]> asedeno.scripts.mit.edu Git - linux.git/blob - fs/nfs/nfs4proc.c
Merge branches 'pm-core', 'pm-qos', 'pm-domains' and 'pm-opp'
[linux.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57
58 #include "nfs4_fs.h"
59 #include "delegation.h"
60 #include "internal.h"
61 #include "iostat.h"
62 #include "callback.h"
63 #include "pnfs.h"
64 #include "netns.h"
65 #include "nfs4idmap.h"
66 #include "nfs4session.h"
67 #include "fscache.h"
68
69 #include "nfs4trace.h"
70
71 #define NFSDBG_FACILITY         NFSDBG_PROC
72
73 #define NFS4_POLL_RETRY_MIN     (HZ/10)
74 #define NFS4_POLL_RETRY_MAX     (15*HZ)
75
76 /* file attributes which can be mapped to nfs attributes */
77 #define NFS4_VALID_ATTRS (ATTR_MODE \
78         | ATTR_UID \
79         | ATTR_GID \
80         | ATTR_SIZE \
81         | ATTR_ATIME \
82         | ATTR_MTIME \
83         | ATTR_CTIME \
84         | ATTR_ATIME_SET \
85         | ATTR_MTIME_SET)
86
87 struct nfs4_opendata;
88 static int _nfs4_proc_open(struct nfs4_opendata *data);
89 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
90 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
91 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
92 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
93 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
94 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
95                             struct nfs_fattr *fattr, struct iattr *sattr,
96                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
97                             struct nfs4_label *olabel);
98 #ifdef CONFIG_NFS_V4_1
99 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
100                 struct rpc_cred *);
101 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
102                 struct rpc_cred *, bool);
103 #endif
104
105 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
106 static inline struct nfs4_label *
107 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
108         struct iattr *sattr, struct nfs4_label *label)
109 {
110         int err;
111
112         if (label == NULL)
113                 return NULL;
114
115         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
116                 return NULL;
117
118         err = security_dentry_init_security(dentry, sattr->ia_mode,
119                                 &dentry->d_name, (void **)&label->label, &label->len);
120         if (err == 0)
121                 return label;
122
123         return NULL;
124 }
125 static inline void
126 nfs4_label_release_security(struct nfs4_label *label)
127 {
128         if (label)
129                 security_release_secctx(label->label, label->len);
130 }
131 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
132 {
133         if (label)
134                 return server->attr_bitmask;
135
136         return server->attr_bitmask_nl;
137 }
138 #else
139 static inline struct nfs4_label *
140 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
141         struct iattr *sattr, struct nfs4_label *l)
142 { return NULL; }
143 static inline void
144 nfs4_label_release_security(struct nfs4_label *label)
145 { return; }
146 static inline u32 *
147 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
148 { return server->attr_bitmask; }
149 #endif
150
151 /* Prevent leaks of NFSv4 errors into userland */
152 static int nfs4_map_errors(int err)
153 {
154         if (err >= -1000)
155                 return err;
156         switch (err) {
157         case -NFS4ERR_RESOURCE:
158         case -NFS4ERR_LAYOUTTRYLATER:
159         case -NFS4ERR_RECALLCONFLICT:
160                 return -EREMOTEIO;
161         case -NFS4ERR_WRONGSEC:
162         case -NFS4ERR_WRONG_CRED:
163                 return -EPERM;
164         case -NFS4ERR_BADOWNER:
165         case -NFS4ERR_BADNAME:
166                 return -EINVAL;
167         case -NFS4ERR_SHARE_DENIED:
168                 return -EACCES;
169         case -NFS4ERR_MINOR_VERS_MISMATCH:
170                 return -EPROTONOSUPPORT;
171         case -NFS4ERR_FILE_OPEN:
172                 return -EBUSY;
173         default:
174                 dprintk("%s could not handle NFSv4 error %d\n",
175                                 __func__, -err);
176                 break;
177         }
178         return -EIO;
179 }
180
181 /*
182  * This is our standard bitmap for GETATTR requests.
183  */
184 const u32 nfs4_fattr_bitmap[3] = {
185         FATTR4_WORD0_TYPE
186         | FATTR4_WORD0_CHANGE
187         | FATTR4_WORD0_SIZE
188         | FATTR4_WORD0_FSID
189         | FATTR4_WORD0_FILEID,
190         FATTR4_WORD1_MODE
191         | FATTR4_WORD1_NUMLINKS
192         | FATTR4_WORD1_OWNER
193         | FATTR4_WORD1_OWNER_GROUP
194         | FATTR4_WORD1_RAWDEV
195         | FATTR4_WORD1_SPACE_USED
196         | FATTR4_WORD1_TIME_ACCESS
197         | FATTR4_WORD1_TIME_METADATA
198         | FATTR4_WORD1_TIME_MODIFY
199         | FATTR4_WORD1_MOUNTED_ON_FILEID,
200 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
201         FATTR4_WORD2_SECURITY_LABEL
202 #endif
203 };
204
205 static const u32 nfs4_pnfs_open_bitmap[3] = {
206         FATTR4_WORD0_TYPE
207         | FATTR4_WORD0_CHANGE
208         | FATTR4_WORD0_SIZE
209         | FATTR4_WORD0_FSID
210         | FATTR4_WORD0_FILEID,
211         FATTR4_WORD1_MODE
212         | FATTR4_WORD1_NUMLINKS
213         | FATTR4_WORD1_OWNER
214         | FATTR4_WORD1_OWNER_GROUP
215         | FATTR4_WORD1_RAWDEV
216         | FATTR4_WORD1_SPACE_USED
217         | FATTR4_WORD1_TIME_ACCESS
218         | FATTR4_WORD1_TIME_METADATA
219         | FATTR4_WORD1_TIME_MODIFY,
220         FATTR4_WORD2_MDSTHRESHOLD
221 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
222         | FATTR4_WORD2_SECURITY_LABEL
223 #endif
224 };
225
226 static const u32 nfs4_open_noattr_bitmap[3] = {
227         FATTR4_WORD0_TYPE
228         | FATTR4_WORD0_FILEID,
229 };
230
231 const u32 nfs4_statfs_bitmap[3] = {
232         FATTR4_WORD0_FILES_AVAIL
233         | FATTR4_WORD0_FILES_FREE
234         | FATTR4_WORD0_FILES_TOTAL,
235         FATTR4_WORD1_SPACE_AVAIL
236         | FATTR4_WORD1_SPACE_FREE
237         | FATTR4_WORD1_SPACE_TOTAL
238 };
239
240 const u32 nfs4_pathconf_bitmap[3] = {
241         FATTR4_WORD0_MAXLINK
242         | FATTR4_WORD0_MAXNAME,
243         0
244 };
245
246 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
247                         | FATTR4_WORD0_MAXREAD
248                         | FATTR4_WORD0_MAXWRITE
249                         | FATTR4_WORD0_LEASE_TIME,
250                         FATTR4_WORD1_TIME_DELTA
251                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
252                         FATTR4_WORD2_LAYOUT_BLKSIZE
253                         | FATTR4_WORD2_CLONE_BLKSIZE
254 };
255
256 const u32 nfs4_fs_locations_bitmap[3] = {
257         FATTR4_WORD0_TYPE
258         | FATTR4_WORD0_CHANGE
259         | FATTR4_WORD0_SIZE
260         | FATTR4_WORD0_FSID
261         | FATTR4_WORD0_FILEID
262         | FATTR4_WORD0_FS_LOCATIONS,
263         FATTR4_WORD1_MODE
264         | FATTR4_WORD1_NUMLINKS
265         | FATTR4_WORD1_OWNER
266         | FATTR4_WORD1_OWNER_GROUP
267         | FATTR4_WORD1_RAWDEV
268         | FATTR4_WORD1_SPACE_USED
269         | FATTR4_WORD1_TIME_ACCESS
270         | FATTR4_WORD1_TIME_METADATA
271         | FATTR4_WORD1_TIME_MODIFY
272         | FATTR4_WORD1_MOUNTED_ON_FILEID,
273 };
274
275 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
276                 struct nfs4_readdir_arg *readdir)
277 {
278         __be32 *start, *p;
279
280         if (cookie > 2) {
281                 readdir->cookie = cookie;
282                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
283                 return;
284         }
285
286         readdir->cookie = 0;
287         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
288         if (cookie == 2)
289                 return;
290         
291         /*
292          * NFSv4 servers do not return entries for '.' and '..'
293          * Therefore, we fake these entries here.  We let '.'
294          * have cookie 0 and '..' have cookie 1.  Note that
295          * when talking to the server, we always send cookie 0
296          * instead of 1 or 2.
297          */
298         start = p = kmap_atomic(*readdir->pages);
299         
300         if (cookie == 0) {
301                 *p++ = xdr_one;                                  /* next */
302                 *p++ = xdr_zero;                   /* cookie, first word */
303                 *p++ = xdr_one;                   /* cookie, second word */
304                 *p++ = xdr_one;                             /* entry len */
305                 memcpy(p, ".\0\0\0", 4);                        /* entry */
306                 p++;
307                 *p++ = xdr_one;                         /* bitmap length */
308                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
309                 *p++ = htonl(8);              /* attribute buffer length */
310                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
311         }
312         
313         *p++ = xdr_one;                                  /* next */
314         *p++ = xdr_zero;                   /* cookie, first word */
315         *p++ = xdr_two;                   /* cookie, second word */
316         *p++ = xdr_two;                             /* entry len */
317         memcpy(p, "..\0\0", 4);                         /* entry */
318         p++;
319         *p++ = xdr_one;                         /* bitmap length */
320         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
321         *p++ = htonl(8);              /* attribute buffer length */
322         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
323
324         readdir->pgbase = (char *)p - (char *)start;
325         readdir->count -= readdir->pgbase;
326         kunmap_atomic(start);
327 }
328
329 static void nfs4_test_and_free_stateid(struct nfs_server *server,
330                 nfs4_stateid *stateid,
331                 struct rpc_cred *cred)
332 {
333         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
334
335         ops->test_and_free_expired(server, stateid, cred);
336 }
337
338 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
339                 nfs4_stateid *stateid,
340                 struct rpc_cred *cred)
341 {
342         stateid->type = NFS4_REVOKED_STATEID_TYPE;
343         nfs4_test_and_free_stateid(server, stateid, cred);
344 }
345
346 static void nfs4_free_revoked_stateid(struct nfs_server *server,
347                 const nfs4_stateid *stateid,
348                 struct rpc_cred *cred)
349 {
350         nfs4_stateid tmp;
351
352         nfs4_stateid_copy(&tmp, stateid);
353         __nfs4_free_revoked_stateid(server, &tmp, cred);
354 }
355
356 static long nfs4_update_delay(long *timeout)
357 {
358         long ret;
359         if (!timeout)
360                 return NFS4_POLL_RETRY_MAX;
361         if (*timeout <= 0)
362                 *timeout = NFS4_POLL_RETRY_MIN;
363         if (*timeout > NFS4_POLL_RETRY_MAX)
364                 *timeout = NFS4_POLL_RETRY_MAX;
365         ret = *timeout;
366         *timeout <<= 1;
367         return ret;
368 }
369
370 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
371 {
372         int res = 0;
373
374         might_sleep();
375
376         freezable_schedule_timeout_killable_unsafe(
377                 nfs4_update_delay(timeout));
378         if (fatal_signal_pending(current))
379                 res = -ERESTARTSYS;
380         return res;
381 }
382
383 /* This is the error handling routine for processes that are allowed
384  * to sleep.
385  */
386 static int nfs4_do_handle_exception(struct nfs_server *server,
387                 int errorcode, struct nfs4_exception *exception)
388 {
389         struct nfs_client *clp = server->nfs_client;
390         struct nfs4_state *state = exception->state;
391         const nfs4_stateid *stateid = exception->stateid;
392         struct inode *inode = exception->inode;
393         int ret = errorcode;
394
395         exception->delay = 0;
396         exception->recovering = 0;
397         exception->retry = 0;
398
399         if (stateid == NULL && state != NULL)
400                 stateid = &state->stateid;
401
402         switch(errorcode) {
403                 case 0:
404                         return 0;
405                 case -NFS4ERR_DELEG_REVOKED:
406                 case -NFS4ERR_ADMIN_REVOKED:
407                 case -NFS4ERR_EXPIRED:
408                 case -NFS4ERR_BAD_STATEID:
409                         if (inode != NULL && stateid != NULL) {
410                                 nfs_inode_find_state_and_recover(inode,
411                                                 stateid);
412                                 goto wait_on_recovery;
413                         }
414                 case -NFS4ERR_OPENMODE:
415                         if (inode) {
416                                 int err;
417
418                                 err = nfs_async_inode_return_delegation(inode,
419                                                 stateid);
420                                 if (err == 0)
421                                         goto wait_on_recovery;
422                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
423                                         exception->retry = 1;
424                                         break;
425                                 }
426                         }
427                         if (state == NULL)
428                                 break;
429                         ret = nfs4_schedule_stateid_recovery(server, state);
430                         if (ret < 0)
431                                 break;
432                         goto wait_on_recovery;
433                 case -NFS4ERR_STALE_STATEID:
434                 case -NFS4ERR_STALE_CLIENTID:
435                         nfs4_schedule_lease_recovery(clp);
436                         goto wait_on_recovery;
437                 case -NFS4ERR_MOVED:
438                         ret = nfs4_schedule_migration_recovery(server);
439                         if (ret < 0)
440                                 break;
441                         goto wait_on_recovery;
442                 case -NFS4ERR_LEASE_MOVED:
443                         nfs4_schedule_lease_moved_recovery(clp);
444                         goto wait_on_recovery;
445 #if defined(CONFIG_NFS_V4_1)
446                 case -NFS4ERR_BADSESSION:
447                 case -NFS4ERR_BADSLOT:
448                 case -NFS4ERR_BAD_HIGH_SLOT:
449                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
450                 case -NFS4ERR_DEADSESSION:
451                 case -NFS4ERR_SEQ_FALSE_RETRY:
452                 case -NFS4ERR_SEQ_MISORDERED:
453                         dprintk("%s ERROR: %d Reset session\n", __func__,
454                                 errorcode);
455                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
456                         goto wait_on_recovery;
457 #endif /* defined(CONFIG_NFS_V4_1) */
458                 case -NFS4ERR_FILE_OPEN:
459                         if (exception->timeout > HZ) {
460                                 /* We have retried a decent amount, time to
461                                  * fail
462                                  */
463                                 ret = -EBUSY;
464                                 break;
465                         }
466                 case -NFS4ERR_DELAY:
467                         nfs_inc_server_stats(server, NFSIOS_DELAY);
468                 case -NFS4ERR_GRACE:
469                 case -NFS4ERR_LAYOUTTRYLATER:
470                 case -NFS4ERR_RECALLCONFLICT:
471                         exception->delay = 1;
472                         return 0;
473
474                 case -NFS4ERR_RETRY_UNCACHED_REP:
475                 case -NFS4ERR_OLD_STATEID:
476                         exception->retry = 1;
477                         break;
478                 case -NFS4ERR_BADOWNER:
479                         /* The following works around a Linux server bug! */
480                 case -NFS4ERR_BADNAME:
481                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
482                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
483                                 exception->retry = 1;
484                                 printk(KERN_WARNING "NFS: v4 server %s "
485                                                 "does not accept raw "
486                                                 "uid/gids. "
487                                                 "Reenabling the idmapper.\n",
488                                                 server->nfs_client->cl_hostname);
489                         }
490         }
491         /* We failed to handle the error */
492         return nfs4_map_errors(ret);
493 wait_on_recovery:
494         exception->recovering = 1;
495         return 0;
496 }
497
498 /* This is the error handling routine for processes that are allowed
499  * to sleep.
500  */
501 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
502 {
503         struct nfs_client *clp = server->nfs_client;
504         int ret;
505
506         ret = nfs4_do_handle_exception(server, errorcode, exception);
507         if (exception->delay) {
508                 ret = nfs4_delay(server->client, &exception->timeout);
509                 goto out_retry;
510         }
511         if (exception->recovering) {
512                 ret = nfs4_wait_clnt_recover(clp);
513                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
514                         return -EIO;
515                 goto out_retry;
516         }
517         return ret;
518 out_retry:
519         if (ret == 0)
520                 exception->retry = 1;
521         return ret;
522 }
523
524 static int
525 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
526                 int errorcode, struct nfs4_exception *exception)
527 {
528         struct nfs_client *clp = server->nfs_client;
529         int ret;
530
531         ret = nfs4_do_handle_exception(server, errorcode, exception);
532         if (exception->delay) {
533                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
534                 goto out_retry;
535         }
536         if (exception->recovering) {
537                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
538                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
539                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
540                 goto out_retry;
541         }
542         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
543                 ret = -EIO;
544         return ret;
545 out_retry:
546         if (ret == 0)
547                 exception->retry = 1;
548         return ret;
549 }
550
551 static int
552 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
553                         struct nfs4_state *state, long *timeout)
554 {
555         struct nfs4_exception exception = {
556                 .state = state,
557         };
558
559         if (task->tk_status >= 0)
560                 return 0;
561         if (timeout)
562                 exception.timeout = *timeout;
563         task->tk_status = nfs4_async_handle_exception(task, server,
564                         task->tk_status,
565                         &exception);
566         if (exception.delay && timeout)
567                 *timeout = exception.timeout;
568         if (exception.retry)
569                 return -EAGAIN;
570         return 0;
571 }
572
573 /*
574  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
575  * or 'false' otherwise.
576  */
577 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
578 {
579         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
580
581         if (flavor == RPC_AUTH_GSS_KRB5I ||
582             flavor == RPC_AUTH_GSS_KRB5P)
583                 return true;
584
585         return false;
586 }
587
588 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
589 {
590         spin_lock(&clp->cl_lock);
591         if (time_before(clp->cl_last_renewal,timestamp))
592                 clp->cl_last_renewal = timestamp;
593         spin_unlock(&clp->cl_lock);
594 }
595
596 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
597 {
598         struct nfs_client *clp = server->nfs_client;
599
600         if (!nfs4_has_session(clp))
601                 do_renew_lease(clp, timestamp);
602 }
603
604 struct nfs4_call_sync_data {
605         const struct nfs_server *seq_server;
606         struct nfs4_sequence_args *seq_args;
607         struct nfs4_sequence_res *seq_res;
608 };
609
610 void nfs4_init_sequence(struct nfs4_sequence_args *args,
611                         struct nfs4_sequence_res *res, int cache_reply)
612 {
613         args->sa_slot = NULL;
614         args->sa_cache_this = cache_reply;
615         args->sa_privileged = 0;
616
617         res->sr_slot = NULL;
618 }
619
620 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
621 {
622         args->sa_privileged = 1;
623 }
624
625 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
626                          struct nfs4_sequence_args *args,
627                          struct nfs4_sequence_res *res,
628                          struct rpc_task *task)
629 {
630         struct nfs4_slot *slot;
631
632         /* slot already allocated? */
633         if (res->sr_slot != NULL)
634                 goto out_start;
635
636         spin_lock(&tbl->slot_tbl_lock);
637         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
638                 goto out_sleep;
639
640         slot = nfs4_alloc_slot(tbl);
641         if (IS_ERR(slot)) {
642                 if (slot == ERR_PTR(-ENOMEM))
643                         task->tk_timeout = HZ >> 2;
644                 goto out_sleep;
645         }
646         spin_unlock(&tbl->slot_tbl_lock);
647
648         slot->privileged = args->sa_privileged ? 1 : 0;
649         args->sa_slot = slot;
650         res->sr_slot = slot;
651
652 out_start:
653         rpc_call_start(task);
654         return 0;
655
656 out_sleep:
657         if (args->sa_privileged)
658                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
659                                 NULL, RPC_PRIORITY_PRIVILEGED);
660         else
661                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
662         spin_unlock(&tbl->slot_tbl_lock);
663         return -EAGAIN;
664 }
665 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
666
667 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
668 {
669         struct nfs4_slot *slot = res->sr_slot;
670         struct nfs4_slot_table *tbl;
671
672         tbl = slot->table;
673         spin_lock(&tbl->slot_tbl_lock);
674         if (!nfs41_wake_and_assign_slot(tbl, slot))
675                 nfs4_free_slot(tbl, slot);
676         spin_unlock(&tbl->slot_tbl_lock);
677
678         res->sr_slot = NULL;
679 }
680
681 static int nfs40_sequence_done(struct rpc_task *task,
682                                struct nfs4_sequence_res *res)
683 {
684         if (res->sr_slot != NULL)
685                 nfs40_sequence_free_slot(res);
686         return 1;
687 }
688
689 #if defined(CONFIG_NFS_V4_1)
690
691 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
692 {
693         struct nfs4_session *session;
694         struct nfs4_slot_table *tbl;
695         struct nfs4_slot *slot = res->sr_slot;
696         bool send_new_highest_used_slotid = false;
697
698         tbl = slot->table;
699         session = tbl->session;
700
701         /* Bump the slot sequence number */
702         if (slot->seq_done)
703                 slot->seq_nr++;
704         slot->seq_done = 0;
705
706         spin_lock(&tbl->slot_tbl_lock);
707         /* Be nice to the server: try to ensure that the last transmitted
708          * value for highest_user_slotid <= target_highest_slotid
709          */
710         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
711                 send_new_highest_used_slotid = true;
712
713         if (nfs41_wake_and_assign_slot(tbl, slot)) {
714                 send_new_highest_used_slotid = false;
715                 goto out_unlock;
716         }
717         nfs4_free_slot(tbl, slot);
718
719         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
720                 send_new_highest_used_slotid = false;
721 out_unlock:
722         spin_unlock(&tbl->slot_tbl_lock);
723         res->sr_slot = NULL;
724         if (send_new_highest_used_slotid)
725                 nfs41_notify_server(session->clp);
726         if (waitqueue_active(&tbl->slot_waitq))
727                 wake_up_all(&tbl->slot_waitq);
728 }
729
730 static int nfs41_sequence_process(struct rpc_task *task,
731                 struct nfs4_sequence_res *res)
732 {
733         struct nfs4_session *session;
734         struct nfs4_slot *slot = res->sr_slot;
735         struct nfs_client *clp;
736         bool interrupted = false;
737         int ret = 1;
738
739         if (slot == NULL)
740                 goto out_noaction;
741         /* don't increment the sequence number if the task wasn't sent */
742         if (!RPC_WAS_SENT(task))
743                 goto out;
744
745         session = slot->table->session;
746
747         if (slot->interrupted) {
748                 slot->interrupted = 0;
749                 interrupted = true;
750         }
751
752         trace_nfs4_sequence_done(session, res);
753         /* Check the SEQUENCE operation status */
754         switch (res->sr_status) {
755         case 0:
756                 /* If previous op on slot was interrupted and we reused
757                  * the seq# and got a reply from the cache, then retry
758                  */
759                 if (task->tk_status == -EREMOTEIO && interrupted) {
760                         ++slot->seq_nr;
761                         goto retry_nowait;
762                 }
763                 /* Update the slot's sequence and clientid lease timer */
764                 slot->seq_done = 1;
765                 clp = session->clp;
766                 do_renew_lease(clp, res->sr_timestamp);
767                 /* Check sequence flags */
768                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
769                                 !!slot->privileged);
770                 nfs41_update_target_slotid(slot->table, slot, res);
771                 break;
772         case 1:
773                 /*
774                  * sr_status remains 1 if an RPC level error occurred.
775                  * The server may or may not have processed the sequence
776                  * operation..
777                  * Mark the slot as having hosted an interrupted RPC call.
778                  */
779                 slot->interrupted = 1;
780                 goto out;
781         case -NFS4ERR_DELAY:
782                 /* The server detected a resend of the RPC call and
783                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
784                  * of RFC5661.
785                  */
786                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
787                         __func__,
788                         slot->slot_nr,
789                         slot->seq_nr);
790                 goto out_retry;
791         case -NFS4ERR_BADSLOT:
792                 /*
793                  * The slot id we used was probably retired. Try again
794                  * using a different slot id.
795                  */
796                 goto retry_nowait;
797         case -NFS4ERR_SEQ_MISORDERED:
798                 /*
799                  * Was the last operation on this sequence interrupted?
800                  * If so, retry after bumping the sequence number.
801                  */
802                 if (interrupted) {
803                         ++slot->seq_nr;
804                         goto retry_nowait;
805                 }
806                 /*
807                  * Could this slot have been previously retired?
808                  * If so, then the server may be expecting seq_nr = 1!
809                  */
810                 if (slot->seq_nr != 1) {
811                         slot->seq_nr = 1;
812                         goto retry_nowait;
813                 }
814                 break;
815         case -NFS4ERR_SEQ_FALSE_RETRY:
816                 ++slot->seq_nr;
817                 goto retry_nowait;
818         case -NFS4ERR_DEADSESSION:
819         case -NFS4ERR_BADSESSION:
820                 nfs4_schedule_session_recovery(session, res->sr_status);
821                 goto retry_nowait;
822         default:
823                 /* Just update the slot sequence no. */
824                 slot->seq_done = 1;
825         }
826 out:
827         /* The session may be reset by one of the error handlers. */
828         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
829 out_noaction:
830         return ret;
831 retry_nowait:
832         if (rpc_restart_call_prepare(task)) {
833                 nfs41_sequence_free_slot(res);
834                 task->tk_status = 0;
835                 ret = 0;
836         }
837         goto out;
838 out_retry:
839         if (!rpc_restart_call(task))
840                 goto out;
841         rpc_delay(task, NFS4_POLL_RETRY_MAX);
842         return 0;
843 }
844
845 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
846 {
847         if (!nfs41_sequence_process(task, res))
848                 return 0;
849         if (res->sr_slot != NULL)
850                 nfs41_sequence_free_slot(res);
851         return 1;
852
853 }
854 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
855
856 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
857 {
858         if (res->sr_slot == NULL)
859                 return 1;
860         if (res->sr_slot->table->session != NULL)
861                 return nfs41_sequence_process(task, res);
862         return nfs40_sequence_done(task, res);
863 }
864
865 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
866 {
867         if (res->sr_slot != NULL) {
868                 if (res->sr_slot->table->session != NULL)
869                         nfs41_sequence_free_slot(res);
870                 else
871                         nfs40_sequence_free_slot(res);
872         }
873 }
874
875 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
876 {
877         if (res->sr_slot == NULL)
878                 return 1;
879         if (!res->sr_slot->table->session)
880                 return nfs40_sequence_done(task, res);
881         return nfs41_sequence_done(task, res);
882 }
883 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
884
885 int nfs41_setup_sequence(struct nfs4_session *session,
886                                 struct nfs4_sequence_args *args,
887                                 struct nfs4_sequence_res *res,
888                                 struct rpc_task *task)
889 {
890         struct nfs4_slot *slot;
891         struct nfs4_slot_table *tbl;
892
893         dprintk("--> %s\n", __func__);
894         /* slot already allocated? */
895         if (res->sr_slot != NULL)
896                 goto out_success;
897
898         tbl = &session->fc_slot_table;
899
900         task->tk_timeout = 0;
901
902         spin_lock(&tbl->slot_tbl_lock);
903         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
904             !args->sa_privileged) {
905                 /* The state manager will wait until the slot table is empty */
906                 dprintk("%s session is draining\n", __func__);
907                 goto out_sleep;
908         }
909
910         slot = nfs4_alloc_slot(tbl);
911         if (IS_ERR(slot)) {
912                 /* If out of memory, try again in 1/4 second */
913                 if (slot == ERR_PTR(-ENOMEM))
914                         task->tk_timeout = HZ >> 2;
915                 dprintk("<-- %s: no free slots\n", __func__);
916                 goto out_sleep;
917         }
918         spin_unlock(&tbl->slot_tbl_lock);
919
920         slot->privileged = args->sa_privileged ? 1 : 0;
921         args->sa_slot = slot;
922
923         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
924                         slot->slot_nr, slot->seq_nr);
925
926         res->sr_slot = slot;
927         res->sr_timestamp = jiffies;
928         res->sr_status_flags = 0;
929         /*
930          * sr_status is only set in decode_sequence, and so will remain
931          * set to 1 if an rpc level failure occurs.
932          */
933         res->sr_status = 1;
934         trace_nfs4_setup_sequence(session, args);
935 out_success:
936         rpc_call_start(task);
937         return 0;
938 out_sleep:
939         /* Privileged tasks are queued with top priority */
940         if (args->sa_privileged)
941                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
942                                 NULL, RPC_PRIORITY_PRIVILEGED);
943         else
944                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
945         spin_unlock(&tbl->slot_tbl_lock);
946         return -EAGAIN;
947 }
948 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
949
950 static int nfs4_setup_sequence(const struct nfs_server *server,
951                                struct nfs4_sequence_args *args,
952                                struct nfs4_sequence_res *res,
953                                struct rpc_task *task)
954 {
955         struct nfs4_session *session = nfs4_get_session(server);
956         int ret = 0;
957
958         if (!session)
959                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
960                                             args, res, task);
961
962         dprintk("--> %s clp %p session %p sr_slot %u\n",
963                 __func__, session->clp, session, res->sr_slot ?
