Merge tag 'pci-v4.8-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci
[sfrench/cifs-2.6.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/file.h>
42 #include <linux/string.h>
43 #include <linux/ratelimit.h>
44 #include <linux/printk.h>
45 #include <linux/slab.h>
46 #include <linux/sunrpc/clnt.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69
70 #include "nfs4trace.h"
71
72 #define NFSDBG_FACILITY         NFSDBG_PROC
73
74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
76
77 /* file attributes which can be mapped to nfs attributes */
78 #define NFS4_VALID_ATTRS (ATTR_MODE \
79         | ATTR_UID \
80         | ATTR_GID \
81         | ATTR_SIZE \
82         | ATTR_ATIME \
83         | ATTR_MTIME \
84         | ATTR_CTIME \
85         | ATTR_ATIME_SET \
86         | ATTR_MTIME_SET)
87
88 struct nfs4_opendata;
89 static int _nfs4_proc_open(struct nfs4_opendata *data);
90 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
91 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
92 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
93 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
95 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
96                             struct nfs_fattr *fattr, struct iattr *sattr,
97                             struct nfs4_state *state, struct nfs4_label *ilabel,
98                             struct nfs4_label *olabel);
99 #ifdef CONFIG_NFS_V4_1
100 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
101                 struct rpc_cred *);
102 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
103                 struct rpc_cred *);
104 #endif
105
106 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
107 static inline struct nfs4_label *
108 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
109         struct iattr *sattr, struct nfs4_label *label)
110 {
111         int err;
112
113         if (label == NULL)
114                 return NULL;
115
116         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
117                 return NULL;
118
119         err = security_dentry_init_security(dentry, sattr->ia_mode,
120                                 &dentry->d_name, (void **)&label->label, &label->len);
121         if (err == 0)
122                 return label;
123
124         return NULL;
125 }
126 static inline void
127 nfs4_label_release_security(struct nfs4_label *label)
128 {
129         if (label)
130                 security_release_secctx(label->label, label->len);
131 }
132 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
133 {
134         if (label)
135                 return server->attr_bitmask;
136
137         return server->attr_bitmask_nl;
138 }
139 #else
140 static inline struct nfs4_label *
141 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
142         struct iattr *sattr, struct nfs4_label *l)
143 { return NULL; }
144 static inline void
145 nfs4_label_release_security(struct nfs4_label *label)
146 { return; }
147 static inline u32 *
148 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
149 { return server->attr_bitmask; }
150 #endif
151
152 /* Prevent leaks of NFSv4 errors into userland */
153 static int nfs4_map_errors(int err)
154 {
155         if (err >= -1000)
156                 return err;
157         switch (err) {
158         case -NFS4ERR_RESOURCE:
159         case -NFS4ERR_LAYOUTTRYLATER:
160         case -NFS4ERR_RECALLCONFLICT:
161                 return -EREMOTEIO;
162         case -NFS4ERR_WRONGSEC:
163         case -NFS4ERR_WRONG_CRED:
164                 return -EPERM;
165         case -NFS4ERR_BADOWNER:
166         case -NFS4ERR_BADNAME:
167                 return -EINVAL;
168         case -NFS4ERR_SHARE_DENIED:
169                 return -EACCES;
170         case -NFS4ERR_MINOR_VERS_MISMATCH:
171                 return -EPROTONOSUPPORT;
172         case -NFS4ERR_FILE_OPEN:
173                 return -EBUSY;
174         default:
175                 dprintk("%s could not handle NFSv4 error %d\n",
176                                 __func__, -err);
177                 break;
178         }
179         return -EIO;
180 }
181
182 /*
183  * This is our standard bitmap for GETATTR requests.
184  */
185 const u32 nfs4_fattr_bitmap[3] = {
186         FATTR4_WORD0_TYPE
187         | FATTR4_WORD0_CHANGE
188         | FATTR4_WORD0_SIZE
189         | FATTR4_WORD0_FSID
190         | FATTR4_WORD0_FILEID,
191         FATTR4_WORD1_MODE
192         | FATTR4_WORD1_NUMLINKS
193         | FATTR4_WORD1_OWNER
194         | FATTR4_WORD1_OWNER_GROUP
195         | FATTR4_WORD1_RAWDEV
196         | FATTR4_WORD1_SPACE_USED
197         | FATTR4_WORD1_TIME_ACCESS
198         | FATTR4_WORD1_TIME_METADATA
199         | FATTR4_WORD1_TIME_MODIFY
200         | FATTR4_WORD1_MOUNTED_ON_FILEID,
201 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
202         FATTR4_WORD2_SECURITY_LABEL
203 #endif
204 };
205
206 static const u32 nfs4_pnfs_open_bitmap[3] = {
207         FATTR4_WORD0_TYPE
208         | FATTR4_WORD0_CHANGE
209         | FATTR4_WORD0_SIZE
210         | FATTR4_WORD0_FSID
211         | FATTR4_WORD0_FILEID,
212         FATTR4_WORD1_MODE
213         | FATTR4_WORD1_NUMLINKS
214         | FATTR4_WORD1_OWNER
215         | FATTR4_WORD1_OWNER_GROUP
216         | FATTR4_WORD1_RAWDEV
217         | FATTR4_WORD1_SPACE_USED
218         | FATTR4_WORD1_TIME_ACCESS
219         | FATTR4_WORD1_TIME_METADATA
220         | FATTR4_WORD1_TIME_MODIFY,
221         FATTR4_WORD2_MDSTHRESHOLD
222 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
223         | FATTR4_WORD2_SECURITY_LABEL
224 #endif
225 };
226
227 static const u32 nfs4_open_noattr_bitmap[3] = {
228         FATTR4_WORD0_TYPE
229         | FATTR4_WORD0_CHANGE
230         | FATTR4_WORD0_FILEID,
231 };
232
233 const u32 nfs4_statfs_bitmap[3] = {
234         FATTR4_WORD0_FILES_AVAIL
235         | FATTR4_WORD0_FILES_FREE
236         | FATTR4_WORD0_FILES_TOTAL,
237         FATTR4_WORD1_SPACE_AVAIL
238         | FATTR4_WORD1_SPACE_FREE
239         | FATTR4_WORD1_SPACE_TOTAL
240 };
241
242 const u32 nfs4_pathconf_bitmap[3] = {
243         FATTR4_WORD0_MAXLINK
244         | FATTR4_WORD0_MAXNAME,
245         0
246 };
247
248 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
249                         | FATTR4_WORD0_MAXREAD
250                         | FATTR4_WORD0_MAXWRITE
251                         | FATTR4_WORD0_LEASE_TIME,
252                         FATTR4_WORD1_TIME_DELTA
253                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
254                         FATTR4_WORD2_LAYOUT_BLKSIZE
255                         | FATTR4_WORD2_CLONE_BLKSIZE
256 };
257
258 const u32 nfs4_fs_locations_bitmap[3] = {
259         FATTR4_WORD0_TYPE
260         | FATTR4_WORD0_CHANGE
261         | FATTR4_WORD0_SIZE
262         | FATTR4_WORD0_FSID
263         | FATTR4_WORD0_FILEID
264         | FATTR4_WORD0_FS_LOCATIONS,
265         FATTR4_WORD1_MODE
266         | FATTR4_WORD1_NUMLINKS
267         | FATTR4_WORD1_OWNER
268         | FATTR4_WORD1_OWNER_GROUP
269         | FATTR4_WORD1_RAWDEV
270         | FATTR4_WORD1_SPACE_USED
271         | FATTR4_WORD1_TIME_ACCESS
272         | FATTR4_WORD1_TIME_METADATA
273         | FATTR4_WORD1_TIME_MODIFY
274         | FATTR4_WORD1_MOUNTED_ON_FILEID,
275 };
276
277 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
278                 struct nfs4_readdir_arg *readdir)
279 {
280         __be32 *start, *p;
281
282         if (cookie > 2) {
283                 readdir->cookie = cookie;
284                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
285                 return;
286         }
287
288         readdir->cookie = 0;
289         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
290         if (cookie == 2)
291                 return;
292         
293         /*
294          * NFSv4 servers do not return entries for '.' and '..'
295          * Therefore, we fake these entries here.  We let '.'
296          * have cookie 0 and '..' have cookie 1.  Note that
297          * when talking to the server, we always send cookie 0
298          * instead of 1 or 2.
299          */
300         start = p = kmap_atomic(*readdir->pages);
301         
302         if (cookie == 0) {
303                 *p++ = xdr_one;                                  /* next */
304                 *p++ = xdr_zero;                   /* cookie, first word */
305                 *p++ = xdr_one;                   /* cookie, second word */
306                 *p++ = xdr_one;                             /* entry len */
307                 memcpy(p, ".\0\0\0", 4);                        /* entry */
308                 p++;
309                 *p++ = xdr_one;                         /* bitmap length */
310                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
311                 *p++ = htonl(8);              /* attribute buffer length */
312                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
313         }
314         
315         *p++ = xdr_one;                                  /* next */
316         *p++ = xdr_zero;                   /* cookie, first word */
317         *p++ = xdr_two;                   /* cookie, second word */
318         *p++ = xdr_two;                             /* entry len */
319         memcpy(p, "..\0\0", 4);                         /* entry */
320         p++;
321         *p++ = xdr_one;                         /* bitmap length */
322         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
323         *p++ = htonl(8);              /* attribute buffer length */
324         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
325
326         readdir->pgbase = (char *)p - (char *)start;
327         readdir->count -= readdir->pgbase;
328         kunmap_atomic(start);
329 }
330
331 static long nfs4_update_delay(long *timeout)
332 {
333         long ret;
334         if (!timeout)
335                 return NFS4_POLL_RETRY_MAX;
336         if (*timeout <= 0)
337                 *timeout = NFS4_POLL_RETRY_MIN;
338         if (*timeout > NFS4_POLL_RETRY_MAX)
339                 *timeout = NFS4_POLL_RETRY_MAX;
340         ret = *timeout;
341         *timeout <<= 1;
342         return ret;
343 }
344
345 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
346 {
347         int res = 0;
348
349         might_sleep();
350
351         freezable_schedule_timeout_killable_unsafe(
352                 nfs4_update_delay(timeout));
353         if (fatal_signal_pending(current))
354                 res = -ERESTARTSYS;
355         return res;
356 }
357
358 /* This is the error handling routine for processes that are allowed
359  * to sleep.
360  */
361 static int nfs4_do_handle_exception(struct nfs_server *server,
362                 int errorcode, struct nfs4_exception *exception)
363 {
364         struct nfs_client *clp = server->nfs_client;
365         struct nfs4_state *state = exception->state;
366         const nfs4_stateid *stateid = exception->stateid;
367         struct inode *inode = exception->inode;
368         int ret = errorcode;
369
370         exception->delay = 0;
371         exception->recovering = 0;
372         exception->retry = 0;
373         switch(errorcode) {
374                 case 0:
375                         return 0;
376                 case -NFS4ERR_OPENMODE:
377                 case -NFS4ERR_DELEG_REVOKED:
378                 case -NFS4ERR_ADMIN_REVOKED:
379                 case -NFS4ERR_BAD_STATEID:
380                         if (inode) {
381                                 int err;
382
383                                 err = nfs_async_inode_return_delegation(inode,
384                                                 stateid);
385                                 if (err == 0)
386                                         goto wait_on_recovery;
387                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
388                                         exception->retry = 1;
389                                         break;
390                                 }
391                         }
392                         if (state == NULL)
393                                 break;
394                         ret = nfs4_schedule_stateid_recovery(server, state);
395                         if (ret < 0)
396                                 break;
397                         goto wait_on_recovery;
398                 case -NFS4ERR_EXPIRED:
399                         if (state != NULL) {
400                                 ret = nfs4_schedule_stateid_recovery(server, state);
401                                 if (ret < 0)
402                                         break;
403                         }
404                 case -NFS4ERR_STALE_STATEID:
405                 case -NFS4ERR_STALE_CLIENTID:
406                         nfs4_schedule_lease_recovery(clp);
407                         goto wait_on_recovery;
408                 case -NFS4ERR_MOVED:
409                         ret = nfs4_schedule_migration_recovery(server);
410                         if (ret < 0)
411                                 break;
412                         goto wait_on_recovery;
413                 case -NFS4ERR_LEASE_MOVED:
414                         nfs4_schedule_lease_moved_recovery(clp);
415                         goto wait_on_recovery;
416 #if defined(CONFIG_NFS_V4_1)
417                 case -NFS4ERR_BADSESSION:
418                 case -NFS4ERR_BADSLOT:
419                 case -NFS4ERR_BAD_HIGH_SLOT:
420                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
421                 case -NFS4ERR_DEADSESSION:
422                 case -NFS4ERR_SEQ_FALSE_RETRY:
423                 case -NFS4ERR_SEQ_MISORDERED:
424                         dprintk("%s ERROR: %d Reset session\n", __func__,
425                                 errorcode);
426                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
427                         goto wait_on_recovery;
428 #endif /* defined(CONFIG_NFS_V4_1) */
429                 case -NFS4ERR_FILE_OPEN:
430                         if (exception->timeout > HZ) {
431                                 /* We have retried a decent amount, time to
432                                  * fail
433                                  */
434                                 ret = -EBUSY;
435                                 break;
436                         }
437                 case -NFS4ERR_DELAY:
438                         nfs_inc_server_stats(server, NFSIOS_DELAY);
439                 case -NFS4ERR_GRACE:
440                 case -NFS4ERR_LAYOUTTRYLATER:
441                 case -NFS4ERR_RECALLCONFLICT:
442                         exception->delay = 1;
443                         return 0;
444
445                 case -NFS4ERR_RETRY_UNCACHED_REP:
446                 case -NFS4ERR_OLD_STATEID:
447                         exception->retry = 1;
448                         break;
449                 case -NFS4ERR_BADOWNER:
450                         /* The following works around a Linux server bug! */
451                 case -NFS4ERR_BADNAME:
452                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
453                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
454                                 exception->retry = 1;
455                                 printk(KERN_WARNING "NFS: v4 server %s "
456                                                 "does not accept raw "
457                                                 "uid/gids. "
458                                                 "Reenabling the idmapper.\n",
459                                                 server->nfs_client->cl_hostname);
460                         }
461         }
462         /* We failed to handle the error */
463         return nfs4_map_errors(ret);
464 wait_on_recovery:
465         exception->recovering = 1;
466         return 0;
467 }
468
469 /* This is the error handling routine for processes that are allowed
470  * to sleep.
471  */
472 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
473 {
474         struct nfs_client *clp = server->nfs_client;
475         int ret;
476
477         ret = nfs4_do_handle_exception(server, errorcode, exception);
478         if (exception->delay) {
479                 ret = nfs4_delay(server->client, &exception->timeout);
480                 goto out_retry;
481         }
482         if (exception->recovering) {
483                 ret = nfs4_wait_clnt_recover(clp);
484                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
485                         return -EIO;
486                 goto out_retry;
487         }
488         return ret;
489 out_retry:
490         if (ret == 0)
491                 exception->retry = 1;
492         return ret;
493 }
494
495 static int
496 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
497                 int errorcode, struct nfs4_exception *exception)
498 {
499         struct nfs_client *clp = server->nfs_client;
500         int ret;
501
502         ret = nfs4_do_handle_exception(server, errorcode, exception);
503         if (exception->delay) {
504                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
505                 goto out_retry;
506         }
507         if (exception->recovering) {
508                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
509                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
510                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
511                 goto out_retry;
512         }
513         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
514                 ret = -EIO;
515         return ret;
516 out_retry:
517         if (ret == 0)
518                 exception->retry = 1;
519         return ret;
520 }
521
522 static int
523 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
524                         struct nfs4_state *state, long *timeout)
525 {
526         struct nfs4_exception exception = {
527                 .state = state,
528         };
529
530         if (task->tk_status >= 0)
531                 return 0;
532         if (timeout)
533                 exception.timeout = *timeout;
534         task->tk_status = nfs4_async_handle_exception(task, server,
535                         task->tk_status,
536                         &exception);
537         if (exception.delay && timeout)
538                 *timeout = exception.timeout;
539         if (exception.retry)
540                 return -EAGAIN;
541         return 0;
542 }
543
544 /*
545  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
546  * or 'false' otherwise.
547  */
548 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
549 {
550         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
551
552         if (flavor == RPC_AUTH_GSS_KRB5I ||
553             flavor == RPC_AUTH_GSS_KRB5P)
554                 return true;
555
556         return false;
557 }
558
559 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
560 {
561         spin_lock(&clp->cl_lock);
562         if (time_before(clp->cl_last_renewal,timestamp))
563                 clp->cl_last_renewal = timestamp;
564         spin_unlock(&clp->cl_lock);
565 }
566
567 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
568 {
569         struct nfs_client *clp = server->nfs_client;
570
571         if (!nfs4_has_session(clp))
572                 do_renew_lease(clp, timestamp);
573 }
574
575 struct nfs4_call_sync_data {
576         const struct nfs_server *seq_server;
577         struct nfs4_sequence_args *seq_args;
578         struct nfs4_sequence_res *seq_res;
579 };
580
581 void nfs4_init_sequence(struct nfs4_sequence_args *args,
582                         struct nfs4_sequence_res *res, int cache_reply)
583 {
584         args->sa_slot = NULL;
585         args->sa_cache_this = cache_reply;
586         args->sa_privileged = 0;
587
588         res->sr_slot = NULL;
589 }
590
591 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
592 {
593         args->sa_privileged = 1;
594 }
595
596 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
597                          struct nfs4_sequence_args *args,
598                          struct nfs4_sequence_res *res,
599                          struct rpc_task *task)
600 {
601         struct nfs4_slot *slot;
602
603         /* slot already allocated? */
604         if (res->sr_slot != NULL)
605                 goto out_start;
606
607         spin_lock(&tbl->slot_tbl_lock);
608         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
609                 goto out_sleep;
610
611         slot = nfs4_alloc_slot(tbl);
612         if (IS_ERR(slot)) {
613                 if (slot == ERR_PTR(-ENOMEM))
614                         task->tk_timeout = HZ >> 2;
615                 goto out_sleep;
616         }
617         spin_unlock(&tbl->slot_tbl_lock);
618
619         args->sa_slot = slot;
620         res->sr_slot = slot;
621
622 out_start:
623         rpc_call_start(task);
624         return 0;
625
626 out_sleep:
627         if (args->sa_privileged)
628                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
629                                 NULL, RPC_PRIORITY_PRIVILEGED);
630         else
631                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
632         spin_unlock(&tbl->slot_tbl_lock);
633         return -EAGAIN;
634 }
635 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
636
637 static int nfs40_sequence_done(struct rpc_task *task,
638                                struct nfs4_sequence_res *res)
639 {
640         struct nfs4_slot *slot = res->sr_slot;
641         struct nfs4_slot_table *tbl;
642
643         if (slot == NULL)
644                 goto out;
645
646         tbl = slot->table;
647         spin_lock(&tbl->slot_tbl_lock);
648         if (!nfs41_wake_and_assign_slot(tbl, slot))
649                 nfs4_free_slot(tbl, slot);
650         spin_unlock(&tbl->slot_tbl_lock);
651
652         res->sr_slot = NULL;
653 out:
654         return 1;
655 }
656
657 #if defined(CONFIG_NFS_V4_1)
658
659 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
660 {
661         struct nfs4_session *session;
662         struct nfs4_slot_table *tbl;
663         struct nfs4_slot *slot = res->sr_slot;
664         bool send_new_highest_used_slotid = false;
665
666         tbl = slot->table;
667         session = tbl->session;
668
669         spin_lock(&tbl->slot_tbl_lock);
670         /* Be nice to the server: try to ensure that the last transmitted
671          * value for highest_user_slotid <= target_highest_slotid
672          */
673         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
674                 send_new_highest_used_slotid = true;
675
676         if (nfs41_wake_and_assign_slot(tbl, slot)) {
677                 send_new_highest_used_slotid = false;
678                 goto out_unlock;
679         }
680         nfs4_free_slot(tbl, slot);
681
682         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
683                 send_new_highest_used_slotid = false;
684 out_unlock:
685         spin_unlock(&tbl->slot_tbl_lock);
686         res->sr_slot = NULL;
687         if (send_new_highest_used_slotid)
688                 nfs41_notify_server(session->clp);
689 }
690
691 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
692 {
693         struct nfs4_session *session;
694         struct nfs4_slot *slot = res->sr_slot;
695         struct nfs_client *clp;
696         bool interrupted = false;
697         int ret = 1;
698
699         if (slot == NULL)
700                 goto out_noaction;
701         /* don't increment the sequence number if the task wasn't sent */
702         if (!RPC_WAS_SENT(task))
703                 goto out;
704
705         session = slot->table->session;
706
707         if (slot->interrupted) {
708                 slot->interrupted = 0;
709                 interrupted = true;
710         }
711
712         trace_nfs4_sequence_done(session, res);
713         /* Check the SEQUENCE operation status */
714         switch (res->sr_status) {
715         case 0:
716                 /* Update the slot's sequence and clientid lease timer */
717                 ++slot->seq_nr;
718                 clp = session->clp;
719                 do_renew_lease(clp, res->sr_timestamp);
720                 /* Check sequence flags */
721                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
722                 nfs41_update_target_slotid(slot->table, slot, res);
723                 break;
724         case 1:
725                 /*
726                  * sr_status remains 1 if an RPC level error occurred.
727                  * The server may or may not have processed the sequence
728                  * operation..
729                  * Mark the slot as having hosted an interrupted RPC call.
730                  */
731                 slot->interrupted = 1;
732                 goto out;
733         case -NFS4ERR_DELAY:
734                 /* The server detected a resend of the RPC call and
735                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
736                  * of RFC5661.
737                  */
738                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
739                         __func__,
740                         slot->slot_nr,
741                         slot->seq_nr);
742                 goto out_retry;
743         case -NFS4ERR_BADSLOT:
744                 /*
745                  * The slot id we used was probably retired. Try again
746                  * using a different slot id.
747                  */
748                 goto retry_nowait;
749         case -NFS4ERR_SEQ_MISORDERED:
750                 /*
751                  * Was the last operation on this sequence interrupted?
752                  * If so, retry after bumping the sequence number.
753                  */
754                 if (interrupted) {
755                         ++slot->seq_nr;
756                         goto retry_nowait;
757                 }
758                 /*
759                  * Could this slot have been previously retired?
760                  * If so, then the server may be expecting seq_nr = 1!
761                  */
762                 if (slot->seq_nr != 1) {
763                         slot->seq_nr = 1;
764                         goto retry_nowait;
765                 }
766                 break;
767         case -NFS4ERR_SEQ_FALSE_RETRY:
768                 ++slot->seq_nr;
769                 goto retry_nowait;
770         default:
771                 /* Just update the slot sequence no. */
772                 ++slot->seq_nr;
773         }
774 out:
775         /* The session may be reset by one of the error handlers. */
776         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
777         nfs41_sequence_free_slot(res);
778 out_noaction:
779         return ret;
780 retry_nowait:
781         if (rpc_restart_call_prepare(task)) {
782                 task->tk_status = 0;
783                 ret = 0;
784         }
785         goto out;
786 out_retry:
787         if (!rpc_restart_call(task))
788                 goto out;
789         rpc_delay(task, NFS4_POLL_RETRY_MAX);
790         return 0;
791 }
792 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
793
794 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
795 {
796         if (res->sr_slot == NULL)
797                 return 1;
798         if (!res->sr_slot->table->session)
799                 return nfs40_sequence_done(task, res);
800         return nfs41_sequence_done(task, res);
801 }
802 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
803
804 int nfs41_setup_sequence(struct nfs4_session *session,
805                                 struct nfs4_sequence_args *args,
806                                 struct nfs4_sequence_res *res,
807                                 struct rpc_task *task)
808 {
809         struct nfs4_slot *slot;
810         struct nfs4_slot_table *tbl;
811
812         dprintk("--> %s\n", __func__);
813         /* slot already allocated? */
814         if (res->sr_slot != NULL)
815                 goto out_success;
816
817         tbl = &session->fc_slot_table;
818
819         task->tk_timeout = 0;
820
821         spin_lock(&tbl->slot_tbl_lock);
822         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
823             !args->sa_privileged) {
824                 /* The state manager will wait until the slot table is empty */
825                 dprintk("%s session is draining\n", __func__);
826                 goto out_sleep;
827         }
828
829         slot = nfs4_alloc_slot(tbl);
830         if (IS_ERR(slot)) {
831                 /* If out of memory, try again in 1/4 second */
832                 if (slot == ERR_PTR(-ENOMEM))
833                         task->tk_timeout = HZ >> 2;
834                 dprintk("<-- %s: no free slots\n", __func__);
835                 goto out_sleep;
836         }
837         spin_unlock(&tbl->slot_tbl_lock);
838
839         args->sa_slot = slot;
840
841         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
842                         slot->slot_nr, slot->seq_nr);
843
844         res->sr_slot = slot;
845         res->sr_timestamp = jiffies;
846         res->sr_status_flags = 0;
847         /*
848          * sr_status is only set in decode_sequence, and so will remain
849          * set to 1 if an rpc level failure occurs.
850          */
851         res->sr_status = 1;
852         trace_nfs4_setup_sequence(session, args);
853 out_success:
854         rpc_call_start(task);
855         return 0;
856 out_sleep:
857         /* Privileged tasks are queued with top priority */
858         if (args->sa_privileged)
859                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
860                                 NULL, RPC_PRIORITY_PRIVILEGED);
861         else
862                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
863         spin_unlock(&tbl->slot_tbl_lock);
864         return -EAGAIN;
865 }
866 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
867
868 static int nfs4_setup_sequence(const struct nfs_server *server,
869                                struct nfs4_sequence_args *args,
870                                struct nfs4_sequence_res *res,
871                                struct rpc_task *task)
872 {
873         struct nfs4_session *session = nfs4_get_session(server);
874         int ret = 0;
875
876         if (!session)
877                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
878                                             args, res, task);
879
880         dprintk("--> %s clp %p session %p sr_slot %u\n",
881                 __func__, session->clp, session, res->sr_slot ?
