]> asedeno.scripts.mit.edu Git - linux.git/blob - fs/nfs/nfs4proc.c
PM / QoS: Remove global notifiers
[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 ((attrset[1] & FATTR4_WORD1_MODE))
2704                 sattr->ia_valid &= ~ATTR_MODE;
2705
2706         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2707                 *label = NULL;
2708 }
2709
2710 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2711                 fmode_t fmode,
2712                 int flags,
2713                 struct nfs_open_context *ctx)
2714 {
2715         struct nfs4_state_owner *sp = opendata->owner;
2716         struct nfs_server *server = sp->so_server;
2717         struct dentry *dentry;
2718         struct nfs4_state *state;
2719         unsigned int seq;
2720         int ret;
2721
2722         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2723
2724         ret = _nfs4_proc_open(opendata);
2725         if (ret != 0)
2726                 goto out;
2727
2728         state = nfs4_opendata_to_nfs4_state(opendata);
2729         ret = PTR_ERR(state);
2730         if (IS_ERR(state))
2731                 goto out;
2732         if (server->caps & NFS_CAP_POSIX_LOCK)
2733                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2734         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2735                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2736
2737         dentry = opendata->dentry;
2738         if (d_really_is_negative(dentry)) {
2739                 struct dentry *alias;
2740                 d_drop(dentry);
2741                 alias = d_exact_alias(dentry, state->inode);
2742                 if (!alias)
2743                         alias = d_splice_alias(igrab(state->inode), dentry);
2744                 /* d_splice_alias() can't fail here - it's a non-directory */
2745                 if (alias) {
2746                         dput(ctx->dentry);
2747                         ctx->dentry = dentry = alias;
2748                 }
2749                 nfs_set_verifier(dentry,
2750                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2751         }
2752
2753         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2754         if (ret != 0)
2755                 goto out;
2756
2757         ctx->state = state;
2758         if (d_inode(dentry) == state->inode) {
2759                 nfs_inode_attach_open_context(ctx);
2760                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2761                         nfs4_schedule_stateid_recovery(server, state);
2762         }
2763 out:
2764         return ret;
2765 }
2766
2767 /*
2768  * Returns a referenced nfs4_state
2769  */
2770 static int _nfs4_do_open(struct inode *dir,
2771                         struct nfs_open_context *ctx,
2772                         int flags,
2773                         struct iattr *sattr,
2774                         struct nfs4_label *label,
2775                         int *opened)
2776 {
2777         struct nfs4_state_owner  *sp;
2778         struct nfs4_state     *state = NULL;
2779         struct nfs_server       *server = NFS_SERVER(dir);
2780         struct nfs4_opendata *opendata;
2781         struct dentry *dentry = ctx->dentry;
2782         struct rpc_cred *cred = ctx->cred;
2783         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2784         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2785         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2786         struct nfs4_label *olabel = NULL;
2787         int status;
2788
2789         /* Protect against reboot recovery conflicts */
2790         status = -ENOMEM;
2791         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2792         if (sp == NULL) {
2793                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2794                 goto out_err;
2795         }
2796         status = nfs4_recover_expired_lease(server);
2797         if (status != 0)
2798                 goto err_put_state_owner;
2799         if (d_really_is_positive(dentry))
2800                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2801         status = -ENOMEM;
2802         if (d_really_is_positive(dentry))
2803                 claim = NFS4_OPEN_CLAIM_FH;
2804         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2805                         label, claim, GFP_KERNEL);
2806         if (opendata == NULL)
2807                 goto err_put_state_owner;
2808
2809         if (label) {
2810                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2811                 if (IS_ERR(olabel)) {
2812                         status = PTR_ERR(olabel);
2813                         goto err_opendata_put;
2814                 }
2815         }
2816
2817         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2818                 if (!opendata->f_attr.mdsthreshold) {
2819                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2820                         if (!opendata->f_attr.mdsthreshold)
2821                                 goto err_free_label;
2822                 }
2823                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2824         }
2825         if (d_really_is_positive(dentry))
2826                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2827
2828         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2829         if (status != 0)
2830                 goto err_free_label;
2831         state = ctx->state;
2832
2833         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2834             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2835                 nfs4_exclusive_attrset(opendata, sattr, &label);
2836                 /*
2837                  * send create attributes which was not set by open
2838                  * with an extra setattr.
2839                  */
2840                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2841                         nfs_fattr_init(opendata->o_res.f_attr);
2842                         status = nfs4_do_setattr(state->inode, cred,
2843                                         opendata->o_res.f_attr, sattr,
2844                                         ctx, label, olabel);
2845                         if (status == 0) {
2846                                 nfs_setattr_update_inode(state->inode, sattr,
2847                                                 opendata->o_res.f_attr);
2848                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2849                         }
2850                 }
2851         }
2852         if (opened && opendata->file_created)
2853                 *opened |= FILE_CREATED;
2854
2855         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2856                 *ctx_th = opendata->f_attr.mdsthreshold;
2857                 opendata->f_attr.mdsthreshold = NULL;
2858         }
2859
2860         nfs4_label_free(olabel);
2861
2862         nfs4_opendata_put(opendata);
2863         nfs4_put_state_owner(sp);
2864         return 0;
2865 err_free_label:
2866         nfs4_label_free(olabel);
2867 err_opendata_put:
2868         nfs4_opendata_put(opendata);
2869 err_put_state_owner:
2870         nfs4_put_state_owner(sp);
2871 out_err:
2872         return status;
2873 }
2874
2875
2876 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2877                                         struct nfs_open_context *ctx,
2878                                         int flags,
2879                                         struct iattr *sattr,
2880                                         struct nfs4_label *label,
2881                                         int *opened)
2882 {
2883         struct nfs_server *server = NFS_SERVER(dir);
2884         struct nfs4_exception exception = { };
2885         struct nfs4_state *res;
2886         int status;
2887
2888         do {
2889                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2890                 res = ctx->state;
2891                 trace_nfs4_open_file(ctx, flags, status);
2892                 if (status == 0)
2893                         break;
2894                 /* NOTE: BAD_SEQID means the server and client disagree about the
2895                  * book-keeping w.r.t. state-changing operations
2896                  * (OPEN/CLOSE/LOCK/LOCKU...)
2897                  * It is actually a sign of a bug on the client or on the server.
2898                  *
2899                  * If we receive a BAD_SEQID error in the particular case of
2900                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2901                  * have unhashed the old state_owner for us, and that we can
2902                  * therefore safely retry using a new one. We should still warn
2903                  * the user though...
2904                  */
2905                 if (status == -NFS4ERR_BAD_SEQID) {
2906                         pr_warn_ratelimited("NFS: v4 server %s "
2907                                         " returned a bad sequence-id error!\n",
2908                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2909                         exception.retry = 1;
2910                         continue;
2911                 }
2912                 /*
2913                  * BAD_STATEID on OPEN means that the server cancelled our
2914                  * state before it received the OPEN_CONFIRM.
2915                  * Recover by retrying the request as per the discussion
2916                  * on Page 181 of RFC3530.
2917                  */
2918                 if (status == -NFS4ERR_BAD_STATEID) {
2919                         exception.retry = 1;
2920                         continue;
2921                 }
2922                 if (status == -EAGAIN) {
2923                         /* We must have found a delegation */
2924                         exception.retry = 1;
2925                         continue;
2926                 }
2927                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2928                         continue;
2929                 res = ERR_PTR(nfs4_handle_exception(server,
2930                                         status, &exception));
2931         } while (exception.retry);
2932         return res;
2933 }
2934
2935 static int _nfs4_do_setattr(struct inode *inode,
2936                             struct nfs_setattrargs *arg,
2937                             struct nfs_setattrres *res,
2938                             struct rpc_cred *cred,
2939                             struct nfs_open_context *ctx)
2940 {
2941         struct nfs_server *server = NFS_SERVER(inode);
2942         struct rpc_message msg = {
2943                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2944                 .rpc_argp       = arg,
2945                 .rpc_resp       = res,
2946                 .rpc_cred       = cred,
2947         };
2948         struct rpc_cred *delegation_cred = NULL;
2949         unsigned long timestamp = jiffies;
2950         fmode_t fmode;
2951         bool truncate;
2952         int status;
2953
2954         nfs_fattr_init(res->fattr);
2955
2956         /* Servers should only apply open mode checks for file size changes */
2957         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2958         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2959
2960         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2961                 /* Use that stateid */
2962         } else if (truncate && ctx != NULL) {
2963                 struct nfs_lock_context *l_ctx;
2964                 if (!nfs4_valid_open_stateid(ctx->state))
2965                         return -EBADF;
2966                 l_ctx = nfs_get_lock_context(ctx);
2967                 if (IS_ERR(l_ctx))
2968                         return PTR_ERR(l_ctx);
2969                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2970                                                 &arg->stateid, &delegation_cred);
2971                 nfs_put_lock_context(l_ctx);
2972                 if (status == -EIO)
2973                         return -EBADF;
2974         } else
2975                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2976         if (delegation_cred)
2977                 msg.rpc_cred = delegation_cred;
2978
2979         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2980
2981         put_rpccred(delegation_cred);
2982         if (status == 0 && ctx != NULL)
2983                 renew_lease(server, timestamp);
2984         trace_nfs4_setattr(inode, &arg->stateid, status);
2985         return status;
2986 }
2987
2988 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2989                            struct nfs_fattr *fattr, struct iattr *sattr,
2990                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2991                            struct nfs4_label *olabel)
2992 {
2993         struct nfs_server *server = NFS_SERVER(inode);
2994         struct nfs4_state *state = ctx ? ctx->state : NULL;
2995         struct nfs_setattrargs  arg = {
2996                 .fh             = NFS_FH(inode),
2997                 .iap            = sattr,
2998                 .server         = server,
2999                 .bitmask = server->attr_bitmask,
3000                 .label          = ilabel,
3001         };
3002         struct nfs_setattrres  res = {
3003                 .fattr          = fattr,
3004                 .label          = olabel,
3005                 .server         = server,
3006         };
3007         struct nfs4_exception exception = {
3008                 .state = state,
3009                 .inode = inode,
3010                 .stateid = &arg.stateid,
3011         };
3012         int err;
3013
3014         arg.bitmask = nfs4_bitmask(server, ilabel);
3015         if (ilabel)
3016                 arg.bitmask = nfs4_bitmask(server, olabel);
3017
3018         do {
3019                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3020                 switch (err) {
3021                 case -NFS4ERR_OPENMODE:
3022                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3023                                 pr_warn_once("NFSv4: server %s is incorrectly "
3024                                                 "applying open mode checks to "
3025                                                 "a SETATTR that is not "
3026                                                 "changing file size.\n",
3027                                                 server->nfs_client->cl_hostname);
3028                         }
3029                         if (state && !(state->state & FMODE_WRITE)) {
3030                                 err = -EBADF;
3031                                 if (sattr->ia_valid & ATTR_OPEN)
3032                                         err = -EACCES;
3033                                 goto out;
3034                         }
3035                 }
3036                 err = nfs4_handle_exception(server, err, &exception);
3037         } while (exception.retry);
3038 out:
3039         return err;
3040 }
3041
3042 static bool
3043 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3044 {
3045         if (inode == NULL || !nfs_have_layout(inode))
3046                 return false;
3047
3048         return pnfs_wait_on_layoutreturn(inode, task);
3049 }
3050
3051 struct nfs4_closedata {
3052         struct inode *inode;
3053         struct nfs4_state *state;
3054         struct nfs_closeargs arg;
3055         struct nfs_closeres res;
3056         struct {
3057                 struct nfs4_layoutreturn_args arg;
3058                 struct nfs4_layoutreturn_res res;
3059                 struct nfs4_xdr_opaque_data ld_private;
3060                 u32 roc_barrier;
3061                 bool roc;
3062         } lr;
3063         struct nfs_fattr fattr;
3064         unsigned long timestamp;
3065 };
3066
3067 static void nfs4_free_closedata(void *data)
3068 {
3069         struct nfs4_closedata *calldata = data;
3070         struct nfs4_state_owner *sp = calldata->state->owner;
3071         struct super_block *sb = calldata->state->inode->i_sb;
3072
3073         if (calldata->lr.roc)
3074                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3075                                 calldata->res.lr_ret);
3076         nfs4_put_open_state(calldata->state);
3077         nfs_free_seqid(calldata->arg.seqid);
3078         nfs4_put_state_owner(sp);
3079         nfs_sb_deactive(sb);
3080         kfree(calldata);
3081 }
3082
3083 static void nfs4_close_done(struct rpc_task *task, void *data)
3084 {
3085         struct nfs4_closedata *calldata = data;
3086         struct nfs4_state *state = calldata->state;
3087         struct nfs_server *server = NFS_SERVER(calldata->inode);
3088         nfs4_stateid *res_stateid = NULL;
3089
3090         dprintk("%s: begin!\n", __func__);
3091         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3092                 return;
3093         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3094
3095         /* Handle Layoutreturn errors */
3096         if (calldata->arg.lr_args && task->tk_status != 0) {
3097                 switch (calldata->res.lr_ret) {
3098                 default:
3099                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3100                         break;
3101                 case 0:
3102                         calldata->arg.lr_args = NULL;
3103                         calldata->res.lr_res = NULL;
3104                         break;
3105                 case -NFS4ERR_ADMIN_REVOKED:
3106                 case -NFS4ERR_DELEG_REVOKED:
3107                 case -NFS4ERR_EXPIRED:
3108                 case -NFS4ERR_BAD_STATEID:
3109                 case -NFS4ERR_OLD_STATEID:
3110                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3111                 case -NFS4ERR_WRONG_CRED:
3112                         calldata->arg.lr_args = NULL;
3113                         calldata->res.lr_res = NULL;
3114                         calldata->res.lr_ret = 0;
3115                         rpc_restart_call_prepare(task);
3116                         return;
3117                 }
3118         }
3119
3120         /* hmm. we are done with the inode, and in the process of freeing
3121          * the state_owner. we keep this around to process errors
3122          */
3123         switch (task->tk_status) {
3124                 case 0:
3125                         res_stateid = &calldata->res.stateid;
3126                         renew_lease(server, calldata->timestamp);
3127                         break;
3128                 case -NFS4ERR_ACCESS:
3129                         if (calldata->arg.bitmask != NULL) {
3130                                 calldata->arg.bitmask = NULL;
3131                                 calldata->res.fattr = NULL;
3132                                 task->tk_status = 0;
3133                                 rpc_restart_call_prepare(task);
3134                                 goto out_release;
3135
3136                         }
3137                         break;
3138                 case -NFS4ERR_ADMIN_REVOKED:
3139                 case -NFS4ERR_STALE_STATEID:
3140                 case -NFS4ERR_EXPIRED:
3141                         nfs4_free_revoked_stateid(server,
3142                                         &calldata->arg.stateid,
3143                                         task->tk_msg.rpc_cred);
3144                 case -NFS4ERR_OLD_STATEID:
3145                 case -NFS4ERR_BAD_STATEID:
3146                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3147                                                 &state->open_stateid)) {
3148                                 rpc_restart_call_prepare(task);
3149                                 goto out_release;
3150                         }
3151                         if (calldata->arg.fmode == 0)
3152                                 break;
3153                 default:
3154                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3155                                 rpc_restart_call_prepare(task);
3156                                 goto out_release;
3157                         }
3158         }
3159         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3160                         res_stateid, calldata->arg.fmode);
3161 out_release:
3162         nfs_release_seqid(calldata->arg.seqid);
3163         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3164         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3165 }
3166
3167 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3168 {
3169         struct nfs4_closedata *calldata = data;
3170         struct nfs4_state *state = calldata->state;
3171         struct inode *inode = calldata->inode;
3172         bool is_rdonly, is_wronly, is_rdwr;
3173         int call_close = 0;
3174
3175         dprintk("%s: begin!\n", __func__);
3176         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3177                 goto out_wait;
3178
3179         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3180         spin_lock(&state->owner->so_lock);
3181         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3182         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3183         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3184         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3185         /* Calculate the change in open mode */
3186         calldata->arg.fmode = 0;
3187         if (state->n_rdwr == 0) {
3188                 if (state->n_rdonly == 0)
3189                         call_close |= is_rdonly;
3190                 else if (is_rdonly)
3191                         calldata->arg.fmode |= FMODE_READ;
3192                 if (state->n_wronly == 0)
3193                         call_close |= is_wronly;
3194                 else if (is_wronly)
3195                         calldata->arg.fmode |= FMODE_WRITE;
3196                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3197                         call_close |= is_rdwr;
3198         } else if (is_rdwr)
3199                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3200
3201         if (!nfs4_valid_open_stateid(state) ||
3202             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3203                 call_close = 0;
3204         spin_unlock(&state->owner->so_lock);
3205
3206         if (!call_close) {
3207                 /* Note: exit _without_ calling nfs4_close_done */
3208                 goto out_no_action;
3209         }
3210
3211         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3212                 nfs_release_seqid(calldata->arg.seqid);
3213                 goto out_wait;
3214         }
3215
3216         if (calldata->arg.fmode == 0)
3217                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3218
3219         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3220                 /* Close-to-open cache consistency revalidation */
3221                 if (!nfs4_have_delegation(inode, FMODE_READ))
3222                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3223                 else
3224                         calldata->arg.bitmask = NULL;
3225         }
3226
3227         calldata->arg.share_access =
3228                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3229                                 calldata->arg.fmode, 0);
3230
3231         if (calldata->res.fattr == NULL)
3232                 calldata->arg.bitmask = NULL;
3233         else if (calldata->arg.bitmask == NULL)
3234                 calldata->res.fattr = NULL;
3235         calldata->timestamp = jiffies;
3236         if (nfs4_setup_sequence(NFS_SERVER(inode),
3237                                 &calldata->arg.seq_args,
3238                                 &calldata->res.seq_res,
3239                                 task) != 0)
3240                 nfs_release_seqid(calldata->arg.seqid);
3241         dprintk("%s: done!\n", __func__);
3242         return;
3243 out_no_action:
3244         task->tk_action = NULL;
3245 out_wait:
3246         nfs4_sequence_done(task, &calldata->res.seq_res);
3247 }
3248
3249 static const struct rpc_call_ops nfs4_close_ops = {
3250         .rpc_call_prepare = nfs4_close_prepare,
3251         .rpc_call_done = nfs4_close_done,
3252         .rpc_release = nfs4_free_closedata,
3253 };
3254
3255 /* 
3256  * It is possible for data to be read/written from a mem-mapped file 
3257  * after the sys_close call (which hits the vfs layer as a flush).
3258  * This means that we can't safely call nfsv4 close on a file until 
3259  * the inode is cleared. This in turn means that we are not good
3260  * NFSv4 citizens - we do not indicate to the server to update the file's 
3261  * share state even when we are done with one of the three share 
3262  * stateid's in the inode.
3263  *
3264  * NOTE: Caller must be holding the sp->so_owner semaphore!
3265  */
3266 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3267 {
3268         struct nfs_server *server = NFS_SERVER(state->inode);
3269         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3270         struct nfs4_closedata *calldata;
3271         struct nfs4_state_owner *sp = state->owner;
3272         struct rpc_task *task;
3273         struct rpc_message msg = {
3274                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3275                 .rpc_cred = state->owner->so_cred,
3276         };
3277         struct rpc_task_setup task_setup_data = {
3278                 .rpc_client = server->client,
3279                 .rpc_message = &msg,
3280                 .callback_ops = &nfs4_close_ops,
3281                 .workqueue = nfsiod_workqueue,
3282                 .flags = RPC_TASK_ASYNC,
3283         };
3284         int status = -ENOMEM;
3285
3286         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3287                 &task_setup_data.rpc_client, &msg);
3288
3289         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3290         if (calldata == NULL)
3291                 goto out;
3292         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3293         calldata->inode = state->inode;
3294         calldata->state = state;
3295         calldata->arg.fh = NFS_FH(state->inode);
3296         /* Serialization for the sequence id */
3297         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3298         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3299         if (IS_ERR(calldata->arg.seqid))
3300                 goto out_free_calldata;
3301         nfs_fattr_init(&calldata->fattr);
3302         calldata->arg.fmode = 0;
3303         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3304         calldata->res.fattr = &calldata->fattr;
3305         calldata->res.seqid = calldata->arg.seqid;
3306         calldata->res.server = server;
3307         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3308         calldata->lr.roc = pnfs_roc(state->inode,
3309                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3310         if (calldata->lr.roc) {
3311                 calldata->arg.lr_args = &calldata->lr.arg;
3312                 calldata->res.lr_res = &calldata->lr.res;
3313         }
3314         nfs_sb_active(calldata->inode->i_sb);
3315
3316         msg.rpc_argp = &calldata->arg;
3317         msg.rpc_resp = &calldata->res;
3318         task_setup_data.callback_data = calldata;
3319         task = rpc_run_task(&task_setup_data);
3320         if (IS_ERR(task))
3321                 return PTR_ERR(task);
3322         status = 0;
3323         if (wait)
3324                 status = rpc_wait_for_completion_task(task);
3325         rpc_put_task(task);
3326         return status;
3327 out_free_calldata:
3328         kfree(calldata);
3329 out:
3330         nfs4_put_open_state(state);
3331         nfs4_put_state_owner(sp);
3332         return status;
3333 }
3334
3335 static struct inode *
3336 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3337                 int open_flags, struct iattr *attr, int *opened)
3338 {
3339         struct nfs4_state *state;
3340         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3341
3342         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3343
3344         /* Protect against concurrent sillydeletes */
3345         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3346
3347         nfs4_label_release_security(label);
3348
3349         if (IS_ERR(state))
3350                 return ERR_CAST(state);
3351         return state->inode;
3352 }
3353
3354 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3355 {
3356         if (ctx->state == NULL)
3357                 return;
3358         if (is_sync)
3359                 nfs4_close_sync(ctx->state, ctx->mode);
3360         else
3361                 nfs4_close_state(ctx->state, ctx->mode);
3362 }
3363
3364 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3365 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3366 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3367
3368 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3369 {
3370         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3371         struct nfs4_server_caps_arg args = {
3372                 .fhandle = fhandle,
3373                 .bitmask = bitmask,
3374         };
3375         struct nfs4_server_caps_res res = {};
3376         struct rpc_message msg = {
3377                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3378                 .rpc_argp = &args,
3379                 .rpc_resp = &res,
3380         };
3381         int status;
3382
3383         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3384                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3385                      FATTR4_WORD0_LINK_SUPPORT |
3386                      FATTR4_WORD0_SYMLINK_SUPPORT |
3387                      FATTR4_WORD0_ACLSUPPORT;
3388         if (minorversion)
3389                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3390
3391         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3392         if (status == 0) {
3393                 /* Sanity check the server answers */
3394                 switch (minorversion) {
3395                 case 0:
3396                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3397                         res.attr_bitmask[2] = 0;
3398                         break;
3399                 case 1:
3400                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3401                         break;
3402                 case 2:
3403                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3404                 }
3405                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3406                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3407                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3408                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3409                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3410                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3411                                 NFS_CAP_SECURITY_LABEL);
3412                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3413                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3414                         server->caps |= NFS_CAP_ACLS;
3415                 if (res.has_links != 0)
3416                         server->caps |= NFS_CAP_HARDLINKS;
3417                 if (res.has_symlinks != 0)
3418                         server->caps |= NFS_CAP_SYMLINKS;
3419                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3420                         server->caps |= NFS_CAP_FILEID;
3421                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3422                         server->caps |= NFS_CAP_MODE;
3423                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3424                         server->caps |= NFS_CAP_NLINK;
3425                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3426                         server->caps |= NFS_CAP_OWNER;
3427                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3428                         server->caps |= NFS_CAP_OWNER_GROUP;
3429                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3430                         server->caps |= NFS_CAP_ATIME;
3431                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3432                         server->caps |= NFS_CAP_CTIME;
3433                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3434                         server->caps |= NFS_CAP_MTIME;
3435 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3436                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3437                         server->caps |= NFS_CAP_SECURITY_LABEL;
3438 #endif
3439                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3440                                 sizeof(server->attr_bitmask));
3441                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3442
3443                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3444                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3445                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3446                 server->cache_consistency_bitmask[2] = 0;
3447                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3448                         sizeof(server->exclcreat_bitmask));
3449                 server->acl_bitmask = res.acl_bitmask;
3450                 server->fh_expire_type = res.fh_expire_type;
3451         }
3452
3453         return status;
3454 }
3455
3456 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3457 {
3458         struct nfs4_exception exception = { };
3459         int err;
3460         do {
3461                 err = nfs4_handle_exception(server,
3462                                 _nfs4_server_capabilities(server, fhandle),
3463                                 &exception);
3464         } while (exception.retry);
3465         return err;
3466 }
3467
3468 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3469                 struct nfs_fsinfo *info)
3470 {
3471         u32 bitmask[3];
3472         struct nfs4_lookup_root_arg args = {
3473                 .bitmask = bitmask,
3474         };
3475         struct nfs4_lookup_res res = {
3476                 .server = server,
3477                 .fattr = info->fattr,
3478                 .fh = fhandle,
3479         };
3480         struct rpc_message msg = {
3481                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3482                 .rpc_argp = &args,
3483                 .rpc_resp = &res,
3484         };
3485
3486         bitmask[0] = nfs4_fattr_bitmap[0];
3487         bitmask[1] = nfs4_fattr_bitmap[1];
3488         /*
3489          * Process the label in the upcoming getfattr
3490          */
3491         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3492
3493         nfs_fattr_init(info->fattr);
3494         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3495 }
3496
3497 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3498                 struct nfs_fsinfo *info)
3499 {
3500         struct nfs4_exception exception = { };
3501         int err;
3502         do {
3503                 err = _nfs4_lookup_root(server, fhandle, info);
3504                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3505                 switch (err) {
3506                 case 0:
3507                 case -NFS4ERR_WRONGSEC:
3508                         goto out;
3509                 default:
3510                         err = nfs4_handle_exception(server, err, &exception);
3511                 }
3512         } while (exception.retry);
3513 out:
3514         return err;
3515 }
3516
3517 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3518                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3519 {
3520         struct rpc_auth_create_args auth_args = {
3521                 .pseudoflavor = flavor,
3522         };
3523         struct rpc_auth *auth;
3524         int ret;
3525
3526         auth = rpcauth_create(&auth_args, server->client);
3527         if (IS_ERR(auth)) {
3528                 ret = -EACCES;
3529                 goto out;
3530         }
3531         ret = nfs4_lookup_root(server, fhandle, info);
3532 out:
3533         return ret;
3534 }
3535
3536 /*
3537  * Retry pseudoroot lookup with various security flavors.  We do this when:
3538  *
3539  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3540  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3541  *
3542  * Returns zero on success, or a negative NFS4ERR value, or a
3543  * negative errno value.