964                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
965
966         ret = nfs41_setup_sequence(session, args, res, task);
967
968         dprintk("<-- %s status=%d\n", __func__, ret);
969         return ret;
970 }
971
972 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
973 {
974         struct nfs4_call_sync_data *data = calldata;
975         struct nfs4_session *session = nfs4_get_session(data->seq_server);
976
977         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
978
979         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
980 }
981
982 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
983 {
984         struct nfs4_call_sync_data *data = calldata;
985
986         nfs41_sequence_done(task, data->seq_res);
987 }
988
989 static const struct rpc_call_ops nfs41_call_sync_ops = {
990         .rpc_call_prepare = nfs41_call_sync_prepare,
991         .rpc_call_done = nfs41_call_sync_done,
992 };
993
994 #else   /* !CONFIG_NFS_V4_1 */
995
996 static int nfs4_setup_sequence(const struct nfs_server *server,
997                                struct nfs4_sequence_args *args,
998                                struct nfs4_sequence_res *res,
999                                struct rpc_task *task)
1000 {
1001         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
1002                                     args, res, task);
1003 }
1004
1005 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1006 {
1007         return nfs40_sequence_done(task, res);
1008 }
1009
1010 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1011 {
1012         if (res->sr_slot != NULL)
1013                 nfs40_sequence_free_slot(res);
1014 }
1015
1016 int nfs4_sequence_done(struct rpc_task *task,
1017                        struct nfs4_sequence_res *res)
1018 {
1019         return nfs40_sequence_done(task, res);
1020 }
1021 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1022
1023 #endif  /* !CONFIG_NFS_V4_1 */
1024
1025 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1026 {
1027         struct nfs4_call_sync_data *data = calldata;
1028         nfs4_setup_sequence(data->seq_server,
1029                                 data->seq_args, data->seq_res, task);
1030 }
1031
1032 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1033 {
1034         struct nfs4_call_sync_data *data = calldata;
1035         nfs4_sequence_done(task, data->seq_res);
1036 }
1037
1038 static const struct rpc_call_ops nfs40_call_sync_ops = {
1039         .rpc_call_prepare = nfs40_call_sync_prepare,
1040         .rpc_call_done = nfs40_call_sync_done,
1041 };
1042
1043 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1044                                    struct nfs_server *server,
1045                                    struct rpc_message *msg,
1046                                    struct nfs4_sequence_args *args,
1047                                    struct nfs4_sequence_res *res)
1048 {
1049         int ret;
1050         struct rpc_task *task;
1051         struct nfs_client *clp = server->nfs_client;
1052         struct nfs4_call_sync_data data = {
1053                 .seq_server = server,
1054                 .seq_args = args,
1055                 .seq_res = res,
1056         };
1057         struct rpc_task_setup task_setup = {
1058                 .rpc_client = clnt,
1059                 .rpc_message = msg,
1060                 .callback_ops = clp->cl_mvops->call_sync_ops,
1061                 .callback_data = &data
1062         };
1063
1064         task = rpc_run_task(&task_setup);
1065         if (IS_ERR(task))
1066                 ret = PTR_ERR(task);
1067         else {
1068                 ret = task->tk_status;
1069                 rpc_put_task(task);
1070         }
1071         return ret;
1072 }
1073
1074 int nfs4_call_sync(struct rpc_clnt *clnt,
1075                    struct nfs_server *server,
1076                    struct rpc_message *msg,
1077                    struct nfs4_sequence_args *args,
1078                    struct nfs4_sequence_res *res,
1079                    int cache_reply)
1080 {
1081         nfs4_init_sequence(args, res, cache_reply);
1082         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1083 }
1084
1085 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1086                 unsigned long timestamp)
1087 {
1088         struct nfs_inode *nfsi = NFS_I(dir);
1089
1090         spin_lock(&dir->i_lock);
1091         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1092         if (cinfo->atomic && cinfo->before == dir->i_version) {
1093                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1094                 nfsi->attrtimeo_timestamp = jiffies;
1095         } else {
1096                 nfs_force_lookup_revalidate(dir);
1097                 if (cinfo->before != dir->i_version)
1098                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1099                                 NFS_INO_INVALID_ACL;
1100         }
1101         dir->i_version = cinfo->after;
1102         nfsi->read_cache_jiffies = timestamp;
1103         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1104         nfs_fscache_invalidate(dir);
1105         spin_unlock(&dir->i_lock);
1106 }
1107
1108 struct nfs4_opendata {
1109         struct kref kref;
1110         struct nfs_openargs o_arg;
1111         struct nfs_openres o_res;
1112         struct nfs_open_confirmargs c_arg;
1113         struct nfs_open_confirmres c_res;
1114         struct nfs4_string owner_name;
1115         struct nfs4_string group_name;
1116         struct nfs4_label *a_label;
1117         struct nfs_fattr f_attr;
1118         struct nfs4_label *f_label;
1119         struct dentry *dir;
1120         struct dentry *dentry;
1121         struct nfs4_state_owner *owner;
1122         struct nfs4_state *state;
1123         struct iattr attrs;
1124         unsigned long timestamp;
1125         unsigned int rpc_done : 1;
1126         unsigned int file_created : 1;
1127         unsigned int is_recover : 1;
1128         int rpc_status;
1129         int cancelled;
1130 };
1131
1132 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1133                 int err, struct nfs4_exception *exception)
1134 {
1135         if (err != -EINVAL)
1136                 return false;
1137         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1138                 return false;
1139         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1140         exception->retry = 1;
1141         return true;
1142 }
1143
1144 static u32
1145 nfs4_map_atomic_open_share(struct nfs_server *server,
1146                 fmode_t fmode, int openflags)
1147 {
1148         u32 res = 0;
1149
1150         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1151         case FMODE_READ:
1152                 res = NFS4_SHARE_ACCESS_READ;
1153                 break;
1154         case FMODE_WRITE:
1155                 res = NFS4_SHARE_ACCESS_WRITE;
1156                 break;
1157         case FMODE_READ|FMODE_WRITE:
1158                 res = NFS4_SHARE_ACCESS_BOTH;
1159         }
1160         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1161                 goto out;
1162         /* Want no delegation if we're using O_DIRECT */
1163         if (openflags & O_DIRECT)
1164                 res |= NFS4_SHARE_WANT_NO_DELEG;
1165 out:
1166         return res;
1167 }
1168
1169 static enum open_claim_type4
1170 nfs4_map_atomic_open_claim(struct nfs_server *server,
1171                 enum open_claim_type4 claim)
1172 {
1173         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1174                 return claim;
1175         switch (claim) {
1176         default:
1177                 return claim;
1178         case NFS4_OPEN_CLAIM_FH:
1179                 return NFS4_OPEN_CLAIM_NULL;
1180         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1181                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1182         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1183                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1184         }
1185 }
1186
1187 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1188 {
1189         p->o_res.f_attr = &p->f_attr;
1190         p->o_res.f_label = p->f_label;
1191         p->o_res.seqid = p->o_arg.seqid;
1192         p->c_res.seqid = p->c_arg.seqid;
1193         p->o_res.server = p->o_arg.server;
1194         p->o_res.access_request = p->o_arg.access;
1195         nfs_fattr_init(&p->f_attr);
1196         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1197 }
1198
1199 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1200                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1201                 const struct iattr *attrs,
1202                 struct nfs4_label *label,
1203                 enum open_claim_type4 claim,
1204                 gfp_t gfp_mask)
1205 {
1206         struct dentry *parent = dget_parent(dentry);
1207         struct inode *dir = d_inode(parent);
1208         struct nfs_server *server = NFS_SERVER(dir);
1209         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1210         struct nfs4_opendata *p;
1211
1212         p = kzalloc(sizeof(*p), gfp_mask);
1213         if (p == NULL)
1214                 goto err;
1215
1216         p->f_label = nfs4_label_alloc(server, gfp_mask);
1217         if (IS_ERR(p->f_label))
1218                 goto err_free_p;
1219
1220         p->a_label = nfs4_label_alloc(server, gfp_mask);
1221         if (IS_ERR(p->a_label))
1222                 goto err_free_f;
1223
1224         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1225         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1226         if (IS_ERR(p->o_arg.seqid))
1227                 goto err_free_label;
1228         nfs_sb_active(dentry->d_sb);
1229         p->dentry = dget(dentry);
1230         p->dir = parent;
1231         p->owner = sp;
1232         atomic_inc(&sp->so_count);
1233         p->o_arg.open_flags = flags;
1234         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1235         p->o_arg.umask = current_umask();
1236         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1237         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1238                         fmode, flags);
1239         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1240          * will return permission denied for all bits until close */
1241         if (!(flags & O_EXCL)) {
1242                 /* ask server to check for all possible rights as results
1243                  * are cached */
1244                 switch (p->o_arg.claim) {
1245                 default:
1246                         break;
1247                 case NFS4_OPEN_CLAIM_NULL:
1248                 case NFS4_OPEN_CLAIM_FH:
1249                         p->o_arg.access = NFS4_ACCESS_READ |
1250                                 NFS4_ACCESS_MODIFY |
1251                                 NFS4_ACCESS_EXTEND |
1252                                 NFS4_ACCESS_EXECUTE;
1253                 }
1254         }
1255         p->o_arg.clientid = server->nfs_client->cl_clientid;
1256         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1257         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1258         p->o_arg.name = &dentry->d_name;
1259         p->o_arg.server = server;
1260         p->o_arg.bitmask = nfs4_bitmask(server, label);
1261         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1262         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1263         switch (p->o_arg.claim) {
1264         case NFS4_OPEN_CLAIM_NULL:
1265         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1266         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1267                 p->o_arg.fh = NFS_FH(dir);
1268                 break;
1269         case NFS4_OPEN_CLAIM_PREVIOUS:
1270         case NFS4_OPEN_CLAIM_FH:
1271         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1272         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1273                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1274         }
1275         if (attrs != NULL && attrs->ia_valid != 0) {
1276                 __u32 verf[2];
1277
1278                 p->o_arg.u.attrs = &p->attrs;
1279                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1280
1281                 verf[0] = jiffies;
1282                 verf[1] = current->pid;
1283                 memcpy(p->o_arg.u.verifier.data, verf,
1284                                 sizeof(p->o_arg.u.verifier.data));
1285         }
1286         p->c_arg.fh = &p->o_res.fh;
1287         p->c_arg.stateid = &p->o_res.stateid;
1288         p->c_arg.seqid = p->o_arg.seqid;
1289         nfs4_init_opendata_res(p);
1290         kref_init(&p->kref);
1291         return p;
1292
1293 err_free_label:
1294         nfs4_label_free(p->a_label);
1295 err_free_f:
1296         nfs4_label_free(p->f_label);
1297 err_free_p:
1298         kfree(p);
1299 err:
1300         dput(parent);
1301         return NULL;
1302 }
1303
1304 static void nfs4_opendata_free(struct kref *kref)
1305 {
1306         struct nfs4_opendata *p = container_of(kref,
1307                         struct nfs4_opendata, kref);
1308         struct super_block *sb = p->dentry->d_sb;
1309
1310         nfs_free_seqid(p->o_arg.seqid);
1311         nfs4_sequence_free_slot(&p->o_res.seq_res);
1312         if (p->state != NULL)
1313                 nfs4_put_open_state(p->state);
1314         nfs4_put_state_owner(p->owner);
1315
1316         nfs4_label_free(p->a_label);
1317         nfs4_label_free(p->f_label);
1318
1319         dput(p->dir);
1320         dput(p->dentry);
1321         nfs_sb_deactive(sb);
1322         nfs_fattr_free_names(&p->f_attr);
1323         kfree(p->f_attr.mdsthreshold);
1324         kfree(p);
1325 }
1326
1327 static void nfs4_opendata_put(struct nfs4_opendata *p)
1328 {
1329         if (p != NULL)
1330                 kref_put(&p->kref, nfs4_opendata_free);
1331 }
1332
1333 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1334 {
1335         int ret;
1336
1337         ret = rpc_wait_for_completion_task(task);
1338         return ret;
1339 }
1340
1341 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1342                 fmode_t fmode)
1343 {
1344         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1345         case FMODE_READ|FMODE_WRITE:
1346                 return state->n_rdwr != 0;
1347         case FMODE_WRITE:
1348                 return state->n_wronly != 0;
1349         case FMODE_READ:
1350                 return state->n_rdonly != 0;
1351         }
1352         WARN_ON_ONCE(1);
1353         return false;
1354 }
1355
1356 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1357 {
1358         int ret = 0;
1359
1360         if (open_mode & (O_EXCL|O_TRUNC))
1361                 goto out;
1362         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1363                 case FMODE_READ:
1364                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1365                                 && state->n_rdonly != 0;
1366                         break;
1367                 case FMODE_WRITE:
1368                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1369                                 && state->n_wronly != 0;
1370                         break;
1371                 case FMODE_READ|FMODE_WRITE:
1372                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1373                                 && state->n_rdwr != 0;
1374         }
1375 out:
1376         return ret;
1377 }
1378
1379 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1380                 enum open_claim_type4 claim)
1381 {
1382         if (delegation == NULL)
1383                 return 0;
1384         if ((delegation->type & fmode) != fmode)
1385                 return 0;
1386         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1387                 return 0;
1388         switch (claim) {
1389         case NFS4_OPEN_CLAIM_NULL:
1390         case NFS4_OPEN_CLAIM_FH:
1391                 break;
1392         case NFS4_OPEN_CLAIM_PREVIOUS:
1393                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1394                         break;
1395         default:
1396                 return 0;
1397         }
1398         nfs_mark_delegation_referenced(delegation);
1399         return 1;
1400 }
1401
1402 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1403 {
1404         switch (fmode) {
1405                 case FMODE_WRITE:
1406                         state->n_wronly++;
1407                         break;
1408                 case FMODE_READ:
1409                         state->n_rdonly++;
1410                         break;
1411                 case FMODE_READ|FMODE_WRITE:
1412                         state->n_rdwr++;
1413         }
1414         nfs4_state_set_mode_locked(state, state->state | fmode);
1415 }
1416
1417 #ifdef CONFIG_NFS_V4_1
1418 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1419 {
1420         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1421                 return true;
1422         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1423                 return true;
1424         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1425                 return true;
1426         return false;
1427 }
1428 #endif /* CONFIG_NFS_V4_1 */
1429
1430 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1431 {
1432         struct nfs_client *clp = state->owner->so_server->nfs_client;
1433         bool need_recover = false;
1434
1435         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1436                 need_recover = true;
1437         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1438                 need_recover = true;
1439         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1440                 need_recover = true;
1441         if (need_recover)
1442                 nfs4_state_mark_reclaim_nograce(clp, state);
1443 }
1444
1445 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1446                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1447 {
1448         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1449                 return true;
1450         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1451                 nfs4_stateid_copy(freeme, &state->open_stateid);
1452                 nfs_test_and_clear_all_open_stateid(state);
1453                 return true;
1454         }
1455         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1456                 return true;
1457         return false;
1458 }
1459
1460 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1461 {
1462         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1463                 return;
1464         if (state->n_wronly)
1465                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1466         if (state->n_rdonly)
1467                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1468         if (state->n_rdwr)
1469                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1470         set_bit(NFS_OPEN_STATE, &state->flags);
1471 }
1472
1473 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1474                 nfs4_stateid *stateid, fmode_t fmode)
1475 {
1476         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1477         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1478         case FMODE_WRITE:
1479                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1480                 break;
1481         case FMODE_READ:
1482                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1483                 break;
1484         case 0:
1485                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1486                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1487                 clear_bit(NFS_OPEN_STATE, &state->flags);
1488         }
1489         if (stateid == NULL)
1490                 return;
1491         /* Handle OPEN+OPEN_DOWNGRADE races */
1492         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1493             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1494                 nfs_resync_open_stateid_locked(state);
1495                 return;
1496         }
1497         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1498                 nfs4_stateid_copy(&state->stateid, stateid);
1499         nfs4_stateid_copy(&state->open_stateid, stateid);
1500 }
1501
1502 static void nfs_clear_open_stateid(struct nfs4_state *state,
1503         nfs4_stateid *arg_stateid,
1504         nfs4_stateid *stateid, fmode_t fmode)
1505 {
1506         write_seqlock(&state->seqlock);
1507         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1508         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1509                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1510         write_sequnlock(&state->seqlock);
1511         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1512                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1513 }
1514
1515 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1516                 const nfs4_stateid *stateid, fmode_t fmode,
1517                 nfs4_stateid *freeme)
1518 {
1519         switch (fmode) {
1520                 case FMODE_READ:
1521                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1522                         break;
1523                 case FMODE_WRITE:
1524                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1525                         break;
1526                 case FMODE_READ|FMODE_WRITE:
1527                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1528         }
1529         if (!nfs_need_update_open_stateid(state, stateid, freeme))
1530                 return;
1531         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1532                 nfs4_stateid_copy(&state->stateid, stateid);
1533         nfs4_stateid_copy(&state->open_stateid, stateid);
1534 }
1535
1536 static void __update_open_stateid(struct nfs4_state *state,
1537                 const nfs4_stateid *open_stateid,
1538                 const nfs4_stateid *deleg_stateid,
1539                 fmode_t fmode,
1540                 nfs4_stateid *freeme)
1541 {
1542         /*
1543          * Protect the call to nfs4_state_set_mode_locked and
1544          * serialise the stateid update
1545          */
1546         spin_lock(&state->owner->so_lock);
1547         write_seqlock(&state->seqlock);
1548         if (deleg_stateid != NULL) {
1549                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1550                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1551         }
1552         if (open_stateid != NULL)
1553                 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1554         write_sequnlock(&state->seqlock);
1555         update_open_stateflags(state, fmode);
1556         spin_unlock(&state->owner->so_lock);
1557 }
1558
1559 static int update_open_stateid(struct nfs4_state *state,
1560                 const nfs4_stateid *open_stateid,
1561                 const nfs4_stateid *delegation,
1562                 fmode_t fmode)
1563 {
1564         struct nfs_server *server = NFS_SERVER(state->inode);
1565         struct nfs_client *clp = server->nfs_client;
1566         struct nfs_inode *nfsi = NFS_I(state->inode);
1567         struct nfs_delegation *deleg_cur;
1568         nfs4_stateid freeme = { };
1569         int ret = 0;
1570
1571         fmode &= (FMODE_READ|FMODE_WRITE);
1572
1573         rcu_read_lock();
1574         deleg_cur = rcu_dereference(nfsi->delegation);
1575         if (deleg_cur == NULL)
1576                 goto no_delegation;
1577
1578         spin_lock(&deleg_cur->lock);
1579         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1580            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1581             (deleg_cur->type & fmode) != fmode)
1582                 goto no_delegation_unlock;
1583
1584         if (delegation == NULL)
1585                 delegation = &deleg_cur->stateid;
1586         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1587                 goto no_delegation_unlock;
1588
1589         nfs_mark_delegation_referenced(deleg_cur);
1590         __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1591                         fmode, &freeme);
1592         ret = 1;
1593 no_delegation_unlock:
1594         spin_unlock(&deleg_cur->lock);
1595 no_delegation:
1596         rcu_read_unlock();
1597
1598         if (!ret && open_stateid != NULL) {
1599                 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1600                 ret = 1;
1601         }
1602         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1603                 nfs4_schedule_state_manager(clp);
1604         if (freeme.type != 0)
1605                 nfs4_test_and_free_stateid(server, &freeme,
1606                                 state->owner->so_cred);
1607
1608         return ret;
1609 }
1610
1611 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1612                 const nfs4_stateid *stateid)
1613 {
1614         struct nfs4_state *state = lsp->ls_state;
1615         bool ret = false;
1616
1617         spin_lock(&state->state_lock);
1618         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1619                 goto out_noupdate;
1620         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1621                 goto out_noupdate;
1622         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1623         ret = true;
1624 out_noupdate:
1625         spin_unlock(&state->state_lock);
1626         return ret;
1627 }
1628
1629 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1630 {
1631         struct nfs_delegation *delegation;
1632
1633         rcu_read_lock();
1634         delegation = rcu_dereference(NFS_I(inode)->delegation);
1635         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1636                 rcu_read_unlock();
1637                 return;
1638         }
1639         rcu_read_unlock();
1640         nfs4_inode_return_delegation(inode);
1641 }
1642
1643 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1644 {
1645         struct nfs4_state *state = opendata->state;
1646         struct nfs_inode *nfsi = NFS_I(state->inode);
1647         struct nfs_delegation *delegation;
1648         int open_mode = opendata->o_arg.open_flags;
1649         fmode_t fmode = opendata->o_arg.fmode;
1650         enum open_claim_type4 claim = opendata->o_arg.claim;
1651         nfs4_stateid stateid;
1652         int ret = -EAGAIN;
1653
1654         for (;;) {
1655                 spin_lock(&state->owner->so_lock);
1656                 if (can_open_cached(state, fmode, open_mode)) {
1657                         update_open_stateflags(state, fmode);
1658                         spin_unlock(&state->owner->so_lock);
1659                         goto out_return_state;
1660                 }
1661                 spin_unlock(&state->owner->so_lock);
1662                 rcu_read_lock();
1663                 delegation = rcu_dereference(nfsi->delegation);
1664                 if (!can_open_delegated(delegation, fmode, claim)) {
1665                         rcu_read_unlock();
1666                         break;
1667                 }
1668                 /* Save the delegation */
1669                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1670                 rcu_read_unlock();
1671                 nfs_release_seqid(opendata->o_arg.seqid);
1672                 if (!opendata->is_recover) {
1673                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1674                         if (ret != 0)
1675                                 goto out;
1676                 }
1677                 ret = -EAGAIN;
1678
1679                 /* Try to update the stateid using the delegation */
1680                 if (update_open_stateid(state, NULL, &stateid, fmode))
1681                         goto out_return_state;
1682         }
1683 out:
1684         return ERR_PTR(ret);
1685 out_return_state:
1686         atomic_inc(&state->count);
1687         return state;
1688 }
1689
1690 static void
1691 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1692 {
1693         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1694         struct nfs_delegation *delegation;
1695         int delegation_flags = 0;
1696
1697         rcu_read_lock();
1698         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1699         if (delegation)
1700                 delegation_flags = delegation->flags;
1701         rcu_read_unlock();
1702         switch (data->o_arg.claim) {
1703         default:
1704                 break;
1705         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1706         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1707                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1708                                    "returning a delegation for "
1709                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1710                                    clp->cl_hostname);
1711                 return;
1712         }
1713         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1714                 nfs_inode_set_delegation(state->inode,
1715                                          data->owner->so_cred,
1716                                          &data->o_res);
1717         else
1718                 nfs_inode_reclaim_delegation(state->inode,
1719                                              data->owner->so_cred,
1720                                              &data->o_res);
1721 }
1722
1723 /*
1724  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1725  * and update the nfs4_state.
1726  */
1727 static struct nfs4_state *
1728 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1729 {
1730         struct inode *inode = data->state->inode;
1731         struct nfs4_state *state = data->state;
1732         int ret;
1733
1734         if (!data->rpc_done) {
1735                 if (data->rpc_status) {
1736                         ret = data->rpc_status;
1737                         goto err;
1738                 }
1739                 /* cached opens have already been processed */
1740                 goto update;
1741         }
1742
1743         ret = nfs_refresh_inode(inode, &data->f_attr);
1744         if (ret)
1745                 goto err;
1746
1747         if (data->o_res.delegation_type != 0)
1748                 nfs4_opendata_check_deleg(data, state);
1749 update:
1750         update_open_stateid(state, &data->o_res.stateid, NULL,
1751                             data->o_arg.fmode);
1752         atomic_inc(&state->count);
1753
1754         return state;
1755 err:
1756         return ERR_PTR(ret);
1757
1758 }
1759
1760 static struct nfs4_state *
1761 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1762 {
1763         struct inode *inode;
1764         struct nfs4_state *state = NULL;
1765         int ret;
1766
1767         if (!data->rpc_done) {
1768                 state = nfs4_try_open_cached(data);
1769                 trace_nfs4_cached_open(data->state);
1770                 goto out;
1771         }
1772
1773         ret = -EAGAIN;
1774         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1775                 goto err;
1776         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1777         ret = PTR_ERR(inode);
1778         if (IS_ERR(inode))
1779                 goto err;
1780         ret = -ENOMEM;
1781         state = nfs4_get_open_state(inode, data->owner);
1782         if (state == NULL)
1783                 goto err_put_inode;
1784         if (data->o_res.delegation_type != 0)
1785                 nfs4_opendata_check_deleg(data, state);
1786         update_open_stateid(state, &data->o_res.stateid, NULL,
1787                         data->o_arg.fmode);
1788         iput(inode);
1789 out:
1790         nfs_release_seqid(data->o_arg.seqid);
1791         return state;
1792 err_put_inode:
1793         iput(inode);
1794 err:
1795         return ERR_PTR(ret);
1796 }
1797
1798 static struct nfs4_state *
1799 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1800 {
1801         struct nfs4_state *ret;
1802
1803         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1804                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1805         else
1806                 ret = _nfs4_opendata_to_nfs4_state(data);
1807         nfs4_sequence_free_slot(&data->o_res.seq_res);
1808         return ret;
1809 }
1810
1811 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1812 {
1813         struct nfs_inode *nfsi = NFS_I(state->inode);
1814         struct nfs_open_context *ctx;
1815
1816         spin_lock(&state->inode->i_lock);
1817         list_for_each_entry(ctx, &nfsi->open_files, list) {
1818                 if (ctx->state != state)
1819                         continue;
1820                 get_nfs_open_context(ctx);
1821                 spin_unlock(&state->inode->i_lock);
1822                 return ctx;
1823         }
1824         spin_unlock(&state->inode->i_lock);
1825         return ERR_PTR(-ENOENT);
1826 }
1827
1828 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1829                 struct nfs4_state *state, enum open_claim_type4 claim)
1830 {
1831         struct nfs4_opendata *opendata;
1832
1833         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1834                         NULL, NULL, claim, GFP_NOFS);
1835         if (opendata == NULL)
1836                 return ERR_PTR(-ENOMEM);
1837         opendata->state = state;
1838         atomic_inc(&state->count);
1839         return opendata;
1840 }
1841
1842 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1843                 fmode_t fmode)
1844 {
1845         struct nfs4_state *newstate;
1846         int ret;
1847
1848         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1849                 return 0;
1850         opendata->o_arg.open_flags = 0;
1851         opendata->o_arg.fmode = fmode;
1852         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1853                         NFS_SB(opendata->dentry->d_sb),
1854                         fmode, 0);
1855         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1856         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1857         nfs4_init_opendata_res(opendata);
1858         ret = _nfs4_recover_proc_open(opendata);
1859         if (ret != 0)
1860                 return ret; 
1861         newstate = nfs4_opendata_to_nfs4_state(opendata);
1862         if (IS_ERR(newstate))
1863                 return PTR_ERR(newstate);
1864         if (newstate != opendata->state)
1865                 ret = -ESTALE;
1866         nfs4_close_state(newstate, fmode);
1867         return ret;
1868 }
1869
1870 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1871 {
1872         int ret;
1873
1874         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1875         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1876         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1877         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1878         /* memory barrier prior to reading state->n_* */
1879         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1880         clear_bit(NFS_OPEN_STATE, &state->flags);
1881         smp_rmb();
1882         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1883         if (ret != 0)
1884                 return ret;
1885         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1886         if (ret != 0)
1887                 return ret;
1888         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1889         if (ret != 0)
1890                 return ret;
1891         /*
1892          * We may have performed cached opens for all three recoveries.
1893          * Check if we need to update the current stateid.
1894          */
1895         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1896             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1897                 write_seqlock(&state->seqlock);
1898                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1899                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1900                 write_sequnlock(&state->seqlock);
1901         }
1902         return 0;
1903 }
1904
1905 /*
1906  * OPEN_RECLAIM:
1907  *      reclaim state on the server after a reboot.
1908  */
1909 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1910 {
1911         struct nfs_delegation *delegation;
1912         struct nfs4_opendata *opendata;
1913         fmode_t delegation_type = 0;
1914         int status;
1915
1916         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1917                         NFS4_OPEN_CLAIM_PREVIOUS);
1918         if (IS_ERR(opendata))
1919                 return PTR_ERR(opendata);
1920         rcu_read_lock();
1921         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1922         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1923                 delegation_type = delegation->type;
1924         rcu_read_unlock();
1925         opendata->o_arg.u.delegation_type = delegation_type;
1926         status = nfs4_open_recover(opendata, state);
1927         nfs4_opendata_put(opendata);
1928         return status;
1929 }
1930
1931 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1932 {
1933         struct nfs_server *server = NFS_SERVER(state->inode);
1934         struct nfs4_exception exception = { };
1935         int err;
1936         do {
1937                 err = _nfs4_do_open_reclaim(ctx, state);
1938                 trace_nfs4_open_reclaim(ctx, 0, err);
1939                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1940                         continue;
1941                 if (err != -NFS4ERR_DELAY)
1942                         break;
1943                 nfs4_handle_exception(server, err, &exception);
1944         } while (exception.retry);
1945         return err;
1946 }
1947
1948 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1949 {
1950         struct nfs_open_context *ctx;
1951         int ret;
1952
1953         ctx = nfs4_state_find_open_context(state);
1954         if (IS_ERR(ctx))
1955                 return -EAGAIN;
1956         ret = nfs4_do_open_reclaim(ctx, state);
1957         put_nfs_open_context(ctx);
1958         return ret;
1959 }
1960
1961 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1962 {
1963         switch (err) {
1964                 default:
1965                         printk(KERN_ERR "NFS: %s: unhandled error "
1966                                         "%d.\n", __func__, err);
1967                 case 0:
1968                 case -ENOENT:
1969                 case -EAGAIN:
1970                 case -ESTALE:
1971                         break;
1972                 case -NFS4ERR_BADSESSION:
1973                 case -NFS4ERR_BADSLOT:
1974                 case -NFS4ERR_BAD_HIGH_SLOT:
1975                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1976                 case -NFS4ERR_DEADSESSION:
1977                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1978                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1979                         return -EAGAIN;
1980                 case -NFS4ERR_STALE_CLIENTID:
1981                 case -NFS4ERR_STALE_STATEID:
1982                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1983                         /* Don't recall a delegation if it was lost */
1984                         nfs4_schedule_lease_recovery(server->nfs_client);
1985                         return -EAGAIN;
1986                 case -NFS4ERR_MOVED:
1987                         nfs4_schedule_migration_recovery(server);
1988                         return -EAGAIN;
1989                 case -NFS4ERR_LEASE_MOVED:
1990                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1991                         return -EAGAIN;
1992                 case -NFS4ERR_DELEG_REVOKED:
1993                 case -NFS4ERR_ADMIN_REVOKED:
1994                 case -NFS4ERR_EXPIRED:
1995                 case -NFS4ERR_BAD_STATEID:
1996                 case -NFS4ERR_OPENMODE:
1997                         nfs_inode_find_state_and_recover(state->inode,
1998                                         stateid);
1999                         nfs4_schedule_stateid_recovery(server, state);
2000                         return -EAGAIN;
2001                 case -NFS4ERR_DELAY:
2002                 case -NFS4ERR_GRACE:
2003                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2004                         ssleep(1);
2005                         return -EAGAIN;
2006                 case -ENOMEM:
2007                 case -NFS4ERR_DENIED:
2008                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
2009                         return 0;
2010         }
2011         return err;
2012 }
2013
2014 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2015                 struct nfs4_state *state, const nfs4_stateid *stateid,
2016                 fmode_t type)
2017 {
2018         struct nfs_server *server = NFS_SERVER(state->inode);
2019         struct nfs4_opendata *opendata;
2020         int err = 0;
2021
2022         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2023                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2024         if (IS_ERR(opendata))
2025                 return PTR_ERR(opendata);
2026         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2027         write_seqlock(&state->seqlock);
2028         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2029         write_sequnlock(&state->seqlock);
2030         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2031         switch (type & (FMODE_READ|FMODE_WRITE)) {
2032         case FMODE_READ|FMODE_WRITE:
2033         case FMODE_WRITE:
2034                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2035                 if (err)
2036                         break;
2037                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2038                 if (err)
2039                         break;
2040         case FMODE_READ:
2041                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2042         }
2043         nfs4_opendata_put(opendata);
2044         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
2045 }
2046
2047 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2048 {
2049         struct nfs4_opendata *data = calldata;
2050
2051         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
2052                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
2053 }
2054
2055 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2056 {
2057         struct nfs4_opendata *data = calldata;
2058
2059         nfs40_sequence_done(task, &data->c_res.seq_res);
2060
2061         data->rpc_status = task->tk_status;
2062         if (data->rpc_status == 0) {
2063                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2064                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2065                 renew_lease(data->o_res.server, data->timestamp);
2066                 data->rpc_done = 1;
2067         }
2068 }
2069
2070 static void nfs4_open_confirm_release(void *calldata)
2071 {
2072         struct nfs4_opendata *data = calldata;
2073         struct nfs4_state *state = NULL;
2074
2075         /* If this request hasn't been cancelled, do nothing */
2076         if (data->cancelled == 0)
2077                 goto out_free;
2078         /* In case of error, no cleanup! */
2079         if (!data->rpc_done)
2080                 goto out_free;
2081         state = nfs4_opendata_to_nfs4_state(data);
2082         if (!IS_ERR(state))
2083                 nfs4_close_state(state, data->o_arg.fmode);
2084 out_free:
2085         nfs4_opendata_put(data);
2086 }
2087
2088 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2089         .rpc_call_prepare = nfs4_open_confirm_prepare,
2090         .rpc_call_done = nfs4_open_confirm_done,
2091         .rpc_release = nfs4_open_confirm_release,
2092 };
2093
2094 /*
2095  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2096  */
2097 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2098 {
2099         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2100         struct rpc_task *task;
2101         struct  rpc_message msg = {
2102                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2103                 .rpc_argp = &data->c_arg,
2104                 .rpc_resp = &data->c_res,
2105                 .rpc_cred = data->owner->so_cred,
2106         };
2107         struct rpc_task_setup task_setup_data = {
2108                 .rpc_client = server->client,
2109                 .rpc_message = &msg,
2110                 .callback_ops = &nfs4_open_confirm_ops,
2111                 .callback_data = data,
2112                 .workqueue = nfsiod_workqueue,
2113                 .flags = RPC_TASK_ASYNC,
2114         };
2115         int status;
2116
2117         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2118         kref_get(&data->kref);
2119         data->rpc_done = 0;
2120         data->rpc_status = 0;
2121         data->timestamp = jiffies;
2122         if (data->is_recover)
2123                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2124         task = rpc_run_task(&task_setup_data);
2125         if (IS_ERR(task))
2126                 return PTR_ERR(task);
2127         status = nfs4_wait_for_completion_rpc_task(task);
2128         if (status != 0) {
2129                 data->cancelled = 1;
2130                 smp_wmb();
2131         } else
2132                 status = data->rpc_status;
2133         rpc_put_task(task);
2134         return status;
2135 }
2136
2137 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2138 {
2139         struct nfs4_opendata *data = calldata;
2140         struct nfs4_state_owner *sp = data->owner;
2141         struct nfs_client *clp = sp->so_server->nfs_client;
2142         enum open_claim_type4 claim = data->o_arg.claim;
2143
2144         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2145                 goto out_wait;
2146         /*
2147          * Check if we still need to send an OPEN call, or if we can use
2148          * a delegation instead.
2149          */
2150         if (data->state != NULL) {
2151                 struct nfs_delegation *delegation;
2152
2153                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2154                         goto out_no_action;
2155                 rcu_read_lock();
2156                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2157                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2158                         goto unlock_no_action;
2159                 rcu_read_unlock();
2160         }
2161         /* Update client id. */
2162         data->o_arg.clientid = clp->cl_clientid;
2163         switch (claim) {
2164         default:
2165                 break;
2166         case NFS4_OPEN_CLAIM_PREVIOUS:
2167         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2168         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2169                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2170         case NFS4_OPEN_CLAIM_FH:
2171                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2172                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2173         }
2174         data->timestamp = jiffies;
2175         if (nfs4_setup_sequence(data->o_arg.server,
2176                                 &data->o_arg.seq_args,
2177                                 &data->o_res.seq_res,
2178                                 task) != 0)
2179                 nfs_release_seqid(data->o_arg.seqid);
2180
2181         /* Set the create mode (note dependency on the session type) */
2182         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2183         if (data->o_arg.open_flags & O_EXCL) {
2184                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2185                 if (nfs4_has_persistent_session(clp))
2186                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2187                 else if (clp->cl_mvops->minor_version > 0)
2188                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2189         }
2190         return;
2191 unlock_no_action:
2192         trace_nfs4_cached_open(data->state);
2193         rcu_read_unlock();
2194 out_no_action:
2195         task->tk_action = NULL;
2196 out_wait:
2197         nfs4_sequence_done(task, &data->o_res.seq_res);
2198 }
2199
2200 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2201 {
2202         struct nfs4_opendata *data = calldata;
2203
2204         data->rpc_status = task->tk_status;
2205
2206         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2207                 return;
2208
2209         if (task->tk_status == 0) {
2210                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2211                         switch (data->o_res.f_attr->mode & S_IFMT) {
2212                         case S_IFREG:
2213                                 break;
2214                         case S_IFLNK:
2215                                 data->rpc_status = -ELOOP;
2216                                 break;
2217                         case S_IFDIR:
2218                                 data->rpc_status = -EISDIR;
2219                                 break;
2220                         default:
2221                                 data->rpc_status = -ENOTDIR;
2222                         }
2223                 }
2224                 renew_lease(data->o_res.server, data->timestamp);
2225                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2226                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2227         }
2228         data->rpc_done = 1;
2229 }
2230
2231 static void nfs4_open_release(void *calldata)
2232 {
2233         struct nfs4_opendata *data = calldata;
2234         struct nfs4_state *state = NULL;
2235
2236         /* If this request hasn't been cancelled, do nothing */
2237         if (data->cancelled == 0)
2238                 goto out_free;
2239         /* In case of error, no cleanup! */
2240         if (data->rpc_status != 0 || !data->rpc_done)
2241                 goto out_free;
2242         /* In case we need an open_confirm, no cleanup! */
2243         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2244                 goto out_free;
2245         state = nfs4_opendata_to_nfs4_state(data);
2246         if (!IS_ERR(state))
2247                 nfs4_close_state(state, data->o_arg.fmode);
2248 out_free:
2249         nfs4_opendata_put(data);
2250 }
2251
2252 static const struct rpc_call_ops nfs4_open_ops = {
2253         .rpc_call_prepare = nfs4_open_prepare,
2254         .rpc_call_done = nfs4_open_done,
2255         .rpc_release = nfs4_open_release,
2256 };
2257
2258 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2259 {
2260         struct inode *dir = d_inode(data->dir);
2261         struct nfs_server *server = NFS_SERVER(dir);
2262         struct nfs_openargs *o_arg = &data->o_arg;
2263         struct nfs_openres *o_res = &data->o_res;
2264         struct rpc_task *task;
2265         struct rpc_message msg = {
2266                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2267                 .rpc_argp = o_arg,
2268                 .rpc_resp = o_res,
2269                 .rpc_cred = data->owner->so_cred,
2270         };
2271         struct rpc_task_setup task_setup_data = {
2272                 .rpc_client = server->client,
2273                 .rpc_message = &msg,
2274                 .callback_ops = &nfs4_open_ops,
2275                 .callback_data = data,
2276                 .workqueue = nfsiod_workqueue,
2277                 .flags = RPC_TASK_ASYNC,
2278         };
2279         int status;
2280
2281         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2282         kref_get(&data->kref);
2283         data->rpc_done = 0;
2284         data->rpc_status = 0;
2285         data->cancelled = 0;
2286         data->is_recover = 0;
2287         if (isrecover) {
2288                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2289                 data->is_recover = 1;
2290         }
2291         task = rpc_run_task(&task_setup_data);
2292         if (IS_ERR(task))
2293                 return PTR_ERR(task);
2294         status = nfs4_wait_for_completion_rpc_task(task);
2295         if (status != 0) {
2296                 data->cancelled = 1;
2297                 smp_wmb();
2298         } else
2299                 status = data->rpc_status;
2300         rpc_put_task(task);
2301
2302         return status;
2303 }
2304
2305 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2306 {
2307         struct inode *dir = d_inode(data->dir);
2308         struct nfs_openres *o_res = &data->o_res;
2309         int status;
2310
2311         status = nfs4_run_open_task(data, 1);
2312         if (status != 0 || !data->rpc_done)
2313                 return status;
2314
2315         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2316
2317         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2318                 status = _nfs4_proc_open_confirm(data);
2319                 if (status != 0)
2320                         return status;
2321         }
2322
2323         return status;
2324 }
2325
2326 /*
2327  * Additional permission checks in order to distinguish between an
2328  * open for read, and an open for execute. This works around the
2329  * fact that NFSv4 OPEN treats read and execute permissions as being
2330  * the same.
2331  * Note that in the non-execute case, we want to turn off permission
2332  * checking if we just created a new file (POSIX open() semantics).
2333  */
2334 static int nfs4_opendata_access(struct rpc_cred *cred,
2335                                 struct nfs4_opendata *opendata,
2336                                 struct nfs4_state *state, fmode_t fmode,
2337                                 int openflags)
2338 {
2339         struct nfs_access_entry cache;
2340         u32 mask;
2341
2342         /* access call failed or for some reason the server doesn't
2343          * support any access modes -- defer access call until later */
2344         if (opendata->o_res.access_supported == 0)
2345                 return 0;
2346
2347         mask = 0;
2348         /*
2349          * Use openflags to check for exec, because fmode won't
2350          * always have FMODE_EXEC set when file open for exec.