882                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
883
884         ret = nfs41_setup_sequence(session, args, res, task);
885
886         dprintk("<-- %s status=%d\n", __func__, ret);
887         return ret;
888 }
889
890 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
891 {
892         struct nfs4_call_sync_data *data = calldata;
893         struct nfs4_session *session = nfs4_get_session(data->seq_server);
894
895         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
896
897         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
898 }
899
900 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
901 {
902         struct nfs4_call_sync_data *data = calldata;
903
904         nfs41_sequence_done(task, data->seq_res);
905 }
906
907 static const struct rpc_call_ops nfs41_call_sync_ops = {
908         .rpc_call_prepare = nfs41_call_sync_prepare,
909         .rpc_call_done = nfs41_call_sync_done,
910 };
911
912 #else   /* !CONFIG_NFS_V4_1 */
913
914 static int nfs4_setup_sequence(const struct nfs_server *server,
915                                struct nfs4_sequence_args *args,
916                                struct nfs4_sequence_res *res,
917                                struct rpc_task *task)
918 {
919         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
920                                     args, res, task);
921 }
922
923 int nfs4_sequence_done(struct rpc_task *task,
924                        struct nfs4_sequence_res *res)
925 {
926         return nfs40_sequence_done(task, res);
927 }
928 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
929
930 #endif  /* !CONFIG_NFS_V4_1 */
931
932 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
933 {
934         struct nfs4_call_sync_data *data = calldata;
935         nfs4_setup_sequence(data->seq_server,
936                                 data->seq_args, data->seq_res, task);
937 }
938
939 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
940 {
941         struct nfs4_call_sync_data *data = calldata;
942         nfs4_sequence_done(task, data->seq_res);
943 }
944
945 static const struct rpc_call_ops nfs40_call_sync_ops = {
946         .rpc_call_prepare = nfs40_call_sync_prepare,
947         .rpc_call_done = nfs40_call_sync_done,
948 };
949
950 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
951                                    struct nfs_server *server,
952                                    struct rpc_message *msg,
953                                    struct nfs4_sequence_args *args,
954                                    struct nfs4_sequence_res *res)
955 {
956         int ret;
957         struct rpc_task *task;
958         struct nfs_client *clp = server->nfs_client;
959         struct nfs4_call_sync_data data = {
960                 .seq_server = server,
961                 .seq_args = args,
962                 .seq_res = res,
963         };
964         struct rpc_task_setup task_setup = {
965                 .rpc_client = clnt,
966                 .rpc_message = msg,
967                 .callback_ops = clp->cl_mvops->call_sync_ops,
968                 .callback_data = &data
969         };
970
971         task = rpc_run_task(&task_setup);
972         if (IS_ERR(task))
973                 ret = PTR_ERR(task);
974         else {
975                 ret = task->tk_status;
976                 rpc_put_task(task);
977         }
978         return ret;
979 }
980
981 int nfs4_call_sync(struct rpc_clnt *clnt,
982                    struct nfs_server *server,
983                    struct rpc_message *msg,
984                    struct nfs4_sequence_args *args,
985                    struct nfs4_sequence_res *res,
986                    int cache_reply)
987 {
988         nfs4_init_sequence(args, res, cache_reply);
989         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
990 }
991
992 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
993 {
994         struct nfs_inode *nfsi = NFS_I(dir);
995
996         spin_lock(&dir->i_lock);
997         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
998         if (!cinfo->atomic || cinfo->before != dir->i_version)
999                 nfs_force_lookup_revalidate(dir);
1000         dir->i_version = cinfo->after;
1001         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1002         nfs_fscache_invalidate(dir);
1003         spin_unlock(&dir->i_lock);
1004 }
1005
1006 struct nfs4_opendata {
1007         struct kref kref;
1008         struct nfs_openargs o_arg;
1009         struct nfs_openres o_res;
1010         struct nfs_open_confirmargs c_arg;
1011         struct nfs_open_confirmres c_res;
1012         struct nfs4_string owner_name;
1013         struct nfs4_string group_name;
1014         struct nfs4_label *a_label;
1015         struct nfs_fattr f_attr;
1016         struct nfs4_label *f_label;
1017         struct dentry *dir;
1018         struct dentry *dentry;
1019         struct nfs4_state_owner *owner;
1020         struct nfs4_state *state;
1021         struct iattr attrs;
1022         unsigned long timestamp;
1023         unsigned int rpc_done : 1;
1024         unsigned int file_created : 1;
1025         unsigned int is_recover : 1;
1026         int rpc_status;
1027         int cancelled;
1028 };
1029
1030 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1031                 int err, struct nfs4_exception *exception)
1032 {
1033         if (err != -EINVAL)
1034                 return false;
1035         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1036                 return false;
1037         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1038         exception->retry = 1;
1039         return true;
1040 }
1041
1042 static u32
1043 nfs4_map_atomic_open_share(struct nfs_server *server,
1044                 fmode_t fmode, int openflags)
1045 {
1046         u32 res = 0;
1047
1048         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1049         case FMODE_READ:
1050                 res = NFS4_SHARE_ACCESS_READ;
1051                 break;
1052         case FMODE_WRITE:
1053                 res = NFS4_SHARE_ACCESS_WRITE;
1054                 break;
1055         case FMODE_READ|FMODE_WRITE:
1056                 res = NFS4_SHARE_ACCESS_BOTH;
1057         }
1058         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1059                 goto out;
1060         /* Want no delegation if we're using O_DIRECT */
1061         if (openflags & O_DIRECT)
1062                 res |= NFS4_SHARE_WANT_NO_DELEG;
1063 out:
1064         return res;
1065 }
1066
1067 static enum open_claim_type4
1068 nfs4_map_atomic_open_claim(struct nfs_server *server,
1069                 enum open_claim_type4 claim)
1070 {
1071         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1072                 return claim;
1073         switch (claim) {
1074         default:
1075                 return claim;
1076         case NFS4_OPEN_CLAIM_FH:
1077                 return NFS4_OPEN_CLAIM_NULL;
1078         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1079                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1080         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1081                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1082         }
1083 }
1084
1085 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1086 {
1087         p->o_res.f_attr = &p->f_attr;
1088         p->o_res.f_label = p->f_label;
1089         p->o_res.seqid = p->o_arg.seqid;
1090         p->c_res.seqid = p->c_arg.seqid;
1091         p->o_res.server = p->o_arg.server;
1092         p->o_res.access_request = p->o_arg.access;
1093         nfs_fattr_init(&p->f_attr);
1094         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1095 }
1096
1097 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1098                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1099                 const struct iattr *attrs,
1100                 struct nfs4_label *label,
1101                 enum open_claim_type4 claim,
1102                 gfp_t gfp_mask)
1103 {
1104         struct dentry *parent = dget_parent(dentry);
1105         struct inode *dir = d_inode(parent);
1106         struct nfs_server *server = NFS_SERVER(dir);
1107         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1108         struct nfs4_opendata *p;
1109
1110         p = kzalloc(sizeof(*p), gfp_mask);
1111         if (p == NULL)
1112                 goto err;
1113
1114         p->f_label = nfs4_label_alloc(server, gfp_mask);
1115         if (IS_ERR(p->f_label))
1116                 goto err_free_p;
1117
1118         p->a_label = nfs4_label_alloc(server, gfp_mask);
1119         if (IS_ERR(p->a_label))
1120                 goto err_free_f;
1121
1122         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1123         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1124         if (IS_ERR(p->o_arg.seqid))
1125                 goto err_free_label;
1126         nfs_sb_active(dentry->d_sb);
1127         p->dentry = dget(dentry);
1128         p->dir = parent;
1129         p->owner = sp;
1130         atomic_inc(&sp->so_count);
1131         p->o_arg.open_flags = flags;
1132         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1133         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1134                         fmode, flags);
1135         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1136          * will return permission denied for all bits until close */
1137         if (!(flags & O_EXCL)) {
1138                 /* ask server to check for all possible rights as results
1139                  * are cached */
1140                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1141                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1142         }
1143         p->o_arg.clientid = server->nfs_client->cl_clientid;
1144         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1145         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1146         p->o_arg.name = &dentry->d_name;
1147         p->o_arg.server = server;
1148         p->o_arg.bitmask = nfs4_bitmask(server, label);
1149         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1150         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1151         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1152         switch (p->o_arg.claim) {
1153         case NFS4_OPEN_CLAIM_NULL:
1154         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1155         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1156                 p->o_arg.fh = NFS_FH(dir);
1157                 break;
1158         case NFS4_OPEN_CLAIM_PREVIOUS:
1159         case NFS4_OPEN_CLAIM_FH:
1160         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1161         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1162                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1163         }
1164         if (attrs != NULL && attrs->ia_valid != 0) {
1165                 __u32 verf[2];
1166
1167                 p->o_arg.u.attrs = &p->attrs;
1168                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1169
1170                 verf[0] = jiffies;
1171                 verf[1] = current->pid;
1172                 memcpy(p->o_arg.u.verifier.data, verf,
1173                                 sizeof(p->o_arg.u.verifier.data));
1174         }
1175         p->c_arg.fh = &p->o_res.fh;
1176         p->c_arg.stateid = &p->o_res.stateid;
1177         p->c_arg.seqid = p->o_arg.seqid;
1178         nfs4_init_opendata_res(p);
1179         kref_init(&p->kref);
1180         return p;
1181
1182 err_free_label:
1183         nfs4_label_free(p->a_label);
1184 err_free_f:
1185         nfs4_label_free(p->f_label);
1186 err_free_p:
1187         kfree(p);
1188 err:
1189         dput(parent);
1190         return NULL;
1191 }
1192
1193 static void nfs4_opendata_free(struct kref *kref)
1194 {
1195         struct nfs4_opendata *p = container_of(kref,
1196                         struct nfs4_opendata, kref);
1197         struct super_block *sb = p->dentry->d_sb;
1198
1199         nfs_free_seqid(p->o_arg.seqid);
1200         if (p->state != NULL)
1201                 nfs4_put_open_state(p->state);
1202         nfs4_put_state_owner(p->owner);
1203
1204         nfs4_label_free(p->a_label);
1205         nfs4_label_free(p->f_label);
1206
1207         dput(p->dir);
1208         dput(p->dentry);
1209         nfs_sb_deactive(sb);
1210         nfs_fattr_free_names(&p->f_attr);
1211         kfree(p->f_attr.mdsthreshold);
1212         kfree(p);
1213 }
1214
1215 static void nfs4_opendata_put(struct nfs4_opendata *p)
1216 {
1217         if (p != NULL)
1218                 kref_put(&p->kref, nfs4_opendata_free);
1219 }
1220
1221 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1222 {
1223         int ret;
1224
1225         ret = rpc_wait_for_completion_task(task);
1226         return ret;
1227 }
1228
1229 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1230                 fmode_t fmode)
1231 {
1232         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1233         case FMODE_READ|FMODE_WRITE:
1234                 return state->n_rdwr != 0;
1235         case FMODE_WRITE:
1236                 return state->n_wronly != 0;
1237         case FMODE_READ:
1238                 return state->n_rdonly != 0;
1239         }
1240         WARN_ON_ONCE(1);
1241         return false;
1242 }
1243
1244 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1245 {
1246         int ret = 0;
1247
1248         if (open_mode & (O_EXCL|O_TRUNC))
1249                 goto out;
1250         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1251                 case FMODE_READ:
1252                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1253                                 && state->n_rdonly != 0;
1254                         break;
1255                 case FMODE_WRITE:
1256                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1257                                 && state->n_wronly != 0;
1258                         break;
1259                 case FMODE_READ|FMODE_WRITE:
1260                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1261                                 && state->n_rdwr != 0;
1262         }
1263 out:
1264         return ret;
1265 }
1266
1267 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1268                 enum open_claim_type4 claim)
1269 {
1270         if (delegation == NULL)
1271                 return 0;
1272         if ((delegation->type & fmode) != fmode)
1273                 return 0;
1274         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1275                 return 0;
1276         switch (claim) {
1277         case NFS4_OPEN_CLAIM_NULL:
1278         case NFS4_OPEN_CLAIM_FH:
1279                 break;
1280         case NFS4_OPEN_CLAIM_PREVIOUS:
1281                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1282                         break;
1283         default:
1284                 return 0;
1285         }
1286         nfs_mark_delegation_referenced(delegation);
1287         return 1;
1288 }
1289
1290 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1291 {
1292         switch (fmode) {
1293                 case FMODE_WRITE:
1294                         state->n_wronly++;
1295                         break;
1296                 case FMODE_READ:
1297                         state->n_rdonly++;
1298                         break;
1299                 case FMODE_READ|FMODE_WRITE:
1300                         state->n_rdwr++;
1301         }
1302         nfs4_state_set_mode_locked(state, state->state | fmode);
1303 }
1304
1305 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1306 {
1307         struct nfs_client *clp = state->owner->so_server->nfs_client;
1308         bool need_recover = false;
1309
1310         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1311                 need_recover = true;
1312         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1313                 need_recover = true;
1314         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1315                 need_recover = true;
1316         if (need_recover)
1317                 nfs4_state_mark_reclaim_nograce(clp, state);
1318 }
1319
1320 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1321                 nfs4_stateid *stateid)
1322 {
1323         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1324                 return true;
1325         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1326                 nfs_test_and_clear_all_open_stateid(state);
1327                 return true;
1328         }
1329         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1330                 return true;
1331         return false;
1332 }
1333
1334 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1335 {
1336         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1337                 return;
1338         if (state->n_wronly)
1339                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1340         if (state->n_rdonly)
1341                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1342         if (state->n_rdwr)
1343                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1344         set_bit(NFS_OPEN_STATE, &state->flags);
1345 }
1346
1347 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1348                 nfs4_stateid *arg_stateid,
1349                 nfs4_stateid *stateid, fmode_t fmode)
1350 {
1351         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1352         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1353         case FMODE_WRITE:
1354                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1355                 break;
1356         case FMODE_READ:
1357                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1358                 break;
1359         case 0:
1360                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1361                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1362                 clear_bit(NFS_OPEN_STATE, &state->flags);
1363         }
1364         if (stateid == NULL)
1365                 return;
1366         /* Handle races with OPEN */
1367         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1368             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1369             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1370                 nfs_resync_open_stateid_locked(state);
1371                 return;
1372         }
1373         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1374                 nfs4_stateid_copy(&state->stateid, stateid);
1375         nfs4_stateid_copy(&state->open_stateid, stateid);
1376 }
1377
1378 static void nfs_clear_open_stateid(struct nfs4_state *state,
1379         nfs4_stateid *arg_stateid,
1380         nfs4_stateid *stateid, fmode_t fmode)
1381 {
1382         write_seqlock(&state->seqlock);
1383         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1384         write_sequnlock(&state->seqlock);
1385         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1386                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1387 }
1388
1389 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1390 {
1391         switch (fmode) {
1392                 case FMODE_READ:
1393                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1394                         break;
1395                 case FMODE_WRITE:
1396                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1397                         break;
1398                 case FMODE_READ|FMODE_WRITE:
1399                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1400         }
1401         if (!nfs_need_update_open_stateid(state, stateid))
1402                 return;
1403         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1404                 nfs4_stateid_copy(&state->stateid, stateid);
1405         nfs4_stateid_copy(&state->open_stateid, stateid);
1406 }
1407
1408 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1409 {
1410         /*
1411          * Protect the call to nfs4_state_set_mode_locked and
1412          * serialise the stateid update
1413          */
1414         spin_lock(&state->owner->so_lock);
1415         write_seqlock(&state->seqlock);
1416         if (deleg_stateid != NULL) {
1417                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1418                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1419         }
1420         if (open_stateid != NULL)
1421                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1422         write_sequnlock(&state->seqlock);
1423         update_open_stateflags(state, fmode);
1424         spin_unlock(&state->owner->so_lock);
1425 }
1426
1427 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1428 {
1429         struct nfs_inode *nfsi = NFS_I(state->inode);
1430         struct nfs_delegation *deleg_cur;
1431         int ret = 0;
1432
1433         fmode &= (FMODE_READ|FMODE_WRITE);
1434
1435         rcu_read_lock();
1436         deleg_cur = rcu_dereference(nfsi->delegation);
1437         if (deleg_cur == NULL)
1438                 goto no_delegation;
1439
1440         spin_lock(&deleg_cur->lock);
1441         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1442            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1443             (deleg_cur->type & fmode) != fmode)
1444                 goto no_delegation_unlock;
1445
1446         if (delegation == NULL)
1447                 delegation = &deleg_cur->stateid;
1448         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1449                 goto no_delegation_unlock;
1450
1451         nfs_mark_delegation_referenced(deleg_cur);
1452         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1453         ret = 1;
1454 no_delegation_unlock:
1455         spin_unlock(&deleg_cur->lock);
1456 no_delegation:
1457         rcu_read_unlock();
1458
1459         if (!ret && open_stateid != NULL) {
1460                 __update_open_stateid(state, open_stateid, NULL, fmode);
1461                 ret = 1;
1462         }
1463         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1464                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1465
1466         return ret;
1467 }
1468
1469 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1470                 const nfs4_stateid *stateid)
1471 {
1472         struct nfs4_state *state = lsp->ls_state;
1473         bool ret = false;
1474
1475         spin_lock(&state->state_lock);
1476         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1477                 goto out_noupdate;
1478         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1479                 goto out_noupdate;
1480         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1481         ret = true;
1482 out_noupdate:
1483         spin_unlock(&state->state_lock);
1484         return ret;
1485 }
1486
1487 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1488 {
1489         struct nfs_delegation *delegation;
1490
1491         rcu_read_lock();
1492         delegation = rcu_dereference(NFS_I(inode)->delegation);
1493         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1494                 rcu_read_unlock();
1495                 return;
1496         }
1497         rcu_read_unlock();
1498         nfs4_inode_return_delegation(inode);
1499 }
1500
1501 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1502 {
1503         struct nfs4_state *state = opendata->state;
1504         struct nfs_inode *nfsi = NFS_I(state->inode);
1505         struct nfs_delegation *delegation;
1506         int open_mode = opendata->o_arg.open_flags;
1507         fmode_t fmode = opendata->o_arg.fmode;
1508         enum open_claim_type4 claim = opendata->o_arg.claim;
1509         nfs4_stateid stateid;
1510         int ret = -EAGAIN;
1511
1512         for (;;) {
1513                 spin_lock(&state->owner->so_lock);
1514                 if (can_open_cached(state, fmode, open_mode)) {
1515                         update_open_stateflags(state, fmode);
1516                         spin_unlock(&state->owner->so_lock);
1517                         goto out_return_state;
1518                 }
1519                 spin_unlock(&state->owner->so_lock);
1520                 rcu_read_lock();
1521                 delegation = rcu_dereference(nfsi->delegation);
1522                 if (!can_open_delegated(delegation, fmode, claim)) {
1523                         rcu_read_unlock();
1524                         break;
1525                 }
1526                 /* Save the delegation */
1527                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1528                 rcu_read_unlock();
1529                 nfs_release_seqid(opendata->o_arg.seqid);
1530                 if (!opendata->is_recover) {
1531                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1532                         if (ret != 0)
1533                                 goto out;
1534                 }
1535                 ret = -EAGAIN;
1536
1537                 /* Try to update the stateid using the delegation */
1538                 if (update_open_stateid(state, NULL, &stateid, fmode))
1539                         goto out_return_state;
1540         }
1541 out:
1542         return ERR_PTR(ret);
1543 out_return_state:
1544         atomic_inc(&state->count);
1545         return state;
1546 }
1547
1548 static void
1549 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1550 {
1551         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1552         struct nfs_delegation *delegation;
1553         int delegation_flags = 0;
1554
1555         rcu_read_lock();
1556         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1557         if (delegation)
1558                 delegation_flags = delegation->flags;
1559         rcu_read_unlock();
1560         switch (data->o_arg.claim) {
1561         default:
1562                 break;
1563         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1564         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1565                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1566                                    "returning a delegation for "
1567                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1568                                    clp->cl_hostname);
1569                 return;
1570         }
1571         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1572                 nfs_inode_set_delegation(state->inode,
1573                                          data->owner->so_cred,
1574                                          &data->o_res);
1575         else
1576                 nfs_inode_reclaim_delegation(state->inode,
1577                                              data->owner->so_cred,
1578                                              &data->o_res);
1579 }
1580
1581 /*
1582  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1583  * and update the nfs4_state.
1584  */
1585 static struct nfs4_state *
1586 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1587 {
1588         struct inode *inode = data->state->inode;
1589         struct nfs4_state *state = data->state;
1590         int ret;
1591
1592         if (!data->rpc_done) {
1593                 if (data->rpc_status) {
1594                         ret = data->rpc_status;
1595                         goto err;
1596                 }
1597                 /* cached opens have already been processed */
1598                 goto update;
1599         }
1600
1601         ret = nfs_refresh_inode(inode, &data->f_attr);
1602         if (ret)
1603                 goto err;
1604
1605         if (data->o_res.delegation_type != 0)
1606                 nfs4_opendata_check_deleg(data, state);
1607 update:
1608         update_open_stateid(state, &data->o_res.stateid, NULL,
1609                             data->o_arg.fmode);
1610         atomic_inc(&state->count);
1611
1612         return state;
1613 err:
1614         return ERR_PTR(ret);
1615
1616 }
1617
1618 static struct nfs4_state *
1619 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1620 {
1621         struct inode *inode;
1622         struct nfs4_state *state = NULL;
1623         int ret;
1624
1625         if (!data->rpc_done) {
1626                 state = nfs4_try_open_cached(data);
1627                 trace_nfs4_cached_open(data->state);
1628                 goto out;
1629         }
1630
1631         ret = -EAGAIN;
1632         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1633                 goto err;
1634         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1635         ret = PTR_ERR(inode);
1636         if (IS_ERR(inode))
1637                 goto err;
1638         ret = -ENOMEM;
1639         state = nfs4_get_open_state(inode, data->owner);
1640         if (state == NULL)
1641                 goto err_put_inode;
1642         if (data->o_res.delegation_type != 0)
1643                 nfs4_opendata_check_deleg(data, state);
1644         update_open_stateid(state, &data->o_res.stateid, NULL,
1645                         data->o_arg.fmode);
1646         iput(inode);
1647 out:
1648         nfs_release_seqid(data->o_arg.seqid);
1649         return state;
1650 err_put_inode:
1651         iput(inode);
1652 err:
1653         return ERR_PTR(ret);
1654 }
1655
1656 static struct nfs4_state *
1657 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1658 {
1659         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1660                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1661         return _nfs4_opendata_to_nfs4_state(data);
1662 }
1663
1664 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1665 {
1666         struct nfs_inode *nfsi = NFS_I(state->inode);
1667         struct nfs_open_context *ctx;
1668
1669         spin_lock(&state->inode->i_lock);
1670         list_for_each_entry(ctx, &nfsi->open_files, list) {
1671                 if (ctx->state != state)
1672                         continue;
1673                 get_nfs_open_context(ctx);
1674                 spin_unlock(&state->inode->i_lock);
1675                 return ctx;
1676         }
1677         spin_unlock(&state->inode->i_lock);
1678         return ERR_PTR(-ENOENT);
1679 }
1680
1681 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1682                 struct nfs4_state *state, enum open_claim_type4 claim)
1683 {
1684         struct nfs4_opendata *opendata;
1685
1686         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1687                         NULL, NULL, claim, GFP_NOFS);
1688         if (opendata == NULL)
1689                 return ERR_PTR(-ENOMEM);
1690         opendata->state = state;
1691         atomic_inc(&state->count);
1692         return opendata;
1693 }
1694
1695 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1696                 fmode_t fmode)
1697 {
1698         struct nfs4_state *newstate;
1699         int ret;
1700
1701         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1702                 return 0;
1703         opendata->o_arg.open_flags = 0;
1704         opendata->o_arg.fmode = fmode;
1705         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1706                         NFS_SB(opendata->dentry->d_sb),
1707                         fmode, 0);
1708         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1709         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1710         nfs4_init_opendata_res(opendata);
1711         ret = _nfs4_recover_proc_open(opendata);
1712         if (ret != 0)
1713                 return ret; 
1714         newstate = nfs4_opendata_to_nfs4_state(opendata);
1715         if (IS_ERR(newstate))
1716                 return PTR_ERR(newstate);
1717         if (newstate != opendata->state)
1718                 ret = -ESTALE;
1719         nfs4_close_state(newstate, fmode);
1720         return ret;
1721 }
1722
1723 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1724 {
1725         int ret;
1726
1727         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1728         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1729         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1730         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1731         /* memory barrier prior to reading state->n_* */
1732         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1733         clear_bit(NFS_OPEN_STATE, &state->flags);
1734         smp_rmb();
1735         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1736         if (ret != 0)
1737                 return ret;
1738         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1739         if (ret != 0)
1740                 return ret;
1741         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1742         if (ret != 0)
1743                 return ret;
1744         /*
1745          * We may have performed cached opens for all three recoveries.
1746          * Check if we need to update the current stateid.
1747          */
1748         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1749             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1750                 write_seqlock(&state->seqlock);
1751                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1752                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1753                 write_sequnlock(&state->seqlock);
1754         }
1755         return 0;
1756 }
1757
1758 /*
1759  * OPEN_RECLAIM:
1760  *      reclaim state on the server after a reboot.
1761  */
1762 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1763 {
1764         struct nfs_delegation *delegation;
1765         struct nfs4_opendata *opendata;
1766         fmode_t delegation_type = 0;
1767         int status;
1768
1769         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1770                         NFS4_OPEN_CLAIM_PREVIOUS);
1771         if (IS_ERR(opendata))
1772                 return PTR_ERR(opendata);
1773         rcu_read_lock();
1774         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1775         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1776                 delegation_type = delegation->type;
1777         rcu_read_unlock();
1778         opendata->o_arg.u.delegation_type = delegation_type;
1779         status = nfs4_open_recover(opendata, state);
1780         nfs4_opendata_put(opendata);
1781         return status;
1782 }
1783
1784 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1785 {
1786         struct nfs_server *server = NFS_SERVER(state->inode);
1787         struct nfs4_exception exception = { };
1788         int err;
1789         do {
1790                 err = _nfs4_do_open_reclaim(ctx, state);
1791                 trace_nfs4_open_reclaim(ctx, 0, err);
1792                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1793                         continue;
1794                 if (err != -NFS4ERR_DELAY)
1795                         break;
1796                 nfs4_handle_exception(server, err, &exception);
1797         } while (exception.retry);
1798         return err;
1799 }
1800
1801 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1802 {
1803         struct nfs_open_context *ctx;
1804         int ret;
1805
1806         ctx = nfs4_state_find_open_context(state);
1807         if (IS_ERR(ctx))
1808                 return -EAGAIN;
1809         ret = nfs4_do_open_reclaim(ctx, state);
1810         put_nfs_open_context(ctx);
1811         return ret;
1812 }
1813
1814 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1815 {
1816         switch (err) {
1817                 default:
1818                         printk(KERN_ERR "NFS: %s: unhandled error "
1819                                         "%d.\n", __func__, err);
1820                 case 0:
1821                 case -ENOENT:
1822                 case -EAGAIN:
1823                 case -ESTALE:
1824                         break;
1825                 case -NFS4ERR_BADSESSION:
1826                 case -NFS4ERR_BADSLOT:
1827                 case -NFS4ERR_BAD_HIGH_SLOT:
1828                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1829                 case -NFS4ERR_DEADSESSION:
1830                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1831                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1832                         return -EAGAIN;
1833                 case -NFS4ERR_STALE_CLIENTID:
1834                 case -NFS4ERR_STALE_STATEID:
1835                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1836                 case -NFS4ERR_EXPIRED:
1837                         /* Don't recall a delegation if it was lost */
1838                         nfs4_schedule_lease_recovery(server->nfs_client);
1839                         return -EAGAIN;
1840                 case -NFS4ERR_MOVED:
1841                         nfs4_schedule_migration_recovery(server);
1842                         return -EAGAIN;
1843                 case -NFS4ERR_LEASE_MOVED:
1844                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1845                         return -EAGAIN;
1846                 case -NFS4ERR_DELEG_REVOKED:
1847                 case -NFS4ERR_ADMIN_REVOKED:
1848                 case -NFS4ERR_BAD_STATEID:
1849                 case -NFS4ERR_OPENMODE:
1850                         nfs_inode_find_state_and_recover(state->inode,
1851                                         stateid);
1852                         nfs4_schedule_stateid_recovery(server, state);
1853                         return -EAGAIN;
1854                 case -NFS4ERR_DELAY:
1855                 case -NFS4ERR_GRACE:
1856                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1857                         ssleep(1);
1858                         return -EAGAIN;
1859                 case -ENOMEM:
1860                 case -NFS4ERR_DENIED:
1861                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1862                         return 0;
1863         }
1864         return err;
1865 }
1866
1867 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1868                 struct nfs4_state *state, const nfs4_stateid *stateid,
1869                 fmode_t type)
1870 {
1871         struct nfs_server *server = NFS_SERVER(state->inode);
1872         struct nfs4_opendata *opendata;
1873         int err = 0;
1874
1875         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1876                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1877         if (IS_ERR(opendata))
1878                 return PTR_ERR(opendata);
1879         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1880         write_seqlock(&state->seqlock);
1881         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1882         write_sequnlock(&state->seqlock);
1883         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1884         switch (type & (FMODE_READ|FMODE_WRITE)) {
1885         case FMODE_READ|FMODE_WRITE:
1886         case FMODE_WRITE:
1887                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1888                 if (err)
1889                         break;
1890                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1891                 if (err)
1892                         break;
1893         case FMODE_READ:
1894                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1895         }
1896         nfs4_opendata_put(opendata);
1897         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1898 }
1899
1900 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1901 {
1902         struct nfs4_opendata *data = calldata;
1903
1904         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1905                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1906 }
1907
1908 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1909 {
1910         struct nfs4_opendata *data = calldata;
1911
1912         nfs40_sequence_done(task, &data->c_res.seq_res);
1913
1914         data->rpc_status = task->tk_status;
1915         if (data->rpc_status == 0) {
1916                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1917                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1918                 renew_lease(data->o_res.server, data->timestamp);
1919                 data->rpc_done = 1;
1920         }
1921 }
1922
1923 static void nfs4_open_confirm_release(void *calldata)
1924 {
1925         struct nfs4_opendata *data = calldata;
1926         struct nfs4_state *state = NULL;
1927
1928         /* If this request hasn't been cancelled, do nothing */
1929         if (data->cancelled == 0)
1930                 goto out_free;
1931         /* In case of error, no cleanup! */
1932         if (!data->rpc_done)
1933                 goto out_free;
1934         state = nfs4_opendata_to_nfs4_state(data);
1935         if (!IS_ERR(state))
1936                 nfs4_close_state(state, data->o_arg.fmode);
1937 out_free:
1938         nfs4_opendata_put(data);
1939 }
1940
1941 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1942         .rpc_call_prepare = nfs4_open_confirm_prepare,
1943         .rpc_call_done = nfs4_open_confirm_done,
1944         .rpc_release = nfs4_open_confirm_release,
1945 };
1946
1947 /*
1948  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1949  */
1950 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1951 {
1952         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1953         struct rpc_task *task;
1954         struct  rpc_message msg = {
1955                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1956                 .rpc_argp = &data->c_arg,
1957                 .rpc_resp = &data->c_res,
1958                 .rpc_cred = data->owner->so_cred,
1959         };
1960         struct rpc_task_setup task_setup_data = {
1961                 .rpc_client = server->client,
1962                 .rpc_message = &msg,
1963                 .callback_ops = &nfs4_open_confirm_ops,
1964                 .callback_data = data,
1965                 .workqueue = nfsiod_workqueue,
1966                 .flags = RPC_TASK_ASYNC,
1967         };
1968         int status;
1969
1970         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1971         kref_get(&data->kref);
1972         data->rpc_done = 0;
1973         data->rpc_status = 0;
1974         data->timestamp = jiffies;
1975         if (data->is_recover)
1976                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1977         task = rpc_run_task(&task_setup_data);
1978         if (IS_ERR(task))
1979                 return PTR_ERR(task);
1980         status = nfs4_wait_for_completion_rpc_task(task);
1981         if (status != 0) {
1982                 data->cancelled = 1;
1983                 smp_wmb();
1984         } else
1985                 status = data->rpc_status;
1986         rpc_put_task(task);
1987         return status;
1988 }
1989
1990 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1991 {
1992         struct nfs4_opendata *data = calldata;
1993         struct nfs4_state_owner *sp = data->owner;
1994         struct nfs_client *clp = sp->so_server->nfs_client;
1995         enum open_claim_type4 claim = data->o_arg.claim;
1996
1997         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1998                 goto out_wait;
1999         /*
2000          * Check if we still need to send an OPEN call, or if we can use
2001          * a delegation instead.
2002          */
2003         if (data->state != NULL) {
2004                 struct nfs_delegation *delegation;
2005
2006                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2007                         goto out_no_action;
2008                 rcu_read_lock();
2009                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2010                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2011                         goto unlock_no_action;
2012                 rcu_read_unlock();
2013         }
2014         /* Update client id. */
2015         data->o_arg.clientid = clp->cl_clientid;
2016         switch (claim) {
2017         default:
2018                 break;
2019         case NFS4_OPEN_CLAIM_PREVIOUS:
2020         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2021         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2022                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2023         case NFS4_OPEN_CLAIM_FH:
2024                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2025                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2026         }
2027         data->timestamp = jiffies;
2028         if (nfs4_setup_sequence(data->o_arg.server,
2029                                 &data->o_arg.seq_args,
2030                                 &data->o_res.seq_res,
2031                                 task) != 0)
2032                 nfs_release_seqid(data->o_arg.seqid);
2033
2034         /* Set the create mode (note dependency on the session type) */
2035         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2036         if (data->o_arg.open_flags & O_EXCL) {
2037                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2038                 if (nfs4_has_persistent_session(clp))
2039                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2040                 else if (clp->cl_mvops->minor_version > 0)
2041                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2042         }
2043         return;
2044 unlock_no_action:
2045         trace_nfs4_cached_open(data->state);
2046         rcu_read_unlock();
2047 out_no_action:
2048         task->tk_action = NULL;
2049 out_wait:
2050         nfs4_sequence_done(task, &data->o_res.seq_res);
2051 }
2052
2053 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2054 {
2055         struct nfs4_opendata *data = calldata;
2056
2057         data->rpc_status = task->tk_status;
2058
2059         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2060                 return;
2061
2062         if (task->tk_status == 0) {
2063                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2064                         switch (data->o_res.f_attr->mode & S_IFMT) {
2065                         case S_IFREG:
2066                                 break;
2067                         case S_IFLNK:
2068                                 data->rpc_status = -ELOOP;
2069                                 break;
2070                         case S_IFDIR:
2071                                 data->rpc_status = -EISDIR;
2072                                 break;
2073                         default:
2074                                 data->rpc_status = -ENOTDIR;
2075                         }
2076                 }
2077                 renew_lease(data->o_res.server, data->timestamp);
2078                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2079                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2080         }
2081         data->rpc_done = 1;
2082 }
2083
2084 static void nfs4_open_release(void *calldata)
2085 {
2086         struct nfs4_opendata *data = calldata;
2087         struct nfs4_state *state = NULL;
2088
2089         /* If this request hasn't been cancelled, do nothing */
2090         if (data->cancelled == 0)
2091                 goto out_free;
2092         /* In case of error, no cleanup! */
2093         if (data->rpc_status != 0 || !data->rpc_done)
2094                 goto out_free;
2095         /* In case we need an open_confirm, no cleanup! */
2096         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2097                 goto out_free;
2098         state = nfs4_opendata_to_nfs4_state(data);
2099         if (!IS_ERR(state))
2100                 nfs4_close_state(state, data->o_arg.fmode);
2101 out_free:
2102         nfs4_opendata_put(data);
2103 }
2104
2105 static const struct rpc_call_ops nfs4_open_ops = {
2106         .rpc_call_prepare = nfs4_open_prepare,
2107         .rpc_call_done = nfs4_open_done,
2108         .rpc_release = nfs4_open_release,
2109 };
2110
2111 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2112 {
2113         struct inode *dir = d_inode(data->dir);
2114         struct nfs_server *server = NFS_SERVER(dir);
2115         struct nfs_openargs *o_arg = &data->o_arg;
2116         struct nfs_openres *o_res = &data->o_res;
2117         struct rpc_task *task;
2118         struct rpc_message msg = {
2119                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2120                 .rpc_argp = o_arg,
2121                 .rpc_resp = o_res,
2122                 .rpc_cred = data->owner->so_cred,
2123         };
2124         struct rpc_task_setup task_setup_data = {
2125                 .rpc_client = server->client,
2126                 .rpc_message = &msg,
2127                 .callback_ops = &nfs4_open_ops,
2128                 .callback_data = data,
2129                 .workqueue = nfsiod_workqueue,
2130                 .flags = RPC_TASK_ASYNC,
2131         };
2132         int status;
2133
2134         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2135         kref_get(&data->kref);
2136         data->rpc_done = 0;
2137         data->rpc_status = 0;
2138         data->cancelled = 0;
2139         data->is_recover = 0;
2140         if (isrecover) {
2141                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2142                 data->is_recover = 1;
2143         }
2144         task = rpc_run_task(&task_setup_data);
2145         if (IS_ERR(task))
2146                 return PTR_ERR(task);
2147         status = nfs4_wait_for_completion_rpc_task(task);
2148         if (status != 0) {
2149                 data->cancelled = 1;
2150                 smp_wmb();
2151         } else
2152                 status = data->rpc_status;
2153         rpc_put_task(task);
2154
2155         return status;
2156 }
2157
2158 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2159 {
2160         struct inode *dir = d_inode(data->dir);
2161         struct nfs_openres *o_res = &data->o_res;
2162         int status;
2163
2164         status = nfs4_run_open_task(data, 1);
2165         if (status != 0 || !data->rpc_done)
2166                 return status;
2167
2168         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2169
2170         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2171                 status = _nfs4_proc_open_confirm(data);
2172                 if (status != 0)
2173                         return status;
2174         }
2175
2176         return status;
2177 }
2178
2179 /*
2180  * Additional permission checks in order to distinguish between an
2181  * open for read, and an open for execute. This works around the
2182  * fact that NFSv4 OPEN treats read and execute permissions as being
2183  * the same.
2184  * Note that in the non-execute case, we want to turn off permission
2185  * checking if we just created a new file (POSIX open() semantics).
2186  */
2187 static int nfs4_opendata_access(struct rpc_cred *cred,
2188                                 struct nfs4_opendata *opendata,
2189                                 struct nfs4_state *state, fmode_t fmode,
2190                                 int openflags)
2191 {
2192         struct nfs_access_entry cache;
2193         u32 mask;
2194
2195         /* access call failed or for some reason the server doesn't
2196          * support any access modes -- defer access call until later */
2197         if (opendata->o_res.access_supported == 0)
2198                 return 0;
2199
2200         mask = 0;
2201         /*
2202          * Use openflags to check for exec, because fmode won't
2203          * always have FMODE_EXEC set when file open for exec.
2204          */
2205         if (openflags & __FMODE_EXEC) {
2206                 /* ONLY check for exec rights */
2207                 mask = MAY_EXEC;
2208         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2209                 mask = MAY_READ;
2210
2211         cache.cred = cred;
2212         cache.jiffies = jiffies;
2213         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2214         nfs_access_add_cache(state->inode, &cache);
2215
2216         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2217                 return 0;
2218
2219         /* even though OPEN succeeded, access is denied. Close the file */
2220         nfs4_close_state(state, fmode);
2221         return -EACCES;
2222 }
2223
2224 /*
2225  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2226  */
2227 static int _nfs4_proc_open(struct nfs4_opendata *data)
2228 {
2229         struct inode *dir = d_inode(data->dir);
2230         struct nfs_server *server = NFS_SERVER(dir);
2231         struct nfs_openargs *o_arg = &data->o_arg;
2232         struct nfs_openres *o_res = &data->o_res;
2233         int status;
2234
2235         status = nfs4_run_open_task(data, 0);
2236         if (!data->rpc_done)
2237                 return status;
2238         if (status != 0) {
2239                 if (status == -NFS4ERR_BADNAME &&
2240                                 !(o_arg->open_flags & O_CREAT))
2241                         return -ENOENT;
2242                 return status;
2243         }
2244
2245         nfs_fattr_map_and_free_names(server, &data->f_attr);
2246
2247         if (o_arg->open_flags & O_CREAT) {
2248                 update_changeattr(dir, &o_res->cinfo);
2249                 if (o_arg->open_flags & O_EXCL)
2250                         data->file_created = 1;
2251                 else if (o_res->cinfo.before != o_res->cinfo.after)
2252                         data->file_created = 1;
2253         }
2254         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2255                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2256         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2257                 status = _nfs4_proc_open_confirm(data);
2258                 if (status != 0)
2259                         return status;
2260         }
2261         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2262                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2263         return 0;
2264 }
2265
2266 static int nfs4_recover_expired_lease(struct nfs_server *server)
2267 {
2268         return nfs4_client_recover_expired_lease(server->nfs_client);
2269 }
2270
2271 /*
2272  * OPEN_EXPIRED:
2273  *      reclaim state on the server after a network partition.