3544  */
3545 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3546                               struct nfs_fsinfo *info)
3547 {
3548         /* Per 3530bis 15.33.5 */
3549         static const rpc_authflavor_t flav_array[] = {
3550                 RPC_AUTH_GSS_KRB5P,
3551                 RPC_AUTH_GSS_KRB5I,
3552                 RPC_AUTH_GSS_KRB5,
3553                 RPC_AUTH_UNIX,                  /* courtesy */
3554                 RPC_AUTH_NULL,
3555         };
3556         int status = -EPERM;
3557         size_t i;
3558
3559         if (server->auth_info.flavor_len > 0) {
3560                 /* try each flavor specified by user */
3561                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3562                         status = nfs4_lookup_root_sec(server, fhandle, info,
3563                                                 server->auth_info.flavors[i]);
3564                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3565                                 continue;
3566                         break;
3567                 }
3568         } else {
3569                 /* no flavors specified by user, try default list */
3570                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3571                         status = nfs4_lookup_root_sec(server, fhandle, info,
3572                                                       flav_array[i]);
3573                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3574                                 continue;
3575                         break;
3576                 }
3577         }
3578
3579         /*
3580          * -EACCESS could mean that the user doesn't have correct permissions
3581          * to access the mount.  It could also mean that we tried to mount
3582          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3583          * existing mount programs don't handle -EACCES very well so it should
3584          * be mapped to -EPERM instead.
3585          */
3586         if (status == -EACCES)
3587                 status = -EPERM;
3588         return status;
3589 }
3590
3591 /**
3592  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3593  * @server: initialized nfs_server handle
3594  * @fhandle: we fill in the pseudo-fs root file handle
3595  * @info: we fill in an FSINFO struct
3596  * @auth_probe: probe the auth flavours
3597  *
3598  * Returns zero on success, or a negative errno.
3599  */
3600 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3601                          struct nfs_fsinfo *info,
3602                          bool auth_probe)
3603 {
3604         int status = 0;
3605
3606         if (!auth_probe)
3607                 status = nfs4_lookup_root(server, fhandle, info);
3608
3609         if (auth_probe || status == NFS4ERR_WRONGSEC)
3610                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3611                                 fhandle, info);
3612
3613         if (status == 0)
3614                 status = nfs4_server_capabilities(server, fhandle);
3615         if (status == 0)
3616                 status = nfs4_do_fsinfo(server, fhandle, info);
3617
3618         return nfs4_map_errors(status);
3619 }
3620
3621 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3622                               struct nfs_fsinfo *info)
3623 {
3624         int error;
3625         struct nfs_fattr *fattr = info->fattr;
3626         struct nfs4_label *label = NULL;
3627
3628         error = nfs4_server_capabilities(server, mntfh);
3629         if (error < 0) {
3630                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3631                 return error;
3632         }
3633
3634         label = nfs4_label_alloc(server, GFP_KERNEL);
3635         if (IS_ERR(label))
3636                 return PTR_ERR(label);
3637
3638         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3639         if (error < 0) {
3640                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3641                 goto err_free_label;
3642         }
3643
3644         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3645             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3646                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3647
3648 err_free_label:
3649         nfs4_label_free(label);
3650
3651         return error;
3652 }
3653
3654 /*
3655  * Get locations and (maybe) other attributes of a referral.
3656  * Note that we'll actually follow the referral later when
3657  * we detect fsid mismatch in inode revalidation
3658  */
3659 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3660                              const struct qstr *name, struct nfs_fattr *fattr,
3661                              struct nfs_fh *fhandle)
3662 {
3663         int status = -ENOMEM;
3664         struct page *page = NULL;
3665         struct nfs4_fs_locations *locations = NULL;
3666
3667         page = alloc_page(GFP_KERNEL);
3668         if (page == NULL)
3669                 goto out;
3670         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3671         if (locations == NULL)
3672                 goto out;
3673
3674         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3675         if (status != 0)
3676                 goto out;
3677
3678         /*
3679          * If the fsid didn't change, this is a migration event, not a
3680          * referral.  Cause us to drop into the exception handler, which
3681          * will kick off migration recovery.
3682          */
3683         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3684                 dprintk("%s: server did not return a different fsid for"
3685                         " a referral at %s\n", __func__, name->name);
3686                 status = -NFS4ERR_MOVED;
3687                 goto out;
3688         }
3689         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3690         nfs_fixup_referral_attributes(&locations->fattr);
3691
3692         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3693         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3694         memset(fhandle, 0, sizeof(struct nfs_fh));
3695 out:
3696         if (page)
3697                 __free_page(page);
3698         kfree(locations);
3699         return status;
3700 }
3701
3702 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3703                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3704 {
3705         struct nfs4_getattr_arg args = {
3706                 .fh = fhandle,
3707                 .bitmask = server->attr_bitmask,
3708         };
3709         struct nfs4_getattr_res res = {
3710                 .fattr = fattr,
3711                 .label = label,
3712                 .server = server,
3713         };
3714         struct rpc_message msg = {
3715                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3716                 .rpc_argp = &args,
3717                 .rpc_resp = &res,
3718         };
3719
3720         args.bitmask = nfs4_bitmask(server, label);
3721
3722         nfs_fattr_init(fattr);
3723         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3724 }
3725
3726 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3727                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3728 {
3729         struct nfs4_exception exception = { };
3730         int err;
3731         do {
3732                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3733                 trace_nfs4_getattr(server, fhandle, fattr, err);
3734                 err = nfs4_handle_exception(server, err,
3735                                 &exception);
3736         } while (exception.retry);
3737         return err;
3738 }
3739
3740 /* 
3741  * The file is not closed if it is opened due to the a request to change
3742  * the size of the file. The open call will not be needed once the
3743  * VFS layer lookup-intents are implemented.
3744  *
3745  * Close is called when the inode is destroyed.
3746  * If we haven't opened the file for O_WRONLY, we
3747  * need to in the size_change case to obtain a stateid.
3748  *
3749  * Got race?
3750  * Because OPEN is always done by name in nfsv4, it is
3751  * possible that we opened a different file by the same
3752  * name.  We can recognize this race condition, but we
3753  * can't do anything about it besides returning an error.
3754  *
3755  * This will be fixed with VFS changes (lookup-intent).
3756  */
3757 static int
3758 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3759                   struct iattr *sattr)
3760 {
3761         struct inode *inode = d_inode(dentry);
3762         struct rpc_cred *cred = NULL;
3763         struct nfs_open_context *ctx = NULL;
3764         struct nfs4_label *label = NULL;
3765         int status;
3766
3767         if (pnfs_ld_layoutret_on_setattr(inode) &&
3768             sattr->ia_valid & ATTR_SIZE &&
3769             sattr->ia_size < i_size_read(inode))
3770                 pnfs_commit_and_return_layout(inode);
3771
3772         nfs_fattr_init(fattr);
3773         
3774         /* Deal with open(O_TRUNC) */
3775         if (sattr->ia_valid & ATTR_OPEN)
3776                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3777
3778         /* Optimization: if the end result is no change, don't RPC */
3779         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3780                 return 0;
3781
3782         /* Search for an existing open(O_WRITE) file */
3783         if (sattr->ia_valid & ATTR_FILE) {
3784
3785                 ctx = nfs_file_open_context(sattr->ia_file);
3786                 if (ctx)
3787                         cred = ctx->cred;
3788         }
3789
3790         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3791         if (IS_ERR(label))
3792                 return PTR_ERR(label);
3793
3794         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3795         if (status == 0) {
3796                 nfs_setattr_update_inode(inode, sattr, fattr);
3797                 nfs_setsecurity(inode, fattr, label);
3798         }
3799         nfs4_label_free(label);
3800         return status;
3801 }
3802
3803 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3804                 const struct qstr *name, struct nfs_fh *fhandle,
3805                 struct nfs_fattr *fattr, struct nfs4_label *label)
3806 {
3807         struct nfs_server *server = NFS_SERVER(dir);
3808         int                    status;
3809         struct nfs4_lookup_arg args = {
3810                 .bitmask = server->attr_bitmask,
3811                 .dir_fh = NFS_FH(dir),
3812                 .name = name,
3813         };
3814         struct nfs4_lookup_res res = {
3815                 .server = server,
3816                 .fattr = fattr,
3817                 .label = label,
3818                 .fh = fhandle,
3819         };
3820         struct rpc_message msg = {
3821                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3822                 .rpc_argp = &args,
3823                 .rpc_resp = &res,
3824         };
3825
3826         args.bitmask = nfs4_bitmask(server, label);
3827
3828         nfs_fattr_init(fattr);
3829
3830         dprintk("NFS call  lookup %s\n", name->name);
3831         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3832         dprintk("NFS reply lookup: %d\n", status);
3833         return status;
3834 }
3835
3836 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3837 {
3838         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3839                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3840         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3841         fattr->nlink = 2;
3842 }
3843
3844 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3845                                    const struct qstr *name, struct nfs_fh *fhandle,
3846                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3847 {
3848         struct nfs4_exception exception = { };
3849         struct rpc_clnt *client = *clnt;
3850         int err;
3851         do {
3852                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3853                 trace_nfs4_lookup(dir, name, err);
3854                 switch (err) {
3855                 case -NFS4ERR_BADNAME:
3856                         err = -ENOENT;
3857                         goto out;
3858                 case -NFS4ERR_MOVED:
3859                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3860                         if (err == -NFS4ERR_MOVED)
3861                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3862                         goto out;
3863                 case -NFS4ERR_WRONGSEC:
3864                         err = -EPERM;
3865                         if (client != *clnt)
3866                                 goto out;
3867                         client = nfs4_negotiate_security(client, dir, name);
3868                         if (IS_ERR(client))
3869                                 return PTR_ERR(client);
3870
3871                         exception.retry = 1;
3872                         break;
3873                 default:
3874                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3875                 }
3876         } while (exception.retry);
3877
3878 out:
3879         if (err == 0)
3880                 *clnt = client;
3881         else if (client != *clnt)
3882                 rpc_shutdown_client(client);
3883
3884         return err;
3885 }
3886
3887 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3888                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3889                             struct nfs4_label *label)
3890 {
3891         int status;
3892         struct rpc_clnt *client = NFS_CLIENT(dir);
3893
3894         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3895         if (client != NFS_CLIENT(dir)) {
3896                 rpc_shutdown_client(client);
3897                 nfs_fixup_secinfo_attributes(fattr);
3898         }
3899         return status;
3900 }
3901
3902 struct rpc_clnt *
3903 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3904                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3905 {
3906         struct rpc_clnt *client = NFS_CLIENT(dir);
3907         int status;
3908
3909         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3910         if (status < 0)
3911                 return ERR_PTR(status);
3912         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3913 }
3914
3915 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3916 {
3917         struct nfs_server *server = NFS_SERVER(inode);
3918         struct nfs4_accessargs args = {
3919                 .fh = NFS_FH(inode),
3920                 .bitmask = server->cache_consistency_bitmask,
3921         };
3922         struct nfs4_accessres res = {
3923                 .server = server,
3924         };
3925         struct rpc_message msg = {
3926                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3927                 .rpc_argp = &args,
3928                 .rpc_resp = &res,
3929                 .rpc_cred = entry->cred,
3930         };
3931         int mode = entry->mask;
3932         int status = 0;
3933
3934         /*
3935          * Determine which access bits we want to ask for...
3936          */
3937         if (mode & MAY_READ)
3938                 args.access |= NFS4_ACCESS_READ;
3939         if (S_ISDIR(inode->i_mode)) {
3940                 if (mode & MAY_WRITE)
3941                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3942                 if (mode & MAY_EXEC)
3943                         args.access |= NFS4_ACCESS_LOOKUP;
3944         } else {
3945                 if (mode & MAY_WRITE)
3946                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3947                 if (mode & MAY_EXEC)
3948                         args.access |= NFS4_ACCESS_EXECUTE;
3949         }
3950
3951         res.fattr = nfs_alloc_fattr();
3952         if (res.fattr == NULL)
3953                 return -ENOMEM;
3954
3955         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3956         if (!status) {
3957                 nfs_access_set_mask(entry, res.access);
3958                 nfs_refresh_inode(inode, res.fattr);
3959         }
3960         nfs_free_fattr(res.fattr);
3961         return status;
3962 }
3963
3964 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3965 {
3966         struct nfs4_exception exception = { };
3967         int err;
3968         do {
3969                 err = _nfs4_proc_access(inode, entry);
3970                 trace_nfs4_access(inode, err);
3971                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3972                                 &exception);
3973         } while (exception.retry);
3974         return err;
3975 }
3976
3977 /*
3978  * TODO: For the time being, we don't try to get any attributes
3979  * along with any of the zero-copy operations READ, READDIR,
3980  * READLINK, WRITE.
3981  *
3982  * In the case of the first three, we want to put the GETATTR
3983  * after the read-type operation -- this is because it is hard
3984  * to predict the length of a GETATTR response in v4, and thus
3985  * align the READ data correctly.  This means that the GETATTR
3986  * may end up partially falling into the page cache, and we should
3987  * shift it into the 'tail' of the xdr_buf before processing.
3988  * To do this efficiently, we need to know the total length
3989  * of data received, which doesn't seem to be available outside
3990  * of the RPC layer.
3991  *
3992  * In the case of WRITE, we also want to put the GETATTR after
3993  * the operation -- in this case because we want to make sure
3994  * we get the post-operation mtime and size.
3995  *
3996  * Both of these changes to the XDR layer would in fact be quite
3997  * minor, but I decided to leave them for a subsequent patch.
3998  */
3999 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4000                 unsigned int pgbase, unsigned int pglen)
4001 {
4002         struct nfs4_readlink args = {
4003                 .fh       = NFS_FH(inode),
4004                 .pgbase   = pgbase,
4005                 .pglen    = pglen,
4006                 .pages    = &page,
4007         };
4008         struct nfs4_readlink_res res;
4009         struct rpc_message msg = {
4010                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4011                 .rpc_argp = &args,
4012                 .rpc_resp = &res,
4013         };
4014
4015         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4016 }
4017
4018 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4019                 unsigned int pgbase, unsigned int pglen)
4020 {
4021         struct nfs4_exception exception = { };
4022         int err;
4023         do {
4024                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4025                 trace_nfs4_readlink(inode, err);
4026                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4027                                 &exception);
4028         } while (exception.retry);
4029         return err;
4030 }
4031
4032 /*
4033  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4034  */
4035 static int
4036 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4037                  int flags)
4038 {
4039         struct nfs_server *server = NFS_SERVER(dir);
4040         struct nfs4_label l, *ilabel = NULL;
4041         struct nfs_open_context *ctx;
4042         struct nfs4_state *state;
4043         int status = 0;
4044
4045         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4046         if (IS_ERR(ctx))
4047                 return PTR_ERR(ctx);
4048
4049         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4050
4051         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4052                 sattr->ia_mode &= ~current_umask();
4053         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4054         if (IS_ERR(state)) {
4055                 status = PTR_ERR(state);
4056                 goto out;
4057         }
4058 out:
4059         nfs4_label_release_security(ilabel);
4060         put_nfs_open_context(ctx);
4061         return status;
4062 }
4063
4064 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4065 {
4066         struct nfs_server *server = NFS_SERVER(dir);
4067         struct nfs_removeargs args = {
4068                 .fh = NFS_FH(dir),
4069                 .name = *name,
4070         };
4071         struct nfs_removeres res = {
4072                 .server = server,
4073         };
4074         struct rpc_message msg = {
4075                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4076                 .rpc_argp = &args,
4077                 .rpc_resp = &res,
4078         };
4079         unsigned long timestamp = jiffies;
4080         int status;
4081
4082         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4083         if (status == 0)
4084                 update_changeattr(dir, &res.cinfo, timestamp);
4085         return status;
4086 }
4087
4088 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4089 {
4090         struct nfs4_exception exception = { };
4091         int err;
4092         do {
4093                 err = _nfs4_proc_remove(dir, name);
4094                 trace_nfs4_remove(dir, name, err);
4095                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4096                                 &exception);
4097         } while (exception.retry);
4098         return err;
4099 }
4100
4101 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4102 {
4103         struct nfs_server *server = NFS_SERVER(dir);
4104         struct nfs_removeargs *args = msg->rpc_argp;
4105         struct nfs_removeres *res = msg->rpc_resp;
4106
4107         res->server = server;
4108         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4109         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4110
4111         nfs_fattr_init(res->dir_attr);
4112 }
4113
4114 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4115 {
4116         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
4117                         &data->args.seq_args,
4118                         &data->res.seq_res,
4119                         task);
4120 }
4121
4122 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4123 {
4124         struct nfs_unlinkdata *data = task->tk_calldata;
4125         struct nfs_removeres *res = &data->res;
4126
4127         if (!nfs4_sequence_done(task, &res->seq_res))
4128                 return 0;
4129         if (nfs4_async_handle_error(task, res->server, NULL,
4130                                     &data->timeout) == -EAGAIN)
4131                 return 0;
4132         if (task->tk_status == 0)
4133                 update_changeattr(dir, &res->cinfo, res->dir_attr->time_start);
4134         return 1;
4135 }
4136
4137 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4138 {
4139         struct nfs_server *server = NFS_SERVER(dir);
4140         struct nfs_renameargs *arg = msg->rpc_argp;
4141         struct nfs_renameres *res = msg->rpc_resp;
4142
4143         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4144         res->server = server;
4145         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4146 }
4147
4148 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4149 {
4150         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
4151                         &data->args.seq_args,
4152                         &data->res.seq_res,
4153                         task);
4154 }
4155
4156 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4157                                  struct inode *new_dir)
4158 {
4159         struct nfs_renamedata *data = task->tk_calldata;
4160         struct nfs_renameres *res = &data->res;
4161
4162         if (!nfs4_sequence_done(task, &res->seq_res))
4163                 return 0;
4164         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4165                 return 0;
4166
4167         if (task->tk_status == 0) {
4168                 update_changeattr(old_dir, &res->old_cinfo, res->old_fattr->time_start);
4169                 if (new_dir != old_dir)
4170                         update_changeattr(new_dir, &res->new_cinfo, res->new_fattr->time_start);
4171         }
4172         return 1;
4173 }
4174
4175 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4176 {
4177         struct nfs_server *server = NFS_SERVER(inode);
4178         struct nfs4_link_arg arg = {
4179                 .fh     = NFS_FH(inode),
4180                 .dir_fh = NFS_FH(dir),
4181                 .name   = name,
4182                 .bitmask = server->attr_bitmask,
4183         };
4184         struct nfs4_link_res res = {
4185                 .server = server,
4186                 .label = NULL,
4187         };
4188         struct rpc_message msg = {
4189                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4190                 .rpc_argp = &arg,
4191                 .rpc_resp = &res,
4192         };
4193         int status = -ENOMEM;
4194
4195         res.fattr = nfs_alloc_fattr();
4196         if (res.fattr == NULL)
4197                 goto out;
4198
4199         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4200         if (IS_ERR(res.label)) {
4201                 status = PTR_ERR(res.label);
4202                 goto out;
4203         }
4204         arg.bitmask = nfs4_bitmask(server, res.label);
4205
4206         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4207         if (!status) {
4208                 update_changeattr(dir, &res.cinfo, res.fattr->time_start);
4209                 status = nfs_post_op_update_inode(inode, res.fattr);
4210                 if (!status)
4211                         nfs_setsecurity(inode, res.fattr, res.label);
4212         }
4213
4214
4215         nfs4_label_free(res.label);
4216
4217 out:
4218         nfs_free_fattr(res.fattr);
4219         return status;
4220 }
4221
4222 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4223 {
4224         struct nfs4_exception exception = { };
4225         int err;
4226         do {
4227                 err = nfs4_handle_exception(NFS_SERVER(inode),
4228                                 _nfs4_proc_link(inode, dir, name),
4229                                 &exception);
4230         } while (exception.retry);
4231         return err;
4232 }
4233
4234 struct nfs4_createdata {
4235         struct rpc_message msg;
4236         struct nfs4_create_arg arg;
4237         struct nfs4_create_res res;
4238         struct nfs_fh fh;
4239         struct nfs_fattr fattr;
4240         struct nfs4_label *label;
4241 };
4242
4243 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4244                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4245 {
4246         struct nfs4_createdata *data;
4247
4248         data = kzalloc(sizeof(*data), GFP_KERNEL);
4249         if (data != NULL) {
4250                 struct nfs_server *server = NFS_SERVER(dir);
4251
4252                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4253                 if (IS_ERR(data->label))
4254                         goto out_free;
4255
4256                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4257                 data->msg.rpc_argp = &data->arg;
4258                 data->msg.rpc_resp = &data->res;
4259                 data->arg.dir_fh = NFS_FH(dir);
4260                 data->arg.server = server;
4261                 data->arg.name = name;
4262                 data->arg.attrs = sattr;
4263                 data->arg.ftype = ftype;
4264                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4265                 data->arg.umask = current_umask();
4266                 data->res.server = server;
4267                 data->res.fh = &data->fh;
4268                 data->res.fattr = &data->fattr;
4269                 data->res.label = data->label;
4270                 nfs_fattr_init(data->res.fattr);
4271         }
4272         return data;
4273 out_free:
4274         kfree(data);
4275         return NULL;
4276 }
4277
4278 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4279 {
4280         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4281                                     &data->arg.seq_args, &data->res.seq_res, 1);
4282         if (status == 0) {
4283                 update_changeattr(dir, &data->res.dir_cinfo,
4284                                 data->res.fattr->time_start);
4285                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4286         }
4287         return status;
4288 }
4289
4290 static void nfs4_free_createdata(struct nfs4_createdata *data)
4291 {
4292         nfs4_label_free(data->label);
4293         kfree(data);
4294 }
4295
4296 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4297                 struct page *page, unsigned int len, struct iattr *sattr,
4298                 struct nfs4_label *label)
4299 {
4300         struct nfs4_createdata *data;
4301         int status = -ENAMETOOLONG;
4302
4303         if (len > NFS4_MAXPATHLEN)
4304                 goto out;
4305
4306         status = -ENOMEM;
4307         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4308         if (data == NULL)
4309                 goto out;
4310
4311         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4312         data->arg.u.symlink.pages = &page;
4313         data->arg.u.symlink.len = len;
4314         data->arg.label = label;
4315         
4316         status = nfs4_do_create(dir, dentry, data);
4317
4318         nfs4_free_createdata(data);
4319 out:
4320         return status;
4321 }
4322
4323 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4324                 struct page *page, unsigned int len, struct iattr *sattr)
4325 {
4326         struct nfs4_exception exception = { };
4327         struct nfs4_label l, *label = NULL;
4328         int err;
4329
4330         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4331
4332         do {
4333                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4334                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4335                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4336                                 &exception);
4337         } while (exception.retry);
4338
4339         nfs4_label_release_security(label);
4340         return err;
4341 }
4342
4343 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4344                 struct iattr *sattr, struct nfs4_label *label)
4345 {
4346         struct nfs4_createdata *data;
4347         int status = -ENOMEM;
4348
4349         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4350         if (data == NULL)
4351                 goto out;
4352
4353         data->arg.label = label;
4354         status = nfs4_do_create(dir, dentry, data);
4355
4356         nfs4_free_createdata(data);
4357 out:
4358         return status;
4359 }
4360
4361 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4362                 struct iattr *sattr)
4363 {
4364         struct nfs_server *server = NFS_SERVER(dir);
4365         struct nfs4_exception exception = { };
4366         struct nfs4_label l, *label = NULL;
4367         int err;