2351          */
2352         if (openflags & __FMODE_EXEC) {
2353                 /* ONLY check for exec rights */
2354                 mask = MAY_EXEC;
2355         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2356                 mask = MAY_READ;
2357
2358         cache.cred = cred;
2359         cache.jiffies = jiffies;
2360         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2361         nfs_access_add_cache(state->inode, &cache);
2362
2363         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2364                 return 0;
2365
2366         /* even though OPEN succeeded, access is denied. Close the file */
2367         nfs4_close_state(state, fmode);
2368         return -EACCES;
2369 }
2370
2371 /*
2372  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2373  */
2374 static int _nfs4_proc_open(struct nfs4_opendata *data)
2375 {
2376         struct inode *dir = d_inode(data->dir);
2377         struct nfs_server *server = NFS_SERVER(dir);
2378         struct nfs_openargs *o_arg = &data->o_arg;
2379         struct nfs_openres *o_res = &data->o_res;
2380         int status;
2381
2382         status = nfs4_run_open_task(data, 0);
2383         if (!data->rpc_done)
2384                 return status;
2385         if (status != 0) {
2386                 if (status == -NFS4ERR_BADNAME &&
2387                                 !(o_arg->open_flags & O_CREAT))
2388                         return -ENOENT;
2389                 return status;
2390         }
2391
2392         nfs_fattr_map_and_free_names(server, &data->f_attr);
2393
2394         if (o_arg->open_flags & O_CREAT) {
2395                 if (o_arg->open_flags & O_EXCL)
2396                         data->file_created = 1;
2397                 else if (o_res->cinfo.before != o_res->cinfo.after)
2398                         data->file_created = 1;
2399                 if (data->file_created || dir->i_version != o_res->cinfo.after)
2400                         update_changeattr(dir, &o_res->cinfo,
2401                                         o_res->f_attr->time_start);
2402         }
2403         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2404                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2405         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2406                 status = _nfs4_proc_open_confirm(data);
2407                 if (status != 0)
2408                         return status;
2409         }
2410         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2411                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2412         return 0;
2413 }
2414
2415 static int nfs4_recover_expired_lease(struct nfs_server *server)
2416 {
2417         return nfs4_client_recover_expired_lease(server->nfs_client);
2418 }
2419
2420 /*
2421  * OPEN_EXPIRED:
2422  *      reclaim state on the server after a network partition.
2423  *      Assumes caller holds the appropriate lock
2424  */
2425 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2426 {
2427         struct nfs4_opendata *opendata;
2428         int ret;
2429
2430         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2431                         NFS4_OPEN_CLAIM_FH);
2432         if (IS_ERR(opendata))
2433                 return PTR_ERR(opendata);
2434         ret = nfs4_open_recover(opendata, state);
2435         if (ret == -ESTALE)
2436                 d_drop(ctx->dentry);
2437         nfs4_opendata_put(opendata);
2438         return ret;
2439 }
2440
2441 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2442 {
2443         struct nfs_server *server = NFS_SERVER(state->inode);
2444         struct nfs4_exception exception = { };
2445         int err;
2446
2447         do {
2448                 err = _nfs4_open_expired(ctx, state);
2449                 trace_nfs4_open_expired(ctx, 0, err);
2450                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2451                         continue;
2452                 switch (err) {
2453                 default:
2454                         goto out;
2455                 case -NFS4ERR_GRACE:
2456                 case -NFS4ERR_DELAY:
2457                         nfs4_handle_exception(server, err, &exception);
2458                         err = 0;
2459                 }
2460         } while (exception.retry);
2461 out:
2462         return err;
2463 }
2464
2465 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2466 {
2467         struct nfs_open_context *ctx;
2468         int ret;
2469
2470         ctx = nfs4_state_find_open_context(state);
2471         if (IS_ERR(ctx))
2472                 return -EAGAIN;
2473         ret = nfs4_do_open_expired(ctx, state);
2474         put_nfs_open_context(ctx);
2475         return ret;
2476 }
2477
2478 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2479                 const nfs4_stateid *stateid)
2480 {
2481         nfs_remove_bad_delegation(state->inode, stateid);
2482         write_seqlock(&state->seqlock);
2483         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2484         write_sequnlock(&state->seqlock);
2485         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2486 }
2487
2488 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2489 {
2490         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2491                 nfs_finish_clear_delegation_stateid(state, NULL);
2492 }
2493
2494 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2495 {
2496         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2497         nfs40_clear_delegation_stateid(state);
2498         return nfs4_open_expired(sp, state);
2499 }
2500
2501 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2502                 nfs4_stateid *stateid,
2503                 struct rpc_cred *cred)
2504 {
2505         return -NFS4ERR_BAD_STATEID;
2506 }
2507
2508 #if defined(CONFIG_NFS_V4_1)
2509 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2510                 nfs4_stateid *stateid,
2511                 struct rpc_cred *cred)
2512 {
2513         int status;
2514
2515         switch (stateid->type) {
2516         default:
2517                 break;
2518         case NFS4_INVALID_STATEID_TYPE:
2519         case NFS4_SPECIAL_STATEID_TYPE:
2520                 return -NFS4ERR_BAD_STATEID;
2521         case NFS4_REVOKED_STATEID_TYPE:
2522                 goto out_free;
2523         }
2524
2525         status = nfs41_test_stateid(server, stateid, cred);
2526         switch (status) {
2527         case -NFS4ERR_EXPIRED:
2528         case -NFS4ERR_ADMIN_REVOKED:
2529         case -NFS4ERR_DELEG_REVOKED:
2530                 break;
2531         default:
2532                 return status;
2533         }
2534 out_free:
2535         /* Ack the revoked state to the server */
2536         nfs41_free_stateid(server, stateid, cred, true);
2537         return -NFS4ERR_EXPIRED;
2538 }
2539
2540 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2541 {
2542         struct nfs_server *server = NFS_SERVER(state->inode);
2543         nfs4_stateid stateid;
2544         struct nfs_delegation *delegation;
2545         struct rpc_cred *cred;
2546         int status;
2547
2548         /* Get the delegation credential for use by test/free_stateid */
2549         rcu_read_lock();
2550         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2551         if (delegation == NULL) {
2552                 rcu_read_unlock();
2553                 return;
2554         }
2555
2556         nfs4_stateid_copy(&stateid, &delegation->stateid);
2557         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2558                 rcu_read_unlock();
2559                 nfs_finish_clear_delegation_stateid(state, &stateid);
2560                 return;
2561         }
2562
2563         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED, &delegation->flags)) {
2564                 rcu_read_unlock();
2565                 return;
2566         }
2567
2568         cred = get_rpccred(delegation->cred);
2569         rcu_read_unlock();
2570         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2571         trace_nfs4_test_delegation_stateid(state, NULL, status);
2572         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2573                 nfs_finish_clear_delegation_stateid(state, &stateid);
2574
2575         put_rpccred(cred);
2576 }
2577
2578 /**
2579  * nfs41_check_expired_locks - possibly free a lock stateid
2580  *
2581  * @state: NFSv4 state for an inode
2582  *
2583  * Returns NFS_OK if recovery for this stateid is now finished.
2584  * Otherwise a negative NFS4ERR value is returned.
2585  */
2586 static int nfs41_check_expired_locks(struct nfs4_state *state)
2587 {
2588         int status, ret = NFS_OK;
2589         struct nfs4_lock_state *lsp, *prev = NULL;
2590         struct nfs_server *server = NFS_SERVER(state->inode);
2591
2592         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2593                 goto out;
2594
2595         spin_lock(&state->state_lock);
2596         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2597                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2598                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2599
2600                         atomic_inc(&lsp->ls_count);
2601                         spin_unlock(&state->state_lock);
2602
2603                         nfs4_put_lock_state(prev);
2604                         prev = lsp;
2605
2606                         status = nfs41_test_and_free_expired_stateid(server,
2607                                         &lsp->ls_stateid,
2608                                         cred);
2609                         trace_nfs4_test_lock_stateid(state, lsp, status);
2610                         if (status == -NFS4ERR_EXPIRED ||
2611                             status == -NFS4ERR_BAD_STATEID) {
2612                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2613                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2614                                 if (!recover_lost_locks)
2615                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2616                         } else if (status != NFS_OK) {
2617                                 ret = status;
2618                                 nfs4_put_lock_state(prev);
2619                                 goto out;
2620                         }
2621                         spin_lock(&state->state_lock);
2622                 }
2623         }
2624         spin_unlock(&state->state_lock);
2625         nfs4_put_lock_state(prev);
2626 out:
2627         return ret;
2628 }
2629
2630 /**
2631  * nfs41_check_open_stateid - possibly free an open stateid
2632  *
2633  * @state: NFSv4 state for an inode
2634  *
2635  * Returns NFS_OK if recovery for this stateid is now finished.
2636  * Otherwise a negative NFS4ERR value is returned.
2637  */
2638 static int nfs41_check_open_stateid(struct nfs4_state *state)
2639 {
2640         struct nfs_server *server = NFS_SERVER(state->inode);
2641         nfs4_stateid *stateid = &state->open_stateid;
2642         struct rpc_cred *cred = state->owner->so_cred;
2643         int status;
2644
2645         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2646                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2647                         if (nfs4_have_delegation(state->inode, state->state))
2648                                 return NFS_OK;
2649                         return -NFS4ERR_OPENMODE;
2650                 }
2651                 return -NFS4ERR_BAD_STATEID;
2652         }
2653         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2654         trace_nfs4_test_open_stateid(state, NULL, status);
2655         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2656                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2657                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2658                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2659                 clear_bit(NFS_OPEN_STATE, &state->flags);
2660                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2661         }
2662         if (status != NFS_OK)
2663                 return status;
2664         if (nfs_open_stateid_recover_openmode(state))
2665                 return -NFS4ERR_OPENMODE;
2666         return NFS_OK;
2667 }
2668
2669 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2670 {
2671         int status;
2672
2673         nfs41_check_delegation_stateid(state);
2674         status = nfs41_check_expired_locks(state);
2675         if (status != NFS_OK)
2676                 return status;
2677         status = nfs41_check_open_stateid(state);
2678         if (status != NFS_OK)
2679                 status = nfs4_open_expired(sp, state);
2680         return status;
2681 }
2682 #endif
2683
2684 /*
2685  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2686  * fields corresponding to attributes that were used to store the verifier.
2687  * Make sure we clobber those fields in the later setattr call
2688  */
2689 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2690                                 struct iattr *sattr, struct nfs4_label **label)
2691 {
2692         const u32 *attrset = opendata->o_res.attrset;
2693
2694         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2695             !(sattr->ia_valid & ATTR_ATIME_SET))
2696                 sattr->ia_valid |= ATTR_ATIME;
2697
2698         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2699             !(sattr->ia_valid & ATTR_MTIME_SET))
2700                 sattr->ia_valid |= ATTR_MTIME;
2701
2702         /* Except MODE, it seems harmless of setting twice. */
2703         if (opendata->o_arg.createmode != NFS4_CREATE_EXCLUSIVE &&
2704                 attrset[1] & FATTR4_WORD1_MODE)
2705                 sattr->ia_valid &= ~ATTR_MODE;
2706
2707         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2708                 *label = NULL;
2709 }
2710
2711 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2712                 fmode_t fmode,
2713                 int flags,
2714                 struct nfs_open_context *ctx)
2715 {
2716         struct nfs4_state_owner *sp = opendata->owner;
2717         struct nfs_server *server = sp->so_server;
2718         struct dentry *dentry;
2719         struct nfs4_state *state;
2720         unsigned int seq;
2721         int ret;
2722
2723         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2724
2725         ret = _nfs4_proc_open(opendata);
2726         if (ret != 0)
2727                 goto out;
2728
2729         state = nfs4_opendata_to_nfs4_state(opendata);
2730         ret = PTR_ERR(state);
2731         if (IS_ERR(state))
2732                 goto out;
2733         if (server->caps & NFS_CAP_POSIX_LOCK)
2734                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2735         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2736                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2737
2738         dentry = opendata->dentry;
2739         if (d_really_is_negative(dentry)) {
2740                 struct dentry *alias;
2741                 d_drop(dentry);
2742                 alias = d_exact_alias(dentry, state->inode);
2743                 if (!alias)
2744                         alias = d_splice_alias(igrab(state->inode), dentry);
2745                 /* d_splice_alias() can't fail here - it's a non-directory */
2746                 if (alias) {
2747                         dput(ctx->dentry);
2748                         ctx->dentry = dentry = alias;
2749                 }
2750                 nfs_set_verifier(dentry,
2751                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2752         }
2753
2754         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2755         if (ret != 0)
2756                 goto out;
2757
2758         ctx->state = state;
2759         if (d_inode(dentry) == state->inode) {
2760                 nfs_inode_attach_open_context(ctx);
2761                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2762                         nfs4_schedule_stateid_recovery(server, state);
2763         }
2764 out:
2765         return ret;
2766 }
2767
2768 /*
2769  * Returns a referenced nfs4_state
2770  */
2771 static int _nfs4_do_open(struct inode *dir,
2772                         struct nfs_open_context *ctx,
2773                         int flags,
2774                         struct iattr *sattr,
2775                         struct nfs4_label *label,
2776                         int *opened)
2777 {
2778         struct nfs4_state_owner  *sp;
2779         struct nfs4_state     *state = NULL;
2780         struct nfs_server       *server = NFS_SERVER(dir);
2781         struct nfs4_opendata *opendata;
2782         struct dentry *dentry = ctx->dentry;
2783         struct rpc_cred *cred = ctx->cred;
2784         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2785         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2786         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2787         struct nfs4_label *olabel = NULL;
2788         int status;
2789
2790         /* Protect against reboot recovery conflicts */
2791         status = -ENOMEM;
2792         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2793         if (sp == NULL) {
2794                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2795                 goto out_err;
2796         }
2797         status = nfs4_recover_expired_lease(server);
2798         if (status != 0)
2799                 goto err_put_state_owner;
2800         if (d_really_is_positive(dentry))
2801                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2802         status = -ENOMEM;
2803         if (d_really_is_positive(dentry))
2804                 claim = NFS4_OPEN_CLAIM_FH;
2805         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2806                         label, claim, GFP_KERNEL);
2807         if (opendata == NULL)
2808                 goto err_put_state_owner;
2809
2810         if (label) {
2811                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2812                 if (IS_ERR(olabel)) {
2813                         status = PTR_ERR(olabel);
2814                         goto err_opendata_put;
2815                 }
2816         }
2817
2818         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2819                 if (!opendata->f_attr.mdsthreshold) {
2820                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2821                         if (!opendata->f_attr.mdsthreshold)
2822                                 goto err_free_label;
2823                 }
2824                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2825         }
2826         if (d_really_is_positive(dentry))
2827                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2828
2829         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2830         if (status != 0)
2831                 goto err_free_label;
2832         state = ctx->state;
2833
2834         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2835             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2836                 nfs4_exclusive_attrset(opendata, sattr, &label);
2837                 /*
2838                  * send create attributes which was not set by open
2839                  * with an extra setattr.
2840                  */
2841                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2842                         nfs_fattr_init(opendata->o_res.f_attr);
2843                         status = nfs4_do_setattr(state->inode, cred,
2844                                         opendata->o_res.f_attr, sattr,
2845                                         ctx, label, olabel);
2846                         if (status == 0) {
2847                                 nfs_setattr_update_inode(state->inode, sattr,
2848                                                 opendata->o_res.f_attr);
2849                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2850                         }
2851                 }
2852         }
2853         if (opened && opendata->file_created)
2854                 *opened |= FILE_CREATED;
2855
2856         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2857                 *ctx_th = opendata->f_attr.mdsthreshold;
2858                 opendata->f_attr.mdsthreshold = NULL;
2859         }
2860
2861         nfs4_label_free(olabel);
2862
2863         nfs4_opendata_put(opendata);
2864         nfs4_put_state_owner(sp);
2865         return 0;
2866 err_free_label:
2867         nfs4_label_free(olabel);
2868 err_opendata_put:
2869         nfs4_opendata_put(opendata);
2870 err_put_state_owner:
2871         nfs4_put_state_owner(sp);
2872 out_err:
2873         return status;
2874 }
2875
2876
2877 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2878                                         struct nfs_open_context *ctx,
2879                                         int flags,
2880                                         struct iattr *sattr,
2881                                         struct nfs4_label *label,
2882                                         int *opened)
2883 {
2884         struct nfs_server *server = NFS_SERVER(dir);
2885         struct nfs4_exception exception = { };
2886         struct nfs4_state *res;
2887         int status;
2888
2889         do {
2890                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2891                 res = ctx->state;
2892                 trace_nfs4_open_file(ctx, flags, status);
2893                 if (status == 0)
2894                         break;
2895                 /* NOTE: BAD_SEQID means the server and client disagree about the
2896                  * book-keeping w.r.t. state-changing operations
2897                  * (OPEN/CLOSE/LOCK/LOCKU...)
2898                  * It is actually a sign of a bug on the client or on the server.
2899                  *
2900                  * If we receive a BAD_SEQID error in the particular case of
2901                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2902                  * have unhashed the old state_owner for us, and that we can
2903                  * therefore safely retry using a new one. We should still warn
2904                  * the user though...
2905                  */
2906                 if (status == -NFS4ERR_BAD_SEQID) {
2907                         pr_warn_ratelimited("NFS: v4 server %s "
2908                                         " returned a bad sequence-id error!\n",
2909                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2910                         exception.retry = 1;
2911                         continue;
2912                 }
2913                 /*
2914                  * BAD_STATEID on OPEN means that the server cancelled our
2915                  * state before it received the OPEN_CONFIRM.
2916                  * Recover by retrying the request as per the discussion
2917                  * on Page 181 of RFC3530.
2918                  */
2919                 if (status == -NFS4ERR_BAD_STATEID) {
2920                         exception.retry = 1;
2921                         continue;
2922                 }
2923                 if (status == -EAGAIN) {
2924                         /* We must have found a delegation */
2925                         exception.retry = 1;
2926                         continue;
2927                 }
2928                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2929                         continue;
2930                 res = ERR_PTR(nfs4_handle_exception(server,
2931                                         status, &exception));
2932         } while (exception.retry);
2933         return res;
2934 }
2935
2936 static int _nfs4_do_setattr(struct inode *inode,
2937                             struct nfs_setattrargs *arg,
2938                             struct nfs_setattrres *res,
2939                             struct rpc_cred *cred,
2940                             struct nfs_open_context *ctx)
2941 {
2942         struct nfs_server *server = NFS_SERVER(inode);
2943         struct rpc_message msg = {
2944                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2945                 .rpc_argp       = arg,
2946                 .rpc_resp       = res,
2947                 .rpc_cred       = cred,
2948         };
2949         struct rpc_cred *delegation_cred = NULL;
2950         unsigned long timestamp = jiffies;
2951         fmode_t fmode;
2952         bool truncate;
2953         int status;
2954
2955         nfs_fattr_init(res->fattr);
2956
2957         /* Servers should only apply open mode checks for file size changes */
2958         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2959         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2960
2961         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2962                 /* Use that stateid */
2963         } else if (truncate && ctx != NULL) {
2964                 struct nfs_lock_context *l_ctx;
2965                 if (!nfs4_valid_open_stateid(ctx->state))
2966                         return -EBADF;
2967                 l_ctx = nfs_get_lock_context(ctx);
2968                 if (IS_ERR(l_ctx))
2969                         return PTR_ERR(l_ctx);
2970                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2971                                                 &arg->stateid, &delegation_cred);
2972                 nfs_put_lock_context(l_ctx);
2973                 if (status == -EIO)
2974                         return -EBADF;
2975         } else
2976                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2977         if (delegation_cred)
2978                 msg.rpc_cred = delegation_cred;
2979
2980         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2981
2982         put_rpccred(delegation_cred);
2983         if (status == 0 && ctx != NULL)
2984                 renew_lease(server, timestamp);
2985         trace_nfs4_setattr(inode, &arg->stateid, status);
2986         return status;
2987 }
2988
2989 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2990                            struct nfs_fattr *fattr, struct iattr *sattr,
2991                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2992                            struct nfs4_label *olabel)
2993 {
2994         struct nfs_server *server = NFS_SERVER(inode);
2995         struct nfs4_state *state = ctx ? ctx->state : NULL;
2996         struct nfs_setattrargs  arg = {
2997                 .fh             = NFS_FH(inode),
2998                 .iap            = sattr,
2999                 .server         = server,
3000                 .bitmask = server->attr_bitmask,
3001                 .label          = ilabel,
3002         };
3003         struct nfs_setattrres  res = {
3004                 .fattr          = fattr,
3005                 .label          = olabel,
3006                 .server         = server,
3007         };
3008         struct nfs4_exception exception = {
3009                 .state = state,
3010                 .inode = inode,
3011                 .stateid = &arg.stateid,
3012         };
3013         int err;
3014
3015         arg.bitmask = nfs4_bitmask(server, ilabel);
3016         if (ilabel)
3017                 arg.bitmask = nfs4_bitmask(server, olabel);
3018
3019         do {
3020                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3021                 switch (err) {
3022                 case -NFS4ERR_OPENMODE:
3023                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3024                                 pr_warn_once("NFSv4: server %s is incorrectly "
3025                                                 "applying open mode checks to "
3026                                                 "a SETATTR that is not "
3027                                                 "changing file size.\n",
3028                                                 server->nfs_client->cl_hostname);
3029                         }
3030                         if (state && !(state->state & FMODE_WRITE)) {
3031                                 err = -EBADF;
3032                                 if (sattr->ia_valid & ATTR_OPEN)
3033                                         err = -EACCES;
3034                                 goto out;
3035                         }
3036                 }
3037                 err = nfs4_handle_exception(server, err, &exception);
3038         } while (exception.retry);
3039 out:
3040         return err;
3041 }
3042
3043 static bool
3044 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3045 {
3046         if (inode == NULL || !nfs_have_layout(inode))
3047                 return false;
3048
3049         return pnfs_wait_on_layoutreturn(inode, task);
3050 }
3051
3052 struct nfs4_closedata {
3053         struct inode *inode;
3054         struct nfs4_state *state;
3055         struct nfs_closeargs arg;
3056         struct nfs_closeres res;
3057         struct {
3058                 struct nfs4_layoutreturn_args arg;
3059                 struct nfs4_layoutreturn_res res;
3060                 struct nfs4_xdr_opaque_data ld_private;
3061                 u32 roc_barrier;
3062                 bool roc;
3063         } lr;
3064         struct nfs_fattr fattr;
3065         unsigned long timestamp;
3066 };
3067
3068 static void nfs4_free_closedata(void *data)
3069 {
3070         struct nfs4_closedata *calldata = data;
3071         struct nfs4_state_owner *sp = calldata->state->owner;
3072         struct super_block *sb = calldata->state->inode->i_sb;
3073
3074         if (calldata->lr.roc)
3075                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3076                                 calldata->res.lr_ret);
3077         nfs4_put_open_state(calldata->state);
3078         nfs_free_seqid(calldata->arg.seqid);
3079         nfs4_put_state_owner(sp);
3080         nfs_sb_deactive(sb);
3081         kfree(calldata);
3082 }
3083
3084 static void nfs4_close_done(struct rpc_task *task, void *data)
3085 {
3086         struct nfs4_closedata *calldata = data;
3087         struct nfs4_state *state = calldata->state;
3088         struct nfs_server *server = NFS_SERVER(calldata->inode);
3089         nfs4_stateid *res_stateid = NULL;
3090
3091         dprintk("%s: begin!\n", __func__);
3092         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3093                 return;
3094         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3095
3096         /* Handle Layoutreturn errors */
3097         if (calldata->arg.lr_args && task->tk_status != 0) {
3098                 switch (calldata->res.lr_ret) {
3099                 default:
3100                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3101                         break;
3102                 case 0:
3103                         calldata->arg.lr_args = NULL;
3104                         calldata->res.lr_res = NULL;
3105                         break;
3106                 case -NFS4ERR_ADMIN_REVOKED:
3107                 case -NFS4ERR_DELEG_REVOKED:
3108                 case -NFS4ERR_EXPIRED:
3109                 case -NFS4ERR_BAD_STATEID:
3110                 case -NFS4ERR_OLD_STATEID:
3111                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3112                 case -NFS4ERR_WRONG_CRED:
3113                         calldata->arg.lr_args = NULL;
3114                         calldata->res.lr_res = NULL;
3115                         calldata->res.lr_ret = 0;
3116                         rpc_restart_call_prepare(task);
3117                         return;
3118                 }
3119         }
3120
3121         /* hmm. we are done with the inode, and in the process of freeing
3122          * the state_owner. we keep this around to process errors
3123          */
3124         switch (task->tk_status) {
3125                 case 0:
3126                         res_stateid = &calldata->res.stateid;
3127                         renew_lease(server, calldata->timestamp);
3128                         break;
3129                 case -NFS4ERR_ACCESS:
3130                         if (calldata->arg.bitmask != NULL) {
3131                                 calldata->arg.bitmask = NULL;
3132                                 calldata->res.fattr = NULL;
3133                                 task->tk_status = 0;
3134                                 rpc_restart_call_prepare(task);
3135                                 goto out_release;
3136
3137                         }
3138                         break;
3139                 case -NFS4ERR_ADMIN_REVOKED:
3140                 case -NFS4ERR_STALE_STATEID:
3141                 case -NFS4ERR_EXPIRED:
3142                         nfs4_free_revoked_stateid(server,
3143                                         &calldata->arg.stateid,
3144                                         task->tk_msg.rpc_cred);
3145                 case -NFS4ERR_OLD_STATEID:
3146                 case -NFS4ERR_BAD_STATEID:
3147                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3148                                                 &state->open_stateid)) {
3149                                 rpc_restart_call_prepare(task);
3150                                 goto out_release;
3151                         }
3152                         if (calldata->arg.fmode == 0)
3153                                 break;
3154                 default:
3155                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3156                                 rpc_restart_call_prepare(task);
3157                                 goto out_release;
3158                         }
3159         }
3160         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3161                         res_stateid, calldata->arg.fmode);
3162 out_release:
3163         nfs_release_seqid(calldata->arg.seqid);
3164         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3165         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3166 }
3167
3168 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3169 {
3170         struct nfs4_closedata *calldata = data;
3171         struct nfs4_state *state = calldata->state;
3172         struct inode *inode = calldata->inode;
3173         bool is_rdonly, is_wronly, is_rdwr;
3174         int call_close = 0;
3175
3176         dprintk("%s: begin!\n", __func__);
3177         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3178                 goto out_wait;
3179
3180         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3181         spin_lock(&state->owner->so_lock);
3182         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3183         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3184         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3185         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3186         /* Calculate the change in open mode */
3187         calldata->arg.fmode = 0;
3188         if (state->n_rdwr == 0) {
3189                 if (state->n_rdonly == 0)
3190                         call_close |= is_rdonly;
3191                 else if (is_rdonly)
3192                         calldata->arg.fmode |= FMODE_READ;
3193                 if (state->n_wronly == 0)
3194                         call_close |= is_wronly;
3195                 else if (is_wronly)
3196                         calldata->arg.fmode |= FMODE_WRITE;
3197                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3198                         call_close |= is_rdwr;
3199         } else if (is_rdwr)
3200                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3201
3202         if (!nfs4_valid_open_stateid(state) ||
3203             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3204                 call_close = 0;
3205         spin_unlock(&state->owner->so_lock);
3206
3207         if (!call_close) {
3208                 /* Note: exit _without_ calling nfs4_close_done */
3209                 goto out_no_action;
3210         }
3211
3212         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3213                 nfs_release_seqid(calldata->arg.seqid);
3214                 goto out_wait;
3215         }
3216
3217         if (calldata->arg.fmode == 0)
3218                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3219
3220         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3221                 /* Close-to-open cache consistency revalidation */
3222                 if (!nfs4_have_delegation(inode, FMODE_READ))
3223                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3224                 else
3225                         calldata->arg.bitmask = NULL;
3226         }
3227
3228         calldata->arg.share_access =
3229                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3230                                 calldata->arg.fmode, 0);
3231
3232         if (calldata->res.fattr == NULL)
3233                 calldata->arg.bitmask = NULL;
3234         else if (calldata->arg.bitmask == NULL)
3235                 calldata->res.fattr = NULL;
3236         calldata->timestamp = jiffies;
3237         if (nfs4_setup_sequence(NFS_SERVER(inode),
3238                                 &calldata->arg.seq_args,
3239                                 &calldata->res.seq_res,
3240                                 task) != 0)
3241                 nfs_release_seqid(calldata->arg.seqid);
3242         dprintk("%s: done!\n", __func__);
3243         return;
3244 out_no_action:
3245         task->tk_action = NULL;
3246 out_wait:
3247         nfs4_sequence_done(task, &calldata->res.seq_res);
3248 }
3249
3250 static const struct rpc_call_ops nfs4_close_ops = {
3251         .rpc_call_prepare = nfs4_close_prepare,
3252         .rpc_call_done = nfs4_close_done,
3253         .rpc_release = nfs4_free_closedata,
3254 };
3255
3256 /* 
3257  * It is possible for data to be read/written from a mem-mapped file 
3258  * after the sys_close call (which hits the vfs layer as a flush).
3259  * This means that we can't safely call nfsv4 close on a file until 
3260  * the inode is cleared. This in turn means that we are not good
3261  * NFSv4 citizens - we do not indicate to the server to update the file's 
3262  * share state even when we are done with one of the three share 
3263  * stateid's in the inode.
3264  *
3265  * NOTE: Caller must be holding the sp->so_owner semaphore!
3266  */
3267 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3268 {
3269         struct nfs_server *server = NFS_SERVER(state->inode);
3270         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3271         struct nfs4_closedata *calldata;
3272         struct nfs4_state_owner *sp = state->owner;
3273         struct rpc_task *task;
3274         struct rpc_message msg = {
3275                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3276                 .rpc_cred = state->owner->so_cred,
3277         };
3278         struct rpc_task_setup task_setup_data = {
3279                 .rpc_client = server->client,
3280                 .rpc_message = &msg,
3281                 .callback_ops = &nfs4_close_ops,
3282                 .workqueue = nfsiod_workqueue,
3283                 .flags = RPC_TASK_ASYNC,
3284         };
3285         int status = -ENOMEM;
3286
3287         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3288                 &task_setup_data.rpc_client, &msg);
3289
3290         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3291         if (calldata == NULL)
3292                 goto out;
3293         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3294         calldata->inode = state->inode;
3295         calldata->state = state;
3296         calldata->arg.fh = NFS_FH(state->inode);
3297         /* Serialization for the sequence id */
3298         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3299         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3300         if (IS_ERR(calldata->arg.seqid))
3301                 goto out_free_calldata;
3302         nfs_fattr_init(&calldata->fattr);
3303         calldata->arg.fmode = 0;
3304         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3305         calldata->res.fattr = &calldata->fattr;
3306         calldata->res.seqid = calldata->arg.seqid;
3307         calldata->res.server = server;
3308         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3309         calldata->lr.roc = pnfs_roc(state->inode,
3310                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3311         if (calldata->lr.roc) {
3312                 calldata->arg.lr_args = &calldata->lr.arg;
3313                 calldata->res.lr_res = &calldata->lr.res;
3314         }
3315         nfs_sb_active(calldata->inode->i_sb);
3316
3317         msg.rpc_argp = &calldata->arg;
3318         msg.rpc_resp = &calldata->res;
3319         task_setup_data.callback_data = calldata;
3320         task = rpc_run_task(&task_setup_data);
3321         if (IS_ERR(task))
3322                 return PTR_ERR(task);
3323         status = 0;
3324         if (wait)
3325                 status = rpc_wait_for_completion_task(task);
3326         rpc_put_task(task);
3327         return status;
3328 out_free_calldata:
3329         kfree(calldata);
3330 out:
3331         nfs4_put_open_state(state);
3332         nfs4_put_state_owner(sp);
3333         return status;
3334 }
3335
3336 static struct inode *
3337 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3338                 int open_flags, struct iattr *attr, int *opened)
3339 {
3340         struct nfs4_state *state;
3341         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3342
3343         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3344
3345         /* Protect against concurrent sillydeletes */
3346         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3347
3348         nfs4_label_release_security(label);
3349
3350         if (IS_ERR(state))
3351                 return ERR_CAST(state);
3352         return state->inode;
3353 }
3354
3355 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3356 {
3357         if (ctx->state == NULL)
3358                 return;
3359         if (is_sync)
3360                 nfs4_close_sync(ctx->state, ctx->mode);
3361         else
3362                 nfs4_close_state(ctx->state, ctx->mode);
3363 }
3364
3365 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3366 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3367 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3368
3369 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3370 {
3371         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3372         struct nfs4_server_caps_arg args = {
3373                 .fhandle = fhandle,
3374                 .bitmask = bitmask,
3375         };
3376         struct nfs4_server_caps_res res = {};
3377         struct rpc_message msg = {
3378                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3379                 .rpc_argp = &args,
3380                 .rpc_resp = &res,
3381         };
3382         int status;
3383
3384         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3385                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3386                      FATTR4_WORD0_LINK_SUPPORT |
3387                      FATTR4_WORD0_SYMLINK_SUPPORT |
3388                      FATTR4_WORD0_ACLSUPPORT;
3389         if (minorversion)
3390                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3391
3392         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3393         if (status == 0) {
3394                 /* Sanity check the server answers */
3395                 switch (minorversion) {
3396                 case 0:
3397                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3398                         res.attr_bitmask[2] = 0;
3399                         break;
3400                 case 1:
3401                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3402                         break;
3403                 case 2:
3404                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3405                 }
3406                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3407                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3408                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3409                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3410                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3411                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3412                                 NFS_CAP_SECURITY_LABEL);
3413                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3414                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3415                         server->caps |= NFS_CAP_ACLS;
3416                 if (res.has_links != 0)
3417                         server->caps |= NFS_CAP_HARDLINKS;
3418                 if (res.has_symlinks != 0)
3419                         server->caps |= NFS_CAP_SYMLINKS;
3420                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3421                         server->caps |= NFS_CAP_FILEID;
3422                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3423                         server->caps |= NFS_CAP_MODE;
3424                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3425                         server->caps |= NFS_CAP_NLINK;
3426                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3427                         server->caps |= NFS_CAP_OWNER;
3428                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3429                         server->caps |= NFS_CAP_OWNER_GROUP;
3430                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3431                         server->caps |= NFS_CAP_ATIME;
3432                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3433                         server->caps |= NFS_CAP_CTIME;
3434                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3435                         server->caps |= NFS_CAP_MTIME;
3436 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3437                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3438                         server->caps |= NFS_CAP_SECURITY_LABEL;
3439 #endif
3440                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3441                                 sizeof(server->attr_bitmask));
3442                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3443
3444                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3445                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3446                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3447                 server->cache_consistency_bitmask[2] = 0;
3448                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3449                         sizeof(server->exclcreat_bitmask));
3450                 server->acl_bitmask = res.acl_bitmask;
3451                 server->fh_expire_type = res.fh_expire_type;
3452         }
3453
3454         return status;
3455 }
3456
3457 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3458 {
3459         struct nfs4_exception exception = { };
3460         int err;
3461         do {
3462                 err = nfs4_handle_exception(server,
3463                                 _nfs4_server_capabilities(server, fhandle),
3464                                 &exception);
3465         } while (exception.retry);
3466         return err;
3467 }
3468
3469 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3470                 struct nfs_fsinfo *info)
3471 {
3472         u32 bitmask[3];
3473         struct nfs4_lookup_root_arg args = {
3474                 .bitmask = bitmask,
3475         };
3476         struct nfs4_lookup_res res = {
3477                 .server = server,
3478                 .fattr = info->fattr,
3479                 .fh = fhandle,
3480         };
3481         struct rpc_message msg = {
3482                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3483                 .rpc_argp = &args,
3484                 .rpc_resp = &res,
3485         };
3486
3487         bitmask[0] = nfs4_fattr_bitmap[0];
3488         bitmask[1] = nfs4_fattr_bitmap[1];
3489         /*
3490          * Process the label in the upcoming getfattr
3491          */
3492         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3493
3494         nfs_fattr_init(info->fattr);
3495         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3496 }
3497
3498 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3499                 struct nfs_fsinfo *info)
3500 {
3501         struct nfs4_exception exception = { };
3502         int err;
3503         do {
3504                 err = _nfs4_lookup_root(server, fhandle, info);
3505                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3506                 switch (err) {
3507                 case 0:
3508                 case -NFS4ERR_WRONGSEC:
3509                         goto out;
3510                 default:
3511                         err = nfs4_handle_exception(server, err, &exception);
3512                 }
3513         } while (exception.retry);
3514 out:
3515         return err;
3516 }
3517
3518 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3519                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3520 {
3521         struct rpc_auth_create_args auth_args = {
3522                 .pseudoflavor = flavor,
3523         };
3524         struct rpc_auth *auth;
3525         int ret;
3526
3527         auth = rpcauth_create(&auth_args, server->client);
3528         if (IS_ERR(auth)) {
3529                 ret = -EACCES;
3530                 goto out;
3531         }
3532         ret = nfs4_lookup_root(server, fhandle, info);
3533 out:
3534         return ret;
3535 }
3536
3537 /*
3538  * Retry pseudoroot lookup with various security flavors.  We do this when:
3539  *
3540  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3541  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3542  *
3543  * Returns zero on success, or a negative NFS4ERR value, or a
3544  * negative errno value.