2274  *      Assumes caller holds the appropriate lock
2275  */
2276 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2277 {
2278         struct nfs4_opendata *opendata;
2279         int ret;
2280
2281         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2282                         NFS4_OPEN_CLAIM_FH);
2283         if (IS_ERR(opendata))
2284                 return PTR_ERR(opendata);
2285         ret = nfs4_open_recover(opendata, state);
2286         if (ret == -ESTALE)
2287                 d_drop(ctx->dentry);
2288         nfs4_opendata_put(opendata);
2289         return ret;
2290 }
2291
2292 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2293 {
2294         struct nfs_server *server = NFS_SERVER(state->inode);
2295         struct nfs4_exception exception = { };
2296         int err;
2297
2298         do {
2299                 err = _nfs4_open_expired(ctx, state);
2300                 trace_nfs4_open_expired(ctx, 0, err);
2301                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2302                         continue;
2303                 switch (err) {
2304                 default:
2305                         goto out;
2306                 case -NFS4ERR_GRACE:
2307                 case -NFS4ERR_DELAY:
2308                         nfs4_handle_exception(server, err, &exception);
2309                         err = 0;
2310                 }
2311         } while (exception.retry);
2312 out:
2313         return err;
2314 }
2315
2316 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2317 {
2318         struct nfs_open_context *ctx;
2319         int ret;
2320
2321         ctx = nfs4_state_find_open_context(state);
2322         if (IS_ERR(ctx))
2323                 return -EAGAIN;
2324         ret = nfs4_do_open_expired(ctx, state);
2325         put_nfs_open_context(ctx);
2326         return ret;
2327 }
2328
2329 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2330 {
2331         nfs_remove_bad_delegation(state->inode);
2332         write_seqlock(&state->seqlock);
2333         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2334         write_sequnlock(&state->seqlock);
2335         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2336 }
2337
2338 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2339 {
2340         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2341                 nfs_finish_clear_delegation_stateid(state);
2342 }
2343
2344 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2345 {
2346         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2347         nfs40_clear_delegation_stateid(state);
2348         return nfs4_open_expired(sp, state);
2349 }
2350
2351 #if defined(CONFIG_NFS_V4_1)
2352 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2353 {
2354         struct nfs_server *server = NFS_SERVER(state->inode);
2355         nfs4_stateid stateid;
2356         struct nfs_delegation *delegation;
2357         struct rpc_cred *cred;
2358         int status;
2359
2360         /* Get the delegation credential for use by test/free_stateid */
2361         rcu_read_lock();
2362         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2363         if (delegation == NULL) {
2364                 rcu_read_unlock();
2365                 return;
2366         }
2367
2368         nfs4_stateid_copy(&stateid, &delegation->stateid);
2369         cred = get_rpccred(delegation->cred);
2370         rcu_read_unlock();
2371         status = nfs41_test_stateid(server, &stateid, cred);
2372         trace_nfs4_test_delegation_stateid(state, NULL, status);
2373
2374         if (status != NFS_OK) {
2375                 /* Free the stateid unless the server explicitly
2376                  * informs us the stateid is unrecognized. */
2377                 if (status != -NFS4ERR_BAD_STATEID)
2378                         nfs41_free_stateid(server, &stateid, cred);
2379                 nfs_finish_clear_delegation_stateid(state);
2380         }
2381
2382         put_rpccred(cred);
2383 }
2384
2385 /**
2386  * nfs41_check_open_stateid - possibly free an open stateid
2387  *
2388  * @state: NFSv4 state for an inode
2389  *
2390  * Returns NFS_OK if recovery for this stateid is now finished.
2391  * Otherwise a negative NFS4ERR value is returned.
2392  */
2393 static int nfs41_check_open_stateid(struct nfs4_state *state)
2394 {
2395         struct nfs_server *server = NFS_SERVER(state->inode);
2396         nfs4_stateid *stateid = &state->open_stateid;
2397         struct rpc_cred *cred = state->owner->so_cred;
2398         int status;
2399
2400         /* If a state reset has been done, test_stateid is unneeded */
2401         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2402             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2403             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2404                 return -NFS4ERR_BAD_STATEID;
2405
2406         status = nfs41_test_stateid(server, stateid, cred);
2407         trace_nfs4_test_open_stateid(state, NULL, status);
2408         if (status != NFS_OK) {
2409                 /* Free the stateid unless the server explicitly
2410                  * informs us the stateid is unrecognized. */
2411                 if (status != -NFS4ERR_BAD_STATEID)
2412                         nfs41_free_stateid(server, stateid, cred);
2413
2414                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2415                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2416                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2417                 clear_bit(NFS_OPEN_STATE, &state->flags);
2418         }
2419         return status;
2420 }
2421
2422 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2423 {
2424         int status;
2425
2426         nfs41_check_delegation_stateid(state);
2427         status = nfs41_check_open_stateid(state);
2428         if (status != NFS_OK)
2429                 status = nfs4_open_expired(sp, state);
2430         return status;
2431 }
2432 #endif
2433
2434 /*
2435  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2436  * fields corresponding to attributes that were used to store the verifier.
2437  * Make sure we clobber those fields in the later setattr call
2438  */
2439 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2440                                 struct iattr *sattr, struct nfs4_label **label)
2441 {
2442         const u32 *attrset = opendata->o_res.attrset;
2443
2444         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2445             !(sattr->ia_valid & ATTR_ATIME_SET))
2446                 sattr->ia_valid |= ATTR_ATIME;
2447
2448         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2449             !(sattr->ia_valid & ATTR_MTIME_SET))
2450                 sattr->ia_valid |= ATTR_MTIME;
2451
2452         /* Except MODE, it seems harmless of setting twice. */
2453         if ((attrset[1] & FATTR4_WORD1_MODE))
2454                 sattr->ia_valid &= ~ATTR_MODE;
2455
2456         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2457                 *label = NULL;
2458 }
2459
2460 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2461                 fmode_t fmode,
2462                 int flags,
2463                 struct nfs_open_context *ctx)
2464 {
2465         struct nfs4_state_owner *sp = opendata->owner;
2466         struct nfs_server *server = sp->so_server;
2467         struct dentry *dentry;
2468         struct nfs4_state *state;
2469         unsigned int seq;
2470         int ret;
2471
2472         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2473
2474         ret = _nfs4_proc_open(opendata);
2475         if (ret != 0)
2476                 goto out;
2477
2478         state = nfs4_opendata_to_nfs4_state(opendata);
2479         ret = PTR_ERR(state);
2480         if (IS_ERR(state))
2481                 goto out;
2482         if (server->caps & NFS_CAP_POSIX_LOCK)
2483                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2484
2485         dentry = opendata->dentry;
2486         if (d_really_is_negative(dentry)) {
2487                 struct dentry *alias;
2488                 d_drop(dentry);
2489                 alias = d_exact_alias(dentry, state->inode);
2490                 if (!alias)
2491                         alias = d_splice_alias(igrab(state->inode), dentry);
2492                 /* d_splice_alias() can't fail here - it's a non-directory */
2493                 if (alias) {
2494                         dput(ctx->dentry);
2495                         ctx->dentry = dentry = alias;
2496                 }
2497                 nfs_set_verifier(dentry,
2498                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2499         }
2500
2501         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2502         if (ret != 0)
2503                 goto out;
2504
2505         ctx->state = state;
2506         if (d_inode(dentry) == state->inode) {
2507                 nfs_inode_attach_open_context(ctx);
2508                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2509                         nfs4_schedule_stateid_recovery(server, state);
2510         }
2511 out:
2512         return ret;
2513 }
2514
2515 /*
2516  * Returns a referenced nfs4_state
2517  */
2518 static int _nfs4_do_open(struct inode *dir,
2519                         struct nfs_open_context *ctx,
2520                         int flags,
2521                         struct iattr *sattr,
2522                         struct nfs4_label *label,
2523                         int *opened)
2524 {
2525         struct nfs4_state_owner  *sp;
2526         struct nfs4_state     *state = NULL;
2527         struct nfs_server       *server = NFS_SERVER(dir);
2528         struct nfs4_opendata *opendata;
2529         struct dentry *dentry = ctx->dentry;
2530         struct rpc_cred *cred = ctx->cred;
2531         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2532         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2533         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2534         struct nfs4_label *olabel = NULL;
2535         int status;
2536
2537         /* Protect against reboot recovery conflicts */
2538         status = -ENOMEM;
2539         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2540         if (sp == NULL) {
2541                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2542                 goto out_err;
2543         }
2544         status = nfs4_recover_expired_lease(server);
2545         if (status != 0)
2546                 goto err_put_state_owner;
2547         if (d_really_is_positive(dentry))
2548                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2549         status = -ENOMEM;
2550         if (d_really_is_positive(dentry))
2551                 claim = NFS4_OPEN_CLAIM_FH;
2552         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2553                         label, claim, GFP_KERNEL);
2554         if (opendata == NULL)
2555                 goto err_put_state_owner;
2556
2557         if (label) {
2558                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2559                 if (IS_ERR(olabel)) {
2560                         status = PTR_ERR(olabel);
2561                         goto err_opendata_put;
2562                 }
2563         }
2564
2565         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2566                 if (!opendata->f_attr.mdsthreshold) {
2567                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2568                         if (!opendata->f_attr.mdsthreshold)
2569                                 goto err_free_label;
2570                 }
2571                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2572         }
2573         if (d_really_is_positive(dentry))
2574                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2575
2576         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2577         if (status != 0)
2578                 goto err_free_label;
2579         state = ctx->state;
2580
2581         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2582             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2583                 nfs4_exclusive_attrset(opendata, sattr, &label);
2584                 /*
2585                  * send create attributes which was not set by open
2586                  * with an extra setattr.
2587                  */
2588                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2589                         nfs_fattr_init(opendata->o_res.f_attr);
2590                         status = nfs4_do_setattr(state->inode, cred,
2591                                         opendata->o_res.f_attr, sattr,
2592                                         state, label, olabel);
2593                         if (status == 0) {
2594                                 nfs_setattr_update_inode(state->inode, sattr,
2595                                                 opendata->o_res.f_attr);
2596                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2597                         }
2598                 }
2599         }
2600         if (opened && opendata->file_created)
2601                 *opened |= FILE_CREATED;
2602
2603         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2604                 *ctx_th = opendata->f_attr.mdsthreshold;
2605                 opendata->f_attr.mdsthreshold = NULL;
2606         }
2607
2608         nfs4_label_free(olabel);
2609
2610         nfs4_opendata_put(opendata);
2611         nfs4_put_state_owner(sp);
2612         return 0;
2613 err_free_label:
2614         nfs4_label_free(olabel);
2615 err_opendata_put:
2616         nfs4_opendata_put(opendata);
2617 err_put_state_owner:
2618         nfs4_put_state_owner(sp);
2619 out_err:
2620         return status;
2621 }
2622
2623
2624 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2625                                         struct nfs_open_context *ctx,
2626                                         int flags,
2627                                         struct iattr *sattr,
2628                                         struct nfs4_label *label,
2629                                         int *opened)
2630 {
2631         struct nfs_server *server = NFS_SERVER(dir);
2632         struct nfs4_exception exception = { };
2633         struct nfs4_state *res;
2634         int status;
2635
2636         do {
2637                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2638                 res = ctx->state;
2639                 trace_nfs4_open_file(ctx, flags, status);
2640                 if (status == 0)
2641                         break;
2642                 /* NOTE: BAD_SEQID means the server and client disagree about the
2643                  * book-keeping w.r.t. state-changing operations
2644                  * (OPEN/CLOSE/LOCK/LOCKU...)
2645                  * It is actually a sign of a bug on the client or on the server.
2646                  *
2647                  * If we receive a BAD_SEQID error in the particular case of
2648                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2649                  * have unhashed the old state_owner for us, and that we can
2650                  * therefore safely retry using a new one. We should still warn
2651                  * the user though...
2652                  */
2653                 if (status == -NFS4ERR_BAD_SEQID) {
2654                         pr_warn_ratelimited("NFS: v4 server %s "
2655                                         " returned a bad sequence-id error!\n",
2656                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2657                         exception.retry = 1;
2658                         continue;
2659                 }
2660                 /*
2661                  * BAD_STATEID on OPEN means that the server cancelled our
2662                  * state before it received the OPEN_CONFIRM.
2663                  * Recover by retrying the request as per the discussion
2664                  * on Page 181 of RFC3530.
2665                  */
2666                 if (status == -NFS4ERR_BAD_STATEID) {
2667                         exception.retry = 1;
2668                         continue;
2669                 }
2670                 if (status == -EAGAIN) {
2671                         /* We must have found a delegation */
2672                         exception.retry = 1;
2673                         continue;
2674                 }
2675                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2676                         continue;
2677                 res = ERR_PTR(nfs4_handle_exception(server,
2678                                         status, &exception));
2679         } while (exception.retry);
2680         return res;
2681 }
2682
2683 static int _nfs4_do_setattr(struct inode *inode,
2684                             struct nfs_setattrargs *arg,
2685                             struct nfs_setattrres *res,
2686                             struct rpc_cred *cred,
2687                             struct nfs4_state *state)
2688 {
2689         struct nfs_server *server = NFS_SERVER(inode);
2690         struct rpc_message msg = {
2691                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2692                 .rpc_argp       = arg,
2693                 .rpc_resp       = res,
2694                 .rpc_cred       = cred,
2695         };
2696         struct rpc_cred *delegation_cred = NULL;
2697         unsigned long timestamp = jiffies;
2698         fmode_t fmode;
2699         bool truncate;
2700         int status;
2701
2702         nfs_fattr_init(res->fattr);
2703
2704         /* Servers should only apply open mode checks for file size changes */
2705         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2706         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2707
2708         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2709                 /* Use that stateid */
2710         } else if (truncate && state != NULL) {
2711                 struct nfs_lockowner lockowner = {
2712                         .l_owner = current->files,
2713                         .l_pid = current->tgid,
2714                 };
2715                 if (!nfs4_valid_open_stateid(state))
2716                         return -EBADF;
2717                 if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner,
2718                                 &arg->stateid, &delegation_cred) == -EIO)
2719                         return -EBADF;
2720         } else
2721                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2722         if (delegation_cred)
2723                 msg.rpc_cred = delegation_cred;
2724
2725         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2726
2727         put_rpccred(delegation_cred);
2728         if (status == 0 && state != NULL)
2729                 renew_lease(server, timestamp);
2730         trace_nfs4_setattr(inode, &arg->stateid, status);
2731         return status;
2732 }
2733
2734 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2735                            struct nfs_fattr *fattr, struct iattr *sattr,
2736                            struct nfs4_state *state, struct nfs4_label *ilabel,
2737                            struct nfs4_label *olabel)
2738 {
2739         struct nfs_server *server = NFS_SERVER(inode);
2740         struct nfs_setattrargs  arg = {
2741                 .fh             = NFS_FH(inode),
2742                 .iap            = sattr,
2743                 .server         = server,
2744                 .bitmask = server->attr_bitmask,
2745                 .label          = ilabel,
2746         };
2747         struct nfs_setattrres  res = {
2748                 .fattr          = fattr,
2749                 .label          = olabel,
2750                 .server         = server,
2751         };
2752         struct nfs4_exception exception = {
2753                 .state = state,
2754                 .inode = inode,
2755                 .stateid = &arg.stateid,
2756         };
2757         int err;
2758
2759         arg.bitmask = nfs4_bitmask(server, ilabel);
2760         if (ilabel)
2761                 arg.bitmask = nfs4_bitmask(server, olabel);
2762
2763         do {
2764                 err = _nfs4_do_setattr(inode, &arg, &res, cred, state);
2765                 switch (err) {
2766                 case -NFS4ERR_OPENMODE:
2767                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2768                                 pr_warn_once("NFSv4: server %s is incorrectly "
2769                                                 "applying open mode checks to "
2770                                                 "a SETATTR that is not "
2771                                                 "changing file size.\n",
2772                                                 server->nfs_client->cl_hostname);
2773                         }
2774                         if (state && !(state->state & FMODE_WRITE)) {
2775                                 err = -EBADF;
2776                                 if (sattr->ia_valid & ATTR_OPEN)
2777                                         err = -EACCES;
2778                                 goto out;
2779                         }
2780                 }
2781                 err = nfs4_handle_exception(server, err, &exception);
2782         } while (exception.retry);
2783 out:
2784         return err;
2785 }
2786
2787 static bool
2788 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2789 {
2790         if (inode == NULL || !nfs_have_layout(inode))
2791                 return false;
2792
2793         return pnfs_wait_on_layoutreturn(inode, task);
2794 }
2795
2796 struct nfs4_closedata {
2797         struct inode *inode;
2798         struct nfs4_state *state;
2799         struct nfs_closeargs arg;
2800         struct nfs_closeres res;
2801         struct nfs_fattr fattr;
2802         unsigned long timestamp;
2803         bool roc;
2804         u32 roc_barrier;
2805 };
2806
2807 static void nfs4_free_closedata(void *data)
2808 {
2809         struct nfs4_closedata *calldata = data;
2810         struct nfs4_state_owner *sp = calldata->state->owner;
2811         struct super_block *sb = calldata->state->inode->i_sb;
2812
2813         if (calldata->roc)
2814                 pnfs_roc_release(calldata->state->inode);
2815         nfs4_put_open_state(calldata->state);
2816         nfs_free_seqid(calldata->arg.seqid);
2817         nfs4_put_state_owner(sp);
2818         nfs_sb_deactive(sb);
2819         kfree(calldata);
2820 }
2821
2822 static void nfs4_close_done(struct rpc_task *task, void *data)
2823 {
2824         struct nfs4_closedata *calldata = data;
2825         struct nfs4_state *state = calldata->state;
2826         struct nfs_server *server = NFS_SERVER(calldata->inode);
2827         nfs4_stateid *res_stateid = NULL;
2828
2829         dprintk("%s: begin!\n", __func__);
2830         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2831                 return;
2832         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2833         /* hmm. we are done with the inode, and in the process of freeing
2834          * the state_owner. we keep this around to process errors
2835          */
2836         switch (task->tk_status) {
2837                 case 0:
2838                         res_stateid = &calldata->res.stateid;
2839                         if (calldata->roc)
2840                                 pnfs_roc_set_barrier(state->inode,
2841                                                      calldata->roc_barrier);
2842                         renew_lease(server, calldata->timestamp);
2843                         break;
2844                 case -NFS4ERR_ADMIN_REVOKED:
2845                 case -NFS4ERR_STALE_STATEID:
2846                 case -NFS4ERR_OLD_STATEID:
2847                 case -NFS4ERR_BAD_STATEID:
2848                 case -NFS4ERR_EXPIRED:
2849                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2850                                                 &state->open_stateid)) {
2851                                 rpc_restart_call_prepare(task);
2852                                 goto out_release;
2853                         }
2854                         if (calldata->arg.fmode == 0)
2855                                 break;
2856                 default:
2857                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2858                                 rpc_restart_call_prepare(task);
2859                                 goto out_release;
2860                         }
2861         }
2862         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2863                         res_stateid, calldata->arg.fmode);
2864 out_release:
2865         nfs_release_seqid(calldata->arg.seqid);
2866         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2867         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2868 }
2869
2870 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2871 {
2872         struct nfs4_closedata *calldata = data;
2873         struct nfs4_state *state = calldata->state;
2874         struct inode *inode = calldata->inode;
2875         bool is_rdonly, is_wronly, is_rdwr;
2876         int call_close = 0;
2877
2878         dprintk("%s: begin!\n", __func__);
2879         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2880                 goto out_wait;
2881
2882         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2883         spin_lock(&state->owner->so_lock);
2884         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2885         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2886         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2887         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2888         /* Calculate the change in open mode */
2889         calldata->arg.fmode = 0;
2890         if (state->n_rdwr == 0) {
2891                 if (state->n_rdonly == 0)
2892                         call_close |= is_rdonly;
2893                 else if (is_rdonly)
2894                         calldata->arg.fmode |= FMODE_READ;
2895                 if (state->n_wronly == 0)
2896                         call_close |= is_wronly;
2897                 else if (is_wronly)
2898                         calldata->arg.fmode |= FMODE_WRITE;
2899                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
2900                         call_close |= is_rdwr;
2901         } else if (is_rdwr)
2902                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2903
2904         if (!nfs4_valid_open_stateid(state))
2905                 call_close = 0;
2906         spin_unlock(&state->owner->so_lock);
2907
2908         if (!call_close) {
2909                 /* Note: exit _without_ calling nfs4_close_done */
2910                 goto out_no_action;
2911         }
2912
2913         if (nfs4_wait_on_layoutreturn(inode, task)) {
2914                 nfs_release_seqid(calldata->arg.seqid);
2915                 goto out_wait;
2916         }
2917
2918         if (calldata->arg.fmode == 0)
2919                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2920         if (calldata->roc)
2921                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2922
2923         calldata->arg.share_access =
2924                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2925                                 calldata->arg.fmode, 0);
2926
2927         nfs_fattr_init(calldata->res.fattr);
2928         calldata->timestamp = jiffies;
2929         if (nfs4_setup_sequence(NFS_SERVER(inode),
2930                                 &calldata->arg.seq_args,
2931                                 &calldata->res.seq_res,
2932                                 task) != 0)
2933                 nfs_release_seqid(calldata->arg.seqid);
2934         dprintk("%s: done!\n", __func__);
2935         return;
2936 out_no_action:
2937         task->tk_action = NULL;
2938 out_wait:
2939         nfs4_sequence_done(task, &calldata->res.seq_res);
2940 }
2941
2942 static const struct rpc_call_ops nfs4_close_ops = {
2943         .rpc_call_prepare = nfs4_close_prepare,
2944         .rpc_call_done = nfs4_close_done,
2945         .rpc_release = nfs4_free_closedata,
2946 };
2947
2948 static bool nfs4_roc(struct inode *inode)
2949 {
2950         if (!nfs_have_layout(inode))
2951                 return false;
2952         return pnfs_roc(inode);
2953 }
2954
2955 /* 
2956  * It is possible for data to be read/written from a mem-mapped file 
2957  * after the sys_close call (which hits the vfs layer as a flush).
2958  * This means that we can't safely call nfsv4 close on a file until 
2959  * the inode is cleared. This in turn means that we are not good
2960  * NFSv4 citizens - we do not indicate to the server to update the file's 
2961  * share state even when we are done with one of the three share 
2962  * stateid's in the inode.
2963  *
2964  * NOTE: Caller must be holding the sp->so_owner semaphore!
2965  */
2966 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2967 {
2968         struct nfs_server *server = NFS_SERVER(state->inode);
2969         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2970         struct nfs4_closedata *calldata;
2971         struct nfs4_state_owner *sp = state->owner;
2972         struct rpc_task *task;
2973         struct rpc_message msg = {
2974                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2975                 .rpc_cred = state->owner->so_cred,
2976         };
2977         struct rpc_task_setup task_setup_data = {
2978                 .rpc_client = server->client,
2979                 .rpc_message = &msg,
2980                 .callback_ops = &nfs4_close_ops,
2981                 .workqueue = nfsiod_workqueue,
2982                 .flags = RPC_TASK_ASYNC,
2983         };
2984         int status = -ENOMEM;
2985
2986         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2987                 &task_setup_data.rpc_client, &msg);
2988
2989         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2990         if (calldata == NULL)
2991                 goto out;
2992         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2993         calldata->inode = state->inode;
2994         calldata->state = state;
2995         calldata->arg.fh = NFS_FH(state->inode);
2996         /* Serialization for the sequence id */
2997         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2998         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2999         if (IS_ERR(calldata->arg.seqid))
3000                 goto out_free_calldata;
3001         calldata->arg.fmode = 0;
3002         calldata->arg.bitmask = server->cache_consistency_bitmask;
3003         calldata->res.fattr = &calldata->fattr;
3004         calldata->res.seqid = calldata->arg.seqid;
3005         calldata->res.server = server;
3006         calldata->roc = nfs4_roc(state->inode);
3007         nfs_sb_active(calldata->inode->i_sb);
3008
3009         msg.rpc_argp = &calldata->arg;
3010         msg.rpc_resp = &calldata->res;
3011         task_setup_data.callback_data = calldata;
3012         task = rpc_run_task(&task_setup_data);
3013         if (IS_ERR(task))
3014                 return PTR_ERR(task);
3015         status = 0;
3016         if (wait)
3017                 status = rpc_wait_for_completion_task(task);
3018         rpc_put_task(task);
3019         return status;
3020 out_free_calldata:
3021         kfree(calldata);
3022 out:
3023         nfs4_put_open_state(state);
3024         nfs4_put_state_owner(sp);
3025         return status;
3026 }
3027
3028 static struct inode *
3029 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3030                 int open_flags, struct iattr *attr, int *opened)
3031 {
3032         struct nfs4_state *state;
3033         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3034
3035         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3036
3037         /* Protect against concurrent sillydeletes */
3038         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3039
3040         nfs4_label_release_security(label);
3041
3042         if (IS_ERR(state))
3043                 return ERR_CAST(state);
3044         return state->inode;
3045 }
3046
3047 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3048 {
3049         if (ctx->state == NULL)
3050                 return;
3051         if (is_sync)
3052                 nfs4_close_sync(ctx->state, ctx->mode);
3053         else
3054                 nfs4_close_state(ctx->state, ctx->mode);
3055 }
3056
3057 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3058 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3059 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3060
3061 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3062 {
3063         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3064         struct nfs4_server_caps_arg args = {
3065                 .fhandle = fhandle,
3066                 .bitmask = bitmask,
3067         };
3068         struct nfs4_server_caps_res res = {};
3069         struct rpc_message msg = {
3070                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3071                 .rpc_argp = &args,
3072                 .rpc_resp = &res,
3073         };
3074         int status;
3075
3076         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3077                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3078                      FATTR4_WORD0_LINK_SUPPORT |
3079                      FATTR4_WORD0_SYMLINK_SUPPORT |
3080                      FATTR4_WORD0_ACLSUPPORT;
3081         if (minorversion)
3082                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3083
3084         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3085         if (status == 0) {
3086                 /* Sanity check the server answers */
3087                 switch (minorversion) {
3088                 case 0:
3089                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3090                         res.attr_bitmask[2] = 0;
3091                         break;
3092                 case 1:
3093                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3094                         break;
3095                 case 2:
3096                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3097                 }
3098                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3099                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3100                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3101                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3102                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3103                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3104                                 NFS_CAP_SECURITY_LABEL);
3105                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3106                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3107                         server->caps |= NFS_CAP_ACLS;
3108                 if (res.has_links != 0)
3109                         server->caps |= NFS_CAP_HARDLINKS;
3110                 if (res.has_symlinks != 0)
3111                         server->caps |= NFS_CAP_SYMLINKS;
3112                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3113                         server->caps |= NFS_CAP_FILEID;
3114                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3115                         server->caps |= NFS_CAP_MODE;
3116                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3117                         server->caps |= NFS_CAP_NLINK;
3118                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3119                         server->caps |= NFS_CAP_OWNER;
3120                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3121                         server->caps |= NFS_CAP_OWNER_GROUP;
3122                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3123                         server->caps |= NFS_CAP_ATIME;
3124                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3125                         server->caps |= NFS_CAP_CTIME;
3126                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3127                         server->caps |= NFS_CAP_MTIME;
3128 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3129                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3130                         server->caps |= NFS_CAP_SECURITY_LABEL;
3131 #endif
3132                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3133                                 sizeof(server->attr_bitmask));
3134                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3135
3136                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3137                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3138                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3139                 server->cache_consistency_bitmask[2] = 0;
3140                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3141                         sizeof(server->exclcreat_bitmask));
3142                 server->acl_bitmask = res.acl_bitmask;
3143                 server->fh_expire_type = res.fh_expire_type;
3144         }
3145
3146         return status;
3147 }
3148
3149 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3150 {
3151         struct nfs4_exception exception = { };
3152         int err;
3153         do {
3154                 err = nfs4_handle_exception(server,
3155                                 _nfs4_server_capabilities(server, fhandle),
3156                                 &exception);
3157         } while (exception.retry);
3158         return err;
3159 }
3160
3161 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3162                 struct nfs_fsinfo *info)
3163 {
3164         u32 bitmask[3];
3165         struct nfs4_lookup_root_arg args = {
3166                 .bitmask = bitmask,
3167         };
3168         struct nfs4_lookup_res res = {
3169                 .server = server,
3170                 .fattr = info->fattr,
3171                 .fh = fhandle,
3172         };
3173         struct rpc_message msg = {
3174                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3175                 .rpc_argp = &args,
3176                 .rpc_resp = &res,
3177         };
3178
3179         bitmask[0] = nfs4_fattr_bitmap[0];
3180         bitmask[1] = nfs4_fattr_bitmap[1];
3181         /*
3182          * Process the label in the upcoming getfattr
3183          */
3184         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3185
3186         nfs_fattr_init(info->fattr);
3187         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3188 }
3189
3190 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3191                 struct nfs_fsinfo *info)
3192 {
3193         struct nfs4_exception exception = { };
3194         int err;
3195         do {
3196                 err = _nfs4_lookup_root(server, fhandle, info);
3197                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3198                 switch (err) {
3199                 case 0:
3200                 case -NFS4ERR_WRONGSEC:
3201                         goto out;
3202                 default:
3203                         err = nfs4_handle_exception(server, err, &exception);
3204                 }
3205         } while (exception.retry);
3206 out:
3207         return err;
3208 }
3209
3210 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3211                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3212 {
3213         struct rpc_auth_create_args auth_args = {
3214                 .pseudoflavor = flavor,
3215         };
3216         struct rpc_auth *auth;
3217         int ret;
3218
3219         auth = rpcauth_create(&auth_args, server->client);
3220         if (IS_ERR(auth)) {
3221                 ret = -EACCES;
3222                 goto out;
3223         }
3224         ret = nfs4_lookup_root(server, fhandle, info);
3225 out:
3226         return ret;
3227 }
3228
3229 /*
3230  * Retry pseudoroot lookup with various security flavors.  We do this when:
3231  *
3232  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3233  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3234  *
3235  * Returns zero on success, or a negative NFS4ERR value, or a
3236  * negative errno value.
3237  */
3238 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3239                               struct nfs_fsinfo *info)
3240 {
3241         /* Per 3530bis 15.33.5 */
3242         static const rpc_authflavor_t flav_array[] = {
3243                 RPC_AUTH_GSS_KRB5P,
3244                 RPC_AUTH_GSS_KRB5I,
3245                 RPC_AUTH_GSS_KRB5,
3246                 RPC_AUTH_UNIX,                  /* courtesy */
3247                 RPC_AUTH_NULL,
3248         };
3249         int status = -EPERM;
3250         size_t i;
3251
3252         if (server->auth_info.flavor_len > 0) {
3253                 /* try each flavor specified by user */
3254                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3255                         status = nfs4_lookup_root_sec(server, fhandle, info,
3256                                                 server->auth_info.flavors[i]);
3257                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3258                                 continue;
3259                         break;
3260                 }
3261         } else {
3262                 /* no flavors specified by user, try default list */
3263                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3264                         status = nfs4_lookup_root_sec(server, fhandle, info,
3265                                                       flav_array[i]);
3266                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3267                                 continue;
3268                         break;
3269                 }
3270         }
3271
3272         /*
3273          * -EACCESS could mean that the user doesn't have correct permissions
3274          * to access the mount.  It could also mean that we tried to mount
3275          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3276          * existing mount programs don't handle -EACCES very well so it should
3277          * be mapped to -EPERM instead.
3278          */
3279         if (status == -EACCES)
3280                 status = -EPERM;
3281         return status;
3282 }
3283
3284 /**
3285  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3286  * @server: initialized nfs_server handle
3287  * @fhandle: we fill in the pseudo-fs root file handle
3288  * @info: we fill in an FSINFO struct
3289  * @auth_probe: probe the auth flavours
3290  *
3291  * Returns zero on success, or a negative errno.
3292  */
3293 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3294                          struct nfs_fsinfo *info,
3295                          bool auth_probe)
3296 {
3297         int status = 0;
3298
3299         if (!auth_probe)
3300                 status = nfs4_lookup_root(server, fhandle, info);
3301
3302         if (auth_probe || status == NFS4ERR_WRONGSEC)
3303                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3304                                 fhandle, info);
3305
3306         if (status == 0)
3307                 status = nfs4_server_capabilities(server, fhandle);
3308         if (status == 0)
3309                 status = nfs4_do_fsinfo(server, fhandle, info);
3310
3311         return nfs4_map_errors(status);
3312 }
3313
3314 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3315                               struct nfs_fsinfo *info)
3316 {
3317         int error;
3318         struct nfs_fattr *fattr = info->fattr;
3319         struct nfs4_label *label = NULL;
3320
3321         error = nfs4_server_capabilities(server, mntfh);
3322         if (error < 0) {
3323                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3324                 return error;
3325         }
3326
3327         label = nfs4_label_alloc(server, GFP_KERNEL);
3328         if (IS_ERR(label))
3329                 return PTR_ERR(label);
3330
3331         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3332         if (error < 0) {
3333                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3334                 goto err_free_label;
3335         }
3336
3337         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3338             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3339                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3340
3341 err_free_label:
3342         nfs4_label_free(label);
3343
3344         return error;
3345 }
3346
3347 /*
3348  * Get locations and (maybe) other attributes of a referral.
