1018
1019 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1020                 int err, struct nfs4_exception *exception)
1021 {
1022         if (err != -EINVAL)
1023                 return false;
1024         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1025                 return false;
1026         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1027         exception->retry = 1;
1028         return true;
1029 }
1030
1031 static u32
1032 nfs4_map_atomic_open_share(struct nfs_server *server,
1033                 fmode_t fmode, int openflags)
1034 {
1035         u32 res = 0;
1036
1037         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1038         case FMODE_READ:
1039                 res = NFS4_SHARE_ACCESS_READ;
1040                 break;
1041         case FMODE_WRITE:
1042                 res = NFS4_SHARE_ACCESS_WRITE;
1043                 break;
1044         case FMODE_READ|FMODE_WRITE:
1045                 res = NFS4_SHARE_ACCESS_BOTH;
1046         }
1047         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1048                 goto out;
1049         /* Want no delegation if we're using O_DIRECT */
1050         if (openflags & O_DIRECT)
1051                 res |= NFS4_SHARE_WANT_NO_DELEG;
1052 out:
1053         return res;
1054 }
1055
1056 static enum open_claim_type4
1057 nfs4_map_atomic_open_claim(struct nfs_server *server,
1058                 enum open_claim_type4 claim)
1059 {
1060         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1061                 return claim;
1062         switch (claim) {
1063         default:
1064                 return claim;
1065         case NFS4_OPEN_CLAIM_FH:
1066                 return NFS4_OPEN_CLAIM_NULL;
1067         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1068                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1069         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1070                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1071         }
1072 }
1073
1074 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1075 {
1076         p->o_res.f_attr = &p->f_attr;
1077         p->o_res.f_label = p->f_label;
1078         p->o_res.seqid = p->o_arg.seqid;
1079         p->c_res.seqid = p->c_arg.seqid;
1080         p->o_res.server = p->o_arg.server;
1081         p->o_res.access_request = p->o_arg.access;
1082         nfs_fattr_init(&p->f_attr);
1083         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1084 }
1085
1086 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1087                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1088                 const struct iattr *attrs,
1089                 struct nfs4_label *label,
1090                 enum open_claim_type4 claim,
1091                 gfp_t gfp_mask)
1092 {
1093         struct dentry *parent = dget_parent(dentry);
1094         struct inode *dir = d_inode(parent);
1095         struct nfs_server *server = NFS_SERVER(dir);
1096         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1097         struct nfs4_opendata *p;
1098
1099         p = kzalloc(sizeof(*p), gfp_mask);
1100         if (p == NULL)
1101                 goto err;
1102
1103         p->f_label = nfs4_label_alloc(server, gfp_mask);
1104         if (IS_ERR(p->f_label))
1105                 goto err_free_p;
1106
1107         p->a_label = nfs4_label_alloc(server, gfp_mask);
1108         if (IS_ERR(p->a_label))
1109                 goto err_free_f;
1110
1111         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1112         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1113         if (IS_ERR(p->o_arg.seqid))
1114                 goto err_free_label;
1115         nfs_sb_active(dentry->d_sb);
1116         p->dentry = dget(dentry);
1117         p->dir = parent;
1118         p->owner = sp;
1119         atomic_inc(&sp->so_count);
1120         p->o_arg.open_flags = flags;
1121         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1122         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1123                         fmode, flags);
1124         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1125          * will return permission denied for all bits until close */
1126         if (!(flags & O_EXCL)) {
1127                 /* ask server to check for all possible rights as results
1128                  * are cached */
1129                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1130                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1131         }
1132         p->o_arg.clientid = server->nfs_client->cl_clientid;
1133         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1134         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1135         p->o_arg.name = &dentry->d_name;
1136         p->o_arg.server = server;
1137         p->o_arg.bitmask = nfs4_bitmask(server, label);
1138         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1139         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1140         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1141         switch (p->o_arg.claim) {
1142         case NFS4_OPEN_CLAIM_NULL:
1143         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1144         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1145                 p->o_arg.fh = NFS_FH(dir);
1146                 break;
1147         case NFS4_OPEN_CLAIM_PREVIOUS:
1148         case NFS4_OPEN_CLAIM_FH:
1149         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1150         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1151                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1152         }
1153         if (attrs != NULL && attrs->ia_valid != 0) {
1154                 __u32 verf[2];
1155
1156                 p->o_arg.u.attrs = &p->attrs;
1157                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1158
1159                 verf[0] = jiffies;
1160                 verf[1] = current->pid;
1161                 memcpy(p->o_arg.u.verifier.data, verf,
1162                                 sizeof(p->o_arg.u.verifier.data));
1163         }
1164         p->c_arg.fh = &p->o_res.fh;
1165         p->c_arg.stateid = &p->o_res.stateid;
1166         p->c_arg.seqid = p->o_arg.seqid;
1167         nfs4_init_opendata_res(p);
1168         kref_init(&p->kref);
1169         return p;
1170
1171 err_free_label:
1172         nfs4_label_free(p->a_label);
1173 err_free_f:
1174         nfs4_label_free(p->f_label);
1175 err_free_p:
1176         kfree(p);
1177 err:
1178         dput(parent);
1179         return NULL;
1180 }
1181
1182 static void nfs4_opendata_free(struct kref *kref)
1183 {
1184         struct nfs4_opendata *p = container_of(kref,
1185                         struct nfs4_opendata, kref);
1186         struct super_block *sb = p->dentry->d_sb;
1187
1188         nfs_free_seqid(p->o_arg.seqid);
1189         if (p->state != NULL)
1190                 nfs4_put_open_state(p->state);
1191         nfs4_put_state_owner(p->owner);
1192
1193         nfs4_label_free(p->a_label);
1194         nfs4_label_free(p->f_label);
1195
1196         dput(p->dir);
1197         dput(p->dentry);
1198         nfs_sb_deactive(sb);
1199         nfs_fattr_free_names(&p->f_attr);
1200         kfree(p->f_attr.mdsthreshold);
1201         kfree(p);
1202 }
1203
1204 static void nfs4_opendata_put(struct nfs4_opendata *p)
1205 {
1206         if (p != NULL)
1207                 kref_put(&p->kref, nfs4_opendata_free);
1208 }
1209
1210 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1211 {
1212         int ret;
1213
1214         ret = rpc_wait_for_completion_task(task);
1215         return ret;
1216 }
1217
1218 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1219                 fmode_t fmode)
1220 {
1221         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1222         case FMODE_READ|FMODE_WRITE:
1223                 return state->n_rdwr != 0;
1224         case FMODE_WRITE:
1225                 return state->n_wronly != 0;
1226         case FMODE_READ:
1227                 return state->n_rdonly != 0;
1228         }
1229         WARN_ON_ONCE(1);
1230         return false;
1231 }
1232
1233 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1234 {
1235         int ret = 0;
1236
1237         if (open_mode & (O_EXCL|O_TRUNC))
1238                 goto out;
1239         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1240                 case FMODE_READ:
1241                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1242                                 && state->n_rdonly != 0;
1243                         break;
1244                 case FMODE_WRITE:
1245                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1246                                 && state->n_wronly != 0;
1247                         break;
1248                 case FMODE_READ|FMODE_WRITE:
1249                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1250                                 && state->n_rdwr != 0;
1251         }
1252 out:
1253         return ret;
1254 }
1255
1256 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1257                 enum open_claim_type4 claim)
1258 {
1259         if (delegation == NULL)
1260                 return 0;
1261         if ((delegation->type & fmode) != fmode)
1262                 return 0;
1263         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1264                 return 0;
1265         switch (claim) {
1266         case NFS4_OPEN_CLAIM_NULL:
1267         case NFS4_OPEN_CLAIM_FH:
1268                 break;
1269         case NFS4_OPEN_CLAIM_PREVIOUS:
1270                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1271                         break;
1272         default:
1273                 return 0;
1274         }
1275         nfs_mark_delegation_referenced(delegation);
1276         return 1;
1277 }
1278
1279 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1280 {
1281         switch (fmode) {
1282                 case FMODE_WRITE:
1283                         state->n_wronly++;
1284                         break;
1285                 case FMODE_READ:
1286                         state->n_rdonly++;
1287                         break;
1288                 case FMODE_READ|FMODE_WRITE:
1289                         state->n_rdwr++;
1290         }
1291         nfs4_state_set_mode_locked(state, state->state | fmode);
1292 }
1293
1294 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1295 {
1296         struct nfs_client *clp = state->owner->so_server->nfs_client;
1297         bool need_recover = false;
1298
1299         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1300                 need_recover = true;
1301         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1302                 need_recover = true;
1303         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1304                 need_recover = true;
1305         if (need_recover)
1306                 nfs4_state_mark_reclaim_nograce(clp, state);
1307 }
1308
1309 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1310                 nfs4_stateid *stateid)
1311 {
1312         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1313                 return true;
1314         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1315                 nfs_test_and_clear_all_open_stateid(state);
1316                 return true;
1317         }
1318         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1319                 return true;
1320         return false;
1321 }
1322
1323 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1324 {
1325         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1326                 return;
1327         if (state->n_wronly)
1328                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1329         if (state->n_rdonly)
1330                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1331         if (state->n_rdwr)
1332                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1333         set_bit(NFS_OPEN_STATE, &state->flags);
1334 }
1335
1336 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1337                 nfs4_stateid *arg_stateid,
1338                 nfs4_stateid *stateid, fmode_t fmode)
1339 {
1340         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1341         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1342         case FMODE_WRITE:
1343                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1344                 break;
1345         case FMODE_READ:
1346                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1347                 break;
1348         case 0:
1349                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1350                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1351                 clear_bit(NFS_OPEN_STATE, &state->flags);
1352         }
1353         if (stateid == NULL)
1354                 return;
1355         /* Handle races with OPEN */
1356         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1357             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1358             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1359                 nfs_resync_open_stateid_locked(state);
1360                 return;
1361         }
1362         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1363                 nfs4_stateid_copy(&state->stateid, stateid);
1364         nfs4_stateid_copy(&state->open_stateid, stateid);
1365 }
1366
1367 static void nfs_clear_open_stateid(struct nfs4_state *state,
1368         nfs4_stateid *arg_stateid,
1369         nfs4_stateid *stateid, fmode_t fmode)
1370 {
1371         write_seqlock(&state->seqlock);
1372         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1373         write_sequnlock(&state->seqlock);
1374         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1375                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1376 }
1377
1378 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1379 {
1380         switch (fmode) {
1381                 case FMODE_READ:
1382                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1383                         break;
1384                 case FMODE_WRITE:
1385                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1386                         break;
1387                 case FMODE_READ|FMODE_WRITE:
1388                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1389         }
1390         if (!nfs_need_update_open_stateid(state, stateid))
1391                 return;
1392         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1393                 nfs4_stateid_copy(&state->stateid, stateid);
1394         nfs4_stateid_copy(&state->open_stateid, stateid);
1395 }
1396
1397 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1398 {
1399         /*
1400          * Protect the call to nfs4_state_set_mode_locked and
1401          * serialise the stateid update
1402          */
1403         spin_lock(&state->owner->so_lock);
1404         write_seqlock(&state->seqlock);
1405         if (deleg_stateid != NULL) {
1406                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1407                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1408         }
1409         if (open_stateid != NULL)
1410                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1411         write_sequnlock(&state->seqlock);
1412         update_open_stateflags(state, fmode);
1413         spin_unlock(&state->owner->so_lock);
1414 }
1415
1416 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1417 {
1418         struct nfs_inode *nfsi = NFS_I(state->inode);
1419         struct nfs_delegation *deleg_cur;
1420         int ret = 0;
1421
1422         fmode &= (FMODE_READ|FMODE_WRITE);
1423
1424         rcu_read_lock();
1425         deleg_cur = rcu_dereference(nfsi->delegation);
1426         if (deleg_cur == NULL)
1427                 goto no_delegation;
1428
1429         spin_lock(&deleg_cur->lock);
1430         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1431            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1432             (deleg_cur->type & fmode) != fmode)
1433                 goto no_delegation_unlock;
1434
1435         if (delegation == NULL)
1436                 delegation = &deleg_cur->stateid;
1437         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1438                 goto no_delegation_unlock;
1439
1440         nfs_mark_delegation_referenced(deleg_cur);
1441         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1442         ret = 1;
1443 no_delegation_unlock:
1444         spin_unlock(&deleg_cur->lock);
1445 no_delegation:
1446         rcu_read_unlock();
1447
1448         if (!ret && open_stateid != NULL) {
1449                 __update_open_stateid(state, open_stateid, NULL, fmode);
1450                 ret = 1;
1451         }
1452         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1453                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1454
1455         return ret;
1456 }
1457
1458 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1459                 const nfs4_stateid *stateid)
1460 {
1461         struct nfs4_state *state = lsp->ls_state;
1462         bool ret = false;
1463
1464         spin_lock(&state->state_lock);
1465         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1466                 goto out_noupdate;
1467         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1468                 goto out_noupdate;
1469         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1470         ret = true;
1471 out_noupdate:
1472         spin_unlock(&state->state_lock);
1473         return ret;
1474 }
1475
1476 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1477 {
1478         struct nfs_delegation *delegation;
1479
1480         rcu_read_lock();
1481         delegation = rcu_dereference(NFS_I(inode)->delegation);
1482         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1483                 rcu_read_unlock();
1484                 return;
1485         }
1486         rcu_read_unlock();
1487         nfs4_inode_return_delegation(inode);
1488 }
1489
1490 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1491 {
1492         struct nfs4_state *state = opendata->state;
1493         struct nfs_inode *nfsi = NFS_I(state->inode);
1494         struct nfs_delegation *delegation;
1495         int open_mode = opendata->o_arg.open_flags;
1496         fmode_t fmode = opendata->o_arg.fmode;
1497         enum open_claim_type4 claim = opendata->o_arg.claim;
1498         nfs4_stateid stateid;
1499         int ret = -EAGAIN;
1500
1501         for (;;) {
1502                 spin_lock(&state->owner->so_lock);
1503                 if (can_open_cached(state, fmode, open_mode)) {
1504                         update_open_stateflags(state, fmode);
1505                         spin_unlock(&state->owner->so_lock);
1506                         goto out_return_state;
1507                 }
1508                 spin_unlock(&state->owner->so_lock);
1509                 rcu_read_lock();
1510                 delegation = rcu_dereference(nfsi->delegation);
1511                 if (!can_open_delegated(delegation, fmode, claim)) {
1512                         rcu_read_unlock();
1513                         break;
1514                 }
1515                 /* Save the delegation */
1516                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1517                 rcu_read_unlock();
1518                 nfs_release_seqid(opendata->o_arg.seqid);
1519                 if (!opendata->is_recover) {
1520                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1521                         if (ret != 0)
1522                                 goto out;
1523                 }
1524                 ret = -EAGAIN;
1525
1526                 /* Try to update the stateid using the delegation */
1527                 if (update_open_stateid(state, NULL, &stateid, fmode))
1528                         goto out_return_state;
1529         }
1530 out:
1531         return ERR_PTR(ret);
1532 out_return_state:
1533         atomic_inc(&state->count);
1534         return state;
1535 }
1536
1537 static void
1538 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1539 {
1540         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1541         struct nfs_delegation *delegation;
1542         int delegation_flags = 0;
1543
1544         rcu_read_lock();
1545         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1546         if (delegation)
1547                 delegation_flags = delegation->flags;
1548         rcu_read_unlock();
1549         switch (data->o_arg.claim) {
1550         default:
1551                 break;
1552         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1553         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1554                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1555                                    "returning a delegation for "
1556                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1557                                    clp->cl_hostname);
1558                 return;
1559         }
1560         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1561                 nfs_inode_set_delegation(state->inode,
1562                                          data->owner->so_cred,
1563                                          &data->o_res);
1564         else
1565                 nfs_inode_reclaim_delegation(state->inode,
1566                                              data->owner->so_cred,
1567                                              &data->o_res);
1568 }
1569
1570 /*
1571  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1572  * and update the nfs4_state.
1573  */
1574 static struct nfs4_state *
1575 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1576 {
1577         struct inode *inode = data->state->inode;
1578         struct nfs4_state *state = data->state;
1579         int ret;
1580
1581         if (!data->rpc_done) {
1582                 if (data->rpc_status) {
1583                         ret = data->rpc_status;
1584                         goto err;
1585                 }
1586                 /* cached opens have already been processed */
1587                 goto update;
1588         }
1589
1590         ret = nfs_refresh_inode(inode, &data->f_attr);
1591         if (ret)
1592                 goto err;
1593
1594         if (data->o_res.delegation_type != 0)
1595                 nfs4_opendata_check_deleg(data, state);
1596 update:
1597         update_open_stateid(state, &data->o_res.stateid, NULL,
1598                             data->o_arg.fmode);
1599         atomic_inc(&state->count);
1600
1601         return state;
1602 err:
1603         return ERR_PTR(ret);
1604
1605 }
1606
1607 static struct nfs4_state *
1608 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1609 {
1610         struct inode *inode;
1611         struct nfs4_state *state = NULL;
1612         int ret;
1613
1614         if (!data->rpc_done) {
1615                 state = nfs4_try_open_cached(data);
1616                 trace_nfs4_cached_open(data->state);
1617                 goto out;
1618         }
1619
1620         ret = -EAGAIN;
1621         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1622                 goto err;
1623         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1624         ret = PTR_ERR(inode);
1625         if (IS_ERR(inode))
1626                 goto err;
1627         ret = -ENOMEM;
1628         state = nfs4_get_open_state(inode, data->owner);
1629         if (state == NULL)
1630                 goto err_put_inode;
1631         if (data->o_res.delegation_type != 0)
1632                 nfs4_opendata_check_deleg(data, state);
1633         update_open_stateid(state, &data->o_res.stateid, NULL,
1634                         data->o_arg.fmode);
1635         iput(inode);
1636 out:
1637         nfs_release_seqid(data->o_arg.seqid);
1638         return state;
1639 err_put_inode:
1640         iput(inode);
1641 err:
1642         return ERR_PTR(ret);
1643 }
1644
1645 static struct nfs4_state *
1646 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1647 {
1648         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1649                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1650         return _nfs4_opendata_to_nfs4_state(data);
1651 }
1652
1653 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1654 {
1655         struct nfs_inode *nfsi = NFS_I(state->inode);
1656         struct nfs_open_context *ctx;
1657
1658         spin_lock(&state->inode->i_lock);
1659         list_for_each_entry(ctx, &nfsi->open_files, list) {
1660                 if (ctx->state != state)
1661                         continue;
1662                 get_nfs_open_context(ctx);
1663                 spin_unlock(&state->inode->i_lock);
1664                 return ctx;
1665         }
1666         spin_unlock(&state->inode->i_lock);
1667         return ERR_PTR(-ENOENT);
1668 }
1669
1670 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1671                 struct nfs4_state *state, enum open_claim_type4 claim)
1672 {
1673         struct nfs4_opendata *opendata;
1674
1675         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1676                         NULL, NULL, claim, GFP_NOFS);
1677         if (opendata == NULL)
1678                 return ERR_PTR(-ENOMEM);
1679         opendata->state = state;
1680         atomic_inc(&state->count);
1681         return opendata;
1682 }
1683
1684 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1685                 fmode_t fmode)
1686 {
1687         struct nfs4_state *newstate;
1688         int ret;
1689
1690         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1691                 return 0;
1692         opendata->o_arg.open_flags = 0;
1693         opendata->o_arg.fmode = fmode;
1694         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1695                         NFS_SB(opendata->dentry->d_sb),
1696                         fmode, 0);
1697         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1698         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1699         nfs4_init_opendata_res(opendata);
1700         ret = _nfs4_recover_proc_open(opendata);
1701         if (ret != 0)
1702                 return ret; 
1703         newstate = nfs4_opendata_to_nfs4_state(opendata);
1704         if (IS_ERR(newstate))
1705                 return PTR_ERR(newstate);
1706         if (newstate != opendata->state)
1707                 ret = -ESTALE;
1708         nfs4_close_state(newstate, fmode);
1709         return ret;
1710 }
1711
1712 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1713 {
1714         int ret;
1715
1716         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1717         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1718         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1719         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1720         /* memory barrier prior to reading state->n_* */
1721         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1722         clear_bit(NFS_OPEN_STATE, &state->flags);
1723         smp_rmb();
1724         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1725         if (ret != 0)
1726                 return ret;
1727         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1728         if (ret != 0)
1729                 return ret;
1730         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1731         if (ret != 0)
1732                 return ret;
1733         /*
1734          * We may have performed cached opens for all three recoveries.
1735          * Check if we need to update the current stateid.
1736          */
1737         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1738             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1739                 write_seqlock(&state->seqlock);
1740                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1741                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1742                 write_sequnlock(&state->seqlock);
1743         }
1744         return 0;
1745 }
1746
1747 /*
1748  * OPEN_RECLAIM:
1749  *      reclaim state on the server after a reboot.
1750  */
1751 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1752 {
1753         struct nfs_delegation *delegation;
1754         struct nfs4_opendata *opendata;
1755         fmode_t delegation_type = 0;
1756         int status;
1757
1758         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1759                         NFS4_OPEN_CLAIM_PREVIOUS);
1760         if (IS_ERR(opendata))
1761                 return PTR_ERR(opendata);
1762         rcu_read_lock();
1763         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1764         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1765                 delegation_type = delegation->type;
1766         rcu_read_unlock();
1767         opendata->o_arg.u.delegation_type = delegation_type;
1768         status = nfs4_open_recover(opendata, state);
1769         nfs4_opendata_put(opendata);
1770         return status;
1771 }
1772
1773 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1774 {
1775         struct nfs_server *server = NFS_SERVER(state->inode);
1776         struct nfs4_exception exception = { };
1777         int err;
1778         do {
1779                 err = _nfs4_do_open_reclaim(ctx, state);
1780                 trace_nfs4_open_reclaim(ctx, 0, err);
1781                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1782                         continue;
1783                 if (err != -NFS4ERR_DELAY)
1784                         break;
1785                 nfs4_handle_exception(server, err, &exception);
1786         } while (exception.retry);
1787         return err;
1788 }
1789
1790 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1791 {
1792         struct nfs_open_context *ctx;
1793         int ret;
1794
1795         ctx = nfs4_state_find_open_context(state);
1796         if (IS_ERR(ctx))
1797                 return -EAGAIN;
1798         ret = nfs4_do_open_reclaim(ctx, state);
1799         put_nfs_open_context(ctx);
1800         return ret;
1801 }
1802
1803 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1804 {
1805         switch (err) {
1806                 default:
1807                         printk(KERN_ERR "NFS: %s: unhandled error "
1808                                         "%d.\n", __func__, err);
1809                 case 0:
1810                 case -ENOENT:
1811                 case -EAGAIN:
1812                 case -ESTALE:
1813                         break;
1814                 case -NFS4ERR_BADSESSION:
1815                 case -NFS4ERR_BADSLOT:
1816                 case -NFS4ERR_BAD_HIGH_SLOT:
1817                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1818                 case -NFS4ERR_DEADSESSION:
1819                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1820                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1821                         return -EAGAIN;
1822                 case -NFS4ERR_STALE_CLIENTID:
1823                 case -NFS4ERR_STALE_STATEID:
1824                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1825                 case -NFS4ERR_EXPIRED:
1826                         /* Don't recall a delegation if it was lost */
1827                         nfs4_schedule_lease_recovery(server->nfs_client);
1828                         return -EAGAIN;
1829                 case -NFS4ERR_MOVED:
1830                         nfs4_schedule_migration_recovery(server);
1831                         return -EAGAIN;
1832                 case -NFS4ERR_LEASE_MOVED:
1833                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1834                         return -EAGAIN;
1835                 case -NFS4ERR_DELEG_REVOKED:
1836                 case -NFS4ERR_ADMIN_REVOKED:
1837                 case -NFS4ERR_BAD_STATEID:
1838                 case -NFS4ERR_OPENMODE:
1839                         nfs_inode_find_state_and_recover(state->inode,
1840                                         stateid);
1841                         nfs4_schedule_stateid_recovery(server, state);
1842                         return -EAGAIN;
1843                 case -NFS4ERR_DELAY:
1844                 case -NFS4ERR_GRACE:
1845                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1846                         ssleep(1);
1847                         return -EAGAIN;
1848                 case -ENOMEM:
1849                 case -NFS4ERR_DENIED:
1850                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1851                         return 0;
1852         }
1853         return err;
1854 }
1855
1856 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1857                 struct nfs4_state *state, const nfs4_stateid *stateid,
1858                 fmode_t type)
1859 {
1860         struct nfs_server *server = NFS_SERVER(state->inode);
1861         struct nfs4_opendata *opendata;
1862         int err = 0;
1863
1864         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1865                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1866         if (IS_ERR(opendata))
1867                 return PTR_ERR(opendata);
1868         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1869         write_seqlock(&state->seqlock);
1870         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1871         write_sequnlock(&state->seqlock);
1872         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1873         switch (type & (FMODE_READ|FMODE_WRITE)) {
1874         case FMODE_READ|FMODE_WRITE:
1875         case FMODE_WRITE:
1876                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1877                 if (err)
1878                         break;
1879                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1880                 if (err)
1881                         break;
1882         case FMODE_READ:
1883                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1884         }
1885         nfs4_opendata_put(opendata);
1886         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1887 }
1888
1889 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1890 {
1891         struct nfs4_opendata *data = calldata;
1892
1893         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1894                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1895 }
1896
1897 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1898 {
1899         struct nfs4_opendata *data = calldata;
1900
1901         nfs40_sequence_done(task, &data->c_res.seq_res);
1902
1903         data->rpc_status = task->tk_status;
1904         if (data->rpc_status == 0) {
1905                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1906                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1907                 renew_lease(data->o_res.server, data->timestamp);
1908                 data->rpc_done = 1;
1909         }
1910 }
1911
1912 static void nfs4_open_confirm_release(void *calldata)
1913 {
1914         struct nfs4_opendata *data = calldata;
1915         struct nfs4_state *state = NULL;
1916
1917         /* If this request hasn't been cancelled, do nothing */
1918         if (data->cancelled == 0)
1919                 goto out_free;
1920         /* In case of error, no cleanup! */
1921         if (!data->rpc_done)
1922                 goto out_free;
1923         state = nfs4_opendata_to_nfs4_state(data);
1924         if (!IS_ERR(state))
1925                 nfs4_close_state(state, data->o_arg.fmode);
1926 out_free:
1927         nfs4_opendata_put(data);
1928 }
1929
1930 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1931         .rpc_call_prepare = nfs4_open_confirm_prepare,
1932         .rpc_call_done = nfs4_open_confirm_done,
1933         .rpc_release = nfs4_open_confirm_release,
1934 };
1935
1936 /*
1937  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1938  */
1939 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1940 {
1941         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1942         struct rpc_task *task;
1943         struct  rpc_message msg = {
1944                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1945                 .rpc_argp = &data->c_arg,
1946                 .rpc_resp = &data->c_res,
1947                 .rpc_cred = data->owner->so_cred,
1948         };
1949         struct rpc_task_setup task_setup_data = {
1950                 .rpc_client = server->client,
1951                 .rpc_message = &msg,
1952                 .callback_ops = &nfs4_open_confirm_ops,
1953                 .callback_data = data,
1954                 .workqueue = nfsiod_workqueue,
1955                 .flags = RPC_TASK_ASYNC,
1956         };
1957         int status;
1958
1959         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1960         kref_get(&data->kref);
1961         data->rpc_done = 0;
1962         data->rpc_status = 0;
1963         data->timestamp = jiffies;
1964         if (data->is_recover)
1965                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1966         task = rpc_run_task(&task_setup_data);
1967         if (IS_ERR(task))
1968                 return PTR_ERR(task);
1969         status = nfs4_wait_for_completion_rpc_task(task);
1970         if (status != 0) {
1971                 data->cancelled = 1;
1972                 smp_wmb();
1973         } else
1974                 status = data->rpc_status;
1975         rpc_put_task(task);
1976         return status;
1977 }
1978
1979 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1980 {
1981         struct nfs4_opendata *data = calldata;
1982         struct nfs4_state_owner *sp = data->owner;
1983         struct nfs_client *clp = sp->so_server->nfs_client;
1984         enum open_claim_type4 claim = data->o_arg.claim;
1985
1986         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1987                 goto out_wait;
1988         /*
1989          * Check if we still need to send an OPEN call, or if we can use
1990          * a delegation instead.