4368
4369         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4370
4371         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4372                 sattr->ia_mode &= ~current_umask();
4373         do {
4374                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4375                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4376                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4377                                 &exception);
4378         } while (exception.retry);
4379         nfs4_label_release_security(label);
4380
4381         return err;
4382 }
4383
4384 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4385                 u64 cookie, struct page **pages, unsigned int count, int plus)
4386 {
4387         struct inode            *dir = d_inode(dentry);
4388         struct nfs4_readdir_arg args = {
4389                 .fh = NFS_FH(dir),
4390                 .pages = pages,
4391                 .pgbase = 0,
4392                 .count = count,
4393                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4394                 .plus = plus,
4395         };
4396         struct nfs4_readdir_res res;
4397         struct rpc_message msg = {
4398                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4399                 .rpc_argp = &args,
4400                 .rpc_resp = &res,
4401                 .rpc_cred = cred,
4402         };
4403         int                     status;
4404
4405         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4406                         dentry,
4407                         (unsigned long long)cookie);
4408         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4409         res.pgbase = args.pgbase;
4410         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4411         if (status >= 0) {
4412                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4413                 status += args.pgbase;
4414         }
4415
4416         nfs_invalidate_atime(dir);
4417
4418         dprintk("%s: returns %d\n", __func__, status);
4419         return status;
4420 }
4421
4422 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4423                 u64 cookie, struct page **pages, unsigned int count, int plus)
4424 {
4425         struct nfs4_exception exception = { };
4426         int err;
4427         do {
4428                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4429                                 pages, count, plus);
4430                 trace_nfs4_readdir(d_inode(dentry), err);
4431                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4432                                 &exception);
4433         } while (exception.retry);
4434         return err;
4435 }
4436
4437 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4438                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4439 {
4440         struct nfs4_createdata *data;
4441         int mode = sattr->ia_mode;
4442         int status = -ENOMEM;
4443
4444         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4445         if (data == NULL)
4446                 goto out;
4447
4448         if (S_ISFIFO(mode))
4449                 data->arg.ftype = NF4FIFO;
4450         else if (S_ISBLK(mode)) {
4451                 data->arg.ftype = NF4BLK;
4452                 data->arg.u.device.specdata1 = MAJOR(rdev);
4453                 data->arg.u.device.specdata2 = MINOR(rdev);
4454         }
4455         else if (S_ISCHR(mode)) {
4456                 data->arg.ftype = NF4CHR;
4457                 data->arg.u.device.specdata1 = MAJOR(rdev);
4458                 data->arg.u.device.specdata2 = MINOR(rdev);
4459         } else if (!S_ISSOCK(mode)) {
4460                 status = -EINVAL;
4461                 goto out_free;
4462         }
4463
4464         data->arg.label = label;
4465         status = nfs4_do_create(dir, dentry, data);
4466 out_free:
4467         nfs4_free_createdata(data);
4468 out:
4469         return status;
4470 }
4471
4472 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4473                 struct iattr *sattr, dev_t rdev)
4474 {
4475         struct nfs_server *server = NFS_SERVER(dir);
4476         struct nfs4_exception exception = { };
4477         struct nfs4_label l, *label = NULL;
4478         int err;
4479
4480         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4481
4482         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4483                 sattr->ia_mode &= ~current_umask();
4484         do {
4485                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4486                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4487                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4488                                 &exception);
4489         } while (exception.retry);
4490
4491         nfs4_label_release_security(label);
4492
4493         return err;
4494 }
4495
4496 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4497                  struct nfs_fsstat *fsstat)
4498 {
4499         struct nfs4_statfs_arg args = {
4500                 .fh = fhandle,
4501                 .bitmask = server->attr_bitmask,
4502         };
4503         struct nfs4_statfs_res res = {
4504                 .fsstat = fsstat,
4505         };
4506         struct rpc_message msg = {
4507                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4508                 .rpc_argp = &args,
4509                 .rpc_resp = &res,
4510         };
4511
4512         nfs_fattr_init(fsstat->fattr);
4513         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4514 }
4515
4516 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4517 {
4518         struct nfs4_exception exception = { };
4519         int err;
4520         do {
4521                 err = nfs4_handle_exception(server,
4522                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4523                                 &exception);
4524         } while (exception.retry);
4525         return err;
4526 }
4527
4528 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4529                 struct nfs_fsinfo *fsinfo)
4530 {
4531         struct nfs4_fsinfo_arg args = {
4532                 .fh = fhandle,
4533                 .bitmask = server->attr_bitmask,
4534         };
4535         struct nfs4_fsinfo_res res = {
4536                 .fsinfo = fsinfo,
4537         };
4538         struct rpc_message msg = {
4539                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4540                 .rpc_argp = &args,
4541                 .rpc_resp = &res,
4542         };
4543
4544         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4545 }
4546
4547 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4548 {
4549         struct nfs4_exception exception = { };
4550         unsigned long now = jiffies;
4551         int err;
4552
4553         do {
4554                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4555                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4556                 if (err == 0) {
4557                         nfs4_set_lease_period(server->nfs_client,
4558                                         fsinfo->lease_time * HZ,
4559                                         now);
4560                         break;
4561                 }
4562                 err = nfs4_handle_exception(server, err, &exception);
4563         } while (exception.retry);
4564         return err;
4565 }
4566
4567 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4568 {
4569         int error;
4570
4571         nfs_fattr_init(fsinfo->fattr);
4572         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4573         if (error == 0) {
4574                 /* block layout checks this! */
4575                 server->pnfs_blksize = fsinfo->blksize;
4576                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4577         }
4578
4579         return error;
4580 }
4581
4582 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4583                 struct nfs_pathconf *pathconf)
4584 {
4585         struct nfs4_pathconf_arg args = {
4586                 .fh = fhandle,
4587                 .bitmask = server->attr_bitmask,
4588         };
4589         struct nfs4_pathconf_res res = {
4590                 .pathconf = pathconf,
4591         };
4592         struct rpc_message msg = {
4593                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4594                 .rpc_argp = &args,
4595                 .rpc_resp = &res,
4596         };
4597
4598         /* None of the pathconf attributes are mandatory to implement */
4599         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4600                 memset(pathconf, 0, sizeof(*pathconf));
4601                 return 0;
4602         }
4603
4604         nfs_fattr_init(pathconf->fattr);
4605         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4606 }
4607
4608 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4609                 struct nfs_pathconf *pathconf)
4610 {
4611         struct nfs4_exception exception = { };
4612         int err;
4613
4614         do {
4615                 err = nfs4_handle_exception(server,
4616                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4617                                 &exception);
4618         } while (exception.retry);
4619         return err;
4620 }
4621
4622 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4623                 const struct nfs_open_context *ctx,
4624                 const struct nfs_lock_context *l_ctx,
4625                 fmode_t fmode)
4626 {
4627         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4628 }
4629 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4630
4631 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4632                 const struct nfs_open_context *ctx,
4633                 const struct nfs_lock_context *l_ctx,
4634                 fmode_t fmode)
4635 {
4636         nfs4_stateid current_stateid;
4637
4638         /* If the current stateid represents a lost lock, then exit */
4639         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4640                 return true;
4641         return nfs4_stateid_match(stateid, &current_stateid);
4642 }
4643
4644 static bool nfs4_error_stateid_expired(int err)
4645 {
4646         switch (err) {
4647         case -NFS4ERR_DELEG_REVOKED:
4648         case -NFS4ERR_ADMIN_REVOKED:
4649         case -NFS4ERR_BAD_STATEID:
4650         case -NFS4ERR_STALE_STATEID:
4651         case -NFS4ERR_OLD_STATEID:
4652         case -NFS4ERR_OPENMODE:
4653         case -NFS4ERR_EXPIRED:
4654                 return true;
4655         }
4656         return false;
4657 }
4658
4659 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4660 {
4661         struct nfs_server *server = NFS_SERVER(hdr->inode);
4662
4663         trace_nfs4_read(hdr, task->tk_status);
4664         if (task->tk_status < 0) {
4665                 struct nfs4_exception exception = {
4666                         .inode = hdr->inode,
4667                         .state = hdr->args.context->state,
4668                         .stateid = &hdr->args.stateid,
4669                 };
4670                 task->tk_status = nfs4_async_handle_exception(task,
4671                                 server, task->tk_status, &exception);
4672                 if (exception.retry) {
4673                         rpc_restart_call_prepare(task);
4674                         return -EAGAIN;
4675                 }
4676         }
4677
4678         if (task->tk_status > 0)
4679                 renew_lease(server, hdr->timestamp);
4680         return 0;
4681 }
4682
4683 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4684                 struct nfs_pgio_args *args)
4685 {
4686
4687         if (!nfs4_error_stateid_expired(task->tk_status) ||
4688                 nfs4_stateid_is_current(&args->stateid,
4689                                 args->context,
4690                                 args->lock_context,
4691                                 FMODE_READ))
4692                 return false;
4693         rpc_restart_call_prepare(task);
4694         return true;
4695 }
4696
4697 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4698 {
4699
4700         dprintk("--> %s\n", __func__);
4701
4702         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4703                 return -EAGAIN;
4704         if (nfs4_read_stateid_changed(task, &hdr->args))
4705                 return -EAGAIN;
4706         if (task->tk_status > 0)
4707                 nfs_invalidate_atime(hdr->inode);
4708         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4709                                     nfs4_read_done_cb(task, hdr);
4710 }
4711
4712 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4713                                  struct rpc_message *msg)
4714 {
4715         hdr->timestamp   = jiffies;
4716         if (!hdr->pgio_done_cb)
4717                 hdr->pgio_done_cb = nfs4_read_done_cb;
4718         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4719         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4720 }
4721
4722 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4723                                       struct nfs_pgio_header *hdr)
4724 {
4725         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4726                         &hdr->args.seq_args,
4727                         &hdr->res.seq_res,
4728                         task))
4729                 return 0;
4730         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4731                                 hdr->args.lock_context,
4732                                 hdr->rw_ops->rw_mode) == -EIO)
4733                 return -EIO;
4734         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4735                 return -EIO;
4736         return 0;
4737 }
4738
4739 static int nfs4_write_done_cb(struct rpc_task *task,
4740                               struct nfs_pgio_header *hdr)
4741 {
4742         struct inode *inode = hdr->inode;
4743
4744         trace_nfs4_write(hdr, task->tk_status);
4745         if (task->tk_status < 0) {
4746                 struct nfs4_exception exception = {
4747                         .inode = hdr->inode,
4748                         .state = hdr->args.context->state,
4749                         .stateid = &hdr->args.stateid,
4750                 };
4751                 task->tk_status = nfs4_async_handle_exception(task,
4752                                 NFS_SERVER(inode), task->tk_status,
4753                                 &exception);
4754                 if (exception.retry) {
4755                         rpc_restart_call_prepare(task);
4756                         return -EAGAIN;
4757                 }
4758         }
4759         if (task->tk_status >= 0) {
4760                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4761                 nfs_writeback_update_inode(hdr);
4762         }
4763         return 0;
4764 }
4765
4766 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4767                 struct nfs_pgio_args *args)
4768 {
4769
4770         if (!nfs4_error_stateid_expired(task->tk_status) ||
4771                 nfs4_stateid_is_current(&args->stateid,
4772                                 args->context,
4773                                 args->lock_context,
4774                                 FMODE_WRITE))
4775                 return false;
4776         rpc_restart_call_prepare(task);
4777         return true;
4778 }
4779
4780 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4781 {
4782         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4783                 return -EAGAIN;
4784         if (nfs4_write_stateid_changed(task, &hdr->args))
4785                 return -EAGAIN;
4786         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4787                 nfs4_write_done_cb(task, hdr);
4788 }
4789
4790 static
4791 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4792 {
4793         /* Don't request attributes for pNFS or O_DIRECT writes */
4794         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4795                 return false;
4796         /* Otherwise, request attributes if and only if we don't hold
4797          * a delegation
4798          */
4799         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4800 }
4801
4802 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4803                                   struct rpc_message *msg)
4804 {
4805         struct nfs_server *server = NFS_SERVER(hdr->inode);
4806
4807         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4808                 hdr->args.bitmask = NULL;
4809                 hdr->res.fattr = NULL;
4810         } else
4811                 hdr->args.bitmask = server->cache_consistency_bitmask;
4812
4813         if (!hdr->pgio_done_cb)
4814                 hdr->pgio_done_cb = nfs4_write_done_cb;
4815         hdr->res.server = server;
4816         hdr->timestamp   = jiffies;
4817
4818         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4819         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4820 }
4821
4822 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4823 {
4824         nfs4_setup_sequence(NFS_SERVER(data->inode),
4825                         &data->args.seq_args,
4826                         &data->res.seq_res,
4827                         task);
4828 }
4829
4830 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4831 {
4832         struct inode *inode = data->inode;
4833
4834         trace_nfs4_commit(data, task->tk_status);
4835         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4836                                     NULL, NULL) == -EAGAIN) {
4837                 rpc_restart_call_prepare(task);
4838                 return -EAGAIN;
4839         }
4840         return 0;
4841 }
4842
4843 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4844 {
4845         if (!nfs4_sequence_done(task, &data->res.seq_res))
4846                 return -EAGAIN;
4847         return data->commit_done_cb(task, data);
4848 }
4849
4850 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4851 {
4852         struct nfs_server *server = NFS_SERVER(data->inode);
4853
4854         if (data->commit_done_cb == NULL)
4855                 data->commit_done_cb = nfs4_commit_done_cb;
4856         data->res.server = server;
4857         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4858         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4859 }
4860
4861 struct nfs4_renewdata {
4862         struct nfs_client       *client;
4863         unsigned long           timestamp;
4864 };
4865
4866 /*
4867  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4868  * standalone procedure for queueing an asynchronous RENEW.
4869  */
4870 static void nfs4_renew_release(void *calldata)
4871 {
4872         struct nfs4_renewdata *data = calldata;
4873         struct nfs_client *clp = data->client;
4874
4875         if (atomic_read(&clp->cl_count) > 1)
4876                 nfs4_schedule_state_renewal(clp);
4877         nfs_put_client(clp);
4878         kfree(data);
4879 }
4880
4881 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4882 {
4883         struct nfs4_renewdata *data = calldata;
4884         struct nfs_client *clp = data->client;
4885         unsigned long timestamp = data->timestamp;
4886
4887         trace_nfs4_renew_async(clp, task->tk_status);
4888         switch (task->tk_status) {
4889         case 0:
4890                 break;
4891         case -NFS4ERR_LEASE_MOVED:
4892                 nfs4_schedule_lease_moved_recovery(clp);
4893                 break;
4894         default:
4895                 /* Unless we're shutting down, schedule state recovery! */
4896                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4897                         return;
4898                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4899                         nfs4_schedule_lease_recovery(clp);
4900                         return;
4901                 }
4902                 nfs4_schedule_path_down_recovery(clp);
4903         }
4904         do_renew_lease(clp, timestamp);
4905 }
4906
4907 static const struct rpc_call_ops nfs4_renew_ops = {
4908         .rpc_call_done = nfs4_renew_done,
4909         .rpc_release = nfs4_renew_release,
4910 };
4911
4912 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4913 {
4914         struct rpc_message msg = {
4915                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4916                 .rpc_argp       = clp,
4917                 .rpc_cred       = cred,
4918         };
4919         struct nfs4_renewdata *data;
4920
4921         if (renew_flags == 0)
4922                 return 0;
4923         if (!atomic_inc_not_zero(&clp->cl_count))
4924                 return -EIO;
4925         data = kmalloc(sizeof(*data), GFP_NOFS);
4926         if (data == NULL)
4927                 return -ENOMEM;
4928         data->client = clp;
4929         data->timestamp = jiffies;
4930         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4931                         &nfs4_renew_ops, data);
4932 }
4933
4934 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4935 {
4936         struct rpc_message msg = {
4937                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4938                 .rpc_argp       = clp,
4939                 .rpc_cred       = cred,
4940         };
4941         unsigned long now = jiffies;
4942         int status;
4943
4944         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4945         if (status < 0)
4946                 return status;
4947         do_renew_lease(clp, now);
4948         return 0;
4949 }
4950
4951 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4952 {
4953         return server->caps & NFS_CAP_ACLS;
4954 }
4955
4956 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4957  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4958  * the stack.
4959  */
4960 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4961
4962 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4963                 struct page **pages)
4964 {
4965         struct page *newpage, **spages;
4966         int rc = 0;
4967         size_t len;
4968         spages = pages;
4969
4970         do {
4971                 len = min_t(size_t, PAGE_SIZE, buflen);
4972                 newpage = alloc_page(GFP_KERNEL);
4973
4974                 if (newpage == NULL)
4975                         goto unwind;
4976                 memcpy(page_address(newpage), buf, len);
4977                 buf += len;
4978                 buflen -= len;
4979                 *pages++ = newpage;
4980                 rc++;
4981         } while (buflen != 0);
4982
4983         return rc;
4984
4985 unwind:
4986         for(; rc > 0; rc--)
4987                 __free_page(spages[rc-1]);
4988         return -ENOMEM;
4989 }
4990
4991 struct nfs4_cached_acl {
4992         int cached;
4993         size_t len;
4994         char data[0];
4995 };
4996
4997 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4998 {
4999         struct nfs_inode *nfsi = NFS_I(inode);
5000
5001         spin_lock(&inode->i_lock);
5002         kfree(nfsi->nfs4_acl);
5003         nfsi->nfs4_acl = acl;
5004         spin_unlock(&inode->i_lock);
5005 }
5006
5007 static void nfs4_zap_acl_attr(struct inode *inode)
5008 {
5009         nfs4_set_cached_acl(inode, NULL);
5010 }
5011
5012 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5013 {
5014         struct nfs_inode *nfsi = NFS_I(inode);
5015         struct nfs4_cached_acl *acl;
5016         int ret = -ENOENT;
5017
5018         spin_lock(&inode->i_lock);
5019         acl = nfsi->nfs4_acl;
5020         if (acl == NULL)
5021                 goto out;
5022         if (buf == NULL) /* user is just asking for length */
5023                 goto out_len;
5024         if (acl->cached == 0)
5025                 goto out;
5026         ret = -ERANGE; /* see getxattr(2) man page */
5027         if (acl->len > buflen)
5028                 goto out;
5029         memcpy(buf, acl->data, acl->len);
5030 out_len:
5031         ret = acl->len;
5032 out:
5033         spin_unlock(&inode->i_lock);
5034         return ret;
5035 }
5036
5037 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5038 {
5039         struct nfs4_cached_acl *acl;
5040         size_t buflen = sizeof(*acl) + acl_len;
5041
5042         if (buflen <= PAGE_SIZE) {
5043                 acl = kmalloc(buflen, GFP_KERNEL);
5044                 if (acl == NULL)
5045                         goto out;
5046                 acl->cached = 1;
5047                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5048         } else {
5049                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5050                 if (acl == NULL)
5051                         goto out;
5052                 acl->cached = 0;
5053         }
5054         acl->len = acl_len;
5055 out:
5056         nfs4_set_cached_acl(inode, acl);
5057 }
5058
5059 /*
5060  * The getxattr API returns the required buffer length when called with a
5061  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5062  * the required buf.  On a NULL buf, we send a page of data to the server
5063  * guessing that the ACL request can be serviced by a page. If so, we cache
5064  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5065  * the cache. If not so, we throw away the page, and cache the required
5066  * length. The next getxattr call will then produce another round trip to
5067  * the server, this time with the input buf of the required size.
5068  */
5069 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5070 {
5071         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
5072         struct nfs_getaclargs args = {
5073                 .fh = NFS_FH(inode),
5074                 .acl_pages = pages,
5075                 .acl_len = buflen,
5076         };
5077         struct nfs_getaclres res = {
5078                 .acl_len = buflen,
5079         };
5080         struct rpc_message msg = {
5081                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5082                 .rpc_argp = &args,
5083                 .rpc_resp = &res,
5084         };
5085         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5086         int ret = -ENOMEM, i;
5087
5088         /* As long as we're doing a round trip to the server anyway,
5089          * let's be prepared for a page of acl data. */
5090         if (npages == 0)
5091                 npages = 1;
5092         if (npages > ARRAY_SIZE(pages))
5093                 return -ERANGE;
5094
5095         for (i = 0; i < npages; i++) {
5096                 pages[i] = alloc_page(GFP_KERNEL);
5097                 if (!pages[i])
5098                         goto out_free;
5099         }
5100
5101         /* for decoding across pages */
5102         res.acl_scratch = alloc_page(GFP_KERNEL);
5103         if (!res.acl_scratch)
5104                 goto out_free;
5105
5106         args.acl_len = npages * PAGE_SIZE;
5107
5108         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5109                 __func__, buf, buflen, npages, args.acl_len);
5110         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5111                              &msg, &args.seq_args, &res.seq_res, 0);
5112         if (ret)
5113                 goto out_free;
5114
5115         /* Handle the case where the passed-in buffer is too short */
5116         if (res.acl_flags & NFS4_ACL_TRUNC) {
5117                 /* Did the user only issue a request for the acl length? */
5118                 if (buf == NULL)
5119                         goto out_ok;
5120                 ret = -ERANGE;
5121                 goto out_free;
5122         }
5123         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5124         if (buf) {
5125                 if (res.acl_len > buflen) {
5126                         ret = -ERANGE;
5127                         goto out_free;
5128                 }
5129                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5130         }
5131 out_ok:
5132         ret = res.acl_len;
5133 out_free:
5134         for (i = 0; i < npages; i++)
5135                 if (pages[i])
5136                         __free_page(pages[i]);
5137         if (res.acl_scratch)
5138                 __free_page(res.acl_scratch);
5139         return ret;
5140 }
5141
5142 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5143 {
5144         struct nfs4_exception exception = { };
5145         ssize_t ret;
5146         do {
5147                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5148                 trace_nfs4_get_acl(inode, ret);
5149                 if (ret >= 0)
5150                         break;
5151                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5152         } while (exception.retry);
5153         return ret;
5154 }
5155
5156 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5157 {
5158         struct nfs_server *server = NFS_SERVER(inode);
5159         int ret;
5160
5161         if (!nfs4_server_supports_acls(server))
5162                 return -EOPNOTSUPP;
5163         ret = nfs_revalidate_inode(server, inode);
5164         if (ret < 0)
5165                 return ret;
5166         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5167                 nfs_zap_acl_cache(inode);
5168         ret = nfs4_read_cached_acl(inode, buf, buflen);
5169         if (ret != -ENOENT)
5170                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5171                  * but no cached acl data, just the acl length */
5172                 return ret;
5173         return nfs4_get_acl_uncached(inode, buf, buflen);
5174 }
5175
5176 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5177 {
5178         struct nfs_server *server = NFS_SERVER(inode);
5179         struct page *pages[NFS4ACL_MAXPAGES];
5180         struct nfs_setaclargs arg = {
5181                 .fh             = NFS_FH(inode),
5182                 .acl_pages      = pages,
5183                 .acl_len        = buflen,
5184         };
5185         struct nfs_setaclres res;
5186         struct rpc_message msg = {
5187                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5188                 .rpc_argp       = &arg,
5189                 .rpc_resp       = &res,
5190         };
5191         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5192         int ret, i;
5193
5194         if (!nfs4_server_supports_acls(server))
5195                 return -EOPNOTSUPP;
5196         if (npages > ARRAY_SIZE(pages))
5197                 return -ERANGE;
5198         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5199         if (i < 0)
5200                 return i;
5201         nfs4_inode_return_delegation(inode);
5202         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5203
5204         /*
5205          * Free each page after tx, so the only ref left is
5206          * held by the network stack
5207          */
5208         for (; i > 0; i--)
5209                 put_page(pages[i-1]);
5210
5211         /*
5212          * Acl update can result in inode attribute update.