3349  * Note that we'll actually follow the referral later when
3350  * we detect fsid mismatch in inode revalidation
3351  */
3352 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3353                              const struct qstr *name, struct nfs_fattr *fattr,
3354                              struct nfs_fh *fhandle)
3355 {
3356         int status = -ENOMEM;
3357         struct page *page = NULL;
3358         struct nfs4_fs_locations *locations = NULL;
3359
3360         page = alloc_page(GFP_KERNEL);
3361         if (page == NULL)
3362                 goto out;
3363         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3364         if (locations == NULL)
3365                 goto out;
3366
3367         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3368         if (status != 0)
3369                 goto out;
3370
3371         /*
3372          * If the fsid didn't change, this is a migration event, not a
3373          * referral.  Cause us to drop into the exception handler, which
3374          * will kick off migration recovery.
3375          */
3376         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3377                 dprintk("%s: server did not return a different fsid for"
3378                         " a referral at %s\n", __func__, name->name);
3379                 status = -NFS4ERR_MOVED;
3380                 goto out;
3381         }
3382         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3383         nfs_fixup_referral_attributes(&locations->fattr);
3384
3385         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3386         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3387         memset(fhandle, 0, sizeof(struct nfs_fh));
3388 out:
3389         if (page)
3390                 __free_page(page);
3391         kfree(locations);
3392         return status;
3393 }
3394
3395 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3396                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3397 {
3398         struct nfs4_getattr_arg args = {
3399                 .fh = fhandle,
3400                 .bitmask = server->attr_bitmask,
3401         };
3402         struct nfs4_getattr_res res = {
3403                 .fattr = fattr,
3404                 .label = label,
3405                 .server = server,
3406         };
3407         struct rpc_message msg = {
3408                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3409                 .rpc_argp = &args,
3410                 .rpc_resp = &res,
3411         };
3412
3413         args.bitmask = nfs4_bitmask(server, label);
3414
3415         nfs_fattr_init(fattr);
3416         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3417 }
3418
3419 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3420                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3421 {
3422         struct nfs4_exception exception = { };
3423         int err;
3424         do {
3425                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3426                 trace_nfs4_getattr(server, fhandle, fattr, err);
3427                 err = nfs4_handle_exception(server, err,
3428                                 &exception);
3429         } while (exception.retry);
3430         return err;
3431 }
3432
3433 /* 
3434  * The file is not closed if it is opened due to the a request to change
3435  * the size of the file. The open call will not be needed once the
3436  * VFS layer lookup-intents are implemented.
3437  *
3438  * Close is called when the inode is destroyed.
3439  * If we haven't opened the file for O_WRONLY, we
3440  * need to in the size_change case to obtain a stateid.
3441  *
3442  * Got race?
3443  * Because OPEN is always done by name in nfsv4, it is
3444  * possible that we opened a different file by the same
3445  * name.  We can recognize this race condition, but we
3446  * can't do anything about it besides returning an error.
3447  *
3448  * This will be fixed with VFS changes (lookup-intent).
3449  */
3450 static int
3451 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3452                   struct iattr *sattr)
3453 {
3454         struct inode *inode = d_inode(dentry);
3455         struct rpc_cred *cred = NULL;
3456         struct nfs4_state *state = NULL;
3457         struct nfs4_label *label = NULL;
3458         int status;
3459
3460         if (pnfs_ld_layoutret_on_setattr(inode) &&
3461             sattr->ia_valid & ATTR_SIZE &&
3462             sattr->ia_size < i_size_read(inode))
3463                 pnfs_commit_and_return_layout(inode);
3464
3465         nfs_fattr_init(fattr);
3466         
3467         /* Deal with open(O_TRUNC) */
3468         if (sattr->ia_valid & ATTR_OPEN)
3469                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3470
3471         /* Optimization: if the end result is no change, don't RPC */
3472         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3473                 return 0;
3474
3475         /* Search for an existing open(O_WRITE) file */
3476         if (sattr->ia_valid & ATTR_FILE) {
3477                 struct nfs_open_context *ctx;
3478
3479                 ctx = nfs_file_open_context(sattr->ia_file);
3480                 if (ctx) {
3481                         cred = ctx->cred;
3482                         state = ctx->state;
3483                 }
3484         }
3485
3486         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3487         if (IS_ERR(label))
3488                 return PTR_ERR(label);
3489
3490         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3491         if (status == 0) {
3492                 nfs_setattr_update_inode(inode, sattr, fattr);
3493                 nfs_setsecurity(inode, fattr, label);
3494         }
3495         nfs4_label_free(label);
3496         return status;
3497 }
3498
3499 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3500                 const struct qstr *name, struct nfs_fh *fhandle,
3501                 struct nfs_fattr *fattr, struct nfs4_label *label)
3502 {
3503         struct nfs_server *server = NFS_SERVER(dir);
3504         int                    status;
3505         struct nfs4_lookup_arg args = {
3506                 .bitmask = server->attr_bitmask,
3507                 .dir_fh = NFS_FH(dir),
3508                 .name = name,
3509         };
3510         struct nfs4_lookup_res res = {
3511                 .server = server,
3512                 .fattr = fattr,
3513                 .label = label,
3514                 .fh = fhandle,
3515         };
3516         struct rpc_message msg = {
3517                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3518                 .rpc_argp = &args,
3519                 .rpc_resp = &res,
3520         };
3521
3522         args.bitmask = nfs4_bitmask(server, label);
3523
3524         nfs_fattr_init(fattr);
3525
3526         dprintk("NFS call  lookup %s\n", name->name);
3527         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3528         dprintk("NFS reply lookup: %d\n", status);
3529         return status;
3530 }
3531
3532 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3533 {
3534         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3535                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3536         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3537         fattr->nlink = 2;
3538 }
3539
3540 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3541                                    struct qstr *name, struct nfs_fh *fhandle,
3542                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3543 {
3544         struct nfs4_exception exception = { };
3545         struct rpc_clnt *client = *clnt;
3546         int err;
3547         do {
3548                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3549                 trace_nfs4_lookup(dir, name, err);
3550                 switch (err) {
3551                 case -NFS4ERR_BADNAME:
3552                         err = -ENOENT;
3553                         goto out;
3554                 case -NFS4ERR_MOVED:
3555                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3556                         if (err == -NFS4ERR_MOVED)
3557                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3558                         goto out;
3559                 case -NFS4ERR_WRONGSEC:
3560                         err = -EPERM;
3561                         if (client != *clnt)
3562                                 goto out;
3563                         client = nfs4_negotiate_security(client, dir, name);
3564                         if (IS_ERR(client))
3565                                 return PTR_ERR(client);
3566
3567                         exception.retry = 1;
3568                         break;
3569                 default:
3570                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3571                 }
3572         } while (exception.retry);
3573
3574 out:
3575         if (err == 0)
3576                 *clnt = client;
3577         else if (client != *clnt)
3578                 rpc_shutdown_client(client);
3579
3580         return err;
3581 }
3582
3583 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3584                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3585                             struct nfs4_label *label)
3586 {
3587         int status;
3588         struct rpc_clnt *client = NFS_CLIENT(dir);
3589
3590         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3591         if (client != NFS_CLIENT(dir)) {
3592                 rpc_shutdown_client(client);
3593                 nfs_fixup_secinfo_attributes(fattr);
3594         }
3595         return status;
3596 }
3597
3598 struct rpc_clnt *
3599 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3600                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3601 {
3602         struct rpc_clnt *client = NFS_CLIENT(dir);
3603         int status;
3604
3605         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3606         if (status < 0)
3607                 return ERR_PTR(status);
3608         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3609 }
3610
3611 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3612 {
3613         struct nfs_server *server = NFS_SERVER(inode);
3614         struct nfs4_accessargs args = {
3615                 .fh = NFS_FH(inode),
3616                 .bitmask = server->cache_consistency_bitmask,
3617         };
3618         struct nfs4_accessres res = {
3619                 .server = server,
3620         };
3621         struct rpc_message msg = {
3622                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3623                 .rpc_argp = &args,
3624                 .rpc_resp = &res,
3625                 .rpc_cred = entry->cred,
3626         };
3627         int mode = entry->mask;
3628         int status = 0;
3629
3630         /*
3631          * Determine which access bits we want to ask for...
3632          */
3633         if (mode & MAY_READ)
3634                 args.access |= NFS4_ACCESS_READ;
3635         if (S_ISDIR(inode->i_mode)) {
3636                 if (mode & MAY_WRITE)
3637                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3638                 if (mode & MAY_EXEC)
3639                         args.access |= NFS4_ACCESS_LOOKUP;
3640         } else {
3641                 if (mode & MAY_WRITE)
3642                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3643                 if (mode & MAY_EXEC)
3644                         args.access |= NFS4_ACCESS_EXECUTE;
3645         }
3646
3647         res.fattr = nfs_alloc_fattr();
3648         if (res.fattr == NULL)
3649                 return -ENOMEM;
3650
3651         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3652         if (!status) {
3653                 nfs_access_set_mask(entry, res.access);
3654                 nfs_refresh_inode(inode, res.fattr);
3655         }
3656         nfs_free_fattr(res.fattr);
3657         return status;
3658 }
3659
3660 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3661 {
3662         struct nfs4_exception exception = { };
3663         int err;
3664         do {
3665                 err = _nfs4_proc_access(inode, entry);
3666                 trace_nfs4_access(inode, err);
3667                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3668                                 &exception);
3669         } while (exception.retry);
3670         return err;
3671 }
3672
3673 /*
3674  * TODO: For the time being, we don't try to get any attributes
3675  * along with any of the zero-copy operations READ, READDIR,
3676  * READLINK, WRITE.
3677  *
3678  * In the case of the first three, we want to put the GETATTR
3679  * after the read-type operation -- this is because it is hard
3680  * to predict the length of a GETATTR response in v4, and thus
3681  * align the READ data correctly.  This means that the GETATTR
3682  * may end up partially falling into the page cache, and we should
3683  * shift it into the 'tail' of the xdr_buf before processing.
3684  * To do this efficiently, we need to know the total length
3685  * of data received, which doesn't seem to be available outside
3686  * of the RPC layer.
3687  *
3688  * In the case of WRITE, we also want to put the GETATTR after
3689  * the operation -- in this case because we want to make sure
3690  * we get the post-operation mtime and size.
3691  *
3692  * Both of these changes to the XDR layer would in fact be quite
3693  * minor, but I decided to leave them for a subsequent patch.
3694  */
3695 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3696                 unsigned int pgbase, unsigned int pglen)
3697 {
3698         struct nfs4_readlink args = {
3699                 .fh       = NFS_FH(inode),
3700                 .pgbase   = pgbase,
3701                 .pglen    = pglen,
3702                 .pages    = &page,
3703         };
3704         struct nfs4_readlink_res res;
3705         struct rpc_message msg = {
3706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3707                 .rpc_argp = &args,
3708                 .rpc_resp = &res,
3709         };
3710
3711         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3712 }
3713
3714 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3715                 unsigned int pgbase, unsigned int pglen)
3716 {
3717         struct nfs4_exception exception = { };
3718         int err;
3719         do {
3720                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3721                 trace_nfs4_readlink(inode, err);
3722                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3723                                 &exception);
3724         } while (exception.retry);
3725         return err;
3726 }
3727
3728 /*
3729  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3730  */
3731 static int
3732 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3733                  int flags)
3734 {
3735         struct nfs4_label l, *ilabel = NULL;
3736         struct nfs_open_context *ctx;
3737         struct nfs4_state *state;
3738         int status = 0;
3739
3740         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3741         if (IS_ERR(ctx))
3742                 return PTR_ERR(ctx);
3743
3744         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3745
3746         sattr->ia_mode &= ~current_umask();
3747         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3748         if (IS_ERR(state)) {
3749                 status = PTR_ERR(state);
3750                 goto out;
3751         }
3752 out:
3753         nfs4_label_release_security(ilabel);
3754         put_nfs_open_context(ctx);
3755         return status;
3756 }
3757
3758 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3759 {
3760         struct nfs_server *server = NFS_SERVER(dir);
3761         struct nfs_removeargs args = {
3762                 .fh = NFS_FH(dir),
3763                 .name = *name,
3764         };
3765         struct nfs_removeres res = {
3766                 .server = server,
3767         };
3768         struct rpc_message msg = {
3769                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3770                 .rpc_argp = &args,
3771                 .rpc_resp = &res,
3772         };
3773         int status;
3774
3775         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3776         if (status == 0)
3777                 update_changeattr(dir, &res.cinfo);
3778         return status;
3779 }
3780
3781 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3782 {
3783         struct nfs4_exception exception = { };
3784         int err;
3785         do {
3786                 err = _nfs4_proc_remove(dir, name);
3787                 trace_nfs4_remove(dir, name, err);
3788                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3789                                 &exception);
3790         } while (exception.retry);
3791         return err;
3792 }
3793
3794 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3795 {
3796         struct nfs_server *server = NFS_SERVER(dir);
3797         struct nfs_removeargs *args = msg->rpc_argp;
3798         struct nfs_removeres *res = msg->rpc_resp;
3799
3800         res->server = server;
3801         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3802         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3803
3804         nfs_fattr_init(res->dir_attr);
3805 }
3806
3807 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3808 {
3809         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
3810                         &data->args.seq_args,
3811                         &data->res.seq_res,
3812                         task);
3813 }
3814
3815 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3816 {
3817         struct nfs_unlinkdata *data = task->tk_calldata;
3818         struct nfs_removeres *res = &data->res;
3819
3820         if (!nfs4_sequence_done(task, &res->seq_res))
3821                 return 0;
3822         if (nfs4_async_handle_error(task, res->server, NULL,
3823                                     &data->timeout) == -EAGAIN)
3824                 return 0;
3825         update_changeattr(dir, &res->cinfo);
3826         return 1;
3827 }
3828
3829 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3830 {
3831         struct nfs_server *server = NFS_SERVER(dir);
3832         struct nfs_renameargs *arg = msg->rpc_argp;
3833         struct nfs_renameres *res = msg->rpc_resp;
3834
3835         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3836         res->server = server;
3837         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3838 }
3839
3840 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3841 {
3842         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3843                         &data->args.seq_args,
3844                         &data->res.seq_res,
3845                         task);
3846 }
3847
3848 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3849                                  struct inode *new_dir)
3850 {
3851         struct nfs_renamedata *data = task->tk_calldata;
3852         struct nfs_renameres *res = &data->res;
3853
3854         if (!nfs4_sequence_done(task, &res->seq_res))
3855                 return 0;
3856         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3857                 return 0;
3858
3859         update_changeattr(old_dir, &res->old_cinfo);
3860         update_changeattr(new_dir, &res->new_cinfo);
3861         return 1;
3862 }
3863
3864 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3865 {
3866         struct nfs_server *server = NFS_SERVER(inode);
3867         struct nfs4_link_arg arg = {
3868                 .fh     = NFS_FH(inode),
3869                 .dir_fh = NFS_FH(dir),
3870                 .name   = name,
3871                 .bitmask = server->attr_bitmask,
3872         };
3873         struct nfs4_link_res res = {
3874                 .server = server,
3875                 .label = NULL,
3876         };
3877         struct rpc_message msg = {
3878                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3879                 .rpc_argp = &arg,
3880                 .rpc_resp = &res,
3881         };
3882         int status = -ENOMEM;
3883
3884         res.fattr = nfs_alloc_fattr();
3885         if (res.fattr == NULL)
3886                 goto out;
3887
3888         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3889         if (IS_ERR(res.label)) {
3890                 status = PTR_ERR(res.label);
3891                 goto out;
3892         }
3893         arg.bitmask = nfs4_bitmask(server, res.label);
3894
3895         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3896         if (!status) {
3897                 update_changeattr(dir, &res.cinfo);
3898                 status = nfs_post_op_update_inode(inode, res.fattr);
3899                 if (!status)
3900                         nfs_setsecurity(inode, res.fattr, res.label);
3901         }
3902
3903
3904         nfs4_label_free(res.label);
3905
3906 out:
3907         nfs_free_fattr(res.fattr);
3908         return status;
3909 }
3910
3911 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3912 {
3913         struct nfs4_exception exception = { };
3914         int err;
3915         do {
3916                 err = nfs4_handle_exception(NFS_SERVER(inode),
3917                                 _nfs4_proc_link(inode, dir, name),
3918                                 &exception);
3919         } while (exception.retry);
3920         return err;
3921 }
3922
3923 struct nfs4_createdata {
3924         struct rpc_message msg;
3925         struct nfs4_create_arg arg;
3926         struct nfs4_create_res res;
3927         struct nfs_fh fh;
3928         struct nfs_fattr fattr;
3929         struct nfs4_label *label;
3930 };
3931
3932 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3933                 struct qstr *name, struct iattr *sattr, u32 ftype)
3934 {
3935         struct nfs4_createdata *data;
3936
3937         data = kzalloc(sizeof(*data), GFP_KERNEL);
3938         if (data != NULL) {
3939                 struct nfs_server *server = NFS_SERVER(dir);
3940
3941                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3942                 if (IS_ERR(data->label))
3943                         goto out_free;
3944
3945                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3946                 data->msg.rpc_argp = &data->arg;
3947                 data->msg.rpc_resp = &data->res;
3948                 data->arg.dir_fh = NFS_FH(dir);
3949                 data->arg.server = server;
3950                 data->arg.name = name;
3951                 data->arg.attrs = sattr;
3952                 data->arg.ftype = ftype;
3953                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3954                 data->res.server = server;
3955                 data->res.fh = &data->fh;
3956                 data->res.fattr = &data->fattr;
3957                 data->res.label = data->label;
3958                 nfs_fattr_init(data->res.fattr);
3959         }
3960         return data;
3961 out_free:
3962         kfree(data);
3963         return NULL;
3964 }
3965
3966 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3967 {
3968         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3969                                     &data->arg.seq_args, &data->res.seq_res, 1);
3970         if (status == 0) {
3971                 update_changeattr(dir, &data->res.dir_cinfo);
3972                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3973         }
3974         return status;
3975 }
3976
3977 static void nfs4_free_createdata(struct nfs4_createdata *data)
3978 {
3979         nfs4_label_free(data->label);
3980         kfree(data);
3981 }
3982
3983 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3984                 struct page *page, unsigned int len, struct iattr *sattr,
3985                 struct nfs4_label *label)
3986 {
3987         struct nfs4_createdata *data;
3988         int status = -ENAMETOOLONG;
3989
3990         if (len > NFS4_MAXPATHLEN)
3991                 goto out;
3992
3993         status = -ENOMEM;
3994         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3995         if (data == NULL)
3996                 goto out;
3997
3998         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3999         data->arg.u.symlink.pages = &page;
4000         data->arg.u.symlink.len = len;
4001         data->arg.label = label;
4002         
4003         status = nfs4_do_create(dir, dentry, data);
4004
4005         nfs4_free_createdata(data);
4006 out:
4007         return status;
4008 }
4009
4010 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4011                 struct page *page, unsigned int len, struct iattr *sattr)
4012 {
4013         struct nfs4_exception exception = { };
4014         struct nfs4_label l, *label = NULL;
4015         int err;
4016
4017         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4018
4019         do {
4020                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4021                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4022                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4023                                 &exception);
4024         } while (exception.retry);
4025
4026         nfs4_label_release_security(label);
4027         return err;
4028 }
4029
4030 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4031                 struct iattr *sattr, struct nfs4_label *label)
4032 {
4033         struct nfs4_createdata *data;
4034         int status = -ENOMEM;
4035
4036         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4037         if (data == NULL)
4038                 goto out;
4039
4040         data->arg.label = label;
4041         status = nfs4_do_create(dir, dentry, data);
4042
4043         nfs4_free_createdata(data);
4044 out:
4045         return status;
4046 }
4047
4048 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4049                 struct iattr *sattr)
4050 {
4051         struct nfs4_exception exception = { };
4052         struct nfs4_label l, *label = NULL;
4053         int err;
4054
4055         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4056
4057         sattr->ia_mode &= ~current_umask();
4058         do {
4059                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4060                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4061                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4062                                 &exception);
4063         } while (exception.retry);
4064         nfs4_label_release_security(label);
4065
4066         return err;
4067 }
4068
4069 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4070                 u64 cookie, struct page **pages, unsigned int count, int plus)
4071 {
4072         struct inode            *dir = d_inode(dentry);
4073         struct nfs4_readdir_arg args = {
4074                 .fh = NFS_FH(dir),
4075                 .pages = pages,
4076                 .pgbase = 0,
4077                 .count = count,
4078                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4079                 .plus = plus,
4080         };
4081         struct nfs4_readdir_res res;
4082         struct rpc_message msg = {
4083                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4084                 .rpc_argp = &args,
4085                 .rpc_resp = &res,
4086                 .rpc_cred = cred,
4087         };
4088         int                     status;
4089
4090         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4091                         dentry,
4092                         (unsigned long long)cookie);
4093         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4094         res.pgbase = args.pgbase;
4095         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4096         if (status >= 0) {
4097                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4098                 status += args.pgbase;
4099         }
4100
4101         nfs_invalidate_atime(dir);
4102
4103         dprintk("%s: returns %d\n", __func__, status);
4104         return status;
4105 }
4106
4107 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4108                 u64 cookie, struct page **pages, unsigned int count, int plus)
4109 {
4110         struct nfs4_exception exception = { };
4111         int err;
4112         do {
4113                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4114                                 pages, count, plus);
4115                 trace_nfs4_readdir(d_inode(dentry), err);
4116                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4117                                 &exception);
4118         } while (exception.retry);
4119         return err;
4120 }
4121
4122 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4123                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4124 {
4125         struct nfs4_createdata *data;
4126         int mode = sattr->ia_mode;
4127         int status = -ENOMEM;
4128
4129         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4130         if (data == NULL)
4131                 goto out;
4132
4133         if (S_ISFIFO(mode))
4134                 data->arg.ftype = NF4FIFO;
4135         else if (S_ISBLK(mode)) {
4136                 data->arg.ftype = NF4BLK;
4137                 data->arg.u.device.specdata1 = MAJOR(rdev);
4138                 data->arg.u.device.specdata2 = MINOR(rdev);
4139         }
4140         else if (S_ISCHR(mode)) {
4141                 data->arg.ftype = NF4CHR;
4142                 data->arg.u.device.specdata1 = MAJOR(rdev);
4143                 data->arg.u.device.specdata2 = MINOR(rdev);
4144         } else if (!S_ISSOCK(mode)) {
4145                 status = -EINVAL;
4146                 goto out_free;
4147         }
4148
4149         data->arg.label = label;
4150         status = nfs4_do_create(dir, dentry, data);
4151 out_free:
4152         nfs4_free_createdata(data);
4153 out:
4154         return status;
4155 }
4156
4157 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4158                 struct iattr *sattr, dev_t rdev)
4159 {
4160         struct nfs4_exception exception = { };
4161         struct nfs4_label l, *label = NULL;
4162         int err;
4163
4164         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4165
4166         sattr->ia_mode &= ~current_umask();
4167         do {
4168                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4169                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4170                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4171                                 &exception);
4172         } while (exception.retry);
4173
4174         nfs4_label_release_security(label);
4175
4176         return err;
4177 }
4178
4179 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4180                  struct nfs_fsstat *fsstat)
4181 {
4182         struct nfs4_statfs_arg args = {
4183                 .fh = fhandle,
4184                 .bitmask = server->attr_bitmask,
4185         };
4186         struct nfs4_statfs_res res = {
4187                 .fsstat = fsstat,
4188         };
4189         struct rpc_message msg = {
4190                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4191                 .rpc_argp = &args,
4192                 .rpc_resp = &res,
4193         };
4194
4195         nfs_fattr_init(fsstat->fattr);
4196         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4197 }
4198
4199 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4200 {
4201         struct nfs4_exception exception = { };
4202         int err;
4203         do {
4204                 err = nfs4_handle_exception(server,
4205                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4206                                 &exception);
4207         } while (exception.retry);
4208         return err;
4209 }
4210
4211 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4212                 struct nfs_fsinfo *fsinfo)
4213 {
4214         struct nfs4_fsinfo_arg args = {
4215                 .fh = fhandle,
4216                 .bitmask = server->attr_bitmask,
4217         };
4218         struct nfs4_fsinfo_res res = {
4219                 .fsinfo = fsinfo,
4220         };
4221         struct rpc_message msg = {
4222                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4223                 .rpc_argp = &args,
4224                 .rpc_resp = &res,
4225         };
4226
4227         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4228 }
4229
4230 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4231 {
4232         struct nfs4_exception exception = { };
4233         unsigned long now = jiffies;
4234         int err;
4235
4236         do {
4237                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4238                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4239                 if (err == 0) {
4240                         struct nfs_client *clp = server->nfs_client;
4241
4242                         spin_lock(&clp->cl_lock);
4243                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4244                         clp->cl_last_renewal = now;
4245                         spin_unlock(&clp->cl_lock);
4246                         break;
4247                 }
4248                 err = nfs4_handle_exception(server, err, &exception);
4249         } while (exception.retry);
4250         return err;
4251 }
4252
4253 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4254 {
4255         int error;
4256
4257         nfs_fattr_init(fsinfo->fattr);
4258         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4259         if (error == 0) {
4260                 /* block layout checks this! */
4261                 server->pnfs_blksize = fsinfo->blksize;
4262                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4263         }
4264
4265         return error;
4266 }
4267
4268 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4269                 struct nfs_pathconf *pathconf)
4270 {
4271         struct nfs4_pathconf_arg args = {
4272                 .fh = fhandle,
4273                 .bitmask = server->attr_bitmask,
4274         };
4275         struct nfs4_pathconf_res res = {
4276                 .pathconf = pathconf,
4277         };
4278         struct rpc_message msg = {
4279                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4280                 .rpc_argp = &args,
4281                 .rpc_resp = &res,
4282         };
4283
4284         /* None of the pathconf attributes are mandatory to implement */
4285         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4286                 memset(pathconf, 0, sizeof(*pathconf));
4287                 return 0;
4288         }
4289
4290         nfs_fattr_init(pathconf->fattr);
4291         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4292 }
4293
4294 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4295                 struct nfs_pathconf *pathconf)
4296 {
4297         struct nfs4_exception exception = { };
4298         int err;
4299
4300         do {
4301                 err = nfs4_handle_exception(server,
4302                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4303                                 &exception);
4304         } while (exception.retry);
4305         return err;
4306 }
4307
4308 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4309                 const struct nfs_open_context *ctx,
4310                 const struct nfs_lock_context *l_ctx,
4311                 fmode_t fmode)
4312 {
4313         const struct nfs_lockowner *lockowner = NULL;
4314
4315         if (l_ctx != NULL)
4316                 lockowner = &l_ctx->lockowner;
4317         return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL);
4318 }
4319 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4320
4321 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4322                 const struct nfs_open_context *ctx,
4323                 const struct nfs_lock_context *l_ctx,
4324                 fmode_t fmode)
4325 {
4326         nfs4_stateid current_stateid;
4327
4328         /* If the current stateid represents a lost lock, then exit */
4329         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4330                 return true;
4331         return nfs4_stateid_match(stateid, &current_stateid);
4332 }
4333
4334 static bool nfs4_error_stateid_expired(int err)
4335 {
4336         switch (err) {
4337         case -NFS4ERR_DELEG_REVOKED:
4338         case -NFS4ERR_ADMIN_REVOKED:
4339         case -NFS4ERR_BAD_STATEID:
4340         case -NFS4ERR_STALE_STATEID:
4341         case -NFS4ERR_OLD_STATEID:
4342         case -NFS4ERR_OPENMODE:
4343         case -NFS4ERR_EXPIRED:
4344                 return true;
4345         }
4346         return false;
4347 }
4348
4349 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4350 {
4351         nfs_invalidate_atime(hdr->inode);
4352 }
4353
4354 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4355 {
4356         struct nfs_server *server = NFS_SERVER(hdr->inode);
4357
4358         trace_nfs4_read(hdr, task->tk_status);
4359         if (nfs4_async_handle_error(task, server,
4360                                     hdr->args.context->state,
4361                                     NULL) == -EAGAIN) {
4362                 rpc_restart_call_prepare(task);
4363                 return -EAGAIN;
4364         }
4365
4366         __nfs4_read_done_cb(hdr);
4367         if (task->tk_status > 0)
4368                 renew_lease(server, hdr->timestamp);
4369         return 0;
4370 }
4371
4372 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4373                 struct nfs_pgio_args *args)
4374 {
4375
4376         if (!nfs4_error_stateid_expired(task->tk_status) ||
4377                 nfs4_stateid_is_current(&args->stateid,
4378                                 args->context,
4379                                 args->lock_context,
4380                                 FMODE_READ))
4381                 return false;
4382         rpc_restart_call_prepare(task);
4383         return true;
4384 }
4385
4386 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4387 {
4388
4389         dprintk("--> %s\n", __func__);
4390
4391         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4392                 return -EAGAIN;
4393         if (nfs4_read_stateid_changed(task, &hdr->args))
4394                 return -EAGAIN;
4395         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4396                                     nfs4_read_done_cb(task, hdr);
4397 }
4398
4399 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4400                                  struct rpc_message *msg)
4401 {
4402         hdr->timestamp   = jiffies;
4403         if (!hdr->pgio_done_cb)
4404                 hdr->pgio_done_cb = nfs4_read_done_cb;
4405         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4406         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4407 }
4408
4409 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4410                                       struct nfs_pgio_header *hdr)
4411 {
4412         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4413                         &hdr->args.seq_args,
4414                         &hdr->res.seq_res,
4415                         task))
4416                 return 0;
4417         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4418                                 hdr->args.lock_context,
4419                                 hdr->rw_ops->rw_mode) == -EIO)
4420                 return -EIO;
4421         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4422                 return -EIO;
4423         return 0;
4424 }
4425
4426 static int nfs4_write_done_cb(struct rpc_task *task,
4427                               struct nfs_pgio_header *hdr)
4428 {
4429         struct inode *inode = hdr->inode;
4430
4431         trace_nfs4_write(hdr, task->tk_status);
4432         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4433                                     hdr->args.context->state,
4434                                     NULL) == -EAGAIN) {
4435                 rpc_restart_call_prepare(task);
4436                 return -EAGAIN;
4437         }
4438         if (task->tk_status >= 0) {
4439                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4440                 nfs_writeback_update_inode(hdr);
4441         }
4442         return 0;
4443 }
4444
4445 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4446                 struct nfs_pgio_args *args)
4447 {
4448
4449         if (!nfs4_error_stateid_expired(task->tk_status) ||
4450                 nfs4_stateid_is_current(&args->stateid,
4451                                 args->context,
4452                                 args->lock_context,
4453                                 FMODE_WRITE))
4454                 return false;
4455         rpc_restart_call_prepare(task);
4456         return true;
4457 }
4458
4459 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4460 {
4461         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4462                 return -EAGAIN;
4463         if (nfs4_write_stateid_changed(task, &hdr->args))
4464                 return -EAGAIN;
4465         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4466                 nfs4_write_done_cb(task, hdr);
4467 }
4468
4469 static
4470 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4471 {
4472         /* Don't request attributes for pNFS or O_DIRECT writes */
4473         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4474                 return false;
4475         /* Otherwise, request attributes if and only if we don't hold
4476          * a delegation
4477          */
4478         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4479 }
4480
4481 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4482                                   struct rpc_message *msg)
4483 {
4484         struct nfs_server *server = NFS_SERVER(hdr->inode);
4485
4486         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4487                 hdr->args.bitmask = NULL;
4488                 hdr->res.fattr = NULL;
4489         } else
4490                 hdr->args.bitmask = server->cache_consistency_bitmask;
4491
4492         if (!hdr->pgio_done_cb)
4493                 hdr->pgio_done_cb = nfs4_write_done_cb;
4494         hdr->res.server = server;
4495         hdr->timestamp   = jiffies;
4496
4497         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4498         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4499 }
4500
4501 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4502 {
4503         nfs4_setup_sequence(NFS_SERVER(data->inode),
4504                         &data->args.seq_args,
4505                         &data->res.seq_res,
4506                         task);
4507 }
4508
4509 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4510 {
4511         struct inode *inode = data->inode;
4512
4513         trace_nfs4_commit(data, task->tk_status);
4514         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4515                                     NULL, NULL) == -EAGAIN) {
4516                 rpc_restart_call_prepare(task);
4517                 return -EAGAIN;
4518         }
4519         return 0;
4520 }
4521
4522 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4523 {
4524         if (!nfs4_sequence_done(task, &data->res.seq_res))
4525                 return -EAGAIN;
4526         return data->commit_done_cb(task, data);
4527 }
4528
4529 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4530 {
4531         struct nfs_server *server = NFS_SERVER(data->inode);
4532
4533         if (data->commit_done_cb == NULL)
4534                 data->commit_done_cb = nfs4_commit_done_cb;
4535         data->res.server = server;
4536         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4537         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4538 }
4539
4540 struct nfs4_renewdata {
4541         struct nfs_client       *client;
4542         unsigned long           timestamp;
4543 };
4544
4545 /*
4546  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4547  * standalone procedure for queueing an asynchronous RENEW.