3545  */
3546 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3547                               struct nfs_fsinfo *info)
3548 {
3549         /* Per 3530bis 15.33.5 */
3550         static const rpc_authflavor_t flav_array[] = {
3551                 RPC_AUTH_GSS_KRB5P,
3552                 RPC_AUTH_GSS_KRB5I,
3553                 RPC_AUTH_GSS_KRB5,
3554                 RPC_AUTH_UNIX,                  /* courtesy */
3555                 RPC_AUTH_NULL,
3556         };
3557         int status = -EPERM;
3558         size_t i;
3559
3560         if (server->auth_info.flavor_len > 0) {
3561                 /* try each flavor specified by user */
3562                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3563                         status = nfs4_lookup_root_sec(server, fhandle, info,
3564                                                 server->auth_info.flavors[i]);
3565                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3566                                 continue;
3567                         break;
3568                 }
3569         } else {
3570                 /* no flavors specified by user, try default list */
3571                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3572                         status = nfs4_lookup_root_sec(server, fhandle, info,
3573                                                       flav_array[i]);
3574                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3575                                 continue;
3576                         break;
3577                 }
3578         }
3579
3580         /*
3581          * -EACCESS could mean that the user doesn't have correct permissions
3582          * to access the mount.  It could also mean that we tried to mount
3583          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3584          * existing mount programs don't handle -EACCES very well so it should
3585          * be mapped to -EPERM instead.
3586          */
3587         if (status == -EACCES)
3588                 status = -EPERM;
3589         return status;
3590 }
3591
3592 /**
3593  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3594  * @server: initialized nfs_server handle
3595  * @fhandle: we fill in the pseudo-fs root file handle
3596  * @info: we fill in an FSINFO struct
3597  * @auth_probe: probe the auth flavours
3598  *
3599  * Returns zero on success, or a negative errno.
3600  */
3601 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3602                          struct nfs_fsinfo *info,
3603                          bool auth_probe)
3604 {
3605         int status = 0;
3606
3607         if (!auth_probe)
3608                 status = nfs4_lookup_root(server, fhandle, info);
3609
3610         if (auth_probe || status == NFS4ERR_WRONGSEC)
3611                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3612                                 fhandle, info);
3613
3614         if (status == 0)
3615                 status = nfs4_server_capabilities(server, fhandle);
3616         if (status == 0)
3617                 status = nfs4_do_fsinfo(server, fhandle, info);
3618
3619         return nfs4_map_errors(status);
3620 }
3621
3622 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3623                               struct nfs_fsinfo *info)
3624 {
3625         int error;
3626         struct nfs_fattr *fattr = info->fattr;
3627         struct nfs4_label *label = NULL;
3628
3629         error = nfs4_server_capabilities(server, mntfh);
3630         if (error < 0) {
3631                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3632                 return error;
3633         }
3634
3635         label = nfs4_label_alloc(server, GFP_KERNEL);
3636         if (IS_ERR(label))
3637                 return PTR_ERR(label);
3638
3639         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3640         if (error < 0) {
3641                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3642                 goto err_free_label;
3643         }
3644
3645         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3646             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3647                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3648
3649 err_free_label:
3650         nfs4_label_free(label);
3651
3652         return error;
3653 }
3654
3655 /*
3656  * Get locations and (maybe) other attributes of a referral.
3657  * Note that we'll actually follow the referral later when
3658  * we detect fsid mismatch in inode revalidation
3659  */
3660 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3661                              const struct qstr *name, struct nfs_fattr *fattr,
3662                              struct nfs_fh *fhandle)
3663 {
3664         int status = -ENOMEM;
3665         struct page *page = NULL;
3666         struct nfs4_fs_locations *locations = NULL;
3667
3668         page = alloc_page(GFP_KERNEL);
3669         if (page == NULL)
3670                 goto out;
3671         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3672         if (locations == NULL)
3673                 goto out;
3674
3675         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3676         if (status != 0)
3677                 goto out;
3678
3679         /*
3680          * If the fsid didn't change, this is a migration event, not a
3681          * referral.  Cause us to drop into the exception handler, which
3682          * will kick off migration recovery.
3683          */
3684         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3685                 dprintk("%s: server did not return a different fsid for"
3686                         " a referral at %s\n", __func__, name->name);
3687                 status = -NFS4ERR_MOVED;
3688                 goto out;
3689         }
3690         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3691         nfs_fixup_referral_attributes(&locations->fattr);
3692
3693         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3694         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3695         memset(fhandle, 0, sizeof(struct nfs_fh));
3696 out:
3697         if (page)
3698                 __free_page(page);
3699         kfree(locations);
3700         return status;
3701 }
3702
3703 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3704                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3705 {
3706         struct nfs4_getattr_arg args = {
3707                 .fh = fhandle,
3708                 .bitmask = server->attr_bitmask,
3709         };
3710         struct nfs4_getattr_res res = {
3711                 .fattr = fattr,
3712                 .label = label,
3713                 .server = server,
3714         };
3715         struct rpc_message msg = {
3716                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3717                 .rpc_argp = &args,
3718                 .rpc_resp = &res,
3719         };
3720
3721         args.bitmask = nfs4_bitmask(server, label);
3722
3723         nfs_fattr_init(fattr);
3724         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3725 }
3726
3727 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3728                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3729 {
3730         struct nfs4_exception exception = { };
3731         int err;
3732         do {
3733                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3734                 trace_nfs4_getattr(server, fhandle, fattr, err);
3735                 err = nfs4_handle_exception(server, err,
3736                                 &exception);
3737         } while (exception.retry);
3738         return err;
3739 }
3740
3741 /* 
3742  * The file is not closed if it is opened due to the a request to change
3743  * the size of the file. The open call will not be needed once the
3744  * VFS layer lookup-intents are implemented.
3745  *
3746  * Close is called when the inode is destroyed.
3747  * If we haven't opened the file for O_WRONLY, we
3748  * need to in the size_change case to obtain a stateid.
3749  *
3750  * Got race?
3751  * Because OPEN is always done by name in nfsv4, it is
3752  * possible that we opened a different file by the same
3753  * name.  We can recognize this race condition, but we
3754  * can't do anything about it besides returning an error.
3755  *
3756  * This will be fixed with VFS changes (lookup-intent).
3757  */
3758 static int
3759 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3760                   struct iattr *sattr)
3761 {
3762         struct inode *inode = d_inode(dentry);
3763         struct rpc_cred *cred = NULL;
3764         struct nfs_open_context *ctx = NULL;
3765         struct nfs4_label *label = NULL;
3766         int status;
3767
3768         if (pnfs_ld_layoutret_on_setattr(inode) &&
3769             sattr->ia_valid & ATTR_SIZE &&
3770             sattr->ia_size < i_size_read(inode))
3771                 pnfs_commit_and_return_layout(inode);
3772
3773         nfs_fattr_init(fattr);
3774         
3775         /* Deal with open(O_TRUNC) */
3776         if (sattr->ia_valid & ATTR_OPEN)
3777                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3778
3779         /* Optimization: if the end result is no change, don't RPC */
3780         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3781                 return 0;
3782
3783         /* Search for an existing open(O_WRITE) file */
3784         if (sattr->ia_valid & ATTR_FILE) {
3785
3786                 ctx = nfs_file_open_context(sattr->ia_file);
3787                 if (ctx)
3788                         cred = ctx->cred;
3789         }
3790
3791         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3792         if (IS_ERR(label))
3793                 return PTR_ERR(label);
3794
3795         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3796         if (status == 0) {
3797                 nfs_setattr_update_inode(inode, sattr, fattr);
3798                 nfs_setsecurity(inode, fattr, label);
3799         }
3800         nfs4_label_free(label);
3801         return status;
3802 }
3803
3804 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3805                 const struct qstr *name, struct nfs_fh *fhandle,
3806                 struct nfs_fattr *fattr, struct nfs4_label *label)
3807 {
3808         struct nfs_server *server = NFS_SERVER(dir);
3809         int                    status;
3810         struct nfs4_lookup_arg args = {
3811                 .bitmask = server->attr_bitmask,
3812                 .dir_fh = NFS_FH(dir),
3813                 .name = name,
3814         };
3815         struct nfs4_lookup_res res = {
3816                 .server = server,
3817                 .fattr = fattr,
3818                 .label = label,
3819                 .fh = fhandle,
3820         };
3821         struct rpc_message msg = {
3822                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3823                 .rpc_argp = &args,
3824                 .rpc_resp = &res,
3825         };
3826
3827         args.bitmask = nfs4_bitmask(server, label);
3828
3829         nfs_fattr_init(fattr);
3830
3831         dprintk("NFS call  lookup %s\n", name->name);
3832         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3833         dprintk("NFS reply lookup: %d\n", status);
3834         return status;
3835 }
3836
3837 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3838 {
3839         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3840                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3841         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3842         fattr->nlink = 2;
3843 }
3844
3845 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3846                                    const struct qstr *name, struct nfs_fh *fhandle,
3847                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3848 {
3849         struct nfs4_exception exception = { };
3850         struct rpc_clnt *client = *clnt;
3851         int err;
3852         do {
3853                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3854                 trace_nfs4_lookup(dir, name, err);
3855                 switch (err) {
3856                 case -NFS4ERR_BADNAME:
3857                         err = -ENOENT;
3858                         goto out;
3859                 case -NFS4ERR_MOVED:
3860                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3861                         if (err == -NFS4ERR_MOVED)
3862                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3863                         goto out;
3864                 case -NFS4ERR_WRONGSEC:
3865                         err = -EPERM;
3866                         if (client != *clnt)
3867                                 goto out;
3868                         client = nfs4_negotiate_security(client, dir, name);
3869                         if (IS_ERR(client))
3870                                 return PTR_ERR(client);
3871
3872                         exception.retry = 1;
3873                         break;
3874                 default:
3875                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3876                 }
3877         } while (exception.retry);
3878
3879 out:
3880         if (err == 0)
3881                 *clnt = client;
3882         else if (client != *clnt)
3883                 rpc_shutdown_client(client);
3884
3885         return err;
3886 }
3887
3888 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3889                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3890                             struct nfs4_label *label)
3891 {
3892         int status;
3893         struct rpc_clnt *client = NFS_CLIENT(dir);
3894
3895         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3896         if (client != NFS_CLIENT(dir)) {
3897                 rpc_shutdown_client(client);
3898                 nfs_fixup_secinfo_attributes(fattr);
3899         }
3900         return status;
3901 }
3902
3903 struct rpc_clnt *
3904 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3905                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3906 {
3907         struct rpc_clnt *client = NFS_CLIENT(dir);
3908         int status;
3909
3910         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3911         if (status < 0)
3912                 return ERR_PTR(status);
3913         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3914 }
3915
3916 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3917 {
3918         struct nfs_server *server = NFS_SERVER(inode);
3919         struct nfs4_accessargs args = {
3920                 .fh = NFS_FH(inode),
3921                 .bitmask = server->cache_consistency_bitmask,
3922         };
3923         struct nfs4_accessres res = {
3924                 .server = server,
3925         };
3926         struct rpc_message msg = {
3927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3928                 .rpc_argp = &args,
3929                 .rpc_resp = &res,
3930                 .rpc_cred = entry->cred,
3931         };
3932         int mode = entry->mask;
3933         int status = 0;
3934
3935         /*
3936          * Determine which access bits we want to ask for...
3937          */
3938         if (mode & MAY_READ)
3939                 args.access |= NFS4_ACCESS_READ;
3940         if (S_ISDIR(inode->i_mode)) {
3941                 if (mode & MAY_WRITE)
3942                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3943                 if (mode & MAY_EXEC)
3944                         args.access |= NFS4_ACCESS_LOOKUP;
3945         } else {
3946                 if (mode & MAY_WRITE)
3947                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3948                 if (mode & MAY_EXEC)
3949                         args.access |= NFS4_ACCESS_EXECUTE;
3950         }
3951
3952         res.fattr = nfs_alloc_fattr();
3953         if (res.fattr == NULL)
3954                 return -ENOMEM;
3955
3956         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3957         if (!status) {
3958                 nfs_access_set_mask(entry, res.access);
3959                 nfs_refresh_inode(inode, res.fattr);
3960         }
3961         nfs_free_fattr(res.fattr);
3962         return status;
3963 }
3964
3965 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3966 {
3967         struct nfs4_exception exception = { };
3968         int err;
3969         do {
3970                 err = _nfs4_proc_access(inode, entry);
3971                 trace_nfs4_access(inode, err);
3972                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3973                                 &exception);
3974         } while (exception.retry);
3975         return err;
3976 }
3977
3978 /*
3979  * TODO: For the time being, we don't try to get any attributes
3980  * along with any of the zero-copy operations READ, READDIR,
3981  * READLINK, WRITE.
3982  *
3983  * In the case of the first three, we want to put the GETATTR
3984  * after the read-type operation -- this is because it is hard
3985  * to predict the length of a GETATTR response in v4, and thus
3986  * align the READ data correctly.  This means that the GETATTR
3987  * may end up partially falling into the page cache, and we should
3988  * shift it into the 'tail' of the xdr_buf before processing.
3989  * To do this efficiently, we need to know the total length
3990  * of data received, which doesn't seem to be available outside
3991  * of the RPC layer.
3992  *
3993  * In the case of WRITE, we also want to put the GETATTR after
3994  * the operation -- in this case because we want to make sure
3995  * we get the post-operation mtime and size.
3996  *
3997  * Both of these changes to the XDR layer would in fact be quite
3998  * minor, but I decided to leave them for a subsequent patch.
3999  */
4000 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4001                 unsigned int pgbase, unsigned int pglen)
4002 {
4003         struct nfs4_readlink args = {
4004                 .fh       = NFS_FH(inode),
4005                 .pgbase   = pgbase,
4006                 .pglen    = pglen,
4007                 .pages    = &page,
4008         };
4009         struct nfs4_readlink_res res;
4010         struct rpc_message msg = {
4011                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4012                 .rpc_argp = &args,
4013                 .rpc_resp = &res,
4014         };
4015
4016         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4017 }
4018
4019 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4020                 unsigned int pgbase, unsigned int pglen)
4021 {
4022         struct nfs4_exception exception = { };
4023         int err;
4024         do {
4025                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4026                 trace_nfs4_readlink(inode, err);
4027                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4028                                 &exception);
4029         } while (exception.retry);
4030         return err;
4031 }
4032
4033 /*
4034  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4035  */
4036 static int
4037 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4038                  int flags)
4039 {
4040         struct nfs_server *server = NFS_SERVER(dir);
4041         struct nfs4_label l, *ilabel = NULL;
4042         struct nfs_open_context *ctx;
4043         struct nfs4_state *state;
4044         int status = 0;
4045
4046         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4047         if (IS_ERR(ctx))
4048                 return PTR_ERR(ctx);
4049
4050         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4051
4052         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4053                 sattr->ia_mode &= ~current_umask();
4054         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4055         if (IS_ERR(state)) {
4056                 status = PTR_ERR(state);
4057                 goto out;
4058         }
4059 out:
4060         nfs4_label_release_security(ilabel);
4061         put_nfs_open_context(ctx);
4062         return status;
4063 }
4064
4065 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4066 {
4067         struct nfs_server *server = NFS_SERVER(dir);
4068         struct nfs_removeargs args = {
4069                 .fh = NFS_FH(dir),
4070                 .name = *name,
4071         };
4072         struct nfs_removeres res = {
4073                 .server = server,
4074         };
4075         struct rpc_message msg = {
4076                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4077                 .rpc_argp = &args,
4078                 .rpc_resp = &res,
4079         };
4080         unsigned long timestamp = jiffies;
4081         int status;
4082
4083         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4084         if (status == 0)
4085                 update_changeattr(dir, &res.cinfo, timestamp);
4086         return status;
4087 }
4088
4089 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4090 {
4091         struct nfs4_exception exception = { };
4092         int err;
4093         do {
4094                 err = _nfs4_proc_remove(dir, name);
4095                 trace_nfs4_remove(dir, name, err);
4096                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4097                                 &exception);
4098         } while (exception.retry);
4099         return err;
4100 }
4101
4102 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4103 {
4104         struct nfs_server *server = NFS_SERVER(dir);
4105         struct nfs_removeargs *args = msg->rpc_argp;
4106         struct nfs_removeres *res = msg->rpc_resp;
4107
4108         res->server = server;
4109         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4110         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4111
4112         nfs_fattr_init(res->dir_attr);
4113 }
4114
4115 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4116 {
4117         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
4118                         &data->args.seq_args,
4119                         &data->res.seq_res,
4120                         task);
4121 }
4122
4123 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4124 {
4125         struct nfs_unlinkdata *data = task->tk_calldata;
4126         struct nfs_removeres *res = &data->res;
4127
4128         if (!nfs4_sequence_done(task, &res->seq_res))
4129                 return 0;
4130         if (nfs4_async_handle_error(task, res->server, NULL,
4131                                     &data->timeout) == -EAGAIN)
4132                 return 0;
4133         if (task->tk_status == 0)
4134                 update_changeattr(dir, &res->cinfo, res->dir_attr->time_start);
4135         return 1;
4136 }
4137
4138 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4139 {
4140         struct nfs_server *server = NFS_SERVER(dir);
4141         struct nfs_renameargs *arg = msg->rpc_argp;
4142         struct nfs_renameres *res = msg->rpc_resp;
4143
4144         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4145         res->server = server;
4146         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4147 }
4148
4149 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4150 {
4151         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
4152                         &data->args.seq_args,
4153                         &data->res.seq_res,
4154                         task);
4155 }
4156
4157 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4158                                  struct inode *new_dir)
4159 {
4160         struct nfs_renamedata *data = task->tk_calldata;
4161         struct nfs_renameres *res = &data->res;
4162
4163         if (!nfs4_sequence_done(task, &res->seq_res))
4164                 return 0;
4165         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4166                 return 0;
4167
4168         if (task->tk_status == 0) {
4169                 update_changeattr(old_dir, &res->old_cinfo, res->old_fattr->time_start);
4170                 if (new_dir != old_dir)
4171                         update_changeattr(new_dir, &res->new_cinfo, res->new_fattr->time_start);
4172         }
4173         return 1;
4174 }
4175
4176 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4177 {
4178         struct nfs_server *server = NFS_SERVER(inode);
4179         struct nfs4_link_arg arg = {
4180                 .fh     = NFS_FH(inode),
4181                 .dir_fh = NFS_FH(dir),
4182                 .name   = name,
4183                 .bitmask = server->attr_bitmask,
4184         };
4185         struct nfs4_link_res res = {
4186                 .server = server,
4187                 .label = NULL,
4188         };
4189         struct rpc_message msg = {
4190                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4191                 .rpc_argp = &arg,
4192                 .rpc_resp = &res,
4193         };
4194         int status = -ENOMEM;
4195
4196         res.fattr = nfs_alloc_fattr();
4197         if (res.fattr == NULL)
4198                 goto out;
4199
4200         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4201         if (IS_ERR(res.label)) {
4202                 status = PTR_ERR(res.label);
4203                 goto out;
4204         }
4205         arg.bitmask = nfs4_bitmask(server, res.label);
4206
4207         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4208         if (!status) {
4209                 update_changeattr(dir, &res.cinfo, res.fattr->time_start);
4210                 status = nfs_post_op_update_inode(inode, res.fattr);
4211                 if (!status)
4212                         nfs_setsecurity(inode, res.fattr, res.label);
4213         }
4214
4215
4216         nfs4_label_free(res.label);
4217
4218 out:
4219         nfs_free_fattr(res.fattr);
4220         return status;
4221 }
4222
4223 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4224 {
4225         struct nfs4_exception exception = { };
4226         int err;
4227         do {
4228                 err = nfs4_handle_exception(NFS_SERVER(inode),
4229                                 _nfs4_proc_link(inode, dir, name),
4230                                 &exception);
4231         } while (exception.retry);
4232         return err;
4233 }
4234
4235 struct nfs4_createdata {
4236         struct rpc_message msg;
4237         struct nfs4_create_arg arg;
4238         struct nfs4_create_res res;
4239         struct nfs_fh fh;
4240         struct nfs_fattr fattr;
4241         struct nfs4_label *label;
4242 };
4243
4244 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4245                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4246 {
4247         struct nfs4_createdata *data;
4248
4249         data = kzalloc(sizeof(*data), GFP_KERNEL);
4250         if (data != NULL) {
4251                 struct nfs_server *server = NFS_SERVER(dir);
4252
4253                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4254                 if (IS_ERR(data->label))
4255                         goto out_free;
4256
4257                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4258                 data->msg.rpc_argp = &data->arg;
4259                 data->msg.rpc_resp = &data->res;
4260                 data->arg.dir_fh = NFS_FH(dir);
4261                 data->arg.server = server;
4262                 data->arg.name = name;
4263                 data->arg.attrs = sattr;
4264                 data->arg.ftype = ftype;
4265                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4266                 data->arg.umask = current_umask();
4267                 data->res.server = server;
4268                 data->res.fh = &data->fh;
4269                 data->res.fattr = &data->fattr;
4270                 data->res.label = data->label;
4271                 nfs_fattr_init(data->res.fattr);
4272         }
4273         return data;
4274 out_free:
4275         kfree(data);
4276         return NULL;
4277 }
4278
4279 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4280 {
4281         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4282                                     &data->arg.seq_args, &data->res.seq_res, 1);
4283         if (status == 0) {
4284                 update_changeattr(dir, &data->res.dir_cinfo,
4285                                 data->res.fattr->time_start);
4286                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4287         }
4288         return status;
4289 }
4290
4291 static void nfs4_free_createdata(struct nfs4_createdata *data)
4292 {
4293         nfs4_label_free(data->label);
4294         kfree(data);
4295 }
4296
4297 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4298                 struct page *page, unsigned int len, struct iattr *sattr,
4299                 struct nfs4_label *label)
4300 {
4301         struct nfs4_createdata *data;
4302         int status = -ENAMETOOLONG;
4303
4304         if (len > NFS4_MAXPATHLEN)
4305                 goto out;
4306
4307         status = -ENOMEM;
4308         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4309         if (data == NULL)
4310                 goto out;
4311
4312         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4313         data->arg.u.symlink.pages = &page;
4314         data->arg.u.symlink.len = len;
4315         data->arg.label = label;
4316         
4317         status = nfs4_do_create(dir, dentry, data);
4318
4319         nfs4_free_createdata(data);
4320 out:
4321         return status;
4322 }
4323
4324 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4325                 struct page *page, unsigned int len, struct iattr *sattr)
4326 {
4327         struct nfs4_exception exception = { };
4328         struct nfs4_label l, *label = NULL;
4329         int err;
4330
4331         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4332
4333         do {
4334                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4335                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4336                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4337                                 &exception);
4338         } while (exception.retry);
4339
4340         nfs4_label_release_security(label);
4341         return err;
4342 }
4343
4344 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4345                 struct iattr *sattr, struct nfs4_label *label)
4346 {
4347         struct nfs4_createdata *data;
4348         int status = -ENOMEM;
4349
4350         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4351         if (data == NULL)
4352                 goto out;
4353
4354         data->arg.label = label;
4355         status = nfs4_do_create(dir, dentry, data);
4356
4357         nfs4_free_createdata(data);
4358 out:
4359         return status;
4360 }
4361
4362 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4363                 struct iattr *sattr)
4364 {
4365         struct nfs_server *server = NFS_SERVER(dir);
4366         struct nfs4_exception exception = { };
4367         struct nfs4_label l, *label = NULL;
4368         int err;
4369
4370         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4371
4372         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4373                 sattr->ia_mode &= ~current_umask();
4374         do {
4375                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4376                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4377                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4378                                 &exception);
4379         } while (exception.retry);
4380         nfs4_label_release_security(label);
4381
4382         return err;
4383 }
4384
4385 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4386                 u64 cookie, struct page **pages, unsigned int count, int plus)
4387 {
4388         struct inode            *dir = d_inode(dentry);
4389         struct nfs4_readdir_arg args = {
4390                 .fh = NFS_FH(dir),
4391                 .pages = pages,
4392                 .pgbase = 0,
4393                 .count = count,
4394                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4395                 .plus = plus,
4396         };
4397         struct nfs4_readdir_res res;
4398         struct rpc_message msg = {
4399                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4400                 .rpc_argp = &args,
4401                 .rpc_resp = &res,
4402                 .rpc_cred = cred,
4403         };
4404         int                     status;
4405
4406         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4407                         dentry,
4408                         (unsigned long long)cookie);
4409         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4410         res.pgbase = args.pgbase;
4411         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4412         if (status >= 0) {
4413                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4414                 status += args.pgbase;
4415         }
4416
4417         nfs_invalidate_atime(dir);
4418
4419         dprintk("%s: returns %d\n", __func__, status);
4420         return status;
4421 }
4422
4423 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4424                 u64 cookie, struct page **pages, unsigned int count, int plus)
4425 {
4426         struct nfs4_exception exception = { };
4427         int err;
4428         do {
4429                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4430                                 pages, count, plus);
4431                 trace_nfs4_readdir(d_inode(dentry), err);
4432                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4433                                 &exception);
4434         } while (exception.retry);
4435         return err;
4436 }
4437
4438 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4439                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4440 {
4441         struct nfs4_createdata *data;
4442         int mode = sattr->ia_mode;
4443         int status = -ENOMEM;
4444
4445         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4446         if (data == NULL)
4447                 goto out;
4448
4449         if (S_ISFIFO(mode))
4450                 data->arg.ftype = NF4FIFO;
4451         else if (S_ISBLK(mode)) {
4452                 data->arg.ftype = NF4BLK;
4453                 data->arg.u.device.specdata1 = MAJOR(rdev);
4454                 data->arg.u.device.specdata2 = MINOR(rdev);
4455         }
4456         else if (S_ISCHR(mode)) {
4457                 data->arg.ftype = NF4CHR;
4458                 data->arg.u.device.specdata1 = MAJOR(rdev);
4459                 data->arg.u.device.specdata2 = MINOR(rdev);
4460         } else if (!S_ISSOCK(mode)) {
4461                 status = -EINVAL;
4462                 goto out_free;
4463         }
4464
4465         data->arg.label = label;
4466         status = nfs4_do_create(dir, dentry, data);
4467 out_free:
4468         nfs4_free_createdata(data);
4469 out:
4470         return status;
4471 }
4472
4473 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4474                 struct iattr *sattr, dev_t rdev)
4475 {
4476         struct nfs_server *server = NFS_SERVER(dir);
4477         struct nfs4_exception exception = { };
4478         struct nfs4_label l, *label = NULL;
4479         int err;
4480
4481         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4482
4483         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4484                 sattr->ia_mode &= ~current_umask();
4485         do {
4486                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4487                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4488                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4489                                 &exception);
4490         } while (exception.retry);
4491
4492         nfs4_label_release_security(label);
4493
4494         return err;
4495 }
4496
4497 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4498                  struct nfs_fsstat *fsstat)
4499 {
4500         struct nfs4_statfs_arg args = {
4501                 .fh = fhandle,
4502                 .bitmask = server->attr_bitmask,
4503         };
4504         struct nfs4_statfs_res res = {
4505                 .fsstat = fsstat,
4506         };
4507         struct rpc_message msg = {
4508                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4509                 .rpc_argp = &args,
4510                 .rpc_resp = &res,
4511         };
4512
4513         nfs_fattr_init(fsstat->fattr);
4514         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4515 }
4516
4517 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4518 {
4519         struct nfs4_exception exception = { };
4520         int err;
4521         do {
4522                 err = nfs4_handle_exception(server,
4523                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4524                                 &exception);
4525         } while (exception.retry);
4526         return err;
4527 }
4528
4529 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4530                 struct nfs_fsinfo *fsinfo)
4531 {
4532         struct nfs4_fsinfo_arg args = {
4533                 .fh = fhandle,
4534                 .bitmask = server->attr_bitmask,
4535         };
4536         struct nfs4_fsinfo_res res = {
4537                 .fsinfo = fsinfo,
4538         };
4539         struct rpc_message msg = {
4540                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4541                 .rpc_argp = &args,
4542                 .rpc_resp = &res,
4543         };
4544
4545         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4546 }
4547
4548 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4549 {
4550         struct nfs4_exception exception = { };
4551         unsigned long now = jiffies;
4552         int err;
4553
4554         do {
4555                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4556                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4557                 if (err == 0) {
4558                         nfs4_set_lease_period(server->nfs_client,
4559                                         fsinfo->lease_time * HZ,
4560                                         now);
4561                         break;
4562                 }
4563                 err = nfs4_handle_exception(server, err, &exception);
4564         } while (exception.retry);
4565         return err;
4566 }
4567
4568 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4569 {
4570         int error;
4571
4572         nfs_fattr_init(fsinfo->fattr);
4573         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4574         if (error == 0) {
4575                 /* block layout checks this! */
4576                 server->pnfs_blksize = fsinfo->blksize;
4577                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4578         }
4579
4580         return error;
4581 }
4582
4583 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4584                 struct nfs_pathconf *pathconf)
4585 {
4586         struct nfs4_pathconf_arg args = {
4587                 .fh = fhandle,
4588                 .bitmask = server->attr_bitmask,
4589         };
4590         struct nfs4_pathconf_res res = {
4591                 .pathconf = pathconf,
4592         };
4593         struct rpc_message msg = {
4594                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4595                 .rpc_argp = &args,
4596                 .rpc_resp = &res,
4597         };
4598
4599         /* None of the pathconf attributes are mandatory to implement */
4600         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4601                 memset(pathconf, 0, sizeof(*pathconf));
4602                 return 0;
4603         }
4604
4605         nfs_fattr_init(pathconf->fattr);
4606         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4607 }
4608
4609 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4610                 struct nfs_pathconf *pathconf)
4611 {
4612         struct nfs4_exception exception = { };
4613         int err;
4614
4615         do {
4616                 err = nfs4_handle_exception(server,
4617                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4618                                 &exception);
4619         } while (exception.retry);
4620         return err;
4621 }
4622
4623 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4624                 const struct nfs_open_context *ctx,
4625                 const struct nfs_lock_context *l_ctx,
4626                 fmode_t fmode)
4627 {
4628         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4629 }
4630 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4631
4632 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4633                 const struct nfs_open_context *ctx,
4634                 const struct nfs_lock_context *l_ctx,
4635                 fmode_t fmode)
4636 {
4637         nfs4_stateid current_stateid;
4638
4639         /* If the current stateid represents a lost lock, then exit */
4640         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4641                 return true;
4642         return nfs4_stateid_match(stateid, &current_stateid);
4643 }
4644
4645 static bool nfs4_error_stateid_expired(int err)
4646 {
4647         switch (err) {
4648         case -NFS4ERR_DELEG_REVOKED:
4649         case -NFS4ERR_ADMIN_REVOKED:
4650         case -NFS4ERR_BAD_STATEID:
4651         case -NFS4ERR_STALE_STATEID:
4652         case -NFS4ERR_OLD_STATEID:
4653         case -NFS4ERR_OPENMODE:
4654         case -NFS4ERR_EXPIRED:
4655                 return true;
4656         }
4657         return false;
4658 }
4659
4660 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4661 {
4662         struct nfs_server *server = NFS_SERVER(hdr->inode);
4663
4664         trace_nfs4_read(hdr, task->tk_status);
4665         if (task->tk_status < 0) {
4666                 struct nfs4_exception exception = {
4667                         .inode = hdr->inode,
4668                         .state = hdr->args.context->state,
4669                         .stateid = &hdr->args.stateid,
4670                 };
4671                 task->tk_status = nfs4_async_handle_exception(task,
4672                                 server, task->tk_status, &exception);
4673                 if (exception.retry) {
4674                         rpc_restart_call_prepare(task);
4675                         return -EAGAIN;
4676                 }
4677         }
4678
4679         if (task->tk_status > 0)
4680                 renew_lease(server, hdr->timestamp);
4681         return 0;
4682 }
4683
4684 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4685                 struct nfs_pgio_args *args)
4686 {
4687
4688         if (!nfs4_error_stateid_expired(task->tk_status) ||
4689                 nfs4_stateid_is_current(&args->stateid,
4690                                 args->context,
4691                                 args->lock_context,
4692                                 FMODE_READ))
4693                 return false;
4694         rpc_restart_call_prepare(task);
4695         return true;
4696 }
4697
4698 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4699 {
4700
4701         dprintk("--> %s\n", __func__);
4702
4703         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4704                 return -EAGAIN;
4705         if (nfs4_read_stateid_changed(task, &hdr->args))
4706                 return -EAGAIN;
4707         if (task->tk_status > 0)
4708                 nfs_invalidate_atime(hdr->inode);
4709         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4710                                     nfs4_read_done_cb(task, hdr);
4711 }
4712
4713 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4714                                  struct rpc_message *msg)
4715 {
4716         hdr->timestamp   = jiffies;
4717         if (!hdr->pgio_done_cb)
4718                 hdr->pgio_done_cb = nfs4_read_done_cb;
4719         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4720         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4721 }
4722
4723 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4724                                       struct nfs_pgio_header *hdr)
4725 {
4726         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4727                         &hdr->args.seq_args,
4728                         &hdr->res.seq_res,
4729                         task))
4730                 return 0;
4731         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4732                                 hdr->args.lock_context,
4733                                 hdr->rw_ops->rw_mode) == -EIO)
4734                 return -EIO;
4735         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4736                 return -EIO;
4737         return 0;
4738 }
4739
4740 static int nfs4_write_done_cb(struct rpc_task *task,
4741                               struct nfs_pgio_header *hdr)
4742 {
4743         struct inode *inode = hdr->inode;
4744
4745         trace_nfs4_write(hdr, task->tk_status);
4746         if (task->tk_status < 0) {
4747                 struct nfs4_exception exception = {
4748                         .inode = hdr->inode,
4749                         .state = hdr->args.context->state,
4750                         .stateid = &hdr->args.stateid,
4751                 };
4752                 task->tk_status = nfs4_async_handle_exception(task,
4753                                 NFS_SERVER(inode), task->tk_status,
4754                                 &exception);
4755                 if (exception.retry) {
4756                         rpc_restart_call_prepare(task);
4757                         return -EAGAIN;
4758                 }
4759         }
4760         if (task->tk_status >= 0) {
4761                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4762                 nfs_writeback_update_inode(hdr);
4763         }
4764         return 0;
4765 }
4766
4767 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4768                 struct nfs_pgio_args *args)
4769 {
4770
4771         if (!nfs4_error_stateid_expired(task->tk_status) ||
4772                 nfs4_stateid_is_current(&args->stateid,
4773                                 args->context,
4774                                 args->lock_context,
4775                                 FMODE_WRITE))
4776                 return false;
4777         rpc_restart_call_prepare(task);
4778         return true;
4779 }
4780
4781 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4782 {
4783         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4784                 return -EAGAIN;
4785         if (nfs4_write_stateid_changed(task, &hdr->args))
4786                 return -EAGAIN;
4787         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4788                 nfs4_write_done_cb(task, hdr);
4789 }
4790
4791 static
4792 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4793 {
4794         /* Don't request attributes for pNFS or O_DIRECT writes */
4795         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4796                 return false;
4797         /* Otherwise, request attributes if and only if we don't hold
4798          * a delegation
4799          */
4800         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4801 }
4802
4803 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4804                                   struct rpc_message *msg)
4805 {
4806         struct nfs_server *server = NFS_SERVER(hdr->inode);
4807
4808         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4809                 hdr->args.bitmask = NULL;
4810                 hdr->res.fattr = NULL;
4811         } else
4812                 hdr->args.bitmask = server->cache_consistency_bitmask;
4813
4814         if (!hdr->pgio_done_cb)
4815                 hdr->pgio_done_cb = nfs4_write_done_cb;
4816         hdr->res.server = server;
4817         hdr->timestamp   = jiffies;
4818
4819         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4820         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4821 }
4822
4823 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4824 {
4825         nfs4_setup_sequence(NFS_SERVER(data->inode),
4826                         &data->args.seq_args,
4827                         &data->res.seq_res,
4828                         task);
4829 }
4830
4831 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4832 {
4833         struct inode *inode = data->inode;
4834
4835         trace_nfs4_commit(data, task->tk_status);
4836         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4837                                     NULL, NULL) == -EAGAIN) {
4838                 rpc_restart_call_prepare(task);
4839                 return -EAGAIN;
4840         }
4841         return 0;
4842 }
4843
4844 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4845 {
4846         if (!nfs4_sequence_done(task, &data->res.seq_res))
4847                 return -EAGAIN;
4848         return data->commit_done_cb(task, data);
4849 }
4850
4851 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4852 {
4853         struct nfs_server *server = NFS_SERVER(data->inode);
4854
4855         if (data->commit_done_cb == NULL)
4856                 data->commit_done_cb = nfs4_commit_done_cb;
4857         data->res.server = server;
4858         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4859         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4860 }
4861
4862 struct nfs4_renewdata {
4863         struct nfs_client       *client;
4864         unsigned long           timestamp;
4865 };
4866
4867 /*
4868  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4869  * standalone procedure for queueing an asynchronous RENEW.