1991          */
1992         if (data->state != NULL) {
1993                 struct nfs_delegation *delegation;
1994
1995                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1996                         goto out_no_action;
1997                 rcu_read_lock();
1998                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1999                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2000                         goto unlock_no_action;
2001                 rcu_read_unlock();
2002         }
2003         /* Update client id. */
2004         data->o_arg.clientid = clp->cl_clientid;
2005         switch (claim) {
2006         default:
2007                 break;
2008         case NFS4_OPEN_CLAIM_PREVIOUS:
2009         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2010         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2011                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2012         case NFS4_OPEN_CLAIM_FH:
2013                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2014                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2015         }
2016         data->timestamp = jiffies;
2017         if (nfs4_setup_sequence(data->o_arg.server,
2018                                 &data->o_arg.seq_args,
2019                                 &data->o_res.seq_res,
2020                                 task) != 0)
2021                 nfs_release_seqid(data->o_arg.seqid);
2022
2023         /* Set the create mode (note dependency on the session type) */
2024         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2025         if (data->o_arg.open_flags & O_EXCL) {
2026                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2027                 if (nfs4_has_persistent_session(clp))
2028                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2029                 else if (clp->cl_mvops->minor_version > 0)
2030                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2031         }
2032         return;
2033 unlock_no_action:
2034         trace_nfs4_cached_open(data->state);
2035         rcu_read_unlock();
2036 out_no_action:
2037         task->tk_action = NULL;
2038 out_wait:
2039         nfs4_sequence_done(task, &data->o_res.seq_res);
2040 }
2041
2042 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2043 {
2044         struct nfs4_opendata *data = calldata;
2045
2046         data->rpc_status = task->tk_status;
2047
2048         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2049                 return;
2050
2051         if (task->tk_status == 0) {
2052                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2053                         switch (data->o_res.f_attr->mode & S_IFMT) {
2054                         case S_IFREG:
2055                                 break;
2056                         case S_IFLNK:
2057                                 data->rpc_status = -ELOOP;
2058                                 break;
2059                         case S_IFDIR:
2060                                 data->rpc_status = -EISDIR;
2061                                 break;
2062                         default:
2063                                 data->rpc_status = -ENOTDIR;
2064                         }
2065                 }
2066                 renew_lease(data->o_res.server, data->timestamp);
2067                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2068                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2069         }
2070         data->rpc_done = 1;
2071 }
2072
2073 static void nfs4_open_release(void *calldata)
2074 {
2075         struct nfs4_opendata *data = calldata;
2076         struct nfs4_state *state = NULL;
2077
2078         /* If this request hasn't been cancelled, do nothing */
2079         if (data->cancelled == 0)
2080                 goto out_free;
2081         /* In case of error, no cleanup! */
2082         if (data->rpc_status != 0 || !data->rpc_done)
2083                 goto out_free;
2084         /* In case we need an open_confirm, no cleanup! */
2085         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2086                 goto out_free;
2087         state = nfs4_opendata_to_nfs4_state(data);
2088         if (!IS_ERR(state))
2089                 nfs4_close_state(state, data->o_arg.fmode);
2090 out_free:
2091         nfs4_opendata_put(data);
2092 }
2093
2094 static const struct rpc_call_ops nfs4_open_ops = {
2095         .rpc_call_prepare = nfs4_open_prepare,
2096         .rpc_call_done = nfs4_open_done,
2097         .rpc_release = nfs4_open_release,
2098 };
2099
2100 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2101 {
2102         struct inode *dir = d_inode(data->dir);
2103         struct nfs_server *server = NFS_SERVER(dir);
2104         struct nfs_openargs *o_arg = &data->o_arg;
2105         struct nfs_openres *o_res = &data->o_res;
2106         struct rpc_task *task;
2107         struct rpc_message msg = {
2108                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2109                 .rpc_argp = o_arg,
2110                 .rpc_resp = o_res,
2111                 .rpc_cred = data->owner->so_cred,
2112         };
2113         struct rpc_task_setup task_setup_data = {
2114                 .rpc_client = server->client,
2115                 .rpc_message = &msg,
2116                 .callback_ops = &nfs4_open_ops,
2117                 .callback_data = data,
2118                 .workqueue = nfsiod_workqueue,
2119                 .flags = RPC_TASK_ASYNC,
2120         };
2121         int status;
2122
2123         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2124         kref_get(&data->kref);
2125         data->rpc_done = 0;
2126         data->rpc_status = 0;
2127         data->cancelled = 0;
2128         data->is_recover = 0;
2129         if (isrecover) {
2130                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2131                 data->is_recover = 1;
2132         }
2133         task = rpc_run_task(&task_setup_data);
2134         if (IS_ERR(task))
2135                 return PTR_ERR(task);
2136         status = nfs4_wait_for_completion_rpc_task(task);
2137         if (status != 0) {
2138                 data->cancelled = 1;
2139                 smp_wmb();
2140         } else
2141                 status = data->rpc_status;
2142         rpc_put_task(task);
2143
2144         return status;
2145 }
2146
2147 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2148 {
2149         struct inode *dir = d_inode(data->dir);
2150         struct nfs_openres *o_res = &data->o_res;
2151         int status;
2152
2153         status = nfs4_run_open_task(data, 1);
2154         if (status != 0 || !data->rpc_done)
2155                 return status;
2156
2157         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2158
2159         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2160                 status = _nfs4_proc_open_confirm(data);
2161                 if (status != 0)
2162                         return status;
2163         }
2164
2165         return status;
2166 }
2167
2168 /*
2169  * Additional permission checks in order to distinguish between an
2170  * open for read, and an open for execute. This works around the
2171  * fact that NFSv4 OPEN treats read and execute permissions as being
2172  * the same.
2173  * Note that in the non-execute case, we want to turn off permission
2174  * checking if we just created a new file (POSIX open() semantics).
2175  */
2176 static int nfs4_opendata_access(struct rpc_cred *cred,
2177                                 struct nfs4_opendata *opendata,
2178                                 struct nfs4_state *state, fmode_t fmode,
2179                                 int openflags)
2180 {
2181         struct nfs_access_entry cache;
2182         u32 mask;
2183
2184         /* access call failed or for some reason the server doesn't
2185          * support any access modes -- defer access call until later */
2186         if (opendata->o_res.access_supported == 0)
2187                 return 0;
2188
2189         mask = 0;
2190         /*
2191          * Use openflags to check for exec, because fmode won't
2192          * always have FMODE_EXEC set when file open for exec.
2193          */
2194         if (openflags & __FMODE_EXEC) {
2195                 /* ONLY check for exec rights */
2196                 mask = MAY_EXEC;
2197         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2198                 mask = MAY_READ;
2199
2200         cache.cred = cred;
2201         cache.jiffies = jiffies;
2202         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2203         nfs_access_add_cache(state->inode, &cache);
2204
2205         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2206                 return 0;
2207
2208         /* even though OPEN succeeded, access is denied. Close the file */
2209         nfs4_close_state(state, fmode);
2210         return -EACCES;
2211 }
2212
2213 /*
2214  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2215  */
2216 static int _nfs4_proc_open(struct nfs4_opendata *data)
2217 {
2218         struct inode *dir = d_inode(data->dir);
2219         struct nfs_server *server = NFS_SERVER(dir);
2220         struct nfs_openargs *o_arg = &data->o_arg;
2221         struct nfs_openres *o_res = &data->o_res;
2222         int status;
2223
2224         status = nfs4_run_open_task(data, 0);
2225         if (!data->rpc_done)
2226                 return status;
2227         if (status != 0) {
2228                 if (status == -NFS4ERR_BADNAME &&
2229                                 !(o_arg->open_flags & O_CREAT))
2230                         return -ENOENT;
2231                 return status;
2232         }
2233
2234         nfs_fattr_map_and_free_names(server, &data->f_attr);
2235
2236         if (o_arg->open_flags & O_CREAT) {
2237                 update_changeattr(dir, &o_res->cinfo);
2238                 if (o_arg->open_flags & O_EXCL)
2239                         data->file_created = 1;
2240                 else if (o_res->cinfo.before != o_res->cinfo.after)
2241                         data->file_created = 1;
2242         }
2243         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2244                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2245         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2246                 status = _nfs4_proc_open_confirm(data);
2247                 if (status != 0)
2248                         return status;
2249         }
2250         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2251                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2252         return 0;
2253 }
2254
2255 static int nfs4_recover_expired_lease(struct nfs_server *server)
2256 {
2257         return nfs4_client_recover_expired_lease(server->nfs_client);
2258 }
2259
2260 /*
2261  * OPEN_EXPIRED:
2262  *      reclaim state on the server after a network partition.
2263  *      Assumes caller holds the appropriate lock
2264  */
2265 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2266 {
2267         struct nfs4_opendata *opendata;
2268         int ret;
2269
2270         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2271                         NFS4_OPEN_CLAIM_FH);
2272         if (IS_ERR(opendata))
2273                 return PTR_ERR(opendata);
2274         ret = nfs4_open_recover(opendata, state);
2275         if (ret == -ESTALE)
2276                 d_drop(ctx->dentry);
2277         nfs4_opendata_put(opendata);
2278         return ret;
2279 }
2280
2281 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2282 {
2283         struct nfs_server *server = NFS_SERVER(state->inode);
2284         struct nfs4_exception exception = { };
2285         int err;
2286
2287         do {
2288                 err = _nfs4_open_expired(ctx, state);
2289                 trace_nfs4_open_expired(ctx, 0, err);
2290                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2291                         continue;
2292                 switch (err) {
2293                 default:
2294                         goto out;
2295                 case -NFS4ERR_GRACE:
2296                 case -NFS4ERR_DELAY:
2297                         nfs4_handle_exception(server, err, &exception);
2298                         err = 0;
2299                 }
2300         } while (exception.retry);
2301 out:
2302         return err;
2303 }
2304
2305 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2306 {
2307         struct nfs_open_context *ctx;
2308         int ret;
2309
2310         ctx = nfs4_state_find_open_context(state);
2311         if (IS_ERR(ctx))
2312                 return -EAGAIN;
2313         ret = nfs4_do_open_expired(ctx, state);
2314         put_nfs_open_context(ctx);
2315         return ret;
2316 }
2317
2318 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2319 {
2320         nfs_remove_bad_delegation(state->inode);
2321         write_seqlock(&state->seqlock);
2322         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2323         write_sequnlock(&state->seqlock);
2324         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2325 }
2326
2327 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2328 {
2329         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2330                 nfs_finish_clear_delegation_stateid(state);
2331 }
2332
2333 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2334 {
2335         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2336         nfs40_clear_delegation_stateid(state);
2337         return nfs4_open_expired(sp, state);
2338 }
2339
2340 #if defined(CONFIG_NFS_V4_1)
2341 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2342 {
2343         struct nfs_server *server = NFS_SERVER(state->inode);
2344         nfs4_stateid stateid;
2345         struct nfs_delegation *delegation;
2346         struct rpc_cred *cred;
2347         int status;
2348
2349         /* Get the delegation credential for use by test/free_stateid */
2350         rcu_read_lock();
2351         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2352         if (delegation == NULL) {
2353                 rcu_read_unlock();
2354                 return;
2355         }
2356
2357         nfs4_stateid_copy(&stateid, &delegation->stateid);
2358         cred = get_rpccred(delegation->cred);
2359         rcu_read_unlock();
2360         status = nfs41_test_stateid(server, &stateid, cred);
2361         trace_nfs4_test_delegation_stateid(state, NULL, status);
2362
2363         if (status != NFS_OK) {
2364                 /* Free the stateid unless the server explicitly
2365                  * informs us the stateid is unrecognized. */
2366                 if (status != -NFS4ERR_BAD_STATEID)
2367                         nfs41_free_stateid(server, &stateid, cred);
2368                 nfs_finish_clear_delegation_stateid(state);
2369         }
2370
2371         put_rpccred(cred);
2372 }
2373
2374 /**
2375  * nfs41_check_open_stateid - possibly free an open stateid
2376  *
2377  * @state: NFSv4 state for an inode
2378  *
2379  * Returns NFS_OK if recovery for this stateid is now finished.
2380  * Otherwise a negative NFS4ERR value is returned.
2381  */
2382 static int nfs41_check_open_stateid(struct nfs4_state *state)
2383 {
2384         struct nfs_server *server = NFS_SERVER(state->inode);
2385         nfs4_stateid *stateid = &state->open_stateid;
2386         struct rpc_cred *cred = state->owner->so_cred;
2387         int status;
2388
2389         /* If a state reset has been done, test_stateid is unneeded */
2390         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2391             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2392             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2393                 return -NFS4ERR_BAD_STATEID;
2394
2395         status = nfs41_test_stateid(server, stateid, cred);
2396         trace_nfs4_test_open_stateid(state, NULL, status);
2397         if (status != NFS_OK) {
2398                 /* Free the stateid unless the server explicitly
2399                  * informs us the stateid is unrecognized. */
2400                 if (status != -NFS4ERR_BAD_STATEID)
2401                         nfs41_free_stateid(server, stateid, cred);
2402
2403                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2404                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2405                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2406                 clear_bit(NFS_OPEN_STATE, &state->flags);
2407         }
2408         return status;
2409 }
2410
2411 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2412 {
2413         int status;
2414
2415         nfs41_check_delegation_stateid(state);
2416         status = nfs41_check_open_stateid(state);
2417         if (status != NFS_OK)
2418                 status = nfs4_open_expired(sp, state);
2419         return status;
2420 }
2421 #endif
2422
2423 /*
2424  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2425  * fields corresponding to attributes that were used to store the verifier.
2426  * Make sure we clobber those fields in the later setattr call
2427  */
2428 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2429                                 struct iattr *sattr, struct nfs4_label **label)
2430 {
2431         const u32 *attrset = opendata->o_res.attrset;
2432
2433         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2434             !(sattr->ia_valid & ATTR_ATIME_SET))
2435                 sattr->ia_valid |= ATTR_ATIME;
2436
2437         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2438             !(sattr->ia_valid & ATTR_MTIME_SET))
2439                 sattr->ia_valid |= ATTR_MTIME;
2440
2441         /* Except MODE, it seems harmless of setting twice. */
2442         if ((attrset[1] & FATTR4_WORD1_MODE))
2443                 sattr->ia_valid &= ~ATTR_MODE;
2444
2445         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2446                 *label = NULL;
2447 }
2448
2449 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2450                 fmode_t fmode,
2451                 int flags,
2452                 struct nfs_open_context *ctx)
2453 {
2454         struct nfs4_state_owner *sp = opendata->owner;
2455         struct nfs_server *server = sp->so_server;
2456         struct dentry *dentry;
2457         struct nfs4_state *state;
2458         unsigned int seq;
2459         int ret;
2460
2461         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2462
2463         ret = _nfs4_proc_open(opendata);
2464         if (ret != 0)
2465                 goto out;
2466
2467         state = nfs4_opendata_to_nfs4_state(opendata);
2468         ret = PTR_ERR(state);
2469         if (IS_ERR(state))
2470                 goto out;
2471         if (server->caps & NFS_CAP_POSIX_LOCK)
2472                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2473
2474         dentry = opendata->dentry;
2475         if (d_really_is_negative(dentry)) {
2476                 struct dentry *alias;
2477                 d_drop(dentry);
2478                 alias = d_exact_alias(dentry, state->inode);
2479                 if (!alias)
2480                         alias = d_splice_alias(igrab(state->inode), dentry);
2481                 /* d_splice_alias() can't fail here - it's a non-directory */
2482                 if (alias) {
2483                         dput(ctx->dentry);
2484                         ctx->dentry = dentry = alias;
2485                 }
2486                 nfs_set_verifier(dentry,
2487                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2488         }
2489
2490         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2491         if (ret != 0)
2492                 goto out;
2493
2494         ctx->state = state;
2495         if (d_inode(dentry) == state->inode) {
2496                 nfs_inode_attach_open_context(ctx);
2497                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2498                         nfs4_schedule_stateid_recovery(server, state);
2499         }
2500 out:
2501         return ret;
2502 }
2503
2504 /*
2505  * Returns a referenced nfs4_state
2506  */
2507 static int _nfs4_do_open(struct inode *dir,
2508                         struct nfs_open_context *ctx,
2509                         int flags,
2510                         struct iattr *sattr,
2511                         struct nfs4_label *label,
2512                         int *opened)
2513 {
2514         struct nfs4_state_owner  *sp;
2515         struct nfs4_state     *state = NULL;
2516         struct nfs_server       *server = NFS_SERVER(dir);
2517         struct nfs4_opendata *opendata;
2518         struct dentry *dentry = ctx->dentry;
2519         struct rpc_cred *cred = ctx->cred;
2520         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2521         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2522         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2523         struct nfs4_label *olabel = NULL;
2524         int status;
2525
2526         /* Protect against reboot recovery conflicts */
2527         status = -ENOMEM;
2528         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2529         if (sp == NULL) {
2530                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2531                 goto out_err;
2532         }
2533         status = nfs4_recover_expired_lease(server);
2534         if (status != 0)
2535                 goto err_put_state_owner;
2536         if (d_really_is_positive(dentry))
2537                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2538         status = -ENOMEM;
2539         if (d_really_is_positive(dentry))
2540                 claim = NFS4_OPEN_CLAIM_FH;
2541         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2542                         label, claim, GFP_KERNEL);
2543         if (opendata == NULL)
2544                 goto err_put_state_owner;
2545
2546         if (label) {
2547                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2548                 if (IS_ERR(olabel)) {
2549                         status = PTR_ERR(olabel);
2550                         goto err_opendata_put;
2551                 }
2552         }
2553
2554         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2555                 if (!opendata->f_attr.mdsthreshold) {
2556                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2557                         if (!opendata->f_attr.mdsthreshold)
2558                                 goto err_free_label;
2559                 }
2560                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2561         }
2562         if (d_really_is_positive(dentry))
2563                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2564
2565         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2566         if (status != 0)
2567                 goto err_free_label;
2568         state = ctx->state;
2569
2570         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2571             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2572                 nfs4_exclusive_attrset(opendata, sattr, &label);
2573                 /*
2574                  * send create attributes which was not set by open
2575                  * with an extra setattr.
2576                  */
2577                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2578                         nfs_fattr_init(opendata->o_res.f_attr);
2579                         status = nfs4_do_setattr(state->inode, cred,
2580                                         opendata->o_res.f_attr, sattr,
2581                                         state, label, olabel);
2582                         if (status == 0) {
2583                                 nfs_setattr_update_inode(state->inode, sattr,
2584                                                 opendata->o_res.f_attr);
2585                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2586                         }
2587                 }
2588         }
2589         if (opened && opendata->file_created)
2590                 *opened |= FILE_CREATED;
2591
2592         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2593                 *ctx_th = opendata->f_attr.mdsthreshold;
2594                 opendata->f_attr.mdsthreshold = NULL;
2595         }
2596
2597         nfs4_label_free(olabel);
2598
2599         nfs4_opendata_put(opendata);
2600         nfs4_put_state_owner(sp);
2601         return 0;
2602 err_free_label:
2603         nfs4_label_free(olabel);
2604 err_opendata_put:
2605         nfs4_opendata_put(opendata);
2606 err_put_state_owner:
2607         nfs4_put_state_owner(sp);
2608 out_err:
2609         return status;
2610 }
2611
2612
2613 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2614                                         struct nfs_open_context *ctx,
2615                                         int flags,
2616                                         struct iattr *sattr,
2617                                         struct nfs4_label *label,
2618                                         int *opened)
2619 {
2620         struct nfs_server *server = NFS_SERVER(dir);
2621         struct nfs4_exception exception = { };
2622         struct nfs4_state *res;
2623         int status;
2624
2625         do {
2626                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2627                 res = ctx->state;
2628                 trace_nfs4_open_file(ctx, flags, status);
2629                 if (status == 0)
2630                         break;
2631                 /* NOTE: BAD_SEQID means the server and client disagree about the
2632                  * book-keeping w.r.t. state-changing operations
2633                  * (OPEN/CLOSE/LOCK/LOCKU...)
2634                  * It is actually a sign of a bug on the client or on the server.
2635                  *
2636                  * If we receive a BAD_SEQID error in the particular case of
2637                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2638                  * have unhashed the old state_owner for us, and that we can
2639                  * therefore safely retry using a new one. We should still warn
2640                  * the user though...
2641                  */
2642                 if (status == -NFS4ERR_BAD_SEQID) {
2643                         pr_warn_ratelimited("NFS: v4 server %s "
2644                                         " returned a bad sequence-id error!\n",
2645                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2646                         exception.retry = 1;
2647                         continue;
2648                 }
2649                 /*
2650                  * BAD_STATEID on OPEN means that the server cancelled our
2651                  * state before it received the OPEN_CONFIRM.
2652                  * Recover by retrying the request as per the discussion
2653                  * on Page 181 of RFC3530.
2654                  */
2655                 if (status == -NFS4ERR_BAD_STATEID) {
2656                         exception.retry = 1;
2657                         continue;
2658                 }
2659                 if (status == -EAGAIN) {
2660                         /* We must have found a delegation */
2661                         exception.retry = 1;
2662                         continue;
2663                 }
2664                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2665                         continue;
2666                 res = ERR_PTR(nfs4_handle_exception(server,
2667                                         status, &exception));
2668         } while (exception.retry);
2669         return res;
2670 }
2671
2672 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2673                             struct nfs_fattr *fattr, struct iattr *sattr,
2674                             struct nfs4_state *state, struct nfs4_label *ilabel,
2675                             struct nfs4_label *olabel)
2676 {
2677         struct nfs_server *server = NFS_SERVER(inode);
2678         struct nfs_setattrargs  arg = {
2679                 .fh             = NFS_FH(inode),
2680                 .iap            = sattr,
2681                 .server         = server,
2682                 .bitmask = server->attr_bitmask,
2683                 .label          = ilabel,
2684         };
2685         struct nfs_setattrres  res = {
2686                 .fattr          = fattr,
2687                 .label          = olabel,
2688                 .server         = server,
2689         };
2690         struct rpc_message msg = {
2691                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2692                 .rpc_argp       = &arg,
2693                 .rpc_resp       = &res,
2694                 .rpc_cred       = cred,
2695         };
2696         struct rpc_cred *delegation_cred = NULL;
2697         unsigned long timestamp = jiffies;
2698         fmode_t fmode;
2699         bool truncate;
2700         int status;
2701
2702         arg.bitmask = nfs4_bitmask(server, ilabel);
2703         if (ilabel)
2704                 arg.bitmask = nfs4_bitmask(server, olabel);
2705
2706         nfs_fattr_init(fattr);
2707
2708         /* Servers should only apply open mode checks for file size changes */
2709         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2710         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2711
2712         if (nfs4_copy_delegation_stateid(inode, fmode, &arg.stateid, &delegation_cred)) {
2713                 /* Use that stateid */
2714         } else if (truncate && state != NULL) {
2715                 struct nfs_lockowner lockowner = {
2716                         .l_owner = current->files,
2717                         .l_pid = current->tgid,
2718                 };
2719                 if (!nfs4_valid_open_stateid(state))
2720                         return -EBADF;
2721                 if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner,
2722                                 &arg.stateid, &delegation_cred) == -EIO)
2723                         return -EBADF;
2724         } else
2725                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2726         if (delegation_cred)
2727                 msg.rpc_cred = delegation_cred;
2728
2729         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2730
2731         put_rpccred(delegation_cred);
2732         if (status == 0 && state != NULL)
2733                 renew_lease(server, timestamp);
2734         trace_nfs4_setattr(inode, &arg.stateid, status);
2735         return status;
2736 }
2737
2738 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2739                            struct nfs_fattr *fattr, struct iattr *sattr,
2740                            struct nfs4_state *state, struct nfs4_label *ilabel,
2741                            struct nfs4_label *olabel)
2742 {
2743         struct nfs_server *server = NFS_SERVER(inode);
2744         struct nfs4_exception exception = {
2745                 .state = state,
2746                 .inode = inode,
2747         };
2748         int err;
2749         do {
2750                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2751                 switch (err) {
2752                 case -NFS4ERR_OPENMODE:
2753                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2754                                 pr_warn_once("NFSv4: server %s is incorrectly "
2755                                                 "applying open mode checks to "
2756                                                 "a SETATTR that is not "
2757                                                 "changing file size.\n",
2758                                                 server->nfs_client->cl_hostname);
2759                         }
2760                         if (state && !(state->state & FMODE_WRITE)) {
2761                                 err = -EBADF;
2762                                 if (sattr->ia_valid & ATTR_OPEN)
2763                                         err = -EACCES;
2764                                 goto out;
2765                         }
2766                 }
2767                 err = nfs4_handle_exception(server, err, &exception);
2768         } while (exception.retry);
2769 out:
2770         return err;
2771 }
2772
2773 static bool
2774 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2775 {
2776         if (inode == NULL || !nfs_have_layout(inode))
2777                 return false;
2778
2779         return pnfs_wait_on_layoutreturn(inode, task);
2780 }
2781
2782 struct nfs4_closedata {
2783         struct inode *inode;
2784         struct nfs4_state *state;
2785         struct nfs_closeargs arg;
2786         struct nfs_closeres res;
2787         struct nfs_fattr fattr;
2788         unsigned long timestamp;
2789         bool roc;
2790         u32 roc_barrier;
2791 };
2792
2793 static void nfs4_free_closedata(void *data)
2794 {
2795         struct nfs4_closedata *calldata = data;
2796         struct nfs4_state_owner *sp = calldata->state->owner;
2797         struct super_block *sb = calldata->state->inode->i_sb;
2798
2799         if (calldata->roc)
2800                 pnfs_roc_release(calldata->state->inode);
2801         nfs4_put_open_state(calldata->state);
2802         nfs_free_seqid(calldata->arg.seqid);
2803         nfs4_put_state_owner(sp);
2804         nfs_sb_deactive(sb);
2805         kfree(calldata);
2806 }
2807
2808 static void nfs4_close_done(struct rpc_task *task, void *data)
2809 {
2810         struct nfs4_closedata *calldata = data;
2811         struct nfs4_state *state = calldata->state;
2812         struct nfs_server *server = NFS_SERVER(calldata->inode);
2813         nfs4_stateid *res_stateid = NULL;
2814
2815         dprintk("%s: begin!\n", __func__);
2816         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2817                 return;
2818         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2819         /* hmm. we are done with the inode, and in the process of freeing
2820          * the state_owner. we keep this around to process errors
2821          */
2822         switch (task->tk_status) {
2823                 case 0:
2824                         res_stateid = &calldata->res.stateid;
2825                         if (calldata->roc)
2826                                 pnfs_roc_set_barrier(state->inode,
2827                                                      calldata->roc_barrier);
2828                         renew_lease(server, calldata->timestamp);
2829                         break;
2830                 case -NFS4ERR_ADMIN_REVOKED:
2831                 case -NFS4ERR_STALE_STATEID:
2832                 case -NFS4ERR_OLD_STATEID:
2833                 case -NFS4ERR_BAD_STATEID:
2834                 case -NFS4ERR_EXPIRED:
2835                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2836                                                 &state->open_stateid)) {
2837                                 rpc_restart_call_prepare(task);
2838                                 goto out_release;
2839                         }
2840                         if (calldata->arg.fmode == 0)
2841                                 break;
2842                 default:
2843                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2844                                 rpc_restart_call_prepare(task);
2845                                 goto out_release;
2846                         }
2847         }
2848         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2849                         res_stateid, calldata->arg.fmode);
2850 out_release:
2851         nfs_release_seqid(calldata->arg.seqid);
2852         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2853         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2854 }
2855
2856 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2857 {
2858         struct nfs4_closedata *calldata = data;
2859         struct nfs4_state *state = calldata->state;
2860         struct inode *inode = calldata->inode;
2861         bool is_rdonly, is_wronly, is_rdwr;
2862         int call_close = 0;
2863
2864         dprintk("%s: begin!\n", __func__);
2865         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2866                 goto out_wait;
2867
2868         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2869         spin_lock(&state->owner->so_lock);
2870         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2871         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2872         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2873         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2874         /* Calculate the change in open mode */
2875         calldata->arg.fmode = 0;
2876         if (state->n_rdwr == 0) {
2877                 if (state->n_rdonly == 0)
2878                         call_close |= is_rdonly;
2879                 else if (is_rdonly)
2880                         calldata->arg.fmode |= FMODE_READ;
2881                 if (state->n_wronly == 0)
2882                         call_close |= is_wronly;
2883                 else if (is_wronly)
2884                         calldata->arg.fmode |= FMODE_WRITE;
2885         } else if (is_rdwr)
2886                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2887
2888         if (calldata->arg.fmode == 0)
2889                 call_close |= is_rdwr;
2890
2891         if (!nfs4_valid_open_stateid(state))
2892                 call_close = 0;
2893         spin_unlock(&state->owner->so_lock);
2894
2895         if (!call_close) {
2896                 /* Note: exit _without_ calling nfs4_close_done */
2897                 goto out_no_action;
2898         }
2899
2900         if (nfs4_wait_on_layoutreturn(inode, task)) {
2901                 nfs_release_seqid(calldata->arg.seqid);
2902                 goto out_wait;
2903         }
2904
2905         if (calldata->arg.fmode == 0)
2906                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2907         if (calldata->roc)
2908                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2909
2910         calldata->arg.share_access =
2911                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2912                                 calldata->arg.fmode, 0);
2913
2914         nfs_fattr_init(calldata->res.fattr);
2915         calldata->timestamp = jiffies;
2916         if (nfs4_setup_sequence(NFS_SERVER(inode),
2917                                 &calldata->arg.seq_args,
2918                                 &calldata->res.seq_res,
2919                                 task) != 0)
2920                 nfs_release_seqid(calldata->arg.seqid);
2921         dprintk("%s: done!\n", __func__);
2922         return;
2923 out_no_action:
2924         task->tk_action = NULL;
2925 out_wait:
2926         nfs4_sequence_done(task, &calldata->res.seq_res);
2927 }
2928
2929 static const struct rpc_call_ops nfs4_close_ops = {
2930         .rpc_call_prepare = nfs4_close_prepare,
2931         .rpc_call_done = nfs4_close_done,
2932         .rpc_release = nfs4_free_closedata,
2933 };
2934
2935 static bool nfs4_roc(struct inode *inode)
2936 {
2937         if (!nfs_have_layout(inode))
2938                 return false;
2939         return pnfs_roc(inode);
2940 }
2941
2942 /* 
2943  * It is possible for data to be read/written from a mem-mapped file 
2944  * after the sys_close call (which hits the vfs layer as a flush).
2945  * This means that we can't safely call nfsv4 close on a file until 
2946  * the inode is cleared. This in turn means that we are not good
2947  * NFSv4 citizens - we do not indicate to the server to update the file's 
2948  * share state even when we are done with one of the three share 
2949  * stateid's in the inode.
2950  *
2951  * NOTE: Caller must be holding the sp->so_owner semaphore!