4548  */
4549 static void nfs4_renew_release(void *calldata)
4550 {
4551         struct nfs4_renewdata *data = calldata;
4552         struct nfs_client *clp = data->client;
4553
4554         if (atomic_read(&clp->cl_count) > 1)
4555                 nfs4_schedule_state_renewal(clp);
4556         nfs_put_client(clp);
4557         kfree(data);
4558 }
4559
4560 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4561 {
4562         struct nfs4_renewdata *data = calldata;
4563         struct nfs_client *clp = data->client;
4564         unsigned long timestamp = data->timestamp;
4565
4566         trace_nfs4_renew_async(clp, task->tk_status);
4567         switch (task->tk_status) {
4568         case 0:
4569                 break;
4570         case -NFS4ERR_LEASE_MOVED:
4571                 nfs4_schedule_lease_moved_recovery(clp);
4572                 break;
4573         default:
4574                 /* Unless we're shutting down, schedule state recovery! */
4575                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4576                         return;
4577                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4578                         nfs4_schedule_lease_recovery(clp);
4579                         return;
4580                 }
4581                 nfs4_schedule_path_down_recovery(clp);
4582         }
4583         do_renew_lease(clp, timestamp);
4584 }
4585
4586 static const struct rpc_call_ops nfs4_renew_ops = {
4587         .rpc_call_done = nfs4_renew_done,
4588         .rpc_release = nfs4_renew_release,
4589 };
4590
4591 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4592 {
4593         struct rpc_message msg = {
4594                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4595                 .rpc_argp       = clp,
4596                 .rpc_cred       = cred,
4597         };
4598         struct nfs4_renewdata *data;
4599
4600         if (renew_flags == 0)
4601                 return 0;
4602         if (!atomic_inc_not_zero(&clp->cl_count))
4603                 return -EIO;
4604         data = kmalloc(sizeof(*data), GFP_NOFS);
4605         if (data == NULL)
4606                 return -ENOMEM;
4607         data->client = clp;
4608         data->timestamp = jiffies;
4609         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4610                         &nfs4_renew_ops, data);
4611 }
4612
4613 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4614 {
4615         struct rpc_message msg = {
4616                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4617                 .rpc_argp       = clp,
4618                 .rpc_cred       = cred,
4619         };
4620         unsigned long now = jiffies;
4621         int status;
4622
4623         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4624         if (status < 0)
4625                 return status;
4626         do_renew_lease(clp, now);
4627         return 0;
4628 }
4629
4630 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4631 {
4632         return server->caps & NFS_CAP_ACLS;
4633 }
4634
4635 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4636  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4637  * the stack.
4638  */
4639 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4640
4641 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4642                 struct page **pages)
4643 {
4644         struct page *newpage, **spages;
4645         int rc = 0;
4646         size_t len;
4647         spages = pages;
4648
4649         do {
4650                 len = min_t(size_t, PAGE_SIZE, buflen);
4651                 newpage = alloc_page(GFP_KERNEL);
4652
4653                 if (newpage == NULL)
4654                         goto unwind;
4655                 memcpy(page_address(newpage), buf, len);
4656                 buf += len;
4657                 buflen -= len;
4658                 *pages++ = newpage;
4659                 rc++;
4660         } while (buflen != 0);
4661
4662         return rc;
4663
4664 unwind:
4665         for(; rc > 0; rc--)
4666                 __free_page(spages[rc-1]);
4667         return -ENOMEM;
4668 }
4669
4670 struct nfs4_cached_acl {
4671         int cached;
4672         size_t len;
4673         char data[0];
4674 };
4675
4676 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4677 {
4678         struct nfs_inode *nfsi = NFS_I(inode);
4679
4680         spin_lock(&inode->i_lock);
4681         kfree(nfsi->nfs4_acl);
4682         nfsi->nfs4_acl = acl;
4683         spin_unlock(&inode->i_lock);
4684 }
4685
4686 static void nfs4_zap_acl_attr(struct inode *inode)
4687 {
4688         nfs4_set_cached_acl(inode, NULL);
4689 }
4690
4691 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4692 {
4693         struct nfs_inode *nfsi = NFS_I(inode);
4694         struct nfs4_cached_acl *acl;
4695         int ret = -ENOENT;
4696
4697         spin_lock(&inode->i_lock);
4698         acl = nfsi->nfs4_acl;
4699         if (acl == NULL)
4700                 goto out;
4701         if (buf == NULL) /* user is just asking for length */
4702                 goto out_len;
4703         if (acl->cached == 0)
4704                 goto out;
4705         ret = -ERANGE; /* see getxattr(2) man page */
4706         if (acl->len > buflen)
4707                 goto out;
4708         memcpy(buf, acl->data, acl->len);
4709 out_len:
4710         ret = acl->len;
4711 out:
4712         spin_unlock(&inode->i_lock);
4713         return ret;
4714 }
4715
4716 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4717 {
4718         struct nfs4_cached_acl *acl;
4719         size_t buflen = sizeof(*acl) + acl_len;
4720
4721         if (buflen <= PAGE_SIZE) {
4722                 acl = kmalloc(buflen, GFP_KERNEL);
4723                 if (acl == NULL)
4724                         goto out;
4725                 acl->cached = 1;
4726                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4727         } else {
4728                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4729                 if (acl == NULL)
4730                         goto out;
4731                 acl->cached = 0;
4732         }
4733         acl->len = acl_len;
4734 out:
4735         nfs4_set_cached_acl(inode, acl);
4736 }
4737
4738 /*
4739  * The getxattr API returns the required buffer length when called with a
4740  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4741  * the required buf.  On a NULL buf, we send a page of data to the server
4742  * guessing that the ACL request can be serviced by a page. If so, we cache
4743  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4744  * the cache. If not so, we throw away the page, and cache the required
4745  * length. The next getxattr call will then produce another round trip to
4746  * the server, this time with the input buf of the required size.
4747  */
4748 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4749 {
4750         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4751         struct nfs_getaclargs args = {
4752                 .fh = NFS_FH(inode),
4753                 .acl_pages = pages,
4754                 .acl_len = buflen,
4755         };
4756         struct nfs_getaclres res = {
4757                 .acl_len = buflen,
4758         };
4759         struct rpc_message msg = {
4760                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4761                 .rpc_argp = &args,
4762                 .rpc_resp = &res,
4763         };
4764         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4765         int ret = -ENOMEM, i;
4766
4767         /* As long as we're doing a round trip to the server anyway,
4768          * let's be prepared for a page of acl data. */
4769         if (npages == 0)
4770                 npages = 1;
4771         if (npages > ARRAY_SIZE(pages))
4772                 return -ERANGE;
4773
4774         for (i = 0; i < npages; i++) {
4775                 pages[i] = alloc_page(GFP_KERNEL);
4776                 if (!pages[i])
4777                         goto out_free;
4778         }
4779
4780         /* for decoding across pages */
4781         res.acl_scratch = alloc_page(GFP_KERNEL);
4782         if (!res.acl_scratch)
4783                 goto out_free;
4784
4785         args.acl_len = npages * PAGE_SIZE;
4786
4787         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4788                 __func__, buf, buflen, npages, args.acl_len);
4789         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4790                              &msg, &args.seq_args, &res.seq_res, 0);
4791         if (ret)
4792                 goto out_free;
4793
4794         /* Handle the case where the passed-in buffer is too short */
4795         if (res.acl_flags & NFS4_ACL_TRUNC) {
4796                 /* Did the user only issue a request for the acl length? */
4797                 if (buf == NULL)
4798                         goto out_ok;
4799                 ret = -ERANGE;
4800                 goto out_free;
4801         }
4802         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4803         if (buf) {
4804                 if (res.acl_len > buflen) {
4805                         ret = -ERANGE;
4806                         goto out_free;
4807                 }
4808                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4809         }
4810 out_ok:
4811         ret = res.acl_len;
4812 out_free:
4813         for (i = 0; i < npages; i++)
4814                 if (pages[i])
4815                         __free_page(pages[i]);
4816         if (res.acl_scratch)
4817                 __free_page(res.acl_scratch);
4818         return ret;
4819 }
4820
4821 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4822 {
4823         struct nfs4_exception exception = { };
4824         ssize_t ret;
4825         do {
4826                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4827                 trace_nfs4_get_acl(inode, ret);
4828                 if (ret >= 0)
4829                         break;
4830                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4831         } while (exception.retry);
4832         return ret;
4833 }
4834
4835 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4836 {
4837         struct nfs_server *server = NFS_SERVER(inode);
4838         int ret;
4839
4840         if (!nfs4_server_supports_acls(server))
4841                 return -EOPNOTSUPP;
4842         ret = nfs_revalidate_inode(server, inode);
4843         if (ret < 0)
4844                 return ret;
4845         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4846                 nfs_zap_acl_cache(inode);
4847         ret = nfs4_read_cached_acl(inode, buf, buflen);
4848         if (ret != -ENOENT)
4849                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4850                  * but no cached acl data, just the acl length */
4851                 return ret;
4852         return nfs4_get_acl_uncached(inode, buf, buflen);
4853 }
4854
4855 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4856 {
4857         struct nfs_server *server = NFS_SERVER(inode);
4858         struct page *pages[NFS4ACL_MAXPAGES];
4859         struct nfs_setaclargs arg = {
4860                 .fh             = NFS_FH(inode),
4861                 .acl_pages      = pages,
4862                 .acl_len        = buflen,
4863         };
4864         struct nfs_setaclres res;
4865         struct rpc_message msg = {
4866                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4867                 .rpc_argp       = &arg,
4868                 .rpc_resp       = &res,
4869         };
4870         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4871         int ret, i;
4872
4873         if (!nfs4_server_supports_acls(server))
4874                 return -EOPNOTSUPP;
4875         if (npages > ARRAY_SIZE(pages))
4876                 return -ERANGE;
4877         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4878         if (i < 0)
4879                 return i;
4880         nfs4_inode_return_delegation(inode);
4881         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4882
4883         /*
4884          * Free each page after tx, so the only ref left is
4885          * held by the network stack
4886          */
4887         for (; i > 0; i--)
4888                 put_page(pages[i-1]);
4889
4890         /*
4891          * Acl update can result in inode attribute update.
4892          * so mark the attribute cache invalid.
4893          */
4894         spin_lock(&inode->i_lock);
4895         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4896         spin_unlock(&inode->i_lock);
4897         nfs_access_zap_cache(inode);
4898         nfs_zap_acl_cache(inode);
4899         return ret;
4900 }
4901
4902 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4903 {
4904         struct nfs4_exception exception = { };
4905         int err;
4906         do {
4907                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4908                 trace_nfs4_set_acl(inode, err);
4909                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4910                                 &exception);
4911         } while (exception.retry);
4912         return err;
4913 }
4914
4915 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4916 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4917                                         size_t buflen)
4918 {
4919         struct nfs_server *server = NFS_SERVER(inode);
4920         struct nfs_fattr fattr;
4921         struct nfs4_label label = {0, 0, buflen, buf};
4922
4923         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4924         struct nfs4_getattr_arg arg = {
4925                 .fh             = NFS_FH(inode),
4926                 .bitmask        = bitmask,
4927         };
4928         struct nfs4_getattr_res res = {
4929                 .fattr          = &fattr,
4930                 .label          = &label,
4931                 .server         = server,
4932         };
4933         struct rpc_message msg = {
4934                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4935                 .rpc_argp       = &arg,
4936                 .rpc_resp       = &res,
4937         };
4938         int ret;
4939
4940         nfs_fattr_init(&fattr);
4941
4942         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4943         if (ret)
4944                 return ret;
4945         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4946                 return -ENOENT;
4947         if (buflen < label.len)
4948                 return -ERANGE;
4949         return 0;
4950 }
4951
4952 static int nfs4_get_security_label(struct inode *inode, void *buf,
4953                                         size_t buflen)
4954 {
4955         struct nfs4_exception exception = { };
4956         int err;
4957
4958         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4959                 return -EOPNOTSUPP;
4960
4961         do {
4962                 err = _nfs4_get_security_label(inode, buf, buflen);
4963                 trace_nfs4_get_security_label(inode, err);
4964                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4965                                 &exception);
4966         } while (exception.retry);
4967         return err;
4968 }
4969
4970 static int _nfs4_do_set_security_label(struct inode *inode,
4971                 struct nfs4_label *ilabel,
4972                 struct nfs_fattr *fattr,
4973                 struct nfs4_label *olabel)
4974 {
4975
4976         struct iattr sattr = {0};
4977         struct nfs_server *server = NFS_SERVER(inode);
4978         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4979         struct nfs_setattrargs arg = {
4980                 .fh             = NFS_FH(inode),
4981                 .iap            = &sattr,
4982                 .server         = server,
4983                 .bitmask        = bitmask,
4984                 .label          = ilabel,
4985         };
4986         struct nfs_setattrres res = {
4987                 .fattr          = fattr,
4988                 .label          = olabel,
4989                 .server         = server,
4990         };
4991         struct rpc_message msg = {
4992                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4993                 .rpc_argp       = &arg,
4994                 .rpc_resp       = &res,
4995         };
4996         int status;
4997
4998         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4999
5000         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5001         if (status)
5002                 dprintk("%s failed: %d\n", __func__, status);
5003
5004         return status;
5005 }
5006
5007 static int nfs4_do_set_security_label(struct inode *inode,
5008                 struct nfs4_label *ilabel,
5009                 struct nfs_fattr *fattr,
5010                 struct nfs4_label *olabel)
5011 {
5012         struct nfs4_exception exception = { };
5013         int err;
5014
5015         do {
5016                 err = _nfs4_do_set_security_label(inode, ilabel,
5017                                 fattr, olabel);
5018                 trace_nfs4_set_security_label(inode, err);
5019                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5020                                 &exception);
5021         } while (exception.retry);
5022         return err;
5023 }
5024
5025 static int
5026 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5027 {
5028         struct nfs4_label ilabel, *olabel = NULL;
5029         struct nfs_fattr fattr;
5030         struct rpc_cred *cred;
5031         int status;
5032
5033         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5034                 return -EOPNOTSUPP;
5035
5036         nfs_fattr_init(&fattr);
5037
5038         ilabel.pi = 0;
5039         ilabel.lfs = 0;
5040         ilabel.label = (char *)buf;
5041         ilabel.len = buflen;
5042
5043         cred = rpc_lookup_cred();
5044         if (IS_ERR(cred))
5045                 return PTR_ERR(cred);
5046
5047         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5048         if (IS_ERR(olabel)) {
5049                 status = -PTR_ERR(olabel);
5050                 goto out;
5051         }
5052
5053         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5054         if (status == 0)
5055                 nfs_setsecurity(inode, &fattr, olabel);
5056
5057         nfs4_label_free(olabel);
5058 out:
5059         put_rpccred(cred);
5060         return status;
5061 }
5062 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5063
5064
5065 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5066                                     nfs4_verifier *bootverf)
5067 {
5068         __be32 verf[2];
5069
5070         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5071                 /* An impossible timestamp guarantees this value
5072                  * will never match a generated boot time. */
5073                 verf[0] = 0;
5074                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5075         } else {
5076                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5077                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5078                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5079         }
5080         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5081 }
5082
5083 static int
5084 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5085 {
5086         size_t len;
5087         char *str;
5088
5089         if (clp->cl_owner_id != NULL)
5090                 return 0;
5091
5092         rcu_read_lock();
5093         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5094                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5095                 1 +
5096                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5097                 1;
5098         rcu_read_unlock();
5099
5100         if (len > NFS4_OPAQUE_LIMIT + 1)
5101                 return -EINVAL;
5102
5103         /*
5104          * Since this string is allocated at mount time, and held until the
5105          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5106          * about a memory-reclaim deadlock.
5107          */
5108         str = kmalloc(len, GFP_KERNEL);
5109         if (!str)
5110                 return -ENOMEM;
5111
5112         rcu_read_lock();
5113         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5114                         clp->cl_ipaddr,
5115                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5116                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5117         rcu_read_unlock();
5118
5119         clp->cl_owner_id = str;
5120         return 0;
5121 }
5122
5123 static int
5124 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5125 {
5126         size_t len;
5127         char *str;
5128
5129         len = 10 + 10 + 1 + 10 + 1 +
5130                 strlen(nfs4_client_id_uniquifier) + 1 +
5131                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5132
5133         if (len > NFS4_OPAQUE_LIMIT + 1)
5134                 return -EINVAL;
5135
5136         /*
5137          * Since this string is allocated at mount time, and held until the
5138          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5139          * about a memory-reclaim deadlock.
5140          */
5141         str = kmalloc(len, GFP_KERNEL);
5142         if (!str)
5143                 return -ENOMEM;
5144
5145         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5146                         clp->rpc_ops->version, clp->cl_minorversion,
5147                         nfs4_client_id_uniquifier,
5148                         clp->cl_rpcclient->cl_nodename);
5149         clp->cl_owner_id = str;
5150         return 0;
5151 }
5152
5153 static int
5154 nfs4_init_uniform_client_string(struct nfs_client *clp)
5155 {
5156         size_t len;
5157         char *str;
5158
5159         if (clp->cl_owner_id != NULL)
5160                 return 0;
5161
5162         if (nfs4_client_id_uniquifier[0] != '\0')
5163                 return nfs4_init_uniquifier_client_string(clp);
5164
5165         len = 10 + 10 + 1 + 10 + 1 +
5166                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5167
5168         if (len > NFS4_OPAQUE_LIMIT + 1)
5169                 return -EINVAL;
5170
5171         /*
5172          * Since this string is allocated at mount time, and held until the
5173          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5174          * about a memory-reclaim deadlock.
5175          */
5176         str = kmalloc(len, GFP_KERNEL);
5177         if (!str)
5178                 return -ENOMEM;
5179
5180         scnprintf(str, len, "Linux NFSv%u.%u %s",
5181                         clp->rpc_ops->version, clp->cl_minorversion,
5182                         clp->cl_rpcclient->cl_nodename);
5183         clp->cl_owner_id = str;
5184         return 0;
5185 }
5186
5187 /*
5188  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5189  * services.  Advertise one based on the address family of the
5190  * clientaddr.
5191  */
5192 static unsigned int
5193 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5194 {
5195         if (strchr(clp->cl_ipaddr, ':') != NULL)
5196                 return scnprintf(buf, len, "tcp6");
5197         else
5198                 return scnprintf(buf, len, "tcp");
5199 }
5200
5201 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5202 {
5203         struct nfs4_setclientid *sc = calldata;
5204
5205         if (task->tk_status == 0)
5206                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5207 }
5208
5209 static const struct rpc_call_ops nfs4_setclientid_ops = {
5210         .rpc_call_done = nfs4_setclientid_done,
5211 };
5212
5213 /**
5214  * nfs4_proc_setclientid - Negotiate client ID
5215  * @clp: state data structure
5216  * @program: RPC program for NFSv4 callback service
5217  * @port: IP port number for NFS4 callback service
5218  * @cred: RPC credential to use for this call
5219  * @res: where to place the result
5220  *
5221  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5222  */
5223 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5224                 unsigned short port, struct rpc_cred *cred,
5225                 struct nfs4_setclientid_res *res)
5226 {
5227         nfs4_verifier sc_verifier;
5228         struct nfs4_setclientid setclientid = {
5229                 .sc_verifier = &sc_verifier,
5230                 .sc_prog = program,
5231                 .sc_clnt = clp,
5232         };
5233         struct rpc_message msg = {
5234                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5235                 .rpc_argp = &setclientid,
5236                 .rpc_resp = res,
5237                 .rpc_cred = cred,
5238         };
5239         struct rpc_task *task;
5240         struct rpc_task_setup task_setup_data = {
5241                 .rpc_client = clp->cl_rpcclient,
5242                 .rpc_message = &msg,
5243                 .callback_ops = &nfs4_setclientid_ops,
5244                 .callback_data = &setclientid,
5245                 .flags = RPC_TASK_TIMEOUT,
5246         };
5247         int status;
5248
5249         /* nfs_client_id4 */
5250         nfs4_init_boot_verifier(clp, &sc_verifier);
5251
5252         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5253                 status = nfs4_init_uniform_client_string(clp);
5254         else
5255                 status = nfs4_init_nonuniform_client_string(clp);
5256
5257         if (status)
5258                 goto out;
5259
5260         /* cb_client4 */
5261         setclientid.sc_netid_len =
5262                                 nfs4_init_callback_netid(clp,
5263                                                 setclientid.sc_netid,
5264                                                 sizeof(setclientid.sc_netid));
5265         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5266                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5267                                 clp->cl_ipaddr, port >> 8, port & 255);
5268
5269         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5270                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5271                 clp->cl_owner_id);
5272         task = rpc_run_task(&task_setup_data);
5273         if (IS_ERR(task)) {
5274                 status = PTR_ERR(task);
5275                 goto out;
5276         }
5277         status = task->tk_status;
5278         if (setclientid.sc_cred) {
5279                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5280                 put_rpccred(setclientid.sc_cred);
5281         }
5282         rpc_put_task(task);
5283 out:
5284         trace_nfs4_setclientid(clp, status);
5285         dprintk("NFS reply setclientid: %d\n", status);
5286         return status;
5287 }
5288
5289 /**
5290  * nfs4_proc_setclientid_confirm - Confirm client ID
5291  * @clp: state data structure
5292  * @res: result of a previous SETCLIENTID
5293  * @cred: RPC credential to use for this call
5294  *
5295  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5296  */
5297 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5298                 struct nfs4_setclientid_res *arg,
5299                 struct rpc_cred *cred)
5300 {
5301         struct rpc_message msg = {
5302                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5303                 .rpc_argp = arg,
5304                 .rpc_cred = cred,
5305         };
5306         int status;
5307
5308         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5309                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5310                 clp->cl_clientid);
5311         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5312         trace_nfs4_setclientid_confirm(clp, status);
5313         dprintk("NFS reply setclientid_confirm: %d\n", status);
5314         return status;
5315 }
5316
5317 struct nfs4_delegreturndata {
5318         struct nfs4_delegreturnargs args;
5319         struct nfs4_delegreturnres res;
5320         struct nfs_fh fh;
5321         nfs4_stateid stateid;
5322         unsigned long timestamp;
5323         struct nfs_fattr fattr;
5324         int rpc_status;
5325         struct inode *inode;
5326         bool roc;
5327         u32 roc_barrier;
5328 };
5329
5330 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5331 {
5332         struct nfs4_delegreturndata *data = calldata;
5333
5334         if (!nfs4_sequence_done(task, &data->res.seq_res))
5335                 return;
5336
5337         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5338         switch (task->tk_status) {
5339         case 0:
5340                 renew_lease(data->res.server, data->timestamp);
5341         case -NFS4ERR_ADMIN_REVOKED:
5342         case -NFS4ERR_DELEG_REVOKED:
5343         case -NFS4ERR_BAD_STATEID:
5344         case -NFS4ERR_OLD_STATEID:
5345         case -NFS4ERR_STALE_STATEID:
5346         case -NFS4ERR_EXPIRED:
5347                 task->tk_status = 0;
5348                 if (data->roc)
5349                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5350                 break;
5351         default:
5352                 if (nfs4_async_handle_error(task, data->res.server,
5353                                             NULL, NULL) == -EAGAIN) {
5354                         rpc_restart_call_prepare(task);
5355                         return;
5356                 }
5357         }
5358         data->rpc_status = task->tk_status;
5359 }
5360
5361 static void nfs4_delegreturn_release(void *calldata)
5362 {
5363         struct nfs4_delegreturndata *data = calldata;
5364         struct inode *inode = data->inode;
5365
5366         if (inode) {
5367                 if (data->roc)
5368                         pnfs_roc_release(inode);
5369                 nfs_iput_and_deactive(inode);
5370         }
5371         kfree(calldata);
5372 }
5373
5374 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5375 {
5376         struct nfs4_delegreturndata *d_data;
5377
5378         d_data = (struct nfs4_delegreturndata *)data;
5379
5380         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5381                 return;
5382
5383         if (d_data->roc)
5384                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5385
5386         nfs4_setup_sequence(d_data->res.server,
5387                         &d_data->args.seq_args,
5388                         &d_data->res.seq_res,
5389                         task);
5390 }
5391
5392 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5393         .rpc_call_prepare = nfs4_delegreturn_prepare,
5394         .rpc_call_done = nfs4_delegreturn_done,
5395         .rpc_release = nfs4_delegreturn_release,
5396 };
5397
5398 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5399 {
5400         struct nfs4_delegreturndata *data;
5401         struct nfs_server *server = NFS_SERVER(inode);
5402         struct rpc_task *task;
5403         struct rpc_message msg = {
5404                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5405                 .rpc_cred = cred,
5406         };
5407         struct rpc_task_setup task_setup_data = {
5408                 .rpc_client = server->client,
5409                 .rpc_message = &msg,
5410                 .callback_ops = &nfs4_delegreturn_ops,
5411                 .flags = RPC_TASK_ASYNC,
5412         };
5413         int status = 0;
5414
5415         data = kzalloc(sizeof(*data), GFP_NOFS);
5416         if (data == NULL)
5417                 return -ENOMEM;
5418         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5419
5420         nfs4_state_protect(server->nfs_client,
5421                         NFS_SP4_MACH_CRED_CLEANUP,
5422                         &task_setup_data.rpc_client, &msg);
5423
5424         data->args.fhandle = &data->fh;
5425         data->args.stateid = &data->stateid;
5426         data->args.bitmask = server->cache_consistency_bitmask;
5427         nfs_copy_fh(&data->fh, NFS_FH(inode));
5428         nfs4_stateid_copy(&data->stateid, stateid);
5429         data->res.fattr = &data->fattr;
5430         data->res.server = server;
5431         nfs_fattr_init(data->res.fattr);
5432         data->timestamp = jiffies;
5433         data->rpc_status = 0;
5434         data->inode = nfs_igrab_and_active(inode);
5435         if (data->inode)
5436                 data->roc = nfs4_roc(inode);
5437
5438         task_setup_data.callback_data = data;
5439         msg.rpc_argp = &data->args;
5440         msg.rpc_resp = &data->res;
5441         task = rpc_run_task(&task_setup_data);
5442         if (IS_ERR(task))
5443                 return PTR_ERR(task);
5444         if (!issync)
5445                 goto out;
5446         status = nfs4_wait_for_completion_rpc_task(task);
5447         if (status != 0)
5448                 goto out;
5449         status = data->rpc_status;
5450         if (status == 0)
5451                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5452         else
5453                 nfs_refresh_inode(inode, &data->fattr);
5454 out:
5455         rpc_put_task(task);
5456         return status;
5457 }
5458
5459 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5460 {
5461         struct nfs_server *server = NFS_SERVER(inode);
5462         struct nfs4_exception exception = { };
5463         int err;
5464         do {
5465                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5466                 trace_nfs4_delegreturn(inode, stateid, err);
5467                 switch (err) {
5468                         case -NFS4ERR_STALE_STATEID:
5469                         case -NFS4ERR_EXPIRED:
5470                         case 0:
5471                                 return 0;
5472                 }
5473                 err = nfs4_handle_exception(server, err, &exception);
5474         } while (exception.retry);
5475         return err;
5476 }
5477
5478 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5479 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5480
5481 /* 
5482  * sleep, with exponential backoff, and retry the LOCK operation. 
5483  */
5484 static unsigned long
5485 nfs4_set_lock_task_retry(unsigned long timeout)
5486 {
5487         freezable_schedule_timeout_killable_unsafe(timeout);
5488         timeout <<= 1;
5489         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5490                 return NFS4_LOCK_MAXTIMEOUT;
5491         return timeout;
5492 }
5493
5494 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5495 {
5496         struct inode *inode = state->inode;
5497         struct nfs_server *server = NFS_SERVER(inode);
5498         struct nfs_client *clp = server->nfs_client;
5499         struct nfs_lockt_args arg = {
5500                 .fh = NFS_FH(inode),
5501                 .fl = request,
5502         };
5503         struct nfs_lockt_res res = {
5504                 .denied = request,
5505         };
5506         struct rpc_message msg = {
5507                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5508                 .rpc_argp       = &arg,
5509                 .rpc_resp       = &res,
5510                 .rpc_cred       = state->owner->so_cred,
5511         };
5512         struct nfs4_lock_state *lsp;
5513         int status;
5514
5515         arg.lock_owner.clientid = clp->cl_clientid;
5516         status = nfs4_set_lock_state(state, request);
5517         if (status != 0)
5518                 goto out;
5519         lsp = request->fl_u.nfs4_fl.owner;
5520         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5521         arg.lock_owner.s_dev = server->s_dev;
5522         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5523         switch (status) {
5524                 case 0:
5525                         request->fl_type = F_UNLCK;
5526                         break;
5527                 case -NFS4ERR_DENIED:
5528                         status = 0;
5529         }
5530         request->fl_ops->fl_release_private(request);
5531         request->fl_ops = NULL;
5532 out:
5533         return status;
5534 }
5535
5536 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5537 {
5538         struct nfs4_exception exception = { };
5539         int err;
5540
5541         do {
5542                 err = _nfs4_proc_getlk(state, cmd, request);
5543                 trace_nfs4_get_lock(request, state, cmd, err);
5544                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5545                                 &exception);
5546         } while (exception.retry);
5547         return err;
5548 }
5549
5550 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5551 {
5552         return locks_lock_inode_wait(inode, fl);
5553 }
5554
5555 struct nfs4_unlockdata {
5556         struct nfs_locku_args arg;
5557         struct nfs_locku_res res;
5558         struct nfs4_lock_state *lsp;
5559         struct nfs_open_context *ctx;
5560         struct file_lock fl;
5561         struct nfs_server *server;
5562         unsigned long timestamp;
5563 };
5564
5565 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5566                 struct nfs_open_context *ctx,
5567                 struct nfs4_lock_state *lsp,
5568                 struct nfs_seqid *seqid)
5569 {
5570         struct nfs4_unlockdata *p;
5571         struct inode *inode = lsp->ls_state->inode;
5572
5573         p = kzalloc(sizeof(*p), GFP_NOFS);
5574         if (p == NULL)
5575                 return NULL;
5576         p->arg.fh = NFS_FH(inode);
5577         p->arg.fl = &p->fl;
5578         p->arg.seqid = seqid;
5579         p->res.seqid = seqid;
5580         p->lsp = lsp;
5581         atomic_inc(&lsp->ls_count);
5582         /* Ensure we don't close file until we're done freeing locks! */
5583         p->ctx = get_nfs_open_context(ctx);
5584         memcpy(&p->fl, fl, sizeof(p->fl));
5585         p->server = NFS_SERVER(inode);
5586         return p;
5587 }
5588
5589 static void nfs4_locku_release_calldata(void *data)
5590 {
5591         struct nfs4_unlockdata *calldata = data;
5592         nfs_free_seqid(calldata->arg.seqid);
5593         nfs4_put_lock_state(calldata->lsp);
5594         put_nfs_open_context(calldata->ctx);
5595         kfree(calldata);
5596 }
5597
5598 static void nfs4_locku_done(struct rpc_task *task, void *data)
5599 {
5600         struct nfs4_unlockdata *calldata = data;
5601
5602         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5603                 return;
5604         switch (task->tk_status) {
5605                 case 0:
5606                         renew_lease(calldata->server, calldata->timestamp);
5607                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5608                         if (nfs4_update_lock_stateid(calldata->lsp,
5609                                         &calldata->res.stateid))
5610                                 break;
5611                 case -NFS4ERR_BAD_STATEID:
5612                 case -NFS4ERR_OLD_STATEID:
5613                 case -NFS4ERR_STALE_STATEID:
5614                 case -NFS4ERR_EXPIRED:
5615                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5616                                                 &calldata->lsp->ls_stateid))
5617                                 rpc_restart_call_prepare(task);
5618                         break;
5619                 default:
5620                         if (nfs4_async_handle_error(task, calldata->server,
5621                                                     NULL, NULL) == -EAGAIN)
5622                                 rpc_restart_call_prepare(task);
5623         }
5624         nfs_release_seqid(calldata->arg.seqid);
5625 }
5626
5627 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5628 {
5629         struct nfs4_unlockdata *calldata = data;
5630
5631         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5632                 goto out_wait;
5633         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5634         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5635                 /* Note: exit _without_ running nfs4_locku_done */
5636                 goto out_no_action;
5637         }
5638         calldata->timestamp = jiffies;
5639         if (nfs4_setup_sequence(calldata->server,
5640                                 &calldata->arg.seq_args,
5641                                 &calldata->res.seq_res,
5642                                 task) != 0)
5643                 nfs_release_seqid(calldata->arg.seqid);
5644         return;
5645 out_no_action:
5646         task->tk_action = NULL;
5647 out_wait:
5648         nfs4_sequence_done(task, &calldata->res.seq_res);
5649 }
5650
5651 static const struct rpc_call_ops nfs4_locku_ops = {
5652         .rpc_call_prepare = nfs4_locku_prepare,
5653         .rpc_call_done = nfs4_locku_done,
5654         .rpc_release = nfs4_locku_release_calldata,
5655 };
5656
5657 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5658                 struct nfs_open_context *ctx,
5659                 struct nfs4_lock_state *lsp,
5660                 struct nfs_seqid *seqid)
5661 {
5662         struct nfs4_unlockdata *data;
5663         struct rpc_message msg = {
5664                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5665                 .rpc_cred = ctx->cred,
5666         };
5667         struct rpc_task_setup task_setup_data = {
5668                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5669                 .rpc_message = &msg,
5670                 .callback_ops = &nfs4_locku_ops,
5671                 .workqueue = nfsiod_workqueue,
5672                 .flags = RPC_TASK_ASYNC,
5673         };
5674
5675         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5676                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5677
5678         /* Ensure this is an unlock - when canceling a lock, the
5679          * canceled lock is passed in, and it won't be an unlock.