4870  */
4871 static void nfs4_renew_release(void *calldata)
4872 {
4873         struct nfs4_renewdata *data = calldata;
4874         struct nfs_client *clp = data->client;
4875
4876         if (atomic_read(&clp->cl_count) > 1)
4877                 nfs4_schedule_state_renewal(clp);
4878         nfs_put_client(clp);
4879         kfree(data);
4880 }
4881
4882 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4883 {
4884         struct nfs4_renewdata *data = calldata;
4885         struct nfs_client *clp = data->client;
4886         unsigned long timestamp = data->timestamp;
4887
4888         trace_nfs4_renew_async(clp, task->tk_status);
4889         switch (task->tk_status) {
4890         case 0:
4891                 break;
4892         case -NFS4ERR_LEASE_MOVED:
4893                 nfs4_schedule_lease_moved_recovery(clp);
4894                 break;
4895         default:
4896                 /* Unless we're shutting down, schedule state recovery! */
4897                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4898                         return;
4899                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4900                         nfs4_schedule_lease_recovery(clp);
4901                         return;
4902                 }
4903                 nfs4_schedule_path_down_recovery(clp);
4904         }
4905         do_renew_lease(clp, timestamp);
4906 }
4907
4908 static const struct rpc_call_ops nfs4_renew_ops = {
4909         .rpc_call_done = nfs4_renew_done,
4910         .rpc_release = nfs4_renew_release,
4911 };
4912
4913 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4914 {
4915         struct rpc_message msg = {
4916                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4917                 .rpc_argp       = clp,
4918                 .rpc_cred       = cred,
4919         };
4920         struct nfs4_renewdata *data;
4921
4922         if (renew_flags == 0)
4923                 return 0;
4924         if (!atomic_inc_not_zero(&clp->cl_count))
4925                 return -EIO;
4926         data = kmalloc(sizeof(*data), GFP_NOFS);
4927         if (data == NULL)
4928                 return -ENOMEM;
4929         data->client = clp;
4930         data->timestamp = jiffies;
4931         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4932                         &nfs4_renew_ops, data);
4933 }
4934
4935 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4936 {
4937         struct rpc_message msg = {
4938                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4939                 .rpc_argp       = clp,
4940                 .rpc_cred       = cred,
4941         };
4942         unsigned long now = jiffies;
4943         int status;
4944
4945         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4946         if (status < 0)
4947                 return status;
4948         do_renew_lease(clp, now);
4949         return 0;
4950 }
4951
4952 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4953 {
4954         return server->caps & NFS_CAP_ACLS;
4955 }
4956
4957 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4958  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4959  * the stack.
4960  */
4961 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4962
4963 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4964                 struct page **pages)
4965 {
4966         struct page *newpage, **spages;
4967         int rc = 0;
4968         size_t len;
4969         spages = pages;
4970
4971         do {
4972                 len = min_t(size_t, PAGE_SIZE, buflen);
4973                 newpage = alloc_page(GFP_KERNEL);
4974
4975                 if (newpage == NULL)
4976                         goto unwind;
4977                 memcpy(page_address(newpage), buf, len);
4978                 buf += len;
4979                 buflen -= len;
4980                 *pages++ = newpage;
4981                 rc++;
4982         } while (buflen != 0);
4983
4984         return rc;
4985
4986 unwind:
4987         for(; rc > 0; rc--)
4988                 __free_page(spages[rc-1]);
4989         return -ENOMEM;
4990 }
4991
4992 struct nfs4_cached_acl {
4993         int cached;
4994         size_t len;
4995         char data[0];
4996 };
4997
4998 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4999 {
5000         struct nfs_inode *nfsi = NFS_I(inode);
5001
5002         spin_lock(&inode->i_lock);
5003         kfree(nfsi->nfs4_acl);
5004         nfsi->nfs4_acl = acl;
5005         spin_unlock(&inode->i_lock);
5006 }
5007
5008 static void nfs4_zap_acl_attr(struct inode *inode)
5009 {
5010         nfs4_set_cached_acl(inode, NULL);
5011 }
5012
5013 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5014 {
5015         struct nfs_inode *nfsi = NFS_I(inode);
5016         struct nfs4_cached_acl *acl;
5017         int ret = -ENOENT;
5018
5019         spin_lock(&inode->i_lock);
5020         acl = nfsi->nfs4_acl;
5021         if (acl == NULL)
5022                 goto out;
5023         if (buf == NULL) /* user is just asking for length */
5024                 goto out_len;
5025         if (acl->cached == 0)
5026                 goto out;
5027         ret = -ERANGE; /* see getxattr(2) man page */
5028         if (acl->len > buflen)
5029                 goto out;
5030         memcpy(buf, acl->data, acl->len);
5031 out_len:
5032         ret = acl->len;
5033 out:
5034         spin_unlock(&inode->i_lock);
5035         return ret;
5036 }
5037
5038 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5039 {
5040         struct nfs4_cached_acl *acl;
5041         size_t buflen = sizeof(*acl) + acl_len;
5042
5043         if (buflen <= PAGE_SIZE) {
5044                 acl = kmalloc(buflen, GFP_KERNEL);
5045                 if (acl == NULL)
5046                         goto out;
5047                 acl->cached = 1;
5048                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5049         } else {
5050                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5051                 if (acl == NULL)
5052                         goto out;
5053                 acl->cached = 0;
5054         }
5055         acl->len = acl_len;
5056 out:
5057         nfs4_set_cached_acl(inode, acl);
5058 }
5059
5060 /*
5061  * The getxattr API returns the required buffer length when called with a
5062  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5063  * the required buf.  On a NULL buf, we send a page of data to the server
5064  * guessing that the ACL request can be serviced by a page. If so, we cache
5065  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5066  * the cache. If not so, we throw away the page, and cache the required
5067  * length. The next getxattr call will then produce another round trip to
5068  * the server, this time with the input buf of the required size.
5069  */
5070 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5071 {
5072         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
5073         struct nfs_getaclargs args = {
5074                 .fh = NFS_FH(inode),
5075                 .acl_pages = pages,
5076                 .acl_len = buflen,
5077         };
5078         struct nfs_getaclres res = {
5079                 .acl_len = buflen,
5080         };
5081         struct rpc_message msg = {
5082                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5083                 .rpc_argp = &args,
5084                 .rpc_resp = &res,
5085         };
5086         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5087         int ret = -ENOMEM, i;
5088
5089         /* As long as we're doing a round trip to the server anyway,
5090          * let's be prepared for a page of acl data. */
5091         if (npages == 0)
5092                 npages = 1;
5093         if (npages > ARRAY_SIZE(pages))
5094                 return -ERANGE;
5095
5096         for (i = 0; i < npages; i++) {
5097                 pages[i] = alloc_page(GFP_KERNEL);
5098                 if (!pages[i])
5099                         goto out_free;
5100         }
5101
5102         /* for decoding across pages */
5103         res.acl_scratch = alloc_page(GFP_KERNEL);
5104         if (!res.acl_scratch)
5105                 goto out_free;
5106
5107         args.acl_len = npages * PAGE_SIZE;
5108
5109         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5110                 __func__, buf, buflen, npages, args.acl_len);
5111         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5112                              &msg, &args.seq_args, &res.seq_res, 0);
5113         if (ret)
5114                 goto out_free;
5115
5116         /* Handle the case where the passed-in buffer is too short */
5117         if (res.acl_flags & NFS4_ACL_TRUNC) {
5118                 /* Did the user only issue a request for the acl length? */
5119                 if (buf == NULL)
5120                         goto out_ok;
5121                 ret = -ERANGE;
5122                 goto out_free;
5123         }
5124         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5125         if (buf) {
5126                 if (res.acl_len > buflen) {
5127                         ret = -ERANGE;
5128                         goto out_free;
5129                 }
5130                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5131         }
5132 out_ok:
5133         ret = res.acl_len;
5134 out_free:
5135         for (i = 0; i < npages; i++)
5136                 if (pages[i])
5137                         __free_page(pages[i]);
5138         if (res.acl_scratch)
5139                 __free_page(res.acl_scratch);
5140         return ret;
5141 }
5142
5143 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5144 {
5145         struct nfs4_exception exception = { };
5146         ssize_t ret;
5147         do {
5148                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5149                 trace_nfs4_get_acl(inode, ret);
5150                 if (ret >= 0)
5151                         break;
5152                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5153         } while (exception.retry);
5154         return ret;
5155 }
5156
5157 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5158 {
5159         struct nfs_server *server = NFS_SERVER(inode);
5160         int ret;
5161
5162         if (!nfs4_server_supports_acls(server))
5163                 return -EOPNOTSUPP;
5164         ret = nfs_revalidate_inode(server, inode);
5165         if (ret < 0)
5166                 return ret;
5167         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5168                 nfs_zap_acl_cache(inode);
5169         ret = nfs4_read_cached_acl(inode, buf, buflen);
5170         if (ret != -ENOENT)
5171                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5172                  * but no cached acl data, just the acl length */
5173                 return ret;
5174         return nfs4_get_acl_uncached(inode, buf, buflen);
5175 }
5176
5177 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5178 {
5179         struct nfs_server *server = NFS_SERVER(inode);
5180         struct page *pages[NFS4ACL_MAXPAGES];
5181         struct nfs_setaclargs arg = {
5182                 .fh             = NFS_FH(inode),
5183                 .acl_pages      = pages,
5184                 .acl_len        = buflen,
5185         };
5186         struct nfs_setaclres res;
5187         struct rpc_message msg = {
5188                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5189                 .rpc_argp       = &arg,
5190                 .rpc_resp       = &res,
5191         };
5192         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5193         int ret, i;
5194
5195         if (!nfs4_server_supports_acls(server))
5196                 return -EOPNOTSUPP;
5197         if (npages > ARRAY_SIZE(pages))
5198                 return -ERANGE;
5199         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5200         if (i < 0)
5201                 return i;
5202         nfs4_inode_return_delegation(inode);
5203         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5204
5205         /*
5206          * Free each page after tx, so the only ref left is
5207          * held by the network stack
5208          */
5209         for (; i > 0; i--)
5210                 put_page(pages[i-1]);
5211
5212         /*
5213          * Acl update can result in inode attribute update.
5214          * so mark the attribute cache invalid.
5215          */
5216         spin_lock(&inode->i_lock);
5217         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5218         spin_unlock(&inode->i_lock);
5219         nfs_access_zap_cache(inode);
5220         nfs_zap_acl_cache(inode);
5221         return ret;
5222 }
5223
5224 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5225 {
5226         struct nfs4_exception exception = { };
5227         int err;
5228         do {
5229                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5230                 trace_nfs4_set_acl(inode, err);
5231                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5232                                 &exception);
5233         } while (exception.retry);
5234         return err;
5235 }
5236
5237 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5238 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5239                                         size_t buflen)
5240 {
5241         struct nfs_server *server = NFS_SERVER(inode);
5242         struct nfs_fattr fattr;
5243         struct nfs4_label label = {0, 0, buflen, buf};
5244
5245         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5246         struct nfs4_getattr_arg arg = {
5247                 .fh             = NFS_FH(inode),
5248                 .bitmask        = bitmask,
5249         };
5250         struct nfs4_getattr_res res = {
5251                 .fattr          = &fattr,
5252                 .label          = &label,
5253                 .server         = server,
5254         };
5255         struct rpc_message msg = {
5256                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5257                 .rpc_argp       = &arg,
5258                 .rpc_resp       = &res,
5259         };
5260         int ret;
5261
5262         nfs_fattr_init(&fattr);
5263
5264         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5265         if (ret)
5266                 return ret;
5267         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5268                 return -ENOENT;
5269         if (buflen < label.len)
5270                 return -ERANGE;
5271         return 0;
5272 }
5273
5274 static int nfs4_get_security_label(struct inode *inode, void *buf,
5275                                         size_t buflen)
5276 {
5277         struct nfs4_exception exception = { };
5278         int err;
5279
5280         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5281                 return -EOPNOTSUPP;
5282
5283         do {
5284                 err = _nfs4_get_security_label(inode, buf, buflen);
5285                 trace_nfs4_get_security_label(inode, err);
5286                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5287                                 &exception);
5288         } while (exception.retry);
5289         return err;
5290 }
5291
5292 static int _nfs4_do_set_security_label(struct inode *inode,
5293                 struct nfs4_label *ilabel,
5294                 struct nfs_fattr *fattr,
5295                 struct nfs4_label *olabel)
5296 {
5297
5298         struct iattr sattr = {0};
5299         struct nfs_server *server = NFS_SERVER(inode);
5300         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5301         struct nfs_setattrargs arg = {
5302                 .fh             = NFS_FH(inode),
5303                 .iap            = &sattr,
5304                 .server         = server,
5305                 .bitmask        = bitmask,
5306                 .label          = ilabel,
5307         };
5308         struct nfs_setattrres res = {
5309                 .fattr          = fattr,
5310                 .label          = olabel,
5311                 .server         = server,
5312         };
5313         struct rpc_message msg = {
5314                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5315                 .rpc_argp       = &arg,
5316                 .rpc_resp       = &res,
5317         };
5318         int status;
5319
5320         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5321
5322         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5323         if (status)
5324                 dprintk("%s failed: %d\n", __func__, status);
5325
5326         return status;
5327 }
5328
5329 static int nfs4_do_set_security_label(struct inode *inode,
5330                 struct nfs4_label *ilabel,
5331                 struct nfs_fattr *fattr,
5332                 struct nfs4_label *olabel)
5333 {
5334         struct nfs4_exception exception = { };
5335         int err;
5336
5337         do {
5338                 err = _nfs4_do_set_security_label(inode, ilabel,
5339                                 fattr, olabel);
5340                 trace_nfs4_set_security_label(inode, err);
5341                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5342                                 &exception);
5343         } while (exception.retry);
5344         return err;
5345 }
5346
5347 static int
5348 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5349 {
5350         struct nfs4_label ilabel, *olabel = NULL;
5351         struct nfs_fattr fattr;
5352         struct rpc_cred *cred;
5353         int status;
5354
5355         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5356                 return -EOPNOTSUPP;
5357
5358         nfs_fattr_init(&fattr);
5359
5360         ilabel.pi = 0;
5361         ilabel.lfs = 0;
5362         ilabel.label = (char *)buf;
5363         ilabel.len = buflen;
5364
5365         cred = rpc_lookup_cred();
5366         if (IS_ERR(cred))
5367                 return PTR_ERR(cred);
5368
5369         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5370         if (IS_ERR(olabel)) {
5371                 status = -PTR_ERR(olabel);
5372                 goto out;
5373         }
5374
5375         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5376         if (status == 0)
5377                 nfs_setsecurity(inode, &fattr, olabel);
5378
5379         nfs4_label_free(olabel);
5380 out:
5381         put_rpccred(cred);
5382         return status;
5383 }
5384 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5385
5386
5387 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5388                                     nfs4_verifier *bootverf)
5389 {
5390         __be32 verf[2];
5391
5392         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5393                 /* An impossible timestamp guarantees this value
5394                  * will never match a generated boot time. */
5395                 verf[0] = cpu_to_be32(U32_MAX);
5396                 verf[1] = cpu_to_be32(U32_MAX);
5397         } else {
5398                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5399                 u64 ns = ktime_to_ns(nn->boot_time);
5400
5401                 verf[0] = cpu_to_be32(ns >> 32);
5402                 verf[1] = cpu_to_be32(ns);
5403         }
5404         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5405 }
5406
5407 static int
5408 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5409 {
5410         size_t len;
5411         char *str;
5412
5413         if (clp->cl_owner_id != NULL)
5414                 return 0;
5415
5416         rcu_read_lock();
5417         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5418                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5419                 1 +
5420                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5421                 1;
5422         rcu_read_unlock();
5423
5424         if (len > NFS4_OPAQUE_LIMIT + 1)
5425                 return -EINVAL;
5426
5427         /*
5428          * Since this string is allocated at mount time, and held until the
5429          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5430          * about a memory-reclaim deadlock.
5431          */
5432         str = kmalloc(len, GFP_KERNEL);
5433         if (!str)
5434                 return -ENOMEM;
5435
5436         rcu_read_lock();
5437         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5438                         clp->cl_ipaddr,
5439                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5440                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5441         rcu_read_unlock();
5442
5443         clp->cl_owner_id = str;
5444         return 0;
5445 }
5446
5447 static int
5448 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5449 {
5450         size_t len;
5451         char *str;
5452
5453         len = 10 + 10 + 1 + 10 + 1 +
5454                 strlen(nfs4_client_id_uniquifier) + 1 +
5455                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5456
5457         if (len > NFS4_OPAQUE_LIMIT + 1)
5458                 return -EINVAL;
5459
5460         /*
5461          * Since this string is allocated at mount time, and held until the
5462          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5463          * about a memory-reclaim deadlock.
5464          */
5465         str = kmalloc(len, GFP_KERNEL);
5466         if (!str)
5467                 return -ENOMEM;
5468
5469         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5470                         clp->rpc_ops->version, clp->cl_minorversion,
5471                         nfs4_client_id_uniquifier,
5472                         clp->cl_rpcclient->cl_nodename);
5473         clp->cl_owner_id = str;
5474         return 0;
5475 }
5476
5477 static int
5478 nfs4_init_uniform_client_string(struct nfs_client *clp)
5479 {
5480         size_t len;
5481         char *str;
5482
5483         if (clp->cl_owner_id != NULL)
5484                 return 0;
5485
5486         if (nfs4_client_id_uniquifier[0] != '\0')
5487                 return nfs4_init_uniquifier_client_string(clp);
5488
5489         len = 10 + 10 + 1 + 10 + 1 +
5490                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5491
5492         if (len > NFS4_OPAQUE_LIMIT + 1)
5493                 return -EINVAL;
5494
5495         /*
5496          * Since this string is allocated at mount time, and held until the
5497          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5498          * about a memory-reclaim deadlock.
5499          */
5500         str = kmalloc(len, GFP_KERNEL);
5501         if (!str)
5502                 return -ENOMEM;
5503
5504         scnprintf(str, len, "Linux NFSv%u.%u %s",
5505                         clp->rpc_ops->version, clp->cl_minorversion,
5506                         clp->cl_rpcclient->cl_nodename);
5507         clp->cl_owner_id = str;
5508         return 0;
5509 }
5510
5511 /*
5512  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5513  * services.  Advertise one based on the address family of the
5514  * clientaddr.
5515  */
5516 static unsigned int
5517 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5518 {
5519         if (strchr(clp->cl_ipaddr, ':') != NULL)
5520                 return scnprintf(buf, len, "tcp6");
5521         else
5522                 return scnprintf(buf, len, "tcp");
5523 }
5524
5525 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5526 {
5527         struct nfs4_setclientid *sc = calldata;
5528
5529         if (task->tk_status == 0)
5530                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5531 }
5532
5533 static const struct rpc_call_ops nfs4_setclientid_ops = {
5534         .rpc_call_done = nfs4_setclientid_done,
5535 };
5536
5537 /**
5538  * nfs4_proc_setclientid - Negotiate client ID
5539  * @clp: state data structure
5540  * @program: RPC program for NFSv4 callback service
5541  * @port: IP port number for NFS4 callback service
5542  * @cred: RPC credential to use for this call
5543  * @res: where to place the result
5544  *
5545  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5546  */
5547 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5548                 unsigned short port, struct rpc_cred *cred,
5549                 struct nfs4_setclientid_res *res)
5550 {
5551         nfs4_verifier sc_verifier;
5552         struct nfs4_setclientid setclientid = {
5553                 .sc_verifier = &sc_verifier,
5554                 .sc_prog = program,
5555                 .sc_clnt = clp,
5556         };
5557         struct rpc_message msg = {
5558                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5559                 .rpc_argp = &setclientid,
5560                 .rpc_resp = res,
5561                 .rpc_cred = cred,
5562         };
5563         struct rpc_task *task;
5564         struct rpc_task_setup task_setup_data = {
5565                 .rpc_client = clp->cl_rpcclient,
5566                 .rpc_message = &msg,
5567                 .callback_ops = &nfs4_setclientid_ops,
5568                 .callback_data = &setclientid,
5569                 .flags = RPC_TASK_TIMEOUT,
5570         };
5571         int status;
5572
5573         /* nfs_client_id4 */
5574         nfs4_init_boot_verifier(clp, &sc_verifier);
5575
5576         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5577                 status = nfs4_init_uniform_client_string(clp);
5578         else
5579                 status = nfs4_init_nonuniform_client_string(clp);
5580
5581         if (status)
5582                 goto out;
5583
5584         /* cb_client4 */
5585         setclientid.sc_netid_len =
5586                                 nfs4_init_callback_netid(clp,
5587                                                 setclientid.sc_netid,
5588                                                 sizeof(setclientid.sc_netid));
5589         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5590                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5591                                 clp->cl_ipaddr, port >> 8, port & 255);
5592
5593         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5594                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5595                 clp->cl_owner_id);
5596         task = rpc_run_task(&task_setup_data);
5597         if (IS_ERR(task)) {
5598                 status = PTR_ERR(task);
5599                 goto out;
5600         }
5601         status = task->tk_status;
5602         if (setclientid.sc_cred) {
5603                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5604                 put_rpccred(setclientid.sc_cred);
5605         }
5606         rpc_put_task(task);
5607 out:
5608         trace_nfs4_setclientid(clp, status);
5609         dprintk("NFS reply setclientid: %d\n", status);
5610         return status;
5611 }
5612
5613 /**
5614  * nfs4_proc_setclientid_confirm - Confirm client ID
5615  * @clp: state data structure
5616  * @res: result of a previous SETCLIENTID
5617  * @cred: RPC credential to use for this call
5618  *
5619  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5620  */
5621 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5622                 struct nfs4_setclientid_res *arg,
5623                 struct rpc_cred *cred)
5624 {
5625         struct rpc_message msg = {
5626                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5627                 .rpc_argp = arg,
5628                 .rpc_cred = cred,
5629         };
5630         int status;
5631
5632         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5633                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5634                 clp->cl_clientid);
5635         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5636         trace_nfs4_setclientid_confirm(clp, status);
5637         dprintk("NFS reply setclientid_confirm: %d\n", status);
5638         return status;
5639 }
5640
5641 struct nfs4_delegreturndata {
5642         struct nfs4_delegreturnargs args;
5643         struct nfs4_delegreturnres res;
5644         struct nfs_fh fh;
5645         nfs4_stateid stateid;
5646         unsigned long timestamp;
5647         struct {
5648                 struct nfs4_layoutreturn_args arg;
5649                 struct nfs4_layoutreturn_res res;
5650                 struct nfs4_xdr_opaque_data ld_private;
5651                 u32 roc_barrier;
5652                 bool roc;
5653         } lr;
5654         struct nfs_fattr fattr;
5655         int rpc_status;
5656         struct inode *inode;
5657 };
5658
5659 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5660 {
5661         struct nfs4_delegreturndata *data = calldata;
5662
5663         if (!nfs4_sequence_done(task, &data->res.seq_res))
5664                 return;
5665
5666         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5667
5668         /* Handle Layoutreturn errors */
5669         if (data->args.lr_args && task->tk_status != 0) {
5670                 switch(data->res.lr_ret) {
5671                 default:
5672                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5673                         break;
5674                 case 0:
5675                         data->args.lr_args = NULL;
5676                         data->res.lr_res = NULL;
5677                         break;
5678                 case -NFS4ERR_ADMIN_REVOKED:
5679                 case -NFS4ERR_DELEG_REVOKED:
5680                 case -NFS4ERR_EXPIRED:
5681                 case -NFS4ERR_BAD_STATEID:
5682                 case -NFS4ERR_OLD_STATEID:
5683                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5684                 case -NFS4ERR_WRONG_CRED:
5685                         data->args.lr_args = NULL;
5686                         data->res.lr_res = NULL;
5687                         data->res.lr_ret = 0;
5688                         rpc_restart_call_prepare(task);
5689                         return;
5690                 }
5691         }
5692
5693         switch (task->tk_status) {
5694         case 0:
5695                 renew_lease(data->res.server, data->timestamp);
5696                 break;
5697         case -NFS4ERR_ADMIN_REVOKED:
5698         case -NFS4ERR_DELEG_REVOKED:
5699         case -NFS4ERR_EXPIRED:
5700                 nfs4_free_revoked_stateid(data->res.server,
5701                                 data->args.stateid,
5702                                 task->tk_msg.rpc_cred);
5703         case -NFS4ERR_BAD_STATEID:
5704         case -NFS4ERR_OLD_STATEID:
5705         case -NFS4ERR_STALE_STATEID:
5706                 task->tk_status = 0;
5707                 break;
5708         case -NFS4ERR_ACCESS:
5709                 if (data->args.bitmask) {
5710                         data->args.bitmask = NULL;
5711                         data->res.fattr = NULL;
5712                         task->tk_status = 0;
5713                         rpc_restart_call_prepare(task);
5714                         return;
5715                 }
5716         default:
5717                 if (nfs4_async_handle_error(task, data->res.server,
5718                                             NULL, NULL) == -EAGAIN) {
5719                         rpc_restart_call_prepare(task);
5720                         return;
5721                 }
5722         }
5723         data->rpc_status = task->tk_status;
5724 }
5725
5726 static void nfs4_delegreturn_release(void *calldata)
5727 {
5728         struct nfs4_delegreturndata *data = calldata;
5729         struct inode *inode = data->inode;
5730
5731         if (inode) {
5732                 if (data->lr.roc)
5733                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
5734                                         data->res.lr_ret);
5735                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5736                 nfs_iput_and_deactive(inode);
5737         }
5738         kfree(calldata);
5739 }
5740
5741 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5742 {
5743         struct nfs4_delegreturndata *d_data;
5744
5745         d_data = (struct nfs4_delegreturndata *)data;
5746
5747         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5748                 return;
5749
5750         nfs4_setup_sequence(d_data->res.server,
5751                         &d_data->args.seq_args,
5752                         &d_data->res.seq_res,
5753                         task);
5754 }
5755
5756 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5757         .rpc_call_prepare = nfs4_delegreturn_prepare,
5758         .rpc_call_done = nfs4_delegreturn_done,
5759         .rpc_release = nfs4_delegreturn_release,
5760 };
5761
5762 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5763 {
5764         struct nfs4_delegreturndata *data;
5765         struct nfs_server *server = NFS_SERVER(inode);
5766         struct rpc_task *task;
5767         struct rpc_message msg = {
5768                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5769                 .rpc_cred = cred,
5770         };
5771         struct rpc_task_setup task_setup_data = {
5772                 .rpc_client = server->client,
5773                 .rpc_message = &msg,
5774                 .callback_ops = &nfs4_delegreturn_ops,
5775                 .flags = RPC_TASK_ASYNC,
5776         };
5777         int status = 0;
5778
5779         data = kzalloc(sizeof(*data), GFP_NOFS);
5780         if (data == NULL)
5781                 return -ENOMEM;
5782         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5783
5784         nfs4_state_protect(server->nfs_client,
5785                         NFS_SP4_MACH_CRED_CLEANUP,
5786                         &task_setup_data.rpc_client, &msg);
5787
5788         data->args.fhandle = &data->fh;
5789         data->args.stateid = &data->stateid;
5790         data->args.bitmask = server->cache_consistency_bitmask;
5791         nfs_copy_fh(&data->fh, NFS_FH(inode));
5792         nfs4_stateid_copy(&data->stateid, stateid);
5793         data->res.fattr = &data->fattr;
5794         data->res.server = server;
5795         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5796         data->lr.arg.ld_private = &data->lr.ld_private;
5797         nfs_fattr_init(data->res.fattr);
5798         data->timestamp = jiffies;
5799         data->rpc_status = 0;
5800         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5801         data->inode = nfs_igrab_and_active(inode);
5802         if (data->inode) {
5803                 if (data->lr.roc) {
5804                         data->args.lr_args = &data->lr.arg;
5805                         data->res.lr_res = &data->lr.res;
5806                 }
5807         } else if (data->lr.roc) {
5808                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5809                 data->lr.roc = false;
5810         }
5811
5812         task_setup_data.callback_data = data;
5813         msg.rpc_argp = &data->args;
5814         msg.rpc_resp = &data->res;
5815         task = rpc_run_task(&task_setup_data);
5816         if (IS_ERR(task))
5817                 return PTR_ERR(task);
5818         if (!issync)
5819                 goto out;
5820         status = nfs4_wait_for_completion_rpc_task(task);
5821         if (status != 0)
5822                 goto out;
5823         status = data->rpc_status;
5824 out:
5825         rpc_put_task(task);
5826         return status;
5827 }
5828
5829 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5830 {
5831         struct nfs_server *server = NFS_SERVER(inode);
5832         struct nfs4_exception exception = { };
5833         int err;
5834         do {
5835                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5836                 trace_nfs4_delegreturn(inode, stateid, err);
5837                 switch (err) {
5838                         case -NFS4ERR_STALE_STATEID:
5839                         case -NFS4ERR_EXPIRED:
5840                         case 0:
5841                                 return 0;
5842                 }
5843                 err = nfs4_handle_exception(server, err, &exception);
5844         } while (exception.retry);
5845         return err;
5846 }
5847
5848 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5849 {
5850         struct inode *inode = state->inode;
5851         struct nfs_server *server = NFS_SERVER(inode);
5852         struct nfs_client *clp = server->nfs_client;
5853         struct nfs_lockt_args arg = {
5854                 .fh = NFS_FH(inode),
5855                 .fl = request,
5856         };
5857         struct nfs_lockt_res res = {
5858                 .denied = request,
5859         };
5860         struct rpc_message msg = {
5861                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5862                 .rpc_argp       = &arg,
5863                 .rpc_resp       = &res,
5864                 .rpc_cred       = state->owner->so_cred,
5865         };
5866         struct nfs4_lock_state *lsp;
5867         int status;
5868
5869         arg.lock_owner.clientid = clp->cl_clientid;
5870         status = nfs4_set_lock_state(state, request);
5871         if (status != 0)
5872                 goto out;
5873         lsp = request->fl_u.nfs4_fl.owner;
5874         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5875         arg.lock_owner.s_dev = server->s_dev;
5876         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5877         switch (status) {
5878                 case 0:
5879                         request->fl_type = F_UNLCK;
5880                         break;
5881                 case -NFS4ERR_DENIED:
5882                         status = 0;
5883         }
5884         request->fl_ops->fl_release_private(request);
5885         request->fl_ops = NULL;
5886 out:
5887         return status;
5888 }
5889
5890 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5891 {
5892         struct nfs4_exception exception = { };
5893         int err;
5894
5895         do {
5896                 err = _nfs4_proc_getlk(state, cmd, request);
5897                 trace_nfs4_get_lock(request, state, cmd, err);
5898                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5899                                 &exception);
5900         } while (exception.retry);
5901         return err;
5902 }
5903
5904 struct nfs4_unlockdata {
5905         struct nfs_locku_args arg;
5906         struct nfs_locku_res res;
5907         struct nfs4_lock_state *lsp;
5908         struct nfs_open_context *ctx;
5909         struct file_lock fl;
5910         struct nfs_server *server;
5911         unsigned long timestamp;
5912 };
5913
5914 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5915                 struct nfs_open_context *ctx,
5916                 struct nfs4_lock_state *lsp,
5917                 struct nfs_seqid *seqid)
5918 {
5919         struct nfs4_unlockdata *p;
5920         struct inode *inode = lsp->ls_state->inode;
5921
5922         p = kzalloc(sizeof(*p), GFP_NOFS);
5923         if (p == NULL)
5924                 return NULL;
5925         p->arg.fh = NFS_FH(inode);
5926         p->arg.fl = &p->fl;
5927         p->arg.seqid = seqid;
5928         p->res.seqid = seqid;
5929         p->lsp = lsp;
5930         atomic_inc(&lsp->ls_count);
5931         /* Ensure we don't close file until we're done freeing locks! */
5932         p->ctx = get_nfs_open_context(ctx);
5933         memcpy(&p->fl, fl, sizeof(p->fl));
5934         p->server = NFS_SERVER(inode);
5935         return p;
5936 }
5937
5938 static void nfs4_locku_release_calldata(void *data)
5939 {
5940         struct nfs4_unlockdata *calldata = data;
5941         nfs_free_seqid(calldata->arg.seqid);
5942         nfs4_put_lock_state(calldata->lsp);
5943         put_nfs_open_context(calldata->ctx);
5944         kfree(calldata);
5945 }
5946
5947 static void nfs4_locku_done(struct rpc_task *task, void *data)
5948 {
5949         struct nfs4_unlockdata *calldata = data;
5950
5951         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5952                 return;
5953         switch (task->tk_status) {
5954                 case 0:
5955                         renew_lease(calldata->server, calldata->timestamp);
5956                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5957                         if (nfs4_update_lock_stateid(calldata->lsp,
5958                                         &calldata->res.stateid))
5959                                 break;
5960                 case -NFS4ERR_ADMIN_REVOKED:
5961                 case -NFS4ERR_EXPIRED:
5962                         nfs4_free_revoked_stateid(calldata->server,
5963                                         &calldata->arg.stateid,
5964                                         task->tk_msg.rpc_cred);
5965                 case -NFS4ERR_BAD_STATEID:
5966                 case -NFS4ERR_OLD_STATEID:
5967                 case -NFS4ERR_STALE_STATEID:
5968                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5969                                                 &calldata->lsp->ls_stateid))
5970                                 rpc_restart_call_prepare(task);
5971                         break;
5972                 default:
5973                         if (nfs4_async_handle_error(task, calldata->server,
5974                                                     NULL, NULL) == -EAGAIN)
5975                                 rpc_restart_call_prepare(task);
5976         }
5977         nfs_release_seqid(calldata->arg.seqid);
5978 }
5979
5980 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5981 {
5982         struct nfs4_unlockdata *calldata = data;
5983
5984         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5985                 goto out_wait;
5986         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5987         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5988                 /* Note: exit _without_ running nfs4_locku_done */
5989                 goto out_no_action;
5990         }
5991         calldata->timestamp = jiffies;
5992         if (nfs4_setup_sequence(calldata->server,
5993                                 &calldata->arg.seq_args,
5994                                 &calldata->res.seq_res,
5995                                 task) != 0)
5996                 nfs_release_seqid(calldata->arg.seqid);
5997         return;
5998 out_no_action:
5999         task->tk_action = NULL;
6000 out_wait:
6001         nfs4_sequence_done(task, &calldata->res.seq_res);
6002 }
6003
6004 static const struct rpc_call_ops nfs4_locku_ops = {
6005         .rpc_call_prepare = nfs4_locku_prepare,
6006         .rpc_call_done = nfs4_locku_done,
6007         .rpc_release = nfs4_locku_release_calldata,
6008 };
6009
6010 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6011                 struct nfs_open_context *ctx,
6012                 struct nfs4_lock_state *lsp,
6013                 struct nfs_seqid *seqid)
6014 {
6015         struct nfs4_unlockdata *data;
6016         struct rpc_message msg = {
6017                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6018                 .rpc_cred = ctx->cred,
6019         };
6020         struct rpc_task_setup task_setup_data = {
6021                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6022                 .rpc_message = &msg,
6023                 .callback_ops = &nfs4_locku_ops,
6024                 .workqueue = nfsiod_workqueue,
6025                 .flags = RPC_TASK_ASYNC,
6026         };
6027
6028         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6029                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6030
6031         /* Ensure this is an unlock - when canceling a lock, the
6032          * canceled lock is passed in, and it won't be an unlock.