5213          * so mark the attribute cache invalid.
5214          */
5215         spin_lock(&inode->i_lock);
5216         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5217         spin_unlock(&inode->i_lock);
5218         nfs_access_zap_cache(inode);
5219         nfs_zap_acl_cache(inode);
5220         return ret;
5221 }
5222
5223 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5224 {
5225         struct nfs4_exception exception = { };
5226         int err;
5227         do {
5228                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5229                 trace_nfs4_set_acl(inode, err);
5230                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5231                                 &exception);
5232         } while (exception.retry);
5233         return err;
5234 }
5235
5236 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5237 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5238                                         size_t buflen)
5239 {
5240         struct nfs_server *server = NFS_SERVER(inode);
5241         struct nfs_fattr fattr;
5242         struct nfs4_label label = {0, 0, buflen, buf};
5243
5244         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5245         struct nfs4_getattr_arg arg = {
5246                 .fh             = NFS_FH(inode),
5247                 .bitmask        = bitmask,
5248         };
5249         struct nfs4_getattr_res res = {
5250                 .fattr          = &fattr,
5251                 .label          = &label,
5252                 .server         = server,
5253         };
5254         struct rpc_message msg = {
5255                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5256                 .rpc_argp       = &arg,
5257                 .rpc_resp       = &res,
5258         };
5259         int ret;
5260
5261         nfs_fattr_init(&fattr);
5262
5263         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5264         if (ret)
5265                 return ret;
5266         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5267                 return -ENOENT;
5268         if (buflen < label.len)
5269                 return -ERANGE;
5270         return 0;
5271 }
5272
5273 static int nfs4_get_security_label(struct inode *inode, void *buf,
5274                                         size_t buflen)
5275 {
5276         struct nfs4_exception exception = { };
5277         int err;
5278
5279         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5280                 return -EOPNOTSUPP;
5281
5282         do {
5283                 err = _nfs4_get_security_label(inode, buf, buflen);
5284                 trace_nfs4_get_security_label(inode, err);
5285                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5286                                 &exception);
5287         } while (exception.retry);
5288         return err;
5289 }
5290
5291 static int _nfs4_do_set_security_label(struct inode *inode,
5292                 struct nfs4_label *ilabel,
5293                 struct nfs_fattr *fattr,
5294                 struct nfs4_label *olabel)
5295 {
5296
5297         struct iattr sattr = {0};
5298         struct nfs_server *server = NFS_SERVER(inode);
5299         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5300         struct nfs_setattrargs arg = {
5301                 .fh             = NFS_FH(inode),
5302                 .iap            = &sattr,
5303                 .server         = server,
5304                 .bitmask        = bitmask,
5305                 .label          = ilabel,
5306         };
5307         struct nfs_setattrres res = {
5308                 .fattr          = fattr,
5309                 .label          = olabel,
5310                 .server         = server,
5311         };
5312         struct rpc_message msg = {
5313                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5314                 .rpc_argp       = &arg,
5315                 .rpc_resp       = &res,
5316         };
5317         int status;
5318
5319         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5320
5321         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5322         if (status)
5323                 dprintk("%s failed: %d\n", __func__, status);
5324
5325         return status;
5326 }
5327
5328 static int nfs4_do_set_security_label(struct inode *inode,
5329                 struct nfs4_label *ilabel,
5330                 struct nfs_fattr *fattr,
5331                 struct nfs4_label *olabel)
5332 {
5333         struct nfs4_exception exception = { };
5334         int err;
5335
5336         do {
5337                 err = _nfs4_do_set_security_label(inode, ilabel,
5338                                 fattr, olabel);
5339                 trace_nfs4_set_security_label(inode, err);
5340                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5341                                 &exception);
5342         } while (exception.retry);
5343         return err;
5344 }
5345
5346 static int
5347 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5348 {
5349         struct nfs4_label ilabel, *olabel = NULL;
5350         struct nfs_fattr fattr;
5351         struct rpc_cred *cred;
5352         int status;
5353
5354         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5355                 return -EOPNOTSUPP;
5356
5357         nfs_fattr_init(&fattr);
5358
5359         ilabel.pi = 0;
5360         ilabel.lfs = 0;
5361         ilabel.label = (char *)buf;
5362         ilabel.len = buflen;
5363
5364         cred = rpc_lookup_cred();
5365         if (IS_ERR(cred))
5366                 return PTR_ERR(cred);
5367
5368         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5369         if (IS_ERR(olabel)) {
5370                 status = -PTR_ERR(olabel);
5371                 goto out;
5372         }
5373
5374         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5375         if (status == 0)
5376                 nfs_setsecurity(inode, &fattr, olabel);
5377
5378         nfs4_label_free(olabel);
5379 out:
5380         put_rpccred(cred);
5381         return status;
5382 }
5383 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5384
5385
5386 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5387                                     nfs4_verifier *bootverf)
5388 {
5389         __be32 verf[2];
5390
5391         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5392                 /* An impossible timestamp guarantees this value
5393                  * will never match a generated boot time. */
5394                 verf[0] = cpu_to_be32(U32_MAX);
5395                 verf[1] = cpu_to_be32(U32_MAX);
5396         } else {
5397                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5398                 u64 ns = ktime_to_ns(nn->boot_time);
5399
5400                 verf[0] = cpu_to_be32(ns >> 32);
5401                 verf[1] = cpu_to_be32(ns);
5402         }
5403         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5404 }
5405
5406 static int
5407 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5408 {
5409         size_t len;
5410         char *str;
5411
5412         if (clp->cl_owner_id != NULL)
5413                 return 0;
5414
5415         rcu_read_lock();
5416         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5417                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5418                 1 +
5419                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5420                 1;
5421         rcu_read_unlock();
5422
5423         if (len > NFS4_OPAQUE_LIMIT + 1)
5424                 return -EINVAL;
5425
5426         /*
5427          * Since this string is allocated at mount time, and held until the
5428          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5429          * about a memory-reclaim deadlock.
5430          */
5431         str = kmalloc(len, GFP_KERNEL);
5432         if (!str)
5433                 return -ENOMEM;
5434
5435         rcu_read_lock();
5436         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5437                         clp->cl_ipaddr,
5438                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5439                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5440         rcu_read_unlock();
5441
5442         clp->cl_owner_id = str;
5443         return 0;
5444 }
5445
5446 static int
5447 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5448 {
5449         size_t len;
5450         char *str;
5451
5452         len = 10 + 10 + 1 + 10 + 1 +
5453                 strlen(nfs4_client_id_uniquifier) + 1 +
5454                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5455
5456         if (len > NFS4_OPAQUE_LIMIT + 1)
5457                 return -EINVAL;
5458
5459         /*
5460          * Since this string is allocated at mount time, and held until the
5461          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5462          * about a memory-reclaim deadlock.
5463          */
5464         str = kmalloc(len, GFP_KERNEL);
5465         if (!str)
5466                 return -ENOMEM;
5467
5468         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5469                         clp->rpc_ops->version, clp->cl_minorversion,
5470                         nfs4_client_id_uniquifier,
5471                         clp->cl_rpcclient->cl_nodename);
5472         clp->cl_owner_id = str;
5473         return 0;
5474 }
5475
5476 static int
5477 nfs4_init_uniform_client_string(struct nfs_client *clp)
5478 {
5479         size_t len;
5480         char *str;
5481
5482         if (clp->cl_owner_id != NULL)
5483                 return 0;
5484
5485         if (nfs4_client_id_uniquifier[0] != '\0')
5486                 return nfs4_init_uniquifier_client_string(clp);
5487
5488         len = 10 + 10 + 1 + 10 + 1 +
5489                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5490
5491         if (len > NFS4_OPAQUE_LIMIT + 1)
5492                 return -EINVAL;
5493
5494         /*
5495          * Since this string is allocated at mount time, and held until the
5496          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5497          * about a memory-reclaim deadlock.
5498          */
5499         str = kmalloc(len, GFP_KERNEL);
5500         if (!str)
5501                 return -ENOMEM;
5502
5503         scnprintf(str, len, "Linux NFSv%u.%u %s",
5504                         clp->rpc_ops->version, clp->cl_minorversion,
5505                         clp->cl_rpcclient->cl_nodename);
5506         clp->cl_owner_id = str;
5507         return 0;
5508 }
5509
5510 /*
5511  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5512  * services.  Advertise one based on the address family of the
5513  * clientaddr.
5514  */
5515 static unsigned int
5516 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5517 {
5518         if (strchr(clp->cl_ipaddr, ':') != NULL)
5519                 return scnprintf(buf, len, "tcp6");
5520         else
5521                 return scnprintf(buf, len, "tcp");
5522 }
5523
5524 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5525 {
5526         struct nfs4_setclientid *sc = calldata;
5527
5528         if (task->tk_status == 0)
5529                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5530 }
5531
5532 static const struct rpc_call_ops nfs4_setclientid_ops = {
5533         .rpc_call_done = nfs4_setclientid_done,
5534 };
5535
5536 /**
5537  * nfs4_proc_setclientid - Negotiate client ID
5538  * @clp: state data structure
5539  * @program: RPC program for NFSv4 callback service
5540  * @port: IP port number for NFS4 callback service
5541  * @cred: RPC credential to use for this call
5542  * @res: where to place the result
5543  *
5544  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5545  */
5546 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5547                 unsigned short port, struct rpc_cred *cred,
5548                 struct nfs4_setclientid_res *res)
5549 {
5550         nfs4_verifier sc_verifier;
5551         struct nfs4_setclientid setclientid = {
5552                 .sc_verifier = &sc_verifier,
5553                 .sc_prog = program,
5554                 .sc_clnt = clp,
5555         };
5556         struct rpc_message msg = {
5557                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5558                 .rpc_argp = &setclientid,
5559                 .rpc_resp = res,
5560                 .rpc_cred = cred,
5561         };
5562         struct rpc_task *task;
5563         struct rpc_task_setup task_setup_data = {
5564                 .rpc_client = clp->cl_rpcclient,
5565                 .rpc_message = &msg,
5566                 .callback_ops = &nfs4_setclientid_ops,
5567                 .callback_data = &setclientid,
5568                 .flags = RPC_TASK_TIMEOUT,
5569         };
5570         int status;
5571
5572         /* nfs_client_id4 */
5573         nfs4_init_boot_verifier(clp, &sc_verifier);
5574
5575         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5576                 status = nfs4_init_uniform_client_string(clp);
5577         else
5578                 status = nfs4_init_nonuniform_client_string(clp);
5579
5580         if (status)
5581                 goto out;
5582
5583         /* cb_client4 */
5584         setclientid.sc_netid_len =
5585                                 nfs4_init_callback_netid(clp,
5586                                                 setclientid.sc_netid,
5587                                                 sizeof(setclientid.sc_netid));
5588         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5589                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5590                                 clp->cl_ipaddr, port >> 8, port & 255);
5591
5592         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5593                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5594                 clp->cl_owner_id);
5595         task = rpc_run_task(&task_setup_data);
5596         if (IS_ERR(task)) {
5597                 status = PTR_ERR(task);
5598                 goto out;
5599         }
5600         status = task->tk_status;
5601         if (setclientid.sc_cred) {
5602                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5603                 put_rpccred(setclientid.sc_cred);
5604         }
5605         rpc_put_task(task);
5606 out:
5607         trace_nfs4_setclientid(clp, status);
5608         dprintk("NFS reply setclientid: %d\n", status);
5609         return status;
5610 }
5611
5612 /**
5613  * nfs4_proc_setclientid_confirm - Confirm client ID
5614  * @clp: state data structure
5615  * @res: result of a previous SETCLIENTID
5616  * @cred: RPC credential to use for this call
5617  *
5618  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5619  */
5620 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5621                 struct nfs4_setclientid_res *arg,
5622                 struct rpc_cred *cred)
5623 {
5624         struct rpc_message msg = {
5625                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5626                 .rpc_argp = arg,
5627                 .rpc_cred = cred,
5628         };
5629         int status;
5630
5631         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5632                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5633                 clp->cl_clientid);
5634         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5635         trace_nfs4_setclientid_confirm(clp, status);
5636         dprintk("NFS reply setclientid_confirm: %d\n", status);
5637         return status;
5638 }
5639
5640 struct nfs4_delegreturndata {
5641         struct nfs4_delegreturnargs args;
5642         struct nfs4_delegreturnres res;
5643         struct nfs_fh fh;
5644         nfs4_stateid stateid;
5645         unsigned long timestamp;
5646         struct {
5647                 struct nfs4_layoutreturn_args arg;
5648                 struct nfs4_layoutreturn_res res;
5649                 struct nfs4_xdr_opaque_data ld_private;
5650                 u32 roc_barrier;
5651                 bool roc;
5652         } lr;
5653         struct nfs_fattr fattr;
5654         int rpc_status;
5655         struct inode *inode;
5656 };
5657
5658 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5659 {
5660         struct nfs4_delegreturndata *data = calldata;
5661
5662         if (!nfs4_sequence_done(task, &data->res.seq_res))
5663                 return;
5664
5665         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5666
5667         /* Handle Layoutreturn errors */
5668         if (data->args.lr_args && task->tk_status != 0) {
5669                 switch(data->res.lr_ret) {
5670                 default:
5671                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5672                         break;
5673                 case 0:
5674                         data->args.lr_args = NULL;
5675                         data->res.lr_res = NULL;
5676                         break;
5677                 case -NFS4ERR_ADMIN_REVOKED:
5678                 case -NFS4ERR_DELEG_REVOKED:
5679                 case -NFS4ERR_EXPIRED:
5680                 case -NFS4ERR_BAD_STATEID:
5681                 case -NFS4ERR_OLD_STATEID:
5682                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5683                 case -NFS4ERR_WRONG_CRED:
5684                         data->args.lr_args = NULL;
5685                         data->res.lr_res = NULL;
5686                         data->res.lr_ret = 0;
5687                         rpc_restart_call_prepare(task);
5688                         return;
5689                 }
5690         }
5691
5692         switch (task->tk_status) {
5693         case 0:
5694                 renew_lease(data->res.server, data->timestamp);
5695                 break;
5696         case -NFS4ERR_ADMIN_REVOKED:
5697         case -NFS4ERR_DELEG_REVOKED:
5698         case -NFS4ERR_EXPIRED:
5699                 nfs4_free_revoked_stateid(data->res.server,
5700                                 data->args.stateid,
5701                                 task->tk_msg.rpc_cred);
5702         case -NFS4ERR_BAD_STATEID:
5703         case -NFS4ERR_OLD_STATEID:
5704         case -NFS4ERR_STALE_STATEID:
5705                 task->tk_status = 0;
5706                 break;
5707         case -NFS4ERR_ACCESS:
5708                 if (data->args.bitmask) {
5709                         data->args.bitmask = NULL;
5710                         data->res.fattr = NULL;
5711                         task->tk_status = 0;
5712                         rpc_restart_call_prepare(task);
5713                         return;
5714                 }
5715         default:
5716                 if (nfs4_async_handle_error(task, data->res.server,
5717                                             NULL, NULL) == -EAGAIN) {
5718                         rpc_restart_call_prepare(task);
5719                         return;
5720                 }
5721         }
5722         data->rpc_status = task->tk_status;
5723 }
5724
5725 static void nfs4_delegreturn_release(void *calldata)
5726 {
5727         struct nfs4_delegreturndata *data = calldata;
5728         struct inode *inode = data->inode;
5729
5730         if (inode) {
5731                 if (data->lr.roc)
5732                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
5733                                         data->res.lr_ret);
5734                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5735                 nfs_iput_and_deactive(inode);
5736         }
5737         kfree(calldata);
5738 }
5739
5740 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5741 {
5742         struct nfs4_delegreturndata *d_data;
5743
5744         d_data = (struct nfs4_delegreturndata *)data;
5745
5746         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5747                 return;
5748
5749         nfs4_setup_sequence(d_data->res.server,
5750                         &d_data->args.seq_args,
5751                         &d_data->res.seq_res,
5752                         task);
5753 }
5754
5755 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5756         .rpc_call_prepare = nfs4_delegreturn_prepare,
5757         .rpc_call_done = nfs4_delegreturn_done,
5758         .rpc_release = nfs4_delegreturn_release,
5759 };
5760
5761 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5762 {
5763         struct nfs4_delegreturndata *data;
5764         struct nfs_server *server = NFS_SERVER(inode);
5765         struct rpc_task *task;
5766         struct rpc_message msg = {
5767                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5768                 .rpc_cred = cred,
5769         };
5770         struct rpc_task_setup task_setup_data = {
5771                 .rpc_client = server->client,
5772                 .rpc_message = &msg,
5773                 .callback_ops = &nfs4_delegreturn_ops,
5774                 .flags = RPC_TASK_ASYNC,
5775         };
5776         int status = 0;
5777
5778         data = kzalloc(sizeof(*data), GFP_NOFS);
5779         if (data == NULL)
5780                 return -ENOMEM;
5781         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5782
5783         nfs4_state_protect(server->nfs_client,
5784                         NFS_SP4_MACH_CRED_CLEANUP,
5785                         &task_setup_data.rpc_client, &msg);
5786
5787         data->args.fhandle = &data->fh;
5788         data->args.stateid = &data->stateid;
5789         data->args.bitmask = server->cache_consistency_bitmask;
5790         nfs_copy_fh(&data->fh, NFS_FH(inode));
5791         nfs4_stateid_copy(&data->stateid, stateid);
5792         data->res.fattr = &data->fattr;
5793         data->res.server = server;
5794         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5795         data->lr.arg.ld_private = &data->lr.ld_private;
5796         nfs_fattr_init(data->res.fattr);
5797         data->timestamp = jiffies;
5798         data->rpc_status = 0;
5799         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5800         data->inode = nfs_igrab_and_active(inode);
5801         if (data->inode) {
5802                 if (data->lr.roc) {
5803                         data->args.lr_args = &data->lr.arg;
5804                         data->res.lr_res = &data->lr.res;
5805                 }
5806         } else if (data->lr.roc) {
5807                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5808                 data->lr.roc = false;
5809         }
5810
5811         task_setup_data.callback_data = data;
5812         msg.rpc_argp = &data->args;
5813         msg.rpc_resp = &data->res;
5814         task = rpc_run_task(&task_setup_data);
5815         if (IS_ERR(task))
5816                 return PTR_ERR(task);
5817         if (!issync)
5818                 goto out;
5819         status = nfs4_wait_for_completion_rpc_task(task);
5820         if (status != 0)
5821                 goto out;
5822         status = data->rpc_status;
5823 out:
5824         rpc_put_task(task);
5825         return status;
5826 }
5827
5828 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5829 {
5830         struct nfs_server *server = NFS_SERVER(inode);
5831         struct nfs4_exception exception = { };
5832         int err;
5833         do {
5834                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5835                 trace_nfs4_delegreturn(inode, stateid, err);
5836                 switch (err) {
5837                         case -NFS4ERR_STALE_STATEID:
5838                         case -NFS4ERR_EXPIRED:
5839                         case 0:
5840                                 return 0;
5841                 }
5842                 err = nfs4_handle_exception(server, err, &exception);
5843         } while (exception.retry);
5844         return err;
5845 }
5846
5847 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5848 {
5849         struct inode *inode = state->inode;
5850         struct nfs_server *server = NFS_SERVER(inode);
5851         struct nfs_client *clp = server->nfs_client;
5852         struct nfs_lockt_args arg = {
5853                 .fh = NFS_FH(inode),
5854                 .fl = request,
5855         };
5856         struct nfs_lockt_res res = {
5857                 .denied = request,
5858         };
5859         struct rpc_message msg = {
5860                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5861                 .rpc_argp       = &arg,
5862                 .rpc_resp       = &res,
5863                 .rpc_cred       = state->owner->so_cred,
5864         };
5865         struct nfs4_lock_state *lsp;
5866         int status;
5867
5868         arg.lock_owner.clientid = clp->cl_clientid;
5869         status = nfs4_set_lock_state(state, request);
5870         if (status != 0)
5871                 goto out;
5872         lsp = request->fl_u.nfs4_fl.owner;
5873         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5874         arg.lock_owner.s_dev = server->s_dev;
5875         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5876         switch (status) {
5877                 case 0:
5878                         request->fl_type = F_UNLCK;
5879                         break;
5880                 case -NFS4ERR_DENIED:
5881                         status = 0;
5882         }
5883         request->fl_ops->fl_release_private(request);
5884         request->fl_ops = NULL;
5885 out:
5886         return status;
5887 }
5888
5889 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5890 {
5891         struct nfs4_exception exception = { };
5892         int err;
5893
5894         do {
5895                 err = _nfs4_proc_getlk(state, cmd, request);
5896                 trace_nfs4_get_lock(request, state, cmd, err);
5897                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5898                                 &exception);
5899         } while (exception.retry);
5900         return err;
5901 }
5902
5903 struct nfs4_unlockdata {
5904         struct nfs_locku_args arg;
5905         struct nfs_locku_res res;
5906         struct nfs4_lock_state *lsp;
5907         struct nfs_open_context *ctx;
5908         struct file_lock fl;
5909         struct nfs_server *server;
5910         unsigned long timestamp;
5911 };
5912
5913 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5914                 struct nfs_open_context *ctx,
5915                 struct nfs4_lock_state *lsp,
5916                 struct nfs_seqid *seqid)
5917 {
5918         struct nfs4_unlockdata *p;
5919         struct inode *inode = lsp->ls_state->inode;
5920
5921         p = kzalloc(sizeof(*p), GFP_NOFS);
5922         if (p == NULL)
5923                 return NULL;
5924         p->arg.fh = NFS_FH(inode);
5925         p->arg.fl = &p->fl;
5926         p->arg.seqid = seqid;
5927         p->res.seqid = seqid;
5928         p->lsp = lsp;
5929         atomic_inc(&lsp->ls_count);
5930         /* Ensure we don't close file until we're done freeing locks! */
5931         p->ctx = get_nfs_open_context(ctx);
5932         memcpy(&p->fl, fl, sizeof(p->fl));
5933         p->server = NFS_SERVER(inode);
5934         return p;
5935 }
5936
5937 static void nfs4_locku_release_calldata(void *data)
5938 {
5939         struct nfs4_unlockdata *calldata = data;
5940         nfs_free_seqid(calldata->arg.seqid);
5941         nfs4_put_lock_state(calldata->lsp);
5942         put_nfs_open_context(calldata->ctx);
5943         kfree(calldata);
5944 }
5945
5946 static void nfs4_locku_done(struct rpc_task *task, void *data)
5947 {
5948         struct nfs4_unlockdata *calldata = data;
5949
5950         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5951                 return;
5952         switch (task->tk_status) {
5953                 case 0:
5954                         renew_lease(calldata->server, calldata->timestamp);
5955                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5956                         if (nfs4_update_lock_stateid(calldata->lsp,
5957                                         &calldata->res.stateid))
5958                                 break;
5959                 case -NFS4ERR_ADMIN_REVOKED:
5960                 case -NFS4ERR_EXPIRED:
5961                         nfs4_free_revoked_stateid(calldata->server,
5962                                         &calldata->arg.stateid,
5963                                         task->tk_msg.rpc_cred);
5964                 case -NFS4ERR_BAD_STATEID:
5965                 case -NFS4ERR_OLD_STATEID:
5966                 case -NFS4ERR_STALE_STATEID:
5967                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5968                                                 &calldata->lsp->ls_stateid))
5969                                 rpc_restart_call_prepare(task);
5970                         break;
5971                 default:
5972                         if (nfs4_async_handle_error(task, calldata->server,
5973                                                     NULL, NULL) == -EAGAIN)
5974                                 rpc_restart_call_prepare(task);
5975         }
5976         nfs_release_seqid(calldata->arg.seqid);
5977 }
5978
5979 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5980 {
5981         struct nfs4_unlockdata *calldata = data;
5982
5983         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5984                 goto out_wait;
5985         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5986         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5987                 /* Note: exit _without_ running nfs4_locku_done */
5988                 goto out_no_action;
5989         }
5990         calldata->timestamp = jiffies;
5991         if (nfs4_setup_sequence(calldata->server,
5992                                 &calldata->arg.seq_args,
5993                                 &calldata->res.seq_res,
5994                                 task) != 0)
5995                 nfs_release_seqid(calldata->arg.seqid);
5996         return;
5997 out_no_action:
5998         task->tk_action = NULL;
5999 out_wait:
6000         nfs4_sequence_done(task, &calldata->res.seq_res);
6001 }
6002
6003 static const struct rpc_call_ops nfs4_locku_ops = {
6004         .rpc_call_prepare = nfs4_locku_prepare,
6005         .rpc_call_done = nfs4_locku_done,
6006         .rpc_release = nfs4_locku_release_calldata,
6007 };
6008
6009 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6010                 struct nfs_open_context *ctx,
6011                 struct nfs4_lock_state *lsp,
6012                 struct nfs_seqid *seqid)
6013 {
6014         struct nfs4_unlockdata *data;
6015         struct rpc_message msg = {
6016                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6017                 .rpc_cred = ctx->cred,
6018         };
6019         struct rpc_task_setup task_setup_data = {
6020                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6021                 .rpc_message = &msg,
6022                 .callback_ops = &nfs4_locku_ops,
6023                 .workqueue = nfsiod_workqueue,
6024                 .flags = RPC_TASK_ASYNC,
6025         };
6026
6027         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6028                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6029
6030         /* Ensure this is an unlock - when canceling a lock, the
6031          * canceled lock is passed in, and it won't be an unlock.