5680          */
5681         fl->fl_type = F_UNLCK;
5682
5683         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5684         if (data == NULL) {
5685                 nfs_free_seqid(seqid);
5686                 return ERR_PTR(-ENOMEM);
5687         }
5688
5689         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5690         msg.rpc_argp = &data->arg;
5691         msg.rpc_resp = &data->res;
5692         task_setup_data.callback_data = data;
5693         return rpc_run_task(&task_setup_data);
5694 }
5695
5696 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5697 {
5698         struct inode *inode = state->inode;
5699         struct nfs4_state_owner *sp = state->owner;
5700         struct nfs_inode *nfsi = NFS_I(inode);
5701         struct nfs_seqid *seqid;
5702         struct nfs4_lock_state *lsp;
5703         struct rpc_task *task;
5704         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5705         int status = 0;
5706         unsigned char fl_flags = request->fl_flags;
5707
5708         status = nfs4_set_lock_state(state, request);
5709         /* Unlock _before_ we do the RPC call */
5710         request->fl_flags |= FL_EXISTS;
5711         /* Exclude nfs_delegation_claim_locks() */
5712         mutex_lock(&sp->so_delegreturn_mutex);
5713         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5714         down_read(&nfsi->rwsem);
5715         if (do_vfs_lock(inode, request) == -ENOENT) {
5716                 up_read(&nfsi->rwsem);
5717                 mutex_unlock(&sp->so_delegreturn_mutex);
5718                 goto out;
5719         }
5720         up_read(&nfsi->rwsem);
5721         mutex_unlock(&sp->so_delegreturn_mutex);
5722         if (status != 0)
5723                 goto out;
5724         /* Is this a delegated lock? */
5725         lsp = request->fl_u.nfs4_fl.owner;
5726         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5727                 goto out;
5728         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5729         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5730         status = -ENOMEM;
5731         if (IS_ERR(seqid))
5732                 goto out;
5733         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5734         status = PTR_ERR(task);
5735         if (IS_ERR(task))
5736                 goto out;
5737         status = nfs4_wait_for_completion_rpc_task(task);
5738         rpc_put_task(task);
5739 out:
5740         request->fl_flags = fl_flags;
5741         trace_nfs4_unlock(request, state, F_SETLK, status);
5742         return status;
5743 }
5744
5745 struct nfs4_lockdata {
5746         struct nfs_lock_args arg;
5747         struct nfs_lock_res res;
5748         struct nfs4_lock_state *lsp;
5749         struct nfs_open_context *ctx;
5750         struct file_lock fl;
5751         unsigned long timestamp;
5752         int rpc_status;
5753         int cancelled;
5754         struct nfs_server *server;
5755 };
5756
5757 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5758                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5759                 gfp_t gfp_mask)
5760 {
5761         struct nfs4_lockdata *p;
5762         struct inode *inode = lsp->ls_state->inode;
5763         struct nfs_server *server = NFS_SERVER(inode);
5764         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5765
5766         p = kzalloc(sizeof(*p), gfp_mask);
5767         if (p == NULL)
5768                 return NULL;
5769
5770         p->arg.fh = NFS_FH(inode);
5771         p->arg.fl = &p->fl;
5772         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5773         if (IS_ERR(p->arg.open_seqid))
5774                 goto out_free;
5775         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5776         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5777         if (IS_ERR(p->arg.lock_seqid))
5778                 goto out_free_seqid;
5779         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5780         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5781         p->arg.lock_owner.s_dev = server->s_dev;
5782         p->res.lock_seqid = p->arg.lock_seqid;
5783         p->lsp = lsp;
5784         p->server = server;
5785         atomic_inc(&lsp->ls_count);
5786         p->ctx = get_nfs_open_context(ctx);
5787         get_file(fl->fl_file);
5788         memcpy(&p->fl, fl, sizeof(p->fl));
5789         return p;
5790 out_free_seqid:
5791         nfs_free_seqid(p->arg.open_seqid);
5792 out_free:
5793         kfree(p);
5794         return NULL;
5795 }
5796
5797 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5798 {
5799         struct nfs4_lockdata *data = calldata;
5800         struct nfs4_state *state = data->lsp->ls_state;
5801
5802         dprintk("%s: begin!\n", __func__);
5803         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5804                 goto out_wait;
5805         /* Do we need to do an open_to_lock_owner? */
5806         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5807                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5808                         goto out_release_lock_seqid;
5809                 }
5810                 nfs4_stateid_copy(&data->arg.open_stateid,
5811                                 &state->open_stateid);
5812                 data->arg.new_lock_owner = 1;
5813                 data->res.open_seqid = data->arg.open_seqid;
5814         } else {
5815                 data->arg.new_lock_owner = 0;
5816                 nfs4_stateid_copy(&data->arg.lock_stateid,
5817                                 &data->lsp->ls_stateid);
5818         }
5819         if (!nfs4_valid_open_stateid(state)) {
5820                 data->rpc_status = -EBADF;
5821                 task->tk_action = NULL;
5822                 goto out_release_open_seqid;
5823         }
5824         data->timestamp = jiffies;
5825         if (nfs4_setup_sequence(data->server,
5826                                 &data->arg.seq_args,
5827                                 &data->res.seq_res,
5828                                 task) == 0)
5829                 return;
5830 out_release_open_seqid:
5831         nfs_release_seqid(data->arg.open_seqid);
5832 out_release_lock_seqid:
5833         nfs_release_seqid(data->arg.lock_seqid);
5834 out_wait:
5835         nfs4_sequence_done(task, &data->res.seq_res);
5836         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5837 }
5838
5839 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5840 {
5841         struct nfs4_lockdata *data = calldata;
5842         struct nfs4_lock_state *lsp = data->lsp;
5843
5844         dprintk("%s: begin!\n", __func__);
5845
5846         if (!nfs4_sequence_done(task, &data->res.seq_res))
5847                 return;
5848
5849         data->rpc_status = task->tk_status;
5850         switch (task->tk_status) {
5851         case 0:
5852                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5853                                 data->timestamp);
5854                 if (data->arg.new_lock) {
5855                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5856                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5857                                 rpc_restart_call_prepare(task);
5858                                 break;
5859                         }
5860                 }
5861                 if (data->arg.new_lock_owner != 0) {
5862                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5863                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5864                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5865                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5866                         rpc_restart_call_prepare(task);
5867                 break;
5868         case -NFS4ERR_BAD_STATEID:
5869         case -NFS4ERR_OLD_STATEID:
5870         case -NFS4ERR_STALE_STATEID:
5871         case -NFS4ERR_EXPIRED:
5872                 if (data->arg.new_lock_owner != 0) {
5873                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5874                                                 &lsp->ls_state->open_stateid))
5875                                 rpc_restart_call_prepare(task);
5876                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5877                                                 &lsp->ls_stateid))
5878                                 rpc_restart_call_prepare(task);
5879         }
5880         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5881 }
5882
5883 static void nfs4_lock_release(void *calldata)
5884 {
5885         struct nfs4_lockdata *data = calldata;
5886
5887         dprintk("%s: begin!\n", __func__);
5888         nfs_free_seqid(data->arg.open_seqid);
5889         if (data->cancelled != 0) {
5890                 struct rpc_task *task;
5891                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5892                                 data->arg.lock_seqid);
5893                 if (!IS_ERR(task))
5894                         rpc_put_task_async(task);
5895                 dprintk("%s: cancelling lock!\n", __func__);
5896         } else
5897                 nfs_free_seqid(data->arg.lock_seqid);
5898         nfs4_put_lock_state(data->lsp);
5899         put_nfs_open_context(data->ctx);
5900         fput(data->fl.fl_file);
5901         kfree(data);
5902         dprintk("%s: done!\n", __func__);
5903 }
5904
5905 static const struct rpc_call_ops nfs4_lock_ops = {
5906         .rpc_call_prepare = nfs4_lock_prepare,
5907         .rpc_call_done = nfs4_lock_done,
5908         .rpc_release = nfs4_lock_release,
5909 };
5910
5911 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5912 {
5913         switch (error) {
5914         case -NFS4ERR_ADMIN_REVOKED:
5915         case -NFS4ERR_BAD_STATEID:
5916                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5917                 if (new_lock_owner != 0 ||
5918                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5919                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5920                 break;
5921         case -NFS4ERR_STALE_STATEID:
5922                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5923         case -NFS4ERR_EXPIRED:
5924                 nfs4_schedule_lease_recovery(server->nfs_client);
5925         };
5926 }
5927
5928 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5929 {
5930         struct nfs4_lockdata *data;
5931         struct rpc_task *task;
5932         struct rpc_message msg = {
5933                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5934                 .rpc_cred = state->owner->so_cred,
5935         };
5936         struct rpc_task_setup task_setup_data = {
5937                 .rpc_client = NFS_CLIENT(state->inode),
5938                 .rpc_message = &msg,
5939                 .callback_ops = &nfs4_lock_ops,
5940                 .workqueue = nfsiod_workqueue,
5941                 .flags = RPC_TASK_ASYNC,
5942         };
5943         int ret;
5944
5945         dprintk("%s: begin!\n", __func__);
5946         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5947                         fl->fl_u.nfs4_fl.owner,
5948                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5949         if (data == NULL)
5950                 return -ENOMEM;
5951         if (IS_SETLKW(cmd))
5952                 data->arg.block = 1;
5953         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5954         msg.rpc_argp = &data->arg;
5955         msg.rpc_resp = &data->res;
5956         task_setup_data.callback_data = data;
5957         if (recovery_type > NFS_LOCK_NEW) {
5958                 if (recovery_type == NFS_LOCK_RECLAIM)
5959                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5960                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5961         } else
5962                 data->arg.new_lock = 1;
5963         task = rpc_run_task(&task_setup_data);
5964         if (IS_ERR(task))
5965                 return PTR_ERR(task);
5966         ret = nfs4_wait_for_completion_rpc_task(task);
5967         if (ret == 0) {
5968                 ret = data->rpc_status;
5969                 if (ret)
5970                         nfs4_handle_setlk_error(data->server, data->lsp,
5971                                         data->arg.new_lock_owner, ret);
5972         } else
5973                 data->cancelled = 1;
5974         rpc_put_task(task);
5975         dprintk("%s: done, ret = %d!\n", __func__, ret);
5976         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
5977         return ret;
5978 }
5979
5980 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5981 {
5982         struct nfs_server *server = NFS_SERVER(state->inode);
5983         struct nfs4_exception exception = {
5984                 .inode = state->inode,
5985         };
5986         int err;
5987
5988         do {
5989                 /* Cache the lock if possible... */
5990                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5991                         return 0;
5992                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5993                 if (err != -NFS4ERR_DELAY)
5994                         break;
5995                 nfs4_handle_exception(server, err, &exception);
5996         } while (exception.retry);
5997         return err;
5998 }
5999
6000 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6001 {
6002         struct nfs_server *server = NFS_SERVER(state->inode);
6003         struct nfs4_exception exception = {
6004                 .inode = state->inode,
6005         };
6006         int err;
6007
6008         err = nfs4_set_lock_state(state, request);
6009         if (err != 0)
6010                 return err;
6011         if (!recover_lost_locks) {
6012                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6013                 return 0;
6014         }
6015         do {
6016                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6017                         return 0;
6018                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6019                 switch (err) {
6020                 default:
6021                         goto out;
6022                 case -NFS4ERR_GRACE:
6023                 case -NFS4ERR_DELAY:
6024                         nfs4_handle_exception(server, err, &exception);
6025                         err = 0;
6026                 }
6027         } while (exception.retry);
6028 out:
6029         return err;
6030 }
6031
6032 #if defined(CONFIG_NFS_V4_1)
6033 /**
6034  * nfs41_check_expired_locks - possibly free a lock stateid
6035  *
6036  * @state: NFSv4 state for an inode
6037  *
6038  * Returns NFS_OK if recovery for this stateid is now finished.
6039  * Otherwise a negative NFS4ERR value is returned.
6040  */
6041 static int nfs41_check_expired_locks(struct nfs4_state *state)
6042 {
6043         int status, ret = -NFS4ERR_BAD_STATEID;
6044         struct nfs4_lock_state *lsp;
6045         struct nfs_server *server = NFS_SERVER(state->inode);
6046
6047         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6048                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6049                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6050
6051                         status = nfs41_test_stateid(server,
6052                                         &lsp->ls_stateid,
6053                                         cred);
6054                         trace_nfs4_test_lock_stateid(state, lsp, status);
6055                         if (status != NFS_OK) {
6056                                 /* Free the stateid unless the server
6057                                  * informs us the stateid is unrecognized. */
6058                                 if (status != -NFS4ERR_BAD_STATEID)
6059                                         nfs41_free_stateid(server,
6060                                                         &lsp->ls_stateid,
6061                                                         cred);
6062                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6063                                 ret = status;
6064                         }
6065                 }
6066         };
6067
6068         return ret;
6069 }
6070
6071 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6072 {
6073         int status = NFS_OK;
6074
6075         if (test_bit(LK_STATE_IN_USE, &state->flags))
6076                 status = nfs41_check_expired_locks(state);
6077         if (status != NFS_OK)
6078                 status = nfs4_lock_expired(state, request);
6079         return status;
6080 }
6081 #endif
6082
6083 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6084 {
6085         struct nfs_inode *nfsi = NFS_I(state->inode);
6086         struct nfs4_state_owner *sp = state->owner;
6087         unsigned char fl_flags = request->fl_flags;
6088         int status = -ENOLCK;
6089
6090         if ((fl_flags & FL_POSIX) &&
6091                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6092                 goto out;
6093         /* Is this a delegated open? */
6094         status = nfs4_set_lock_state(state, request);
6095         if (status != 0)
6096                 goto out;
6097         request->fl_flags |= FL_ACCESS;
6098         status = do_vfs_lock(state->inode, request);
6099         if (status < 0)
6100                 goto out;
6101         mutex_lock(&sp->so_delegreturn_mutex);
6102         down_read(&nfsi->rwsem);
6103         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6104                 /* Yes: cache locks! */
6105                 /* ...but avoid races with delegation recall... */
6106                 request->fl_flags = fl_flags & ~FL_SLEEP;
6107                 status = do_vfs_lock(state->inode, request);
6108                 up_read(&nfsi->rwsem);
6109                 mutex_unlock(&sp->so_delegreturn_mutex);
6110                 goto out;
6111         }
6112         up_read(&nfsi->rwsem);
6113         mutex_unlock(&sp->so_delegreturn_mutex);
6114         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6115 out:
6116         request->fl_flags = fl_flags;
6117         return status;
6118 }
6119
6120 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6121 {
6122         struct nfs4_exception exception = {
6123                 .state = state,
6124                 .inode = state->inode,
6125         };
6126         int err;
6127
6128         do {
6129                 err = _nfs4_proc_setlk(state, cmd, request);
6130                 if (err == -NFS4ERR_DENIED)
6131                         err = -EAGAIN;
6132                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6133                                 err, &exception);
6134         } while (exception.retry);
6135         return err;
6136 }
6137
6138 static int
6139 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6140 {
6141         struct nfs_open_context *ctx;
6142         struct nfs4_state *state;
6143         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6144         int status;
6145
6146         /* verify open state */
6147         ctx = nfs_file_open_context(filp);
6148         state = ctx->state;
6149
6150         if (request->fl_start < 0 || request->fl_end < 0)
6151                 return -EINVAL;
6152
6153         if (IS_GETLK(cmd)) {
6154                 if (state != NULL)
6155                         return nfs4_proc_getlk(state, F_GETLK, request);
6156                 return 0;
6157         }
6158
6159         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6160                 return -EINVAL;
6161
6162         if (request->fl_type == F_UNLCK) {
6163                 if (state != NULL)
6164                         return nfs4_proc_unlck(state, cmd, request);
6165                 return 0;
6166         }
6167
6168         if (state == NULL)
6169                 return -ENOLCK;
6170         /*
6171          * Don't rely on the VFS having checked the file open mode,
6172          * since it won't do this for flock() locks.
6173          */
6174         switch (request->fl_type) {
6175         case F_RDLCK:
6176                 if (!(filp->f_mode & FMODE_READ))
6177                         return -EBADF;
6178                 break;
6179         case F_WRLCK:
6180                 if (!(filp->f_mode & FMODE_WRITE))
6181                         return -EBADF;
6182         }
6183
6184         do {
6185                 status = nfs4_proc_setlk(state, cmd, request);
6186                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6187                         break;
6188                 timeout = nfs4_set_lock_task_retry(timeout);
6189                 status = -ERESTARTSYS;
6190                 if (signalled())
6191                         break;
6192         } while(status < 0);
6193         return status;
6194 }
6195
6196 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6197 {
6198         struct nfs_server *server = NFS_SERVER(state->inode);
6199         int err;
6200
6201         err = nfs4_set_lock_state(state, fl);
6202         if (err != 0)
6203                 return err;
6204         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6205         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6206 }
6207
6208 struct nfs_release_lockowner_data {
6209         struct nfs4_lock_state *lsp;
6210         struct nfs_server *server;
6211         struct nfs_release_lockowner_args args;
6212         struct nfs_release_lockowner_res res;
6213         unsigned long timestamp;
6214 };
6215
6216 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6217 {
6218         struct nfs_release_lockowner_data *data = calldata;
6219         struct nfs_server *server = data->server;
6220         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6221                              &data->args.seq_args, &data->res.seq_res, task);
6222         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6223         data->timestamp = jiffies;
6224 }
6225
6226 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6227 {
6228         struct nfs_release_lockowner_data *data = calldata;
6229         struct nfs_server *server = data->server;
6230
6231         nfs40_sequence_done(task, &data->res.seq_res);
6232
6233         switch (task->tk_status) {
6234         case 0:
6235                 renew_lease(server, data->timestamp);
6236                 break;
6237         case -NFS4ERR_STALE_CLIENTID:
6238         case -NFS4ERR_EXPIRED:
6239                 nfs4_schedule_lease_recovery(server->nfs_client);
6240                 break;
6241         case -NFS4ERR_LEASE_MOVED:
6242         case -NFS4ERR_DELAY:
6243                 if (nfs4_async_handle_error(task, server,
6244                                             NULL, NULL) == -EAGAIN)
6245                         rpc_restart_call_prepare(task);
6246         }
6247 }
6248
6249 static void nfs4_release_lockowner_release(void *calldata)
6250 {
6251         struct nfs_release_lockowner_data *data = calldata;
6252         nfs4_free_lock_state(data->server, data->lsp);
6253         kfree(calldata);
6254 }
6255
6256 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6257         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6258         .rpc_call_done = nfs4_release_lockowner_done,
6259         .rpc_release = nfs4_release_lockowner_release,
6260 };
6261
6262 static void
6263 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6264 {
6265         struct nfs_release_lockowner_data *data;
6266         struct rpc_message msg = {
6267                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6268         };
6269
6270         if (server->nfs_client->cl_mvops->minor_version != 0)
6271                 return;
6272
6273         data = kmalloc(sizeof(*data), GFP_NOFS);
6274         if (!data)
6275                 return;
6276         data->lsp = lsp;
6277         data->server = server;
6278         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6279         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6280         data->args.lock_owner.s_dev = server->s_dev;
6281
6282         msg.rpc_argp = &data->args;
6283         msg.rpc_resp = &data->res;
6284         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6285         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6286 }
6287
6288 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6289
6290 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6291                                    struct dentry *unused, struct inode *inode,
6292                                    const char *key, const void *buf,
6293                                    size_t buflen, int flags)
6294 {
6295         return nfs4_proc_set_acl(inode, buf, buflen);
6296 }
6297
6298 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6299                                    struct dentry *unused, struct inode *inode,
6300                                    const char *key, void *buf, size_t buflen)
6301 {
6302         return nfs4_proc_get_acl(inode, buf, buflen);
6303 }
6304
6305 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6306 {
6307         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6308 }
6309
6310 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6311
6312 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6313                                      struct dentry *unused, struct inode *inode,
6314                                      const char *key, const void *buf,
6315                                      size_t buflen, int flags)
6316 {
6317         if (security_ismaclabel(key))
6318                 return nfs4_set_security_label(inode, buf, buflen);
6319
6320         return -EOPNOTSUPP;
6321 }
6322
6323 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6324                                      struct dentry *unused, struct inode *inode,
6325                                      const char *key, void *buf, size_t buflen)
6326 {
6327         if (security_ismaclabel(key))
6328                 return nfs4_get_security_label(inode, buf, buflen);
6329         return -EOPNOTSUPP;
6330 }
6331
6332 static ssize_t
6333 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6334 {
6335         int len = 0;
6336
6337         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6338                 len = security_inode_listsecurity(inode, list, list_len);
6339                 if (list_len && len > list_len)
6340                         return -ERANGE;
6341         }
6342         return len;
6343 }
6344
6345 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6346         .prefix = XATTR_SECURITY_PREFIX,
6347         .get    = nfs4_xattr_get_nfs4_label,
6348         .set    = nfs4_xattr_set_nfs4_label,
6349 };
6350
6351 #else
6352
6353 static ssize_t
6354 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6355 {
6356         return 0;
6357 }
6358
6359 #endif
6360
6361 /*
6362  * nfs_fhget will use either the mounted_on_fileid or the fileid
6363  */
6364 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6365 {
6366         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6367                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6368               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6369               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6370                 return;
6371
6372         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6373                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6374         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6375         fattr->nlink = 2;
6376 }
6377
6378 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6379                                    const struct qstr *name,
6380                                    struct nfs4_fs_locations *fs_locations,
6381                                    struct page *page)
6382 {
6383         struct nfs_server *server = NFS_SERVER(dir);
6384         u32 bitmask[3] = {
6385                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6386         };
6387         struct nfs4_fs_locations_arg args = {
6388                 .dir_fh = NFS_FH(dir),
6389                 .name = name,
6390                 .page = page,
6391                 .bitmask = bitmask,
6392         };
6393         struct nfs4_fs_locations_res res = {
6394                 .fs_locations = fs_locations,
6395         };
6396         struct rpc_message msg = {
6397                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6398                 .rpc_argp = &args,
6399                 .rpc_resp = &res,
6400         };
6401         int status;
6402
6403         dprintk("%s: start\n", __func__);
6404
6405         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6406          * is not supported */
6407         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6408                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6409         else
6410                 bitmask[0] |= FATTR4_WORD0_FILEID;
6411
6412         nfs_fattr_init(&fs_locations->fattr);
6413         fs_locations->server = server;
6414         fs_locations->nlocations = 0;
6415         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6416         dprintk("%s: returned status = %d\n", __func__, status);
6417         return status;
6418 }
6419
6420 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6421                            const struct qstr *name,
6422                            struct nfs4_fs_locations *fs_locations,
6423                            struct page *page)
6424 {
6425         struct nfs4_exception exception = { };
6426         int err;
6427         do {
6428                 err = _nfs4_proc_fs_locations(client, dir, name,
6429                                 fs_locations, page);
6430                 trace_nfs4_get_fs_locations(dir, name, err);
6431                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6432                                 &exception);
6433         } while (exception.retry);
6434         return err;
6435 }
6436
6437 /*
6438  * This operation also signals the server that this client is
6439  * performing migration recovery.  The server can stop returning
6440  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6441  * appended to this compound to identify the client ID which is
6442  * performing recovery.
6443  */
6444 static int _nfs40_proc_get_locations(struct inode *inode,
6445                                      struct nfs4_fs_locations *locations,
6446                                      struct page *page, struct rpc_cred *cred)
6447 {
6448         struct nfs_server *server = NFS_SERVER(inode);
6449         struct rpc_clnt *clnt = server->client;
6450         u32 bitmask[2] = {
6451                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6452         };
6453         struct nfs4_fs_locations_arg args = {
6454                 .clientid       = server->nfs_client->cl_clientid,
6455                 .fh             = NFS_FH(inode),
6456                 .page           = page,
6457                 .bitmask        = bitmask,
6458                 .migration      = 1,            /* skip LOOKUP */
6459                 .renew          = 1,            /* append RENEW */
6460         };
6461         struct nfs4_fs_locations_res res = {
6462                 .fs_locations   = locations,
6463                 .migration      = 1,
6464                 .renew          = 1,
6465         };
6466         struct rpc_message msg = {
6467                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6468                 .rpc_argp       = &args,
6469                 .rpc_resp       = &res,
6470                 .rpc_cred       = cred,
6471         };
6472         unsigned long now = jiffies;
6473         int status;
6474
6475         nfs_fattr_init(&locations->fattr);
6476         locations->server = server;
6477         locations->nlocations = 0;
6478
6479         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6480         nfs4_set_sequence_privileged(&args.seq_args);
6481         status = nfs4_call_sync_sequence(clnt, server, &msg,
6482                                         &args.seq_args, &res.seq_res);
6483         if (status)
6484                 return status;
6485
6486         renew_lease(server, now);
6487         return 0;
6488 }
6489
6490 #ifdef CONFIG_NFS_V4_1
6491
6492 /*
6493  * This operation also signals the server that this client is
6494  * performing migration recovery.  The server can stop asserting
6495  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6496  * performing this operation is identified in the SEQUENCE
6497  * operation in this compound.
6498  *
6499  * When the client supports GETATTR(fs_locations_info), it can
6500  * be plumbed in here.
6501  */
6502 static int _nfs41_proc_get_locations(struct inode *inode,
6503                                      struct nfs4_fs_locations *locations,
6504                                      struct page *page, struct rpc_cred *cred)
6505 {
6506         struct nfs_server *server = NFS_SERVER(inode);
6507         struct rpc_clnt *clnt = server->client;
6508         u32 bitmask[2] = {
6509                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6510         };
6511         struct nfs4_fs_locations_arg args = {
6512                 .fh             = NFS_FH(inode),
6513                 .page           = page,
6514                 .bitmask        = bitmask,
6515                 .migration      = 1,            /* skip LOOKUP */
6516         };
6517         struct nfs4_fs_locations_res res = {
6518                 .fs_locations   = locations,
6519                 .migration      = 1,
6520         };
6521         struct rpc_message msg = {
6522                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6523                 .rpc_argp       = &args,
6524                 .rpc_resp       = &res,
6525                 .rpc_cred       = cred,
6526         };
6527         int status;
6528
6529         nfs_fattr_init(&locations->fattr);
6530         locations->server = server;
6531         locations->nlocations = 0;
6532
6533         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6534         nfs4_set_sequence_privileged(&args.seq_args);
6535         status = nfs4_call_sync_sequence(clnt, server, &msg,
6536                                         &args.seq_args, &res.seq_res);
6537         if (status == NFS4_OK &&
6538             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6539                 status = -NFS4ERR_LEASE_MOVED;
6540         return status;
6541 }
6542
6543 #endif  /* CONFIG_NFS_V4_1 */
6544
6545 /**
6546  * nfs4_proc_get_locations - discover locations for a migrated FSID
6547  * @inode: inode on FSID that is migrating
6548  * @locations: result of query
6549  * @page: buffer
6550  * @cred: credential to use for this operation
6551  *
6552  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6553  * operation failed, or a negative errno if a local error occurred.
6554  *
6555  * On success, "locations" is filled in, but if the server has
6556  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6557  * asserted.
6558  *
6559  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6560  * from this client that require migration recovery.
6561  */
6562 int nfs4_proc_get_locations(struct inode *inode,
6563                             struct nfs4_fs_locations *locations,
6564                             struct page *page, struct rpc_cred *cred)
6565 {
6566         struct nfs_server *server = NFS_SERVER(inode);
6567         struct nfs_client *clp = server->nfs_client;
6568         const struct nfs4_mig_recovery_ops *ops =
6569                                         clp->cl_mvops->mig_recovery_ops;
6570         struct nfs4_exception exception = { };
6571         int status;
6572
6573         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6574                 (unsigned long long)server->fsid.major,
6575                 (unsigned long long)server->fsid.minor,
6576                 clp->cl_hostname);
6577         nfs_display_fhandle(NFS_FH(inode), __func__);
6578
6579         do {
6580                 status = ops->get_locations(inode, locations, page, cred);
6581                 if (status != -NFS4ERR_DELAY)
6582                         break;
6583                 nfs4_handle_exception(server, status, &exception);
6584         } while (exception.retry);
6585         return status;
6586 }
6587
6588 /*
6589  * This operation also signals the server that this client is
6590  * performing "lease moved" recovery.  The server can stop
6591  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6592  * is appended to this compound to identify the client ID which is
6593  * performing recovery.
6594  */
6595 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6596 {
6597         struct nfs_server *server = NFS_SERVER(inode);
6598         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6599         struct rpc_clnt *clnt = server->client;
6600         struct nfs4_fsid_present_arg args = {
6601                 .fh             = NFS_FH(inode),
6602                 .clientid       = clp->cl_clientid,
6603                 .renew          = 1,            /* append RENEW */
6604         };
6605         struct nfs4_fsid_present_res res = {
6606                 .renew          = 1,
6607         };
6608         struct rpc_message msg = {
6609                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6610                 .rpc_argp       = &args,
6611                 .rpc_resp       = &res,
6612                 .rpc_cred       = cred,
6613         };
6614         unsigned long now = jiffies;
6615         int status;
6616
6617         res.fh = nfs_alloc_fhandle();
6618         if (res.fh == NULL)
6619                 return -ENOMEM;
6620
6621         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6622         nfs4_set_sequence_privileged(&args.seq_args);
6623         status = nfs4_call_sync_sequence(clnt, server, &msg,
6624                                                 &args.seq_args, &res.seq_res);
6625         nfs_free_fhandle(res.fh);
6626         if (status)
6627                 return status;
6628
6629         do_renew_lease(clp, now);
6630         return 0;
6631 }
6632
6633 #ifdef CONFIG_NFS_V4_1
6634
6635 /*
6636  * This operation also signals the server that this client is
6637  * performing "lease moved" recovery.  The server can stop asserting
6638  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6639  * this operation is identified in the SEQUENCE operation in this
6640  * compound.
6641  */
6642 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6643 {
6644         struct nfs_server *server = NFS_SERVER(inode);
6645         struct rpc_clnt *clnt = server->client;
6646         struct nfs4_fsid_present_arg args = {
6647                 .fh             = NFS_FH(inode),
6648         };
6649         struct nfs4_fsid_present_res res = {
6650         };
6651         struct rpc_message msg = {
6652                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6653                 .rpc_argp       = &args,
6654                 .rpc_resp       = &res,
6655                 .rpc_cred       = cred,
6656         };
6657         int status;
6658
6659         res.fh = nfs_alloc_fhandle();
6660         if (res.fh == NULL)
6661                 return -ENOMEM;
6662
6663         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6664         nfs4_set_sequence_privileged(&args.seq_args);
6665         status = nfs4_call_sync_sequence(clnt, server, &msg,
6666                                                 &args.seq_args, &res.seq_res);
6667         nfs_free_fhandle(res.fh);
6668         if (status == NFS4_OK &&
6669             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6670                 status = -NFS4ERR_LEASE_MOVED;
6671         return status;
6672 }
6673
6674 #endif  /* CONFIG_NFS_V4_1 */
6675
6676 /**
6677  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6678  * @inode: inode on FSID to check
6679  * @cred: credential to use for this operation
6680  *
6681  * Server indicates whether the FSID is present, moved, or not
6682  * recognized.  This operation is necessary to clear a LEASE_MOVED
6683  * condition for this client ID.
6684  *
6685  * Returns NFS4_OK if the FSID is present on this server,
6686  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6687  *  NFS4ERR code if some error occurred on the server, or a
6688  *  negative errno if a local failure occurred.
6689  */
6690 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6691 {
6692         struct nfs_server *server = NFS_SERVER(inode);
6693         struct nfs_client *clp = server->nfs_client;
6694         const struct nfs4_mig_recovery_ops *ops =
6695                                         clp->cl_mvops->mig_recovery_ops;
6696         struct nfs4_exception exception = { };
6697         int status;
6698
6699         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6700                 (unsigned long long)server->fsid.major,
6701                 (unsigned long long)server->fsid.minor,
6702                 clp->cl_hostname);
6703         nfs_display_fhandle(NFS_FH(inode), __func__);
6704
6705         do {
6706                 status = ops->fsid_present(inode, cred);
6707                 if (status != -NFS4ERR_DELAY)
6708                         break;
6709                 nfs4_handle_exception(server, status, &exception);
6710         } while (exception.retry);
6711         return status;
6712 }
6713
6714 /**
6715  * If 'use_integrity' is true and the state managment nfs_client
6716  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6717  * and the machine credential as per RFC3530bis and RFC5661 Security
6718  * Considerations sections. Otherwise, just use the user cred with the
6719  * filesystem's rpc_client.
6720  */
6721 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6722 {
6723         int status;
6724         struct nfs4_secinfo_arg args = {
6725                 .dir_fh = NFS_FH(dir),
6726                 .name   = name,
6727         };
6728         struct nfs4_secinfo_res res = {
6729                 .flavors     = flavors,
6730         };
6731         struct rpc_message msg = {
6732                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6733                 .rpc_argp = &args,
6734                 .rpc_resp = &res,
6735         };
6736         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6737         struct rpc_cred *cred = NULL;
6738
6739         if (use_integrity) {
6740                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6741                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6742                 msg.rpc_cred = cred;
6743         }
6744
6745         dprintk("NFS call  secinfo %s\n", name->name);
6746
6747         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6748                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6749
6750         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6751                                 &res.seq_res, 0);
6752         dprintk("NFS reply  secinfo: %d\n", status);
6753
6754         if (cred)
6755                 put_rpccred(cred);
6756
6757         return status;
6758 }
6759
6760 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6761                       struct nfs4_secinfo_flavors *flavors)
6762 {
6763         struct nfs4_exception exception = { };
6764         int err;
6765         do {
6766                 err = -NFS4ERR_WRONGSEC;
6767
6768                 /* try to use integrity protection with machine cred */
6769                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6770                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6771
6772                 /*
6773                  * if unable to use integrity protection, or SECINFO with
6774                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6775                  * disallowed by spec, but exists in deployed servers) use
6776                  * the current filesystem's rpc_client and the user cred.