6033          */
6034         fl->fl_type = F_UNLCK;
6035
6036         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6037         if (data == NULL) {
6038                 nfs_free_seqid(seqid);
6039                 return ERR_PTR(-ENOMEM);
6040         }
6041
6042         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6043         msg.rpc_argp = &data->arg;
6044         msg.rpc_resp = &data->res;
6045         task_setup_data.callback_data = data;
6046         return rpc_run_task(&task_setup_data);
6047 }
6048
6049 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6050 {
6051         struct inode *inode = state->inode;
6052         struct nfs4_state_owner *sp = state->owner;
6053         struct nfs_inode *nfsi = NFS_I(inode);
6054         struct nfs_seqid *seqid;
6055         struct nfs4_lock_state *lsp;
6056         struct rpc_task *task;
6057         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6058         int status = 0;
6059         unsigned char fl_flags = request->fl_flags;
6060
6061         status = nfs4_set_lock_state(state, request);
6062         /* Unlock _before_ we do the RPC call */
6063         request->fl_flags |= FL_EXISTS;
6064         /* Exclude nfs_delegation_claim_locks() */
6065         mutex_lock(&sp->so_delegreturn_mutex);
6066         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6067         down_read(&nfsi->rwsem);
6068         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6069                 up_read(&nfsi->rwsem);
6070                 mutex_unlock(&sp->so_delegreturn_mutex);
6071                 goto out;
6072         }
6073         up_read(&nfsi->rwsem);
6074         mutex_unlock(&sp->so_delegreturn_mutex);
6075         if (status != 0)
6076                 goto out;
6077         /* Is this a delegated lock? */
6078         lsp = request->fl_u.nfs4_fl.owner;
6079         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6080                 goto out;
6081         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6082         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6083         status = -ENOMEM;
6084         if (IS_ERR(seqid))
6085                 goto out;
6086         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6087         status = PTR_ERR(task);
6088         if (IS_ERR(task))
6089                 goto out;
6090         status = nfs4_wait_for_completion_rpc_task(task);
6091         rpc_put_task(task);
6092 out:
6093         request->fl_flags = fl_flags;
6094         trace_nfs4_unlock(request, state, F_SETLK, status);
6095         return status;
6096 }
6097
6098 struct nfs4_lockdata {
6099         struct nfs_lock_args arg;
6100         struct nfs_lock_res res;
6101         struct nfs4_lock_state *lsp;
6102         struct nfs_open_context *ctx;
6103         struct file_lock fl;
6104         unsigned long timestamp;
6105         int rpc_status;
6106         int cancelled;
6107         struct nfs_server *server;
6108 };
6109
6110 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6111                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6112                 gfp_t gfp_mask)
6113 {
6114         struct nfs4_lockdata *p;
6115         struct inode *inode = lsp->ls_state->inode;
6116         struct nfs_server *server = NFS_SERVER(inode);
6117         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6118
6119         p = kzalloc(sizeof(*p), gfp_mask);
6120         if (p == NULL)
6121                 return NULL;
6122
6123         p->arg.fh = NFS_FH(inode);
6124         p->arg.fl = &p->fl;
6125         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6126         if (IS_ERR(p->arg.open_seqid))
6127                 goto out_free;
6128         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6129         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6130         if (IS_ERR(p->arg.lock_seqid))
6131                 goto out_free_seqid;
6132         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6133         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6134         p->arg.lock_owner.s_dev = server->s_dev;
6135         p->res.lock_seqid = p->arg.lock_seqid;
6136         p->lsp = lsp;
6137         p->server = server;
6138         atomic_inc(&lsp->ls_count);
6139         p->ctx = get_nfs_open_context(ctx);
6140         memcpy(&p->fl, fl, sizeof(p->fl));
6141         return p;
6142 out_free_seqid:
6143         nfs_free_seqid(p->arg.open_seqid);
6144 out_free:
6145         kfree(p);
6146         return NULL;
6147 }
6148
6149 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6150 {
6151         struct nfs4_lockdata *data = calldata;
6152         struct nfs4_state *state = data->lsp->ls_state;
6153
6154         dprintk("%s: begin!\n", __func__);
6155         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6156                 goto out_wait;
6157         /* Do we need to do an open_to_lock_owner? */
6158         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6159                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6160                         goto out_release_lock_seqid;
6161                 }
6162                 nfs4_stateid_copy(&data->arg.open_stateid,
6163                                 &state->open_stateid);
6164                 data->arg.new_lock_owner = 1;
6165                 data->res.open_seqid = data->arg.open_seqid;
6166         } else {
6167                 data->arg.new_lock_owner = 0;
6168                 nfs4_stateid_copy(&data->arg.lock_stateid,
6169                                 &data->lsp->ls_stateid);
6170         }
6171         if (!nfs4_valid_open_stateid(state)) {
6172                 data->rpc_status = -EBADF;
6173                 task->tk_action = NULL;
6174                 goto out_release_open_seqid;
6175         }
6176         data->timestamp = jiffies;
6177         if (nfs4_setup_sequence(data->server,
6178                                 &data->arg.seq_args,
6179                                 &data->res.seq_res,
6180                                 task) == 0)
6181                 return;
6182 out_release_open_seqid:
6183         nfs_release_seqid(data->arg.open_seqid);
6184 out_release_lock_seqid:
6185         nfs_release_seqid(data->arg.lock_seqid);
6186 out_wait:
6187         nfs4_sequence_done(task, &data->res.seq_res);
6188         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6189 }
6190
6191 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6192 {
6193         struct nfs4_lockdata *data = calldata;
6194         struct nfs4_lock_state *lsp = data->lsp;
6195
6196         dprintk("%s: begin!\n", __func__);
6197
6198         if (!nfs4_sequence_done(task, &data->res.seq_res))
6199                 return;
6200
6201         data->rpc_status = task->tk_status;
6202         switch (task->tk_status) {
6203         case 0:
6204                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6205                                 data->timestamp);
6206                 if (data->arg.new_lock) {
6207                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6208                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6209                                 rpc_restart_call_prepare(task);
6210                                 break;
6211                         }
6212                 }
6213                 if (data->arg.new_lock_owner != 0) {
6214                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6215                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6216                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6217                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6218                         rpc_restart_call_prepare(task);
6219                 break;
6220         case -NFS4ERR_BAD_STATEID:
6221         case -NFS4ERR_OLD_STATEID:
6222         case -NFS4ERR_STALE_STATEID:
6223         case -NFS4ERR_EXPIRED:
6224                 if (data->arg.new_lock_owner != 0) {
6225                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6226                                                 &lsp->ls_state->open_stateid))
6227                                 rpc_restart_call_prepare(task);
6228                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6229                                                 &lsp->ls_stateid))
6230                                 rpc_restart_call_prepare(task);
6231         }
6232         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6233 }
6234
6235 static void nfs4_lock_release(void *calldata)
6236 {
6237         struct nfs4_lockdata *data = calldata;
6238
6239         dprintk("%s: begin!\n", __func__);
6240         nfs_free_seqid(data->arg.open_seqid);
6241         if (data->cancelled != 0) {
6242                 struct rpc_task *task;
6243                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6244                                 data->arg.lock_seqid);
6245                 if (!IS_ERR(task))
6246                         rpc_put_task_async(task);
6247                 dprintk("%s: cancelling lock!\n", __func__);
6248         } else
6249                 nfs_free_seqid(data->arg.lock_seqid);
6250         nfs4_put_lock_state(data->lsp);
6251         put_nfs_open_context(data->ctx);
6252         kfree(data);
6253         dprintk("%s: done!\n", __func__);
6254 }
6255
6256 static const struct rpc_call_ops nfs4_lock_ops = {
6257         .rpc_call_prepare = nfs4_lock_prepare,
6258         .rpc_call_done = nfs4_lock_done,
6259         .rpc_release = nfs4_lock_release,
6260 };
6261
6262 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6263 {
6264         switch (error) {
6265         case -NFS4ERR_ADMIN_REVOKED:
6266         case -NFS4ERR_EXPIRED:
6267         case -NFS4ERR_BAD_STATEID:
6268                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6269                 if (new_lock_owner != 0 ||
6270                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6271                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6272                 break;
6273         case -NFS4ERR_STALE_STATEID:
6274                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6275                 nfs4_schedule_lease_recovery(server->nfs_client);
6276         };
6277 }
6278
6279 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6280 {
6281         struct nfs4_lockdata *data;
6282         struct rpc_task *task;
6283         struct rpc_message msg = {
6284                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6285                 .rpc_cred = state->owner->so_cred,
6286         };
6287         struct rpc_task_setup task_setup_data = {
6288                 .rpc_client = NFS_CLIENT(state->inode),
6289                 .rpc_message = &msg,
6290                 .callback_ops = &nfs4_lock_ops,
6291                 .workqueue = nfsiod_workqueue,
6292                 .flags = RPC_TASK_ASYNC,
6293         };
6294         int ret;
6295
6296         dprintk("%s: begin!\n", __func__);
6297         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6298                         fl->fl_u.nfs4_fl.owner,
6299                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6300         if (data == NULL)
6301                 return -ENOMEM;
6302         if (IS_SETLKW(cmd))
6303                 data->arg.block = 1;
6304         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6305         msg.rpc_argp = &data->arg;
6306         msg.rpc_resp = &data->res;
6307         task_setup_data.callback_data = data;
6308         if (recovery_type > NFS_LOCK_NEW) {
6309                 if (recovery_type == NFS_LOCK_RECLAIM)
6310                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6311                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6312         } else
6313                 data->arg.new_lock = 1;
6314         task = rpc_run_task(&task_setup_data);
6315         if (IS_ERR(task))
6316                 return PTR_ERR(task);
6317         ret = nfs4_wait_for_completion_rpc_task(task);
6318         if (ret == 0) {
6319                 ret = data->rpc_status;
6320                 if (ret)
6321                         nfs4_handle_setlk_error(data->server, data->lsp,
6322                                         data->arg.new_lock_owner, ret);
6323         } else
6324                 data->cancelled = 1;
6325         rpc_put_task(task);
6326         dprintk("%s: done, ret = %d!\n", __func__, ret);
6327         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6328         return ret;
6329 }
6330
6331 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6332 {
6333         struct nfs_server *server = NFS_SERVER(state->inode);
6334         struct nfs4_exception exception = {
6335                 .inode = state->inode,
6336         };
6337         int err;
6338
6339         do {
6340                 /* Cache the lock if possible... */
6341                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6342                         return 0;
6343                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6344                 if (err != -NFS4ERR_DELAY)
6345                         break;
6346                 nfs4_handle_exception(server, err, &exception);
6347         } while (exception.retry);
6348         return err;
6349 }
6350
6351 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6352 {
6353         struct nfs_server *server = NFS_SERVER(state->inode);
6354         struct nfs4_exception exception = {
6355                 .inode = state->inode,
6356         };
6357         int err;
6358
6359         err = nfs4_set_lock_state(state, request);
6360         if (err != 0)
6361                 return err;
6362         if (!recover_lost_locks) {
6363                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6364                 return 0;
6365         }
6366         do {
6367                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6368                         return 0;
6369                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6370                 switch (err) {
6371                 default:
6372                         goto out;
6373                 case -NFS4ERR_GRACE:
6374                 case -NFS4ERR_DELAY:
6375                         nfs4_handle_exception(server, err, &exception);
6376                         err = 0;
6377                 }
6378         } while (exception.retry);
6379 out:
6380         return err;
6381 }
6382
6383 #if defined(CONFIG_NFS_V4_1)
6384 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6385 {
6386         struct nfs4_lock_state *lsp;
6387         int status;
6388
6389         status = nfs4_set_lock_state(state, request);
6390         if (status != 0)
6391                 return status;
6392         lsp = request->fl_u.nfs4_fl.owner;
6393         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6394             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6395                 return 0;
6396         status = nfs4_lock_expired(state, request);
6397         return status;
6398 }
6399 #endif
6400
6401 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6402 {
6403         struct nfs_inode *nfsi = NFS_I(state->inode);
6404         struct nfs4_state_owner *sp = state->owner;
6405         unsigned char fl_flags = request->fl_flags;
6406         int status;
6407
6408         request->fl_flags |= FL_ACCESS;
6409         status = locks_lock_inode_wait(state->inode, request);
6410         if (status < 0)
6411                 goto out;
6412         mutex_lock(&sp->so_delegreturn_mutex);
6413         down_read(&nfsi->rwsem);
6414         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6415                 /* Yes: cache locks! */
6416                 /* ...but avoid races with delegation recall... */
6417                 request->fl_flags = fl_flags & ~FL_SLEEP;
6418                 status = locks_lock_inode_wait(state->inode, request);
6419                 up_read(&nfsi->rwsem);
6420                 mutex_unlock(&sp->so_delegreturn_mutex);
6421                 goto out;
6422         }
6423         up_read(&nfsi->rwsem);
6424         mutex_unlock(&sp->so_delegreturn_mutex);
6425         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6426 out:
6427         request->fl_flags = fl_flags;
6428         return status;
6429 }
6430
6431 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6432 {
6433         struct nfs4_exception exception = {
6434                 .state = state,
6435                 .inode = state->inode,
6436         };
6437         int err;
6438
6439         do {
6440                 err = _nfs4_proc_setlk(state, cmd, request);
6441                 if (err == -NFS4ERR_DENIED)
6442                         err = -EAGAIN;
6443                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6444                                 err, &exception);
6445         } while (exception.retry);
6446         return err;
6447 }
6448
6449 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6450 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6451
6452 static int
6453 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6454                         struct file_lock *request)
6455 {
6456         int             status = -ERESTARTSYS;
6457         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6458
6459         while(!signalled()) {
6460                 status = nfs4_proc_setlk(state, cmd, request);
6461                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6462                         break;
6463                 freezable_schedule_timeout_interruptible(timeout);
6464                 timeout *= 2;
6465                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6466                 status = -ERESTARTSYS;
6467         }
6468         return status;
6469 }
6470
6471 #ifdef CONFIG_NFS_V4_1
6472 struct nfs4_lock_waiter {
6473         struct task_struct      *task;
6474         struct inode            *inode;
6475         struct nfs_lowner       *owner;
6476         bool                    notified;
6477 };
6478
6479 static int
6480 nfs4_wake_lock_waiter(wait_queue_t *wait, unsigned int mode, int flags, void *key)
6481 {
6482         int ret;
6483         struct cb_notify_lock_args *cbnl = key;
6484         struct nfs4_lock_waiter *waiter = wait->private;
6485         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6486                                 *wowner = waiter->owner;
6487
6488         /* Only wake if the callback was for the same owner */
6489         if (lowner->clientid != wowner->clientid ||
6490             lowner->id != wowner->id             ||
6491             lowner->s_dev != wowner->s_dev)
6492                 return 0;
6493
6494         /* Make sure it's for the right inode */
6495         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6496                 return 0;
6497
6498         waiter->notified = true;
6499
6500         /* override "private" so we can use default_wake_function */
6501         wait->private = waiter->task;
6502         ret = autoremove_wake_function(wait, mode, flags, key);
6503         wait->private = waiter;
6504         return ret;
6505 }
6506
6507 static int
6508 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6509 {
6510         int status = -ERESTARTSYS;
6511         unsigned long flags;
6512         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6513         struct nfs_server *server = NFS_SERVER(state->inode);
6514         struct nfs_client *clp = server->nfs_client;
6515         wait_queue_head_t *q = &clp->cl_lock_waitq;
6516         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6517                                     .id = lsp->ls_seqid.owner_id,
6518                                     .s_dev = server->s_dev };
6519         struct nfs4_lock_waiter waiter = { .task  = current,
6520                                            .inode = state->inode,
6521                                            .owner = &owner,
6522                                            .notified = false };
6523         wait_queue_t wait;
6524
6525         /* Don't bother with waitqueue if we don't expect a callback */
6526         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6527                 return nfs4_retry_setlk_simple(state, cmd, request);
6528
6529         init_wait(&wait);
6530         wait.private = &waiter;
6531         wait.func = nfs4_wake_lock_waiter;
6532         add_wait_queue(q, &wait);
6533
6534         while(!signalled()) {
6535                 status = nfs4_proc_setlk(state, cmd, request);
6536                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6537                         break;
6538
6539                 status = -ERESTARTSYS;
6540                 spin_lock_irqsave(&q->lock, flags);
6541                 if (waiter.notified) {
6542                         spin_unlock_irqrestore(&q->lock, flags);
6543                         continue;
6544                 }
6545                 set_current_state(TASK_INTERRUPTIBLE);
6546                 spin_unlock_irqrestore(&q->lock, flags);
6547
6548                 freezable_schedule_timeout_interruptible(NFS4_LOCK_MAXTIMEOUT);
6549         }
6550
6551         finish_wait(q, &wait);
6552         return status;
6553 }
6554 #else /* !CONFIG_NFS_V4_1 */
6555 static inline int
6556 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6557 {
6558         return nfs4_retry_setlk_simple(state, cmd, request);
6559 }
6560 #endif
6561
6562 static int
6563 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6564 {
6565         struct nfs_open_context *ctx;
6566         struct nfs4_state *state;
6567         int status;
6568
6569         /* verify open state */
6570         ctx = nfs_file_open_context(filp);
6571         state = ctx->state;
6572
6573         if (request->fl_start < 0 || request->fl_end < 0)
6574                 return -EINVAL;
6575
6576         if (IS_GETLK(cmd)) {
6577                 if (state != NULL)
6578                         return nfs4_proc_getlk(state, F_GETLK, request);
6579                 return 0;
6580         }
6581
6582         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6583                 return -EINVAL;
6584
6585         if (request->fl_type == F_UNLCK) {
6586                 if (state != NULL)
6587                         return nfs4_proc_unlck(state, cmd, request);
6588                 return 0;
6589         }
6590
6591         if (state == NULL)
6592                 return -ENOLCK;
6593
6594         if ((request->fl_flags & FL_POSIX) &&
6595             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6596                 return -ENOLCK;
6597
6598         /*
6599          * Don't rely on the VFS having checked the file open mode,
6600          * since it won't do this for flock() locks.
6601          */
6602         switch (request->fl_type) {
6603         case F_RDLCK:
6604                 if (!(filp->f_mode & FMODE_READ))
6605                         return -EBADF;
6606                 break;
6607         case F_WRLCK:
6608                 if (!(filp->f_mode & FMODE_WRITE))
6609                         return -EBADF;
6610         }
6611
6612         status = nfs4_set_lock_state(state, request);
6613         if (status != 0)
6614                 return status;
6615
6616         return nfs4_retry_setlk(state, cmd, request);
6617 }
6618
6619 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6620 {
6621         struct nfs_server *server = NFS_SERVER(state->inode);
6622         int err;
6623
6624         err = nfs4_set_lock_state(state, fl);
6625         if (err != 0)
6626                 return err;
6627         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6628         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6629 }
6630
6631 struct nfs_release_lockowner_data {
6632         struct nfs4_lock_state *lsp;
6633         struct nfs_server *server;
6634         struct nfs_release_lockowner_args args;
6635         struct nfs_release_lockowner_res res;
6636         unsigned long timestamp;
6637 };
6638
6639 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6640 {
6641         struct nfs_release_lockowner_data *data = calldata;
6642         struct nfs_server *server = data->server;
6643         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6644                              &data->args.seq_args, &data->res.seq_res, task);
6645         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6646         data->timestamp = jiffies;
6647 }
6648
6649 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6650 {
6651         struct nfs_release_lockowner_data *data = calldata;
6652         struct nfs_server *server = data->server;
6653
6654         nfs40_sequence_done(task, &data->res.seq_res);
6655
6656         switch (task->tk_status) {
6657         case 0:
6658                 renew_lease(server, data->timestamp);
6659                 break;
6660         case -NFS4ERR_STALE_CLIENTID:
6661         case -NFS4ERR_EXPIRED:
6662                 nfs4_schedule_lease_recovery(server->nfs_client);
6663                 break;
6664         case -NFS4ERR_LEASE_MOVED:
6665         case -NFS4ERR_DELAY:
6666                 if (nfs4_async_handle_error(task, server,
6667                                             NULL, NULL) == -EAGAIN)
6668                         rpc_restart_call_prepare(task);
6669         }
6670 }
6671
6672 static void nfs4_release_lockowner_release(void *calldata)
6673 {
6674         struct nfs_release_lockowner_data *data = calldata;
6675         nfs4_free_lock_state(data->server, data->lsp);
6676         kfree(calldata);
6677 }
6678
6679 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6680         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6681         .rpc_call_done = nfs4_release_lockowner_done,
6682         .rpc_release = nfs4_release_lockowner_release,
6683 };
6684
6685 static void
6686 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6687 {
6688         struct nfs_release_lockowner_data *data;
6689         struct rpc_message msg = {
6690                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6691         };
6692
6693         if (server->nfs_client->cl_mvops->minor_version != 0)
6694                 return;
6695
6696         data = kmalloc(sizeof(*data), GFP_NOFS);
6697         if (!data)
6698                 return;
6699         data->lsp = lsp;
6700         data->server = server;
6701         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6702         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6703         data->args.lock_owner.s_dev = server->s_dev;
6704
6705         msg.rpc_argp = &data->args;
6706         msg.rpc_resp = &data->res;
6707         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6708         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6709 }
6710
6711 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6712
6713 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6714                                    struct dentry *unused, struct inode *inode,
6715                                    const char *key, const void *buf,
6716                                    size_t buflen, int flags)
6717 {
6718         return nfs4_proc_set_acl(inode, buf, buflen);
6719 }
6720
6721 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6722                                    struct dentry *unused, struct inode *inode,
6723                                    const char *key, void *buf, size_t buflen)
6724 {
6725         return nfs4_proc_get_acl(inode, buf, buflen);
6726 }
6727
6728 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6729 {
6730         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6731 }
6732
6733 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6734
6735 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6736                                      struct dentry *unused, struct inode *inode,
6737                                      const char *key, const void *buf,
6738                                      size_t buflen, int flags)
6739 {
6740         if (security_ismaclabel(key))
6741                 return nfs4_set_security_label(inode, buf, buflen);
6742
6743         return -EOPNOTSUPP;
6744 }
6745
6746 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6747                                      struct dentry *unused, struct inode *inode,
6748                                      const char *key, void *buf, size_t buflen)
6749 {
6750         if (security_ismaclabel(key))
6751                 return nfs4_get_security_label(inode, buf, buflen);
6752         return -EOPNOTSUPP;
6753 }
6754
6755 static ssize_t
6756 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6757 {
6758         int len = 0;
6759
6760         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6761                 len = security_inode_listsecurity(inode, list, list_len);
6762                 if (list_len && len > list_len)
6763                         return -ERANGE;
6764         }
6765         return len;
6766 }
6767
6768 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6769         .prefix = XATTR_SECURITY_PREFIX,
6770         .get    = nfs4_xattr_get_nfs4_label,
6771         .set    = nfs4_xattr_set_nfs4_label,
6772 };
6773
6774 #else
6775
6776 static ssize_t
6777 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6778 {
6779         return 0;
6780 }
6781
6782 #endif
6783
6784 /*
6785  * nfs_fhget will use either the mounted_on_fileid or the fileid
6786  */
6787 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6788 {
6789         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6790                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6791               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6792               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6793                 return;
6794
6795         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6796                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6797         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6798         fattr->nlink = 2;
6799 }
6800
6801 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6802                                    const struct qstr *name,
6803                                    struct nfs4_fs_locations *fs_locations,
6804                                    struct page *page)
6805 {
6806         struct nfs_server *server = NFS_SERVER(dir);
6807         u32 bitmask[3] = {
6808                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6809         };
6810         struct nfs4_fs_locations_arg args = {
6811                 .dir_fh = NFS_FH(dir),
6812                 .name = name,
6813                 .page = page,
6814                 .bitmask = bitmask,
6815         };
6816         struct nfs4_fs_locations_res res = {
6817                 .fs_locations = fs_locations,
6818         };
6819         struct rpc_message msg = {
6820                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6821                 .rpc_argp = &args,
6822                 .rpc_resp = &res,
6823         };
6824         int status;
6825
6826         dprintk("%s: start\n", __func__);
6827
6828         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6829          * is not supported */
6830         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6831                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6832         else
6833                 bitmask[0] |= FATTR4_WORD0_FILEID;
6834
6835         nfs_fattr_init(&fs_locations->fattr);
6836         fs_locations->server = server;
6837         fs_locations->nlocations = 0;
6838         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6839         dprintk("%s: returned status = %d\n", __func__, status);
6840         return status;
6841 }
6842
6843 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6844                            const struct qstr *name,
6845                            struct nfs4_fs_locations *fs_locations,
6846                            struct page *page)
6847 {
6848         struct nfs4_exception exception = { };
6849         int err;
6850         do {
6851                 err = _nfs4_proc_fs_locations(client, dir, name,
6852                                 fs_locations, page);
6853                 trace_nfs4_get_fs_locations(dir, name, err);
6854                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6855                                 &exception);
6856         } while (exception.retry);
6857         return err;
6858 }
6859
6860 /*
6861  * This operation also signals the server that this client is
6862  * performing migration recovery.  The server can stop returning
6863  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6864  * appended to this compound to identify the client ID which is
6865  * performing recovery.
6866  */
6867 static int _nfs40_proc_get_locations(struct inode *inode,
6868                                      struct nfs4_fs_locations *locations,
6869                                      struct page *page, struct rpc_cred *cred)
6870 {
6871         struct nfs_server *server = NFS_SERVER(inode);
6872         struct rpc_clnt *clnt = server->client;
6873         u32 bitmask[2] = {
6874                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6875         };
6876         struct nfs4_fs_locations_arg args = {
6877                 .clientid       = server->nfs_client->cl_clientid,
6878                 .fh             = NFS_FH(inode),
6879                 .page           = page,
6880                 .bitmask        = bitmask,
6881                 .migration      = 1,            /* skip LOOKUP */
6882                 .renew          = 1,            /* append RENEW */
6883         };
6884         struct nfs4_fs_locations_res res = {
6885                 .fs_locations   = locations,
6886                 .migration      = 1,
6887                 .renew          = 1,
6888         };
6889         struct rpc_message msg = {
6890                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6891                 .rpc_argp       = &args,
6892                 .rpc_resp       = &res,
6893                 .rpc_cred       = cred,
6894         };
6895         unsigned long now = jiffies;
6896         int status;
6897
6898         nfs_fattr_init(&locations->fattr);
6899         locations->server = server;
6900         locations->nlocations = 0;
6901
6902         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6903         nfs4_set_sequence_privileged(&args.seq_args);
6904         status = nfs4_call_sync_sequence(clnt, server, &msg,
6905                                         &args.seq_args, &res.seq_res);
6906         if (status)
6907                 return status;
6908
6909         renew_lease(server, now);
6910         return 0;
6911 }
6912
6913 #ifdef CONFIG_NFS_V4_1
6914
6915 /*
6916  * This operation also signals the server that this client is
6917  * performing migration recovery.  The server can stop asserting
6918  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6919  * performing this operation is identified in the SEQUENCE
6920  * operation in this compound.
6921  *
6922  * When the client supports GETATTR(fs_locations_info), it can
6923  * be plumbed in here.
6924  */
6925 static int _nfs41_proc_get_locations(struct inode *inode,
6926                                      struct nfs4_fs_locations *locations,
6927                                      struct page *page, struct rpc_cred *cred)
6928 {
6929         struct nfs_server *server = NFS_SERVER(inode);
6930         struct rpc_clnt *clnt = server->client;
6931         u32 bitmask[2] = {
6932                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6933         };
6934         struct nfs4_fs_locations_arg args = {
6935                 .fh             = NFS_FH(inode),
6936                 .page           = page,
6937                 .bitmask        = bitmask,
6938                 .migration      = 1,            /* skip LOOKUP */
6939         };
6940         struct nfs4_fs_locations_res res = {
6941                 .fs_locations   = locations,
6942                 .migration      = 1,
6943         };
6944         struct rpc_message msg = {
6945                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6946                 .rpc_argp       = &args,
6947                 .rpc_resp       = &res,
6948                 .rpc_cred       = cred,
6949         };
6950         int status;
6951
6952         nfs_fattr_init(&locations->fattr);
6953         locations->server = server;
6954         locations->nlocations = 0;
6955
6956         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6957         nfs4_set_sequence_privileged(&args.seq_args);
6958         status = nfs4_call_sync_sequence(clnt, server, &msg,
6959                                         &args.seq_args, &res.seq_res);
6960         if (status == NFS4_OK &&
6961             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6962                 status = -NFS4ERR_LEASE_MOVED;
6963         return status;
6964 }
6965
6966 #endif  /* CONFIG_NFS_V4_1 */
6967
6968 /**
6969  * nfs4_proc_get_locations - discover locations for a migrated FSID
6970  * @inode: inode on FSID that is migrating
6971  * @locations: result of query
6972  * @page: buffer
6973  * @cred: credential to use for this operation
6974  *
6975  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6976  * operation failed, or a negative errno if a local error occurred.
6977  *
6978  * On success, "locations" is filled in, but if the server has
6979  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6980  * asserted.
6981  *
6982  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6983  * from this client that require migration recovery.
6984  */
6985 int nfs4_proc_get_locations(struct inode *inode,
6986                             struct nfs4_fs_locations *locations,
6987                             struct page *page, struct rpc_cred *cred)
6988 {
6989         struct nfs_server *server = NFS_SERVER(inode);
6990         struct nfs_client *clp = server->nfs_client;
6991         const struct nfs4_mig_recovery_ops *ops =
6992                                         clp->cl_mvops->mig_recovery_ops;
6993         struct nfs4_exception exception = { };
6994         int status;
6995
6996         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6997                 (unsigned long long)server->fsid.major,
6998                 (unsigned long long)server->fsid.minor,
6999                 clp->cl_hostname);
7000         nfs_display_fhandle(NFS_FH(inode), __func__);
7001
7002         do {
7003                 status = ops->get_locations(inode, locations, page, cred);
7004                 if (status != -NFS4ERR_DELAY)
7005                         break;
7006                 nfs4_handle_exception(server, status, &exception);
7007         } while (exception.retry);
7008         return status;
7009 }
7010
7011 /*
7012  * This operation also signals the server that this client is
7013  * performing "lease moved" recovery.  The server can stop
7014  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7015  * is appended to this compound to identify the client ID which is
7016  * performing recovery.
7017  */
7018 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7019 {
7020         struct nfs_server *server = NFS_SERVER(inode);
7021         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7022         struct rpc_clnt *clnt = server->client;
7023         struct nfs4_fsid_present_arg args = {
7024                 .fh             = NFS_FH(inode),
7025                 .clientid       = clp->cl_clientid,
7026                 .renew          = 1,            /* append RENEW */
7027         };
7028         struct nfs4_fsid_present_res res = {
7029                 .renew          = 1,
7030         };
7031         struct rpc_message msg = {
7032                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7033                 .rpc_argp       = &args,
7034                 .rpc_resp       = &res,
7035                 .rpc_cred       = cred,
7036         };
7037         unsigned long now = jiffies;
7038         int status;
7039
7040         res.fh = nfs_alloc_fhandle();
7041         if (res.fh == NULL)
7042                 return -ENOMEM;
7043
7044         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7045         nfs4_set_sequence_privileged(&args.seq_args);
7046         status = nfs4_call_sync_sequence(clnt, server, &msg,
7047                                                 &args.seq_args, &res.seq_res);
7048         nfs_free_fhandle(res.fh);
7049         if (status)
7050                 return status;
7051
7052         do_renew_lease(clp, now);
7053         return 0;
7054 }
7055
7056 #ifdef CONFIG_NFS_V4_1
7057
7058 /*
7059  * This operation also signals the server that this client is
7060  * performing "lease moved" recovery.  The server can stop asserting
7061  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7062  * this operation is identified in the SEQUENCE operation in this
7063  * compound.
7064  */
7065 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7066 {
7067         struct nfs_server *server = NFS_SERVER(inode);
7068         struct rpc_clnt *clnt = server->client;
7069         struct nfs4_fsid_present_arg args = {
7070                 .fh             = NFS_FH(inode),
7071         };
7072         struct nfs4_fsid_present_res res = {
7073         };
7074         struct rpc_message msg = {
7075                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7076                 .rpc_argp       = &args,
7077                 .rpc_resp       = &res,
7078                 .rpc_cred       = cred,
7079         };
7080         int status;
7081
7082         res.fh = nfs_alloc_fhandle();
7083         if (res.fh == NULL)
7084                 return -ENOMEM;
7085
7086         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7087         nfs4_set_sequence_privileged(&args.seq_args);
7088         status = nfs4_call_sync_sequence(clnt, server, &msg,
7089                                                 &args.seq_args, &res.seq_res);
7090         nfs_free_fhandle(res.fh);
7091         if (status == NFS4_OK &&
7092             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7093                 status = -NFS4ERR_LEASE_MOVED;
7094         return status;
7095 }
7096
7097 #endif  /* CONFIG_NFS_V4_1 */
7098
7099 /**
7100  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7101  * @inode: inode on FSID to check
7102  * @cred: credential to use for this operation
7103  *
7104  * Server indicates whether the FSID is present, moved, or not
7105  * recognized.  This operation is necessary to clear a LEASE_MOVED
7106  * condition for this client ID.
7107  *
7108  * Returns NFS4_OK if the FSID is present on this server,
7109  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7110  *  NFS4ERR code if some error occurred on the server, or a
7111  *  negative errno if a local failure occurred.
7112  */
7113 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7114 {
7115         struct nfs_server *server = NFS_SERVER(inode);
7116         struct nfs_client *clp = server->nfs_client;
7117         const struct nfs4_mig_recovery_ops *ops =
7118                                         clp->cl_mvops->mig_recovery_ops;
7119         struct nfs4_exception exception = { };
7120         int status;
7121
7122         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7123                 (unsigned long long)server->fsid.major,
7124                 (unsigned long long)server->fsid.minor,
7125                 clp->cl_hostname);
7126         nfs_display_fhandle(NFS_FH(inode), __func__);
7127
7128         do {
7129                 status = ops->fsid_present(inode, cred);
7130                 if (status != -NFS4ERR_DELAY)
7131                         break;
7132                 nfs4_handle_exception(server, status, &exception);
7133         } while (exception.retry);
7134         return status;
7135 }
7136
7137 /**
7138  * If 'use_integrity' is true and the state managment nfs_client
7139  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7140  * and the machine credential as per RFC3530bis and RFC5661 Security
7141  * Considerations sections. Otherwise, just use the user cred with the
7142  * filesystem's rpc_client.