6032          */
6033         fl->fl_type = F_UNLCK;
6034
6035         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6036         if (data == NULL) {
6037                 nfs_free_seqid(seqid);
6038                 return ERR_PTR(-ENOMEM);
6039         }
6040
6041         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6042         msg.rpc_argp = &data->arg;
6043         msg.rpc_resp = &data->res;
6044         task_setup_data.callback_data = data;
6045         return rpc_run_task(&task_setup_data);
6046 }
6047
6048 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6049 {
6050         struct inode *inode = state->inode;
6051         struct nfs4_state_owner *sp = state->owner;
6052         struct nfs_inode *nfsi = NFS_I(inode);
6053         struct nfs_seqid *seqid;
6054         struct nfs4_lock_state *lsp;
6055         struct rpc_task *task;
6056         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6057         int status = 0;
6058         unsigned char fl_flags = request->fl_flags;
6059
6060         status = nfs4_set_lock_state(state, request);
6061         /* Unlock _before_ we do the RPC call */
6062         request->fl_flags |= FL_EXISTS;
6063         /* Exclude nfs_delegation_claim_locks() */
6064         mutex_lock(&sp->so_delegreturn_mutex);
6065         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6066         down_read(&nfsi->rwsem);
6067         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6068                 up_read(&nfsi->rwsem);
6069                 mutex_unlock(&sp->so_delegreturn_mutex);
6070                 goto out;
6071         }
6072         up_read(&nfsi->rwsem);
6073         mutex_unlock(&sp->so_delegreturn_mutex);
6074         if (status != 0)
6075                 goto out;
6076         /* Is this a delegated lock? */
6077         lsp = request->fl_u.nfs4_fl.owner;
6078         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6079                 goto out;
6080         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6081         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6082         status = -ENOMEM;
6083         if (IS_ERR(seqid))
6084                 goto out;
6085         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6086         status = PTR_ERR(task);
6087         if (IS_ERR(task))
6088                 goto out;
6089         status = nfs4_wait_for_completion_rpc_task(task);
6090         rpc_put_task(task);
6091 out:
6092         request->fl_flags = fl_flags;
6093         trace_nfs4_unlock(request, state, F_SETLK, status);
6094         return status;
6095 }
6096
6097 struct nfs4_lockdata {
6098         struct nfs_lock_args arg;
6099         struct nfs_lock_res res;
6100         struct nfs4_lock_state *lsp;
6101         struct nfs_open_context *ctx;
6102         struct file_lock fl;
6103         unsigned long timestamp;
6104         int rpc_status;
6105         int cancelled;
6106         struct nfs_server *server;
6107 };
6108
6109 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6110                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6111                 gfp_t gfp_mask)
6112 {
6113         struct nfs4_lockdata *p;
6114         struct inode *inode = lsp->ls_state->inode;
6115         struct nfs_server *server = NFS_SERVER(inode);
6116         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6117
6118         p = kzalloc(sizeof(*p), gfp_mask);
6119         if (p == NULL)
6120                 return NULL;
6121
6122         p->arg.fh = NFS_FH(inode);
6123         p->arg.fl = &p->fl;
6124         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6125         if (IS_ERR(p->arg.open_seqid))
6126                 goto out_free;
6127         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6128         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6129         if (IS_ERR(p->arg.lock_seqid))
6130                 goto out_free_seqid;
6131         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6132         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6133         p->arg.lock_owner.s_dev = server->s_dev;
6134         p->res.lock_seqid = p->arg.lock_seqid;
6135         p->lsp = lsp;
6136         p->server = server;
6137         atomic_inc(&lsp->ls_count);
6138         p->ctx = get_nfs_open_context(ctx);
6139         memcpy(&p->fl, fl, sizeof(p->fl));
6140         return p;
6141 out_free_seqid:
6142         nfs_free_seqid(p->arg.open_seqid);
6143 out_free:
6144         kfree(p);
6145         return NULL;
6146 }
6147
6148 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6149 {
6150         struct nfs4_lockdata *data = calldata;
6151         struct nfs4_state *state = data->lsp->ls_state;
6152
6153         dprintk("%s: begin!\n", __func__);
6154         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6155                 goto out_wait;
6156         /* Do we need to do an open_to_lock_owner? */
6157         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6158                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6159                         goto out_release_lock_seqid;
6160                 }
6161                 nfs4_stateid_copy(&data->arg.open_stateid,
6162                                 &state->open_stateid);
6163                 data->arg.new_lock_owner = 1;
6164                 data->res.open_seqid = data->arg.open_seqid;
6165         } else {
6166                 data->arg.new_lock_owner = 0;
6167                 nfs4_stateid_copy(&data->arg.lock_stateid,
6168                                 &data->lsp->ls_stateid);
6169         }
6170         if (!nfs4_valid_open_stateid(state)) {
6171                 data->rpc_status = -EBADF;
6172                 task->tk_action = NULL;
6173                 goto out_release_open_seqid;
6174         }
6175         data->timestamp = jiffies;
6176         if (nfs4_setup_sequence(data->server,
6177                                 &data->arg.seq_args,
6178                                 &data->res.seq_res,
6179                                 task) == 0)
6180                 return;
6181 out_release_open_seqid:
6182         nfs_release_seqid(data->arg.open_seqid);
6183 out_release_lock_seqid:
6184         nfs_release_seqid(data->arg.lock_seqid);
6185 out_wait:
6186         nfs4_sequence_done(task, &data->res.seq_res);
6187         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6188 }
6189
6190 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6191 {
6192         struct nfs4_lockdata *data = calldata;
6193         struct nfs4_lock_state *lsp = data->lsp;
6194
6195         dprintk("%s: begin!\n", __func__);
6196
6197         if (!nfs4_sequence_done(task, &data->res.seq_res))
6198                 return;
6199
6200         data->rpc_status = task->tk_status;
6201         switch (task->tk_status) {
6202         case 0:
6203                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6204                                 data->timestamp);
6205                 if (data->arg.new_lock) {
6206                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6207                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6208                                 rpc_restart_call_prepare(task);
6209                                 break;
6210                         }
6211                 }
6212                 if (data->arg.new_lock_owner != 0) {
6213                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6214                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6215                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6216                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6217                         rpc_restart_call_prepare(task);
6218                 break;
6219         case -NFS4ERR_BAD_STATEID:
6220         case -NFS4ERR_OLD_STATEID:
6221         case -NFS4ERR_STALE_STATEID:
6222         case -NFS4ERR_EXPIRED:
6223                 if (data->arg.new_lock_owner != 0) {
6224                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6225                                                 &lsp->ls_state->open_stateid))
6226                                 rpc_restart_call_prepare(task);
6227                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6228                                                 &lsp->ls_stateid))
6229                                 rpc_restart_call_prepare(task);
6230         }
6231         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6232 }
6233
6234 static void nfs4_lock_release(void *calldata)
6235 {
6236         struct nfs4_lockdata *data = calldata;
6237
6238         dprintk("%s: begin!\n", __func__);
6239         nfs_free_seqid(data->arg.open_seqid);
6240         if (data->cancelled != 0) {
6241                 struct rpc_task *task;
6242                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6243                                 data->arg.lock_seqid);
6244                 if (!IS_ERR(task))
6245                         rpc_put_task_async(task);
6246                 dprintk("%s: cancelling lock!\n", __func__);
6247         } else
6248                 nfs_free_seqid(data->arg.lock_seqid);
6249         nfs4_put_lock_state(data->lsp);
6250         put_nfs_open_context(data->ctx);
6251         kfree(data);
6252         dprintk("%s: done!\n", __func__);
6253 }
6254
6255 static const struct rpc_call_ops nfs4_lock_ops = {
6256         .rpc_call_prepare = nfs4_lock_prepare,
6257         .rpc_call_done = nfs4_lock_done,
6258         .rpc_release = nfs4_lock_release,
6259 };
6260
6261 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6262 {
6263         switch (error) {
6264         case -NFS4ERR_ADMIN_REVOKED:
6265         case -NFS4ERR_EXPIRED:
6266         case -NFS4ERR_BAD_STATEID:
6267                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6268                 if (new_lock_owner != 0 ||
6269                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6270                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6271                 break;
6272         case -NFS4ERR_STALE_STATEID:
6273                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6274                 nfs4_schedule_lease_recovery(server->nfs_client);
6275         };
6276 }
6277
6278 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6279 {
6280         struct nfs4_lockdata *data;
6281         struct rpc_task *task;
6282         struct rpc_message msg = {
6283                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6284                 .rpc_cred = state->owner->so_cred,
6285         };
6286         struct rpc_task_setup task_setup_data = {
6287                 .rpc_client = NFS_CLIENT(state->inode),
6288                 .rpc_message = &msg,
6289                 .callback_ops = &nfs4_lock_ops,
6290                 .workqueue = nfsiod_workqueue,
6291                 .flags = RPC_TASK_ASYNC,
6292         };
6293         int ret;
6294
6295         dprintk("%s: begin!\n", __func__);
6296         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6297                         fl->fl_u.nfs4_fl.owner,
6298                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6299         if (data == NULL)
6300                 return -ENOMEM;
6301         if (IS_SETLKW(cmd))
6302                 data->arg.block = 1;
6303         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6304         msg.rpc_argp = &data->arg;
6305         msg.rpc_resp = &data->res;
6306         task_setup_data.callback_data = data;
6307         if (recovery_type > NFS_LOCK_NEW) {
6308                 if (recovery_type == NFS_LOCK_RECLAIM)
6309                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6310                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6311         } else
6312                 data->arg.new_lock = 1;
6313         task = rpc_run_task(&task_setup_data);
6314         if (IS_ERR(task))
6315                 return PTR_ERR(task);
6316         ret = nfs4_wait_for_completion_rpc_task(task);
6317         if (ret == 0) {
6318                 ret = data->rpc_status;
6319                 if (ret)
6320                         nfs4_handle_setlk_error(data->server, data->lsp,
6321                                         data->arg.new_lock_owner, ret);
6322         } else
6323                 data->cancelled = 1;
6324         rpc_put_task(task);
6325         dprintk("%s: done, ret = %d!\n", __func__, ret);
6326         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6327         return ret;
6328 }
6329
6330 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6331 {
6332         struct nfs_server *server = NFS_SERVER(state->inode);
6333         struct nfs4_exception exception = {
6334                 .inode = state->inode,
6335         };
6336         int err;
6337
6338         do {
6339                 /* Cache the lock if possible... */
6340                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6341                         return 0;
6342                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6343                 if (err != -NFS4ERR_DELAY)
6344                         break;
6345                 nfs4_handle_exception(server, err, &exception);
6346         } while (exception.retry);
6347         return err;
6348 }
6349
6350 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6351 {
6352         struct nfs_server *server = NFS_SERVER(state->inode);
6353         struct nfs4_exception exception = {
6354                 .inode = state->inode,
6355         };
6356         int err;
6357
6358         err = nfs4_set_lock_state(state, request);
6359         if (err != 0)
6360                 return err;
6361         if (!recover_lost_locks) {
6362                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6363                 return 0;
6364         }
6365         do {
6366                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6367                         return 0;
6368                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6369                 switch (err) {
6370                 default:
6371                         goto out;
6372                 case -NFS4ERR_GRACE:
6373                 case -NFS4ERR_DELAY:
6374                         nfs4_handle_exception(server, err, &exception);
6375                         err = 0;
6376                 }
6377         } while (exception.retry);
6378 out:
6379         return err;
6380 }
6381
6382 #if defined(CONFIG_NFS_V4_1)
6383 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6384 {
6385         struct nfs4_lock_state *lsp;
6386         int status;
6387
6388         status = nfs4_set_lock_state(state, request);
6389         if (status != 0)
6390                 return status;
6391         lsp = request->fl_u.nfs4_fl.owner;
6392         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6393             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6394                 return 0;
6395         status = nfs4_lock_expired(state, request);
6396         return status;
6397 }
6398 #endif
6399
6400 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6401 {
6402         struct nfs_inode *nfsi = NFS_I(state->inode);
6403         struct nfs4_state_owner *sp = state->owner;
6404         unsigned char fl_flags = request->fl_flags;
6405         int status;
6406
6407         request->fl_flags |= FL_ACCESS;
6408         status = locks_lock_inode_wait(state->inode, request);
6409         if (status < 0)
6410                 goto out;
6411         mutex_lock(&sp->so_delegreturn_mutex);
6412         down_read(&nfsi->rwsem);
6413         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6414                 /* Yes: cache locks! */
6415                 /* ...but avoid races with delegation recall... */
6416                 request->fl_flags = fl_flags & ~FL_SLEEP;
6417                 status = locks_lock_inode_wait(state->inode, request);
6418                 up_read(&nfsi->rwsem);
6419                 mutex_unlock(&sp->so_delegreturn_mutex);
6420                 goto out;
6421         }
6422         up_read(&nfsi->rwsem);
6423         mutex_unlock(&sp->so_delegreturn_mutex);
6424         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6425 out:
6426         request->fl_flags = fl_flags;
6427         return status;
6428 }
6429
6430 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6431 {
6432         struct nfs4_exception exception = {
6433                 .state = state,
6434                 .inode = state->inode,
6435         };
6436         int err;
6437
6438         do {
6439                 err = _nfs4_proc_setlk(state, cmd, request);
6440                 if (err == -NFS4ERR_DENIED)
6441                         err = -EAGAIN;
6442                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6443                                 err, &exception);
6444         } while (exception.retry);
6445         return err;
6446 }
6447
6448 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6449 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6450
6451 static int
6452 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6453                         struct file_lock *request)
6454 {
6455         int             status = -ERESTARTSYS;
6456         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6457
6458         while(!signalled()) {
6459                 status = nfs4_proc_setlk(state, cmd, request);
6460                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6461                         break;
6462                 freezable_schedule_timeout_interruptible(timeout);
6463                 timeout *= 2;
6464                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6465                 status = -ERESTARTSYS;
6466         }
6467         return status;
6468 }
6469
6470 #ifdef CONFIG_NFS_V4_1
6471 struct nfs4_lock_waiter {
6472         struct task_struct      *task;
6473         struct inode            *inode;
6474         struct nfs_lowner       *owner;
6475         bool                    notified;
6476 };
6477
6478 static int
6479 nfs4_wake_lock_waiter(wait_queue_t *wait, unsigned int mode, int flags, void *key)
6480 {
6481         int ret;
6482         struct cb_notify_lock_args *cbnl = key;
6483         struct nfs4_lock_waiter *waiter = wait->private;
6484         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6485                                 *wowner = waiter->owner;
6486
6487         /* Only wake if the callback was for the same owner */
6488         if (lowner->clientid != wowner->clientid ||
6489             lowner->id != wowner->id             ||
6490             lowner->s_dev != wowner->s_dev)
6491                 return 0;
6492
6493         /* Make sure it's for the right inode */
6494         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6495                 return 0;
6496
6497         waiter->notified = true;
6498
6499         /* override "private" so we can use default_wake_function */
6500         wait->private = waiter->task;
6501         ret = autoremove_wake_function(wait, mode, flags, key);
6502         wait->private = waiter;
6503         return ret;
6504 }
6505
6506 static int
6507 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6508 {
6509         int status = -ERESTARTSYS;
6510         unsigned long flags;
6511         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6512         struct nfs_server *server = NFS_SERVER(state->inode);
6513         struct nfs_client *clp = server->nfs_client;
6514         wait_queue_head_t *q = &clp->cl_lock_waitq;
6515         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6516                                     .id = lsp->ls_seqid.owner_id,
6517                                     .s_dev = server->s_dev };
6518         struct nfs4_lock_waiter waiter = { .task  = current,
6519                                            .inode = state->inode,
6520                                            .owner = &owner,
6521                                            .notified = false };
6522         wait_queue_t wait;
6523
6524         /* Don't bother with waitqueue if we don't expect a callback */
6525         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6526                 return nfs4_retry_setlk_simple(state, cmd, request);
6527
6528         init_wait(&wait);
6529         wait.private = &waiter;
6530         wait.func = nfs4_wake_lock_waiter;
6531         add_wait_queue(q, &wait);
6532
6533         while(!signalled()) {
6534                 status = nfs4_proc_setlk(state, cmd, request);
6535                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6536                         break;
6537
6538                 status = -ERESTARTSYS;
6539                 spin_lock_irqsave(&q->lock, flags);
6540                 if (waiter.notified) {
6541                         spin_unlock_irqrestore(&q->lock, flags);
6542                         continue;
6543                 }
6544                 set_current_state(TASK_INTERRUPTIBLE);
6545                 spin_unlock_irqrestore(&q->lock, flags);
6546
6547                 freezable_schedule_timeout_interruptible(NFS4_LOCK_MAXTIMEOUT);
6548         }
6549
6550         finish_wait(q, &wait);
6551         return status;
6552 }
6553 #else /* !CONFIG_NFS_V4_1 */
6554 static inline int
6555 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6556 {
6557         return nfs4_retry_setlk_simple(state, cmd, request);
6558 }
6559 #endif
6560
6561 static int
6562 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6563 {
6564         struct nfs_open_context *ctx;
6565         struct nfs4_state *state;
6566         int status;
6567
6568         /* verify open state */
6569         ctx = nfs_file_open_context(filp);
6570         state = ctx->state;
6571
6572         if (request->fl_start < 0 || request->fl_end < 0)
6573                 return -EINVAL;
6574
6575         if (IS_GETLK(cmd)) {
6576                 if (state != NULL)
6577                         return nfs4_proc_getlk(state, F_GETLK, request);
6578                 return 0;
6579         }
6580
6581         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6582                 return -EINVAL;
6583
6584         if (request->fl_type == F_UNLCK) {
6585                 if (state != NULL)
6586                         return nfs4_proc_unlck(state, cmd, request);
6587                 return 0;
6588         }
6589
6590         if (state == NULL)
6591                 return -ENOLCK;
6592
6593         if ((request->fl_flags & FL_POSIX) &&
6594             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6595                 return -ENOLCK;
6596
6597         /*
6598          * Don't rely on the VFS having checked the file open mode,
6599          * since it won't do this for flock() locks.