2952  */
2953 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2954 {
2955         struct nfs_server *server = NFS_SERVER(state->inode);
2956         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2957         struct nfs4_closedata *calldata;
2958         struct nfs4_state_owner *sp = state->owner;
2959         struct rpc_task *task;
2960         struct rpc_message msg = {
2961                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2962                 .rpc_cred = state->owner->so_cred,
2963         };
2964         struct rpc_task_setup task_setup_data = {
2965                 .rpc_client = server->client,
2966                 .rpc_message = &msg,
2967                 .callback_ops = &nfs4_close_ops,
2968                 .workqueue = nfsiod_workqueue,
2969                 .flags = RPC_TASK_ASYNC,
2970         };
2971         int status = -ENOMEM;
2972
2973         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2974                 &task_setup_data.rpc_client, &msg);
2975
2976         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2977         if (calldata == NULL)
2978                 goto out;
2979         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2980         calldata->inode = state->inode;
2981         calldata->state = state;
2982         calldata->arg.fh = NFS_FH(state->inode);
2983         /* Serialization for the sequence id */
2984         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2985         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2986         if (IS_ERR(calldata->arg.seqid))
2987                 goto out_free_calldata;
2988         calldata->arg.fmode = 0;
2989         calldata->arg.bitmask = server->cache_consistency_bitmask;
2990         calldata->res.fattr = &calldata->fattr;
2991         calldata->res.seqid = calldata->arg.seqid;
2992         calldata->res.server = server;
2993         calldata->roc = nfs4_roc(state->inode);
2994         nfs_sb_active(calldata->inode->i_sb);
2995
2996         msg.rpc_argp = &calldata->arg;
2997         msg.rpc_resp = &calldata->res;
2998         task_setup_data.callback_data = calldata;
2999         task = rpc_run_task(&task_setup_data);
3000         if (IS_ERR(task))
3001                 return PTR_ERR(task);
3002         status = 0;
3003         if (wait)
3004                 status = rpc_wait_for_completion_task(task);
3005         rpc_put_task(task);
3006         return status;
3007 out_free_calldata:
3008         kfree(calldata);
3009 out:
3010         nfs4_put_open_state(state);
3011         nfs4_put_state_owner(sp);
3012         return status;
3013 }
3014
3015 static struct inode *
3016 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3017                 int open_flags, struct iattr *attr, int *opened)
3018 {
3019         struct nfs4_state *state;
3020         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3021
3022         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3023
3024         /* Protect against concurrent sillydeletes */
3025         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3026
3027         nfs4_label_release_security(label);
3028
3029         if (IS_ERR(state))
3030                 return ERR_CAST(state);
3031         return state->inode;
3032 }
3033
3034 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3035 {
3036         if (ctx->state == NULL)
3037                 return;
3038         if (is_sync)
3039                 nfs4_close_sync(ctx->state, ctx->mode);
3040         else
3041                 nfs4_close_state(ctx->state, ctx->mode);
3042 }
3043
3044 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3045 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3046 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3047
3048 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3049 {
3050         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3051         struct nfs4_server_caps_arg args = {
3052                 .fhandle = fhandle,
3053                 .bitmask = bitmask,
3054         };
3055         struct nfs4_server_caps_res res = {};
3056         struct rpc_message msg = {
3057                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3058                 .rpc_argp = &args,
3059                 .rpc_resp = &res,
3060         };
3061         int status;
3062
3063         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3064                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3065                      FATTR4_WORD0_LINK_SUPPORT |
3066                      FATTR4_WORD0_SYMLINK_SUPPORT |
3067                      FATTR4_WORD0_ACLSUPPORT;
3068         if (minorversion)
3069                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3070
3071         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3072         if (status == 0) {
3073                 /* Sanity check the server answers */
3074                 switch (minorversion) {
3075                 case 0:
3076                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3077                         res.attr_bitmask[2] = 0;
3078                         break;
3079                 case 1:
3080                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3081                         break;
3082                 case 2:
3083                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3084                 }
3085                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3086                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3087                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3088                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3089                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3090                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3091                                 NFS_CAP_SECURITY_LABEL);
3092                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3093                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3094                         server->caps |= NFS_CAP_ACLS;
3095                 if (res.has_links != 0)
3096                         server->caps |= NFS_CAP_HARDLINKS;
3097                 if (res.has_symlinks != 0)
3098                         server->caps |= NFS_CAP_SYMLINKS;
3099                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3100                         server->caps |= NFS_CAP_FILEID;
3101                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3102                         server->caps |= NFS_CAP_MODE;
3103                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3104                         server->caps |= NFS_CAP_NLINK;
3105                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3106                         server->caps |= NFS_CAP_OWNER;
3107                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3108                         server->caps |= NFS_CAP_OWNER_GROUP;
3109                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3110                         server->caps |= NFS_CAP_ATIME;
3111                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3112                         server->caps |= NFS_CAP_CTIME;
3113                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3114                         server->caps |= NFS_CAP_MTIME;
3115 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3116                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3117                         server->caps |= NFS_CAP_SECURITY_LABEL;
3118 #endif
3119                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3120                                 sizeof(server->attr_bitmask));
3121                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3122
3123                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3124                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3125                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3126                 server->cache_consistency_bitmask[2] = 0;
3127                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3128                         sizeof(server->exclcreat_bitmask));
3129                 server->acl_bitmask = res.acl_bitmask;
3130                 server->fh_expire_type = res.fh_expire_type;
3131         }
3132
3133         return status;
3134 }
3135
3136 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3137 {
3138         struct nfs4_exception exception = { };
3139         int err;
3140         do {
3141                 err = nfs4_handle_exception(server,
3142                                 _nfs4_server_capabilities(server, fhandle),
3143                                 &exception);
3144         } while (exception.retry);
3145         return err;
3146 }
3147
3148 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3149                 struct nfs_fsinfo *info)
3150 {
3151         u32 bitmask[3];
3152         struct nfs4_lookup_root_arg args = {
3153                 .bitmask = bitmask,
3154         };
3155         struct nfs4_lookup_res res = {
3156                 .server = server,
3157                 .fattr = info->fattr,
3158                 .fh = fhandle,
3159         };
3160         struct rpc_message msg = {
3161                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3162                 .rpc_argp = &args,
3163                 .rpc_resp = &res,
3164         };
3165
3166         bitmask[0] = nfs4_fattr_bitmap[0];
3167         bitmask[1] = nfs4_fattr_bitmap[1];
3168         /*
3169          * Process the label in the upcoming getfattr
3170          */
3171         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3172
3173         nfs_fattr_init(info->fattr);
3174         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3175 }
3176
3177 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3178                 struct nfs_fsinfo *info)
3179 {
3180         struct nfs4_exception exception = { };
3181         int err;
3182         do {
3183                 err = _nfs4_lookup_root(server, fhandle, info);
3184                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3185                 switch (err) {
3186                 case 0:
3187                 case -NFS4ERR_WRONGSEC:
3188                         goto out;
3189                 default:
3190                         err = nfs4_handle_exception(server, err, &exception);
3191                 }
3192         } while (exception.retry);
3193 out:
3194         return err;
3195 }
3196
3197 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3198                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3199 {
3200         struct rpc_auth_create_args auth_args = {
3201                 .pseudoflavor = flavor,
3202         };
3203         struct rpc_auth *auth;
3204         int ret;
3205
3206         auth = rpcauth_create(&auth_args, server->client);
3207         if (IS_ERR(auth)) {
3208                 ret = -EACCES;
3209                 goto out;
3210         }
3211         ret = nfs4_lookup_root(server, fhandle, info);
3212 out:
3213         return ret;
3214 }
3215
3216 /*
3217  * Retry pseudoroot lookup with various security flavors.  We do this when:
3218  *
3219  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3220  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3221  *
3222  * Returns zero on success, or a negative NFS4ERR value, or a
3223  * negative errno value.
3224  */
3225 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3226                               struct nfs_fsinfo *info)
3227 {
3228         /* Per 3530bis 15.33.5 */
3229         static const rpc_authflavor_t flav_array[] = {
3230                 RPC_AUTH_GSS_KRB5P,
3231                 RPC_AUTH_GSS_KRB5I,
3232                 RPC_AUTH_GSS_KRB5,
3233                 RPC_AUTH_UNIX,                  /* courtesy */
3234                 RPC_AUTH_NULL,
3235         };
3236         int status = -EPERM;
3237         size_t i;
3238
3239         if (server->auth_info.flavor_len > 0) {
3240                 /* try each flavor specified by user */
3241                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3242                         status = nfs4_lookup_root_sec(server, fhandle, info,
3243                                                 server->auth_info.flavors[i]);
3244                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3245                                 continue;
3246                         break;
3247                 }
3248         } else {
3249                 /* no flavors specified by user, try default list */
3250                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3251                         status = nfs4_lookup_root_sec(server, fhandle, info,
3252                                                       flav_array[i]);
3253                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3254                                 continue;
3255                         break;
3256                 }
3257         }
3258
3259         /*
3260          * -EACCESS could mean that the user doesn't have correct permissions
3261          * to access the mount.  It could also mean that we tried to mount
3262          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3263          * existing mount programs don't handle -EACCES very well so it should
3264          * be mapped to -EPERM instead.
3265          */
3266         if (status == -EACCES)
3267                 status = -EPERM;
3268         return status;
3269 }
3270
3271 static int nfs4_do_find_root_sec(struct nfs_server *server,
3272                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3273 {
3274         int mv = server->nfs_client->cl_minorversion;
3275         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3276 }
3277
3278 /**
3279  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3280  * @server: initialized nfs_server handle
3281  * @fhandle: we fill in the pseudo-fs root file handle
3282  * @info: we fill in an FSINFO struct
3283  * @auth_probe: probe the auth flavours
3284  *
3285  * Returns zero on success, or a negative errno.
3286  */
3287 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3288                          struct nfs_fsinfo *info,
3289                          bool auth_probe)
3290 {
3291         int status = 0;
3292
3293         if (!auth_probe)
3294                 status = nfs4_lookup_root(server, fhandle, info);
3295
3296         if (auth_probe || status == NFS4ERR_WRONGSEC)
3297                 status = nfs4_do_find_root_sec(server, fhandle, info);
3298
3299         if (status == 0)
3300                 status = nfs4_server_capabilities(server, fhandle);
3301         if (status == 0)
3302                 status = nfs4_do_fsinfo(server, fhandle, info);
3303
3304         return nfs4_map_errors(status);
3305 }
3306
3307 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3308                               struct nfs_fsinfo *info)
3309 {
3310         int error;
3311         struct nfs_fattr *fattr = info->fattr;
3312         struct nfs4_label *label = NULL;
3313
3314         error = nfs4_server_capabilities(server, mntfh);
3315         if (error < 0) {
3316                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3317                 return error;
3318         }
3319
3320         label = nfs4_label_alloc(server, GFP_KERNEL);
3321         if (IS_ERR(label))
3322                 return PTR_ERR(label);
3323
3324         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3325         if (error < 0) {
3326                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3327                 goto err_free_label;
3328         }
3329
3330         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3331             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3332                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3333
3334 err_free_label:
3335         nfs4_label_free(label);
3336
3337         return error;
3338 }
3339
3340 /*
3341  * Get locations and (maybe) other attributes of a referral.
3342  * Note that we'll actually follow the referral later when
3343  * we detect fsid mismatch in inode revalidation
3344  */
3345 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3346                              const struct qstr *name, struct nfs_fattr *fattr,
3347                              struct nfs_fh *fhandle)
3348 {
3349         int status = -ENOMEM;
3350         struct page *page = NULL;
3351         struct nfs4_fs_locations *locations = NULL;
3352
3353         page = alloc_page(GFP_KERNEL);
3354         if (page == NULL)
3355                 goto out;
3356         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3357         if (locations == NULL)
3358                 goto out;
3359
3360         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3361         if (status != 0)
3362                 goto out;
3363
3364         /*
3365          * If the fsid didn't change, this is a migration event, not a
3366          * referral.  Cause us to drop into the exception handler, which
3367          * will kick off migration recovery.
3368          */
3369         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3370                 dprintk("%s: server did not return a different fsid for"
3371                         " a referral at %s\n", __func__, name->name);
3372                 status = -NFS4ERR_MOVED;
3373                 goto out;
3374         }
3375         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3376         nfs_fixup_referral_attributes(&locations->fattr);
3377
3378         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3379         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3380         memset(fhandle, 0, sizeof(struct nfs_fh));
3381 out:
3382         if (page)
3383                 __free_page(page);
3384         kfree(locations);
3385         return status;
3386 }
3387
3388 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3389                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3390 {
3391         struct nfs4_getattr_arg args = {
3392                 .fh = fhandle,
3393                 .bitmask = server->attr_bitmask,
3394         };
3395         struct nfs4_getattr_res res = {
3396                 .fattr = fattr,
3397                 .label = label,
3398                 .server = server,
3399         };
3400         struct rpc_message msg = {
3401                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3402                 .rpc_argp = &args,
3403                 .rpc_resp = &res,
3404         };
3405
3406         args.bitmask = nfs4_bitmask(server, label);
3407
3408         nfs_fattr_init(fattr);
3409         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3410 }
3411
3412 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3413                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3414 {
3415         struct nfs4_exception exception = { };
3416         int err;
3417         do {
3418                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3419                 trace_nfs4_getattr(server, fhandle, fattr, err);
3420                 err = nfs4_handle_exception(server, err,
3421                                 &exception);
3422         } while (exception.retry);
3423         return err;
3424 }
3425
3426 /* 
3427  * The file is not closed if it is opened due to the a request to change
3428  * the size of the file. The open call will not be needed once the
3429  * VFS layer lookup-intents are implemented.
3430  *
3431  * Close is called when the inode is destroyed.
3432  * If we haven't opened the file for O_WRONLY, we
3433  * need to in the size_change case to obtain a stateid.
3434  *
3435  * Got race?
3436  * Because OPEN is always done by name in nfsv4, it is
3437  * possible that we opened a different file by the same
3438  * name.  We can recognize this race condition, but we
3439  * can't do anything about it besides returning an error.
3440  *
3441  * This will be fixed with VFS changes (lookup-intent).
3442  */
3443 static int
3444 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3445                   struct iattr *sattr)
3446 {
3447         struct inode *inode = d_inode(dentry);
3448         struct rpc_cred *cred = NULL;
3449         struct nfs4_state *state = NULL;
3450         struct nfs4_label *label = NULL;
3451         int status;
3452
3453         if (pnfs_ld_layoutret_on_setattr(inode) &&
3454             sattr->ia_valid & ATTR_SIZE &&
3455             sattr->ia_size < i_size_read(inode))
3456                 pnfs_commit_and_return_layout(inode);
3457
3458         nfs_fattr_init(fattr);
3459         
3460         /* Deal with open(O_TRUNC) */
3461         if (sattr->ia_valid & ATTR_OPEN)
3462                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3463
3464         /* Optimization: if the end result is no change, don't RPC */
3465         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3466                 return 0;
3467
3468         /* Search for an existing open(O_WRITE) file */
3469         if (sattr->ia_valid & ATTR_FILE) {
3470                 struct nfs_open_context *ctx;
3471
3472                 ctx = nfs_file_open_context(sattr->ia_file);
3473                 if (ctx) {
3474                         cred = ctx->cred;
3475                         state = ctx->state;
3476                 }
3477         }
3478
3479         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3480         if (IS_ERR(label))
3481                 return PTR_ERR(label);
3482
3483         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3484         if (status == 0) {
3485                 nfs_setattr_update_inode(inode, sattr, fattr);
3486                 nfs_setsecurity(inode, fattr, label);
3487         }
3488         nfs4_label_free(label);
3489         return status;
3490 }
3491
3492 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3493                 const struct qstr *name, struct nfs_fh *fhandle,
3494                 struct nfs_fattr *fattr, struct nfs4_label *label)
3495 {
3496         struct nfs_server *server = NFS_SERVER(dir);
3497         int                    status;
3498         struct nfs4_lookup_arg args = {
3499                 .bitmask = server->attr_bitmask,
3500                 .dir_fh = NFS_FH(dir),
3501                 .name = name,
3502         };
3503         struct nfs4_lookup_res res = {
3504                 .server = server,
3505                 .fattr = fattr,
3506                 .label = label,
3507                 .fh = fhandle,
3508         };
3509         struct rpc_message msg = {
3510                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3511                 .rpc_argp = &args,
3512                 .rpc_resp = &res,
3513         };
3514
3515         args.bitmask = nfs4_bitmask(server, label);
3516
3517         nfs_fattr_init(fattr);
3518
3519         dprintk("NFS call  lookup %s\n", name->name);
3520         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3521         dprintk("NFS reply lookup: %d\n", status);
3522         return status;
3523 }
3524
3525 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3526 {
3527         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3528                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3529         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3530         fattr->nlink = 2;
3531 }
3532
3533 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3534                                    struct qstr *name, struct nfs_fh *fhandle,
3535                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3536 {
3537         struct nfs4_exception exception = { };
3538         struct rpc_clnt *client = *clnt;
3539         int err;
3540         do {
3541                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3542                 trace_nfs4_lookup(dir, name, err);
3543                 switch (err) {
3544                 case -NFS4ERR_BADNAME:
3545                         err = -ENOENT;
3546                         goto out;
3547                 case -NFS4ERR_MOVED:
3548                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3549                         if (err == -NFS4ERR_MOVED)
3550                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3551                         goto out;
3552                 case -NFS4ERR_WRONGSEC:
3553                         err = -EPERM;
3554                         if (client != *clnt)
3555                                 goto out;
3556                         client = nfs4_negotiate_security(client, dir, name);
3557                         if (IS_ERR(client))
3558                                 return PTR_ERR(client);
3559
3560                         exception.retry = 1;
3561                         break;
3562                 default:
3563                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3564                 }
3565         } while (exception.retry);
3566
3567 out:
3568         if (err == 0)
3569                 *clnt = client;
3570         else if (client != *clnt)
3571                 rpc_shutdown_client(client);
3572
3573         return err;
3574 }
3575
3576 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3577                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3578                             struct nfs4_label *label)
3579 {
3580         int status;
3581         struct rpc_clnt *client = NFS_CLIENT(dir);
3582
3583         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3584         if (client != NFS_CLIENT(dir)) {
3585                 rpc_shutdown_client(client);
3586                 nfs_fixup_secinfo_attributes(fattr);
3587         }
3588         return status;
3589 }
3590
3591 struct rpc_clnt *
3592 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3593                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3594 {
3595         struct rpc_clnt *client = NFS_CLIENT(dir);
3596         int status;
3597
3598         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3599         if (status < 0)
3600                 return ERR_PTR(status);
3601         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3602 }
3603
3604 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3605 {
3606         struct nfs_server *server = NFS_SERVER(inode);
3607         struct nfs4_accessargs args = {
3608                 .fh = NFS_FH(inode),
3609                 .bitmask = server->cache_consistency_bitmask,
3610         };
3611         struct nfs4_accessres res = {
3612                 .server = server,
3613         };
3614         struct rpc_message msg = {
3615                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3616                 .rpc_argp = &args,
3617                 .rpc_resp = &res,
3618                 .rpc_cred = entry->cred,
3619         };
3620         int mode = entry->mask;
3621         int status = 0;
3622
3623         /*
3624          * Determine which access bits we want to ask for...
3625          */
3626         if (mode & MAY_READ)
3627                 args.access |= NFS4_ACCESS_READ;
3628         if (S_ISDIR(inode->i_mode)) {
3629                 if (mode & MAY_WRITE)
3630                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3631                 if (mode & MAY_EXEC)
3632                         args.access |= NFS4_ACCESS_LOOKUP;
3633         } else {
3634                 if (mode & MAY_WRITE)
3635                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3636                 if (mode & MAY_EXEC)
3637                         args.access |= NFS4_ACCESS_EXECUTE;
3638         }
3639
3640         res.fattr = nfs_alloc_fattr();
3641         if (res.fattr == NULL)
3642                 return -ENOMEM;
3643
3644         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3645         if (!status) {
3646                 nfs_access_set_mask(entry, res.access);
3647                 nfs_refresh_inode(inode, res.fattr);
3648         }
3649         nfs_free_fattr(res.fattr);
3650         return status;
3651 }
3652
3653 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3654 {
3655         struct nfs4_exception exception = { };
3656         int err;
3657         do {
3658                 err = _nfs4_proc_access(inode, entry);
3659                 trace_nfs4_access(inode, err);
3660                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3661                                 &exception);
3662         } while (exception.retry);
3663         return err;
3664 }
3665
3666 /*
3667  * TODO: For the time being, we don't try to get any attributes
3668  * along with any of the zero-copy operations READ, READDIR,
3669  * READLINK, WRITE.
3670  *
3671  * In the case of the first three, we want to put the GETATTR
3672  * after the read-type operation -- this is because it is hard
3673  * to predict the length of a GETATTR response in v4, and thus
3674  * align the READ data correctly.  This means that the GETATTR
3675  * may end up partially falling into the page cache, and we should
3676  * shift it into the 'tail' of the xdr_buf before processing.
3677  * To do this efficiently, we need to know the total length
3678  * of data received, which doesn't seem to be available outside
3679  * of the RPC layer.
3680  *
3681  * In the case of WRITE, we also want to put the GETATTR after
3682  * the operation -- in this case because we want to make sure
3683  * we get the post-operation mtime and size.
3684  *
3685  * Both of these changes to the XDR layer would in fact be quite
3686  * minor, but I decided to leave them for a subsequent patch.