7143  */
7144 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7145 {
7146         int status;
7147         struct nfs4_secinfo_arg args = {
7148                 .dir_fh = NFS_FH(dir),
7149                 .name   = name,
7150         };
7151         struct nfs4_secinfo_res res = {
7152                 .flavors     = flavors,
7153         };
7154         struct rpc_message msg = {
7155                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7156                 .rpc_argp = &args,
7157                 .rpc_resp = &res,
7158         };
7159         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7160         struct rpc_cred *cred = NULL;
7161
7162         if (use_integrity) {
7163                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7164                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7165                 msg.rpc_cred = cred;
7166         }
7167
7168         dprintk("NFS call  secinfo %s\n", name->name);
7169
7170         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7171                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7172
7173         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7174                                 &res.seq_res, 0);
7175         dprintk("NFS reply  secinfo: %d\n", status);
7176
7177         if (cred)
7178                 put_rpccred(cred);
7179
7180         return status;
7181 }
7182
7183 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7184                       struct nfs4_secinfo_flavors *flavors)
7185 {
7186         struct nfs4_exception exception = { };
7187         int err;
7188         do {
7189                 err = -NFS4ERR_WRONGSEC;
7190
7191                 /* try to use integrity protection with machine cred */
7192                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7193                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7194
7195                 /*
7196                  * if unable to use integrity protection, or SECINFO with
7197                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7198                  * disallowed by spec, but exists in deployed servers) use
7199                  * the current filesystem's rpc_client and the user cred.
7200                  */
7201                 if (err == -NFS4ERR_WRONGSEC)
7202                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7203
7204                 trace_nfs4_secinfo(dir, name, err);
7205                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7206                                 &exception);
7207         } while (exception.retry);
7208         return err;
7209 }
7210
7211 #ifdef CONFIG_NFS_V4_1
7212 /*
7213  * Check the exchange flags returned by the server for invalid flags, having
7214  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7215  * DS flags set.
7216  */
7217 static int nfs4_check_cl_exchange_flags(u32 flags)
7218 {
7219         if (flags & ~EXCHGID4_FLAG_MASK_R)
7220                 goto out_inval;
7221         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7222             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7223                 goto out_inval;
7224         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7225                 goto out_inval;
7226         return NFS_OK;
7227 out_inval:
7228         return -NFS4ERR_INVAL;
7229 }
7230
7231 static bool
7232 nfs41_same_server_scope(struct nfs41_server_scope *a,
7233                         struct nfs41_server_scope *b)
7234 {
7235         if (a->server_scope_sz == b->server_scope_sz &&
7236             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
7237                 return true;
7238
7239         return false;
7240 }
7241
7242 static void
7243 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7244 {
7245 }
7246
7247 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7248         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7249 };
7250
7251 /*
7252  * nfs4_proc_bind_one_conn_to_session()
7253  *
7254  * The 4.1 client currently uses the same TCP connection for the
7255  * fore and backchannel.
7256  */
7257 static
7258 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7259                 struct rpc_xprt *xprt,
7260                 struct nfs_client *clp,
7261                 struct rpc_cred *cred)
7262 {
7263         int status;
7264         struct nfs41_bind_conn_to_session_args args = {
7265                 .client = clp,
7266                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7267         };
7268         struct nfs41_bind_conn_to_session_res res;
7269         struct rpc_message msg = {
7270                 .rpc_proc =
7271                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7272                 .rpc_argp = &args,
7273                 .rpc_resp = &res,
7274                 .rpc_cred = cred,
7275         };
7276         struct rpc_task_setup task_setup_data = {
7277                 .rpc_client = clnt,
7278                 .rpc_xprt = xprt,
7279                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7280                 .rpc_message = &msg,
7281                 .flags = RPC_TASK_TIMEOUT,
7282         };
7283         struct rpc_task *task;
7284
7285         dprintk("--> %s\n", __func__);
7286
7287         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7288         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7289                 args.dir = NFS4_CDFC4_FORE;
7290
7291         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7292         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7293                 args.dir = NFS4_CDFC4_FORE;
7294
7295         task = rpc_run_task(&task_setup_data);
7296         if (!IS_ERR(task)) {
7297                 status = task->tk_status;
7298                 rpc_put_task(task);
7299         } else
7300                 status = PTR_ERR(task);
7301         trace_nfs4_bind_conn_to_session(clp, status);
7302         if (status == 0) {
7303                 if (memcmp(res.sessionid.data,
7304                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7305                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7306                         status = -EIO;
7307                         goto out;
7308                 }
7309                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7310                         dprintk("NFS: %s: Unexpected direction from server\n",
7311                                 __func__);
7312                         status = -EIO;
7313                         goto out;
7314                 }
7315                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7316                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7317                                 __func__);
7318                         status = -EIO;
7319                         goto out;
7320                 }
7321         }
7322 out:
7323         dprintk("<-- %s status= %d\n", __func__, status);
7324         return status;
7325 }
7326
7327 struct rpc_bind_conn_calldata {
7328         struct nfs_client *clp;
7329         struct rpc_cred *cred;
7330 };
7331
7332 static int
7333 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7334                 struct rpc_xprt *xprt,
7335                 void *calldata)
7336 {
7337         struct rpc_bind_conn_calldata *p = calldata;
7338
7339         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7340 }
7341
7342 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7343 {
7344         struct rpc_bind_conn_calldata data = {
7345                 .clp = clp,
7346                 .cred = cred,
7347         };
7348         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7349                         nfs4_proc_bind_conn_to_session_callback, &data);
7350 }
7351
7352 /*
7353  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7354  * and operations we'd like to see to enable certain features in the allow map
7355  */
7356 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7357         .how = SP4_MACH_CRED,
7358         .enforce.u.words = {
7359                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7360                       1 << (OP_EXCHANGE_ID - 32) |
7361                       1 << (OP_CREATE_SESSION - 32) |
7362                       1 << (OP_DESTROY_SESSION - 32) |
7363                       1 << (OP_DESTROY_CLIENTID - 32)
7364         },
7365         .allow.u.words = {
7366                 [0] = 1 << (OP_CLOSE) |
7367                       1 << (OP_OPEN_DOWNGRADE) |
7368                       1 << (OP_LOCKU) |
7369                       1 << (OP_DELEGRETURN) |
7370                       1 << (OP_COMMIT),
7371                 [1] = 1 << (OP_SECINFO - 32) |
7372                       1 << (OP_SECINFO_NO_NAME - 32) |
7373                       1 << (OP_LAYOUTRETURN - 32) |
7374                       1 << (OP_TEST_STATEID - 32) |
7375                       1 << (OP_FREE_STATEID - 32) |
7376                       1 << (OP_WRITE - 32)
7377         }
7378 };
7379
7380 /*
7381  * Select the state protection mode for client `clp' given the server results
7382  * from exchange_id in `sp'.
7383  *
7384  * Returns 0 on success, negative errno otherwise.
7385  */
7386 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7387                                  struct nfs41_state_protection *sp)
7388 {
7389         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7390                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7391                       1 << (OP_EXCHANGE_ID - 32) |
7392                       1 << (OP_CREATE_SESSION - 32) |
7393                       1 << (OP_DESTROY_SESSION - 32) |
7394                       1 << (OP_DESTROY_CLIENTID - 32)
7395         };
7396         unsigned int i;
7397
7398         if (sp->how == SP4_MACH_CRED) {
7399                 /* Print state protect result */
7400                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7401                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7402                         if (test_bit(i, sp->enforce.u.longs))
7403                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7404                         if (test_bit(i, sp->allow.u.longs))
7405                                 dfprintk(MOUNT, "  allow op %d\n", i);
7406                 }
7407
7408                 /* make sure nothing is on enforce list that isn't supported */
7409                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7410                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7411                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7412                                 return -EINVAL;
7413                         }
7414                 }
7415
7416                 /*
7417                  * Minimal mode - state operations are allowed to use machine
7418                  * credential.  Note this already happens by default, so the
7419                  * client doesn't have to do anything more than the negotiation.
7420                  *
7421                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7422                  *       we're already using the machine cred for exchange_id
7423                  *       and will never use a different cred.
7424                  */
7425                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7426                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7427                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7428                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7429                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7430                         dfprintk(MOUNT, "  minimal mode enabled\n");
7431                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7432                 } else {
7433                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7434                         return -EINVAL;
7435                 }
7436
7437                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7438                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7439                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7440                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7441                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7442                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7443                 }
7444
7445                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7446                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7447                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7448                                 &clp->cl_sp4_flags);
7449                 }
7450
7451                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7452                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7453                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7454                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7455                 }
7456
7457                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7458                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7459                         dfprintk(MOUNT, "  stateid mode enabled\n");
7460                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7461                 }
7462
7463                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7464                         dfprintk(MOUNT, "  write mode enabled\n");
7465                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7466                 }
7467
7468                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7469                         dfprintk(MOUNT, "  commit mode enabled\n");
7470                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7471                 }
7472         }
7473
7474         return 0;
7475 }
7476
7477 struct nfs41_exchange_id_data {
7478         struct nfs41_exchange_id_res res;
7479         struct nfs41_exchange_id_args args;
7480         struct rpc_xprt *xprt;
7481         int rpc_status;
7482 };
7483
7484 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7485 {
7486         struct nfs41_exchange_id_data *cdata =
7487                                         (struct nfs41_exchange_id_data *)data;
7488         struct nfs_client *clp = cdata->args.client;
7489         int status = task->tk_status;
7490
7491         trace_nfs4_exchange_id(clp, status);
7492
7493         if (status == 0)
7494                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7495
7496         if (cdata->xprt && status == 0) {
7497                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7498                                                       cdata->xprt);
7499                 goto out;
7500         }
7501
7502         if (status  == 0)
7503                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7504
7505         if (status == 0) {
7506                 clp->cl_clientid = cdata->res.clientid;
7507                 clp->cl_exchange_flags = cdata->res.flags;
7508                 /* Client ID is not confirmed */
7509                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7510                         clear_bit(NFS4_SESSION_ESTABLISHED,
7511                         &clp->cl_session->session_state);
7512                         clp->cl_seqid = cdata->res.seqid;
7513                 }
7514
7515                 kfree(clp->cl_serverowner);
7516                 clp->cl_serverowner = cdata->res.server_owner;
7517                 cdata->res.server_owner = NULL;
7518
7519                 /* use the most recent implementation id */
7520                 kfree(clp->cl_implid);
7521                 clp->cl_implid = cdata->res.impl_id;
7522                 cdata->res.impl_id = NULL;
7523
7524                 if (clp->cl_serverscope != NULL &&
7525                     !nfs41_same_server_scope(clp->cl_serverscope,
7526                                         cdata->res.server_scope)) {
7527                         dprintk("%s: server_scope mismatch detected\n",
7528                                 __func__);
7529                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7530                         kfree(clp->cl_serverscope);
7531                         clp->cl_serverscope = NULL;
7532                 }
7533
7534                 if (clp->cl_serverscope == NULL) {
7535                         clp->cl_serverscope = cdata->res.server_scope;
7536                         cdata->res.server_scope = NULL;
7537                 }
7538                 /* Save the EXCHANGE_ID verifier session trunk tests */
7539                 memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7540                        sizeof(clp->cl_confirm.data));
7541         }
7542 out:
7543         cdata->rpc_status = status;
7544         return;
7545 }
7546
7547 static void nfs4_exchange_id_release(void *data)
7548 {
7549         struct nfs41_exchange_id_data *cdata =
7550                                         (struct nfs41_exchange_id_data *)data;
7551
7552         nfs_put_client(cdata->args.client);
7553         if (cdata->xprt) {
7554                 xprt_put(cdata->xprt);
7555                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7556         }
7557         kfree(cdata->res.impl_id);
7558         kfree(cdata->res.server_scope);
7559         kfree(cdata->res.server_owner);
7560         kfree(cdata);
7561 }
7562
7563 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7564         .rpc_call_done = nfs4_exchange_id_done,
7565         .rpc_release = nfs4_exchange_id_release,
7566 };
7567
7568 /*
7569  * _nfs4_proc_exchange_id()
7570  *
7571  * Wrapper for EXCHANGE_ID operation.
7572  */
7573 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7574                         u32 sp4_how, struct rpc_xprt *xprt)
7575 {
7576         nfs4_verifier verifier;
7577         struct rpc_message msg = {
7578                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7579                 .rpc_cred = cred,
7580         };
7581         struct rpc_task_setup task_setup_data = {
7582                 .rpc_client = clp->cl_rpcclient,
7583                 .callback_ops = &nfs4_exchange_id_call_ops,
7584                 .rpc_message = &msg,
7585                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7586         };
7587         struct nfs41_exchange_id_data *calldata;
7588         struct rpc_task *task;
7589         int status = -EIO;
7590
7591         if (!atomic_inc_not_zero(&clp->cl_count))
7592                 goto out;
7593
7594         status = -ENOMEM;
7595         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7596         if (!calldata)
7597                 goto out;
7598
7599         if (!xprt)
7600                 nfs4_init_boot_verifier(clp, &verifier);
7601
7602         status = nfs4_init_uniform_client_string(clp);
7603         if (status)
7604                 goto out_calldata;
7605
7606         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7607                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7608                 clp->cl_owner_id);
7609
7610         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7611                                                 GFP_NOFS);
7612         status = -ENOMEM;
7613         if (unlikely(calldata->res.server_owner == NULL))
7614                 goto out_calldata;
7615
7616         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7617                                         GFP_NOFS);
7618         if (unlikely(calldata->res.server_scope == NULL))
7619                 goto out_server_owner;
7620
7621         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7622         if (unlikely(calldata->res.impl_id == NULL))
7623                 goto out_server_scope;
7624
7625         switch (sp4_how) {
7626         case SP4_NONE:
7627                 calldata->args.state_protect.how = SP4_NONE;
7628                 break;
7629
7630         case SP4_MACH_CRED:
7631                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7632                 break;
7633
7634         default:
7635                 /* unsupported! */
7636                 WARN_ON_ONCE(1);
7637                 status = -EINVAL;
7638                 goto out_impl_id;
7639         }
7640         if (xprt) {
7641                 calldata->xprt = xprt;
7642                 task_setup_data.rpc_xprt = xprt;
7643                 task_setup_data.flags =
7644                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7645                 calldata->args.verifier = &clp->cl_confirm;
7646         } else {
7647                 calldata->args.verifier = &verifier;
7648         }
7649         calldata->args.client = clp;
7650 #ifdef CONFIG_NFS_V4_1_MIGRATION
7651         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7652         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7653         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7654 #else
7655         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7656         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7657 #endif
7658         msg.rpc_argp = &calldata->args;
7659         msg.rpc_resp = &calldata->res;
7660         task_setup_data.callback_data = calldata;
7661
7662         task = rpc_run_task(&task_setup_data);
7663         if (IS_ERR(task)) {
7664         status = PTR_ERR(task);
7665                 goto out_impl_id;
7666         }
7667
7668         if (!xprt) {
7669                 status = rpc_wait_for_completion_task(task);
7670                 if (!status)
7671                         status = calldata->rpc_status;
7672         } else  /* session trunking test */
7673                 status = calldata->rpc_status;
7674
7675         rpc_put_task(task);
7676 out:
7677         if (clp->cl_implid != NULL)
7678                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7679                         "domain: %s, name: %s, date: %llu,%u\n",
7680                         clp->cl_implid->domain, clp->cl_implid->name,
7681                         clp->cl_implid->date.seconds,
7682                         clp->cl_implid->date.nseconds);
7683         dprintk("NFS reply exchange_id: %d\n", status);
7684         return status;
7685
7686 out_impl_id:
7687         kfree(calldata->res.impl_id);
7688 out_server_scope:
7689         kfree(calldata->res.server_scope);
7690 out_server_owner:
7691         kfree(calldata->res.server_owner);
7692 out_calldata:
7693         kfree(calldata);
7694         goto out;
7695 }
7696
7697 /*
7698  * nfs4_proc_exchange_id()
7699  *
7700  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7701  *
7702  * Since the clientid has expired, all compounds using sessions
7703  * associated with the stale clientid will be returning
7704  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7705  * be in some phase of session reset.
7706  *
7707  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7708  */
7709 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7710 {
7711         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7712         int status;
7713
7714         /* try SP4_MACH_CRED if krb5i/p */
7715         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7716             authflavor == RPC_AUTH_GSS_KRB5P) {
7717                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7718                 if (!status)
7719                         return 0;
7720         }
7721
7722         /* try SP4_NONE */
7723         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7724 }
7725
7726 /**
7727  * nfs4_test_session_trunk
7728  *
7729  * This is an add_xprt_test() test function called from
7730  * rpc_clnt_setup_test_and_add_xprt.
7731  *
7732  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7733  * and is dereferrenced in nfs4_exchange_id_release
7734  *
7735  * Upon success, add the new transport to the rpc_clnt
7736  *
7737  * @clnt: struct rpc_clnt to get new transport
7738  * @xprt: the rpc_xprt to test
7739  * @data: call data for _nfs4_proc_exchange_id.
7740  */
7741 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7742                             void *data)
7743 {
7744         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7745         u32 sp4_how;
7746
7747         dprintk("--> %s try %s\n", __func__,
7748                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7749
7750         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7751
7752         /* Test connection for session trunking. Async exchange_id call */
7753         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7754 }
7755 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7756
7757 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7758                 struct rpc_cred *cred)
7759 {
7760         struct rpc_message msg = {
7761                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7762                 .rpc_argp = clp,
7763                 .rpc_cred = cred,
7764         };
7765         int status;
7766
7767         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7768         trace_nfs4_destroy_clientid(clp, status);
7769         if (status)
7770                 dprintk("NFS: Got error %d from the server %s on "
7771                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7772         return status;
7773 }
7774
7775 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7776                 struct rpc_cred *cred)
7777 {
7778         unsigned int loop;
7779         int ret;
7780
7781         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7782                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7783                 switch (ret) {
7784                 case -NFS4ERR_DELAY:
7785                 case -NFS4ERR_CLIENTID_BUSY:
7786                         ssleep(1);
7787                         break;
7788                 default:
7789                         return ret;
7790                 }
7791         }
7792         return 0;
7793 }
7794
7795 int nfs4_destroy_clientid(struct nfs_client *clp)
7796 {
7797         struct rpc_cred *cred;
7798         int ret = 0;
7799
7800         if (clp->cl_mvops->minor_version < 1)
7801                 goto out;
7802         if (clp->cl_exchange_flags == 0)
7803                 goto out;
7804         if (clp->cl_preserve_clid)
7805                 goto out;
7806         cred = nfs4_get_clid_cred(clp);
7807         ret = nfs4_proc_destroy_clientid(clp, cred);
7808         if (cred)
7809                 put_rpccred(cred);
7810         switch (ret) {
7811         case 0:
7812         case -NFS4ERR_STALE_CLIENTID:
7813                 clp->cl_exchange_flags = 0;
7814         }
7815 out:
7816         return ret;
7817 }
7818
7819 struct nfs4_get_lease_time_data {
7820         struct nfs4_get_lease_time_args *args;
7821         struct nfs4_get_lease_time_res *res;
7822         struct nfs_client *clp;
7823 };
7824
7825 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7826                                         void *calldata)
7827 {
7828         struct nfs4_get_lease_time_data *data =
7829                         (struct nfs4_get_lease_time_data *)calldata;
7830
7831         dprintk("--> %s\n", __func__);
7832         /* just setup sequence, do not trigger session recovery
7833            since we're invoked within one */
7834         nfs41_setup_sequence(data->clp->cl_session,
7835                         &data->args->la_seq_args,
7836                         &data->res->lr_seq_res,
7837                         task);
7838         dprintk("<-- %s\n", __func__);
7839 }
7840
7841 /*
7842  * Called from nfs4_state_manager thread for session setup, so don't recover
7843  * from sequence operation or clientid errors.
7844  */
7845 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7846 {
7847         struct nfs4_get_lease_time_data *data =
7848                         (struct nfs4_get_lease_time_data *)calldata;
7849
7850         dprintk("--> %s\n", __func__);
7851         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7852                 return;
7853         switch (task->tk_status) {
7854         case -NFS4ERR_DELAY:
7855         case -NFS4ERR_GRACE:
7856                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7857                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7858                 task->tk_status = 0;
7859                 /* fall through */
7860         case -NFS4ERR_RETRY_UNCACHED_REP:
7861                 rpc_restart_call_prepare(task);
7862                 return;
7863         }
7864         dprintk("<-- %s\n", __func__);
7865 }
7866
7867 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7868         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7869         .rpc_call_done = nfs4_get_lease_time_done,
7870 };
7871
7872 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7873 {
7874         struct rpc_task *task;
7875         struct nfs4_get_lease_time_args args;
7876         struct nfs4_get_lease_time_res res = {
7877                 .lr_fsinfo = fsinfo,
7878         };
7879         struct nfs4_get_lease_time_data data = {
7880                 .args = &args,
7881                 .res = &res,
7882                 .clp = clp,
7883         };
7884         struct rpc_message msg = {
7885                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7886                 .rpc_argp = &args,
7887                 .rpc_resp = &res,
7888         };
7889         struct rpc_task_setup task_setup = {
7890                 .rpc_client = clp->cl_rpcclient,
7891                 .rpc_message = &msg,
7892                 .callback_ops = &nfs4_get_lease_time_ops,
7893                 .callback_data = &data,
7894                 .flags = RPC_TASK_TIMEOUT,
7895         };
7896         int status;
7897
7898         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7899         nfs4_set_sequence_privileged(&args.la_seq_args);
7900         dprintk("--> %s\n", __func__);
7901         task = rpc_run_task(&task_setup);
7902
7903         if (IS_ERR(task))
7904                 status = PTR_ERR(task);
7905         else {
7906                 status = task->tk_status;
7907                 rpc_put_task(task);
7908         }
7909         dprintk("<-- %s return %d\n", __func__, status);
7910
7911         return status;
7912 }
7913
7914 /*
7915  * Initialize the values to be used by the client in CREATE_SESSION
7916  * If nfs4_init_session set the fore channel request and response sizes,
7917  * use them.
7918  *
7919  * Set the back channel max_resp_sz_cached to zero to force the client to
7920  * always set csa_cachethis to FALSE because the current implementation
7921  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7922  */
7923 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7924                                     struct rpc_clnt *clnt)
7925 {
7926         unsigned int max_rqst_sz, max_resp_sz;
7927         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7928
7929         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7930         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7931
7932         /* Fore channel attributes */
7933         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7934         args->fc_attrs.max_resp_sz = max_resp_sz;
7935         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7936         args->fc_attrs.max_reqs = max_session_slots;
7937
7938         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7939                 "max_ops=%u max_reqs=%u\n",
7940                 __func__,
7941                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7942                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7943
7944         /* Back channel attributes */
7945         args->bc_attrs.max_rqst_sz = max_bc_payload;
7946         args->bc_attrs.max_resp_sz = max_bc_payload;
7947         args->bc_attrs.max_resp_sz_cached = 0;
7948         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7949         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7950
7951         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7952                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7953                 __func__,
7954                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7955                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7956                 args->bc_attrs.max_reqs);
7957 }
7958
7959 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7960                 struct nfs41_create_session_res *res)
7961 {
7962         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7963         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7964
7965         if (rcvd->max_resp_sz > sent->max_resp_sz)
7966                 return -EINVAL;
7967         /*
7968          * Our requested max_ops is the minimum we need; we're not
7969          * prepared to break up compounds into smaller pieces than that.
7970          * So, no point even trying to continue if the server won't
7971          * cooperate:
7972          */
7973         if (rcvd->max_ops < sent->max_ops)
7974                 return -EINVAL;
7975         if (rcvd->max_reqs == 0)
7976                 return -EINVAL;
7977         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7978                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7979         return 0;
7980 }
7981
7982 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7983                 struct nfs41_create_session_res *res)
7984 {
7985         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7986         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7987
7988         if (!(res->flags & SESSION4_BACK_CHAN))
7989                 goto out;
7990         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7991                 return -EINVAL;
7992         if (rcvd->max_resp_sz < sent->max_resp_sz)
7993                 return -EINVAL;
7994         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7995                 return -EINVAL;
7996         if (rcvd->max_ops > sent->max_ops)
7997                 return -EINVAL;
7998         if (rcvd->max_reqs > sent->max_reqs)
7999                 return -EINVAL;
8000 out:
8001         return 0;
8002 }
8003
8004 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8005                                      struct nfs41_create_session_res *res)
8006 {
8007         int ret;
8008
8009         ret = nfs4_verify_fore_channel_attrs(args, res);
8010         if (ret)
8011                 return ret;
8012         return nfs4_verify_back_channel_attrs(args, res);
8013 }
8014
8015 static void nfs4_update_session(struct nfs4_session *session,
8016                 struct nfs41_create_session_res *res)
8017 {
8018         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8019         /* Mark client id and session as being confirmed */
8020         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8021         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8022         session->flags = res->flags;
8023         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8024         if (res->flags & SESSION4_BACK_CHAN)
8025                 memcpy(&session->bc_attrs, &res->bc_attrs,
8026                                 sizeof(session->bc_attrs));
8027 }
8028
8029 static int _nfs4_proc_create_session(struct nfs_client *clp,
8030                 struct rpc_cred *cred)
8031 {
8032         struct nfs4_session *session = clp->cl_session;
8033         struct nfs41_create_session_args args = {
8034                 .client = clp,
8035                 .clientid = clp->cl_clientid,
8036                 .seqid = clp->cl_seqid,
8037                 .cb_program = NFS4_CALLBACK,
8038         };
8039         struct nfs41_create_session_res res;
8040
8041         struct rpc_message msg = {
8042                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8043                 .rpc_argp = &args,
8044                 .rpc_resp = &res,
8045                 .rpc_cred = cred,
8046         };
8047         int status;
8048
8049         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8050         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8051
8052         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8053         trace_nfs4_create_session(clp, status);
8054
8055         switch (status) {
8056         case -NFS4ERR_STALE_CLIENTID:
8057         case -NFS4ERR_DELAY:
8058         case -ETIMEDOUT:
8059         case -EACCES:
8060         case -EAGAIN:
8061                 goto out;
8062         };
8063
8064         clp->cl_seqid++;
8065         if (!status) {
8066                 /* Verify the session's negotiated channel_attrs values */
8067                 status = nfs4_verify_channel_attrs(&args, &res);
8068                 /* Increment the clientid slot sequence id */
8069                 if (status)
8070                         goto out;
8071                 nfs4_update_session(session, &res);
8072         }
8073 out:
8074         return status;
8075 }
8076
8077 /*
8078  * Issues a CREATE_SESSION operation to the server.
8079  * It is the responsibility of the caller to verify the session is
8080  * expired before calling this routine.
8081  */
8082 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8083 {
8084         int status;
8085         unsigned *ptr;
8086         struct nfs4_session *session = clp->cl_session;
8087
8088         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8089
8090         status = _nfs4_proc_create_session(clp, cred);
8091         if (status)
8092                 goto out;
8093
8094         /* Init or reset the session slot tables */
8095         status = nfs4_setup_session_slot_tables(session);
8096         dprintk("slot table setup returned %d\n", status);
8097         if (status)
8098                 goto out;
8099
8100         ptr = (unsigned *)&session->sess_id.data[0];
8101         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8102                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8103 out:
8104         dprintk("<-- %s\n", __func__);
8105         return status;
8106 }
8107
8108 /*
8109  * Issue the over-the-wire RPC DESTROY_SESSION.
8110  * The caller must serialize access to this routine.
8111  */
8112 int nfs4_proc_destroy_session(struct nfs4_session *session,
8113                 struct rpc_cred *cred)
8114 {
8115         struct rpc_message msg = {
8116                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8117                 .rpc_argp = session,
8118                 .rpc_cred = cred,
8119         };
8120         int status = 0;
8121
8122         dprintk("--> nfs4_proc_destroy_session\n");
8123
8124         /* session is still being setup */
8125         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8126                 return 0;
8127
8128         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8129         trace_nfs4_destroy_session(session->clp, status);
8130
8131         if (status)
8132                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8133                         "Session has been destroyed regardless...\n", status);
8134
8135         dprintk("<-- nfs4_proc_destroy_session\n");
8136         return status;
8137 }
8138
8139 /*
8140  * Renew the cl_session lease.
8141  */
8142 struct nfs4_sequence_data {
8143         struct nfs_client *clp;
8144         struct nfs4_sequence_args args;
8145         struct nfs4_sequence_res res;
8146 };
8147
8148 static void nfs41_sequence_release(void *data)
8149 {
8150         struct nfs4_sequence_data *calldata = data;
8151         struct nfs_client *clp = calldata->clp;
8152
8153         if (atomic_read(&clp->cl_count) > 1)
8154                 nfs4_schedule_state_renewal(clp);
8155         nfs_put_client(clp);
8156         kfree(calldata);
8157 }
8158
8159 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8160 {
8161         switch(task->tk_status) {
8162         case -NFS4ERR_DELAY:
8163                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8164                 return -EAGAIN;
8165         default:
8166                 nfs4_schedule_lease_recovery(clp);
8167         }
8168         return 0;
8169 }
8170
8171 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8172 {
8173         struct nfs4_sequence_data *calldata = data;
8174         struct nfs_client *clp = calldata->clp;
8175
8176         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8177                 return;
8178
8179         trace_nfs4_sequence(clp, task->tk_status);
8180         if (task->tk_status < 0) {
8181                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8182                 if (atomic_read(&clp->cl_count) == 1)
8183                         goto out;
8184
8185                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8186                         rpc_restart_call_prepare(task);
8187                         return;
8188                 }
8189         }
8190         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8191 out:
8192         dprintk("<-- %s\n", __func__);
8193 }
8194
8195 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8196 {
8197         struct nfs4_sequence_data *calldata = data;
8198         struct nfs_client *clp = calldata->clp;
8199         struct nfs4_sequence_args *args;
8200         struct nfs4_sequence_res *res;
8201
8202         args = task->tk_msg.rpc_argp;
8203         res = task->tk_msg.rpc_resp;
8204
8205         nfs41_setup_sequence(clp->cl_session, args, res, task);
8206 }
8207
8208 static const struct rpc_call_ops nfs41_sequence_ops = {
8209         .rpc_call_done = nfs41_sequence_call_done,
8210         .rpc_call_prepare = nfs41_sequence_prepare,
8211         .rpc_release = nfs41_sequence_release,
8212 };
8213
8214 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8215                 struct rpc_cred *cred,
8216                 bool is_privileged)
8217 {
8218         struct nfs4_sequence_data *calldata;
8219         struct rpc_message msg = {
8220                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8221                 .rpc_cred = cred,
8222         };
8223         struct rpc_task_setup task_setup_data = {
8224                 .rpc_client = clp->cl_rpcclient,
8225                 .rpc_message = &msg,
8226                 .callback_ops = &nfs41_sequence_ops,
8227                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8228         };
8229
8230         if (!atomic_inc_not_zero(&clp->cl_count))
8231                 return ERR_PTR(-EIO);
8232         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8233         if (calldata == NULL) {
8234                 nfs_put_client(clp);
8235                 return ERR_PTR(-ENOMEM);
8236         }
8237         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8238         if (is_privileged)
8239                 nfs4_set_sequence_privileged(&calldata->args);
8240         msg.rpc_argp = &calldata->args;
8241         msg.rpc_resp = &calldata->res;
8242         calldata->clp = clp;
8243         task_setup_data.callback_data = calldata;
8244
8245         return rpc_run_task(&task_setup_data);
8246 }
8247
8248 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8249 {
8250         struct rpc_task *task;
8251         int ret = 0;
8252
8253         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8254                 return -EAGAIN;
8255         task = _nfs41_proc_sequence(clp, cred, false);
8256         if (IS_ERR(task))
8257                 ret = PTR_ERR(task);
8258         else
8259                 rpc_put_task_async(task);
8260         dprintk("<-- %s status=%d\n", __func__, ret);
8261         return ret;
8262 }
8263
8264 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8265 {
8266         struct rpc_task *task;
8267         int ret;
8268
8269         task = _nfs41_proc_sequence(clp, cred, true);
8270         if (IS_ERR(task)) {
8271                 ret = PTR_ERR(task);
8272                 goto out;
8273         }
8274         ret = rpc_wait_for_completion_task(task);
8275         if (!ret)
8276                 ret = task->tk_status;
8277         rpc_put_task(task);
8278 out:
8279         dprintk("<-- %s status=%d\n", __func__, ret);
8280         return ret;
8281 }
8282
8283 struct nfs4_reclaim_complete_data {
8284         struct nfs_client *clp;
8285         struct nfs41_reclaim_complete_args arg;
8286         struct nfs41_reclaim_complete_res res;
8287 };
8288
8289 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8290 {
8291         struct nfs4_reclaim_complete_data *calldata = data;
8292
8293         nfs41_setup_sequence(calldata->clp->cl_session,
8294                         &calldata->arg.seq_args,
8295                         &calldata->res.seq_res,
8296                         task);
8297 }
8298
8299 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8300 {
8301         switch(task->tk_status) {
8302         case 0:
8303         case -NFS4ERR_COMPLETE_ALREADY:
8304         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8305                 break;
8306         case -NFS4ERR_DELAY:
8307                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8308                 /* fall through */
8309         case -NFS4ERR_RETRY_UNCACHED_REP:
8310                 return -EAGAIN;
8311         default:
8312                 nfs4_schedule_lease_recovery(clp);
8313         }
8314         return 0;
8315 }
8316
8317 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8318 {
8319         struct nfs4_reclaim_complete_data *calldata = data;
8320         struct nfs_client *clp = calldata->clp;
8321         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8322
8323         dprintk("--> %s\n", __func__);
8324         if (!nfs41_sequence_done(task, res))
8325                 return;
8326
8327         trace_nfs4_reclaim_complete(clp, task->tk_status);
8328         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8329                 rpc_restart_call_prepare(task);
8330                 return;
8331         }
8332         dprintk("<-- %s\n", __func__);
8333 }
8334
8335 static void nfs4_free_reclaim_complete_data(void *data)
8336 {
8337         struct nfs4_reclaim_complete_data *calldata = data;
8338
8339         kfree(calldata);
8340 }
8341
8342 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8343         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8344         .rpc_call_done = nfs4_reclaim_complete_done,
8345         .rpc_release = nfs4_free_reclaim_complete_data,
8346 };
8347
8348 /*
8349  * Issue a global reclaim complete.
8350  */
8351 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8352                 struct rpc_cred *cred)
8353 {
8354         struct nfs4_reclaim_complete_data *calldata;
8355         struct rpc_task *task;
8356         struct rpc_message msg = {
8357                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8358                 .rpc_cred = cred,
8359         };
8360         struct rpc_task_setup task_setup_data = {
8361                 .rpc_client = clp->cl_rpcclient,
8362                 .rpc_message = &msg,
8363                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8364                 .flags = RPC_TASK_ASYNC,
8365         };
8366         int status = -ENOMEM;
8367
8368         dprintk("--> %s\n", __func__);
8369         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8370         if (calldata == NULL)
8371                 goto out;
8372         calldata->clp = clp;
8373         calldata->arg.one_fs = 0;
8374
8375         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8376         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8377         msg.rpc_argp = &calldata->arg;
8378         msg.rpc_resp = &calldata->res;
8379         task_setup_data.callback_data = calldata;
8380         task = rpc_run_task(&task_setup_data);
8381         if (IS_ERR(task)) {
8382                 status = PTR_ERR(task);
8383                 goto out;
8384         }
8385         status = nfs4_wait_for_completion_rpc_task(task);
8386         if (status == 0)
8387                 status = task->tk_status;
8388         rpc_put_task(task);
8389         return 0;
8390 out:
8391         dprintk("<-- %s status=%d\n", __func__, status);
8392         return status;
8393 }
8394
8395 static void
8396 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8397 {
8398         struct nfs4_layoutget *lgp = calldata;
8399         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8400         struct nfs4_session *session = nfs4_get_session(server);
8401
8402         dprintk("--> %s\n", __func__);
8403         nfs41_setup_sequence(session, &lgp->args.seq_args,
8404                                 &lgp->res.seq_res, task);
8405         dprintk("<-- %s\n", __func__);
8406 }
8407
8408 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8409 {
8410         struct nfs4_layoutget *lgp = calldata;
8411
8412         dprintk("--> %s\n", __func__);
8413         nfs41_sequence_process(task, &lgp->res.seq_res);
8414         dprintk("<-- %s\n", __func__);
8415 }
8416
8417 static int
8418 nfs4_layoutget_handle_exception(struct rpc_task *task,
8419                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8420 {
8421         struct inode *inode = lgp->args.inode;
8422         struct nfs_server *server = NFS_SERVER(inode);
8423         struct pnfs_layout_hdr *lo;
8424         int nfs4err = task->tk_status;
8425         int err, status = 0;
8426         LIST_HEAD(head);
8427
8428         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8429
8430         switch (nfs4err) {
8431         case 0:
8432                 goto out;
8433
8434         /*
8435          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8436          * on the file. set tk_status to -ENODATA to tell upper layer to
8437          * retry go inband.