6600          */
6601         switch (request->fl_type) {
6602         case F_RDLCK:
6603                 if (!(filp->f_mode & FMODE_READ))
6604                         return -EBADF;
6605                 break;
6606         case F_WRLCK:
6607                 if (!(filp->f_mode & FMODE_WRITE))
6608                         return -EBADF;
6609         }
6610
6611         status = nfs4_set_lock_state(state, request);
6612         if (status != 0)
6613                 return status;
6614
6615         return nfs4_retry_setlk(state, cmd, request);
6616 }
6617
6618 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6619 {
6620         struct nfs_server *server = NFS_SERVER(state->inode);
6621         int err;
6622
6623         err = nfs4_set_lock_state(state, fl);
6624         if (err != 0)
6625                 return err;
6626         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6627         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6628 }
6629
6630 struct nfs_release_lockowner_data {
6631         struct nfs4_lock_state *lsp;
6632         struct nfs_server *server;
6633         struct nfs_release_lockowner_args args;
6634         struct nfs_release_lockowner_res res;
6635         unsigned long timestamp;
6636 };
6637
6638 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6639 {
6640         struct nfs_release_lockowner_data *data = calldata;
6641         struct nfs_server *server = data->server;
6642         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6643                              &data->args.seq_args, &data->res.seq_res, task);
6644         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6645         data->timestamp = jiffies;
6646 }
6647
6648 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6649 {
6650         struct nfs_release_lockowner_data *data = calldata;
6651         struct nfs_server *server = data->server;
6652
6653         nfs40_sequence_done(task, &data->res.seq_res);
6654
6655         switch (task->tk_status) {
6656         case 0:
6657                 renew_lease(server, data->timestamp);
6658                 break;
6659         case -NFS4ERR_STALE_CLIENTID:
6660         case -NFS4ERR_EXPIRED:
6661                 nfs4_schedule_lease_recovery(server->nfs_client);
6662                 break;
6663         case -NFS4ERR_LEASE_MOVED:
6664         case -NFS4ERR_DELAY:
6665                 if (nfs4_async_handle_error(task, server,
6666                                             NULL, NULL) == -EAGAIN)
6667                         rpc_restart_call_prepare(task);
6668         }
6669 }
6670
6671 static void nfs4_release_lockowner_release(void *calldata)
6672 {
6673         struct nfs_release_lockowner_data *data = calldata;
6674         nfs4_free_lock_state(data->server, data->lsp);
6675         kfree(calldata);
6676 }
6677
6678 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6679         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6680         .rpc_call_done = nfs4_release_lockowner_done,
6681         .rpc_release = nfs4_release_lockowner_release,
6682 };
6683
6684 static void
6685 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6686 {
6687         struct nfs_release_lockowner_data *data;
6688         struct rpc_message msg = {
6689                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6690         };
6691
6692         if (server->nfs_client->cl_mvops->minor_version != 0)
6693                 return;
6694
6695         data = kmalloc(sizeof(*data), GFP_NOFS);
6696         if (!data)
6697                 return;
6698         data->lsp = lsp;
6699         data->server = server;
6700         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6701         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6702         data->args.lock_owner.s_dev = server->s_dev;
6703
6704         msg.rpc_argp = &data->args;
6705         msg.rpc_resp = &data->res;
6706         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6707         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6708 }
6709
6710 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6711
6712 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6713                                    struct dentry *unused, struct inode *inode,
6714                                    const char *key, const void *buf,
6715                                    size_t buflen, int flags)
6716 {
6717         return nfs4_proc_set_acl(inode, buf, buflen);
6718 }
6719
6720 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6721                                    struct dentry *unused, struct inode *inode,
6722                                    const char *key, void *buf, size_t buflen)
6723 {
6724         return nfs4_proc_get_acl(inode, buf, buflen);
6725 }
6726
6727 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6728 {
6729         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6730 }
6731
6732 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6733
6734 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6735                                      struct dentry *unused, struct inode *inode,
6736                                      const char *key, const void *buf,
6737                                      size_t buflen, int flags)
6738 {
6739         if (security_ismaclabel(key))
6740                 return nfs4_set_security_label(inode, buf, buflen);
6741
6742         return -EOPNOTSUPP;
6743 }
6744
6745 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6746                                      struct dentry *unused, struct inode *inode,
6747                                      const char *key, void *buf, size_t buflen)
6748 {
6749         if (security_ismaclabel(key))
6750                 return nfs4_get_security_label(inode, buf, buflen);
6751         return -EOPNOTSUPP;
6752 }
6753
6754 static ssize_t
6755 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6756 {
6757         int len = 0;
6758
6759         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6760                 len = security_inode_listsecurity(inode, list, list_len);
6761                 if (list_len && len > list_len)
6762                         return -ERANGE;
6763         }
6764         return len;
6765 }
6766
6767 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6768         .prefix = XATTR_SECURITY_PREFIX,
6769         .get    = nfs4_xattr_get_nfs4_label,
6770         .set    = nfs4_xattr_set_nfs4_label,
6771 };
6772
6773 #else
6774
6775 static ssize_t
6776 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6777 {
6778         return 0;
6779 }
6780
6781 #endif
6782
6783 /*
6784  * nfs_fhget will use either the mounted_on_fileid or the fileid
6785  */
6786 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6787 {
6788         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6789                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6790               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6791               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6792                 return;
6793
6794         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6795                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6796         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6797         fattr->nlink = 2;
6798 }
6799
6800 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6801                                    const struct qstr *name,
6802                                    struct nfs4_fs_locations *fs_locations,
6803                                    struct page *page)
6804 {
6805         struct nfs_server *server = NFS_SERVER(dir);
6806         u32 bitmask[3] = {
6807                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6808         };
6809         struct nfs4_fs_locations_arg args = {
6810                 .dir_fh = NFS_FH(dir),
6811                 .name = name,
6812                 .page = page,
6813                 .bitmask = bitmask,
6814         };
6815         struct nfs4_fs_locations_res res = {
6816                 .fs_locations = fs_locations,
6817         };
6818         struct rpc_message msg = {
6819                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6820                 .rpc_argp = &args,
6821                 .rpc_resp = &res,
6822         };
6823         int status;
6824
6825         dprintk("%s: start\n", __func__);
6826
6827         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6828          * is not supported */
6829         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6830                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6831         else
6832                 bitmask[0] |= FATTR4_WORD0_FILEID;
6833
6834         nfs_fattr_init(&fs_locations->fattr);
6835         fs_locations->server = server;
6836         fs_locations->nlocations = 0;
6837         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6838         dprintk("%s: returned status = %d\n", __func__, status);
6839         return status;
6840 }
6841
6842 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6843                            const struct qstr *name,
6844                            struct nfs4_fs_locations *fs_locations,
6845                            struct page *page)
6846 {
6847         struct nfs4_exception exception = { };
6848         int err;
6849         do {
6850                 err = _nfs4_proc_fs_locations(client, dir, name,
6851                                 fs_locations, page);
6852                 trace_nfs4_get_fs_locations(dir, name, err);
6853                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6854                                 &exception);
6855         } while (exception.retry);
6856         return err;
6857 }
6858
6859 /*
6860  * This operation also signals the server that this client is
6861  * performing migration recovery.  The server can stop returning
6862  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6863  * appended to this compound to identify the client ID which is
6864  * performing recovery.
6865  */
6866 static int _nfs40_proc_get_locations(struct inode *inode,
6867                                      struct nfs4_fs_locations *locations,
6868                                      struct page *page, struct rpc_cred *cred)
6869 {
6870         struct nfs_server *server = NFS_SERVER(inode);
6871         struct rpc_clnt *clnt = server->client;
6872         u32 bitmask[2] = {
6873                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6874         };
6875         struct nfs4_fs_locations_arg args = {
6876                 .clientid       = server->nfs_client->cl_clientid,
6877                 .fh             = NFS_FH(inode),
6878                 .page           = page,
6879                 .bitmask        = bitmask,
6880                 .migration      = 1,            /* skip LOOKUP */
6881                 .renew          = 1,            /* append RENEW */
6882         };
6883         struct nfs4_fs_locations_res res = {
6884                 .fs_locations   = locations,
6885                 .migration      = 1,
6886                 .renew          = 1,
6887         };
6888         struct rpc_message msg = {
6889                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6890                 .rpc_argp       = &args,
6891                 .rpc_resp       = &res,
6892                 .rpc_cred       = cred,
6893         };
6894         unsigned long now = jiffies;
6895         int status;
6896
6897         nfs_fattr_init(&locations->fattr);
6898         locations->server = server;
6899         locations->nlocations = 0;
6900
6901         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6902         nfs4_set_sequence_privileged(&args.seq_args);
6903         status = nfs4_call_sync_sequence(clnt, server, &msg,
6904                                         &args.seq_args, &res.seq_res);
6905         if (status)
6906                 return status;
6907
6908         renew_lease(server, now);
6909         return 0;
6910 }
6911
6912 #ifdef CONFIG_NFS_V4_1
6913
6914 /*
6915  * This operation also signals the server that this client is
6916  * performing migration recovery.  The server can stop asserting
6917  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6918  * performing this operation is identified in the SEQUENCE
6919  * operation in this compound.
6920  *
6921  * When the client supports GETATTR(fs_locations_info), it can
6922  * be plumbed in here.
6923  */
6924 static int _nfs41_proc_get_locations(struct inode *inode,
6925                                      struct nfs4_fs_locations *locations,
6926                                      struct page *page, struct rpc_cred *cred)
6927 {
6928         struct nfs_server *server = NFS_SERVER(inode);
6929         struct rpc_clnt *clnt = server->client;
6930         u32 bitmask[2] = {
6931                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6932         };
6933         struct nfs4_fs_locations_arg args = {
6934                 .fh             = NFS_FH(inode),
6935                 .page           = page,
6936                 .bitmask        = bitmask,
6937                 .migration      = 1,            /* skip LOOKUP */
6938         };
6939         struct nfs4_fs_locations_res res = {
6940                 .fs_locations   = locations,
6941                 .migration      = 1,
6942         };
6943         struct rpc_message msg = {
6944                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6945                 .rpc_argp       = &args,
6946                 .rpc_resp       = &res,
6947                 .rpc_cred       = cred,
6948         };
6949         int status;
6950
6951         nfs_fattr_init(&locations->fattr);
6952         locations->server = server;
6953         locations->nlocations = 0;
6954
6955         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6956         nfs4_set_sequence_privileged(&args.seq_args);
6957         status = nfs4_call_sync_sequence(clnt, server, &msg,
6958                                         &args.seq_args, &res.seq_res);
6959         if (status == NFS4_OK &&
6960             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6961                 status = -NFS4ERR_LEASE_MOVED;
6962         return status;
6963 }
6964
6965 #endif  /* CONFIG_NFS_V4_1 */
6966
6967 /**
6968  * nfs4_proc_get_locations - discover locations for a migrated FSID
6969  * @inode: inode on FSID that is migrating
6970  * @locations: result of query
6971  * @page: buffer
6972  * @cred: credential to use for this operation
6973  *
6974  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6975  * operation failed, or a negative errno if a local error occurred.
6976  *
6977  * On success, "locations" is filled in, but if the server has
6978  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6979  * asserted.
6980  *
6981  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6982  * from this client that require migration recovery.
6983  */
6984 int nfs4_proc_get_locations(struct inode *inode,
6985                             struct nfs4_fs_locations *locations,
6986                             struct page *page, struct rpc_cred *cred)
6987 {
6988         struct nfs_server *server = NFS_SERVER(inode);
6989         struct nfs_client *clp = server->nfs_client;
6990         const struct nfs4_mig_recovery_ops *ops =
6991                                         clp->cl_mvops->mig_recovery_ops;
6992         struct nfs4_exception exception = { };
6993         int status;
6994
6995         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6996                 (unsigned long long)server->fsid.major,
6997                 (unsigned long long)server->fsid.minor,
6998                 clp->cl_hostname);
6999         nfs_display_fhandle(NFS_FH(inode), __func__);
7000
7001         do {
7002                 status = ops->get_locations(inode, locations, page, cred);
7003                 if (status != -NFS4ERR_DELAY)
7004                         break;
7005                 nfs4_handle_exception(server, status, &exception);
7006         } while (exception.retry);
7007         return status;
7008 }
7009
7010 /*
7011  * This operation also signals the server that this client is
7012  * performing "lease moved" recovery.  The server can stop
7013  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7014  * is appended to this compound to identify the client ID which is
7015  * performing recovery.
7016  */
7017 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7018 {
7019         struct nfs_server *server = NFS_SERVER(inode);
7020         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7021         struct rpc_clnt *clnt = server->client;
7022         struct nfs4_fsid_present_arg args = {
7023                 .fh             = NFS_FH(inode),
7024                 .clientid       = clp->cl_clientid,
7025                 .renew          = 1,            /* append RENEW */
7026         };
7027         struct nfs4_fsid_present_res res = {
7028                 .renew          = 1,
7029         };
7030         struct rpc_message msg = {
7031                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7032                 .rpc_argp       = &args,
7033                 .rpc_resp       = &res,
7034                 .rpc_cred       = cred,
7035         };
7036         unsigned long now = jiffies;
7037         int status;
7038
7039         res.fh = nfs_alloc_fhandle();
7040         if (res.fh == NULL)
7041                 return -ENOMEM;
7042
7043         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7044         nfs4_set_sequence_privileged(&args.seq_args);
7045         status = nfs4_call_sync_sequence(clnt, server, &msg,
7046                                                 &args.seq_args, &res.seq_res);
7047         nfs_free_fhandle(res.fh);
7048         if (status)
7049                 return status;
7050
7051         do_renew_lease(clp, now);
7052         return 0;
7053 }
7054
7055 #ifdef CONFIG_NFS_V4_1
7056
7057 /*
7058  * This operation also signals the server that this client is
7059  * performing "lease moved" recovery.  The server can stop asserting
7060  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7061  * this operation is identified in the SEQUENCE operation in this
7062  * compound.
7063  */
7064 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7065 {
7066         struct nfs_server *server = NFS_SERVER(inode);
7067         struct rpc_clnt *clnt = server->client;
7068         struct nfs4_fsid_present_arg args = {
7069                 .fh             = NFS_FH(inode),
7070         };
7071         struct nfs4_fsid_present_res res = {
7072         };
7073         struct rpc_message msg = {
7074                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7075                 .rpc_argp       = &args,
7076                 .rpc_resp       = &res,
7077                 .rpc_cred       = cred,
7078         };
7079         int status;
7080
7081         res.fh = nfs_alloc_fhandle();
7082         if (res.fh == NULL)
7083                 return -ENOMEM;
7084
7085         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7086         nfs4_set_sequence_privileged(&args.seq_args);
7087         status = nfs4_call_sync_sequence(clnt, server, &msg,
7088                                                 &args.seq_args, &res.seq_res);
7089         nfs_free_fhandle(res.fh);
7090         if (status == NFS4_OK &&
7091             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7092                 status = -NFS4ERR_LEASE_MOVED;
7093         return status;
7094 }
7095
7096 #endif  /* CONFIG_NFS_V4_1 */
7097
7098 /**
7099  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7100  * @inode: inode on FSID to check
7101  * @cred: credential to use for this operation
7102  *
7103  * Server indicates whether the FSID is present, moved, or not
7104  * recognized.  This operation is necessary to clear a LEASE_MOVED
7105  * condition for this client ID.
7106  *
7107  * Returns NFS4_OK if the FSID is present on this server,
7108  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7109  *  NFS4ERR code if some error occurred on the server, or a
7110  *  negative errno if a local failure occurred.
7111  */
7112 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7113 {
7114         struct nfs_server *server = NFS_SERVER(inode);
7115         struct nfs_client *clp = server->nfs_client;
7116         const struct nfs4_mig_recovery_ops *ops =
7117                                         clp->cl_mvops->mig_recovery_ops;
7118         struct nfs4_exception exception = { };
7119         int status;
7120
7121         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7122                 (unsigned long long)server->fsid.major,
7123                 (unsigned long long)server->fsid.minor,
7124                 clp->cl_hostname);
7125         nfs_display_fhandle(NFS_FH(inode), __func__);
7126
7127         do {
7128                 status = ops->fsid_present(inode, cred);
7129                 if (status != -NFS4ERR_DELAY)
7130                         break;
7131                 nfs4_handle_exception(server, status, &exception);
7132         } while (exception.retry);
7133         return status;
7134 }
7135
7136 /**
7137  * If 'use_integrity' is true and the state managment nfs_client
7138  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7139  * and the machine credential as per RFC3530bis and RFC5661 Security
7140  * Considerations sections. Otherwise, just use the user cred with the
7141  * filesystem's rpc_client.
7142  */
7143 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7144 {
7145         int status;
7146         struct nfs4_secinfo_arg args = {
7147                 .dir_fh = NFS_FH(dir),
7148                 .name   = name,
7149         };
7150         struct nfs4_secinfo_res res = {
7151                 .flavors     = flavors,
7152         };
7153         struct rpc_message msg = {
7154                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7155                 .rpc_argp = &args,
7156                 .rpc_resp = &res,
7157         };
7158         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7159         struct rpc_cred *cred = NULL;
7160
7161         if (use_integrity) {
7162                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7163                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7164                 msg.rpc_cred = cred;
7165         }
7166
7167         dprintk("NFS call  secinfo %s\n", name->name);
7168
7169         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7170                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7171
7172         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7173                                 &res.seq_res, 0);
7174         dprintk("NFS reply  secinfo: %d\n", status);
7175
7176         if (cred)
7177                 put_rpccred(cred);
7178
7179         return status;
7180 }
7181
7182 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7183                       struct nfs4_secinfo_flavors *flavors)
7184 {
7185         struct nfs4_exception exception = { };
7186         int err;
7187         do {
7188                 err = -NFS4ERR_WRONGSEC;
7189
7190                 /* try to use integrity protection with machine cred */
7191                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7192                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7193
7194                 /*
7195                  * if unable to use integrity protection, or SECINFO with
7196                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7197                  * disallowed by spec, but exists in deployed servers) use
7198                  * the current filesystem's rpc_client and the user cred.
7199                  */
7200                 if (err == -NFS4ERR_WRONGSEC)
7201                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7202
7203                 trace_nfs4_secinfo(dir, name, err);
7204                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7205                                 &exception);
7206         } while (exception.retry);
7207         return err;
7208 }
7209
7210 #ifdef CONFIG_NFS_V4_1
7211 /*
7212  * Check the exchange flags returned by the server for invalid flags, having
7213  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7214  * DS flags set.
7215  */
7216 static int nfs4_check_cl_exchange_flags(u32 flags)
7217 {
7218         if (flags & ~EXCHGID4_FLAG_MASK_R)
7219                 goto out_inval;
7220         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7221             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7222                 goto out_inval;
7223         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7224                 goto out_inval;
7225         return NFS_OK;
7226 out_inval:
7227         return -NFS4ERR_INVAL;
7228 }
7229
7230 static bool
7231 nfs41_same_server_scope(struct nfs41_server_scope *a,
7232                         struct nfs41_server_scope *b)
7233 {
7234         if (a->server_scope_sz == b->server_scope_sz &&
7235             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
7236                 return true;
7237
7238         return false;
7239 }
7240
7241 static void
7242 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7243 {
7244 }
7245
7246 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7247         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7248 };
7249
7250 /*
7251  * nfs4_proc_bind_one_conn_to_session()
7252  *
7253  * The 4.1 client currently uses the same TCP connection for the
7254  * fore and backchannel.
7255  */
7256 static
7257 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7258                 struct rpc_xprt *xprt,
7259                 struct nfs_client *clp,
7260                 struct rpc_cred *cred)
7261 {
7262         int status;
7263         struct nfs41_bind_conn_to_session_args args = {
7264                 .client = clp,
7265                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7266         };
7267         struct nfs41_bind_conn_to_session_res res;
7268         struct rpc_message msg = {
7269                 .rpc_proc =
7270                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7271                 .rpc_argp = &args,
7272                 .rpc_resp = &res,
7273                 .rpc_cred = cred,
7274         };
7275         struct rpc_task_setup task_setup_data = {
7276                 .rpc_client = clnt,
7277                 .rpc_xprt = xprt,
7278                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7279                 .rpc_message = &msg,
7280                 .flags = RPC_TASK_TIMEOUT,
7281         };
7282         struct rpc_task *task;
7283
7284         dprintk("--> %s\n", __func__);
7285
7286         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7287         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7288                 args.dir = NFS4_CDFC4_FORE;
7289
7290         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7291         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7292                 args.dir = NFS4_CDFC4_FORE;
7293
7294         task = rpc_run_task(&task_setup_data);
7295         if (!IS_ERR(task)) {
7296                 status = task->tk_status;
7297                 rpc_put_task(task);
7298         } else
7299                 status = PTR_ERR(task);
7300         trace_nfs4_bind_conn_to_session(clp, status);
7301         if (status == 0) {
7302                 if (memcmp(res.sessionid.data,
7303                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7304                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7305                         status = -EIO;
7306                         goto out;
7307                 }
7308                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7309                         dprintk("NFS: %s: Unexpected direction from server\n",
7310                                 __func__);
7311                         status = -EIO;
7312                         goto out;
7313                 }
7314                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7315                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7316                                 __func__);
7317                         status = -EIO;
7318                         goto out;
7319                 }
7320         }
7321 out:
7322         dprintk("<-- %s status= %d\n", __func__, status);
7323         return status;
7324 }
7325
7326 struct rpc_bind_conn_calldata {
7327         struct nfs_client *clp;
7328         struct rpc_cred *cred;
7329 };
7330
7331 static int
7332 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7333                 struct rpc_xprt *xprt,
7334                 void *calldata)
7335 {
7336         struct rpc_bind_conn_calldata *p = calldata;
7337
7338         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7339 }
7340
7341 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7342 {
7343         struct rpc_bind_conn_calldata data = {
7344                 .clp = clp,
7345                 .cred = cred,
7346         };
7347         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7348                         nfs4_proc_bind_conn_to_session_callback, &data);
7349 }
7350
7351 /*
7352  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7353  * and operations we'd like to see to enable certain features in the allow map
7354  */
7355 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7356         .how = SP4_MACH_CRED,
7357         .enforce.u.words = {
7358                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7359                       1 << (OP_EXCHANGE_ID - 32) |
7360                       1 << (OP_CREATE_SESSION - 32) |
7361                       1 << (OP_DESTROY_SESSION - 32) |
7362                       1 << (OP_DESTROY_CLIENTID - 32)
7363         },
7364         .allow.u.words = {
7365                 [0] = 1 << (OP_CLOSE) |
7366                       1 << (OP_OPEN_DOWNGRADE) |
7367                       1 << (OP_LOCKU) |
7368                       1 << (OP_DELEGRETURN) |
7369                       1 << (OP_COMMIT),
7370                 [1] = 1 << (OP_SECINFO - 32) |
7371                       1 << (OP_SECINFO_NO_NAME - 32) |
7372                       1 << (OP_LAYOUTRETURN - 32) |
7373                       1 << (OP_TEST_STATEID - 32) |
7374                       1 << (OP_FREE_STATEID - 32) |
7375                       1 << (OP_WRITE - 32)
7376         }
7377 };
7378
7379 /*
7380  * Select the state protection mode for client `clp' given the server results
7381  * from exchange_id in `sp'.
7382  *
7383  * Returns 0 on success, negative errno otherwise.
7384  */
7385 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7386                                  struct nfs41_state_protection *sp)
7387 {
7388         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7389                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7390                       1 << (OP_EXCHANGE_ID - 32) |
7391                       1 << (OP_CREATE_SESSION - 32) |
7392                       1 << (OP_DESTROY_SESSION - 32) |
7393                       1 << (OP_DESTROY_CLIENTID - 32)
7394         };
7395         unsigned int i;
7396
7397         if (sp->how == SP4_MACH_CRED) {
7398                 /* Print state protect result */
7399                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7400                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7401                         if (test_bit(i, sp->enforce.u.longs))
7402                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7403                         if (test_bit(i, sp->allow.u.longs))
7404                                 dfprintk(MOUNT, "  allow op %d\n", i);
7405                 }
7406
7407                 /* make sure nothing is on enforce list that isn't supported */
7408                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7409                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7410                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7411                                 return -EINVAL;
7412                         }
7413                 }
7414
7415                 /*
7416                  * Minimal mode - state operations are allowed to use machine
7417                  * credential.  Note this already happens by default, so the
7418                  * client doesn't have to do anything more than the negotiation.
7419                  *
7420                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7421                  *       we're already using the machine cred for exchange_id
7422                  *       and will never use a different cred.
7423                  */
7424                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7425                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7426                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7427                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7428                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7429                         dfprintk(MOUNT, "  minimal mode enabled\n");
7430                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7431                 } else {
7432                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7433                         return -EINVAL;
7434                 }
7435
7436                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7437                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7438                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7439                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7440                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7441                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7442                 }
7443
7444                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7445                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7446                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7447                                 &clp->cl_sp4_flags);
7448                 }
7449
7450                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7451                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7452                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7453                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7454                 }
7455
7456                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7457                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7458                         dfprintk(MOUNT, "  stateid mode enabled\n");
7459                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7460                 }
7461
7462                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7463                         dfprintk(MOUNT, "  write mode enabled\n");
7464                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7465                 }
7466
7467                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7468                         dfprintk(MOUNT, "  commit mode enabled\n");
7469                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7470                 }
7471         }
7472
7473         return 0;
7474 }
7475
7476 struct nfs41_exchange_id_data {
7477         struct nfs41_exchange_id_res res;
7478         struct nfs41_exchange_id_args args;
7479         struct rpc_xprt *xprt;
7480         int rpc_status;
7481 };
7482
7483 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7484 {
7485         struct nfs41_exchange_id_data *cdata =
7486                                         (struct nfs41_exchange_id_data *)data;
7487         struct nfs_client *clp = cdata->args.client;
7488         int status = task->tk_status;
7489
7490         trace_nfs4_exchange_id(clp, status);
7491
7492         if (status == 0)
7493                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7494
7495         if (cdata->xprt && status == 0) {
7496                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7497                                                       cdata->xprt);
7498                 goto out;
7499         }
7500
7501         if (status  == 0)
7502                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7503
7504         if (status == 0) {
7505                 clp->cl_clientid = cdata->res.clientid;
7506                 clp->cl_exchange_flags = cdata->res.flags;
7507                 /* Client ID is not confirmed */
7508                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7509                         clear_bit(NFS4_SESSION_ESTABLISHED,
7510                         &clp->cl_session->session_state);
7511                         clp->cl_seqid = cdata->res.seqid;
7512                 }
7513
7514                 kfree(clp->cl_serverowner);
7515                 clp->cl_serverowner = cdata->res.server_owner;
7516                 cdata->res.server_owner = NULL;
7517
7518                 /* use the most recent implementation id */
7519                 kfree(clp->cl_implid);
7520                 clp->cl_implid = cdata->res.impl_id;
7521                 cdata->res.impl_id = NULL;
7522
7523                 if (clp->cl_serverscope != NULL &&
7524                     !nfs41_same_server_scope(clp->cl_serverscope,
7525                                         cdata->res.server_scope)) {
7526                         dprintk("%s: server_scope mismatch detected\n",
7527                                 __func__);
7528                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7529                         kfree(clp->cl_serverscope);
7530                         clp->cl_serverscope = NULL;
7531                 }
7532
7533                 if (clp->cl_serverscope == NULL) {
7534                         clp->cl_serverscope = cdata->res.server_scope;
7535                         cdata->res.server_scope = NULL;
7536                 }
7537                 /* Save the EXCHANGE_ID verifier session trunk tests */
7538                 memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7539                        sizeof(clp->cl_confirm.data));
7540         }
7541 out:
7542         cdata->rpc_status = status;
7543         return;
7544 }
7545
7546 static void nfs4_exchange_id_release(void *data)
7547 {
7548         struct nfs41_exchange_id_data *cdata =
7549                                         (struct nfs41_exchange_id_data *)data;
7550
7551         nfs_put_client(cdata->args.client);
7552         if (cdata->xprt) {
7553                 xprt_put(cdata->xprt);
7554                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7555         }
7556         kfree(cdata->res.impl_id);
7557         kfree(cdata->res.server_scope);
7558         kfree(cdata->res.server_owner);
7559         kfree(cdata);
7560 }
7561
7562 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7563         .rpc_call_done = nfs4_exchange_id_done,
7564         .rpc_release = nfs4_exchange_id_release,
7565 };
7566
7567 /*
7568  * _nfs4_proc_exchange_id()
7569  *
7570  * Wrapper for EXCHANGE_ID operation.