3687  */
3688 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3689                 unsigned int pgbase, unsigned int pglen)
3690 {
3691         struct nfs4_readlink args = {
3692                 .fh       = NFS_FH(inode),
3693                 .pgbase   = pgbase,
3694                 .pglen    = pglen,
3695                 .pages    = &page,
3696         };
3697         struct nfs4_readlink_res res;
3698         struct rpc_message msg = {
3699                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3700                 .rpc_argp = &args,
3701                 .rpc_resp = &res,
3702         };
3703
3704         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3705 }
3706
3707 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3708                 unsigned int pgbase, unsigned int pglen)
3709 {
3710         struct nfs4_exception exception = { };
3711         int err;
3712         do {
3713                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3714                 trace_nfs4_readlink(inode, err);
3715                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3716                                 &exception);
3717         } while (exception.retry);
3718         return err;
3719 }
3720
3721 /*
3722  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3723  */
3724 static int
3725 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3726                  int flags)
3727 {
3728         struct nfs4_label l, *ilabel = NULL;
3729         struct nfs_open_context *ctx;
3730         struct nfs4_state *state;
3731         int status = 0;
3732
3733         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3734         if (IS_ERR(ctx))
3735                 return PTR_ERR(ctx);
3736
3737         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3738
3739         sattr->ia_mode &= ~current_umask();
3740         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3741         if (IS_ERR(state)) {
3742                 status = PTR_ERR(state);
3743                 goto out;
3744         }
3745 out:
3746         nfs4_label_release_security(ilabel);
3747         put_nfs_open_context(ctx);
3748         return status;
3749 }
3750
3751 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3752 {
3753         struct nfs_server *server = NFS_SERVER(dir);
3754         struct nfs_removeargs args = {
3755                 .fh = NFS_FH(dir),
3756                 .name = *name,
3757         };
3758         struct nfs_removeres res = {
3759                 .server = server,
3760         };
3761         struct rpc_message msg = {
3762                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3763                 .rpc_argp = &args,
3764                 .rpc_resp = &res,
3765         };
3766         int status;
3767
3768         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3769         if (status == 0)
3770                 update_changeattr(dir, &res.cinfo);
3771         return status;
3772 }
3773
3774 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3775 {
3776         struct nfs4_exception exception = { };
3777         int err;
3778         do {
3779                 err = _nfs4_proc_remove(dir, name);
3780                 trace_nfs4_remove(dir, name, err);
3781                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3782                                 &exception);
3783         } while (exception.retry);
3784         return err;
3785 }
3786
3787 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3788 {
3789         struct nfs_server *server = NFS_SERVER(dir);
3790         struct nfs_removeargs *args = msg->rpc_argp;
3791         struct nfs_removeres *res = msg->rpc_resp;
3792
3793         res->server = server;
3794         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3795         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3796
3797         nfs_fattr_init(res->dir_attr);
3798 }
3799
3800 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3801 {
3802         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
3803                         &data->args.seq_args,
3804                         &data->res.seq_res,
3805                         task);
3806 }
3807
3808 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3809 {
3810         struct nfs_unlinkdata *data = task->tk_calldata;
3811         struct nfs_removeres *res = &data->res;
3812
3813         if (!nfs4_sequence_done(task, &res->seq_res))
3814                 return 0;
3815         if (nfs4_async_handle_error(task, res->server, NULL,
3816                                     &data->timeout) == -EAGAIN)
3817                 return 0;
3818         update_changeattr(dir, &res->cinfo);
3819         return 1;
3820 }
3821
3822 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3823 {
3824         struct nfs_server *server = NFS_SERVER(dir);
3825         struct nfs_renameargs *arg = msg->rpc_argp;
3826         struct nfs_renameres *res = msg->rpc_resp;
3827
3828         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3829         res->server = server;
3830         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3831 }
3832
3833 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3834 {
3835         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3836                         &data->args.seq_args,
3837                         &data->res.seq_res,
3838                         task);
3839 }
3840
3841 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3842                                  struct inode *new_dir)
3843 {
3844         struct nfs_renamedata *data = task->tk_calldata;
3845         struct nfs_renameres *res = &data->res;
3846
3847         if (!nfs4_sequence_done(task, &res->seq_res))
3848                 return 0;
3849         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3850                 return 0;
3851
3852         update_changeattr(old_dir, &res->old_cinfo);
3853         update_changeattr(new_dir, &res->new_cinfo);
3854         return 1;
3855 }
3856
3857 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3858 {
3859         struct nfs_server *server = NFS_SERVER(inode);
3860         struct nfs4_link_arg arg = {
3861                 .fh     = NFS_FH(inode),
3862                 .dir_fh = NFS_FH(dir),
3863                 .name   = name,
3864                 .bitmask = server->attr_bitmask,
3865         };
3866         struct nfs4_link_res res = {
3867                 .server = server,
3868                 .label = NULL,
3869         };
3870         struct rpc_message msg = {
3871                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3872                 .rpc_argp = &arg,
3873                 .rpc_resp = &res,
3874         };
3875         int status = -ENOMEM;
3876
3877         res.fattr = nfs_alloc_fattr();
3878         if (res.fattr == NULL)
3879                 goto out;
3880
3881         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3882         if (IS_ERR(res.label)) {
3883                 status = PTR_ERR(res.label);
3884                 goto out;
3885         }
3886         arg.bitmask = nfs4_bitmask(server, res.label);
3887
3888         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3889         if (!status) {
3890                 update_changeattr(dir, &res.cinfo);
3891                 status = nfs_post_op_update_inode(inode, res.fattr);
3892                 if (!status)
3893                         nfs_setsecurity(inode, res.fattr, res.label);
3894         }
3895
3896
3897         nfs4_label_free(res.label);
3898
3899 out:
3900         nfs_free_fattr(res.fattr);
3901         return status;
3902 }
3903
3904 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3905 {
3906         struct nfs4_exception exception = { };
3907         int err;
3908         do {
3909                 err = nfs4_handle_exception(NFS_SERVER(inode),
3910                                 _nfs4_proc_link(inode, dir, name),
3911                                 &exception);
3912         } while (exception.retry);
3913         return err;
3914 }
3915
3916 struct nfs4_createdata {
3917         struct rpc_message msg;
3918         struct nfs4_create_arg arg;
3919         struct nfs4_create_res res;
3920         struct nfs_fh fh;
3921         struct nfs_fattr fattr;
3922         struct nfs4_label *label;
3923 };
3924
3925 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3926                 struct qstr *name, struct iattr *sattr, u32 ftype)
3927 {
3928         struct nfs4_createdata *data;
3929
3930         data = kzalloc(sizeof(*data), GFP_KERNEL);
3931         if (data != NULL) {
3932                 struct nfs_server *server = NFS_SERVER(dir);
3933
3934                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3935                 if (IS_ERR(data->label))
3936                         goto out_free;
3937
3938                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3939                 data->msg.rpc_argp = &data->arg;
3940                 data->msg.rpc_resp = &data->res;
3941                 data->arg.dir_fh = NFS_FH(dir);
3942                 data->arg.server = server;
3943                 data->arg.name = name;
3944                 data->arg.attrs = sattr;
3945                 data->arg.ftype = ftype;
3946                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3947                 data->res.server = server;
3948                 data->res.fh = &data->fh;
3949                 data->res.fattr = &data->fattr;
3950                 data->res.label = data->label;
3951                 nfs_fattr_init(data->res.fattr);
3952         }
3953         return data;
3954 out_free:
3955         kfree(data);
3956         return NULL;
3957 }
3958
3959 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3960 {
3961         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3962                                     &data->arg.seq_args, &data->res.seq_res, 1);
3963         if (status == 0) {
3964                 update_changeattr(dir, &data->res.dir_cinfo);
3965                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3966         }
3967         return status;
3968 }
3969
3970 static void nfs4_free_createdata(struct nfs4_createdata *data)
3971 {
3972         nfs4_label_free(data->label);
3973         kfree(data);
3974 }
3975
3976 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3977                 struct page *page, unsigned int len, struct iattr *sattr,
3978                 struct nfs4_label *label)
3979 {
3980         struct nfs4_createdata *data;
3981         int status = -ENAMETOOLONG;
3982
3983         if (len > NFS4_MAXPATHLEN)
3984                 goto out;
3985
3986         status = -ENOMEM;
3987         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3988         if (data == NULL)
3989                 goto out;
3990
3991         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3992         data->arg.u.symlink.pages = &page;
3993         data->arg.u.symlink.len = len;
3994         data->arg.label = label;
3995         
3996         status = nfs4_do_create(dir, dentry, data);
3997
3998         nfs4_free_createdata(data);
3999 out:
4000         return status;
4001 }
4002
4003 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4004                 struct page *page, unsigned int len, struct iattr *sattr)
4005 {
4006         struct nfs4_exception exception = { };
4007         struct nfs4_label l, *label = NULL;
4008         int err;
4009
4010         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4011
4012         do {
4013                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4014                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4015                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4016                                 &exception);
4017         } while (exception.retry);
4018
4019         nfs4_label_release_security(label);
4020         return err;
4021 }
4022
4023 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4024                 struct iattr *sattr, struct nfs4_label *label)
4025 {
4026         struct nfs4_createdata *data;
4027         int status = -ENOMEM;
4028
4029         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4030         if (data == NULL)
4031                 goto out;
4032
4033         data->arg.label = label;
4034         status = nfs4_do_create(dir, dentry, data);
4035
4036         nfs4_free_createdata(data);
4037 out:
4038         return status;
4039 }
4040
4041 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4042                 struct iattr *sattr)
4043 {
4044         struct nfs4_exception exception = { };
4045         struct nfs4_label l, *label = NULL;
4046         int err;
4047
4048         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4049
4050         sattr->ia_mode &= ~current_umask();
4051         do {
4052                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4053                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4054                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4055                                 &exception);
4056         } while (exception.retry);
4057         nfs4_label_release_security(label);
4058
4059         return err;
4060 }
4061
4062 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4063                 u64 cookie, struct page **pages, unsigned int count, int plus)
4064 {
4065         struct inode            *dir = d_inode(dentry);
4066         struct nfs4_readdir_arg args = {
4067                 .fh = NFS_FH(dir),
4068                 .pages = pages,
4069                 .pgbase = 0,
4070                 .count = count,
4071                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4072                 .plus = plus,
4073         };
4074         struct nfs4_readdir_res res;
4075         struct rpc_message msg = {
4076                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4077                 .rpc_argp = &args,
4078                 .rpc_resp = &res,
4079                 .rpc_cred = cred,
4080         };
4081         int                     status;
4082
4083         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4084                         dentry,
4085                         (unsigned long long)cookie);
4086         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4087         res.pgbase = args.pgbase;
4088         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4089         if (status >= 0) {
4090                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4091                 status += args.pgbase;
4092         }
4093
4094         nfs_invalidate_atime(dir);
4095
4096         dprintk("%s: returns %d\n", __func__, status);
4097         return status;
4098 }
4099
4100 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4101                 u64 cookie, struct page **pages, unsigned int count, int plus)
4102 {
4103         struct nfs4_exception exception = { };
4104         int err;
4105         do {
4106                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4107                                 pages, count, plus);
4108                 trace_nfs4_readdir(d_inode(dentry), err);
4109                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4110                                 &exception);
4111         } while (exception.retry);
4112         return err;
4113 }
4114
4115 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4116                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4117 {
4118         struct nfs4_createdata *data;
4119         int mode = sattr->ia_mode;
4120         int status = -ENOMEM;
4121
4122         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4123         if (data == NULL)
4124                 goto out;
4125
4126         if (S_ISFIFO(mode))
4127                 data->arg.ftype = NF4FIFO;
4128         else if (S_ISBLK(mode)) {
4129                 data->arg.ftype = NF4BLK;
4130                 data->arg.u.device.specdata1 = MAJOR(rdev);
4131                 data->arg.u.device.specdata2 = MINOR(rdev);
4132         }
4133         else if (S_ISCHR(mode)) {
4134                 data->arg.ftype = NF4CHR;
4135                 data->arg.u.device.specdata1 = MAJOR(rdev);
4136                 data->arg.u.device.specdata2 = MINOR(rdev);
4137         } else if (!S_ISSOCK(mode)) {
4138                 status = -EINVAL;
4139                 goto out_free;
4140         }
4141
4142         data->arg.label = label;
4143         status = nfs4_do_create(dir, dentry, data);
4144 out_free:
4145         nfs4_free_createdata(data);
4146 out:
4147         return status;
4148 }
4149
4150 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4151                 struct iattr *sattr, dev_t rdev)
4152 {
4153         struct nfs4_exception exception = { };
4154         struct nfs4_label l, *label = NULL;
4155         int err;
4156
4157         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4158
4159         sattr->ia_mode &= ~current_umask();
4160         do {
4161                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4162                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4163                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4164                                 &exception);
4165         } while (exception.retry);
4166
4167         nfs4_label_release_security(label);
4168
4169         return err;
4170 }
4171
4172 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4173                  struct nfs_fsstat *fsstat)
4174 {
4175         struct nfs4_statfs_arg args = {
4176                 .fh = fhandle,
4177                 .bitmask = server->attr_bitmask,
4178         };
4179         struct nfs4_statfs_res res = {
4180                 .fsstat = fsstat,
4181         };
4182         struct rpc_message msg = {
4183                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4184                 .rpc_argp = &args,
4185                 .rpc_resp = &res,
4186         };
4187
4188         nfs_fattr_init(fsstat->fattr);
4189         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4190 }
4191
4192 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4193 {
4194         struct nfs4_exception exception = { };
4195         int err;
4196         do {
4197                 err = nfs4_handle_exception(server,
4198                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4199                                 &exception);
4200         } while (exception.retry);
4201         return err;
4202 }
4203
4204 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4205                 struct nfs_fsinfo *fsinfo)
4206 {
4207         struct nfs4_fsinfo_arg args = {
4208                 .fh = fhandle,
4209                 .bitmask = server->attr_bitmask,
4210         };
4211         struct nfs4_fsinfo_res res = {
4212                 .fsinfo = fsinfo,
4213         };
4214         struct rpc_message msg = {
4215                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4216                 .rpc_argp = &args,
4217                 .rpc_resp = &res,
4218         };
4219
4220         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4221 }
4222
4223 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4224 {
4225         struct nfs4_exception exception = { };
4226         unsigned long now = jiffies;
4227         int err;
4228
4229         do {
4230                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4231                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4232                 if (err == 0) {
4233                         struct nfs_client *clp = server->nfs_client;
4234
4235                         spin_lock(&clp->cl_lock);
4236                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4237                         clp->cl_last_renewal = now;
4238                         spin_unlock(&clp->cl_lock);
4239                         break;
4240                 }
4241                 err = nfs4_handle_exception(server, err, &exception);
4242         } while (exception.retry);
4243         return err;
4244 }
4245
4246 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4247 {
4248         int error;
4249
4250         nfs_fattr_init(fsinfo->fattr);
4251         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4252         if (error == 0) {
4253                 /* block layout checks this! */
4254                 server->pnfs_blksize = fsinfo->blksize;
4255                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4256         }
4257
4258         return error;
4259 }
4260
4261 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4262                 struct nfs_pathconf *pathconf)
4263 {
4264         struct nfs4_pathconf_arg args = {
4265                 .fh = fhandle,
4266                 .bitmask = server->attr_bitmask,
4267         };
4268         struct nfs4_pathconf_res res = {
4269                 .pathconf = pathconf,
4270         };
4271         struct rpc_message msg = {
4272                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4273                 .rpc_argp = &args,
4274                 .rpc_resp = &res,
4275         };
4276
4277         /* None of the pathconf attributes are mandatory to implement */
4278         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4279                 memset(pathconf, 0, sizeof(*pathconf));
4280                 return 0;
4281         }
4282
4283         nfs_fattr_init(pathconf->fattr);
4284         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4285 }
4286
4287 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4288                 struct nfs_pathconf *pathconf)
4289 {
4290         struct nfs4_exception exception = { };
4291         int err;
4292
4293         do {
4294                 err = nfs4_handle_exception(server,
4295                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4296                                 &exception);
4297         } while (exception.retry);
4298         return err;
4299 }
4300
4301 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4302                 const struct nfs_open_context *ctx,
4303                 const struct nfs_lock_context *l_ctx,
4304                 fmode_t fmode)
4305 {
4306         const struct nfs_lockowner *lockowner = NULL;
4307
4308         if (l_ctx != NULL)
4309                 lockowner = &l_ctx->lockowner;
4310         return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL);
4311 }
4312 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4313
4314 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4315                 const struct nfs_open_context *ctx,
4316                 const struct nfs_lock_context *l_ctx,
4317                 fmode_t fmode)
4318 {
4319         nfs4_stateid current_stateid;
4320
4321         /* If the current stateid represents a lost lock, then exit */
4322         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4323                 return true;
4324         return nfs4_stateid_match(stateid, &current_stateid);
4325 }
4326
4327 static bool nfs4_error_stateid_expired(int err)
4328 {
4329         switch (err) {
4330         case -NFS4ERR_DELEG_REVOKED:
4331         case -NFS4ERR_ADMIN_REVOKED:
4332         case -NFS4ERR_BAD_STATEID:
4333         case -NFS4ERR_STALE_STATEID:
4334         case -NFS4ERR_OLD_STATEID:
4335         case -NFS4ERR_OPENMODE:
4336         case -NFS4ERR_EXPIRED:
4337                 return true;
4338         }
4339         return false;
4340 }
4341
4342 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4343 {
4344         nfs_invalidate_atime(hdr->inode);
4345 }
4346
4347 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4348 {
4349         struct nfs_server *server = NFS_SERVER(hdr->inode);
4350
4351         trace_nfs4_read(hdr, task->tk_status);
4352         if (nfs4_async_handle_error(task, server,
4353                                     hdr->args.context->state,
4354                                     NULL) == -EAGAIN) {
4355                 rpc_restart_call_prepare(task);
4356                 return -EAGAIN;
4357         }
4358
4359         __nfs4_read_done_cb(hdr);
4360         if (task->tk_status > 0)
4361                 renew_lease(server, hdr->timestamp);
4362         return 0;
4363 }
4364
4365 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4366                 struct nfs_pgio_args *args)
4367 {
4368
4369         if (!nfs4_error_stateid_expired(task->tk_status) ||
4370                 nfs4_stateid_is_current(&args->stateid,
4371                                 args->context,
4372                                 args->lock_context,
4373                                 FMODE_READ))
4374                 return false;
4375         rpc_restart_call_prepare(task);
4376         return true;
4377 }
4378
4379 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4380 {
4381
4382         dprintk("--> %s\n", __func__);
4383
4384         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4385                 return -EAGAIN;
4386         if (nfs4_read_stateid_changed(task, &hdr->args))
4387                 return -EAGAIN;
4388         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4389                                     nfs4_read_done_cb(task, hdr);
4390 }
4391
4392 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4393                                  struct rpc_message *msg)
4394 {
4395         hdr->timestamp   = jiffies;
4396         hdr->pgio_done_cb = nfs4_read_done_cb;
4397         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4398         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4399 }
4400
4401 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4402                                       struct nfs_pgio_header *hdr)
4403 {
4404         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4405                         &hdr->args.seq_args,
4406                         &hdr->res.seq_res,
4407                         task))
4408                 return 0;
4409         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4410                                 hdr->args.lock_context,
4411                                 hdr->rw_ops->rw_mode) == -EIO)
4412                 return -EIO;
4413         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4414                 return -EIO;
4415         return 0;
4416 }
4417
4418 static int nfs4_write_done_cb(struct rpc_task *task,
4419                               struct nfs_pgio_header *hdr)
4420 {
4421         struct inode *inode = hdr->inode;
4422
4423         trace_nfs4_write(hdr, task->tk_status);
4424         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4425                                     hdr->args.context->state,
4426                                     NULL) == -EAGAIN) {
4427                 rpc_restart_call_prepare(task);
4428                 return -EAGAIN;
4429         }
4430         if (task->tk_status >= 0) {
4431                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4432                 nfs_writeback_update_inode(hdr);
4433         }
4434         return 0;
4435 }
4436
4437 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4438                 struct nfs_pgio_args *args)
4439 {
4440
4441         if (!nfs4_error_stateid_expired(task->tk_status) ||
4442                 nfs4_stateid_is_current(&args->stateid,
4443                                 args->context,
4444                                 args->lock_context,
4445                                 FMODE_WRITE))
4446                 return false;
4447         rpc_restart_call_prepare(task);
4448         return true;
4449 }
4450
4451 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4452 {
4453         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4454                 return -EAGAIN;
4455         if (nfs4_write_stateid_changed(task, &hdr->args))
4456                 return -EAGAIN;
4457         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4458                 nfs4_write_done_cb(task, hdr);
4459 }
4460
4461 static
4462 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4463 {
4464         /* Don't request attributes for pNFS or O_DIRECT writes */
4465         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4466                 return false;
4467         /* Otherwise, request attributes if and only if we don't hold
4468          * a delegation
4469          */
4470         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4471 }
4472
4473 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4474                                   struct rpc_message *msg)
4475 {
4476         struct nfs_server *server = NFS_SERVER(hdr->inode);
4477
4478         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4479                 hdr->args.bitmask = NULL;
4480                 hdr->res.fattr = NULL;
4481         } else
4482                 hdr->args.bitmask = server->cache_consistency_bitmask;
4483
4484         if (!hdr->pgio_done_cb)
4485                 hdr->pgio_done_cb = nfs4_write_done_cb;
4486         hdr->res.server = server;
4487         hdr->timestamp   = jiffies;
4488
4489         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4490         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4491 }
4492
4493 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4494 {
4495         nfs4_setup_sequence(NFS_SERVER(data->inode),
4496                         &data->args.seq_args,
4497                         &data->res.seq_res,
4498                         task);
4499 }
4500
4501 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4502 {
4503         struct inode *inode = data->inode;
4504
4505         trace_nfs4_commit(data, task->tk_status);
4506         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4507                                     NULL, NULL) == -EAGAIN) {
4508                 rpc_restart_call_prepare(task);
4509                 return -EAGAIN;
4510         }
4511         return 0;
4512 }
4513
4514 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4515 {
4516         if (!nfs4_sequence_done(task, &data->res.seq_res))
4517                 return -EAGAIN;
4518         return data->commit_done_cb(task, data);
4519 }
4520
4521 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4522 {
4523         struct nfs_server *server = NFS_SERVER(data->inode);
4524
4525         if (data->commit_done_cb == NULL)
4526                 data->commit_done_cb = nfs4_commit_done_cb;
4527         data->res.server = server;
4528         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4529         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4530 }
4531
4532 struct nfs4_renewdata {
4533         struct nfs_client       *client;
4534         unsigned long           timestamp;
4535 };
4536
4537 /*
4538  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4539  * standalone procedure for queueing an asynchronous RENEW.
4540  */
4541 static void nfs4_renew_release(void *calldata)
4542 {
4543         struct nfs4_renewdata *data = calldata;
4544         struct nfs_client *clp = data->client;
4545
4546         if (atomic_read(&clp->cl_count) > 1)
4547                 nfs4_schedule_state_renewal(clp);
4548         nfs_put_client(clp);
4549         kfree(data);
4550 }
4551
4552 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4553 {
4554         struct nfs4_renewdata *data = calldata;
4555         struct nfs_client *clp = data->client;
4556         unsigned long timestamp = data->timestamp;
4557
4558         trace_nfs4_renew_async(clp, task->tk_status);
4559         switch (task->tk_status) {
4560         case 0:
4561                 break;
4562         case -NFS4ERR_LEASE_MOVED:
4563                 nfs4_schedule_lease_moved_recovery(clp);
4564                 break;
4565         default:
4566                 /* Unless we're shutting down, schedule state recovery! */
4567                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4568                         return;
4569                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4570                         nfs4_schedule_lease_recovery(clp);
4571                         return;
4572                 }
4573                 nfs4_schedule_path_down_recovery(clp);
4574         }
4575         do_renew_lease(clp, timestamp);
4576 }
4577
4578 static const struct rpc_call_ops nfs4_renew_ops = {
4579         .rpc_call_done = nfs4_renew_done,
4580         .rpc_release = nfs4_renew_release,
4581 };
4582
4583 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4584 {
4585         struct rpc_message msg = {
4586                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4587                 .rpc_argp       = clp,
4588                 .rpc_cred       = cred,
4589         };
4590         struct nfs4_renewdata *data;
4591
4592         if (renew_flags == 0)
4593                 return 0;
4594         if (!atomic_inc_not_zero(&clp->cl_count))
4595                 return -EIO;
4596         data = kmalloc(sizeof(*data), GFP_NOFS);
4597         if (data == NULL)
4598                 return -ENOMEM;
4599         data->client = clp;
4600         data->timestamp = jiffies;
4601         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4602                         &nfs4_renew_ops, data);
4603 }
4604
4605 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4606 {
4607         struct rpc_message msg = {
4608                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4609                 .rpc_argp       = clp,
4610                 .rpc_cred       = cred,
4611         };
4612         unsigned long now = jiffies;
4613         int status;
4614
4615         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4616         if (status < 0)
4617                 return status;
4618         do_renew_lease(clp, now);
4619         return 0;
4620 }
4621
4622 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4623 {
4624         return server->caps & NFS_CAP_ACLS;
4625 }
4626
4627 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4628  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4629  * the stack.
4630  */
4631 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4632
4633 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4634                 struct page **pages)
4635 {
4636         struct page *newpage, **spages;
4637         int rc = 0;
4638         size_t len;
4639         spages = pages;
4640
4641         do {
4642                 len = min_t(size_t, PAGE_SIZE, buflen);
4643                 newpage = alloc_page(GFP_KERNEL);
4644
4645                 if (newpage == NULL)
4646                         goto unwind;
4647                 memcpy(page_address(newpage), buf, len);
4648                 buf += len;
4649                 buflen -= len;
4650                 *pages++ = newpage;
4651                 rc++;
4652         } while (buflen != 0);
4653
4654         return rc;
4655
4656 unwind:
4657         for(; rc > 0; rc--)
4658                 __free_page(spages[rc-1]);
4659         return -ENOMEM;
4660 }
4661
4662 struct nfs4_cached_acl {
4663         int cached;
4664         size_t len;
4665         char data[0];
4666 };
4667
4668 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4669 {
4670         struct nfs_inode *nfsi = NFS_I(inode);
4671
4672         spin_lock(&inode->i_lock);
4673         kfree(nfsi->nfs4_acl);
4674         nfsi->nfs4_acl = acl;
4675         spin_unlock(&inode->i_lock);
4676 }
4677
4678 static void nfs4_zap_acl_attr(struct inode *inode)
4679 {
4680         nfs4_set_cached_acl(inode, NULL);
4681 }
4682
4683 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4684 {
4685         struct nfs_inode *nfsi = NFS_I(inode);
4686         struct nfs4_cached_acl *acl;
4687         int ret = -ENOENT;
4688
4689         spin_lock(&inode->i_lock);
4690         acl = nfsi->nfs4_acl;
4691         if (acl == NULL)
4692                 goto out;
4693         if (buf == NULL) /* user is just asking for length */
4694                 goto out_len;
4695         if (acl->cached == 0)
4696                 goto out;
4697         ret = -ERANGE; /* see getxattr(2) man page */
4698         if (acl->len > buflen)
4699                 goto out;
4700         memcpy(buf, acl->data, acl->len);
4701 out_len:
4702         ret = acl->len;
4703 out:
4704         spin_unlock(&inode->i_lock);
4705         return ret;
4706 }
4707
4708 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4709 {
4710         struct nfs4_cached_acl *acl;
4711         size_t buflen = sizeof(*acl) + acl_len;
4712
4713         if (buflen <= PAGE_SIZE) {
4714                 acl = kmalloc(buflen, GFP_KERNEL);
4715                 if (acl == NULL)
4716                         goto out;
4717                 acl->cached = 1;
4718                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4719         } else {
4720                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4721                 if (acl == NULL)
4722                         goto out;
4723                 acl->cached = 0;
4724         }
4725         acl->len = acl_len;
4726 out:
4727         nfs4_set_cached_acl(inode, acl);
4728 }
4729
4730 /*
4731  * The getxattr API returns the required buffer length when called with a
4732  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4733  * the required buf.  On a NULL buf, we send a page of data to the server
4734  * guessing that the ACL request can be serviced by a page. If so, we cache
4735  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4736  * the cache. If not so, we throw away the page, and cache the required
4737  * length. The next getxattr call will then produce another round trip to
4738  * the server, this time with the input buf of the required size.
4739  */
4740 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4741 {
4742         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4743         struct nfs_getaclargs args = {
4744                 .fh = NFS_FH(inode),
4745                 .acl_pages = pages,
4746                 .acl_len = buflen,
4747         };
4748         struct nfs_getaclres res = {
4749                 .acl_len = buflen,
4750         };
4751         struct rpc_message msg = {
4752                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4753                 .rpc_argp = &args,
4754                 .rpc_resp = &res,
4755         };
4756         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4757         int ret = -ENOMEM, i;
4758
4759         /* As long as we're doing a round trip to the server anyway,
4760          * let's be prepared for a page of acl data. */
4761         if (npages == 0)
4762                 npages = 1;
4763         if (npages > ARRAY_SIZE(pages))
4764                 return -ERANGE;
4765
4766         for (i = 0; i < npages; i++) {
4767                 pages[i] = alloc_page(GFP_KERNEL);
4768                 if (!pages[i])
4769                         goto out_free;
4770         }
4771
4772         /* for decoding across pages */
4773         res.acl_scratch = alloc_page(GFP_KERNEL);
4774         if (!res.acl_scratch)
4775                 goto out_free;
4776
4777         args.acl_len = npages * PAGE_SIZE;
4778
4779         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4780                 __func__, buf, buflen, npages, args.acl_len);
4781         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4782                              &msg, &args.seq_args, &res.seq_res, 0);
4783         if (ret)
4784                 goto out_free;
4785
4786         /* Handle the case where the passed-in buffer is too short */
4787         if (res.acl_flags & NFS4_ACL_TRUNC) {
4788                 /* Did the user only issue a request for the acl length? */
4789                 if (buf == NULL)
4790                         goto out_ok;
4791                 ret = -ERANGE;
4792                 goto out_free;
4793         }
4794         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4795         if (buf) {
4796                 if (res.acl_len > buflen) {
4797                         ret = -ERANGE;
4798                         goto out_free;
4799                 }
4800                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4801         }
4802 out_ok:
4803         ret = res.acl_len;
4804 out_free:
4805         for (i = 0; i < npages; i++)
4806                 if (pages[i])
4807                         __free_page(pages[i]);
4808         if (res.acl_scratch)
4809                 __free_page(res.acl_scratch);
4810         return ret;
4811 }
4812
4813 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4814 {
4815         struct nfs4_exception exception = { };
4816         ssize_t ret;
4817         do {
4818                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4819                 trace_nfs4_get_acl(inode, ret);
4820                 if (ret >= 0)
4821                         break;
4822                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4823         } while (exception.retry);
4824         return ret;
4825 }
4826
4827 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4828 {
4829         struct nfs_server *server = NFS_SERVER(inode);
4830         int ret;
4831
4832         if (!nfs4_server_supports_acls(server))
4833                 return -EOPNOTSUPP;
4834         ret = nfs_revalidate_inode(server, inode);
4835         if (ret < 0)
4836                 return ret;
4837         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4838                 nfs_zap_acl_cache(inode);
4839         ret = nfs4_read_cached_acl(inode, buf, buflen);
4840         if (ret != -ENOENT)
4841                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4842                  * but no cached acl data, just the acl length */
4843                 return ret;
4844         return nfs4_get_acl_uncached(inode, buf, buflen);
4845 }
4846
4847 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4848 {
4849         struct nfs_server *server = NFS_SERVER(inode);
4850         struct page *pages[NFS4ACL_MAXPAGES];
4851         struct nfs_setaclargs arg = {
4852                 .fh             = NFS_FH(inode),
4853                 .acl_pages      = pages,
4854                 .acl_len        = buflen,
4855         };
4856         struct nfs_setaclres res;
4857         struct rpc_message msg = {
4858                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4859                 .rpc_argp       = &arg,
4860                 .rpc_resp       = &res,
4861         };
4862         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4863         int ret, i;
4864
4865         if (!nfs4_server_supports_acls(server))
4866                 return -EOPNOTSUPP;
4867         if (npages > ARRAY_SIZE(pages))
4868                 return -ERANGE;
4869         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4870         if (i < 0)
4871                 return i;
4872         nfs4_inode_return_delegation(inode);
4873         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4874
4875         /*
4876          * Free each page after tx, so the only ref left is
4877          * held by the network stack
4878          */
4879         for (; i > 0; i--)
4880                 put_page(pages[i-1]);
4881
4882         /*
4883          * Acl update can result in inode attribute update.
4884          * so mark the attribute cache invalid.
4885          */
4886         spin_lock(&inode->i_lock);
4887         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4888         spin_unlock(&inode->i_lock);
4889         nfs_access_zap_cache(inode);
4890         nfs_zap_acl_cache(inode);
4891         return ret;
4892 }
4893
4894 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4895 {
4896         struct nfs4_exception exception = { };
4897         int err;
4898         do {
4899                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4900                 trace_nfs4_set_acl(inode, err);
4901                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4902                                 &exception);
4903         } while (exception.retry);
4904         return err;
4905 }
4906
4907 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4908 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4909                                         size_t buflen)
4910 {
4911         struct nfs_server *server = NFS_SERVER(inode);
4912         struct nfs_fattr fattr;
4913         struct nfs4_label label = {0, 0, buflen, buf};
4914
4915         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4916         struct nfs4_getattr_arg arg = {
4917                 .fh             = NFS_FH(inode),
4918                 .bitmask        = bitmask,
4919         };
4920         struct nfs4_getattr_res res = {
4921                 .fattr          = &fattr,
4922                 .label          = &label,
4923                 .server         = server,
4924         };
4925         struct rpc_message msg = {
4926                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4927                 .rpc_argp       = &arg,
4928                 .rpc_resp       = &res,
4929         };
4930         int ret;
4931
4932         nfs_fattr_init(&fattr);
4933
4934         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4935         if (ret)
4936                 return ret;
4937         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4938                 return -ENOENT;
4939         if (buflen < label.len)
4940                 return -ERANGE;
4941         return 0;
4942 }
4943
4944 static int nfs4_get_security_label(struct inode *inode, void *buf,
4945                                         size_t buflen)
4946 {
4947         struct nfs4_exception exception = { };
4948         int err;
4949
4950         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4951                 return -EOPNOTSUPP;
4952
4953         do {
4954                 err = _nfs4_get_security_label(inode, buf, buflen);
4955                 trace_nfs4_get_security_label(inode, err);
4956                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4957                                 &exception);
4958         } while (exception.retry);
4959         return err;
4960 }
4961
4962 static int _nfs4_do_set_security_label(struct inode *inode,
4963                 struct nfs4_label *ilabel,
4964                 struct nfs_fattr *fattr,
4965                 struct nfs4_label *olabel)
4966 {
4967
4968         struct iattr sattr = {0};
4969         struct nfs_server *server = NFS_SERVER(inode);
4970         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4971         struct nfs_setattrargs arg = {
4972                 .fh             = NFS_FH(inode),
4973                 .iap            = &sattr,
4974                 .server         = server,
4975                 .bitmask        = bitmask,
4976                 .label          = ilabel,
4977         };
4978         struct nfs_setattrres res = {
4979                 .fattr          = fattr,
4980                 .label          = olabel,
4981                 .server         = server,
4982         };
4983         struct rpc_message msg = {
4984                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4985                 .rpc_argp       = &arg,
4986                 .rpc_resp       = &res,
4987         };
4988         int status;
4989
4990         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4991
4992         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4993         if (status)
4994                 dprintk("%s failed: %d\n", __func__, status);
4995
4996         return status;
4997 }
4998
4999 static int nfs4_do_set_security_label(struct inode *inode,
5000                 struct nfs4_label *ilabel,
5001                 struct nfs_fattr *fattr,
5002                 struct nfs4_label *olabel)
5003 {
5004         struct nfs4_exception exception = { };
5005         int err;
5006
5007         do {
5008                 err = _nfs4_do_set_security_label(inode, ilabel,
5009                                 fattr, olabel);
5010                 trace_nfs4_set_security_label(inode, err);
5011                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5012                                 &exception);
5013         } while (exception.retry);
5014         return err;
5015 }
5016
5017 static int
5018 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5019 {
5020         struct nfs4_label ilabel, *olabel = NULL;
5021         struct nfs_fattr fattr;
5022         struct rpc_cred *cred;
5023         int status;
5024
5025         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5026                 return -EOPNOTSUPP;
5027
5028         nfs_fattr_init(&fattr);
5029
5030         ilabel.pi = 0;
5031         ilabel.lfs = 0;
5032         ilabel.label = (char *)buf;
5033         ilabel.len = buflen;
5034
5035         cred = rpc_lookup_cred();
5036         if (IS_ERR(cred))
5037                 return PTR_ERR(cred);
5038
5039         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5040         if (IS_ERR(olabel)) {
5041                 status = -PTR_ERR(olabel);
5042                 goto out;
5043         }
5044
5045         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5046         if (status == 0)
5047                 nfs_setsecurity(inode, &fattr, olabel);
5048
5049         nfs4_label_free(olabel);
5050 out:
5051         put_rpccred(cred);
5052         return status;
5053 }
5054 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5055
5056
5057 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5058                                     nfs4_verifier *bootverf)
5059 {
5060         __be32 verf[2];
5061
5062         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5063                 /* An impossible timestamp guarantees this value
5064                  * will never match a generated boot time. */
5065                 verf[0] = 0;
5066                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5067         } else {
5068                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5069                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5070                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5071         }
5072         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5073 }
5074
5075 static int
5076 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5077 {
5078         size_t len;
5079         char *str;
5080
5081         if (clp->cl_owner_id != NULL)
5082                 return 0;
5083
5084         rcu_read_lock();
5085         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5086                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5087                 1 +
5088                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5089                 1;
5090         rcu_read_unlock();
5091
5092         if (len > NFS4_OPAQUE_LIMIT + 1)
5093                 return -EINVAL;
5094
5095         /*
5096          * Since this string is allocated at mount time, and held until the
5097          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5098          * about a memory-reclaim deadlock.