8438          */
8439         case -NFS4ERR_LAYOUTUNAVAILABLE:
8440                 status = -ENODATA;
8441                 goto out;
8442         /*
8443          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8444          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8445          */
8446         case -NFS4ERR_BADLAYOUT:
8447                 status = -EOVERFLOW;
8448                 goto out;
8449         /*
8450          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8451          * (or clients) writing to the same RAID stripe except when
8452          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8453          *
8454          * Treat it like we would RECALLCONFLICT -- we retry for a little
8455          * while, and then eventually give up.
8456          */
8457         case -NFS4ERR_LAYOUTTRYLATER:
8458                 if (lgp->args.minlength == 0) {
8459                         status = -EOVERFLOW;
8460                         goto out;
8461                 }
8462                 status = -EBUSY;
8463                 break;
8464         case -NFS4ERR_RECALLCONFLICT:
8465                 status = -ERECALLCONFLICT;
8466                 break;
8467         case -NFS4ERR_DELEG_REVOKED:
8468         case -NFS4ERR_ADMIN_REVOKED:
8469         case -NFS4ERR_EXPIRED:
8470         case -NFS4ERR_BAD_STATEID:
8471                 exception->timeout = 0;
8472                 spin_lock(&inode->i_lock);
8473                 lo = NFS_I(inode)->layout;
8474                 /* If the open stateid was bad, then recover it. */
8475                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8476                     nfs4_stateid_match_other(&lgp->args.stateid,
8477                                         &lgp->args.ctx->state->stateid)) {
8478                         spin_unlock(&inode->i_lock);
8479                         exception->state = lgp->args.ctx->state;
8480                         exception->stateid = &lgp->args.stateid;
8481                         break;
8482                 }
8483
8484                 /*
8485                  * Mark the bad layout state as invalid, then retry
8486                  */
8487                 pnfs_mark_layout_stateid_invalid(lo, &head);
8488                 spin_unlock(&inode->i_lock);
8489                 pnfs_free_lseg_list(&head);
8490                 status = -EAGAIN;
8491                 goto out;
8492         }
8493
8494         nfs4_sequence_free_slot(&lgp->res.seq_res);
8495         err = nfs4_handle_exception(server, nfs4err, exception);
8496         if (!status) {
8497                 if (exception->retry)
8498                         status = -EAGAIN;
8499                 else
8500                         status = err;
8501         }
8502 out:
8503         dprintk("<-- %s\n", __func__);
8504         return status;
8505 }
8506
8507 static size_t max_response_pages(struct nfs_server *server)
8508 {
8509         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8510         return nfs_page_array_len(0, max_resp_sz);
8511 }
8512
8513 static void nfs4_free_pages(struct page **pages, size_t size)
8514 {
8515         int i;
8516
8517         if (!pages)
8518                 return;
8519
8520         for (i = 0; i < size; i++) {
8521                 if (!pages[i])
8522                         break;
8523                 __free_page(pages[i]);
8524         }
8525         kfree(pages);
8526 }
8527
8528 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8529 {
8530         struct page **pages;
8531         int i;
8532
8533         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8534         if (!pages) {
8535                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8536                 return NULL;
8537         }
8538
8539         for (i = 0; i < size; i++) {
8540                 pages[i] = alloc_page(gfp_flags);
8541                 if (!pages[i]) {
8542                         dprintk("%s: failed to allocate page\n", __func__);
8543                         nfs4_free_pages(pages, size);
8544                         return NULL;
8545                 }
8546         }
8547
8548         return pages;
8549 }
8550
8551 static void nfs4_layoutget_release(void *calldata)
8552 {
8553         struct nfs4_layoutget *lgp = calldata;
8554         struct inode *inode = lgp->args.inode;
8555         struct nfs_server *server = NFS_SERVER(inode);
8556         size_t max_pages = max_response_pages(server);
8557
8558         dprintk("--> %s\n", __func__);
8559         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8560         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8561         put_nfs_open_context(lgp->args.ctx);
8562         kfree(calldata);
8563         dprintk("<-- %s\n", __func__);
8564 }
8565
8566 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8567         .rpc_call_prepare = nfs4_layoutget_prepare,
8568         .rpc_call_done = nfs4_layoutget_done,
8569         .rpc_release = nfs4_layoutget_release,
8570 };
8571
8572 struct pnfs_layout_segment *
8573 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8574 {
8575         struct inode *inode = lgp->args.inode;
8576         struct nfs_server *server = NFS_SERVER(inode);
8577         size_t max_pages = max_response_pages(server);
8578         struct rpc_task *task;
8579         struct rpc_message msg = {
8580                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8581                 .rpc_argp = &lgp->args,
8582                 .rpc_resp = &lgp->res,
8583                 .rpc_cred = lgp->cred,
8584         };
8585         struct rpc_task_setup task_setup_data = {
8586                 .rpc_client = server->client,
8587                 .rpc_message = &msg,
8588                 .callback_ops = &nfs4_layoutget_call_ops,
8589                 .callback_data = lgp,
8590                 .flags = RPC_TASK_ASYNC,
8591         };
8592         struct pnfs_layout_segment *lseg = NULL;
8593         struct nfs4_exception exception = {
8594                 .inode = inode,
8595                 .timeout = *timeout,
8596         };
8597         int status = 0;
8598
8599         dprintk("--> %s\n", __func__);
8600
8601         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8602         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8603
8604         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8605         if (!lgp->args.layout.pages) {
8606                 nfs4_layoutget_release(lgp);
8607                 return ERR_PTR(-ENOMEM);
8608         }
8609         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8610
8611         lgp->res.layoutp = &lgp->args.layout;
8612         lgp->res.seq_res.sr_slot = NULL;
8613         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8614
8615         task = rpc_run_task(&task_setup_data);
8616         if (IS_ERR(task))
8617                 return ERR_CAST(task);
8618         status = nfs4_wait_for_completion_rpc_task(task);
8619         if (status == 0) {
8620                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8621                 *timeout = exception.timeout;
8622         }
8623
8624         trace_nfs4_layoutget(lgp->args.ctx,
8625                         &lgp->args.range,
8626                         &lgp->res.range,
8627                         &lgp->res.stateid,
8628                         status);
8629
8630         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8631         if (status == 0 && lgp->res.layoutp->len)
8632                 lseg = pnfs_layout_process(lgp);
8633         nfs4_sequence_free_slot(&lgp->res.seq_res);
8634         rpc_put_task(task);
8635         dprintk("<-- %s status=%d\n", __func__, status);
8636         if (status)
8637                 return ERR_PTR(status);
8638         return lseg;
8639 }
8640
8641 static void
8642 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8643 {
8644         struct nfs4_layoutreturn *lrp = calldata;
8645
8646         dprintk("--> %s\n", __func__);
8647         nfs41_setup_sequence(lrp->clp->cl_session,
8648                         &lrp->args.seq_args,
8649                         &lrp->res.seq_res,
8650                         task);
8651 }
8652
8653 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8654 {
8655         struct nfs4_layoutreturn *lrp = calldata;
8656         struct nfs_server *server;
8657
8658         dprintk("--> %s\n", __func__);
8659
8660         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8661                 return;
8662
8663         server = NFS_SERVER(lrp->args.inode);
8664         switch (task->tk_status) {
8665         default:
8666                 task->tk_status = 0;
8667         case 0:
8668                 break;
8669         case -NFS4ERR_DELAY:
8670                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8671                         break;
8672                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8673                 rpc_restart_call_prepare(task);
8674                 return;
8675         }
8676         dprintk("<-- %s\n", __func__);
8677 }
8678
8679 static void nfs4_layoutreturn_release(void *calldata)
8680 {
8681         struct nfs4_layoutreturn *lrp = calldata;
8682         struct pnfs_layout_hdr *lo = lrp->args.layout;
8683
8684         dprintk("--> %s\n", __func__);
8685         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8686                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8687         nfs4_sequence_free_slot(&lrp->res.seq_res);
8688         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8689                 lrp->ld_private.ops->free(&lrp->ld_private);
8690         pnfs_put_layout_hdr(lrp->args.layout);
8691         nfs_iput_and_deactive(lrp->inode);
8692         kfree(calldata);
8693         dprintk("<-- %s\n", __func__);
8694 }
8695
8696 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8697         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8698         .rpc_call_done = nfs4_layoutreturn_done,
8699         .rpc_release = nfs4_layoutreturn_release,
8700 };
8701
8702 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8703 {
8704         struct rpc_task *task;
8705         struct rpc_message msg = {
8706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8707                 .rpc_argp = &lrp->args,
8708                 .rpc_resp = &lrp->res,
8709                 .rpc_cred = lrp->cred,
8710         };
8711         struct rpc_task_setup task_setup_data = {
8712                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8713                 .rpc_message = &msg,
8714                 .callback_ops = &nfs4_layoutreturn_call_ops,
8715                 .callback_data = lrp,
8716         };
8717         int status = 0;
8718
8719         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8720                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8721                         &task_setup_data.rpc_client, &msg);
8722
8723         dprintk("--> %s\n", __func__);
8724         if (!sync) {
8725                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8726                 if (!lrp->inode) {
8727                         nfs4_layoutreturn_release(lrp);
8728                         return -EAGAIN;
8729                 }
8730                 task_setup_data.flags |= RPC_TASK_ASYNC;
8731         }
8732         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8733         task = rpc_run_task(&task_setup_data);
8734         if (IS_ERR(task))
8735                 return PTR_ERR(task);
8736         if (sync)
8737                 status = task->tk_status;
8738         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8739         dprintk("<-- %s status=%d\n", __func__, status);
8740         rpc_put_task(task);
8741         return status;
8742 }
8743
8744 static int
8745 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8746                 struct pnfs_device *pdev,
8747                 struct rpc_cred *cred)
8748 {
8749         struct nfs4_getdeviceinfo_args args = {
8750                 .pdev = pdev,
8751                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8752                         NOTIFY_DEVICEID4_DELETE,
8753         };
8754         struct nfs4_getdeviceinfo_res res = {
8755                 .pdev = pdev,
8756         };
8757         struct rpc_message msg = {
8758                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8759                 .rpc_argp = &args,
8760                 .rpc_resp = &res,
8761                 .rpc_cred = cred,
8762         };
8763         int status;
8764
8765         dprintk("--> %s\n", __func__);
8766         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8767         if (res.notification & ~args.notify_types)
8768                 dprintk("%s: unsupported notification\n", __func__);
8769         if (res.notification != args.notify_types)
8770                 pdev->nocache = 1;
8771
8772         dprintk("<-- %s status=%d\n", __func__, status);
8773
8774         return status;
8775 }
8776
8777 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8778                 struct pnfs_device *pdev,
8779                 struct rpc_cred *cred)
8780 {
8781         struct nfs4_exception exception = { };
8782         int err;
8783
8784         do {
8785                 err = nfs4_handle_exception(server,
8786                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8787                                         &exception);
8788         } while (exception.retry);
8789         return err;
8790 }
8791 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8792
8793 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8794 {
8795         struct nfs4_layoutcommit_data *data = calldata;
8796         struct nfs_server *server = NFS_SERVER(data->args.inode);
8797         struct nfs4_session *session = nfs4_get_session(server);
8798
8799         nfs41_setup_sequence(session,
8800                         &data->args.seq_args,
8801                         &data->res.seq_res,
8802                         task);
8803 }
8804
8805 static void
8806 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8807 {
8808         struct nfs4_layoutcommit_data *data = calldata;
8809         struct nfs_server *server = NFS_SERVER(data->args.inode);
8810
8811         if (!nfs41_sequence_done(task, &data->res.seq_res))
8812                 return;
8813
8814         switch (task->tk_status) { /* Just ignore these failures */
8815         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8816         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8817         case -NFS4ERR_BADLAYOUT:     /* no layout */
8818         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8819                 task->tk_status = 0;
8820         case 0:
8821                 break;
8822         default:
8823                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8824                         rpc_restart_call_prepare(task);
8825                         return;
8826                 }
8827         }
8828 }
8829
8830 static void nfs4_layoutcommit_release(void *calldata)
8831 {
8832         struct nfs4_layoutcommit_data *data = calldata;
8833
8834         pnfs_cleanup_layoutcommit(data);
8835         nfs_post_op_update_inode_force_wcc(data->args.inode,
8836                                            data->res.fattr);
8837         put_rpccred(data->cred);
8838         nfs_iput_and_deactive(data->inode);
8839         kfree(data);
8840 }
8841
8842 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8843         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8844         .rpc_call_done = nfs4_layoutcommit_done,
8845         .rpc_release = nfs4_layoutcommit_release,
8846 };
8847
8848 int
8849 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8850 {
8851         struct rpc_message msg = {
8852                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8853                 .rpc_argp = &data->args,
8854                 .rpc_resp = &data->res,
8855                 .rpc_cred = data->cred,
8856         };
8857         struct rpc_task_setup task_setup_data = {
8858                 .task = &data->task,
8859                 .rpc_client = NFS_CLIENT(data->args.inode),
8860                 .rpc_message = &msg,
8861                 .callback_ops = &nfs4_layoutcommit_ops,
8862                 .callback_data = data,
8863         };
8864         struct rpc_task *task;
8865         int status = 0;
8866
8867         dprintk("NFS: initiating layoutcommit call. sync %d "
8868                 "lbw: %llu inode %lu\n", sync,
8869                 data->args.lastbytewritten,
8870                 data->args.inode->i_ino);
8871
8872         if (!sync) {
8873                 data->inode = nfs_igrab_and_active(data->args.inode);
8874                 if (data->inode == NULL) {
8875                         nfs4_layoutcommit_release(data);
8876                         return -EAGAIN;
8877                 }
8878                 task_setup_data.flags = RPC_TASK_ASYNC;
8879         }
8880         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8881         task = rpc_run_task(&task_setup_data);
8882         if (IS_ERR(task))
8883                 return PTR_ERR(task);
8884         if (sync)
8885                 status = task->tk_status;
8886         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8887         dprintk("%s: status %d\n", __func__, status);
8888         rpc_put_task(task);
8889         return status;
8890 }
8891
8892 /**
8893  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8894  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8895  */
8896 static int
8897 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8898                     struct nfs_fsinfo *info,
8899                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8900 {
8901         struct nfs41_secinfo_no_name_args args = {
8902                 .style = SECINFO_STYLE_CURRENT_FH,
8903         };
8904         struct nfs4_secinfo_res res = {
8905                 .flavors = flavors,
8906         };
8907         struct rpc_message msg = {
8908                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8909                 .rpc_argp = &args,
8910                 .rpc_resp = &res,
8911         };
8912         struct rpc_clnt *clnt = server->client;
8913         struct rpc_cred *cred = NULL;
8914         int status;
8915
8916         if (use_integrity) {
8917                 clnt = server->nfs_client->cl_rpcclient;
8918                 cred = nfs4_get_clid_cred(server->nfs_client);
8919                 msg.rpc_cred = cred;
8920         }
8921
8922         dprintk("--> %s\n", __func__);
8923         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8924                                 &res.seq_res, 0);
8925         dprintk("<-- %s status=%d\n", __func__, status);
8926
8927         if (cred)
8928                 put_rpccred(cred);
8929
8930         return status;
8931 }
8932
8933 static int
8934 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8935                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8936 {
8937         struct nfs4_exception exception = { };
8938         int err;
8939         do {
8940                 /* first try using integrity protection */
8941                 err = -NFS4ERR_WRONGSEC;
8942
8943                 /* try to use integrity protection with machine cred */
8944                 if (_nfs4_is_integrity_protected(server->nfs_client))
8945                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8946                                                           flavors, true);
8947
8948                 /*
8949                  * if unable to use integrity protection, or SECINFO with
8950                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8951                  * disallowed by spec, but exists in deployed servers) use
8952                  * the current filesystem's rpc_client and the user cred.
8953                  */
8954                 if (err == -NFS4ERR_WRONGSEC)
8955                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8956                                                           flavors, false);
8957
8958                 switch (err) {
8959                 case 0:
8960                 case -NFS4ERR_WRONGSEC:
8961                 case -ENOTSUPP:
8962                         goto out;
8963                 default:
8964                         err = nfs4_handle_exception(server, err, &exception);
8965                 }
8966         } while (exception.retry);
8967 out:
8968         return err;
8969 }
8970
8971 static int
8972 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8973                     struct nfs_fsinfo *info)
8974 {
8975         int err;
8976         struct page *page;
8977         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8978         struct nfs4_secinfo_flavors *flavors;
8979         struct nfs4_secinfo4 *secinfo;
8980         int i;
8981
8982         page = alloc_page(GFP_KERNEL);
8983         if (!page) {
8984                 err = -ENOMEM;
8985                 goto out;
8986         }
8987
8988         flavors = page_address(page);
8989         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8990
8991         /*
8992          * Fall back on "guess and check" method if
8993          * the server doesn't support SECINFO_NO_NAME
8994          */
8995         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8996                 err = nfs4_find_root_sec(server, fhandle, info);
8997                 goto out_freepage;
8998         }
8999         if (err)
9000                 goto out_freepage;
9001
9002         for (i = 0; i < flavors->num_flavors; i++) {
9003                 secinfo = &flavors->flavors[i];
9004
9005                 switch (secinfo->flavor) {
9006                 case RPC_AUTH_NULL:
9007                 case RPC_AUTH_UNIX:
9008                 case RPC_AUTH_GSS:
9009                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9010                                         &secinfo->flavor_info);
9011                         break;
9012                 default:
9013                         flavor = RPC_AUTH_MAXFLAVOR;
9014                         break;
9015                 }
9016
9017                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9018                         flavor = RPC_AUTH_MAXFLAVOR;
9019
9020                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9021                         err = nfs4_lookup_root_sec(server, fhandle,
9022                                                    info, flavor);
9023                         if (!err)
9024                                 break;
9025                 }
9026         }
9027
9028         if (flavor == RPC_AUTH_MAXFLAVOR)
9029                 err = -EPERM;
9030
9031 out_freepage:
9032         put_page(page);
9033         if (err == -EACCES)
9034                 return -EPERM;
9035 out:
9036         return err;
9037 }
9038
9039 static int _nfs41_test_stateid(struct nfs_server *server,
9040                 nfs4_stateid *stateid,
9041                 struct rpc_cred *cred)
9042 {
9043         int status;
9044         struct nfs41_test_stateid_args args = {
9045                 .stateid = stateid,
9046         };
9047         struct nfs41_test_stateid_res res;
9048         struct rpc_message msg = {
9049                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9050                 .rpc_argp = &args,
9051                 .rpc_resp = &res,
9052                 .rpc_cred = cred,
9053         };
9054         struct rpc_clnt *rpc_client = server->client;
9055
9056         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9057                 &rpc_client, &msg);
9058
9059         dprintk("NFS call  test_stateid %p\n", stateid);
9060         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
9061         nfs4_set_sequence_privileged(&args.seq_args);
9062         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9063                         &args.seq_args, &res.seq_res);
9064         if (status != NFS_OK) {
9065                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9066                 return status;
9067         }
9068         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9069         return -res.status;
9070 }
9071
9072 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9073                 int err, struct nfs4_exception *exception)
9074 {
9075         exception->retry = 0;
9076         switch(err) {
9077         case -NFS4ERR_DELAY:
9078         case -NFS4ERR_RETRY_UNCACHED_REP:
9079                 nfs4_handle_exception(server, err, exception);
9080                 break;
9081         case -NFS4ERR_BADSESSION:
9082         case -NFS4ERR_BADSLOT:
9083         case -NFS4ERR_BAD_HIGH_SLOT:
9084         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9085         case -NFS4ERR_DEADSESSION:
9086                 nfs4_do_handle_exception(server, err, exception);
9087         }
9088 }
9089
9090 /**
9091  * nfs41_test_stateid - perform a TEST_STATEID operation
9092  *
9093  * @server: server / transport on which to perform the operation
9094  * @stateid: state ID to test
9095  * @cred: credential
9096  *
9097  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9098  * Otherwise a negative NFS4ERR value is returned if the operation
9099  * failed or the state ID is not currently valid.
9100  */
9101 static int nfs41_test_stateid(struct nfs_server *server,
9102                 nfs4_stateid *stateid,
9103                 struct rpc_cred *cred)
9104 {
9105         struct nfs4_exception exception = { };
9106         int err;
9107         do {
9108                 err = _nfs41_test_stateid(server, stateid, cred);
9109                 nfs4_handle_delay_or_session_error(server, err, &exception);
9110         } while (exception.retry);
9111         return err;
9112 }
9113
9114 struct nfs_free_stateid_data {
9115         struct nfs_server *server;
9116         struct nfs41_free_stateid_args args;
9117         struct nfs41_free_stateid_res res;
9118 };
9119
9120 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9121 {
9122         struct nfs_free_stateid_data *data = calldata;
9123         nfs41_setup_sequence(nfs4_get_session(data->server),
9124                         &data->args.seq_args,
9125                         &data->res.seq_res,
9126                         task);
9127 }
9128
9129 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9130 {
9131         struct nfs_free_stateid_data *data = calldata;
9132
9133         nfs41_sequence_done(task, &data->res.seq_res);
9134
9135         switch (task->tk_status) {
9136         case -NFS4ERR_DELAY:
9137                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9138                         rpc_restart_call_prepare(task);
9139         }
9140 }
9141
9142 static void nfs41_free_stateid_release(void *calldata)
9143 {
9144         kfree(calldata);
9145 }
9146
9147 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9148         .rpc_call_prepare = nfs41_free_stateid_prepare,
9149         .rpc_call_done = nfs41_free_stateid_done,
9150         .rpc_release = nfs41_free_stateid_release,
9151 };
9152
9153 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9154                 const nfs4_stateid *stateid,
9155                 struct rpc_cred *cred,
9156                 bool privileged)
9157 {
9158         struct rpc_message msg = {
9159                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9160                 .rpc_cred = cred,
9161         };
9162         struct rpc_task_setup task_setup = {
9163                 .rpc_client = server->client,
9164                 .rpc_message = &msg,
9165                 .callback_ops = &nfs41_free_stateid_ops,
9166                 .flags = RPC_TASK_ASYNC,
9167         };
9168         struct nfs_free_stateid_data *data;
9169
9170         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9171                 &task_setup.rpc_client, &msg);
9172
9173         dprintk("NFS call  free_stateid %p\n", stateid);
9174         data = kmalloc(sizeof(*data), GFP_NOFS);
9175         if (!data)
9176                 return ERR_PTR(-ENOMEM);
9177         data->server = server;
9178         nfs4_stateid_copy(&data->args.stateid, stateid);
9179
9180         task_setup.callback_data = data;
9181
9182         msg.rpc_argp = &data->args;
9183         msg.rpc_resp = &data->res;
9184         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9185         if (privileged)
9186                 nfs4_set_sequence_privileged(&data->args.seq_args);
9187
9188         return rpc_run_task(&task_setup);
9189 }
9190
9191 /**
9192  * nfs41_free_stateid - perform a FREE_STATEID operation
9193  *
9194  * @server: server / transport on which to perform the operation
9195  * @stateid: state ID to release
9196  * @cred: credential
9197  * @is_recovery: set to true if this call needs to be privileged
9198  *
9199  * Note: this function is always asynchronous.
9200  */
9201 static int nfs41_free_stateid(struct nfs_server *server,
9202                 const nfs4_stateid *stateid,
9203                 struct rpc_cred *cred,
9204                 bool is_recovery)
9205 {
9206         struct rpc_task *task;
9207
9208         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9209         if (IS_ERR(task))
9210                 return PTR_ERR(task);
9211         rpc_put_task(task);
9212         return 0;
9213 }
9214
9215 static void
9216 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9217 {
9218         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9219
9220         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9221         nfs4_free_lock_state(server, lsp);
9222 }
9223
9224 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9225                 const nfs4_stateid *s2)
9226 {
9227         if (s1->type != s2->type)
9228                 return false;
9229
9230         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9231                 return false;
9232
9233         if (s1->seqid == s2->seqid)
9234                 return true;
9235         if (s1->seqid == 0 || s2->seqid == 0)
9236                 return true;
9237
9238         return false;
9239 }
9240
9241 #endif /* CONFIG_NFS_V4_1 */
9242
9243 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9244                 const nfs4_stateid *s2)
9245 {
9246         return nfs4_stateid_match(s1, s2);
9247 }
9248
9249
9250 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9251         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9252         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9253         .recover_open   = nfs4_open_reclaim,
9254         .recover_lock   = nfs4_lock_reclaim,
9255         .establish_clid = nfs4_init_clientid,
9256         .detect_trunking = nfs40_discover_server_trunking,
9257 };
9258
9259 #if defined(CONFIG_NFS_V4_1)
9260 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9261         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9262         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9263         .recover_open   = nfs4_open_reclaim,
9264         .recover_lock   = nfs4_lock_reclaim,
9265         .establish_clid = nfs41_init_clientid,
9266         .reclaim_complete = nfs41_proc_reclaim_complete,
9267         .detect_trunking = nfs41_discover_server_trunking,
9268 };
9269 #endif /* CONFIG_NFS_V4_1 */
9270
9271 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9272         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9273         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9274         .recover_open   = nfs40_open_expired,
9275         .recover_lock   = nfs4_lock_expired,
9276         .establish_clid = nfs4_init_clientid,
9277 };
9278
9279 #if defined(CONFIG_NFS_V4_1)
9280 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9281         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9282         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9283         .recover_open   = nfs41_open_expired,
9284         .recover_lock   = nfs41_lock_expired,
9285         .establish_clid = nfs41_init_clientid,
9286 };
9287 #endif /* CONFIG_NFS_V4_1 */
9288
9289 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9290         .sched_state_renewal = nfs4_proc_async_renew,
9291         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9292         .renew_lease = nfs4_proc_renew,
9293 };
9294
9295 #if defined(CONFIG_NFS_V4_1)
9296 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9297         .sched_state_renewal = nfs41_proc_async_sequence,
9298         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9299         .renew_lease = nfs4_proc_sequence,
9300 };
9301 #endif
9302
9303 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9304         .get_locations = _nfs40_proc_get_locations,
9305         .fsid_present = _nfs40_proc_fsid_present,
9306 };
9307
9308 #if defined(CONFIG_NFS_V4_1)
9309 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9310         .get_locations = _nfs41_proc_get_locations,
9311         .fsid_present = _nfs41_proc_fsid_present,
9312 };
9313 #endif  /* CONFIG_NFS_V4_1 */
9314
9315 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9316         .minor_version = 0,
9317         .init_caps = NFS_CAP_READDIRPLUS
9318                 | NFS_CAP_ATOMIC_OPEN
9319                 | NFS_CAP_POSIX_LOCK,
9320         .init_client = nfs40_init_client,
9321         .shutdown_client = nfs40_shutdown_client,
9322         .match_stateid = nfs4_match_stateid,
9323         .find_root_sec = nfs4_find_root_sec,
9324         .free_lock_state = nfs4_release_lockowner,
9325         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9326         .alloc_seqid = nfs_alloc_seqid,
9327         .call_sync_ops = &nfs40_call_sync_ops,
9328         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9329         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9330         .state_renewal_ops = &nfs40_state_renewal_ops,
9331         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9332 };
9333
9334 #if defined(CONFIG_NFS_V4_1)
9335 static struct nfs_seqid *
9336 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9337 {
9338         return NULL;
9339 }
9340
9341 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9342         .minor_version = 1,
9343         .init_caps = NFS_CAP_READDIRPLUS
9344                 | NFS_CAP_ATOMIC_OPEN
9345                 | NFS_CAP_POSIX_LOCK
9346                 | NFS_CAP_STATEID_NFSV41
9347                 | NFS_CAP_ATOMIC_OPEN_V1,
9348         .init_client = nfs41_init_client,
9349         .shutdown_client = nfs41_shutdown_client,
9350         .match_stateid = nfs41_match_stateid,
9351         .find_root_sec = nfs41_find_root_sec,
9352         .free_lock_state = nfs41_free_lock_state,
9353         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9354         .alloc_seqid = nfs_alloc_no_seqid,
9355         .session_trunk = nfs4_test_session_trunk,
9356         .call_sync_ops = &nfs41_call_sync_ops,
9357         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9358         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9359         .state_renewal_ops = &nfs41_state_renewal_ops,
9360         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9361 };
9362 #endif
9363
9364 #if defined(CONFIG_NFS_V4_2)
9365 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9366         .minor_version = 2,
9367         .init_caps = NFS_CAP_READDIRPLUS
9368                 | NFS_CAP_ATOMIC_OPEN
9369                 | NFS_CAP_POSIX_LOCK
9370                 | NFS_CAP_STATEID_NFSV41
9371                 | NFS_CAP_ATOMIC_OPEN_V1
9372                 | NFS_CAP_ALLOCATE
9373                 | NFS_CAP_COPY
9374                 | NFS_CAP_DEALLOCATE
9375                 | NFS_CAP_SEEK
9376                 | NFS_CAP_LAYOUTSTATS
9377                 | NFS_CAP_CLONE,
9378         .init_client = nfs41_init_client,
9379         .shutdown_client = nfs41_shutdown_client,
9380         .match_stateid = nfs41_match_stateid,
9381         .find_root_sec = nfs41_find_root_sec,
9382         .free_lock_state = nfs41_free_lock_state,
9383         .call_sync_ops = &nfs41_call_sync_ops,
9384         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9385         .alloc_seqid = nfs_alloc_no_seqid,
9386         .session_trunk = nfs4_test_session_trunk,
9387         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9388         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9389         .state_renewal_ops = &nfs41_state_renewal_ops,
9390         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9391 };
9392 #endif
9393
9394 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9395         [0] = &nfs_v4_0_minor_ops,
9396 #if defined(CONFIG_NFS_V4_1)
9397         [1] = &nfs_v4_1_minor_ops,
9398 #endif
9399 #if defined(CONFIG_NFS_V4_2)
9400         [2] = &nfs_v4_2_minor_ops,
9401 #endif
9402 };
9403
9404 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9405 {
9406         ssize_t error, error2;
9407
9408         error = generic_listxattr(dentry, list, size);
9409         if (error < 0)
9410                 return error;
9411         if (list) {
9412                 list += error;
9413                 size -= error;
9414         }
9415
9416         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9417         if (error2 < 0)
9418                 return error2;
9419         return error + error2;
9420 }
9421
9422 static const struct inode_operations nfs4_dir_inode_operations = {
9423         .create         = nfs_create,
9424         .lookup         = nfs_lookup,
9425         .atomic_open    = nfs_atomic_open,
9426         .link           = nfs_link,
9427         .unlink         = nfs_unlink,
9428         .symlink        = nfs_symlink,
9429         .mkdir          = nfs_mkdir,
9430         .rmdir          = nfs_rmdir,
9431         .mknod          = nfs_mknod,
9432         .rename         = nfs_rename,
9433         .permission     = nfs_permission,
9434         .getattr        = nfs_getattr,
9435         .setattr        = nfs_setattr,
9436         .listxattr      = nfs4_listxattr,
9437 };
9438
9439 static const struct inode_operations nfs4_file_inode_operations = {
9440         .permission     = nfs_permission,
9441         .getattr        = nfs_getattr,
9442         .setattr        = nfs_setattr,
9443         .listxattr      = nfs4_listxattr,
9444 };
9445
9446 const struct nfs_rpc_ops nfs_v4_clientops = {
9447         .version        = 4,                    /* protocol version */
9448         .dentry_ops     = &nfs4_dentry_operations,
9449         .dir_inode_ops  = &nfs4_dir_inode_operations,
9450         .file_inode_ops = &nfs4_file_inode_operations,
9451         .file_ops       = &nfs4_file_operations,
9452         .getroot        = nfs4_proc_get_root,
9453         .submount       = nfs4_submount,
9454         .try_mount      = nfs4_try_mount,
9455         .getattr        = nfs4_proc_getattr,
9456         .setattr        = nfs4_proc_setattr,
9457         .lookup         = nfs4_proc_lookup,
9458         .access         = nfs4_proc_access,
9459         .readlink       = nfs4_proc_readlink,
9460         .create         = nfs4_proc_create,
9461         .remove         = nfs4_proc_remove,
9462         .unlink_setup   = nfs4_proc_unlink_setup,
9463         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9464         .unlink_done    = nfs4_proc_unlink_done,
9465         .rename_setup   = nfs4_proc_rename_setup,
9466         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9467         .rename_done    = nfs4_proc_rename_done,
9468         .link           = nfs4_proc_link,
9469         .symlink        = nfs4_proc_symlink,
9470         .mkdir          = nfs4_proc_mkdir,
9471         .rmdir          = nfs4_proc_remove,
9472         .readdir        = nfs4_proc_readdir,
9473         .mknod          = nfs4_proc_mknod,
9474         .statfs         = nfs4_proc_statfs,
9475         .fsinfo         = nfs4_proc_fsinfo,
9476         .pathconf       = nfs4_proc_pathconf,
9477         .set_capabilities = nfs4_server_capabilities,
9478         .decode_dirent  = nfs4_decode_dirent,
9479         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9480         .read_setup     = nfs4_proc_read_setup,
9481         .read_done      = nfs4_read_done,
9482         .write_setup    = nfs4_proc_write_setup,
9483         .write_done     = nfs4_write_done,
9484         .commit_setup   = nfs4_proc_commit_setup,
9485         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9486         .commit_done    = nfs4_commit_done,
9487         .lock           = nfs4_proc_lock,
9488         .clear_acl_cache = nfs4_zap_acl_attr,
9489         .close_context  = nfs4_close_context,
9490         .open_context   = nfs4_atomic_open,
9491         .have_delegation = nfs4_have_delegation,
9492         .return_delegation = nfs4_inode_return_delegation,
9493         .alloc_client   = nfs4_alloc_client,
9494         .init_client    = nfs4_init_client,
9495         .free_client    = nfs4_free_client,
9496         .create_server  = nfs4_create_server,
9497         .clone_server   = nfs_clone_server,
9498 };
9499
9500 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9501         .name   = XATTR_NAME_NFSV4_ACL,
9502         .list   = nfs4_xattr_list_nfs4_acl,
9503         .get    = nfs4_xattr_get_nfs4_acl,
9504         .set    = nfs4_xattr_set_nfs4_acl,
9505 };
9506
9507 const struct xattr_handler *nfs4_xattr_handlers[] = {
9508         &nfs4_xattr_nfs4_acl_handler,
9509 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9510         &nfs4_xattr_nfs4_label_handler,
9511 #endif
9512         NULL
9513 };
9514
9515 /*
9516  * Local variables:
9517  *  c-basic-offset: 8
9518  * End:
9519  */