7571  */
7572 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7573                         u32 sp4_how, struct rpc_xprt *xprt)
7574 {
7575         nfs4_verifier verifier;
7576         struct rpc_message msg = {
7577                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7578                 .rpc_cred = cred,
7579         };
7580         struct rpc_task_setup task_setup_data = {
7581                 .rpc_client = clp->cl_rpcclient,
7582                 .callback_ops = &nfs4_exchange_id_call_ops,
7583                 .rpc_message = &msg,
7584                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7585         };
7586         struct nfs41_exchange_id_data *calldata;
7587         struct rpc_task *task;
7588         int status = -EIO;
7589
7590         if (!atomic_inc_not_zero(&clp->cl_count))
7591                 goto out;
7592
7593         status = -ENOMEM;
7594         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7595         if (!calldata)
7596                 goto out;
7597
7598         if (!xprt)
7599                 nfs4_init_boot_verifier(clp, &verifier);
7600
7601         status = nfs4_init_uniform_client_string(clp);
7602         if (status)
7603                 goto out_calldata;
7604
7605         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7606                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7607                 clp->cl_owner_id);
7608
7609         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7610                                                 GFP_NOFS);
7611         status = -ENOMEM;
7612         if (unlikely(calldata->res.server_owner == NULL))
7613                 goto out_calldata;
7614
7615         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7616                                         GFP_NOFS);
7617         if (unlikely(calldata->res.server_scope == NULL))
7618                 goto out_server_owner;
7619
7620         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7621         if (unlikely(calldata->res.impl_id == NULL))
7622                 goto out_server_scope;
7623
7624         switch (sp4_how) {
7625         case SP4_NONE:
7626                 calldata->args.state_protect.how = SP4_NONE;
7627                 break;
7628
7629         case SP4_MACH_CRED:
7630                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7631                 break;
7632
7633         default:
7634                 /* unsupported! */
7635                 WARN_ON_ONCE(1);
7636                 status = -EINVAL;
7637                 goto out_impl_id;
7638         }
7639         if (xprt) {
7640                 calldata->xprt = xprt;
7641                 task_setup_data.rpc_xprt = xprt;
7642                 task_setup_data.flags =
7643                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7644                 calldata->args.verifier = &clp->cl_confirm;
7645         } else {
7646                 calldata->args.verifier = &verifier;
7647         }
7648         calldata->args.client = clp;
7649 #ifdef CONFIG_NFS_V4_1_MIGRATION
7650         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7651         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7652         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7653 #else
7654         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7655         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7656 #endif
7657         msg.rpc_argp = &calldata->args;
7658         msg.rpc_resp = &calldata->res;
7659         task_setup_data.callback_data = calldata;
7660
7661         task = rpc_run_task(&task_setup_data);
7662         if (IS_ERR(task)) {
7663         status = PTR_ERR(task);
7664                 goto out_impl_id;
7665         }
7666
7667         if (!xprt) {
7668                 status = rpc_wait_for_completion_task(task);
7669                 if (!status)
7670                         status = calldata->rpc_status;
7671         } else  /* session trunking test */
7672                 status = calldata->rpc_status;
7673
7674         rpc_put_task(task);
7675 out:
7676         if (clp->cl_implid != NULL)
7677                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7678                         "domain: %s, name: %s, date: %llu,%u\n",
7679                         clp->cl_implid->domain, clp->cl_implid->name,
7680                         clp->cl_implid->date.seconds,
7681                         clp->cl_implid->date.nseconds);
7682         dprintk("NFS reply exchange_id: %d\n", status);
7683         return status;
7684
7685 out_impl_id:
7686         kfree(calldata->res.impl_id);
7687 out_server_scope:
7688         kfree(calldata->res.server_scope);
7689 out_server_owner:
7690         kfree(calldata->res.server_owner);
7691 out_calldata:
7692         kfree(calldata);
7693         goto out;
7694 }
7695
7696 /*
7697  * nfs4_proc_exchange_id()
7698  *
7699  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7700  *
7701  * Since the clientid has expired, all compounds using sessions
7702  * associated with the stale clientid will be returning
7703  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7704  * be in some phase of session reset.
7705  *
7706  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7707  */
7708 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7709 {
7710         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7711         int status;
7712
7713         /* try SP4_MACH_CRED if krb5i/p */
7714         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7715             authflavor == RPC_AUTH_GSS_KRB5P) {
7716                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7717                 if (!status)
7718                         return 0;
7719         }
7720
7721         /* try SP4_NONE */
7722         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7723 }
7724
7725 /**
7726  * nfs4_test_session_trunk
7727  *
7728  * This is an add_xprt_test() test function called from
7729  * rpc_clnt_setup_test_and_add_xprt.
7730  *
7731  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7732  * and is dereferrenced in nfs4_exchange_id_release
7733  *
7734  * Upon success, add the new transport to the rpc_clnt
7735  *
7736  * @clnt: struct rpc_clnt to get new transport
7737  * @xprt: the rpc_xprt to test
7738  * @data: call data for _nfs4_proc_exchange_id.
7739  */
7740 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7741                             void *data)
7742 {
7743         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7744         u32 sp4_how;
7745
7746         dprintk("--> %s try %s\n", __func__,
7747                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7748
7749         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7750
7751         /* Test connection for session trunking. Async exchange_id call */
7752         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7753 }
7754 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7755
7756 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7757                 struct rpc_cred *cred)
7758 {
7759         struct rpc_message msg = {
7760                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7761                 .rpc_argp = clp,
7762                 .rpc_cred = cred,
7763         };
7764         int status;
7765
7766         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7767         trace_nfs4_destroy_clientid(clp, status);
7768         if (status)
7769                 dprintk("NFS: Got error %d from the server %s on "
7770                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7771         return status;
7772 }
7773
7774 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7775                 struct rpc_cred *cred)
7776 {
7777         unsigned int loop;
7778         int ret;
7779
7780         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7781                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7782                 switch (ret) {
7783                 case -NFS4ERR_DELAY:
7784                 case -NFS4ERR_CLIENTID_BUSY:
7785                         ssleep(1);
7786                         break;
7787                 default:
7788                         return ret;
7789                 }
7790         }
7791         return 0;
7792 }
7793
7794 int nfs4_destroy_clientid(struct nfs_client *clp)
7795 {
7796         struct rpc_cred *cred;
7797         int ret = 0;
7798
7799         if (clp->cl_mvops->minor_version < 1)
7800                 goto out;
7801         if (clp->cl_exchange_flags == 0)
7802                 goto out;
7803         if (clp->cl_preserve_clid)
7804                 goto out;
7805         cred = nfs4_get_clid_cred(clp);
7806         ret = nfs4_proc_destroy_clientid(clp, cred);
7807         if (cred)
7808                 put_rpccred(cred);
7809         switch (ret) {
7810         case 0:
7811         case -NFS4ERR_STALE_CLIENTID:
7812                 clp->cl_exchange_flags = 0;
7813         }
7814 out:
7815         return ret;
7816 }
7817
7818 struct nfs4_get_lease_time_data {
7819         struct nfs4_get_lease_time_args *args;
7820         struct nfs4_get_lease_time_res *res;
7821         struct nfs_client *clp;
7822 };
7823
7824 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7825                                         void *calldata)
7826 {
7827         struct nfs4_get_lease_time_data *data =
7828                         (struct nfs4_get_lease_time_data *)calldata;
7829
7830         dprintk("--> %s\n", __func__);
7831         /* just setup sequence, do not trigger session recovery
7832            since we're invoked within one */
7833         nfs41_setup_sequence(data->clp->cl_session,
7834                         &data->args->la_seq_args,
7835                         &data->res->lr_seq_res,
7836                         task);
7837         dprintk("<-- %s\n", __func__);
7838 }
7839
7840 /*
7841  * Called from nfs4_state_manager thread for session setup, so don't recover
7842  * from sequence operation or clientid errors.
7843  */
7844 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7845 {
7846         struct nfs4_get_lease_time_data *data =
7847                         (struct nfs4_get_lease_time_data *)calldata;
7848
7849         dprintk("--> %s\n", __func__);
7850         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7851                 return;
7852         switch (task->tk_status) {
7853         case -NFS4ERR_DELAY:
7854         case -NFS4ERR_GRACE:
7855                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7856                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7857                 task->tk_status = 0;
7858                 /* fall through */
7859         case -NFS4ERR_RETRY_UNCACHED_REP:
7860                 rpc_restart_call_prepare(task);
7861                 return;
7862         }
7863         dprintk("<-- %s\n", __func__);
7864 }
7865
7866 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7867         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7868         .rpc_call_done = nfs4_get_lease_time_done,
7869 };
7870
7871 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7872 {
7873         struct rpc_task *task;
7874         struct nfs4_get_lease_time_args args;
7875         struct nfs4_get_lease_time_res res = {
7876                 .lr_fsinfo = fsinfo,
7877         };
7878         struct nfs4_get_lease_time_data data = {
7879                 .args = &args,
7880                 .res = &res,
7881                 .clp = clp,
7882         };
7883         struct rpc_message msg = {
7884                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7885                 .rpc_argp = &args,
7886                 .rpc_resp = &res,
7887         };
7888         struct rpc_task_setup task_setup = {
7889                 .rpc_client = clp->cl_rpcclient,
7890                 .rpc_message = &msg,
7891                 .callback_ops = &nfs4_get_lease_time_ops,
7892                 .callback_data = &data,
7893                 .flags = RPC_TASK_TIMEOUT,
7894         };
7895         int status;
7896
7897         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7898         nfs4_set_sequence_privileged(&args.la_seq_args);
7899         dprintk("--> %s\n", __func__);
7900         task = rpc_run_task(&task_setup);
7901
7902         if (IS_ERR(task))
7903                 status = PTR_ERR(task);
7904         else {
7905                 status = task->tk_status;
7906                 rpc_put_task(task);
7907         }
7908         dprintk("<-- %s return %d\n", __func__, status);
7909
7910         return status;
7911 }
7912
7913 /*
7914  * Initialize the values to be used by the client in CREATE_SESSION
7915  * If nfs4_init_session set the fore channel request and response sizes,
7916  * use them.
7917  *
7918  * Set the back channel max_resp_sz_cached to zero to force the client to
7919  * always set csa_cachethis to FALSE because the current implementation
7920  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7921  */
7922 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7923                                     struct rpc_clnt *clnt)
7924 {
7925         unsigned int max_rqst_sz, max_resp_sz;
7926         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7927
7928         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7929         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7930
7931         /* Fore channel attributes */
7932         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7933         args->fc_attrs.max_resp_sz = max_resp_sz;
7934         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7935         args->fc_attrs.max_reqs = max_session_slots;
7936
7937         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7938                 "max_ops=%u max_reqs=%u\n",
7939                 __func__,
7940                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7941                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7942
7943         /* Back channel attributes */
7944         args->bc_attrs.max_rqst_sz = max_bc_payload;
7945         args->bc_attrs.max_resp_sz = max_bc_payload;
7946         args->bc_attrs.max_resp_sz_cached = 0;
7947         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7948         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7949
7950         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7951                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7952                 __func__,
7953                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7954                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7955                 args->bc_attrs.max_reqs);
7956 }
7957
7958 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7959                 struct nfs41_create_session_res *res)
7960 {
7961         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7962         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7963
7964         if (rcvd->max_resp_sz > sent->max_resp_sz)
7965                 return -EINVAL;
7966         /*
7967          * Our requested max_ops is the minimum we need; we're not
7968          * prepared to break up compounds into smaller pieces than that.
7969          * So, no point even trying to continue if the server won't
7970          * cooperate:
7971          */
7972         if (rcvd->max_ops < sent->max_ops)
7973                 return -EINVAL;
7974         if (rcvd->max_reqs == 0)
7975                 return -EINVAL;
7976         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7977                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7978         return 0;
7979 }
7980
7981 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7982                 struct nfs41_create_session_res *res)
7983 {
7984         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7985         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7986
7987         if (!(res->flags & SESSION4_BACK_CHAN))
7988                 goto out;
7989         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7990                 return -EINVAL;
7991         if (rcvd->max_resp_sz < sent->max_resp_sz)
7992                 return -EINVAL;
7993         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7994                 return -EINVAL;
7995         if (rcvd->max_ops > sent->max_ops)
7996                 return -EINVAL;
7997         if (rcvd->max_reqs > sent->max_reqs)
7998                 return -EINVAL;
7999 out:
8000         return 0;
8001 }
8002
8003 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8004                                      struct nfs41_create_session_res *res)
8005 {
8006         int ret;
8007
8008         ret = nfs4_verify_fore_channel_attrs(args, res);
8009         if (ret)
8010                 return ret;
8011         return nfs4_verify_back_channel_attrs(args, res);
8012 }
8013
8014 static void nfs4_update_session(struct nfs4_session *session,
8015                 struct nfs41_create_session_res *res)
8016 {
8017         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8018         /* Mark client id and session as being confirmed */
8019         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8020         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8021         session->flags = res->flags;
8022         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8023         if (res->flags & SESSION4_BACK_CHAN)
8024                 memcpy(&session->bc_attrs, &res->bc_attrs,
8025                                 sizeof(session->bc_attrs));
8026 }
8027
8028 static int _nfs4_proc_create_session(struct nfs_client *clp,
8029                 struct rpc_cred *cred)
8030 {
8031         struct nfs4_session *session = clp->cl_session;
8032         struct nfs41_create_session_args args = {
8033                 .client = clp,
8034                 .clientid = clp->cl_clientid,
8035                 .seqid = clp->cl_seqid,
8036                 .cb_program = NFS4_CALLBACK,
8037         };
8038         struct nfs41_create_session_res res;
8039
8040         struct rpc_message msg = {
8041                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8042                 .rpc_argp = &args,
8043                 .rpc_resp = &res,
8044                 .rpc_cred = cred,
8045         };
8046         int status;
8047
8048         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8049         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8050
8051         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8052         trace_nfs4_create_session(clp, status);
8053
8054         switch (status) {
8055         case -NFS4ERR_STALE_CLIENTID:
8056         case -NFS4ERR_DELAY:
8057         case -ETIMEDOUT:
8058         case -EACCES:
8059         case -EAGAIN:
8060                 goto out;
8061         };
8062
8063         clp->cl_seqid++;
8064         if (!status) {
8065                 /* Verify the session's negotiated channel_attrs values */
8066                 status = nfs4_verify_channel_attrs(&args, &res);
8067                 /* Increment the clientid slot sequence id */
8068                 if (status)
8069                         goto out;
8070                 nfs4_update_session(session, &res);
8071         }
8072 out:
8073         return status;
8074 }
8075
8076 /*
8077  * Issues a CREATE_SESSION operation to the server.
8078  * It is the responsibility of the caller to verify the session is
8079  * expired before calling this routine.
8080  */
8081 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8082 {
8083         int status;
8084         unsigned *ptr;
8085         struct nfs4_session *session = clp->cl_session;
8086
8087         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8088
8089         status = _nfs4_proc_create_session(clp, cred);
8090         if (status)
8091                 goto out;
8092
8093         /* Init or reset the session slot tables */
8094         status = nfs4_setup_session_slot_tables(session);
8095         dprintk("slot table setup returned %d\n", status);
8096         if (status)
8097                 goto out;
8098
8099         ptr = (unsigned *)&session->sess_id.data[0];
8100         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8101                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8102 out:
8103         dprintk("<-- %s\n", __func__);
8104         return status;
8105 }
8106
8107 /*
8108  * Issue the over-the-wire RPC DESTROY_SESSION.
8109  * The caller must serialize access to this routine.
8110  */
8111 int nfs4_proc_destroy_session(struct nfs4_session *session,
8112                 struct rpc_cred *cred)
8113 {
8114         struct rpc_message msg = {
8115                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8116                 .rpc_argp = session,
8117                 .rpc_cred = cred,
8118         };
8119         int status = 0;
8120
8121         dprintk("--> nfs4_proc_destroy_session\n");
8122
8123         /* session is still being setup */
8124         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8125                 return 0;
8126
8127         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8128         trace_nfs4_destroy_session(session->clp, status);
8129
8130         if (status)
8131                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8132                         "Session has been destroyed regardless...\n", status);
8133
8134         dprintk("<-- nfs4_proc_destroy_session\n");
8135         return status;
8136 }
8137
8138 /*
8139  * Renew the cl_session lease.
8140  */
8141 struct nfs4_sequence_data {
8142         struct nfs_client *clp;
8143         struct nfs4_sequence_args args;
8144         struct nfs4_sequence_res res;
8145 };
8146
8147 static void nfs41_sequence_release(void *data)
8148 {
8149         struct nfs4_sequence_data *calldata = data;
8150         struct nfs_client *clp = calldata->clp;
8151
8152         if (atomic_read(&clp->cl_count) > 1)
8153                 nfs4_schedule_state_renewal(clp);
8154         nfs_put_client(clp);
8155         kfree(calldata);
8156 }
8157
8158 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8159 {
8160         switch(task->tk_status) {
8161         case -NFS4ERR_DELAY:
8162                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8163                 return -EAGAIN;
8164         default:
8165                 nfs4_schedule_lease_recovery(clp);
8166         }
8167         return 0;
8168 }
8169
8170 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8171 {
8172         struct nfs4_sequence_data *calldata = data;
8173         struct nfs_client *clp = calldata->clp;
8174
8175         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8176                 return;
8177
8178         trace_nfs4_sequence(clp, task->tk_status);
8179         if (task->tk_status < 0) {
8180                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8181                 if (atomic_read(&clp->cl_count) == 1)
8182                         goto out;
8183
8184                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8185                         rpc_restart_call_prepare(task);
8186                         return;
8187                 }
8188         }
8189         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8190 out:
8191         dprintk("<-- %s\n", __func__);
8192 }
8193
8194 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8195 {
8196         struct nfs4_sequence_data *calldata = data;
8197         struct nfs_client *clp = calldata->clp;
8198         struct nfs4_sequence_args *args;
8199         struct nfs4_sequence_res *res;
8200
8201         args = task->tk_msg.rpc_argp;
8202         res = task->tk_msg.rpc_resp;
8203
8204         nfs41_setup_sequence(clp->cl_session, args, res, task);
8205 }
8206
8207 static const struct rpc_call_ops nfs41_sequence_ops = {
8208         .rpc_call_done = nfs41_sequence_call_done,
8209         .rpc_call_prepare = nfs41_sequence_prepare,
8210         .rpc_release = nfs41_sequence_release,
8211 };
8212
8213 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8214                 struct rpc_cred *cred,
8215                 bool is_privileged)
8216 {
8217         struct nfs4_sequence_data *calldata;
8218         struct rpc_message msg = {
8219                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8220                 .rpc_cred = cred,
8221         };
8222         struct rpc_task_setup task_setup_data = {
8223                 .rpc_client = clp->cl_rpcclient,
8224                 .rpc_message = &msg,
8225                 .callback_ops = &nfs41_sequence_ops,
8226                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8227         };
8228
8229         if (!atomic_inc_not_zero(&clp->cl_count))
8230                 return ERR_PTR(-EIO);
8231         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8232         if (calldata == NULL) {
8233                 nfs_put_client(clp);
8234                 return ERR_PTR(-ENOMEM);
8235         }
8236         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8237         if (is_privileged)
8238                 nfs4_set_sequence_privileged(&calldata->args);
8239         msg.rpc_argp = &calldata->args;
8240         msg.rpc_resp = &calldata->res;
8241         calldata->clp = clp;
8242         task_setup_data.callback_data = calldata;
8243
8244         return rpc_run_task(&task_setup_data);
8245 }
8246
8247 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8248 {
8249         struct rpc_task *task;
8250         int ret = 0;
8251
8252         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8253                 return -EAGAIN;
8254         task = _nfs41_proc_sequence(clp, cred, false);
8255         if (IS_ERR(task))
8256                 ret = PTR_ERR(task);
8257         else
8258                 rpc_put_task_async(task);
8259         dprintk("<-- %s status=%d\n", __func__, ret);
8260         return ret;
8261 }
8262
8263 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8264 {
8265         struct rpc_task *task;
8266         int ret;
8267
8268         task = _nfs41_proc_sequence(clp, cred, true);
8269         if (IS_ERR(task)) {
8270                 ret = PTR_ERR(task);
8271                 goto out;
8272         }
8273         ret = rpc_wait_for_completion_task(task);
8274         if (!ret)
8275                 ret = task->tk_status;
8276         rpc_put_task(task);
8277 out:
8278         dprintk("<-- %s status=%d\n", __func__, ret);
8279         return ret;
8280 }
8281
8282 struct nfs4_reclaim_complete_data {
8283         struct nfs_client *clp;
8284         struct nfs41_reclaim_complete_args arg;
8285         struct nfs41_reclaim_complete_res res;
8286 };
8287
8288 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8289 {
8290         struct nfs4_reclaim_complete_data *calldata = data;
8291
8292         nfs41_setup_sequence(calldata->clp->cl_session,
8293                         &calldata->arg.seq_args,
8294                         &calldata->res.seq_res,
8295                         task);
8296 }
8297
8298 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8299 {
8300         switch(task->tk_status) {
8301         case 0:
8302         case -NFS4ERR_COMPLETE_ALREADY:
8303         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8304                 break;
8305         case -NFS4ERR_DELAY:
8306                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8307                 /* fall through */
8308         case -NFS4ERR_RETRY_UNCACHED_REP:
8309                 return -EAGAIN;
8310         default:
8311                 nfs4_schedule_lease_recovery(clp);
8312         }
8313         return 0;
8314 }
8315
8316 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8317 {
8318         struct nfs4_reclaim_complete_data *calldata = data;
8319         struct nfs_client *clp = calldata->clp;
8320         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8321
8322         dprintk("--> %s\n", __func__);
8323         if (!nfs41_sequence_done(task, res))
8324                 return;
8325
8326         trace_nfs4_reclaim_complete(clp, task->tk_status);
8327         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8328                 rpc_restart_call_prepare(task);
8329                 return;
8330         }
8331         dprintk("<-- %s\n", __func__);
8332 }
8333
8334 static void nfs4_free_reclaim_complete_data(void *data)
8335 {
8336         struct nfs4_reclaim_complete_data *calldata = data;
8337
8338         kfree(calldata);
8339 }
8340
8341 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8342         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8343         .rpc_call_done = nfs4_reclaim_complete_done,
8344         .rpc_release = nfs4_free_reclaim_complete_data,
8345 };
8346
8347 /*
8348  * Issue a global reclaim complete.