5099          */
5100         str = kmalloc(len, GFP_KERNEL);
5101         if (!str)
5102                 return -ENOMEM;
5103
5104         rcu_read_lock();
5105         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5106                         clp->cl_ipaddr,
5107                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5108                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5109         rcu_read_unlock();
5110
5111         clp->cl_owner_id = str;
5112         return 0;
5113 }
5114
5115 static int
5116 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5117 {
5118         size_t len;
5119         char *str;
5120
5121         len = 10 + 10 + 1 + 10 + 1 +
5122                 strlen(nfs4_client_id_uniquifier) + 1 +
5123                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5124
5125         if (len > NFS4_OPAQUE_LIMIT + 1)
5126                 return -EINVAL;
5127
5128         /*
5129          * Since this string is allocated at mount time, and held until the
5130          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5131          * about a memory-reclaim deadlock.
5132          */
5133         str = kmalloc(len, GFP_KERNEL);
5134         if (!str)
5135                 return -ENOMEM;
5136
5137         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5138                         clp->rpc_ops->version, clp->cl_minorversion,
5139                         nfs4_client_id_uniquifier,
5140                         clp->cl_rpcclient->cl_nodename);
5141         clp->cl_owner_id = str;
5142         return 0;
5143 }
5144
5145 static int
5146 nfs4_init_uniform_client_string(struct nfs_client *clp)
5147 {
5148         size_t len;
5149         char *str;
5150
5151         if (clp->cl_owner_id != NULL)
5152                 return 0;
5153
5154         if (nfs4_client_id_uniquifier[0] != '\0')
5155                 return nfs4_init_uniquifier_client_string(clp);
5156
5157         len = 10 + 10 + 1 + 10 + 1 +
5158                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5159
5160         if (len > NFS4_OPAQUE_LIMIT + 1)
5161                 return -EINVAL;
5162
5163         /*
5164          * Since this string is allocated at mount time, and held until the
5165          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5166          * about a memory-reclaim deadlock.
5167          */
5168         str = kmalloc(len, GFP_KERNEL);
5169         if (!str)
5170                 return -ENOMEM;
5171
5172         scnprintf(str, len, "Linux NFSv%u.%u %s",
5173                         clp->rpc_ops->version, clp->cl_minorversion,
5174                         clp->cl_rpcclient->cl_nodename);
5175         clp->cl_owner_id = str;
5176         return 0;
5177 }
5178
5179 /*
5180  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5181  * services.  Advertise one based on the address family of the
5182  * clientaddr.
5183  */
5184 static unsigned int
5185 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5186 {
5187         if (strchr(clp->cl_ipaddr, ':') != NULL)
5188                 return scnprintf(buf, len, "tcp6");
5189         else
5190                 return scnprintf(buf, len, "tcp");
5191 }
5192
5193 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5194 {
5195         struct nfs4_setclientid *sc = calldata;
5196
5197         if (task->tk_status == 0)
5198                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5199 }
5200
5201 static const struct rpc_call_ops nfs4_setclientid_ops = {
5202         .rpc_call_done = nfs4_setclientid_done,
5203 };
5204
5205 /**
5206  * nfs4_proc_setclientid - Negotiate client ID
5207  * @clp: state data structure
5208  * @program: RPC program for NFSv4 callback service
5209  * @port: IP port number for NFS4 callback service
5210  * @cred: RPC credential to use for this call
5211  * @res: where to place the result
5212  *
5213  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5214  */
5215 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5216                 unsigned short port, struct rpc_cred *cred,
5217                 struct nfs4_setclientid_res *res)
5218 {
5219         nfs4_verifier sc_verifier;
5220         struct nfs4_setclientid setclientid = {
5221                 .sc_verifier = &sc_verifier,
5222                 .sc_prog = program,
5223                 .sc_clnt = clp,
5224         };
5225         struct rpc_message msg = {
5226                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5227                 .rpc_argp = &setclientid,
5228                 .rpc_resp = res,
5229                 .rpc_cred = cred,
5230         };
5231         struct rpc_task *task;
5232         struct rpc_task_setup task_setup_data = {
5233                 .rpc_client = clp->cl_rpcclient,
5234                 .rpc_message = &msg,
5235                 .callback_ops = &nfs4_setclientid_ops,
5236                 .callback_data = &setclientid,
5237                 .flags = RPC_TASK_TIMEOUT,
5238         };
5239         int status;
5240
5241         /* nfs_client_id4 */
5242         nfs4_init_boot_verifier(clp, &sc_verifier);
5243
5244         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5245                 status = nfs4_init_uniform_client_string(clp);
5246         else
5247                 status = nfs4_init_nonuniform_client_string(clp);
5248
5249         if (status)
5250                 goto out;
5251
5252         /* cb_client4 */
5253         setclientid.sc_netid_len =
5254                                 nfs4_init_callback_netid(clp,
5255                                                 setclientid.sc_netid,
5256                                                 sizeof(setclientid.sc_netid));
5257         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5258                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5259                                 clp->cl_ipaddr, port >> 8, port & 255);
5260
5261         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5262                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5263                 clp->cl_owner_id);
5264         task = rpc_run_task(&task_setup_data);
5265         if (IS_ERR(task)) {
5266                 status = PTR_ERR(task);
5267                 goto out;
5268         }
5269         status = task->tk_status;
5270         if (setclientid.sc_cred) {
5271                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5272                 put_rpccred(setclientid.sc_cred);
5273         }
5274         rpc_put_task(task);
5275 out:
5276         trace_nfs4_setclientid(clp, status);
5277         dprintk("NFS reply setclientid: %d\n", status);
5278         return status;
5279 }
5280
5281 /**
5282  * nfs4_proc_setclientid_confirm - Confirm client ID
5283  * @clp: state data structure
5284  * @res: result of a previous SETCLIENTID
5285  * @cred: RPC credential to use for this call
5286  *
5287  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5288  */
5289 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5290                 struct nfs4_setclientid_res *arg,
5291                 struct rpc_cred *cred)
5292 {
5293         struct rpc_message msg = {
5294                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5295                 .rpc_argp = arg,
5296                 .rpc_cred = cred,
5297         };
5298         int status;
5299
5300         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5301                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5302                 clp->cl_clientid);
5303         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5304         trace_nfs4_setclientid_confirm(clp, status);
5305         dprintk("NFS reply setclientid_confirm: %d\n", status);
5306         return status;
5307 }
5308
5309 struct nfs4_delegreturndata {
5310         struct nfs4_delegreturnargs args;
5311         struct nfs4_delegreturnres res;
5312         struct nfs_fh fh;
5313         nfs4_stateid stateid;
5314         unsigned long timestamp;
5315         struct nfs_fattr fattr;
5316         int rpc_status;
5317         struct inode *inode;
5318         bool roc;
5319         u32 roc_barrier;
5320 };
5321
5322 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5323 {
5324         struct nfs4_delegreturndata *data = calldata;
5325
5326         if (!nfs4_sequence_done(task, &data->res.seq_res))
5327                 return;
5328
5329         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5330         switch (task->tk_status) {
5331         case 0:
5332                 renew_lease(data->res.server, data->timestamp);
5333         case -NFS4ERR_ADMIN_REVOKED:
5334         case -NFS4ERR_DELEG_REVOKED:
5335         case -NFS4ERR_BAD_STATEID:
5336         case -NFS4ERR_OLD_STATEID:
5337         case -NFS4ERR_STALE_STATEID:
5338         case -NFS4ERR_EXPIRED:
5339                 task->tk_status = 0;
5340                 if (data->roc)
5341                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5342                 break;
5343         default:
5344                 if (nfs4_async_handle_error(task, data->res.server,
5345                                             NULL, NULL) == -EAGAIN) {
5346                         rpc_restart_call_prepare(task);
5347                         return;
5348                 }
5349         }
5350         data->rpc_status = task->tk_status;
5351 }
5352
5353 static void nfs4_delegreturn_release(void *calldata)
5354 {
5355         struct nfs4_delegreturndata *data = calldata;
5356         struct inode *inode = data->inode;
5357
5358         if (inode) {
5359                 if (data->roc)
5360                         pnfs_roc_release(inode);
5361                 nfs_iput_and_deactive(inode);
5362         }
5363         kfree(calldata);
5364 }
5365
5366 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5367 {
5368         struct nfs4_delegreturndata *d_data;
5369
5370         d_data = (struct nfs4_delegreturndata *)data;
5371
5372         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5373                 return;
5374
5375         if (d_data->roc)
5376                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5377
5378         nfs4_setup_sequence(d_data->res.server,
5379                         &d_data->args.seq_args,
5380                         &d_data->res.seq_res,
5381                         task);
5382 }
5383
5384 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5385         .rpc_call_prepare = nfs4_delegreturn_prepare,
5386         .rpc_call_done = nfs4_delegreturn_done,
5387         .rpc_release = nfs4_delegreturn_release,
5388 };
5389
5390 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5391 {
5392         struct nfs4_delegreturndata *data;
5393         struct nfs_server *server = NFS_SERVER(inode);
5394         struct rpc_task *task;
5395         struct rpc_message msg = {
5396                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5397                 .rpc_cred = cred,
5398         };
5399         struct rpc_task_setup task_setup_data = {
5400                 .rpc_client = server->client,
5401                 .rpc_message = &msg,
5402                 .callback_ops = &nfs4_delegreturn_ops,
5403                 .flags = RPC_TASK_ASYNC,
5404         };
5405         int status = 0;
5406
5407         data = kzalloc(sizeof(*data), GFP_NOFS);
5408         if (data == NULL)
5409                 return -ENOMEM;
5410         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5411
5412         nfs4_state_protect(server->nfs_client,
5413                         NFS_SP4_MACH_CRED_CLEANUP,
5414                         &task_setup_data.rpc_client, &msg);
5415
5416         data->args.fhandle = &data->fh;
5417         data->args.stateid = &data->stateid;
5418         data->args.bitmask = server->cache_consistency_bitmask;
5419         nfs_copy_fh(&data->fh, NFS_FH(inode));
5420         nfs4_stateid_copy(&data->stateid, stateid);
5421         data->res.fattr = &data->fattr;
5422         data->res.server = server;
5423         nfs_fattr_init(data->res.fattr);
5424         data->timestamp = jiffies;
5425         data->rpc_status = 0;
5426         data->inode = nfs_igrab_and_active(inode);
5427         if (data->inode)
5428                 data->roc = nfs4_roc(inode);
5429
5430         task_setup_data.callback_data = data;
5431         msg.rpc_argp = &data->args;
5432         msg.rpc_resp = &data->res;
5433         task = rpc_run_task(&task_setup_data);
5434         if (IS_ERR(task))
5435                 return PTR_ERR(task);
5436         if (!issync)
5437                 goto out;
5438         status = nfs4_wait_for_completion_rpc_task(task);
5439         if (status != 0)
5440                 goto out;
5441         status = data->rpc_status;
5442         if (status == 0)
5443                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5444         else
5445                 nfs_refresh_inode(inode, &data->fattr);
5446 out:
5447         rpc_put_task(task);
5448         return status;
5449 }
5450
5451 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5452 {
5453         struct nfs_server *server = NFS_SERVER(inode);
5454         struct nfs4_exception exception = { };
5455         int err;
5456         do {
5457                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5458                 trace_nfs4_delegreturn(inode, stateid, err);
5459                 switch (err) {
5460                         case -NFS4ERR_STALE_STATEID:
5461                         case -NFS4ERR_EXPIRED:
5462                         case 0:
5463                                 return 0;
5464                 }
5465                 err = nfs4_handle_exception(server, err, &exception);
5466         } while (exception.retry);
5467         return err;
5468 }
5469
5470 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5471 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5472
5473 /* 
5474  * sleep, with exponential backoff, and retry the LOCK operation. 
5475  */
5476 static unsigned long
5477 nfs4_set_lock_task_retry(unsigned long timeout)
5478 {
5479         freezable_schedule_timeout_killable_unsafe(timeout);
5480         timeout <<= 1;
5481         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5482                 return NFS4_LOCK_MAXTIMEOUT;
5483         return timeout;
5484 }
5485
5486 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5487 {
5488         struct inode *inode = state->inode;
5489         struct nfs_server *server = NFS_SERVER(inode);
5490         struct nfs_client *clp = server->nfs_client;
5491         struct nfs_lockt_args arg = {
5492                 .fh = NFS_FH(inode),
5493                 .fl = request,
5494         };
5495         struct nfs_lockt_res res = {
5496                 .denied = request,
5497         };
5498         struct rpc_message msg = {
5499                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5500                 .rpc_argp       = &arg,
5501                 .rpc_resp       = &res,
5502                 .rpc_cred       = state->owner->so_cred,
5503         };
5504         struct nfs4_lock_state *lsp;
5505         int status;
5506
5507         arg.lock_owner.clientid = clp->cl_clientid;
5508         status = nfs4_set_lock_state(state, request);
5509         if (status != 0)
5510                 goto out;
5511         lsp = request->fl_u.nfs4_fl.owner;
5512         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5513         arg.lock_owner.s_dev = server->s_dev;
5514         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5515         switch (status) {
5516                 case 0:
5517                         request->fl_type = F_UNLCK;
5518                         break;
5519                 case -NFS4ERR_DENIED:
5520                         status = 0;
5521         }
5522         request->fl_ops->fl_release_private(request);
5523         request->fl_ops = NULL;
5524 out:
5525         return status;
5526 }
5527
5528 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5529 {
5530         struct nfs4_exception exception = { };
5531         int err;
5532
5533         do {
5534                 err = _nfs4_proc_getlk(state, cmd, request);
5535                 trace_nfs4_get_lock(request, state, cmd, err);
5536                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5537                                 &exception);
5538         } while (exception.retry);
5539         return err;
5540 }
5541
5542 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5543 {
5544         return locks_lock_inode_wait(inode, fl);
5545 }
5546
5547 struct nfs4_unlockdata {
5548         struct nfs_locku_args arg;
5549         struct nfs_locku_res res;
5550         struct nfs4_lock_state *lsp;
5551         struct nfs_open_context *ctx;
5552         struct file_lock fl;
5553         struct nfs_server *server;
5554         unsigned long timestamp;
5555 };
5556
5557 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5558                 struct nfs_open_context *ctx,
5559                 struct nfs4_lock_state *lsp,
5560                 struct nfs_seqid *seqid)
5561 {
5562         struct nfs4_unlockdata *p;
5563         struct inode *inode = lsp->ls_state->inode;
5564
5565         p = kzalloc(sizeof(*p), GFP_NOFS);
5566         if (p == NULL)
5567                 return NULL;
5568         p->arg.fh = NFS_FH(inode);
5569         p->arg.fl = &p->fl;
5570         p->arg.seqid = seqid;
5571         p->res.seqid = seqid;
5572         p->lsp = lsp;
5573         atomic_inc(&lsp->ls_count);
5574         /* Ensure we don't close file until we're done freeing locks! */
5575         p->ctx = get_nfs_open_context(ctx);
5576         memcpy(&p->fl, fl, sizeof(p->fl));
5577         p->server = NFS_SERVER(inode);
5578         return p;
5579 }
5580
5581 static void nfs4_locku_release_calldata(void *data)
5582 {
5583         struct nfs4_unlockdata *calldata = data;
5584         nfs_free_seqid(calldata->arg.seqid);
5585         nfs4_put_lock_state(calldata->lsp);
5586         put_nfs_open_context(calldata->ctx);
5587         kfree(calldata);
5588 }
5589
5590 static void nfs4_locku_done(struct rpc_task *task, void *data)
5591 {
5592         struct nfs4_unlockdata *calldata = data;
5593
5594         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5595                 return;
5596         switch (task->tk_status) {
5597                 case 0:
5598                         renew_lease(calldata->server, calldata->timestamp);
5599                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5600                         if (nfs4_update_lock_stateid(calldata->lsp,
5601                                         &calldata->res.stateid))
5602                                 break;
5603                 case -NFS4ERR_BAD_STATEID:
5604                 case -NFS4ERR_OLD_STATEID:
5605                 case -NFS4ERR_STALE_STATEID:
5606                 case -NFS4ERR_EXPIRED:
5607                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5608                                                 &calldata->lsp->ls_stateid))
5609                                 rpc_restart_call_prepare(task);
5610                         break;
5611                 default:
5612                         if (nfs4_async_handle_error(task, calldata->server,
5613                                                     NULL, NULL) == -EAGAIN)
5614                                 rpc_restart_call_prepare(task);
5615         }
5616         nfs_release_seqid(calldata->arg.seqid);
5617 }
5618
5619 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5620 {
5621         struct nfs4_unlockdata *calldata = data;
5622
5623         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5624                 goto out_wait;
5625         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5626         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5627                 /* Note: exit _without_ running nfs4_locku_done */
5628                 goto out_no_action;
5629         }
5630         calldata->timestamp = jiffies;
5631         if (nfs4_setup_sequence(calldata->server,
5632                                 &calldata->arg.seq_args,
5633                                 &calldata->res.seq_res,
5634                                 task) != 0)
5635                 nfs_release_seqid(calldata->arg.seqid);
5636         return;
5637 out_no_action:
5638         task->tk_action = NULL;
5639 out_wait:
5640         nfs4_sequence_done(task, &calldata->res.seq_res);
5641 }
5642
5643 static const struct rpc_call_ops nfs4_locku_ops = {
5644         .rpc_call_prepare = nfs4_locku_prepare,
5645         .rpc_call_done = nfs4_locku_done,
5646         .rpc_release = nfs4_locku_release_calldata,
5647 };
5648
5649 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5650                 struct nfs_open_context *ctx,
5651                 struct nfs4_lock_state *lsp,
5652                 struct nfs_seqid *seqid)
5653 {
5654         struct nfs4_unlockdata *data;
5655         struct rpc_message msg = {
5656                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5657                 .rpc_cred = ctx->cred,
5658         };
5659         struct rpc_task_setup task_setup_data = {
5660                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5661                 .rpc_message = &msg,
5662                 .callback_ops = &nfs4_locku_ops,
5663                 .workqueue = nfsiod_workqueue,
5664                 .flags = RPC_TASK_ASYNC,
5665         };
5666
5667         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5668                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5669
5670         /* Ensure this is an unlock - when canceling a lock, the
5671          * canceled lock is passed in, and it won't be an unlock.
5672          */
5673         fl->fl_type = F_UNLCK;
5674
5675         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5676         if (data == NULL) {
5677                 nfs_free_seqid(seqid);
5678                 return ERR_PTR(-ENOMEM);
5679         }
5680
5681         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5682         msg.rpc_argp = &data->arg;
5683         msg.rpc_resp = &data->res;
5684         task_setup_data.callback_data = data;
5685         return rpc_run_task(&task_setup_data);
5686 }
5687
5688 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5689 {
5690         struct inode *inode = state->inode;
5691         struct nfs4_state_owner *sp = state->owner;
5692         struct nfs_inode *nfsi = NFS_I(inode);
5693         struct nfs_seqid *seqid;
5694         struct nfs4_lock_state *lsp;
5695         struct rpc_task *task;
5696         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5697         int status = 0;
5698         unsigned char fl_flags = request->fl_flags;
5699
5700         status = nfs4_set_lock_state(state, request);
5701         /* Unlock _before_ we do the RPC call */
5702         request->fl_flags |= FL_EXISTS;
5703         /* Exclude nfs_delegation_claim_locks() */
5704         mutex_lock(&sp->so_delegreturn_mutex);
5705         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5706         down_read(&nfsi->rwsem);
5707         if (do_vfs_lock(inode, request) == -ENOENT) {
5708                 up_read(&nfsi->rwsem);
5709                 mutex_unlock(&sp->so_delegreturn_mutex);
5710                 goto out;
5711         }
5712         up_read(&nfsi->rwsem);
5713         mutex_unlock(&sp->so_delegreturn_mutex);
5714         if (status != 0)
5715                 goto out;
5716         /* Is this a delegated lock? */
5717         lsp = request->fl_u.nfs4_fl.owner;
5718         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5719                 goto out;
5720         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5721         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5722         status = -ENOMEM;
5723         if (IS_ERR(seqid))
5724                 goto out;
5725         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5726         status = PTR_ERR(task);
5727         if (IS_ERR(task))
5728                 goto out;
5729         status = nfs4_wait_for_completion_rpc_task(task);
5730         rpc_put_task(task);
5731 out:
5732         request->fl_flags = fl_flags;
5733         trace_nfs4_unlock(request, state, F_SETLK, status);
5734         return status;
5735 }
5736
5737 struct nfs4_lockdata {
5738         struct nfs_lock_args arg;
5739         struct nfs_lock_res res;
5740         struct nfs4_lock_state *lsp;
5741         struct nfs_open_context *ctx;
5742         struct file_lock fl;
5743         unsigned long timestamp;
5744         int rpc_status;
5745         int cancelled;
5746         struct nfs_server *server;
5747 };
5748
5749 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5750                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5751                 gfp_t gfp_mask)
5752 {
5753         struct nfs4_lockdata *p;
5754         struct inode *inode = lsp->ls_state->inode;
5755         struct nfs_server *server = NFS_SERVER(inode);
5756         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5757
5758         p = kzalloc(sizeof(*p), gfp_mask);
5759         if (p == NULL)
5760                 return NULL;
5761
5762         p->arg.fh = NFS_FH(inode);
5763         p->arg.fl = &p->fl;
5764         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5765         if (IS_ERR(p->arg.open_seqid))
5766                 goto out_free;
5767         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5768         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5769         if (IS_ERR(p->arg.lock_seqid))
5770                 goto out_free_seqid;
5771         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5772         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5773         p->arg.lock_owner.s_dev = server->s_dev;
5774         p->res.lock_seqid = p->arg.lock_seqid;
5775         p->lsp = lsp;
5776         p->server = server;
5777         atomic_inc(&lsp->ls_count);
5778         p->ctx = get_nfs_open_context(ctx);
5779         get_file(fl->fl_file);
5780         memcpy(&p->fl, fl, sizeof(p->fl));
5781         return p;
5782 out_free_seqid:
5783         nfs_free_seqid(p->arg.open_seqid);
5784 out_free:
5785         kfree(p);
5786         return NULL;
5787 }
5788
5789 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5790 {
5791         struct nfs4_lockdata *data = calldata;
5792         struct nfs4_state *state = data->lsp->ls_state;
5793
5794         dprintk("%s: begin!\n", __func__);
5795         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5796                 goto out_wait;
5797         /* Do we need to do an open_to_lock_owner? */
5798         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5799                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5800                         goto out_release_lock_seqid;
5801                 }
5802                 nfs4_stateid_copy(&data->arg.open_stateid,
5803                                 &state->open_stateid);
5804                 data->arg.new_lock_owner = 1;
5805                 data->res.open_seqid = data->arg.open_seqid;
5806         } else {
5807                 data->arg.new_lock_owner = 0;
5808                 nfs4_stateid_copy(&data->arg.lock_stateid,
5809                                 &data->lsp->ls_stateid);
5810         }
5811         if (!nfs4_valid_open_stateid(state)) {
5812                 data->rpc_status = -EBADF;
5813                 task->tk_action = NULL;
5814                 goto out_release_open_seqid;
5815         }
5816         data->timestamp = jiffies;
5817         if (nfs4_setup_sequence(data->server,
5818                                 &data->arg.seq_args,
5819                                 &data->res.seq_res,
5820                                 task) == 0)
5821                 return;
5822 out_release_open_seqid:
5823         nfs_release_seqid(data->arg.open_seqid);
5824 out_release_lock_seqid:
5825         nfs_release_seqid(data->arg.lock_seqid);
5826 out_wait:
5827         nfs4_sequence_done(task, &data->res.seq_res);
5828         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5829 }
5830
5831 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5832 {
5833         struct nfs4_lockdata *data = calldata;
5834         struct nfs4_lock_state *lsp = data->lsp;
5835
5836         dprintk("%s: begin!\n", __func__);
5837
5838         if (!nfs4_sequence_done(task, &data->res.seq_res))
5839                 return;
5840
5841         data->rpc_status = task->tk_status;
5842         switch (task->tk_status) {
5843         case 0:
5844                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5845                                 data->timestamp);
5846                 if (data->arg.new_lock) {
5847                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5848                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5849                                 rpc_restart_call_prepare(task);
5850                                 break;
5851                         }
5852                 }
5853                 if (data->arg.new_lock_owner != 0) {
5854                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5855                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5856                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5857                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5858                         rpc_restart_call_prepare(task);
5859                 break;
5860         case -NFS4ERR_BAD_STATEID:
5861         case -NFS4ERR_OLD_STATEID:
5862         case -NFS4ERR_STALE_STATEID:
5863         case -NFS4ERR_EXPIRED:
5864                 if (data->arg.new_lock_owner != 0) {
5865                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5866                                                 &lsp->ls_state->open_stateid))
5867                                 rpc_restart_call_prepare(task);
5868                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5869                                                 &lsp->ls_stateid))
5870                                 rpc_restart_call_prepare(task);
5871         }
5872         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5873 }
5874
5875 static void nfs4_lock_release(void *calldata)
5876 {
5877         struct nfs4_lockdata *data = calldata;
5878
5879         dprintk("%s: begin!\n", __func__);
5880         nfs_free_seqid(data->arg.open_seqid);
5881         if (data->cancelled != 0) {
5882                 struct rpc_task *task;
5883                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5884                                 data->arg.lock_seqid);
5885                 if (!IS_ERR(task))
5886                         rpc_put_task_async(task);
5887                 dprintk("%s: cancelling lock!\n", __func__);
5888         } else
5889                 nfs_free_seqid(data->arg.lock_seqid);
5890         nfs4_put_lock_state(data->lsp);
5891         put_nfs_open_context(data->ctx);
5892         fput(data->fl.fl_file);
5893         kfree(data);
5894         dprintk("%s: done!\n", __func__);
5895 }
5896
5897 static const struct rpc_call_ops nfs4_lock_ops = {
5898         .rpc_call_prepare = nfs4_lock_prepare,
5899         .rpc_call_done = nfs4_lock_done,
5900         .rpc_release = nfs4_lock_release,
5901 };
5902
5903 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5904 {
5905         switch (error) {
5906         case -NFS4ERR_ADMIN_REVOKED:
5907         case -NFS4ERR_BAD_STATEID:
5908                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5909                 if (new_lock_owner != 0 ||
5910                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5911                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5912                 break;
5913         case -NFS4ERR_STALE_STATEID:
5914                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5915         case -NFS4ERR_EXPIRED:
5916                 nfs4_schedule_lease_recovery(server->nfs_client);
5917         };
5918 }
5919
5920 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5921 {
5922         struct nfs4_lockdata *data;
5923         struct rpc_task *task;
5924         struct rpc_message msg = {
5925                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5926                 .rpc_cred = state->owner->so_cred,
5927         };
5928         struct rpc_task_setup task_setup_data = {
5929                 .rpc_client = NFS_CLIENT(state->inode),
5930                 .rpc_message = &msg,
5931                 .callback_ops = &nfs4_lock_ops,
5932                 .workqueue = nfsiod_workqueue,
5933                 .flags = RPC_TASK_ASYNC,
5934         };
5935         int ret;
5936
5937         dprintk("%s: begin!\n", __func__);
5938         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5939                         fl->fl_u.nfs4_fl.owner,
5940                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5941         if (data == NULL)
5942                 return -ENOMEM;
5943         if (IS_SETLKW(cmd))
5944                 data->arg.block = 1;
5945         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5946         msg.rpc_argp = &data->arg;
5947         msg.rpc_resp = &data->res;
5948         task_setup_data.callback_data = data;
5949         if (recovery_type > NFS_LOCK_NEW) {
5950                 if (recovery_type == NFS_LOCK_RECLAIM)
5951                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5952                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5953         } else
5954                 data->arg.new_lock = 1;
5955         task = rpc_run_task(&task_setup_data);
5956         if (IS_ERR(task))
5957                 return PTR_ERR(task);
5958         ret = nfs4_wait_for_completion_rpc_task(task);
5959         if (ret == 0) {
5960                 ret = data->rpc_status;
5961                 if (ret)
5962                         nfs4_handle_setlk_error(data->server, data->lsp,
5963                                         data->arg.new_lock_owner, ret);
5964         } else
5965                 data->cancelled = 1;
5966         rpc_put_task(task);
5967         dprintk("%s: done, ret = %d!\n", __func__, ret);
5968         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
5969         return ret;
5970 }
5971
5972 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5973 {
5974         struct nfs_server *server = NFS_SERVER(state->inode);
5975         struct nfs4_exception exception = {
5976                 .inode = state->inode,
5977         };
5978         int err;
5979
5980         do {
5981                 /* Cache the lock if possible... */
5982                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5983                         return 0;
5984                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5985                 if (err != -NFS4ERR_DELAY)
5986                         break;
5987                 nfs4_handle_exception(server, err, &exception);
5988         } while (exception.retry);
5989         return err;
5990 }
5991
5992 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5993 {
5994         struct nfs_server *server = NFS_SERVER(state->inode);
5995         struct nfs4_exception exception = {
5996                 .inode = state->inode,
5997         };
5998         int err;
5999
6000         err = nfs4_set_lock_state(state, request);
6001         if (err != 0)
6002                 return err;
6003         if (!recover_lost_locks) {
6004                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6005                 return 0;
6006         }
6007         do {
6008                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6009                         return 0;
6010                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6011                 switch (err) {
6012                 default:
6013                         goto out;
6014                 case -NFS4ERR_GRACE:
6015                 case -NFS4ERR_DELAY:
6016                         nfs4_handle_exception(server, err, &exception);
6017                         err = 0;
6018                 }
6019         } while (exception.retry);
6020 out:
6021         return err;
6022 }
6023
6024 #if defined(CONFIG_NFS_V4_1)
6025 /**
6026  * nfs41_check_expired_locks - possibly free a lock stateid
6027  *
6028  * @state: NFSv4 state for an inode
6029  *
6030  * Returns NFS_OK if recovery for this stateid is now finished.