6777                  */
6778                 if (err == -NFS4ERR_WRONGSEC)
6779                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6780
6781                 trace_nfs4_secinfo(dir, name, err);
6782                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6783                                 &exception);
6784         } while (exception.retry);
6785         return err;
6786 }
6787
6788 #ifdef CONFIG_NFS_V4_1
6789 /*
6790  * Check the exchange flags returned by the server for invalid flags, having
6791  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6792  * DS flags set.
6793  */
6794 static int nfs4_check_cl_exchange_flags(u32 flags)
6795 {
6796         if (flags & ~EXCHGID4_FLAG_MASK_R)
6797                 goto out_inval;
6798         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6799             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6800                 goto out_inval;
6801         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6802                 goto out_inval;
6803         return NFS_OK;
6804 out_inval:
6805         return -NFS4ERR_INVAL;
6806 }
6807
6808 static bool
6809 nfs41_same_server_scope(struct nfs41_server_scope *a,
6810                         struct nfs41_server_scope *b)
6811 {
6812         if (a->server_scope_sz == b->server_scope_sz &&
6813             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6814                 return true;
6815
6816         return false;
6817 }
6818
6819 static void
6820 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
6821 {
6822 }
6823
6824 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
6825         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
6826 };
6827
6828 /*
6829  * nfs4_proc_bind_one_conn_to_session()
6830  *
6831  * The 4.1 client currently uses the same TCP connection for the
6832  * fore and backchannel.
6833  */
6834 static
6835 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
6836                 struct rpc_xprt *xprt,
6837                 struct nfs_client *clp,
6838                 struct rpc_cred *cred)
6839 {
6840         int status;
6841         struct nfs41_bind_conn_to_session_args args = {
6842                 .client = clp,
6843                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6844         };
6845         struct nfs41_bind_conn_to_session_res res;
6846         struct rpc_message msg = {
6847                 .rpc_proc =
6848                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6849                 .rpc_argp = &args,
6850                 .rpc_resp = &res,
6851                 .rpc_cred = cred,
6852         };
6853         struct rpc_task_setup task_setup_data = {
6854                 .rpc_client = clnt,
6855                 .rpc_xprt = xprt,
6856                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
6857                 .rpc_message = &msg,
6858                 .flags = RPC_TASK_TIMEOUT,
6859         };
6860         struct rpc_task *task;
6861
6862         dprintk("--> %s\n", __func__);
6863
6864         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6865         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6866                 args.dir = NFS4_CDFC4_FORE;
6867
6868         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
6869         if (xprt != rcu_access_pointer(clnt->cl_xprt))
6870                 args.dir = NFS4_CDFC4_FORE;
6871
6872         task = rpc_run_task(&task_setup_data);
6873         if (!IS_ERR(task)) {
6874                 status = task->tk_status;
6875                 rpc_put_task(task);
6876         } else
6877                 status = PTR_ERR(task);
6878         trace_nfs4_bind_conn_to_session(clp, status);
6879         if (status == 0) {
6880                 if (memcmp(res.sessionid.data,
6881                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6882                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6883                         status = -EIO;
6884                         goto out;
6885                 }
6886                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6887                         dprintk("NFS: %s: Unexpected direction from server\n",
6888                                 __func__);
6889                         status = -EIO;
6890                         goto out;
6891                 }
6892                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6893                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6894                                 __func__);
6895                         status = -EIO;
6896                         goto out;
6897                 }
6898         }
6899 out:
6900         dprintk("<-- %s status= %d\n", __func__, status);
6901         return status;
6902 }
6903
6904 struct rpc_bind_conn_calldata {
6905         struct nfs_client *clp;
6906         struct rpc_cred *cred;
6907 };
6908
6909 static int
6910 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
6911                 struct rpc_xprt *xprt,
6912                 void *calldata)
6913 {
6914         struct rpc_bind_conn_calldata *p = calldata;
6915
6916         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
6917 }
6918
6919 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6920 {
6921         struct rpc_bind_conn_calldata data = {
6922                 .clp = clp,
6923                 .cred = cred,
6924         };
6925         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
6926                         nfs4_proc_bind_conn_to_session_callback, &data);
6927 }
6928
6929 /*
6930  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6931  * and operations we'd like to see to enable certain features in the allow map
6932  */
6933 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6934         .how = SP4_MACH_CRED,
6935         .enforce.u.words = {
6936                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6937                       1 << (OP_EXCHANGE_ID - 32) |
6938                       1 << (OP_CREATE_SESSION - 32) |
6939                       1 << (OP_DESTROY_SESSION - 32) |
6940                       1 << (OP_DESTROY_CLIENTID - 32)
6941         },
6942         .allow.u.words = {
6943                 [0] = 1 << (OP_CLOSE) |
6944                       1 << (OP_OPEN_DOWNGRADE) |
6945                       1 << (OP_LOCKU) |
6946                       1 << (OP_DELEGRETURN) |
6947                       1 << (OP_COMMIT),
6948                 [1] = 1 << (OP_SECINFO - 32) |
6949                       1 << (OP_SECINFO_NO_NAME - 32) |
6950                       1 << (OP_LAYOUTRETURN - 32) |
6951                       1 << (OP_TEST_STATEID - 32) |
6952                       1 << (OP_FREE_STATEID - 32) |
6953                       1 << (OP_WRITE - 32)
6954         }
6955 };
6956
6957 /*
6958  * Select the state protection mode for client `clp' given the server results
6959  * from exchange_id in `sp'.
6960  *
6961  * Returns 0 on success, negative errno otherwise.
6962  */
6963 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6964                                  struct nfs41_state_protection *sp)
6965 {
6966         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6967                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6968                       1 << (OP_EXCHANGE_ID - 32) |
6969                       1 << (OP_CREATE_SESSION - 32) |
6970                       1 << (OP_DESTROY_SESSION - 32) |
6971                       1 << (OP_DESTROY_CLIENTID - 32)
6972         };
6973         unsigned int i;
6974
6975         if (sp->how == SP4_MACH_CRED) {
6976                 /* Print state protect result */
6977                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6978                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6979                         if (test_bit(i, sp->enforce.u.longs))
6980                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6981                         if (test_bit(i, sp->allow.u.longs))
6982                                 dfprintk(MOUNT, "  allow op %d\n", i);
6983                 }
6984
6985                 /* make sure nothing is on enforce list that isn't supported */
6986                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6987                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6988                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6989                                 return -EINVAL;
6990                         }
6991                 }
6992
6993                 /*
6994                  * Minimal mode - state operations are allowed to use machine
6995                  * credential.  Note this already happens by default, so the
6996                  * client doesn't have to do anything more than the negotiation.
6997                  *
6998                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6999                  *       we're already using the machine cred for exchange_id
7000                  *       and will never use a different cred.
7001                  */
7002                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7003                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7004                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7005                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7006                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7007                         dfprintk(MOUNT, "  minimal mode enabled\n");
7008                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7009                 } else {
7010                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7011                         return -EINVAL;
7012                 }
7013
7014                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7015                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7016                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7017                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7018                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7019                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7020                 }
7021
7022                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7023                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7024                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7025                                 &clp->cl_sp4_flags);
7026                 }
7027
7028                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7029                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7030                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7031                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7032                 }
7033
7034                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7035                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7036                         dfprintk(MOUNT, "  stateid mode enabled\n");
7037                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7038                 }
7039
7040                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7041                         dfprintk(MOUNT, "  write mode enabled\n");
7042                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7043                 }
7044
7045                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7046                         dfprintk(MOUNT, "  commit mode enabled\n");
7047                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7048                 }
7049         }
7050
7051         return 0;
7052 }
7053
7054 /*
7055  * _nfs4_proc_exchange_id()
7056  *
7057  * Wrapper for EXCHANGE_ID operation.
7058  */
7059 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7060         u32 sp4_how)
7061 {
7062         nfs4_verifier verifier;
7063         struct nfs41_exchange_id_args args = {
7064                 .verifier = &verifier,
7065                 .client = clp,
7066 #ifdef CONFIG_NFS_V4_1_MIGRATION
7067                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7068                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
7069                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7070 #else
7071                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7072                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
7073 #endif
7074         };
7075         struct nfs41_exchange_id_res res = {
7076                 0
7077         };
7078         int status;
7079         struct rpc_message msg = {
7080                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7081                 .rpc_argp = &args,
7082                 .rpc_resp = &res,
7083                 .rpc_cred = cred,
7084         };
7085
7086         nfs4_init_boot_verifier(clp, &verifier);
7087
7088         status = nfs4_init_uniform_client_string(clp);
7089         if (status)
7090                 goto out;
7091
7092         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7093                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7094                 clp->cl_owner_id);
7095
7096         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7097                                         GFP_NOFS);
7098         if (unlikely(res.server_owner == NULL)) {
7099                 status = -ENOMEM;
7100                 goto out;
7101         }
7102
7103         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7104                                         GFP_NOFS);
7105         if (unlikely(res.server_scope == NULL)) {
7106                 status = -ENOMEM;
7107                 goto out_server_owner;
7108         }
7109
7110         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7111         if (unlikely(res.impl_id == NULL)) {
7112                 status = -ENOMEM;
7113                 goto out_server_scope;
7114         }
7115
7116         switch (sp4_how) {
7117         case SP4_NONE:
7118                 args.state_protect.how = SP4_NONE;
7119                 break;
7120
7121         case SP4_MACH_CRED:
7122                 args.state_protect = nfs4_sp4_mach_cred_request;
7123                 break;
7124
7125         default:
7126                 /* unsupported! */
7127                 WARN_ON_ONCE(1);
7128                 status = -EINVAL;
7129                 goto out_impl_id;
7130         }
7131
7132         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7133         trace_nfs4_exchange_id(clp, status);
7134         if (status == 0)
7135                 status = nfs4_check_cl_exchange_flags(res.flags);
7136
7137         if (status == 0)
7138                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7139
7140         if (status == 0) {
7141                 clp->cl_clientid = res.clientid;
7142                 clp->cl_exchange_flags = res.flags;
7143                 /* Client ID is not confirmed */
7144                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7145                         clear_bit(NFS4_SESSION_ESTABLISHED,
7146                                         &clp->cl_session->session_state);
7147                         clp->cl_seqid = res.seqid;
7148                 }
7149
7150                 kfree(clp->cl_serverowner);
7151                 clp->cl_serverowner = res.server_owner;
7152                 res.server_owner = NULL;
7153
7154                 /* use the most recent implementation id */
7155                 kfree(clp->cl_implid);
7156                 clp->cl_implid = res.impl_id;
7157                 res.impl_id = NULL;
7158
7159                 if (clp->cl_serverscope != NULL &&
7160                     !nfs41_same_server_scope(clp->cl_serverscope,
7161                                              res.server_scope)) {
7162                         dprintk("%s: server_scope mismatch detected\n",
7163                                 __func__);
7164                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7165                         kfree(clp->cl_serverscope);
7166                         clp->cl_serverscope = NULL;
7167                 }
7168
7169                 if (clp->cl_serverscope == NULL) {
7170                         clp->cl_serverscope = res.server_scope;
7171                         res.server_scope = NULL;
7172                 }
7173         }
7174
7175 out_impl_id:
7176         kfree(res.impl_id);
7177 out_server_scope:
7178         kfree(res.server_scope);
7179 out_server_owner:
7180         kfree(res.server_owner);
7181 out:
7182         if (clp->cl_implid != NULL)
7183                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7184                         "domain: %s, name: %s, date: %llu,%u\n",
7185                         clp->cl_implid->domain, clp->cl_implid->name,
7186                         clp->cl_implid->date.seconds,
7187                         clp->cl_implid->date.nseconds);
7188         dprintk("NFS reply exchange_id: %d\n", status);
7189         return status;
7190 }
7191
7192 /*
7193  * nfs4_proc_exchange_id()
7194  *
7195  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7196  *
7197  * Since the clientid has expired, all compounds using sessions
7198  * associated with the stale clientid will be returning
7199  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7200  * be in some phase of session reset.
7201  *
7202  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7203  */
7204 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7205 {
7206         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7207         int status;
7208
7209         /* try SP4_MACH_CRED if krb5i/p */
7210         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7211             authflavor == RPC_AUTH_GSS_KRB5P) {
7212                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7213                 if (!status)
7214                         return 0;
7215         }
7216
7217         /* try SP4_NONE */
7218         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7219 }
7220
7221 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7222                 struct rpc_cred *cred)
7223 {
7224         struct rpc_message msg = {
7225                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7226                 .rpc_argp = clp,
7227                 .rpc_cred = cred,
7228         };
7229         int status;
7230
7231         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7232         trace_nfs4_destroy_clientid(clp, status);
7233         if (status)
7234                 dprintk("NFS: Got error %d from the server %s on "
7235                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7236         return status;
7237 }
7238
7239 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7240                 struct rpc_cred *cred)
7241 {
7242         unsigned int loop;
7243         int ret;
7244
7245         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7246                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7247                 switch (ret) {
7248                 case -NFS4ERR_DELAY:
7249                 case -NFS4ERR_CLIENTID_BUSY:
7250                         ssleep(1);
7251                         break;
7252                 default:
7253                         return ret;
7254                 }
7255         }
7256         return 0;
7257 }
7258
7259 int nfs4_destroy_clientid(struct nfs_client *clp)
7260 {
7261         struct rpc_cred *cred;
7262         int ret = 0;
7263
7264         if (clp->cl_mvops->minor_version < 1)
7265                 goto out;
7266         if (clp->cl_exchange_flags == 0)
7267                 goto out;
7268         if (clp->cl_preserve_clid)
7269                 goto out;
7270         cred = nfs4_get_clid_cred(clp);
7271         ret = nfs4_proc_destroy_clientid(clp, cred);
7272         if (cred)
7273                 put_rpccred(cred);
7274         switch (ret) {
7275         case 0:
7276         case -NFS4ERR_STALE_CLIENTID:
7277                 clp->cl_exchange_flags = 0;
7278         }
7279 out:
7280         return ret;
7281 }
7282
7283 struct nfs4_get_lease_time_data {
7284         struct nfs4_get_lease_time_args *args;
7285         struct nfs4_get_lease_time_res *res;
7286         struct nfs_client *clp;
7287 };
7288
7289 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7290                                         void *calldata)
7291 {
7292         struct nfs4_get_lease_time_data *data =
7293                         (struct nfs4_get_lease_time_data *)calldata;
7294
7295         dprintk("--> %s\n", __func__);
7296         /* just setup sequence, do not trigger session recovery
7297            since we're invoked within one */
7298         nfs41_setup_sequence(data->clp->cl_session,
7299                         &data->args->la_seq_args,
7300                         &data->res->lr_seq_res,
7301                         task);
7302         dprintk("<-- %s\n", __func__);
7303 }
7304
7305 /*
7306  * Called from nfs4_state_manager thread for session setup, so don't recover
7307  * from sequence operation or clientid errors.
7308  */
7309 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7310 {
7311         struct nfs4_get_lease_time_data *data =
7312                         (struct nfs4_get_lease_time_data *)calldata;
7313
7314         dprintk("--> %s\n", __func__);
7315         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7316                 return;
7317         switch (task->tk_status) {
7318         case -NFS4ERR_DELAY:
7319         case -NFS4ERR_GRACE:
7320                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7321                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7322                 task->tk_status = 0;
7323                 /* fall through */
7324         case -NFS4ERR_RETRY_UNCACHED_REP:
7325                 rpc_restart_call_prepare(task);
7326                 return;
7327         }
7328         dprintk("<-- %s\n", __func__);
7329 }
7330
7331 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7332         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7333         .rpc_call_done = nfs4_get_lease_time_done,
7334 };
7335
7336 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7337 {
7338         struct rpc_task *task;
7339         struct nfs4_get_lease_time_args args;
7340         struct nfs4_get_lease_time_res res = {
7341                 .lr_fsinfo = fsinfo,
7342         };
7343         struct nfs4_get_lease_time_data data = {
7344                 .args = &args,
7345                 .res = &res,
7346                 .clp = clp,
7347         };
7348         struct rpc_message msg = {
7349                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7350                 .rpc_argp = &args,
7351                 .rpc_resp = &res,
7352         };
7353         struct rpc_task_setup task_setup = {
7354                 .rpc_client = clp->cl_rpcclient,
7355                 .rpc_message = &msg,
7356                 .callback_ops = &nfs4_get_lease_time_ops,
7357                 .callback_data = &data,
7358                 .flags = RPC_TASK_TIMEOUT,
7359         };
7360         int status;
7361
7362         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7363         nfs4_set_sequence_privileged(&args.la_seq_args);
7364         dprintk("--> %s\n", __func__);
7365         task = rpc_run_task(&task_setup);
7366
7367         if (IS_ERR(task))
7368                 status = PTR_ERR(task);
7369         else {
7370                 status = task->tk_status;
7371                 rpc_put_task(task);
7372         }
7373         dprintk("<-- %s return %d\n", __func__, status);
7374
7375         return status;
7376 }
7377
7378 /*
7379  * Initialize the values to be used by the client in CREATE_SESSION
7380  * If nfs4_init_session set the fore channel request and response sizes,
7381  * use them.
7382  *
7383  * Set the back channel max_resp_sz_cached to zero to force the client to
7384  * always set csa_cachethis to FALSE because the current implementation
7385  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7386  */
7387 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7388                                     struct rpc_clnt *clnt)
7389 {
7390         unsigned int max_rqst_sz, max_resp_sz;
7391         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7392
7393         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7394         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7395
7396         /* Fore channel attributes */
7397         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7398         args->fc_attrs.max_resp_sz = max_resp_sz;
7399         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7400         args->fc_attrs.max_reqs = max_session_slots;
7401
7402         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7403                 "max_ops=%u max_reqs=%u\n",
7404                 __func__,
7405                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7406                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7407
7408         /* Back channel attributes */
7409         args->bc_attrs.max_rqst_sz = max_bc_payload;
7410         args->bc_attrs.max_resp_sz = max_bc_payload;
7411         args->bc_attrs.max_resp_sz_cached = 0;
7412         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7413         args->bc_attrs.max_reqs = NFS41_BC_MAX_CALLBACKS;
7414
7415         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7416                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7417                 __func__,
7418                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7419                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7420                 args->bc_attrs.max_reqs);
7421 }
7422
7423 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7424                 struct nfs41_create_session_res *res)
7425 {
7426         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7427         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7428
7429         if (rcvd->max_resp_sz > sent->max_resp_sz)
7430                 return -EINVAL;
7431         /*
7432          * Our requested max_ops is the minimum we need; we're not
7433          * prepared to break up compounds into smaller pieces than that.
7434          * So, no point even trying to continue if the server won't
7435          * cooperate:
7436          */
7437         if (rcvd->max_ops < sent->max_ops)
7438                 return -EINVAL;
7439         if (rcvd->max_reqs == 0)
7440                 return -EINVAL;
7441         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7442                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7443         return 0;
7444 }
7445
7446 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7447                 struct nfs41_create_session_res *res)
7448 {
7449         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7450         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7451
7452         if (!(res->flags & SESSION4_BACK_CHAN))
7453                 goto out;
7454         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7455                 return -EINVAL;
7456         if (rcvd->max_resp_sz < sent->max_resp_sz)
7457                 return -EINVAL;
7458         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7459                 return -EINVAL;
7460         /* These would render the backchannel useless: */
7461         if (rcvd->max_ops != sent->max_ops)
7462                 return -EINVAL;
7463         if (rcvd->max_reqs != sent->max_reqs)
7464                 return -EINVAL;
7465 out:
7466         return 0;
7467 }
7468
7469 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7470                                      struct nfs41_create_session_res *res)
7471 {
7472         int ret;
7473
7474         ret = nfs4_verify_fore_channel_attrs(args, res);
7475         if (ret)
7476                 return ret;
7477         return nfs4_verify_back_channel_attrs(args, res);
7478 }
7479
7480 static void nfs4_update_session(struct nfs4_session *session,
7481                 struct nfs41_create_session_res *res)
7482 {
7483         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7484         /* Mark client id and session as being confirmed */
7485         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7486         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7487         session->flags = res->flags;
7488         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7489         if (res->flags & SESSION4_BACK_CHAN)
7490                 memcpy(&session->bc_attrs, &res->bc_attrs,
7491                                 sizeof(session->bc_attrs));
7492 }
7493
7494 static int _nfs4_proc_create_session(struct nfs_client *clp,
7495                 struct rpc_cred *cred)
7496 {
7497         struct nfs4_session *session = clp->cl_session;
7498         struct nfs41_create_session_args args = {
7499                 .client = clp,
7500                 .clientid = clp->cl_clientid,
7501                 .seqid = clp->cl_seqid,
7502                 .cb_program = NFS4_CALLBACK,
7503         };
7504         struct nfs41_create_session_res res;
7505
7506         struct rpc_message msg = {
7507                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7508                 .rpc_argp = &args,
7509                 .rpc_resp = &res,
7510                 .rpc_cred = cred,
7511         };
7512         int status;
7513
7514         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7515         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7516
7517         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7518         trace_nfs4_create_session(clp, status);
7519
7520         if (!status) {
7521                 /* Verify the session's negotiated channel_attrs values */
7522                 status = nfs4_verify_channel_attrs(&args, &res);
7523                 /* Increment the clientid slot sequence id */
7524                 if (clp->cl_seqid == res.seqid)
7525                         clp->cl_seqid++;
7526                 if (status)
7527                         goto out;
7528                 nfs4_update_session(session, &res);
7529         }
7530 out:
7531         return status;
7532 }
7533
7534 /*
7535  * Issues a CREATE_SESSION operation to the server.
7536  * It is the responsibility of the caller to verify the session is
7537  * expired before calling this routine.
7538  */
7539 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7540 {
7541         int status;
7542         unsigned *ptr;
7543         struct nfs4_session *session = clp->cl_session;
7544
7545         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7546
7547         status = _nfs4_proc_create_session(clp, cred);
7548         if (status)
7549                 goto out;
7550
7551         /* Init or reset the session slot tables */
7552         status = nfs4_setup_session_slot_tables(session);
7553         dprintk("slot table setup returned %d\n", status);
7554         if (status)
7555                 goto out;
7556
7557         ptr = (unsigned *)&session->sess_id.data[0];
7558         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7559                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7560 out:
7561         dprintk("<-- %s\n", __func__);
7562         return status;
7563 }
7564
7565 /*
7566  * Issue the over-the-wire RPC DESTROY_SESSION.
7567  * The caller must serialize access to this routine.
7568  */
7569 int nfs4_proc_destroy_session(struct nfs4_session *session,
7570                 struct rpc_cred *cred)
7571 {
7572         struct rpc_message msg = {
7573                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7574                 .rpc_argp = session,
7575                 .rpc_cred = cred,
7576         };
7577         int status = 0;
7578
7579         dprintk("--> nfs4_proc_destroy_session\n");
7580
7581         /* session is still being setup */
7582         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7583                 return 0;
7584
7585         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7586         trace_nfs4_destroy_session(session->clp, status);
7587
7588         if (status)
7589                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7590                         "Session has been destroyed regardless...\n", status);
7591
7592         dprintk("<-- nfs4_proc_destroy_session\n");
7593         return status;
7594 }
7595
7596 /*
7597  * Renew the cl_session lease.
7598  */
7599 struct nfs4_sequence_data {
7600         struct nfs_client *clp;
7601         struct nfs4_sequence_args args;
7602         struct nfs4_sequence_res res;
7603 };
7604
7605 static void nfs41_sequence_release(void *data)
7606 {
7607         struct nfs4_sequence_data *calldata = data;
7608         struct nfs_client *clp = calldata->clp;
7609
7610         if (atomic_read(&clp->cl_count) > 1)
7611                 nfs4_schedule_state_renewal(clp);
7612         nfs_put_client(clp);
7613         kfree(calldata);
7614 }
7615
7616 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7617 {
7618         switch(task->tk_status) {
7619         case -NFS4ERR_DELAY:
7620                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7621                 return -EAGAIN;
7622         default:
7623                 nfs4_schedule_lease_recovery(clp);
7624         }
7625         return 0;
7626 }
7627
7628 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7629 {
7630         struct nfs4_sequence_data *calldata = data;
7631         struct nfs_client *clp = calldata->clp;
7632
7633         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7634                 return;
7635
7636         trace_nfs4_sequence(clp, task->tk_status);
7637         if (task->tk_status < 0) {
7638                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7639                 if (atomic_read(&clp->cl_count) == 1)
7640                         goto out;
7641
7642                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7643                         rpc_restart_call_prepare(task);
7644                         return;
7645                 }
7646         }
7647         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7648 out:
7649         dprintk("<-- %s\n", __func__);
7650 }
7651
7652 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7653 {
7654         struct nfs4_sequence_data *calldata = data;
7655         struct nfs_client *clp = calldata->clp;
7656         struct nfs4_sequence_args *args;
7657         struct nfs4_sequence_res *res;
7658
7659         args = task->tk_msg.rpc_argp;
7660         res = task->tk_msg.rpc_resp;
7661
7662         nfs41_setup_sequence(clp->cl_session, args, res, task);
7663 }
7664
7665 static const struct rpc_call_ops nfs41_sequence_ops = {
7666         .rpc_call_done = nfs41_sequence_call_done,
7667         .rpc_call_prepare = nfs41_sequence_prepare,
7668         .rpc_release = nfs41_sequence_release,
7669 };
7670
7671 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7672                 struct rpc_cred *cred,
7673                 bool is_privileged)
7674 {
7675         struct nfs4_sequence_data *calldata;
7676         struct rpc_message msg = {
7677                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7678                 .rpc_cred = cred,
7679         };
7680         struct rpc_task_setup task_setup_data = {
7681                 .rpc_client = clp->cl_rpcclient,
7682                 .rpc_message = &msg,
7683                 .callback_ops = &nfs41_sequence_ops,
7684                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7685         };
7686
7687         if (!atomic_inc_not_zero(&clp->cl_count))
7688                 return ERR_PTR(-EIO);
7689         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7690         if (calldata == NULL) {
7691                 nfs_put_client(clp);
7692                 return ERR_PTR(-ENOMEM);
7693         }
7694         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7695         if (is_privileged)
7696                 nfs4_set_sequence_privileged(&calldata->args);
7697         msg.rpc_argp = &calldata->args;
7698         msg.rpc_resp = &calldata->res;
7699         calldata->clp = clp;
7700         task_setup_data.callback_data = calldata;
7701
7702         return rpc_run_task(&task_setup_data);
7703 }
7704
7705 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7706 {
7707         struct rpc_task *task;
7708         int ret = 0;
7709
7710         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7711                 return -EAGAIN;
7712         task = _nfs41_proc_sequence(clp, cred, false);
7713         if (IS_ERR(task))
7714                 ret = PTR_ERR(task);
7715         else
7716                 rpc_put_task_async(task);
7717         dprintk("<-- %s status=%d\n", __func__, ret);
7718         return ret;
7719 }
7720
7721 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7722 {
7723         struct rpc_task *task;
7724         int ret;
7725
7726         task = _nfs41_proc_sequence(clp, cred, true);
7727         if (IS_ERR(task)) {
7728                 ret = PTR_ERR(task);
7729                 goto out;
7730         }
7731         ret = rpc_wait_for_completion_task(task);
7732         if (!ret)
7733                 ret = task->tk_status;
7734         rpc_put_task(task);
7735 out:
7736         dprintk("<-- %s status=%d\n", __func__, ret);
7737         return ret;
7738 }
7739
7740 struct nfs4_reclaim_complete_data {
7741         struct nfs_client *clp;
7742         struct nfs41_reclaim_complete_args arg;
7743         struct nfs41_reclaim_complete_res res;
7744 };
7745
7746 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7747 {
7748         struct nfs4_reclaim_complete_data *calldata = data;
7749
7750         nfs41_setup_sequence(calldata->clp->cl_session,
7751                         &calldata->arg.seq_args,
7752                         &calldata->res.seq_res,
7753                         task);
7754 }
7755
7756 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7757 {
7758         switch(task->tk_status) {
7759         case 0:
7760         case -NFS4ERR_COMPLETE_ALREADY:
7761         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7762                 break;
7763         case -NFS4ERR_DELAY:
7764                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7765                 /* fall through */
7766         case -NFS4ERR_RETRY_UNCACHED_REP:
7767                 return -EAGAIN;
7768         default:
7769                 nfs4_schedule_lease_recovery(clp);
7770         }
7771         return 0;
7772 }
7773
7774 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7775 {
7776         struct nfs4_reclaim_complete_data *calldata = data;
7777         struct nfs_client *clp = calldata->clp;
7778         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7779
7780         dprintk("--> %s\n", __func__);
7781         if (!nfs41_sequence_done(task, res))
7782                 return;
7783
7784         trace_nfs4_reclaim_complete(clp, task->tk_status);
7785         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7786                 rpc_restart_call_prepare(task);
7787                 return;
7788         }
7789         dprintk("<-- %s\n", __func__);
7790 }
7791
7792 static void nfs4_free_reclaim_complete_data(void *data)
7793 {
7794         struct nfs4_reclaim_complete_data *calldata = data;
7795
7796         kfree(calldata);
7797 }
7798
7799 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7800         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7801         .rpc_call_done = nfs4_reclaim_complete_done,
7802         .rpc_release = nfs4_free_reclaim_complete_data,
7803 };
7804
7805 /*
7806  * Issue a global reclaim complete.
7807  */
7808 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7809                 struct rpc_cred *cred)
7810 {
7811         struct nfs4_reclaim_complete_data *calldata;
7812         struct rpc_task *task;
7813         struct rpc_message msg = {
7814                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7815                 .rpc_cred = cred,
7816         };
7817         struct rpc_task_setup task_setup_data = {
7818                 .rpc_client = clp->cl_rpcclient,
7819                 .rpc_message = &msg,
7820                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7821                 .flags = RPC_TASK_ASYNC,
7822         };
7823         int status = -ENOMEM;
7824
7825         dprintk("--> %s\n", __func__);
7826         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7827         if (calldata == NULL)
7828                 goto out;
7829         calldata->clp = clp;
7830         calldata->arg.one_fs = 0;
7831
7832         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7833         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7834         msg.rpc_argp = &calldata->arg;
7835         msg.rpc_resp = &calldata->res;
7836         task_setup_data.callback_data = calldata;
7837         task = rpc_run_task(&task_setup_data);
7838         if (IS_ERR(task)) {
7839                 status = PTR_ERR(task);
7840                 goto out;
7841         }
7842         status = nfs4_wait_for_completion_rpc_task(task);
7843         if (status == 0)
7844                 status = task->tk_status;
7845         rpc_put_task(task);
7846         return 0;
7847 out:
7848         dprintk("<-- %s status=%d\n", __func__, status);
7849         return status;
7850 }
7851
7852 static void
7853 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7854 {
7855         struct nfs4_layoutget *lgp = calldata;
7856         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7857         struct nfs4_session *session = nfs4_get_session(server);
7858
7859         dprintk("--> %s\n", __func__);
7860         nfs41_setup_sequence(session, &lgp->args.seq_args,
7861                                 &lgp->res.seq_res, task);
7862         dprintk("<-- %s\n", __func__);
7863 }
7864
7865 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7866 {
7867         struct nfs4_layoutget *lgp = calldata;
7868
7869         dprintk("--> %s\n", __func__);
7870         nfs41_sequence_done(task, &lgp->res.seq_res);
7871         dprintk("<-- %s\n", __func__);
7872 }
7873
7874 static int
7875 nfs4_layoutget_handle_exception(struct rpc_task *task,
7876                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
7877 {
7878         struct inode *inode = lgp->args.inode;
7879         struct nfs_server *server = NFS_SERVER(inode);
7880         struct pnfs_layout_hdr *lo;
7881         int nfs4err = task->tk_status;
7882         int err, status = 0;
7883         LIST_HEAD(head);
7884
7885         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7886
7887         switch (nfs4err) {
7888         case 0:
7889                 goto out;
7890
7891         /*
7892          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
7893          * on the file. set tk_status to -ENODATA to tell upper layer to
7894          * retry go inband.
7895          */
7896         case -NFS4ERR_LAYOUTUNAVAILABLE:
7897                 status = -ENODATA;
7898                 goto out;
7899         /*
7900          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7901          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7902          */
7903         case -NFS4ERR_BADLAYOUT:
7904                 status = -EOVERFLOW;
7905                 goto out;
7906         /*
7907          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7908          * (or clients) writing to the same RAID stripe except when
7909          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7910          *
7911          * Treat it like we would RECALLCONFLICT -- we retry for a little
7912          * while, and then eventually give up.