6031  * Otherwise a negative NFS4ERR value is returned.
6032  */
6033 static int nfs41_check_expired_locks(struct nfs4_state *state)
6034 {
6035         int status, ret = -NFS4ERR_BAD_STATEID;
6036         struct nfs4_lock_state *lsp;
6037         struct nfs_server *server = NFS_SERVER(state->inode);
6038
6039         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6040                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6041                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6042
6043                         status = nfs41_test_stateid(server,
6044                                         &lsp->ls_stateid,
6045                                         cred);
6046                         trace_nfs4_test_lock_stateid(state, lsp, status);
6047                         if (status != NFS_OK) {
6048                                 /* Free the stateid unless the server
6049                                  * informs us the stateid is unrecognized. */
6050                                 if (status != -NFS4ERR_BAD_STATEID)
6051                                         nfs41_free_stateid(server,
6052                                                         &lsp->ls_stateid,
6053                                                         cred);
6054                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6055                                 ret = status;
6056                         }
6057                 }
6058         };
6059
6060         return ret;
6061 }
6062
6063 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6064 {
6065         int status = NFS_OK;
6066
6067         if (test_bit(LK_STATE_IN_USE, &state->flags))
6068                 status = nfs41_check_expired_locks(state);
6069         if (status != NFS_OK)
6070                 status = nfs4_lock_expired(state, request);
6071         return status;
6072 }
6073 #endif
6074
6075 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6076 {
6077         struct nfs_inode *nfsi = NFS_I(state->inode);
6078         struct nfs4_state_owner *sp = state->owner;
6079         unsigned char fl_flags = request->fl_flags;
6080         int status = -ENOLCK;
6081
6082         if ((fl_flags & FL_POSIX) &&
6083                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6084                 goto out;
6085         /* Is this a delegated open? */
6086         status = nfs4_set_lock_state(state, request);
6087         if (status != 0)
6088                 goto out;
6089         request->fl_flags |= FL_ACCESS;
6090         status = do_vfs_lock(state->inode, request);
6091         if (status < 0)
6092                 goto out;
6093         mutex_lock(&sp->so_delegreturn_mutex);
6094         down_read(&nfsi->rwsem);
6095         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6096                 /* Yes: cache locks! */
6097                 /* ...but avoid races with delegation recall... */
6098                 request->fl_flags = fl_flags & ~FL_SLEEP;
6099                 status = do_vfs_lock(state->inode, request);
6100                 up_read(&nfsi->rwsem);
6101                 mutex_unlock(&sp->so_delegreturn_mutex);
6102                 goto out;
6103         }
6104         up_read(&nfsi->rwsem);
6105         mutex_unlock(&sp->so_delegreturn_mutex);
6106         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6107 out:
6108         request->fl_flags = fl_flags;
6109         return status;
6110 }
6111
6112 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6113 {
6114         struct nfs4_exception exception = {
6115                 .state = state,
6116                 .inode = state->inode,
6117         };
6118         int err;
6119
6120         do {
6121                 err = _nfs4_proc_setlk(state, cmd, request);
6122                 if (err == -NFS4ERR_DENIED)
6123                         err = -EAGAIN;
6124                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6125                                 err, &exception);
6126         } while (exception.retry);
6127         return err;
6128 }
6129
6130 static int
6131 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6132 {
6133         struct nfs_open_context *ctx;
6134         struct nfs4_state *state;
6135         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6136         int status;
6137
6138         /* verify open state */
6139         ctx = nfs_file_open_context(filp);
6140         state = ctx->state;
6141
6142         if (request->fl_start < 0 || request->fl_end < 0)
6143                 return -EINVAL;
6144
6145         if (IS_GETLK(cmd)) {
6146                 if (state != NULL)
6147                         return nfs4_proc_getlk(state, F_GETLK, request);
6148                 return 0;
6149         }
6150
6151         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6152                 return -EINVAL;
6153
6154         if (request->fl_type == F_UNLCK) {
6155                 if (state != NULL)
6156                         return nfs4_proc_unlck(state, cmd, request);
6157                 return 0;
6158         }
6159
6160         if (state == NULL)
6161                 return -ENOLCK;
6162         /*
6163          * Don't rely on the VFS having checked the file open mode,
6164          * since it won't do this for flock() locks.
6165          */
6166         switch (request->fl_type) {
6167         case F_RDLCK:
6168                 if (!(filp->f_mode & FMODE_READ))
6169                         return -EBADF;
6170                 break;
6171         case F_WRLCK:
6172                 if (!(filp->f_mode & FMODE_WRITE))
6173                         return -EBADF;
6174         }
6175
6176         do {
6177                 status = nfs4_proc_setlk(state, cmd, request);
6178                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6179                         break;
6180                 timeout = nfs4_set_lock_task_retry(timeout);
6181                 status = -ERESTARTSYS;
6182                 if (signalled())
6183                         break;
6184         } while(status < 0);
6185         return status;
6186 }
6187
6188 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6189 {
6190         struct nfs_server *server = NFS_SERVER(state->inode);
6191         int err;
6192
6193         err = nfs4_set_lock_state(state, fl);
6194         if (err != 0)
6195                 return err;
6196         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6197         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6198 }
6199
6200 struct nfs_release_lockowner_data {
6201         struct nfs4_lock_state *lsp;
6202         struct nfs_server *server;
6203         struct nfs_release_lockowner_args args;
6204         struct nfs_release_lockowner_res res;
6205         unsigned long timestamp;
6206 };
6207
6208 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6209 {
6210         struct nfs_release_lockowner_data *data = calldata;
6211         struct nfs_server *server = data->server;
6212         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6213                              &data->args.seq_args, &data->res.seq_res, task);
6214         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6215         data->timestamp = jiffies;
6216 }
6217
6218 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6219 {
6220         struct nfs_release_lockowner_data *data = calldata;
6221         struct nfs_server *server = data->server;
6222
6223         nfs40_sequence_done(task, &data->res.seq_res);
6224
6225         switch (task->tk_status) {
6226         case 0:
6227                 renew_lease(server, data->timestamp);
6228                 break;
6229         case -NFS4ERR_STALE_CLIENTID:
6230         case -NFS4ERR_EXPIRED:
6231                 nfs4_schedule_lease_recovery(server->nfs_client);
6232                 break;
6233         case -NFS4ERR_LEASE_MOVED:
6234         case -NFS4ERR_DELAY:
6235                 if (nfs4_async_handle_error(task, server,
6236                                             NULL, NULL) == -EAGAIN)
6237                         rpc_restart_call_prepare(task);
6238         }
6239 }
6240
6241 static void nfs4_release_lockowner_release(void *calldata)
6242 {
6243         struct nfs_release_lockowner_data *data = calldata;
6244         nfs4_free_lock_state(data->server, data->lsp);
6245         kfree(calldata);
6246 }
6247
6248 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6249         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6250         .rpc_call_done = nfs4_release_lockowner_done,
6251         .rpc_release = nfs4_release_lockowner_release,
6252 };
6253
6254 static void
6255 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6256 {
6257         struct nfs_release_lockowner_data *data;
6258         struct rpc_message msg = {
6259                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6260         };
6261
6262         if (server->nfs_client->cl_mvops->minor_version != 0)
6263                 return;
6264
6265         data = kmalloc(sizeof(*data), GFP_NOFS);
6266         if (!data)
6267                 return;
6268         data->lsp = lsp;
6269         data->server = server;
6270         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6271         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6272         data->args.lock_owner.s_dev = server->s_dev;
6273
6274         msg.rpc_argp = &data->args;
6275         msg.rpc_resp = &data->res;
6276         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6277         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6278 }
6279
6280 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6281
6282 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6283                                    struct dentry *unused, struct inode *inode,
6284                                    const char *key, const void *buf,
6285                                    size_t buflen, int flags)
6286 {
6287         return nfs4_proc_set_acl(inode, buf, buflen);
6288 }
6289
6290 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6291                                    struct dentry *unused, struct inode *inode,
6292                                    const char *key, void *buf, size_t buflen)
6293 {
6294         return nfs4_proc_get_acl(inode, buf, buflen);
6295 }
6296
6297 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6298 {
6299         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6300 }
6301
6302 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6303
6304 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6305                                      struct dentry *unused, struct inode *inode,
6306                                      const char *key, const void *buf,
6307                                      size_t buflen, int flags)
6308 {
6309         if (security_ismaclabel(key))
6310                 return nfs4_set_security_label(inode, buf, buflen);
6311
6312         return -EOPNOTSUPP;
6313 }
6314
6315 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6316                                      struct dentry *unused, struct inode *inode,
6317                                      const char *key, void *buf, size_t buflen)
6318 {
6319         if (security_ismaclabel(key))
6320                 return nfs4_get_security_label(inode, buf, buflen);
6321         return -EOPNOTSUPP;
6322 }
6323
6324 static ssize_t
6325 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6326 {
6327         int len = 0;
6328
6329         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6330                 len = security_inode_listsecurity(inode, list, list_len);
6331                 if (list_len && len > list_len)
6332                         return -ERANGE;
6333         }
6334         return len;
6335 }
6336
6337 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6338         .prefix = XATTR_SECURITY_PREFIX,
6339         .get    = nfs4_xattr_get_nfs4_label,
6340         .set    = nfs4_xattr_set_nfs4_label,
6341 };
6342
6343 #else
6344
6345 static ssize_t
6346 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6347 {
6348         return 0;
6349 }
6350
6351 #endif
6352
6353 /*
6354  * nfs_fhget will use either the mounted_on_fileid or the fileid
6355  */
6356 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6357 {
6358         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6359                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6360               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6361               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6362                 return;
6363
6364         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6365                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6366         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6367         fattr->nlink = 2;
6368 }
6369
6370 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6371                                    const struct qstr *name,
6372                                    struct nfs4_fs_locations *fs_locations,
6373                                    struct page *page)
6374 {
6375         struct nfs_server *server = NFS_SERVER(dir);
6376         u32 bitmask[3] = {
6377                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6378         };
6379         struct nfs4_fs_locations_arg args = {
6380                 .dir_fh = NFS_FH(dir),
6381                 .name = name,
6382                 .page = page,
6383                 .bitmask = bitmask,
6384         };
6385         struct nfs4_fs_locations_res res = {
6386                 .fs_locations = fs_locations,
6387         };
6388         struct rpc_message msg = {
6389                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6390                 .rpc_argp = &args,
6391                 .rpc_resp = &res,
6392         };
6393         int status;
6394
6395         dprintk("%s: start\n", __func__);
6396
6397         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6398          * is not supported */
6399         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6400                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6401         else
6402                 bitmask[0] |= FATTR4_WORD0_FILEID;
6403
6404         nfs_fattr_init(&fs_locations->fattr);
6405         fs_locations->server = server;
6406         fs_locations->nlocations = 0;
6407         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6408         dprintk("%s: returned status = %d\n", __func__, status);
6409         return status;
6410 }
6411
6412 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6413                            const struct qstr *name,
6414                            struct nfs4_fs_locations *fs_locations,
6415                            struct page *page)
6416 {
6417         struct nfs4_exception exception = { };
6418         int err;
6419         do {
6420                 err = _nfs4_proc_fs_locations(client, dir, name,
6421                                 fs_locations, page);
6422                 trace_nfs4_get_fs_locations(dir, name, err);
6423                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6424                                 &exception);
6425         } while (exception.retry);
6426         return err;
6427 }
6428
6429 /*
6430  * This operation also signals the server that this client is
6431  * performing migration recovery.  The server can stop returning
6432  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6433  * appended to this compound to identify the client ID which is
6434  * performing recovery.
6435  */
6436 static int _nfs40_proc_get_locations(struct inode *inode,
6437                                      struct nfs4_fs_locations *locations,
6438                                      struct page *page, struct rpc_cred *cred)
6439 {
6440         struct nfs_server *server = NFS_SERVER(inode);
6441         struct rpc_clnt *clnt = server->client;
6442         u32 bitmask[2] = {
6443                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6444         };
6445         struct nfs4_fs_locations_arg args = {
6446                 .clientid       = server->nfs_client->cl_clientid,
6447                 .fh             = NFS_FH(inode),
6448                 .page           = page,
6449                 .bitmask        = bitmask,
6450                 .migration      = 1,            /* skip LOOKUP */
6451                 .renew          = 1,            /* append RENEW */
6452         };
6453         struct nfs4_fs_locations_res res = {
6454                 .fs_locations   = locations,
6455                 .migration      = 1,
6456                 .renew          = 1,
6457         };
6458         struct rpc_message msg = {
6459                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6460                 .rpc_argp       = &args,
6461                 .rpc_resp       = &res,
6462                 .rpc_cred       = cred,
6463         };
6464         unsigned long now = jiffies;
6465         int status;
6466
6467         nfs_fattr_init(&locations->fattr);
6468         locations->server = server;
6469         locations->nlocations = 0;
6470
6471         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6472         nfs4_set_sequence_privileged(&args.seq_args);
6473         status = nfs4_call_sync_sequence(clnt, server, &msg,
6474                                         &args.seq_args, &res.seq_res);
6475         if (status)
6476                 return status;
6477
6478         renew_lease(server, now);
6479         return 0;
6480 }
6481
6482 #ifdef CONFIG_NFS_V4_1
6483
6484 /*
6485  * This operation also signals the server that this client is
6486  * performing migration recovery.  The server can stop asserting
6487  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6488  * performing this operation is identified in the SEQUENCE
6489  * operation in this compound.
6490  *
6491  * When the client supports GETATTR(fs_locations_info), it can
6492  * be plumbed in here.
6493  */
6494 static int _nfs41_proc_get_locations(struct inode *inode,
6495                                      struct nfs4_fs_locations *locations,
6496                                      struct page *page, struct rpc_cred *cred)
6497 {
6498         struct nfs_server *server = NFS_SERVER(inode);
6499         struct rpc_clnt *clnt = server->client;
6500         u32 bitmask[2] = {
6501                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6502         };
6503         struct nfs4_fs_locations_arg args = {
6504                 .fh             = NFS_FH(inode),
6505                 .page           = page,
6506                 .bitmask        = bitmask,
6507                 .migration      = 1,            /* skip LOOKUP */
6508         };
6509         struct nfs4_fs_locations_res res = {
6510                 .fs_locations   = locations,
6511                 .migration      = 1,
6512         };
6513         struct rpc_message msg = {
6514                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6515                 .rpc_argp       = &args,
6516                 .rpc_resp       = &res,
6517                 .rpc_cred       = cred,
6518         };
6519         int status;
6520
6521         nfs_fattr_init(&locations->fattr);
6522         locations->server = server;
6523         locations->nlocations = 0;
6524
6525         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6526         nfs4_set_sequence_privileged(&args.seq_args);
6527         status = nfs4_call_sync_sequence(clnt, server, &msg,
6528                                         &args.seq_args, &res.seq_res);
6529         if (status == NFS4_OK &&
6530             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6531                 status = -NFS4ERR_LEASE_MOVED;
6532         return status;
6533 }
6534
6535 #endif  /* CONFIG_NFS_V4_1 */
6536
6537 /**
6538  * nfs4_proc_get_locations - discover locations for a migrated FSID
6539  * @inode: inode on FSID that is migrating
6540  * @locations: result of query
6541  * @page: buffer
6542  * @cred: credential to use for this operation
6543  *
6544  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6545  * operation failed, or a negative errno if a local error occurred.
6546  *
6547  * On success, "locations" is filled in, but if the server has
6548  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6549  * asserted.
6550  *
6551  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6552  * from this client that require migration recovery.
6553  */
6554 int nfs4_proc_get_locations(struct inode *inode,
6555                             struct nfs4_fs_locations *locations,
6556                             struct page *page, struct rpc_cred *cred)
6557 {
6558         struct nfs_server *server = NFS_SERVER(inode);
6559         struct nfs_client *clp = server->nfs_client;
6560         const struct nfs4_mig_recovery_ops *ops =
6561                                         clp->cl_mvops->mig_recovery_ops;
6562         struct nfs4_exception exception = { };
6563         int status;
6564
6565         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6566                 (unsigned long long)server->fsid.major,
6567                 (unsigned long long)server->fsid.minor,
6568                 clp->cl_hostname);
6569         nfs_display_fhandle(NFS_FH(inode), __func__);
6570
6571         do {
6572                 status = ops->get_locations(inode, locations, page, cred);
6573                 if (status != -NFS4ERR_DELAY)
6574                         break;
6575                 nfs4_handle_exception(server, status, &exception);
6576         } while (exception.retry);
6577         return status;
6578 }
6579
6580 /*
6581  * This operation also signals the server that this client is
6582  * performing "lease moved" recovery.  The server can stop
6583  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6584  * is appended to this compound to identify the client ID which is
6585  * performing recovery.
6586  */
6587 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6588 {
6589         struct nfs_server *server = NFS_SERVER(inode);
6590         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6591         struct rpc_clnt *clnt = server->client;
6592         struct nfs4_fsid_present_arg args = {
6593                 .fh             = NFS_FH(inode),
6594                 .clientid       = clp->cl_clientid,
6595                 .renew          = 1,            /* append RENEW */
6596         };
6597         struct nfs4_fsid_present_res res = {
6598                 .renew          = 1,
6599         };
6600         struct rpc_message msg = {
6601                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6602                 .rpc_argp       = &args,
6603                 .rpc_resp       = &res,
6604                 .rpc_cred       = cred,
6605         };
6606         unsigned long now = jiffies;
6607         int status;
6608
6609         res.fh = nfs_alloc_fhandle();
6610         if (res.fh == NULL)
6611                 return -ENOMEM;
6612
6613         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6614         nfs4_set_sequence_privileged(&args.seq_args);
6615         status = nfs4_call_sync_sequence(clnt, server, &msg,
6616                                                 &args.seq_args, &res.seq_res);
6617         nfs_free_fhandle(res.fh);
6618         if (status)
6619                 return status;
6620
6621         do_renew_lease(clp, now);
6622         return 0;
6623 }
6624
6625 #ifdef CONFIG_NFS_V4_1
6626
6627 /*
6628  * This operation also signals the server that this client is
6629  * performing "lease moved" recovery.  The server can stop asserting
6630  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6631  * this operation is identified in the SEQUENCE operation in this
6632  * compound.
6633  */
6634 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6635 {
6636         struct nfs_server *server = NFS_SERVER(inode);
6637         struct rpc_clnt *clnt = server->client;
6638         struct nfs4_fsid_present_arg args = {
6639                 .fh             = NFS_FH(inode),
6640         };
6641         struct nfs4_fsid_present_res res = {
6642         };
6643         struct rpc_message msg = {
6644                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6645                 .rpc_argp       = &args,
6646                 .rpc_resp       = &res,
6647                 .rpc_cred       = cred,
6648         };
6649         int status;
6650
6651         res.fh = nfs_alloc_fhandle();
6652         if (res.fh == NULL)
6653                 return -ENOMEM;
6654
6655         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6656         nfs4_set_sequence_privileged(&args.seq_args);
6657         status = nfs4_call_sync_sequence(clnt, server, &msg,
6658                                                 &args.seq_args, &res.seq_res);
6659         nfs_free_fhandle(res.fh);
6660         if (status == NFS4_OK &&
6661             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6662                 status = -NFS4ERR_LEASE_MOVED;
6663         return status;
6664 }
6665
6666 #endif  /* CONFIG_NFS_V4_1 */
6667
6668 /**
6669  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6670  * @inode: inode on FSID to check
6671  * @cred: credential to use for this operation
6672  *
6673  * Server indicates whether the FSID is present, moved, or not
6674  * recognized.  This operation is necessary to clear a LEASE_MOVED
6675  * condition for this client ID.
6676  *
6677  * Returns NFS4_OK if the FSID is present on this server,
6678  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6679  *  NFS4ERR code if some error occurred on the server, or a
6680  *  negative errno if a local failure occurred.
6681  */
6682 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6683 {
6684         struct nfs_server *server = NFS_SERVER(inode);
6685         struct nfs_client *clp = server->nfs_client;
6686         const struct nfs4_mig_recovery_ops *ops =
6687                                         clp->cl_mvops->mig_recovery_ops;
6688         struct nfs4_exception exception = { };
6689         int status;
6690
6691         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6692                 (unsigned long long)server->fsid.major,
6693                 (unsigned long long)server->fsid.minor,
6694                 clp->cl_hostname);
6695         nfs_display_fhandle(NFS_FH(inode), __func__);
6696
6697         do {
6698                 status = ops->fsid_present(inode, cred);
6699                 if (status != -NFS4ERR_DELAY)
6700                         break;
6701                 nfs4_handle_exception(server, status, &exception);
6702         } while (exception.retry);
6703         return status;
6704 }
6705
6706 /**
6707  * If 'use_integrity' is true and the state managment nfs_client
6708  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6709  * and the machine credential as per RFC3530bis and RFC5661 Security
6710  * Considerations sections. Otherwise, just use the user cred with the
6711  * filesystem's rpc_client.
6712  */
6713 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6714 {
6715         int status;
6716         struct nfs4_secinfo_arg args = {
6717                 .dir_fh = NFS_FH(dir),
6718                 .name   = name,
6719         };
6720         struct nfs4_secinfo_res res = {
6721                 .flavors     = flavors,
6722         };
6723         struct rpc_message msg = {
6724                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6725                 .rpc_argp = &args,
6726                 .rpc_resp = &res,
6727         };
6728         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6729         struct rpc_cred *cred = NULL;
6730
6731         if (use_integrity) {
6732                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6733                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6734                 msg.rpc_cred = cred;
6735         }
6736
6737         dprintk("NFS call  secinfo %s\n", name->name);
6738
6739         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6740                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6741
6742         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6743                                 &res.seq_res, 0);
6744         dprintk("NFS reply  secinfo: %d\n", status);
6745
6746         if (cred)
6747                 put_rpccred(cred);
6748
6749         return status;
6750 }
6751
6752 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6753                       struct nfs4_secinfo_flavors *flavors)
6754 {
6755         struct nfs4_exception exception = { };
6756         int err;
6757         do {
6758                 err = -NFS4ERR_WRONGSEC;
6759
6760                 /* try to use integrity protection with machine cred */
6761                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6762                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6763
6764                 /*
6765                  * if unable to use integrity protection, or SECINFO with
6766                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6767                  * disallowed by spec, but exists in deployed servers) use
6768                  * the current filesystem's rpc_client and the user cred.
6769                  */
6770                 if (err == -NFS4ERR_WRONGSEC)
6771                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6772
6773                 trace_nfs4_secinfo(dir, name, err);
6774                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6775                                 &exception);
6776         } while (exception.retry);
6777         return err;
6778 }
6779
6780 #ifdef CONFIG_NFS_V4_1
6781 /*
6782  * Check the exchange flags returned by the server for invalid flags, having
6783  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6784  * DS flags set.
6785  */
6786 static int nfs4_check_cl_exchange_flags(u32 flags)
6787 {
6788         if (flags & ~EXCHGID4_FLAG_MASK_R)
6789                 goto out_inval;
6790         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6791             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6792                 goto out_inval;
6793         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6794                 goto out_inval;
6795         return NFS_OK;
6796 out_inval:
6797         return -NFS4ERR_INVAL;
6798 }
6799
6800 static bool
6801 nfs41_same_server_scope(struct nfs41_server_scope *a,
6802                         struct nfs41_server_scope *b)
6803 {
6804         if (a->server_scope_sz == b->server_scope_sz &&
6805             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6806                 return true;
6807
6808         return false;
6809 }
6810
6811 static void
6812 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
6813 {
6814 }
6815
6816 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
6817         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
6818 };
6819
6820 /*
6821  * nfs4_proc_bind_one_conn_to_session()
6822  *
6823  * The 4.1 client currently uses the same TCP connection for the
6824  * fore and backchannel.
6825  */
6826 static
6827 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
6828                 struct rpc_xprt *xprt,
6829                 struct nfs_client *clp,
6830                 struct rpc_cred *cred)
6831 {
6832         int status;
6833         struct nfs41_bind_conn_to_session_args args = {
6834                 .client = clp,
6835                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6836         };
6837         struct nfs41_bind_conn_to_session_res res;
6838         struct rpc_message msg = {
6839                 .rpc_proc =
6840                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6841                 .rpc_argp = &args,
6842                 .rpc_resp = &res,
6843                 .rpc_cred = cred,
6844         };
6845         struct rpc_task_setup task_setup_data = {
6846                 .rpc_client = clnt,
6847                 .rpc_xprt = xprt,
6848                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
6849                 .rpc_message = &msg,
6850                 .flags = RPC_TASK_TIMEOUT,
6851         };
6852         struct rpc_task *task;
6853
6854         dprintk("--> %s\n", __func__);
6855
6856         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6857         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6858                 args.dir = NFS4_CDFC4_FORE;
6859
6860         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
6861         if (xprt != rcu_access_pointer(clnt->cl_xprt))
6862                 args.dir = NFS4_CDFC4_FORE;
6863
6864         task = rpc_run_task(&task_setup_data);
6865         if (!IS_ERR(task)) {
6866                 status = task->tk_status;
6867                 rpc_put_task(task);
6868         } else
6869                 status = PTR_ERR(task);
6870         trace_nfs4_bind_conn_to_session(clp, status);
6871         if (status == 0) {
6872                 if (memcmp(res.sessionid.data,
6873                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6874                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6875                         status = -EIO;
6876                         goto out;
6877                 }
6878                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6879                         dprintk("NFS: %s: Unexpected direction from server\n",
6880                                 __func__);
6881                         status = -EIO;
6882                         goto out;
6883                 }
6884                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6885                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6886                                 __func__);
6887                         status = -EIO;
6888                         goto out;
6889                 }
6890         }
6891 out:
6892         dprintk("<-- %s status= %d\n", __func__, status);
6893         return status;
6894 }
6895
6896 struct rpc_bind_conn_calldata {
6897         struct nfs_client *clp;
6898         struct rpc_cred *cred;
6899 };
6900
6901 static int
6902 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
6903                 struct rpc_xprt *xprt,
6904                 void *calldata)
6905 {
6906         struct rpc_bind_conn_calldata *p = calldata;
6907
6908         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
6909 }
6910
6911 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6912 {
6913         struct rpc_bind_conn_calldata data = {
6914                 .clp = clp,
6915                 .cred = cred,
6916         };
6917         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
6918                         nfs4_proc_bind_conn_to_session_callback, &data);
6919 }
6920
6921 /*
6922  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6923  * and operations we'd like to see to enable certain features in the allow map
6924  */
6925 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6926         .how = SP4_MACH_CRED,
6927         .enforce.u.words = {
6928                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6929                       1 << (OP_EXCHANGE_ID - 32) |
6930                       1 << (OP_CREATE_SESSION - 32) |
6931                       1 << (OP_DESTROY_SESSION - 32) |
6932                       1 << (OP_DESTROY_CLIENTID - 32)
6933         },
6934         .allow.u.words = {
6935                 [0] = 1 << (OP_CLOSE) |
6936                       1 << (OP_OPEN_DOWNGRADE) |
6937                       1 << (OP_LOCKU) |
6938                       1 << (OP_DELEGRETURN) |
6939                       1 << (OP_COMMIT),
6940                 [1] = 1 << (OP_SECINFO - 32) |
6941                       1 << (OP_SECINFO_NO_NAME - 32) |
6942                       1 << (OP_LAYOUTRETURN - 32) |
6943                       1 << (OP_TEST_STATEID - 32) |
6944                       1 << (OP_FREE_STATEID - 32) |
6945                       1 << (OP_WRITE - 32)
6946         }
6947 };
6948
6949 /*
6950  * Select the state protection mode for client `clp' given the server results
6951  * from exchange_id in `sp'.
6952  *
6953  * Returns 0 on success, negative errno otherwise.
6954  */
6955 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6956                                  struct nfs41_state_protection *sp)
6957 {
6958         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6959                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6960                       1 << (OP_EXCHANGE_ID - 32) |
6961                       1 << (OP_CREATE_SESSION - 32) |
6962                       1 << (OP_DESTROY_SESSION - 32) |
6963                       1 << (OP_DESTROY_CLIENTID - 32)
6964         };
6965         unsigned int i;
6966
6967         if (sp->how == SP4_MACH_CRED) {
6968                 /* Print state protect result */
6969                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6970                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6971                         if (test_bit(i, sp->enforce.u.longs))
6972                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6973                         if (test_bit(i, sp->allow.u.longs))
6974                                 dfprintk(MOUNT, "  allow op %d\n", i);
6975                 }
6976
6977                 /* make sure nothing is on enforce list that isn't supported */
6978                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6979                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6980                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6981                                 return -EINVAL;
6982                         }
6983                 }
6984
6985                 /*
6986                  * Minimal mode - state operations are allowed to use machine
6987                  * credential.  Note this already happens by default, so the
6988                  * client doesn't have to do anything more than the negotiation.
6989                  *
6990                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6991                  *       we're already using the machine cred for exchange_id
6992                  *       and will never use a different cred.