7913          */
7914         case -NFS4ERR_LAYOUTTRYLATER:
7915                 if (lgp->args.minlength == 0) {
7916                         status = -EOVERFLOW;
7917                         goto out;
7918                 }
7919                 status = -EBUSY;
7920                 break;
7921         case -NFS4ERR_RECALLCONFLICT:
7922                 status = -ERECALLCONFLICT;
7923                 break;
7924         case -NFS4ERR_EXPIRED:
7925         case -NFS4ERR_BAD_STATEID:
7926                 exception->timeout = 0;
7927                 spin_lock(&inode->i_lock);
7928                 lo = NFS_I(inode)->layout;
7929                 /* If the open stateid was bad, then recover it. */
7930                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
7931                     nfs4_stateid_match_other(&lgp->args.stateid,
7932                                         &lgp->args.ctx->state->stateid)) {
7933                         spin_unlock(&inode->i_lock);
7934                         exception->state = lgp->args.ctx->state;
7935                         break;
7936                 }
7937
7938                 /*
7939                  * Mark the bad layout state as invalid, then retry
7940                  */
7941                 pnfs_mark_layout_stateid_invalid(lo, &head);
7942                 spin_unlock(&inode->i_lock);
7943                 pnfs_free_lseg_list(&head);
7944                 status = -EAGAIN;
7945                 goto out;
7946         }
7947
7948         err = nfs4_handle_exception(server, nfs4err, exception);
7949         if (!status) {
7950                 if (exception->retry)
7951                         status = -EAGAIN;
7952                 else
7953                         status = err;
7954         }
7955 out:
7956         dprintk("<-- %s\n", __func__);
7957         return status;
7958 }
7959
7960 static size_t max_response_pages(struct nfs_server *server)
7961 {
7962         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7963         return nfs_page_array_len(0, max_resp_sz);
7964 }
7965
7966 static void nfs4_free_pages(struct page **pages, size_t size)
7967 {
7968         int i;
7969
7970         if (!pages)
7971                 return;
7972
7973         for (i = 0; i < size; i++) {
7974                 if (!pages[i])
7975                         break;
7976                 __free_page(pages[i]);
7977         }
7978         kfree(pages);
7979 }
7980
7981 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7982 {
7983         struct page **pages;
7984         int i;
7985
7986         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7987         if (!pages) {
7988                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7989                 return NULL;
7990         }
7991
7992         for (i = 0; i < size; i++) {
7993                 pages[i] = alloc_page(gfp_flags);
7994                 if (!pages[i]) {
7995                         dprintk("%s: failed to allocate page\n", __func__);
7996                         nfs4_free_pages(pages, size);
7997                         return NULL;
7998                 }
7999         }
8000
8001         return pages;
8002 }
8003
8004 static void nfs4_layoutget_release(void *calldata)
8005 {
8006         struct nfs4_layoutget *lgp = calldata;
8007         struct inode *inode = lgp->args.inode;
8008         struct nfs_server *server = NFS_SERVER(inode);
8009         size_t max_pages = max_response_pages(server);
8010
8011         dprintk("--> %s\n", __func__);
8012         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8013         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8014         put_nfs_open_context(lgp->args.ctx);
8015         kfree(calldata);
8016         dprintk("<-- %s\n", __func__);
8017 }
8018
8019 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8020         .rpc_call_prepare = nfs4_layoutget_prepare,
8021         .rpc_call_done = nfs4_layoutget_done,
8022         .rpc_release = nfs4_layoutget_release,
8023 };
8024
8025 struct pnfs_layout_segment *
8026 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8027 {
8028         struct inode *inode = lgp->args.inode;
8029         struct nfs_server *server = NFS_SERVER(inode);
8030         size_t max_pages = max_response_pages(server);
8031         struct rpc_task *task;
8032         struct rpc_message msg = {
8033                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8034                 .rpc_argp = &lgp->args,
8035                 .rpc_resp = &lgp->res,
8036                 .rpc_cred = lgp->cred,
8037         };
8038         struct rpc_task_setup task_setup_data = {
8039                 .rpc_client = server->client,
8040                 .rpc_message = &msg,
8041                 .callback_ops = &nfs4_layoutget_call_ops,
8042                 .callback_data = lgp,
8043                 .flags = RPC_TASK_ASYNC,
8044         };
8045         struct pnfs_layout_segment *lseg = NULL;
8046         struct nfs4_exception exception = {
8047                 .inode = inode,
8048                 .timeout = *timeout,
8049         };
8050         int status = 0;
8051
8052         dprintk("--> %s\n", __func__);
8053
8054         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8055         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8056
8057         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8058         if (!lgp->args.layout.pages) {
8059                 nfs4_layoutget_release(lgp);
8060                 return ERR_PTR(-ENOMEM);
8061         }
8062         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8063
8064         lgp->res.layoutp = &lgp->args.layout;
8065         lgp->res.seq_res.sr_slot = NULL;
8066         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8067
8068         task = rpc_run_task(&task_setup_data);
8069         if (IS_ERR(task))
8070                 return ERR_CAST(task);
8071         status = nfs4_wait_for_completion_rpc_task(task);
8072         if (status == 0) {
8073                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8074                 *timeout = exception.timeout;
8075         }
8076
8077         trace_nfs4_layoutget(lgp->args.ctx,
8078                         &lgp->args.range,
8079                         &lgp->res.range,
8080                         &lgp->res.stateid,
8081                         status);
8082
8083         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8084         if (status == 0 && lgp->res.layoutp->len)
8085                 lseg = pnfs_layout_process(lgp);
8086         rpc_put_task(task);
8087         dprintk("<-- %s status=%d\n", __func__, status);
8088         if (status)
8089                 return ERR_PTR(status);
8090         return lseg;
8091 }
8092
8093 static void
8094 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8095 {
8096         struct nfs4_layoutreturn *lrp = calldata;
8097
8098         dprintk("--> %s\n", __func__);
8099         nfs41_setup_sequence(lrp->clp->cl_session,
8100                         &lrp->args.seq_args,
8101                         &lrp->res.seq_res,
8102                         task);
8103 }
8104
8105 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8106 {
8107         struct nfs4_layoutreturn *lrp = calldata;
8108         struct nfs_server *server;
8109
8110         dprintk("--> %s\n", __func__);
8111
8112         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8113                 return;
8114
8115         server = NFS_SERVER(lrp->args.inode);
8116         switch (task->tk_status) {
8117         default:
8118                 task->tk_status = 0;
8119         case 0:
8120                 break;
8121         case -NFS4ERR_DELAY:
8122                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8123                         break;
8124                 rpc_restart_call_prepare(task);
8125                 return;
8126         }
8127         dprintk("<-- %s\n", __func__);
8128 }
8129
8130 static void nfs4_layoutreturn_release(void *calldata)
8131 {
8132         struct nfs4_layoutreturn *lrp = calldata;
8133         struct pnfs_layout_hdr *lo = lrp->args.layout;
8134         LIST_HEAD(freeme);
8135
8136         dprintk("--> %s\n", __func__);
8137         spin_lock(&lo->plh_inode->i_lock);
8138         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range,
8139                         be32_to_cpu(lrp->args.stateid.seqid));
8140         if (lrp->res.lrs_present && pnfs_layout_is_valid(lo))
8141                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8142         pnfs_clear_layoutreturn_waitbit(lo);
8143         spin_unlock(&lo->plh_inode->i_lock);
8144         pnfs_free_lseg_list(&freeme);
8145         pnfs_put_layout_hdr(lrp->args.layout);
8146         nfs_iput_and_deactive(lrp->inode);
8147         kfree(calldata);
8148         dprintk("<-- %s\n", __func__);
8149 }
8150
8151 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8152         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8153         .rpc_call_done = nfs4_layoutreturn_done,
8154         .rpc_release = nfs4_layoutreturn_release,
8155 };
8156
8157 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8158 {
8159         struct rpc_task *task;
8160         struct rpc_message msg = {
8161                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8162                 .rpc_argp = &lrp->args,
8163                 .rpc_resp = &lrp->res,
8164                 .rpc_cred = lrp->cred,
8165         };
8166         struct rpc_task_setup task_setup_data = {
8167                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8168                 .rpc_message = &msg,
8169                 .callback_ops = &nfs4_layoutreturn_call_ops,
8170                 .callback_data = lrp,
8171         };
8172         int status = 0;
8173
8174         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8175                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8176                         &task_setup_data.rpc_client, &msg);
8177
8178         dprintk("--> %s\n", __func__);
8179         if (!sync) {
8180                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8181                 if (!lrp->inode) {
8182                         nfs4_layoutreturn_release(lrp);
8183                         return -EAGAIN;
8184                 }
8185                 task_setup_data.flags |= RPC_TASK_ASYNC;
8186         }
8187         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8188         task = rpc_run_task(&task_setup_data);
8189         if (IS_ERR(task))
8190                 return PTR_ERR(task);
8191         if (sync)
8192                 status = task->tk_status;
8193         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8194         dprintk("<-- %s status=%d\n", __func__, status);
8195         rpc_put_task(task);
8196         return status;
8197 }
8198
8199 static int
8200 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8201                 struct pnfs_device *pdev,
8202                 struct rpc_cred *cred)
8203 {
8204         struct nfs4_getdeviceinfo_args args = {
8205                 .pdev = pdev,
8206                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8207                         NOTIFY_DEVICEID4_DELETE,
8208         };
8209         struct nfs4_getdeviceinfo_res res = {
8210                 .pdev = pdev,
8211         };
8212         struct rpc_message msg = {
8213                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8214                 .rpc_argp = &args,
8215                 .rpc_resp = &res,
8216                 .rpc_cred = cred,
8217         };
8218         int status;
8219
8220         dprintk("--> %s\n", __func__);
8221         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8222         if (res.notification & ~args.notify_types)
8223                 dprintk("%s: unsupported notification\n", __func__);
8224         if (res.notification != args.notify_types)
8225                 pdev->nocache = 1;
8226
8227         dprintk("<-- %s status=%d\n", __func__, status);
8228
8229         return status;
8230 }
8231
8232 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8233                 struct pnfs_device *pdev,
8234                 struct rpc_cred *cred)
8235 {
8236         struct nfs4_exception exception = { };
8237         int err;
8238
8239         do {
8240                 err = nfs4_handle_exception(server,
8241                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8242                                         &exception);
8243         } while (exception.retry);
8244         return err;
8245 }
8246 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8247
8248 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8249 {
8250         struct nfs4_layoutcommit_data *data = calldata;
8251         struct nfs_server *server = NFS_SERVER(data->args.inode);
8252         struct nfs4_session *session = nfs4_get_session(server);
8253
8254         nfs41_setup_sequence(session,
8255                         &data->args.seq_args,
8256                         &data->res.seq_res,
8257                         task);
8258 }
8259
8260 static void
8261 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8262 {
8263         struct nfs4_layoutcommit_data *data = calldata;
8264         struct nfs_server *server = NFS_SERVER(data->args.inode);
8265
8266         if (!nfs41_sequence_done(task, &data->res.seq_res))
8267                 return;
8268
8269         switch (task->tk_status) { /* Just ignore these failures */
8270         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8271         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8272         case -NFS4ERR_BADLAYOUT:     /* no layout */
8273         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8274                 task->tk_status = 0;
8275         case 0:
8276                 break;
8277         default:
8278                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8279                         rpc_restart_call_prepare(task);
8280                         return;
8281                 }
8282         }
8283 }
8284
8285 static void nfs4_layoutcommit_release(void *calldata)
8286 {
8287         struct nfs4_layoutcommit_data *data = calldata;
8288
8289         pnfs_cleanup_layoutcommit(data);
8290         nfs_post_op_update_inode_force_wcc(data->args.inode,
8291                                            data->res.fattr);
8292         put_rpccred(data->cred);
8293         nfs_iput_and_deactive(data->inode);
8294         kfree(data);
8295 }
8296
8297 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8298         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8299         .rpc_call_done = nfs4_layoutcommit_done,
8300         .rpc_release = nfs4_layoutcommit_release,
8301 };
8302
8303 int
8304 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8305 {
8306         struct rpc_message msg = {
8307                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8308                 .rpc_argp = &data->args,
8309                 .rpc_resp = &data->res,
8310                 .rpc_cred = data->cred,
8311         };
8312         struct rpc_task_setup task_setup_data = {
8313                 .task = &data->task,
8314                 .rpc_client = NFS_CLIENT(data->args.inode),
8315                 .rpc_message = &msg,
8316                 .callback_ops = &nfs4_layoutcommit_ops,
8317                 .callback_data = data,
8318         };
8319         struct rpc_task *task;
8320         int status = 0;
8321
8322         dprintk("NFS: initiating layoutcommit call. sync %d "
8323                 "lbw: %llu inode %lu\n", sync,
8324                 data->args.lastbytewritten,
8325                 data->args.inode->i_ino);
8326
8327         if (!sync) {
8328                 data->inode = nfs_igrab_and_active(data->args.inode);
8329                 if (data->inode == NULL) {
8330                         nfs4_layoutcommit_release(data);
8331                         return -EAGAIN;
8332                 }
8333                 task_setup_data.flags = RPC_TASK_ASYNC;
8334         }
8335         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8336         task = rpc_run_task(&task_setup_data);
8337         if (IS_ERR(task))
8338                 return PTR_ERR(task);
8339         if (sync)
8340                 status = task->tk_status;
8341         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8342         dprintk("%s: status %d\n", __func__, status);
8343         rpc_put_task(task);
8344         return status;
8345 }
8346
8347 /**
8348  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8349  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8350  */
8351 static int
8352 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8353                     struct nfs_fsinfo *info,
8354                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8355 {
8356         struct nfs41_secinfo_no_name_args args = {
8357                 .style = SECINFO_STYLE_CURRENT_FH,
8358         };
8359         struct nfs4_secinfo_res res = {
8360                 .flavors = flavors,
8361         };
8362         struct rpc_message msg = {
8363                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8364                 .rpc_argp = &args,
8365                 .rpc_resp = &res,
8366         };
8367         struct rpc_clnt *clnt = server->client;
8368         struct rpc_cred *cred = NULL;
8369         int status;
8370
8371         if (use_integrity) {
8372                 clnt = server->nfs_client->cl_rpcclient;
8373                 cred = nfs4_get_clid_cred(server->nfs_client);
8374                 msg.rpc_cred = cred;
8375         }
8376
8377         dprintk("--> %s\n", __func__);
8378         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8379                                 &res.seq_res, 0);
8380         dprintk("<-- %s status=%d\n", __func__, status);
8381
8382         if (cred)
8383                 put_rpccred(cred);
8384
8385         return status;
8386 }
8387
8388 static int
8389 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8390                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8391 {
8392         struct nfs4_exception exception = { };
8393         int err;
8394         do {
8395                 /* first try using integrity protection */
8396                 err = -NFS4ERR_WRONGSEC;
8397
8398                 /* try to use integrity protection with machine cred */
8399                 if (_nfs4_is_integrity_protected(server->nfs_client))
8400                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8401                                                           flavors, true);
8402
8403                 /*
8404                  * if unable to use integrity protection, or SECINFO with
8405                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8406                  * disallowed by spec, but exists in deployed servers) use
8407                  * the current filesystem's rpc_client and the user cred.
8408                  */
8409                 if (err == -NFS4ERR_WRONGSEC)
8410                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8411                                                           flavors, false);
8412
8413                 switch (err) {
8414                 case 0:
8415                 case -NFS4ERR_WRONGSEC:
8416                 case -ENOTSUPP:
8417                         goto out;
8418                 default:
8419                         err = nfs4_handle_exception(server, err, &exception);
8420                 }
8421         } while (exception.retry);
8422 out:
8423         return err;
8424 }
8425
8426 static int
8427 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8428                     struct nfs_fsinfo *info)
8429 {
8430         int err;
8431         struct page *page;
8432         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8433         struct nfs4_secinfo_flavors *flavors;
8434         struct nfs4_secinfo4 *secinfo;
8435         int i;
8436
8437         page = alloc_page(GFP_KERNEL);
8438         if (!page) {
8439                 err = -ENOMEM;
8440                 goto out;
8441         }
8442
8443         flavors = page_address(page);
8444         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8445
8446         /*
8447          * Fall back on "guess and check" method if
8448          * the server doesn't support SECINFO_NO_NAME
8449          */
8450         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8451                 err = nfs4_find_root_sec(server, fhandle, info);
8452                 goto out_freepage;
8453         }
8454         if (err)
8455                 goto out_freepage;
8456
8457         for (i = 0; i < flavors->num_flavors; i++) {
8458                 secinfo = &flavors->flavors[i];
8459
8460                 switch (secinfo->flavor) {
8461                 case RPC_AUTH_NULL:
8462                 case RPC_AUTH_UNIX:
8463                 case RPC_AUTH_GSS:
8464                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8465                                         &secinfo->flavor_info);
8466                         break;
8467                 default:
8468                         flavor = RPC_AUTH_MAXFLAVOR;
8469                         break;
8470                 }
8471
8472                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8473                         flavor = RPC_AUTH_MAXFLAVOR;
8474
8475                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8476                         err = nfs4_lookup_root_sec(server, fhandle,
8477                                                    info, flavor);
8478                         if (!err)
8479                                 break;
8480                 }
8481         }
8482
8483         if (flavor == RPC_AUTH_MAXFLAVOR)
8484                 err = -EPERM;
8485
8486 out_freepage:
8487         put_page(page);
8488         if (err == -EACCES)
8489                 return -EPERM;
8490 out:
8491         return err;
8492 }
8493
8494 static int _nfs41_test_stateid(struct nfs_server *server,
8495                 nfs4_stateid *stateid,
8496                 struct rpc_cred *cred)
8497 {
8498         int status;
8499         struct nfs41_test_stateid_args args = {
8500                 .stateid = stateid,
8501         };
8502         struct nfs41_test_stateid_res res;
8503         struct rpc_message msg = {
8504                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8505                 .rpc_argp = &args,
8506                 .rpc_resp = &res,
8507                 .rpc_cred = cred,
8508         };
8509         struct rpc_clnt *rpc_client = server->client;
8510
8511         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8512                 &rpc_client, &msg);
8513
8514         dprintk("NFS call  test_stateid %p\n", stateid);
8515         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8516         nfs4_set_sequence_privileged(&args.seq_args);
8517         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8518                         &args.seq_args, &res.seq_res);
8519         if (status != NFS_OK) {
8520                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8521                 return status;
8522         }
8523         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8524         return -res.status;
8525 }
8526
8527 /**
8528  * nfs41_test_stateid - perform a TEST_STATEID operation
8529  *
8530  * @server: server / transport on which to perform the operation
8531  * @stateid: state ID to test
8532  * @cred: credential
8533  *
8534  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8535  * Otherwise a negative NFS4ERR value is returned if the operation
8536  * failed or the state ID is not currently valid.
8537  */
8538 static int nfs41_test_stateid(struct nfs_server *server,
8539                 nfs4_stateid *stateid,
8540                 struct rpc_cred *cred)
8541 {
8542         struct nfs4_exception exception = { };
8543         int err;
8544         do {
8545                 err = _nfs41_test_stateid(server, stateid, cred);
8546                 if (err != -NFS4ERR_DELAY)
8547                         break;
8548                 nfs4_handle_exception(server, err, &exception);
8549         } while (exception.retry);
8550         return err;
8551 }
8552
8553 struct nfs_free_stateid_data {
8554         struct nfs_server *server;
8555         struct nfs41_free_stateid_args args;
8556         struct nfs41_free_stateid_res res;
8557 };
8558
8559 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8560 {
8561         struct nfs_free_stateid_data *data = calldata;
8562         nfs41_setup_sequence(nfs4_get_session(data->server),
8563                         &data->args.seq_args,
8564                         &data->res.seq_res,
8565                         task);
8566 }
8567
8568 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8569 {
8570         struct nfs_free_stateid_data *data = calldata;
8571
8572         nfs41_sequence_done(task, &data->res.seq_res);
8573
8574         switch (task->tk_status) {
8575         case -NFS4ERR_DELAY:
8576                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8577                         rpc_restart_call_prepare(task);
8578         }
8579 }
8580
8581 static void nfs41_free_stateid_release(void *calldata)
8582 {
8583         kfree(calldata);
8584 }
8585
8586 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8587         .rpc_call_prepare = nfs41_free_stateid_prepare,
8588         .rpc_call_done = nfs41_free_stateid_done,
8589         .rpc_release = nfs41_free_stateid_release,
8590 };
8591
8592 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8593                 nfs4_stateid *stateid,
8594                 struct rpc_cred *cred,
8595                 bool privileged)
8596 {
8597         struct rpc_message msg = {
8598                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8599                 .rpc_cred = cred,
8600         };
8601         struct rpc_task_setup task_setup = {
8602                 .rpc_client = server->client,
8603                 .rpc_message = &msg,
8604                 .callback_ops = &nfs41_free_stateid_ops,
8605                 .flags = RPC_TASK_ASYNC,
8606         };
8607         struct nfs_free_stateid_data *data;
8608
8609         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8610                 &task_setup.rpc_client, &msg);
8611
8612         dprintk("NFS call  free_stateid %p\n", stateid);
8613         data = kmalloc(sizeof(*data), GFP_NOFS);
8614         if (!data)
8615                 return ERR_PTR(-ENOMEM);
8616         data->server = server;
8617         nfs4_stateid_copy(&data->args.stateid, stateid);
8618
8619         task_setup.callback_data = data;
8620
8621         msg.rpc_argp = &data->args;
8622         msg.rpc_resp = &data->res;
8623         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8624         if (privileged)
8625                 nfs4_set_sequence_privileged(&data->args.seq_args);
8626
8627         return rpc_run_task(&task_setup);
8628 }
8629
8630 /**
8631  * nfs41_free_stateid - perform a FREE_STATEID operation
8632  *
8633  * @server: server / transport on which to perform the operation
8634  * @stateid: state ID to release
8635  * @cred: credential
8636  *
8637  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8638  * negative NFS4ERR value is returned.
8639  */
8640 static int nfs41_free_stateid(struct nfs_server *server,
8641                 nfs4_stateid *stateid,
8642                 struct rpc_cred *cred)
8643 {
8644         struct rpc_task *task;
8645         int ret;
8646
8647         task = _nfs41_free_stateid(server, stateid, cred, true);
8648         if (IS_ERR(task))
8649                 return PTR_ERR(task);
8650         ret = rpc_wait_for_completion_task(task);
8651         if (!ret)
8652                 ret = task->tk_status;
8653         rpc_put_task(task);
8654         return ret;
8655 }
8656
8657 static void
8658 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8659 {
8660         struct rpc_task *task;
8661         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8662
8663         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8664         nfs4_free_lock_state(server, lsp);
8665         if (IS_ERR(task))
8666                 return;
8667         rpc_put_task(task);
8668 }
8669
8670 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8671                 const nfs4_stateid *s2)
8672 {
8673         if (s1->type != s2->type)
8674                 return false;
8675
8676         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8677                 return false;
8678
8679         if (s1->seqid == s2->seqid)
8680                 return true;
8681         if (s1->seqid == 0 || s2->seqid == 0)
8682                 return true;
8683
8684         return false;
8685 }
8686
8687 #endif /* CONFIG_NFS_V4_1 */
8688
8689 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8690                 const nfs4_stateid *s2)
8691 {
8692         return nfs4_stateid_match(s1, s2);
8693 }
8694
8695
8696 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8697         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8698         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8699         .recover_open   = nfs4_open_reclaim,
8700         .recover_lock   = nfs4_lock_reclaim,
8701         .establish_clid = nfs4_init_clientid,
8702         .detect_trunking = nfs40_discover_server_trunking,
8703 };
8704
8705 #if defined(CONFIG_NFS_V4_1)
8706 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8707         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8708         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8709         .recover_open   = nfs4_open_reclaim,
8710         .recover_lock   = nfs4_lock_reclaim,
8711         .establish_clid = nfs41_init_clientid,
8712         .reclaim_complete = nfs41_proc_reclaim_complete,
8713         .detect_trunking = nfs41_discover_server_trunking,
8714 };
8715 #endif /* CONFIG_NFS_V4_1 */
8716
8717 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8718         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8719         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8720         .recover_open   = nfs40_open_expired,
8721         .recover_lock   = nfs4_lock_expired,
8722         .establish_clid = nfs4_init_clientid,
8723 };
8724
8725 #if defined(CONFIG_NFS_V4_1)
8726 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8727         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8728         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8729         .recover_open   = nfs41_open_expired,
8730         .recover_lock   = nfs41_lock_expired,
8731         .establish_clid = nfs41_init_clientid,
8732 };
8733 #endif /* CONFIG_NFS_V4_1 */
8734
8735 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8736         .sched_state_renewal = nfs4_proc_async_renew,
8737         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8738         .renew_lease = nfs4_proc_renew,
8739 };
8740
8741 #if defined(CONFIG_NFS_V4_1)
8742 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8743         .sched_state_renewal = nfs41_proc_async_sequence,
8744         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8745         .renew_lease = nfs4_proc_sequence,
8746 };
8747 #endif
8748
8749 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8750         .get_locations = _nfs40_proc_get_locations,
8751         .fsid_present = _nfs40_proc_fsid_present,
8752 };
8753
8754 #if defined(CONFIG_NFS_V4_1)
8755 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8756         .get_locations = _nfs41_proc_get_locations,
8757         .fsid_present = _nfs41_proc_fsid_present,
8758 };
8759 #endif  /* CONFIG_NFS_V4_1 */
8760
8761 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8762         .minor_version = 0,
8763         .init_caps = NFS_CAP_READDIRPLUS
8764                 | NFS_CAP_ATOMIC_OPEN
8765                 | NFS_CAP_POSIX_LOCK,
8766         .init_client = nfs40_init_client,
8767         .shutdown_client = nfs40_shutdown_client,
8768         .match_stateid = nfs4_match_stateid,
8769         .find_root_sec = nfs4_find_root_sec,
8770         .free_lock_state = nfs4_release_lockowner,
8771         .alloc_seqid = nfs_alloc_seqid,
8772         .call_sync_ops = &nfs40_call_sync_ops,
8773         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8774         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8775         .state_renewal_ops = &nfs40_state_renewal_ops,
8776         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8777 };
8778
8779 #if defined(CONFIG_NFS_V4_1)
8780 static struct nfs_seqid *
8781 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8782 {
8783         return NULL;
8784 }
8785
8786 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8787         .minor_version = 1,
8788         .init_caps = NFS_CAP_READDIRPLUS
8789                 | NFS_CAP_ATOMIC_OPEN
8790                 | NFS_CAP_POSIX_LOCK
8791                 | NFS_CAP_STATEID_NFSV41
8792                 | NFS_CAP_ATOMIC_OPEN_V1,
8793         .init_client = nfs41_init_client,
8794         .shutdown_client = nfs41_shutdown_client,
8795         .match_stateid = nfs41_match_stateid,
8796         .find_root_sec = nfs41_find_root_sec,
8797         .free_lock_state = nfs41_free_lock_state,
8798         .alloc_seqid = nfs_alloc_no_seqid,
8799         .call_sync_ops = &nfs41_call_sync_ops,
8800         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8801         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8802         .state_renewal_ops = &nfs41_state_renewal_ops,
8803         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8804 };
8805 #endif
8806
8807 #if defined(CONFIG_NFS_V4_2)
8808 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8809         .minor_version = 2,
8810         .init_caps = NFS_CAP_READDIRPLUS
8811                 | NFS_CAP_ATOMIC_OPEN
8812                 | NFS_CAP_POSIX_LOCK
8813                 | NFS_CAP_STATEID_NFSV41
8814                 | NFS_CAP_ATOMIC_OPEN_V1
8815                 | NFS_CAP_ALLOCATE
8816                 | NFS_CAP_COPY
8817                 | NFS_CAP_DEALLOCATE
8818                 | NFS_CAP_SEEK
8819                 | NFS_CAP_LAYOUTSTATS
8820                 | NFS_CAP_CLONE,
8821         .init_client = nfs41_init_client,
8822         .shutdown_client = nfs41_shutdown_client,
8823         .match_stateid = nfs41_match_stateid,
8824         .find_root_sec = nfs41_find_root_sec,
8825         .free_lock_state = nfs41_free_lock_state,
8826         .call_sync_ops = &nfs41_call_sync_ops,
8827         .alloc_seqid = nfs_alloc_no_seqid,
8828         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8829         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8830         .state_renewal_ops = &nfs41_state_renewal_ops,
8831         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8832 };
8833 #endif
8834
8835 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8836         [0] = &nfs_v4_0_minor_ops,
8837 #if defined(CONFIG_NFS_V4_1)
8838         [1] = &nfs_v4_1_minor_ops,
8839 #endif
8840 #if defined(CONFIG_NFS_V4_2)
8841         [2] = &nfs_v4_2_minor_ops,
8842 #endif
8843 };
8844
8845 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
8846 {
8847         ssize_t error, error2;
8848
8849         error = generic_listxattr(dentry, list, size);
8850         if (error < 0)
8851                 return error;
8852         if (list) {
8853                 list += error;
8854                 size -= error;
8855         }
8856
8857         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
8858         if (error2 < 0)
8859                 return error2;
8860         return error + error2;
8861 }
8862
8863 static const struct inode_operations nfs4_dir_inode_operations = {
8864         .create         = nfs_create,
8865         .lookup         = nfs_lookup,
8866         .atomic_open    = nfs_atomic_open,
8867         .link           = nfs_link,
8868         .unlink         = nfs_unlink,
8869         .symlink        = nfs_symlink,
8870         .mkdir          = nfs_mkdir,
8871         .rmdir          = nfs_rmdir,
8872         .mknod          = nfs_mknod,
8873         .rename         = nfs_rename,
8874         .permission     = nfs_permission,
8875         .getattr        = nfs_getattr,
8876         .setattr        = nfs_setattr,
8877         .getxattr       = generic_getxattr,
8878         .setxattr       = generic_setxattr,
8879         .listxattr      = nfs4_listxattr,
8880         .removexattr    = generic_removexattr,
8881 };
8882
8883 static const struct inode_operations nfs4_file_inode_operations = {
8884         .permission     = nfs_permission,
8885         .getattr        = nfs_getattr,
8886         .setattr        = nfs_setattr,
8887         .getxattr       = generic_getxattr,
8888         .setxattr       = generic_setxattr,
8889         .listxattr      = nfs4_listxattr,
8890         .removexattr    = generic_removexattr,
8891 };
8892
8893 const struct nfs_rpc_ops nfs_v4_clientops = {
8894         .version        = 4,                    /* protocol version */
8895         .dentry_ops     = &nfs4_dentry_operations,
8896         .dir_inode_ops  = &nfs4_dir_inode_operations,
8897         .file_inode_ops = &nfs4_file_inode_operations,
8898         .file_ops       = &nfs4_file_operations,
8899         .getroot        = nfs4_proc_get_root,
8900         .submount       = nfs4_submount,
8901         .try_mount      = nfs4_try_mount,
8902         .getattr        = nfs4_proc_getattr,
8903         .setattr        = nfs4_proc_setattr,
8904         .lookup         = nfs4_proc_lookup,
8905         .access         = nfs4_proc_access,
8906         .readlink       = nfs4_proc_readlink,
8907         .create         = nfs4_proc_create,
8908         .remove         = nfs4_proc_remove,
8909         .unlink_setup   = nfs4_proc_unlink_setup,
8910         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8911         .unlink_done    = nfs4_proc_unlink_done,
8912         .rename_setup   = nfs4_proc_rename_setup,
8913         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8914         .rename_done    = nfs4_proc_rename_done,
8915         .link           = nfs4_proc_link,
8916         .symlink        = nfs4_proc_symlink,
8917         .mkdir          = nfs4_proc_mkdir,
8918         .rmdir          = nfs4_proc_remove,
8919         .readdir        = nfs4_proc_readdir,
8920         .mknod          = nfs4_proc_mknod,
8921         .statfs         = nfs4_proc_statfs,
8922         .fsinfo         = nfs4_proc_fsinfo,
8923         .pathconf       = nfs4_proc_pathconf,
8924         .set_capabilities = nfs4_server_capabilities,
8925         .decode_dirent  = nfs4_decode_dirent,
8926         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8927         .read_setup     = nfs4_proc_read_setup,
8928         .read_done      = nfs4_read_done,
8929         .write_setup    = nfs4_proc_write_setup,
8930         .write_done     = nfs4_write_done,
8931         .commit_setup   = nfs4_proc_commit_setup,
8932         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8933         .commit_done    = nfs4_commit_done,
8934         .lock           = nfs4_proc_lock,
8935         .clear_acl_cache = nfs4_zap_acl_attr,
8936         .close_context  = nfs4_close_context,
8937         .open_context   = nfs4_atomic_open,
8938         .have_delegation = nfs4_have_delegation,
8939         .return_delegation = nfs4_inode_return_delegation,
8940         .alloc_client   = nfs4_alloc_client,
8941         .init_client    = nfs4_init_client,
8942         .free_client    = nfs4_free_client,
8943         .create_server  = nfs4_create_server,
8944         .clone_server   = nfs_clone_server,
8945 };
8946
8947 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8948         .name   = XATTR_NAME_NFSV4_ACL,
8949         .list   = nfs4_xattr_list_nfs4_acl,
8950         .get    = nfs4_xattr_get_nfs4_acl,
8951         .set    = nfs4_xattr_set_nfs4_acl,
8952 };
8953
8954 const struct xattr_handler *nfs4_xattr_handlers[] = {
8955         &nfs4_xattr_nfs4_acl_handler,
8956 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8957         &nfs4_xattr_nfs4_label_handler,
8958 #endif
8959         NULL
8960 };
8961
8962 /*
8963  * Local variables:
8964  *  c-basic-offset: 8
8965  * End:
8966  */