6993                  */
6994                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6995                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6996                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6997                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6998                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6999                         dfprintk(MOUNT, "  minimal mode enabled\n");
7000                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7001                 } else {
7002                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7003                         return -EINVAL;
7004                 }
7005
7006                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7007                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7008                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7009                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7010                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7011                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7012                 }
7013
7014                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7015                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7016                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7017                                 &clp->cl_sp4_flags);
7018                 }
7019
7020                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7021                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7022                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7023                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7024                 }
7025
7026                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7027                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7028                         dfprintk(MOUNT, "  stateid mode enabled\n");
7029                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7030                 }
7031
7032                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7033                         dfprintk(MOUNT, "  write mode enabled\n");
7034                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7035                 }
7036
7037                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7038                         dfprintk(MOUNT, "  commit mode enabled\n");
7039                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7040                 }
7041         }
7042
7043         return 0;
7044 }
7045
7046 /*
7047  * _nfs4_proc_exchange_id()
7048  *
7049  * Wrapper for EXCHANGE_ID operation.
7050  */
7051 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7052         u32 sp4_how)
7053 {
7054         nfs4_verifier verifier;
7055         struct nfs41_exchange_id_args args = {
7056                 .verifier = &verifier,
7057                 .client = clp,
7058 #ifdef CONFIG_NFS_V4_1_MIGRATION
7059                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7060                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
7061                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7062 #else
7063                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7064                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
7065 #endif
7066         };
7067         struct nfs41_exchange_id_res res = {
7068                 0
7069         };
7070         int status;
7071         struct rpc_message msg = {
7072                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7073                 .rpc_argp = &args,
7074                 .rpc_resp = &res,
7075                 .rpc_cred = cred,
7076         };
7077
7078         nfs4_init_boot_verifier(clp, &verifier);
7079
7080         status = nfs4_init_uniform_client_string(clp);
7081         if (status)
7082                 goto out;
7083
7084         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7085                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7086                 clp->cl_owner_id);
7087
7088         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7089                                         GFP_NOFS);
7090         if (unlikely(res.server_owner == NULL)) {
7091                 status = -ENOMEM;
7092                 goto out;
7093         }
7094
7095         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7096                                         GFP_NOFS);
7097         if (unlikely(res.server_scope == NULL)) {
7098                 status = -ENOMEM;
7099                 goto out_server_owner;
7100         }
7101
7102         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7103         if (unlikely(res.impl_id == NULL)) {
7104                 status = -ENOMEM;
7105                 goto out_server_scope;
7106         }
7107
7108         switch (sp4_how) {
7109         case SP4_NONE:
7110                 args.state_protect.how = SP4_NONE;
7111                 break;
7112
7113         case SP4_MACH_CRED:
7114                 args.state_protect = nfs4_sp4_mach_cred_request;
7115                 break;
7116
7117         default:
7118                 /* unsupported! */
7119                 WARN_ON_ONCE(1);
7120                 status = -EINVAL;
7121                 goto out_impl_id;
7122         }
7123
7124         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7125         trace_nfs4_exchange_id(clp, status);
7126         if (status == 0)
7127                 status = nfs4_check_cl_exchange_flags(res.flags);
7128
7129         if (status == 0)
7130                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7131
7132         if (status == 0) {
7133                 clp->cl_clientid = res.clientid;
7134                 clp->cl_exchange_flags = res.flags;
7135                 /* Client ID is not confirmed */
7136                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7137                         clear_bit(NFS4_SESSION_ESTABLISHED,
7138                                         &clp->cl_session->session_state);
7139                         clp->cl_seqid = res.seqid;
7140                 }
7141
7142                 kfree(clp->cl_serverowner);
7143                 clp->cl_serverowner = res.server_owner;
7144                 res.server_owner = NULL;
7145
7146                 /* use the most recent implementation id */
7147                 kfree(clp->cl_implid);
7148                 clp->cl_implid = res.impl_id;
7149                 res.impl_id = NULL;
7150
7151                 if (clp->cl_serverscope != NULL &&
7152                     !nfs41_same_server_scope(clp->cl_serverscope,
7153                                              res.server_scope)) {
7154                         dprintk("%s: server_scope mismatch detected\n",
7155                                 __func__);
7156                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7157                         kfree(clp->cl_serverscope);
7158                         clp->cl_serverscope = NULL;
7159                 }
7160
7161                 if (clp->cl_serverscope == NULL) {
7162                         clp->cl_serverscope = res.server_scope;
7163                         res.server_scope = NULL;
7164                 }
7165         }
7166
7167 out_impl_id:
7168         kfree(res.impl_id);
7169 out_server_scope:
7170         kfree(res.server_scope);
7171 out_server_owner:
7172         kfree(res.server_owner);
7173 out:
7174         if (clp->cl_implid != NULL)
7175                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7176                         "domain: %s, name: %s, date: %llu,%u\n",
7177                         clp->cl_implid->domain, clp->cl_implid->name,
7178                         clp->cl_implid->date.seconds,
7179                         clp->cl_implid->date.nseconds);
7180         dprintk("NFS reply exchange_id: %d\n", status);
7181         return status;
7182 }
7183
7184 /*
7185  * nfs4_proc_exchange_id()
7186  *
7187  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7188  *
7189  * Since the clientid has expired, all compounds using sessions
7190  * associated with the stale clientid will be returning
7191  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7192  * be in some phase of session reset.
7193  *
7194  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7195  */
7196 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7197 {
7198         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7199         int status;
7200
7201         /* try SP4_MACH_CRED if krb5i/p */
7202         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7203             authflavor == RPC_AUTH_GSS_KRB5P) {
7204                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7205                 if (!status)
7206                         return 0;
7207         }
7208
7209         /* try SP4_NONE */
7210         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7211 }
7212
7213 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7214                 struct rpc_cred *cred)
7215 {
7216         struct rpc_message msg = {
7217                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7218                 .rpc_argp = clp,
7219                 .rpc_cred = cred,
7220         };
7221         int status;
7222
7223         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7224         trace_nfs4_destroy_clientid(clp, status);
7225         if (status)
7226                 dprintk("NFS: Got error %d from the server %s on "
7227                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7228         return status;
7229 }
7230
7231 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7232                 struct rpc_cred *cred)
7233 {
7234         unsigned int loop;
7235         int ret;
7236
7237         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7238                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7239                 switch (ret) {
7240                 case -NFS4ERR_DELAY:
7241                 case -NFS4ERR_CLIENTID_BUSY:
7242                         ssleep(1);
7243                         break;
7244                 default:
7245                         return ret;
7246                 }
7247         }
7248         return 0;
7249 }
7250
7251 int nfs4_destroy_clientid(struct nfs_client *clp)
7252 {
7253         struct rpc_cred *cred;
7254         int ret = 0;
7255
7256         if (clp->cl_mvops->minor_version < 1)
7257                 goto out;
7258         if (clp->cl_exchange_flags == 0)
7259                 goto out;
7260         if (clp->cl_preserve_clid)
7261                 goto out;
7262         cred = nfs4_get_clid_cred(clp);
7263         ret = nfs4_proc_destroy_clientid(clp, cred);
7264         if (cred)
7265                 put_rpccred(cred);
7266         switch (ret) {
7267         case 0:
7268         case -NFS4ERR_STALE_CLIENTID:
7269                 clp->cl_exchange_flags = 0;
7270         }
7271 out:
7272         return ret;
7273 }
7274
7275 struct nfs4_get_lease_time_data {
7276         struct nfs4_get_lease_time_args *args;
7277         struct nfs4_get_lease_time_res *res;
7278         struct nfs_client *clp;
7279 };
7280
7281 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7282                                         void *calldata)
7283 {
7284         struct nfs4_get_lease_time_data *data =
7285                         (struct nfs4_get_lease_time_data *)calldata;
7286
7287         dprintk("--> %s\n", __func__);
7288         /* just setup sequence, do not trigger session recovery
7289            since we're invoked within one */
7290         nfs41_setup_sequence(data->clp->cl_session,
7291                         &data->args->la_seq_args,
7292                         &data->res->lr_seq_res,
7293                         task);
7294         dprintk("<-- %s\n", __func__);
7295 }
7296
7297 /*
7298  * Called from nfs4_state_manager thread for session setup, so don't recover
7299  * from sequence operation or clientid errors.
7300  */
7301 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7302 {
7303         struct nfs4_get_lease_time_data *data =
7304                         (struct nfs4_get_lease_time_data *)calldata;
7305
7306         dprintk("--> %s\n", __func__);
7307         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7308                 return;
7309         switch (task->tk_status) {
7310         case -NFS4ERR_DELAY:
7311         case -NFS4ERR_GRACE:
7312                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7313                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7314                 task->tk_status = 0;
7315                 /* fall through */
7316         case -NFS4ERR_RETRY_UNCACHED_REP:
7317                 rpc_restart_call_prepare(task);
7318                 return;
7319         }
7320         dprintk("<-- %s\n", __func__);
7321 }
7322
7323 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7324         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7325         .rpc_call_done = nfs4_get_lease_time_done,
7326 };
7327
7328 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7329 {
7330         struct rpc_task *task;
7331         struct nfs4_get_lease_time_args args;
7332         struct nfs4_get_lease_time_res res = {
7333                 .lr_fsinfo = fsinfo,
7334         };
7335         struct nfs4_get_lease_time_data data = {
7336                 .args = &args,
7337                 .res = &res,
7338                 .clp = clp,
7339         };
7340         struct rpc_message msg = {
7341                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7342                 .rpc_argp = &args,
7343                 .rpc_resp = &res,
7344         };
7345         struct rpc_task_setup task_setup = {
7346                 .rpc_client = clp->cl_rpcclient,
7347                 .rpc_message = &msg,
7348                 .callback_ops = &nfs4_get_lease_time_ops,
7349                 .callback_data = &data,
7350                 .flags = RPC_TASK_TIMEOUT,
7351         };
7352         int status;
7353
7354         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7355         nfs4_set_sequence_privileged(&args.la_seq_args);
7356         dprintk("--> %s\n", __func__);
7357         task = rpc_run_task(&task_setup);
7358
7359         if (IS_ERR(task))
7360                 status = PTR_ERR(task);
7361         else {
7362                 status = task->tk_status;
7363                 rpc_put_task(task);
7364         }
7365         dprintk("<-- %s return %d\n", __func__, status);
7366
7367         return status;
7368 }
7369
7370 /*
7371  * Initialize the values to be used by the client in CREATE_SESSION
7372  * If nfs4_init_session set the fore channel request and response sizes,
7373  * use them.
7374  *
7375  * Set the back channel max_resp_sz_cached to zero to force the client to
7376  * always set csa_cachethis to FALSE because the current implementation
7377  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7378  */
7379 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7380                                     struct rpc_clnt *clnt)
7381 {
7382         unsigned int max_rqst_sz, max_resp_sz;
7383         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7384
7385         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7386         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7387
7388         /* Fore channel attributes */
7389         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7390         args->fc_attrs.max_resp_sz = max_resp_sz;
7391         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7392         args->fc_attrs.max_reqs = max_session_slots;
7393
7394         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7395                 "max_ops=%u max_reqs=%u\n",
7396                 __func__,
7397                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7398                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7399
7400         /* Back channel attributes */
7401         args->bc_attrs.max_rqst_sz = max_bc_payload;
7402         args->bc_attrs.max_resp_sz = max_bc_payload;
7403         args->bc_attrs.max_resp_sz_cached = 0;
7404         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7405         args->bc_attrs.max_reqs = NFS41_BC_MAX_CALLBACKS;
7406
7407         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7408                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7409                 __func__,
7410                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7411                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7412                 args->bc_attrs.max_reqs);
7413 }
7414
7415 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7416                 struct nfs41_create_session_res *res)
7417 {
7418         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7419         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7420
7421         if (rcvd->max_resp_sz > sent->max_resp_sz)
7422                 return -EINVAL;
7423         /*
7424          * Our requested max_ops is the minimum we need; we're not
7425          * prepared to break up compounds into smaller pieces than that.
7426          * So, no point even trying to continue if the server won't
7427          * cooperate:
7428          */
7429         if (rcvd->max_ops < sent->max_ops)
7430                 return -EINVAL;
7431         if (rcvd->max_reqs == 0)
7432                 return -EINVAL;
7433         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7434                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7435         return 0;
7436 }
7437
7438 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7439                 struct nfs41_create_session_res *res)
7440 {
7441         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7442         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7443
7444         if (!(res->flags & SESSION4_BACK_CHAN))
7445                 goto out;
7446         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7447                 return -EINVAL;
7448         if (rcvd->max_resp_sz < sent->max_resp_sz)
7449                 return -EINVAL;
7450         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7451                 return -EINVAL;
7452         /* These would render the backchannel useless: */
7453         if (rcvd->max_ops != sent->max_ops)
7454                 return -EINVAL;
7455         if (rcvd->max_reqs != sent->max_reqs)
7456                 return -EINVAL;
7457 out:
7458         return 0;
7459 }
7460
7461 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7462                                      struct nfs41_create_session_res *res)
7463 {
7464         int ret;
7465
7466         ret = nfs4_verify_fore_channel_attrs(args, res);
7467         if (ret)
7468                 return ret;
7469         return nfs4_verify_back_channel_attrs(args, res);
7470 }
7471
7472 static void nfs4_update_session(struct nfs4_session *session,
7473                 struct nfs41_create_session_res *res)
7474 {
7475         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7476         /* Mark client id and session as being confirmed */
7477         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7478         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7479         session->flags = res->flags;
7480         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7481         if (res->flags & SESSION4_BACK_CHAN)
7482                 memcpy(&session->bc_attrs, &res->bc_attrs,
7483                                 sizeof(session->bc_attrs));
7484 }
7485
7486 static int _nfs4_proc_create_session(struct nfs_client *clp,
7487                 struct rpc_cred *cred)
7488 {
7489         struct nfs4_session *session = clp->cl_session;
7490         struct nfs41_create_session_args args = {
7491                 .client = clp,
7492                 .clientid = clp->cl_clientid,
7493                 .seqid = clp->cl_seqid,
7494                 .cb_program = NFS4_CALLBACK,
7495         };
7496         struct nfs41_create_session_res res;
7497
7498         struct rpc_message msg = {
7499                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7500                 .rpc_argp = &args,
7501                 .rpc_resp = &res,
7502                 .rpc_cred = cred,
7503         };
7504         int status;
7505
7506         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7507         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7508
7509         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7510         trace_nfs4_create_session(clp, status);
7511
7512         if (!status) {
7513                 /* Verify the session's negotiated channel_attrs values */
7514                 status = nfs4_verify_channel_attrs(&args, &res);
7515                 /* Increment the clientid slot sequence id */
7516                 if (clp->cl_seqid == res.seqid)
7517                         clp->cl_seqid++;
7518                 if (status)
7519                         goto out;
7520                 nfs4_update_session(session, &res);
7521         }
7522 out:
7523         return status;
7524 }
7525
7526 /*
7527  * Issues a CREATE_SESSION operation to the server.
7528  * It is the responsibility of the caller to verify the session is
7529  * expired before calling this routine.
7530  */
7531 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7532 {
7533         int status;
7534         unsigned *ptr;
7535         struct nfs4_session *session = clp->cl_session;
7536
7537         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7538
7539         status = _nfs4_proc_create_session(clp, cred);
7540         if (status)
7541                 goto out;
7542
7543         /* Init or reset the session slot tables */
7544         status = nfs4_setup_session_slot_tables(session);
7545         dprintk("slot table setup returned %d\n", status);
7546         if (status)
7547                 goto out;
7548
7549         ptr = (unsigned *)&session->sess_id.data[0];
7550         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7551                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7552 out:
7553         dprintk("<-- %s\n", __func__);
7554         return status;
7555 }
7556
7557 /*
7558  * Issue the over-the-wire RPC DESTROY_SESSION.
7559  * The caller must serialize access to this routine.
7560  */
7561 int nfs4_proc_destroy_session(struct nfs4_session *session,
7562                 struct rpc_cred *cred)
7563 {
7564         struct rpc_message msg = {
7565                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7566                 .rpc_argp = session,
7567                 .rpc_cred = cred,
7568         };
7569         int status = 0;
7570
7571         dprintk("--> nfs4_proc_destroy_session\n");
7572
7573         /* session is still being setup */
7574         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7575                 return 0;
7576
7577         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7578         trace_nfs4_destroy_session(session->clp, status);
7579
7580         if (status)
7581                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7582                         "Session has been destroyed regardless...\n", status);
7583
7584         dprintk("<-- nfs4_proc_destroy_session\n");
7585         return status;
7586 }
7587
7588 /*
7589  * Renew the cl_session lease.
7590  */
7591 struct nfs4_sequence_data {
7592         struct nfs_client *clp;
7593         struct nfs4_sequence_args args;
7594         struct nfs4_sequence_res res;
7595 };
7596
7597 static void nfs41_sequence_release(void *data)
7598 {
7599         struct nfs4_sequence_data *calldata = data;
7600         struct nfs_client *clp = calldata->clp;
7601
7602         if (atomic_read(&clp->cl_count) > 1)
7603                 nfs4_schedule_state_renewal(clp);
7604         nfs_put_client(clp);
7605         kfree(calldata);
7606 }
7607
7608 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7609 {
7610         switch(task->tk_status) {
7611         case -NFS4ERR_DELAY:
7612                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7613                 return -EAGAIN;
7614         default:
7615                 nfs4_schedule_lease_recovery(clp);
7616         }
7617         return 0;
7618 }
7619
7620 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7621 {
7622         struct nfs4_sequence_data *calldata = data;
7623         struct nfs_client *clp = calldata->clp;
7624
7625         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7626                 return;
7627
7628         trace_nfs4_sequence(clp, task->tk_status);
7629         if (task->tk_status < 0) {
7630                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7631                 if (atomic_read(&clp->cl_count) == 1)
7632                         goto out;
7633
7634                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7635                         rpc_restart_call_prepare(task);
7636                         return;
7637                 }
7638         }
7639         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7640 out:
7641         dprintk("<-- %s\n", __func__);
7642 }
7643
7644 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7645 {
7646         struct nfs4_sequence_data *calldata = data;
7647         struct nfs_client *clp = calldata->clp;
7648         struct nfs4_sequence_args *args;
7649         struct nfs4_sequence_res *res;
7650
7651         args = task->tk_msg.rpc_argp;
7652         res = task->tk_msg.rpc_resp;
7653
7654         nfs41_setup_sequence(clp->cl_session, args, res, task);
7655 }
7656
7657 static const struct rpc_call_ops nfs41_sequence_ops = {
7658         .rpc_call_done = nfs41_sequence_call_done,
7659         .rpc_call_prepare = nfs41_sequence_prepare,
7660         .rpc_release = nfs41_sequence_release,
7661 };
7662
7663 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7664                 struct rpc_cred *cred,
7665                 bool is_privileged)
7666 {
7667         struct nfs4_sequence_data *calldata;
7668         struct rpc_message msg = {
7669                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7670                 .rpc_cred = cred,
7671         };
7672         struct rpc_task_setup task_setup_data = {
7673                 .rpc_client = clp->cl_rpcclient,
7674                 .rpc_message = &msg,
7675                 .callback_ops = &nfs41_sequence_ops,
7676                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7677         };
7678
7679         if (!atomic_inc_not_zero(&clp->cl_count))
7680                 return ERR_PTR(-EIO);
7681         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7682         if (calldata == NULL) {
7683                 nfs_put_client(clp);
7684                 return ERR_PTR(-ENOMEM);
7685         }
7686         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7687         if (is_privileged)
7688                 nfs4_set_sequence_privileged(&calldata->args);
7689         msg.rpc_argp = &calldata->args;
7690         msg.rpc_resp = &calldata->res;
7691         calldata->clp = clp;
7692         task_setup_data.callback_data = calldata;
7693
7694         return rpc_run_task(&task_setup_data);
7695 }
7696
7697 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7698 {
7699         struct rpc_task *task;
7700         int ret = 0;
7701
7702         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7703                 return -EAGAIN;
7704         task = _nfs41_proc_sequence(clp, cred, false);
7705         if (IS_ERR(task))
7706                 ret = PTR_ERR(task);
7707         else
7708                 rpc_put_task_async(task);
7709         dprintk("<-- %s status=%d\n", __func__, ret);
7710         return ret;
7711 }
7712
7713 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7714 {
7715         struct rpc_task *task;
7716         int ret;
7717
7718         task = _nfs41_proc_sequence(clp, cred, true);
7719         if (IS_ERR(task)) {
7720                 ret = PTR_ERR(task);
7721                 goto out;
7722         }
7723         ret = rpc_wait_for_completion_task(task);
7724         if (!ret)
7725                 ret = task->tk_status;
7726         rpc_put_task(task);
7727 out:
7728         dprintk("<-- %s status=%d\n", __func__, ret);
7729         return ret;
7730 }
7731
7732 struct nfs4_reclaim_complete_data {
7733         struct nfs_client *clp;
7734         struct nfs41_reclaim_complete_args arg;
7735         struct nfs41_reclaim_complete_res res;
7736 };
7737
7738 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7739 {
7740         struct nfs4_reclaim_complete_data *calldata = data;
7741
7742         nfs41_setup_sequence(calldata->clp->cl_session,
7743                         &calldata->arg.seq_args,
7744                         &calldata->res.seq_res,
7745                         task);
7746 }
7747
7748 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7749 {
7750         switch(task->tk_status) {
7751         case 0:
7752         case -NFS4ERR_COMPLETE_ALREADY:
7753         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7754                 break;
7755         case -NFS4ERR_DELAY:
7756                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7757                 /* fall through */
7758         case -NFS4ERR_RETRY_UNCACHED_REP:
7759                 return -EAGAIN;
7760         default:
7761                 nfs4_schedule_lease_recovery(clp);
7762         }
7763         return 0;
7764 }
7765
7766 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7767 {
7768         struct nfs4_reclaim_complete_data *calldata = data;
7769         struct nfs_client *clp = calldata->clp;
7770         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7771
7772         dprintk("--> %s\n", __func__);
7773         if (!nfs41_sequence_done(task, res))
7774                 return;
7775
7776         trace_nfs4_reclaim_complete(clp, task->tk_status);
7777         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7778                 rpc_restart_call_prepare(task);
7779                 return;
7780         }
7781         dprintk("<-- %s\n", __func__);
7782 }
7783
7784 static void nfs4_free_reclaim_complete_data(void *data)
7785 {
7786         struct nfs4_reclaim_complete_data *calldata = data;
7787
7788         kfree(calldata);
7789 }
7790
7791 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7792         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7793         .rpc_call_done = nfs4_reclaim_complete_done,
7794         .rpc_release = nfs4_free_reclaim_complete_data,
7795 };
7796
7797 /*
7798  * Issue a global reclaim complete.
7799  */
7800 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7801                 struct rpc_cred *cred)
7802 {
7803         struct nfs4_reclaim_complete_data *calldata;
7804         struct rpc_task *task;
7805         struct rpc_message msg = {
7806                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7807                 .rpc_cred = cred,
7808         };
7809         struct rpc_task_setup task_setup_data = {
7810                 .rpc_client = clp->cl_rpcclient,
7811                 .rpc_message = &msg,
7812                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7813                 .flags = RPC_TASK_ASYNC,
7814         };
7815         int status = -ENOMEM;
7816
7817         dprintk("--> %s\n", __func__);
7818         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7819         if (calldata == NULL)
7820                 goto out;
7821         calldata->clp = clp;
7822         calldata->arg.one_fs = 0;
7823
7824         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7825         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7826         msg.rpc_argp = &calldata->arg;
7827         msg.rpc_resp = &calldata->res;
7828         task_setup_data.callback_data = calldata;
7829         task = rpc_run_task(&task_setup_data);
7830         if (IS_ERR(task)) {
7831                 status = PTR_ERR(task);
7832                 goto out;
7833         }
7834         status = nfs4_wait_for_completion_rpc_task(task);
7835         if (status == 0)
7836                 status = task->tk_status;
7837         rpc_put_task(task);
7838         return 0;
7839 out:
7840         dprintk("<-- %s status=%d\n", __func__, status);
7841         return status;
7842 }
7843
7844 static void
7845 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7846 {
7847         struct nfs4_layoutget *lgp = calldata;
7848         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7849         struct nfs4_session *session = nfs4_get_session(server);
7850
7851         dprintk("--> %s\n", __func__);
7852         nfs41_setup_sequence(session, &lgp->args.seq_args,
7853                                 &lgp->res.seq_res, task);
7854         dprintk("<-- %s\n", __func__);
7855 }
7856
7857 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7858 {
7859         struct nfs4_layoutget *lgp = calldata;
7860
7861         dprintk("--> %s\n", __func__);
7862         nfs41_sequence_done(task, &lgp->res.seq_res);
7863         dprintk("<-- %s\n", __func__);
7864 }
7865
7866 static int
7867 nfs4_layoutget_handle_exception(struct rpc_task *task,
7868                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
7869 {
7870         struct inode *inode = lgp->args.inode;
7871         struct nfs_server *server = NFS_SERVER(inode);
7872         struct pnfs_layout_hdr *lo;
7873         int status = task->tk_status;
7874
7875         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7876
7877         switch (status) {
7878         case 0:
7879                 goto out;
7880
7881         /*
7882          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
7883          * on the file. set tk_status to -ENODATA to tell upper layer to
7884          * retry go inband.
7885          */
7886         case -NFS4ERR_LAYOUTUNAVAILABLE:
7887                 status = -ENODATA;
7888                 goto out;
7889         /*
7890          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7891          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7892          */
7893         case -NFS4ERR_BADLAYOUT:
7894                 status = -EOVERFLOW;
7895                 goto out;
7896         /*
7897          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7898          * (or clients) writing to the same RAID stripe except when
7899          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7900          *
7901          * Treat it like we would RECALLCONFLICT -- we retry for a little
7902          * while, and then eventually give up.
7903          */
7904         case -NFS4ERR_LAYOUTTRYLATER:
7905                 if (lgp->args.minlength == 0) {
7906                         status = -EOVERFLOW;
7907                         goto out;
7908                 }
7909                 /* Fallthrough */
7910         case -NFS4ERR_RECALLCONFLICT:
7911                 nfs4_handle_exception(server, -NFS4ERR_RECALLCONFLICT,
7912                                         exception);
7913                 status = -ERECALLCONFLICT;
7914                 goto out;
7915         case -NFS4ERR_EXPIRED:
7916         case -NFS4ERR_BAD_STATEID:
7917                 exception->timeout = 0;
7918                 spin_lock(&inode->i_lock);
7919                 if (nfs4_stateid_match(&lgp->args.stateid,
7920                                         &lgp->args.ctx->state->stateid)) {
7921                         spin_unlock(&inode->i_lock);
7922                         /* If the open stateid was bad, then recover it. */
7923                         exception->state = lgp->args.ctx->state;
7924                         break;
7925                 }
7926                 lo = NFS_I(inode)->layout;
7927                 if (lo && nfs4_stateid_match(&lgp->args.stateid,
7928                                         &lo->plh_stateid)) {
7929                         LIST_HEAD(head);
7930
7931                         /*
7932                          * Mark the bad layout state as invalid, then retry
7933                          * with the current stateid.
7934                          */
7935                         set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7936                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL, 0);
7937                         spin_unlock(&inode->i_lock);
7938                         pnfs_free_lseg_list(&head);
7939                 } else
7940                         spin_unlock(&inode->i_lock);
7941                 status = -EAGAIN;
7942                 goto out;
7943         }
7944
7945         status = nfs4_handle_exception(server, status, exception);
7946         if (exception->retry)
7947                 status = -EAGAIN;
7948 out:
7949         dprintk("<-- %s\n", __func__);
7950         return status;
7951 }
7952
7953 static size_t max_response_pages(struct nfs_server *server)
7954 {
7955         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7956         return nfs_page_array_len(0, max_resp_sz);
7957 }
7958
7959 static void nfs4_free_pages(struct page **pages, size_t size)
7960 {
7961         int i;
7962
7963         if (!pages)
7964                 return;
7965
7966         for (i = 0; i < size; i++) {
7967                 if (!pages[i])
7968                         break;
7969                 __free_page(pages[i]);
7970         }
7971         kfree(pages);
7972 }
7973
7974 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7975 {
7976         struct page **pages;
7977         int i;
7978
7979         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7980         if (!pages) {
7981                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7982                 return NULL;
7983         }
7984
7985         for (i = 0; i < size; i++) {
7986                 pages[i] = alloc_page(gfp_flags);
7987                 if (!pages[i]) {
7988                         dprintk("%s: failed to allocate page\n", __func__);
7989                         nfs4_free_pages(pages, size);
7990                         return NULL;
7991                 }
7992         }
7993
7994         return pages;
7995 }
7996
7997 static void nfs4_layoutget_release(void *calldata)
7998 {
7999         struct nfs4_layoutget *lgp = calldata;
8000         struct inode *inode = lgp->args.inode;
8001         struct nfs_server *server = NFS_SERVER(inode);
8002         size_t max_pages = max_response_pages(server);
8003
8004         dprintk("--> %s\n", __func__);
8005         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8006         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8007         put_nfs_open_context(lgp->args.ctx);
8008         kfree(calldata);
8009         dprintk("<-- %s\n", __func__);
8010 }
8011
8012 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8013         .rpc_call_prepare = nfs4_layoutget_prepare,
8014         .rpc_call_done = nfs4_layoutget_done,
8015         .rpc_release = nfs4_layoutget_release,
8016 };
8017
8018 struct pnfs_layout_segment *
8019 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8020 {
8021         struct inode *inode = lgp->args.inode;
8022         struct nfs_server *server = NFS_SERVER(inode);
8023         size_t max_pages = max_response_pages(server);
8024         struct rpc_task *task;
8025         struct rpc_message msg = {
8026                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8027                 .rpc_argp = &lgp->args,
8028                 .rpc_resp = &lgp->res,
8029                 .rpc_cred = lgp->cred,
8030         };
8031         struct rpc_task_setup task_setup_data = {
8032                 .rpc_client = server->client,
8033                 .rpc_message = &msg,
8034                 .callback_ops = &nfs4_layoutget_call_ops,
8035                 .callback_data = lgp,
8036                 .flags = RPC_TASK_ASYNC,
8037         };
8038         struct pnfs_layout_segment *lseg = NULL;
8039         struct nfs4_exception exception = { .timeout = *timeout };
8040         int status = 0;
8041
8042         dprintk("--> %s\n", __func__);
8043
8044         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8045         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8046
8047         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8048         if (!lgp->args.layout.pages) {
8049                 nfs4_layoutget_release(lgp);
8050                 return ERR_PTR(-ENOMEM);
8051         }
8052         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8053
8054         lgp->res.layoutp = &lgp->args.layout;
8055         lgp->res.seq_res.sr_slot = NULL;
8056         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8057
8058         task = rpc_run_task(&task_setup_data);
8059         if (IS_ERR(task))
8060                 return ERR_CAST(task);
8061         status = nfs4_wait_for_completion_rpc_task(task);
8062         if (status == 0) {
8063                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8064                 *timeout = exception.timeout;
8065         }
8066
8067         trace_nfs4_layoutget(lgp->args.ctx,
8068                         &lgp->args.range,
8069                         &lgp->res.range,
8070                         &lgp->res.stateid,
8071                         status);
8072
8073         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8074         if (status == 0 && lgp->res.layoutp->len)
8075                 lseg = pnfs_layout_process(lgp);
8076         rpc_put_task(task);
8077         dprintk("<-- %s status=%d\n", __func__, status);
8078         if (status)
8079                 return ERR_PTR(status);
8080         return lseg;
8081 }
8082
8083 static void
8084 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8085 {
8086         struct nfs4_layoutreturn *lrp = calldata;
8087
8088         dprintk("--> %s\n", __func__);
8089         nfs41_setup_sequence(lrp->clp->cl_session,
8090                         &lrp->args.seq_args,
8091                         &lrp->res.seq_res,
8092                         task);
8093 }
8094
8095 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8096 {
8097         struct nfs4_layoutreturn *lrp = calldata;
8098         struct nfs_server *server;
8099
8100         dprintk("--> %s\n", __func__);
8101
8102         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8103                 return;
8104
8105         server = NFS_SERVER(lrp->args.inode);
8106         switch (task->tk_status) {
8107         default:
8108                 task->tk_status = 0;
8109         case 0:
8110                 break;
8111         case -NFS4ERR_DELAY:
8112                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8113                         break;
8114                 rpc_restart_call_prepare(task);
8115                 return;
8116         }
8117         dprintk("<-- %s\n", __func__);
8118 }
8119
8120 static void nfs4_layoutreturn_release(void *calldata)
8121 {
8122         struct nfs4_layoutreturn *lrp = calldata;
8123         struct pnfs_layout_hdr *lo = lrp->args.layout;
8124         LIST_HEAD(freeme);
8125
8126         dprintk("--> %s\n", __func__);
8127         spin_lock(&lo->plh_inode->i_lock);
8128         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range,
8129                         be32_to_cpu(lrp->args.stateid.seqid));
8130         pnfs_mark_layout_returned_if_empty(lo);
8131         if (lrp->res.lrs_present)
8132                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8133         pnfs_clear_layoutreturn_waitbit(lo);
8134         spin_unlock(&lo->plh_inode->i_lock);
8135         pnfs_free_lseg_list(&freeme);
8136         pnfs_put_layout_hdr(lrp->args.layout);
8137         nfs_iput_and_deactive(lrp->inode);
8138         kfree(calldata);
8139         dprintk("<-- %s\n", __func__);
8140 }
8141
8142 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8143         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8144         .rpc_call_done = nfs4_layoutreturn_done,
8145         .rpc_release = nfs4_layoutreturn_release,
8146 };
8147
8148 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8149 {
8150         struct rpc_task *task;
8151         struct rpc_message msg = {
8152                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8153                 .rpc_argp = &lrp->args,
8154                 .rpc_resp = &lrp->res,
8155                 .rpc_cred = lrp->cred,
8156         };
8157         struct rpc_task_setup task_setup_data = {
8158                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8159                 .rpc_message = &msg,
8160                 .callback_ops = &nfs4_layoutreturn_call_ops,
8161                 .callback_data = lrp,
8162         };
8163         int status = 0;
8164
8165         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8166                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8167                         &task_setup_data.rpc_client, &msg);
8168
8169         dprintk("--> %s\n", __func__);
8170         if (!sync) {
8171                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8172                 if (!lrp->inode) {
8173                         nfs4_layoutreturn_release(lrp);
8174                         return -EAGAIN;
8175                 }
8176                 task_setup_data.flags |= RPC_TASK_ASYNC;
8177         }
8178         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8179         task = rpc_run_task(&task_setup_data);
8180         if (IS_ERR(task))
8181                 return PTR_ERR(task);
8182         if (sync)
8183                 status = task->tk_status;
8184         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8185         dprintk("<-- %s status=%d\n", __func__, status);
8186         rpc_put_task(task);
8187         return status;
8188 }
8189
8190 static int
8191 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8192                 struct pnfs_device *pdev,
8193                 struct rpc_cred *cred)
8194 {
8195         struct nfs4_getdeviceinfo_args args = {
8196                 .pdev = pdev,
8197                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8198                         NOTIFY_DEVICEID4_DELETE,
8199         };
8200         struct nfs4_getdeviceinfo_res res = {
8201                 .pdev = pdev,
8202         };
8203         struct rpc_message msg = {
8204                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8205                 .rpc_argp = &args,
8206                 .rpc_resp = &res,
8207                 .rpc_cred = cred,
8208         };
8209         int status;
8210
8211         dprintk("--> %s\n", __func__);
8212         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8213         if (res.notification & ~args.notify_types)
8214                 dprintk("%s: unsupported notification\n", __func__);
8215         if (res.notification != args.notify_types)
8216                 pdev->nocache = 1;
8217
8218         dprintk("<-- %s status=%d\n", __func__, status);
8219
8220         return status;
8221 }
8222
8223 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8224                 struct pnfs_device *pdev,
8225                 struct rpc_cred *cred)
8226 {
8227         struct nfs4_exception exception = { };
8228         int err;
8229
8230         do {
8231                 err = nfs4_handle_exception(server,
8232                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8233                                         &exception);
8234         } while (exception.retry);
8235         return err;
8236 }
8237 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8238
8239 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8240 {
8241         struct nfs4_layoutcommit_data *data = calldata;
8242         struct nfs_server *server = NFS_SERVER(data->args.inode);
8243         struct nfs4_session *session = nfs4_get_session(server);
8244
8245         nfs41_setup_sequence(session,
8246                         &data->args.seq_args,
8247                         &data->res.seq_res,
8248                         task);
8249 }
8250
8251 static void
8252 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8253 {
8254         struct nfs4_layoutcommit_data *data = calldata;
8255         struct nfs_server *server = NFS_SERVER(data->args.inode);
8256
8257         if (!nfs41_sequence_done(task, &data->res.seq_res))
8258                 return;
8259
8260         switch (task->tk_status) { /* Just ignore these failures */
8261         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8262         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8263         case -NFS4ERR_BADLAYOUT:     /* no layout */
8264         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8265                 task->tk_status = 0;
8266         case 0:
8267                 break;
8268         default:
8269                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8270                         rpc_restart_call_prepare(task);
8271                         return;
8272                 }
8273         }
8274 }
8275
8276 static void nfs4_layoutcommit_release(void *calldata)
8277 {
8278         struct nfs4_layoutcommit_data *data = calldata;
8279
8280         pnfs_cleanup_layoutcommit(data);
8281         nfs_post_op_update_inode_force_wcc(data->args.inode,
8282                                            data->res.fattr);
8283         put_rpccred(data->cred);
8284         nfs_iput_and_deactive(data->inode);
8285         kfree(data);
8286 }
8287
8288 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8289         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8290         .rpc_call_done = nfs4_layoutcommit_done,
8291         .rpc_release = nfs4_layoutcommit_release,
8292 };
8293
8294 int
8295 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8296 {
8297         struct rpc_message msg = {
8298                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8299                 .rpc_argp = &data->args,
8300                 .rpc_resp = &data->res,
8301                 .rpc_cred = data->cred,
8302         };
8303         struct rpc_task_setup task_setup_data = {
8304                 .task = &data->task,
8305                 .rpc_client = NFS_CLIENT(data->args.inode),
8306                 .rpc_message = &msg,
8307                 .callback_ops = &nfs4_layoutcommit_ops,
8308                 .callback_data = data,
8309         };
8310         struct rpc_task *task;
8311         int status = 0;
8312
8313         dprintk("NFS: initiating layoutcommit call. sync %d "
8314                 "lbw: %llu inode %lu\n", sync,
8315                 data->args.lastbytewritten,
8316                 data->args.inode->i_ino);
8317
8318         if (!sync) {
8319                 data->inode = nfs_igrab_and_active(data->args.inode);
8320                 if (data->inode == NULL) {
8321                         nfs4_layoutcommit_release(data);
8322                         return -EAGAIN;
8323                 }
8324                 task_setup_data.flags = RPC_TASK_ASYNC;
8325         }
8326         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8327         task = rpc_run_task(&task_setup_data);
8328         if (IS_ERR(task))
8329                 return PTR_ERR(task);
8330         if (sync)
8331                 status = task->tk_status;
8332         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8333         dprintk("%s: status %d\n", __func__, status);
8334         rpc_put_task(task);
8335         return status;
8336 }
8337
8338 /**
8339  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8340  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8341  */
8342 static int
8343 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8344                     struct nfs_fsinfo *info,
8345                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8346 {
8347         struct nfs41_secinfo_no_name_args args = {
8348                 .style = SECINFO_STYLE_CURRENT_FH,
8349         };
8350         struct nfs4_secinfo_res res = {
8351                 .flavors = flavors,
8352         };
8353         struct rpc_message msg = {
8354                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8355                 .rpc_argp = &args,
8356                 .rpc_resp = &res,
8357         };
8358         struct rpc_clnt *clnt = server->client;
8359         struct rpc_cred *cred = NULL;
8360         int status;
8361
8362         if (use_integrity) {
8363                 clnt = server->nfs_client->cl_rpcclient;
8364                 cred = nfs4_get_clid_cred(server->nfs_client);
8365                 msg.rpc_cred = cred;
8366         }
8367
8368         dprintk("--> %s\n", __func__);
8369         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8370                                 &res.seq_res, 0);
8371         dprintk("<-- %s status=%d\n", __func__, status);
8372
8373         if (cred)
8374                 put_rpccred(cred);
8375
8376         return status;
8377 }
8378
8379 static int
8380 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8381                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8382 {
8383         struct nfs4_exception exception = { };
8384         int err;
8385         do {
8386                 /* first try using integrity protection */
8387                 err = -NFS4ERR_WRONGSEC;
8388
8389                 /* try to use integrity protection with machine cred */
8390                 if (_nfs4_is_integrity_protected(server->nfs_client))
8391                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8392                                                           flavors, true);
8393
8394                 /*
8395                  * if unable to use integrity protection, or SECINFO with
8396                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8397                  * disallowed by spec, but exists in deployed servers) use
8398                  * the current filesystem's rpc_client and the user cred.
8399                  */
8400                 if (err == -NFS4ERR_WRONGSEC)
8401                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8402                                                           flavors, false);
8403
8404                 switch (err) {
8405                 case 0:
8406                 case -NFS4ERR_WRONGSEC:
8407                 case -ENOTSUPP:
8408                         goto out;
8409                 default:
8410                         err = nfs4_handle_exception(server, err, &exception);
8411                 }
8412         } while (exception.retry);
8413 out:
8414         return err;
8415 }
8416
8417 static int
8418 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8419                     struct nfs_fsinfo *info)
8420 {
8421         int err;
8422         struct page *page;
8423         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8424         struct nfs4_secinfo_flavors *flavors;
8425         struct nfs4_secinfo4 *secinfo;
8426         int i;
8427
8428         page = alloc_page(GFP_KERNEL);
8429         if (!page) {
8430                 err = -ENOMEM;
8431                 goto out;
8432         }
8433
8434         flavors = page_address(page);
8435         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8436
8437         /*
8438          * Fall back on "guess and check" method if
8439          * the server doesn't support SECINFO_NO_NAME
8440          */
8441         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8442                 err = nfs4_find_root_sec(server, fhandle, info);
8443                 goto out_freepage;
8444         }
8445         if (err)
8446                 goto out_freepage;
8447
8448         for (i = 0; i < flavors->num_flavors; i++) {
8449                 secinfo = &flavors->flavors[i];
8450
8451                 switch (secinfo->flavor) {
8452                 case RPC_AUTH_NULL:
8453                 case RPC_AUTH_UNIX:
8454                 case RPC_AUTH_GSS:
8455                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8456                                         &secinfo->flavor_info);
8457                         break;
8458                 default:
8459                         flavor = RPC_AUTH_MAXFLAVOR;
8460                         break;
8461                 }
8462
8463                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8464                         flavor = RPC_AUTH_MAXFLAVOR;
8465
8466                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8467                         err = nfs4_lookup_root_sec(server, fhandle,
8468                                                    info, flavor);
8469                         if (!err)
8470                                 break;
8471                 }
8472         }
8473
8474         if (flavor == RPC_AUTH_MAXFLAVOR)
8475                 err = -EPERM;
8476
8477 out_freepage:
8478         put_page(page);
8479         if (err == -EACCES)
8480                 return -EPERM;
8481 out:
8482         return err;
8483 }
8484
8485 static int _nfs41_test_stateid(struct nfs_server *server,
8486                 nfs4_stateid *stateid,
8487                 struct rpc_cred *cred)
8488 {
8489         int status;
8490         struct nfs41_test_stateid_args args = {
8491                 .stateid = stateid,
8492         };
8493         struct nfs41_test_stateid_res res;
8494         struct rpc_message msg = {
8495                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8496                 .rpc_argp = &args,
8497                 .rpc_resp = &res,
8498                 .rpc_cred = cred,
8499         };
8500         struct rpc_clnt *rpc_client = server->client;
8501
8502         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8503                 &rpc_client, &msg);
8504
8505         dprintk("NFS call  test_stateid %p\n", stateid);
8506         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8507         nfs4_set_sequence_privileged(&args.seq_args);
8508         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8509                         &args.seq_args, &res.seq_res);
8510         if (status != NFS_OK) {
8511                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8512                 return status;
8513         }
8514         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8515         return -res.status;
8516 }
8517
8518 /**
8519  * nfs41_test_stateid - perform a TEST_STATEID operation
8520  *
8521  * @server: server / transport on which to perform the operation
8522  * @stateid: state ID to test
8523  * @cred: credential
8524  *
8525  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8526  * Otherwise a negative NFS4ERR value is returned if the operation
8527  * failed or the state ID is not currently valid.
8528  */
8529 static int nfs41_test_stateid(struct nfs_server *server,
8530                 nfs4_stateid *stateid,
8531                 struct rpc_cred *cred)
8532 {
8533         struct nfs4_exception exception = { };
8534         int err;
8535         do {
8536                 err = _nfs41_test_stateid(server, stateid, cred);
8537                 if (err != -NFS4ERR_DELAY)
8538                         break;
8539                 nfs4_handle_exception(server, err, &exception);
8540         } while (exception.retry);
8541         return err;
8542 }
8543
8544 struct nfs_free_stateid_data {
8545         struct nfs_server *server;
8546         struct nfs41_free_stateid_args args;
8547         struct nfs41_free_stateid_res res;
8548 };
8549
8550 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8551 {
8552         struct nfs_free_stateid_data *data = calldata;
8553         nfs41_setup_sequence(nfs4_get_session(data->server),
8554                         &data->args.seq_args,
8555                         &data->res.seq_res,
8556                         task);
8557 }
8558
8559 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8560 {
8561         struct nfs_free_stateid_data *data = calldata;
8562
8563         nfs41_sequence_done(task, &data->res.seq_res);
8564
8565         switch (task->tk_status) {
8566         case -NFS4ERR_DELAY:
8567                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8568                         rpc_restart_call_prepare(task);
8569         }
8570 }
8571
8572 static void nfs41_free_stateid_release(void *calldata)
8573 {
8574         kfree(calldata);
8575 }
8576
8577 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8578         .rpc_call_prepare = nfs41_free_stateid_prepare,
8579         .rpc_call_done = nfs41_free_stateid_done,
8580         .rpc_release = nfs41_free_stateid_release,
8581 };
8582
8583 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8584                 nfs4_stateid *stateid,
8585                 struct rpc_cred *cred,
8586                 bool privileged)
8587 {
8588         struct rpc_message msg = {
8589                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8590                 .rpc_cred = cred,
8591         };
8592         struct rpc_task_setup task_setup = {
8593                 .rpc_client = server->client,
8594                 .rpc_message = &msg,
8595                 .callback_ops = &nfs41_free_stateid_ops,
8596                 .flags = RPC_TASK_ASYNC,
8597         };
8598         struct nfs_free_stateid_data *data;
8599
8600         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8601                 &task_setup.rpc_client, &msg);
8602
8603         dprintk("NFS call  free_stateid %p\n", stateid);
8604         data = kmalloc(sizeof(*data), GFP_NOFS);
8605         if (!data)
8606                 return ERR_PTR(-ENOMEM);
8607         data->server = server;
8608         nfs4_stateid_copy(&data->args.stateid, stateid);
8609
8610         task_setup.callback_data = data;
8611
8612         msg.rpc_argp = &data->args;
8613         msg.rpc_resp = &data->res;
8614         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8615         if (privileged)
8616                 nfs4_set_sequence_privileged(&data->args.seq_args);
8617
8618         return rpc_run_task(&task_setup);
8619 }
8620
8621 /**
8622  * nfs41_free_stateid - perform a FREE_STATEID operation
8623  *
8624  * @server: server / transport on which to perform the operation
8625  * @stateid: state ID to release
8626  * @cred: credential
8627  *
8628  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8629  * negative NFS4ERR value is returned.
8630  */
8631 static int nfs41_free_stateid(struct nfs_server *server,
8632                 nfs4_stateid *stateid,
8633                 struct rpc_cred *cred)
8634 {
8635         struct rpc_task *task;
8636         int ret;
8637
8638         task = _nfs41_free_stateid(server, stateid, cred, true);
8639         if (IS_ERR(task))
8640                 return PTR_ERR(task);
8641         ret = rpc_wait_for_completion_task(task);
8642         if (!ret)
8643                 ret = task->tk_status;
8644         rpc_put_task(task);
8645         return ret;
8646 }
8647
8648 static void
8649 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8650 {
8651         struct rpc_task *task;
8652         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8653
8654         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8655         nfs4_free_lock_state(server, lsp);
8656         if (IS_ERR(task))
8657                 return;
8658         rpc_put_task(task);
8659 }
8660
8661 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8662                 const nfs4_stateid *s2)
8663 {
8664         if (s1->type != s2->type)
8665                 return false;
8666
8667         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8668                 return false;
8669
8670         if (s1->seqid == s2->seqid)
8671                 return true;
8672         if (s1->seqid == 0 || s2->seqid == 0)
8673                 return true;
8674
8675         return false;
8676 }
8677
8678 #endif /* CONFIG_NFS_V4_1 */
8679
8680 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8681                 const nfs4_stateid *s2)
8682 {
8683         return nfs4_stateid_match(s1, s2);
8684 }
8685
8686
8687 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8688         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8689         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8690         .recover_open   = nfs4_open_reclaim,
8691         .recover_lock   = nfs4_lock_reclaim,
8692         .establish_clid = nfs4_init_clientid,
8693         .detect_trunking = nfs40_discover_server_trunking,
8694 };
8695
8696 #if defined(CONFIG_NFS_V4_1)
8697 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8698         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8699         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8700         .recover_open   = nfs4_open_reclaim,
8701         .recover_lock   = nfs4_lock_reclaim,
8702         .establish_clid = nfs41_init_clientid,
8703         .reclaim_complete = nfs41_proc_reclaim_complete,
8704         .detect_trunking = nfs41_discover_server_trunking,
8705 };
8706 #endif /* CONFIG_NFS_V4_1 */
8707
8708 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8709         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8710         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8711         .recover_open   = nfs40_open_expired,
8712         .recover_lock   = nfs4_lock_expired,
8713         .establish_clid = nfs4_init_clientid,
8714 };
8715
8716 #if defined(CONFIG_NFS_V4_1)
8717 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8718         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8719         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8720         .recover_open   = nfs41_open_expired,
8721         .recover_lock   = nfs41_lock_expired,
8722         .establish_clid = nfs41_init_clientid,
8723 };
8724 #endif /* CONFIG_NFS_V4_1 */
8725
8726 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8727         .sched_state_renewal = nfs4_proc_async_renew,
8728         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8729         .renew_lease = nfs4_proc_renew,
8730 };
8731
8732 #if defined(CONFIG_NFS_V4_1)
8733 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8734         .sched_state_renewal = nfs41_proc_async_sequence,
8735         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8736         .renew_lease = nfs4_proc_sequence,
8737 };
8738 #endif
8739
8740 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8741         .get_locations = _nfs40_proc_get_locations,
8742         .fsid_present = _nfs40_proc_fsid_present,
8743 };
8744
8745 #if defined(CONFIG_NFS_V4_1)
8746 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8747         .get_locations = _nfs41_proc_get_locations,
8748         .fsid_present = _nfs41_proc_fsid_present,
8749 };
8750 #endif  /* CONFIG_NFS_V4_1 */
8751
8752 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8753         .minor_version = 0,
8754         .init_caps = NFS_CAP_READDIRPLUS
8755                 | NFS_CAP_ATOMIC_OPEN
8756                 | NFS_CAP_POSIX_LOCK,
8757         .init_client = nfs40_init_client,
8758         .shutdown_client = nfs40_shutdown_client,
8759         .match_stateid = nfs4_match_stateid,
8760         .find_root_sec = nfs4_find_root_sec,
8761         .free_lock_state = nfs4_release_lockowner,
8762         .alloc_seqid = nfs_alloc_seqid,
8763         .call_sync_ops = &nfs40_call_sync_ops,
8764         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8765         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8766         .state_renewal_ops = &nfs40_state_renewal_ops,
8767         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8768 };
8769
8770 #if defined(CONFIG_NFS_V4_1)
8771 static struct nfs_seqid *
8772 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8773 {
8774         return NULL;
8775 }
8776
8777 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8778         .minor_version = 1,
8779         .init_caps = NFS_CAP_READDIRPLUS
8780                 | NFS_CAP_ATOMIC_OPEN
8781                 | NFS_CAP_POSIX_LOCK
8782                 | NFS_CAP_STATEID_NFSV41
8783                 | NFS_CAP_ATOMIC_OPEN_V1,
8784         .init_client = nfs41_init_client,
8785         .shutdown_client = nfs41_shutdown_client,
8786         .match_stateid = nfs41_match_stateid,
8787         .find_root_sec = nfs41_find_root_sec,
8788         .free_lock_state = nfs41_free_lock_state,
8789         .alloc_seqid = nfs_alloc_no_seqid,
8790         .call_sync_ops = &nfs41_call_sync_ops,
8791         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8792         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8793         .state_renewal_ops = &nfs41_state_renewal_ops,
8794         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8795 };
8796 #endif
8797
8798 #if defined(CONFIG_NFS_V4_2)
8799 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8800         .minor_version = 2,
8801         .init_caps = NFS_CAP_READDIRPLUS
8802                 | NFS_CAP_ATOMIC_OPEN
8803                 | NFS_CAP_POSIX_LOCK
8804                 | NFS_CAP_STATEID_NFSV41
8805                 | NFS_CAP_ATOMIC_OPEN_V1
8806                 | NFS_CAP_ALLOCATE
8807                 | NFS_CAP_COPY
8808                 | NFS_CAP_DEALLOCATE
8809                 | NFS_CAP_SEEK
8810                 | NFS_CAP_LAYOUTSTATS
8811                 | NFS_CAP_CLONE,
8812         .init_client = nfs41_init_client,
8813         .shutdown_client = nfs41_shutdown_client,
8814         .match_stateid = nfs41_match_stateid,
8815         .find_root_sec = nfs41_find_root_sec,
8816         .free_lock_state = nfs41_free_lock_state,
8817         .call_sync_ops = &nfs41_call_sync_ops,
8818         .alloc_seqid = nfs_alloc_no_seqid,
8819         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8820         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8821         .state_renewal_ops = &nfs41_state_renewal_ops,
8822         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8823 };
8824 #endif
8825
8826 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8827         [0] = &nfs_v4_0_minor_ops,
8828 #if defined(CONFIG_NFS_V4_1)
8829         [1] = &nfs_v4_1_minor_ops,
8830 #endif
8831 #if defined(CONFIG_NFS_V4_2)
8832         [2] = &nfs_v4_2_minor_ops,
8833 #endif
8834 };
8835
8836 ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
8837 {
8838         ssize_t error, error2;
8839
8840         error = generic_listxattr(dentry, list, size);
8841         if (error < 0)
8842                 return error;
8843         if (list) {
8844                 list += error;
8845                 size -= error;
8846         }
8847
8848         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
8849         if (error2 < 0)
8850                 return error2;
8851         return error + error2;
8852 }
8853
8854 static const struct inode_operations nfs4_dir_inode_operations = {
8855         .create         = nfs_create,
8856         .lookup         = nfs_lookup,
8857         .atomic_open    = nfs_atomic_open,
8858         .link           = nfs_link,
8859         .unlink         = nfs_unlink,
8860         .symlink        = nfs_symlink,
8861         .mkdir          = nfs_mkdir,
8862         .rmdir          = nfs_rmdir,
8863         .mknod          = nfs_mknod,
8864         .rename         = nfs_rename,
8865         .permission     = nfs_permission,
8866         .getattr        = nfs_getattr,
8867         .setattr        = nfs_setattr,
8868         .getxattr       = generic_getxattr,
8869         .setxattr       = generic_setxattr,
8870         .listxattr      = nfs4_listxattr,
8871         .removexattr    = generic_removexattr,
8872 };
8873
8874 static const struct inode_operations nfs4_file_inode_operations = {
8875         .permission     = nfs_permission,
8876         .getattr        = nfs_getattr,
8877         .setattr        = nfs_setattr,
8878         .getxattr       = generic_getxattr,
8879         .setxattr       = generic_setxattr,
8880         .listxattr      = nfs4_listxattr,
8881         .removexattr    = generic_removexattr,
8882 };
8883
8884 const struct nfs_rpc_ops nfs_v4_clientops = {
8885         .version        = 4,                    /* protocol version */
8886         .dentry_ops     = &nfs4_dentry_operations,
8887         .dir_inode_ops  = &nfs4_dir_inode_operations,
8888         .file_inode_ops = &nfs4_file_inode_operations,
8889         .file_ops       = &nfs4_file_operations,
8890         .getroot        = nfs4_proc_get_root,
8891         .submount       = nfs4_submount,
8892         .try_mount      = nfs4_try_mount,
8893         .getattr        = nfs4_proc_getattr,
8894         .setattr        = nfs4_proc_setattr,
8895         .lookup         = nfs4_proc_lookup,
8896         .access         = nfs4_proc_access,
8897         .readlink       = nfs4_proc_readlink,
8898         .create         = nfs4_proc_create,
8899         .remove         = nfs4_proc_remove,
8900         .unlink_setup   = nfs4_proc_unlink_setup,
8901         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8902         .unlink_done    = nfs4_proc_unlink_done,
8903         .rename_setup   = nfs4_proc_rename_setup,
8904         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8905         .rename_done    = nfs4_proc_rename_done,
8906         .link           = nfs4_proc_link,
8907         .symlink        = nfs4_proc_symlink,
8908         .mkdir          = nfs4_proc_mkdir,
8909         .rmdir          = nfs4_proc_remove,
8910         .readdir        = nfs4_proc_readdir,
8911         .mknod          = nfs4_proc_mknod,
8912         .statfs         = nfs4_proc_statfs,
8913         .fsinfo         = nfs4_proc_fsinfo,
8914         .pathconf       = nfs4_proc_pathconf,
8915         .set_capabilities = nfs4_server_capabilities,
8916         .decode_dirent  = nfs4_decode_dirent,
8917         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8918         .read_setup     = nfs4_proc_read_setup,
8919         .read_done      = nfs4_read_done,
8920         .write_setup    = nfs4_proc_write_setup,
8921         .write_done     = nfs4_write_done,
8922         .commit_setup   = nfs4_proc_commit_setup,
8923         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8924         .commit_done    = nfs4_commit_done,
8925         .lock           = nfs4_proc_lock,
8926         .clear_acl_cache = nfs4_zap_acl_attr,
8927         .close_context  = nfs4_close_context,
8928         .open_context   = nfs4_atomic_open,
8929         .have_delegation = nfs4_have_delegation,
8930         .return_delegation = nfs4_inode_return_delegation,
8931         .alloc_client   = nfs4_alloc_client,
8932         .init_client    = nfs4_init_client,
8933         .free_client    = nfs4_free_client,
8934         .create_server  = nfs4_create_server,
8935         .clone_server   = nfs_clone_server,
8936 };
8937
8938 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8939         .name   = XATTR_NAME_NFSV4_ACL,
8940         .list   = nfs4_xattr_list_nfs4_acl,
8941         .get    = nfs4_xattr_get_nfs4_acl,
8942         .set    = nfs4_xattr_set_nfs4_acl,
8943 };
8944
8945 const struct xattr_handler *nfs4_xattr_handlers[] = {
8946         &nfs4_xattr_nfs4_acl_handler,
8947 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8948         &nfs4_xattr_nfs4_label_handler,
8949 #endif
8950         NULL
8951 };
8952
8953 /*
8954  * Local variables:
8955  *  c-basic-offset: 8
8956  * End:
8957  */