8349  */
8350 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8351                 struct rpc_cred *cred)
8352 {
8353         struct nfs4_reclaim_complete_data *calldata;
8354         struct rpc_task *task;
8355         struct rpc_message msg = {
8356                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8357                 .rpc_cred = cred,
8358         };
8359         struct rpc_task_setup task_setup_data = {
8360                 .rpc_client = clp->cl_rpcclient,
8361                 .rpc_message = &msg,
8362                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8363                 .flags = RPC_TASK_ASYNC,
8364         };
8365         int status = -ENOMEM;
8366
8367         dprintk("--> %s\n", __func__);
8368         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8369         if (calldata == NULL)
8370                 goto out;
8371         calldata->clp = clp;
8372         calldata->arg.one_fs = 0;
8373
8374         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8375         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8376         msg.rpc_argp = &calldata->arg;
8377         msg.rpc_resp = &calldata->res;
8378         task_setup_data.callback_data = calldata;
8379         task = rpc_run_task(&task_setup_data);
8380         if (IS_ERR(task)) {
8381                 status = PTR_ERR(task);
8382                 goto out;
8383         }
8384         status = nfs4_wait_for_completion_rpc_task(task);
8385         if (status == 0)
8386                 status = task->tk_status;
8387         rpc_put_task(task);
8388         return 0;
8389 out:
8390         dprintk("<-- %s status=%d\n", __func__, status);
8391         return status;
8392 }
8393
8394 static void
8395 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8396 {
8397         struct nfs4_layoutget *lgp = calldata;
8398         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8399         struct nfs4_session *session = nfs4_get_session(server);
8400
8401         dprintk("--> %s\n", __func__);
8402         nfs41_setup_sequence(session, &lgp->args.seq_args,
8403                                 &lgp->res.seq_res, task);
8404         dprintk("<-- %s\n", __func__);
8405 }
8406
8407 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8408 {
8409         struct nfs4_layoutget *lgp = calldata;
8410
8411         dprintk("--> %s\n", __func__);
8412         nfs41_sequence_process(task, &lgp->res.seq_res);
8413         dprintk("<-- %s\n", __func__);
8414 }
8415
8416 static int
8417 nfs4_layoutget_handle_exception(struct rpc_task *task,
8418                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8419 {
8420         struct inode *inode = lgp->args.inode;
8421         struct nfs_server *server = NFS_SERVER(inode);
8422         struct pnfs_layout_hdr *lo;
8423         int nfs4err = task->tk_status;
8424         int err, status = 0;
8425         LIST_HEAD(head);
8426
8427         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8428
8429         switch (nfs4err) {
8430         case 0:
8431                 goto out;
8432
8433         /*
8434          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8435          * on the file. set tk_status to -ENODATA to tell upper layer to
8436          * retry go inband.
8437          */
8438         case -NFS4ERR_LAYOUTUNAVAILABLE:
8439                 status = -ENODATA;
8440                 goto out;
8441         /*
8442          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8443          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8444          */
8445         case -NFS4ERR_BADLAYOUT:
8446                 status = -EOVERFLOW;
8447                 goto out;
8448         /*
8449          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8450          * (or clients) writing to the same RAID stripe except when
8451          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8452          *
8453          * Treat it like we would RECALLCONFLICT -- we retry for a little
8454          * while, and then eventually give up.
8455          */
8456         case -NFS4ERR_LAYOUTTRYLATER:
8457                 if (lgp->args.minlength == 0) {
8458                         status = -EOVERFLOW;
8459                         goto out;
8460                 }
8461                 status = -EBUSY;
8462                 break;
8463         case -NFS4ERR_RECALLCONFLICT:
8464                 status = -ERECALLCONFLICT;
8465                 break;
8466         case -NFS4ERR_DELEG_REVOKED:
8467         case -NFS4ERR_ADMIN_REVOKED:
8468         case -NFS4ERR_EXPIRED:
8469         case -NFS4ERR_BAD_STATEID:
8470                 exception->timeout = 0;
8471                 spin_lock(&inode->i_lock);
8472                 lo = NFS_I(inode)->layout;
8473                 /* If the open stateid was bad, then recover it. */
8474                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8475                     nfs4_stateid_match_other(&lgp->args.stateid,
8476                                         &lgp->args.ctx->state->stateid)) {
8477                         spin_unlock(&inode->i_lock);
8478                         exception->state = lgp->args.ctx->state;
8479                         exception->stateid = &lgp->args.stateid;
8480                         break;
8481                 }
8482
8483                 /*
8484                  * Mark the bad layout state as invalid, then retry
8485                  */
8486                 pnfs_mark_layout_stateid_invalid(lo, &head);
8487                 spin_unlock(&inode->i_lock);
8488                 pnfs_free_lseg_list(&head);
8489                 status = -EAGAIN;
8490                 goto out;
8491         }
8492
8493         err = nfs4_handle_exception(server, nfs4err, exception);
8494         if (!status) {
8495                 if (exception->retry)
8496                         status = -EAGAIN;
8497                 else
8498                         status = err;
8499         }
8500 out:
8501         dprintk("<-- %s\n", __func__);
8502         return status;
8503 }
8504
8505 static size_t max_response_pages(struct nfs_server *server)
8506 {
8507         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8508         return nfs_page_array_len(0, max_resp_sz);
8509 }
8510
8511 static void nfs4_free_pages(struct page **pages, size_t size)
8512 {
8513         int i;
8514
8515         if (!pages)
8516                 return;
8517
8518         for (i = 0; i < size; i++) {
8519                 if (!pages[i])
8520                         break;
8521                 __free_page(pages[i]);
8522         }
8523         kfree(pages);
8524 }
8525
8526 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8527 {
8528         struct page **pages;
8529         int i;
8530
8531         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8532         if (!pages) {
8533                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8534                 return NULL;
8535         }
8536
8537         for (i = 0; i < size; i++) {
8538                 pages[i] = alloc_page(gfp_flags);
8539                 if (!pages[i]) {
8540                         dprintk("%s: failed to allocate page\n", __func__);
8541                         nfs4_free_pages(pages, size);
8542                         return NULL;
8543                 }
8544         }
8545
8546         return pages;
8547 }
8548
8549 static void nfs4_layoutget_release(void *calldata)
8550 {
8551         struct nfs4_layoutget *lgp = calldata;
8552         struct inode *inode = lgp->args.inode;
8553         struct nfs_server *server = NFS_SERVER(inode);
8554         size_t max_pages = max_response_pages(server);
8555
8556         dprintk("--> %s\n", __func__);
8557         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8558         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8559         put_nfs_open_context(lgp->args.ctx);
8560         kfree(calldata);
8561         dprintk("<-- %s\n", __func__);
8562 }
8563
8564 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8565         .rpc_call_prepare = nfs4_layoutget_prepare,
8566         .rpc_call_done = nfs4_layoutget_done,
8567         .rpc_release = nfs4_layoutget_release,
8568 };
8569
8570 struct pnfs_layout_segment *
8571 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8572 {
8573         struct inode *inode = lgp->args.inode;
8574         struct nfs_server *server = NFS_SERVER(inode);
8575         size_t max_pages = max_response_pages(server);
8576         struct rpc_task *task;
8577         struct rpc_message msg = {
8578                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8579                 .rpc_argp = &lgp->args,
8580                 .rpc_resp = &lgp->res,
8581                 .rpc_cred = lgp->cred,
8582         };
8583         struct rpc_task_setup task_setup_data = {
8584                 .rpc_client = server->client,
8585                 .rpc_message = &msg,
8586                 .callback_ops = &nfs4_layoutget_call_ops,
8587                 .callback_data = lgp,
8588                 .flags = RPC_TASK_ASYNC,
8589         };
8590         struct pnfs_layout_segment *lseg = NULL;
8591         struct nfs4_exception exception = {
8592                 .inode = inode,
8593                 .timeout = *timeout,
8594         };
8595         int status = 0;
8596
8597         dprintk("--> %s\n", __func__);
8598
8599         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8600         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8601
8602         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8603         if (!lgp->args.layout.pages) {
8604                 nfs4_layoutget_release(lgp);
8605                 return ERR_PTR(-ENOMEM);
8606         }
8607         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8608
8609         lgp->res.layoutp = &lgp->args.layout;
8610         lgp->res.seq_res.sr_slot = NULL;
8611         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8612
8613         task = rpc_run_task(&task_setup_data);
8614         if (IS_ERR(task))
8615                 return ERR_CAST(task);
8616         status = nfs4_wait_for_completion_rpc_task(task);
8617         if (status == 0) {
8618                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8619                 *timeout = exception.timeout;
8620         }
8621
8622         trace_nfs4_layoutget(lgp->args.ctx,
8623                         &lgp->args.range,
8624                         &lgp->res.range,
8625                         &lgp->res.stateid,
8626                         status);
8627
8628         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8629         if (status == 0 && lgp->res.layoutp->len)
8630                 lseg = pnfs_layout_process(lgp);
8631         nfs4_sequence_free_slot(&lgp->res.seq_res);
8632         rpc_put_task(task);
8633         dprintk("<-- %s status=%d\n", __func__, status);
8634         if (status)
8635                 return ERR_PTR(status);
8636         return lseg;
8637 }
8638
8639 static void
8640 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8641 {
8642         struct nfs4_layoutreturn *lrp = calldata;
8643
8644         dprintk("--> %s\n", __func__);
8645         nfs41_setup_sequence(lrp->clp->cl_session,
8646                         &lrp->args.seq_args,
8647                         &lrp->res.seq_res,
8648                         task);
8649 }
8650
8651 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8652 {
8653         struct nfs4_layoutreturn *lrp = calldata;
8654         struct nfs_server *server;
8655
8656         dprintk("--> %s\n", __func__);
8657
8658         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8659                 return;
8660
8661         server = NFS_SERVER(lrp->args.inode);
8662         switch (task->tk_status) {
8663         default:
8664                 task->tk_status = 0;
8665         case 0:
8666                 break;
8667         case -NFS4ERR_DELAY:
8668                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8669                         break;
8670                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8671                 rpc_restart_call_prepare(task);
8672                 return;
8673         }
8674         dprintk("<-- %s\n", __func__);
8675 }
8676
8677 static void nfs4_layoutreturn_release(void *calldata)
8678 {
8679         struct nfs4_layoutreturn *lrp = calldata;
8680         struct pnfs_layout_hdr *lo = lrp->args.layout;
8681
8682         dprintk("--> %s\n", __func__);
8683         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8684                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8685         nfs4_sequence_free_slot(&lrp->res.seq_res);
8686         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8687                 lrp->ld_private.ops->free(&lrp->ld_private);
8688         pnfs_put_layout_hdr(lrp->args.layout);
8689         nfs_iput_and_deactive(lrp->inode);
8690         kfree(calldata);
8691         dprintk("<-- %s\n", __func__);
8692 }
8693
8694 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8695         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8696         .rpc_call_done = nfs4_layoutreturn_done,
8697         .rpc_release = nfs4_layoutreturn_release,
8698 };
8699
8700 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8701 {
8702         struct rpc_task *task;
8703         struct rpc_message msg = {
8704                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8705                 .rpc_argp = &lrp->args,
8706                 .rpc_resp = &lrp->res,
8707                 .rpc_cred = lrp->cred,
8708         };
8709         struct rpc_task_setup task_setup_data = {
8710                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8711                 .rpc_message = &msg,
8712                 .callback_ops = &nfs4_layoutreturn_call_ops,
8713                 .callback_data = lrp,
8714         };
8715         int status = 0;
8716
8717         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8718                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8719                         &task_setup_data.rpc_client, &msg);
8720
8721         dprintk("--> %s\n", __func__);
8722         if (!sync) {
8723                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8724                 if (!lrp->inode) {
8725                         nfs4_layoutreturn_release(lrp);
8726                         return -EAGAIN;
8727                 }
8728                 task_setup_data.flags |= RPC_TASK_ASYNC;
8729         }
8730         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8731         task = rpc_run_task(&task_setup_data);
8732         if (IS_ERR(task))
8733                 return PTR_ERR(task);
8734         if (sync)
8735                 status = task->tk_status;
8736         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8737         dprintk("<-- %s status=%d\n", __func__, status);
8738         rpc_put_task(task);
8739         return status;
8740 }
8741
8742 static int
8743 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8744                 struct pnfs_device *pdev,
8745                 struct rpc_cred *cred)
8746 {
8747         struct nfs4_getdeviceinfo_args args = {
8748                 .pdev = pdev,
8749                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8750                         NOTIFY_DEVICEID4_DELETE,
8751         };
8752         struct nfs4_getdeviceinfo_res res = {
8753                 .pdev = pdev,
8754         };
8755         struct rpc_message msg = {
8756                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8757                 .rpc_argp = &args,
8758                 .rpc_resp = &res,
8759                 .rpc_cred = cred,
8760         };
8761         int status;
8762
8763         dprintk("--> %s\n", __func__);
8764         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8765         if (res.notification & ~args.notify_types)
8766                 dprintk("%s: unsupported notification\n", __func__);
8767         if (res.notification != args.notify_types)
8768                 pdev->nocache = 1;
8769
8770         dprintk("<-- %s status=%d\n", __func__, status);
8771
8772         return status;
8773 }
8774
8775 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8776                 struct pnfs_device *pdev,
8777                 struct rpc_cred *cred)
8778 {
8779         struct nfs4_exception exception = { };
8780         int err;
8781
8782         do {
8783                 err = nfs4_handle_exception(server,
8784                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8785                                         &exception);
8786         } while (exception.retry);
8787         return err;
8788 }
8789 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8790
8791 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8792 {
8793         struct nfs4_layoutcommit_data *data = calldata;
8794         struct nfs_server *server = NFS_SERVER(data->args.inode);
8795         struct nfs4_session *session = nfs4_get_session(server);
8796
8797         nfs41_setup_sequence(session,
8798                         &data->args.seq_args,
8799                         &data->res.seq_res,
8800                         task);
8801 }
8802
8803 static void
8804 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8805 {
8806         struct nfs4_layoutcommit_data *data = calldata;
8807         struct nfs_server *server = NFS_SERVER(data->args.inode);
8808
8809         if (!nfs41_sequence_done(task, &data->res.seq_res))
8810                 return;
8811
8812         switch (task->tk_status) { /* Just ignore these failures */
8813         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8814         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8815         case -NFS4ERR_BADLAYOUT:     /* no layout */
8816         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8817                 task->tk_status = 0;
8818         case 0:
8819                 break;
8820         default:
8821                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8822                         rpc_restart_call_prepare(task);
8823                         return;
8824                 }
8825         }
8826 }
8827
8828 static void nfs4_layoutcommit_release(void *calldata)
8829 {
8830         struct nfs4_layoutcommit_data *data = calldata;
8831
8832         pnfs_cleanup_layoutcommit(data);
8833         nfs_post_op_update_inode_force_wcc(data->args.inode,
8834                                            data->res.fattr);
8835         put_rpccred(data->cred);
8836         nfs_iput_and_deactive(data->inode);
8837         kfree(data);
8838 }
8839
8840 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8841         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8842         .rpc_call_done = nfs4_layoutcommit_done,
8843         .rpc_release = nfs4_layoutcommit_release,
8844 };
8845
8846 int
8847 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8848 {
8849         struct rpc_message msg = {
8850                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8851                 .rpc_argp = &data->args,
8852                 .rpc_resp = &data->res,
8853                 .rpc_cred = data->cred,
8854         };
8855         struct rpc_task_setup task_setup_data = {
8856                 .task = &data->task,
8857                 .rpc_client = NFS_CLIENT(data->args.inode),
8858                 .rpc_message = &msg,
8859                 .callback_ops = &nfs4_layoutcommit_ops,
8860                 .callback_data = data,
8861         };
8862         struct rpc_task *task;
8863         int status = 0;
8864
8865         dprintk("NFS: initiating layoutcommit call. sync %d "
8866                 "lbw: %llu inode %lu\n", sync,
8867                 data->args.lastbytewritten,
8868                 data->args.inode->i_ino);
8869
8870         if (!sync) {
8871                 data->inode = nfs_igrab_and_active(data->args.inode);
8872                 if (data->inode == NULL) {
8873                         nfs4_layoutcommit_release(data);
8874                         return -EAGAIN;
8875                 }
8876                 task_setup_data.flags = RPC_TASK_ASYNC;
8877         }
8878         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8879         task = rpc_run_task(&task_setup_data);
8880         if (IS_ERR(task))
8881                 return PTR_ERR(task);
8882         if (sync)
8883                 status = task->tk_status;
8884         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8885         dprintk("%s: status %d\n", __func__, status);
8886         rpc_put_task(task);
8887         return status;
8888 }
8889
8890 /**
8891  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8892  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8893  */
8894 static int
8895 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8896                     struct nfs_fsinfo *info,
8897                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8898 {
8899         struct nfs41_secinfo_no_name_args args = {
8900                 .style = SECINFO_STYLE_CURRENT_FH,
8901         };
8902         struct nfs4_secinfo_res res = {
8903                 .flavors = flavors,
8904         };
8905         struct rpc_message msg = {
8906                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8907                 .rpc_argp = &args,
8908                 .rpc_resp = &res,
8909         };
8910         struct rpc_clnt *clnt = server->client;
8911         struct rpc_cred *cred = NULL;
8912         int status;
8913
8914         if (use_integrity) {
8915                 clnt = server->nfs_client->cl_rpcclient;
8916                 cred = nfs4_get_clid_cred(server->nfs_client);
8917                 msg.rpc_cred = cred;
8918         }
8919
8920         dprintk("--> %s\n", __func__);
8921         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8922                                 &res.seq_res, 0);
8923         dprintk("<-- %s status=%d\n", __func__, status);
8924
8925         if (cred)
8926                 put_rpccred(cred);
8927
8928         return status;
8929 }
8930
8931 static int
8932 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8933                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8934 {
8935         struct nfs4_exception exception = { };
8936         int err;
8937         do {
8938                 /* first try using integrity protection */
8939                 err = -NFS4ERR_WRONGSEC;
8940
8941                 /* try to use integrity protection with machine cred */
8942                 if (_nfs4_is_integrity_protected(server->nfs_client))
8943                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8944                                                           flavors, true);
8945
8946                 /*
8947                  * if unable to use integrity protection, or SECINFO with
8948                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8949                  * disallowed by spec, but exists in deployed servers) use
8950                  * the current filesystem's rpc_client and the user cred.
8951                  */
8952                 if (err == -NFS4ERR_WRONGSEC)
8953                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8954                                                           flavors, false);
8955
8956                 switch (err) {
8957                 case 0:
8958                 case -NFS4ERR_WRONGSEC:
8959                 case -ENOTSUPP:
8960                         goto out;
8961                 default:
8962                         err = nfs4_handle_exception(server, err, &exception);
8963                 }
8964         } while (exception.retry);
8965 out:
8966         return err;
8967 }
8968
8969 static int
8970 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8971                     struct nfs_fsinfo *info)
8972 {
8973         int err;
8974         struct page *page;
8975         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8976         struct nfs4_secinfo_flavors *flavors;
8977         struct nfs4_secinfo4 *secinfo;
8978         int i;
8979
8980         page = alloc_page(GFP_KERNEL);
8981         if (!page) {
8982                 err = -ENOMEM;
8983                 goto out;
8984         }
8985
8986         flavors = page_address(page);
8987         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8988
8989         /*
8990          * Fall back on "guess and check" method if
8991          * the server doesn't support SECINFO_NO_NAME
8992          */
8993         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8994                 err = nfs4_find_root_sec(server, fhandle, info);
8995                 goto out_freepage;
8996         }
8997         if (err)
8998                 goto out_freepage;
8999
9000         for (i = 0; i < flavors->num_flavors; i++) {
9001                 secinfo = &flavors->flavors[i];
9002
9003                 switch (secinfo->flavor) {
9004                 case RPC_AUTH_NULL:
9005                 case RPC_AUTH_UNIX:
9006                 case RPC_AUTH_GSS:
9007                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9008                                         &secinfo->flavor_info);
9009                         break;
9010                 default:
9011                         flavor = RPC_AUTH_MAXFLAVOR;
9012                         break;
9013                 }
9014
9015                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9016                         flavor = RPC_AUTH_MAXFLAVOR;
9017
9018                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9019                         err = nfs4_lookup_root_sec(server, fhandle,
9020                                                    info, flavor);
9021                         if (!err)
9022                                 break;
9023                 }
9024         }
9025
9026         if (flavor == RPC_AUTH_MAXFLAVOR)
9027                 err = -EPERM;
9028
9029 out_freepage:
9030         put_page(page);
9031         if (err == -EACCES)
9032                 return -EPERM;
9033 out:
9034         return err;
9035 }
9036
9037 static int _nfs41_test_stateid(struct nfs_server *server,
9038                 nfs4_stateid *stateid,
9039                 struct rpc_cred *cred)
9040 {
9041         int status;
9042         struct nfs41_test_stateid_args args = {
9043                 .stateid = stateid,
9044         };
9045         struct nfs41_test_stateid_res res;
9046         struct rpc_message msg = {
9047                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9048                 .rpc_argp = &args,
9049                 .rpc_resp = &res,
9050                 .rpc_cred = cred,
9051         };
9052         struct rpc_clnt *rpc_client = server->client;
9053
9054         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9055                 &rpc_client, &msg);
9056
9057         dprintk("NFS call  test_stateid %p\n", stateid);
9058         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
9059         nfs4_set_sequence_privileged(&args.seq_args);
9060         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9061                         &args.seq_args, &res.seq_res);
9062         if (status != NFS_OK) {
9063                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9064                 return status;
9065         }
9066         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9067         return -res.status;
9068 }
9069
9070 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9071                 int err, struct nfs4_exception *exception)
9072 {
9073         exception->retry = 0;
9074         switch(err) {
9075         case -NFS4ERR_DELAY:
9076         case -NFS4ERR_RETRY_UNCACHED_REP:
9077                 nfs4_handle_exception(server, err, exception);
9078                 break;
9079         case -NFS4ERR_BADSESSION:
9080         case -NFS4ERR_BADSLOT:
9081         case -NFS4ERR_BAD_HIGH_SLOT:
9082         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9083         case -NFS4ERR_DEADSESSION:
9084                 nfs4_do_handle_exception(server, err, exception);
9085         }
9086 }
9087
9088 /**
9089  * nfs41_test_stateid - perform a TEST_STATEID operation
9090  *
9091  * @server: server / transport on which to perform the operation
9092  * @stateid: state ID to test
9093  * @cred: credential
9094  *
9095  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9096  * Otherwise a negative NFS4ERR value is returned if the operation
9097  * failed or the state ID is not currently valid.
9098  */
9099 static int nfs41_test_stateid(struct nfs_server *server,
9100                 nfs4_stateid *stateid,
9101                 struct rpc_cred *cred)
9102 {
9103         struct nfs4_exception exception = { };
9104         int err;
9105         do {
9106                 err = _nfs41_test_stateid(server, stateid, cred);
9107                 nfs4_handle_delay_or_session_error(server, err, &exception);
9108         } while (exception.retry);
9109         return err;
9110 }
9111
9112 struct nfs_free_stateid_data {
9113         struct nfs_server *server;
9114         struct nfs41_free_stateid_args args;
9115         struct nfs41_free_stateid_res res;
9116 };
9117
9118 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9119 {
9120         struct nfs_free_stateid_data *data = calldata;
9121         nfs41_setup_sequence(nfs4_get_session(data->server),
9122                         &data->args.seq_args,
9123                         &data->res.seq_res,
9124                         task);
9125 }
9126
9127 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9128 {
9129         struct nfs_free_stateid_data *data = calldata;
9130
9131         nfs41_sequence_done(task, &data->res.seq_res);
9132
9133         switch (task->tk_status) {
9134         case -NFS4ERR_DELAY:
9135                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9136                         rpc_restart_call_prepare(task);
9137         }
9138 }
9139
9140 static void nfs41_free_stateid_release(void *calldata)
9141 {
9142         kfree(calldata);
9143 }
9144
9145 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9146         .rpc_call_prepare = nfs41_free_stateid_prepare,
9147         .rpc_call_done = nfs41_free_stateid_done,
9148         .rpc_release = nfs41_free_stateid_release,
9149 };
9150
9151 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9152                 const nfs4_stateid *stateid,
9153                 struct rpc_cred *cred,
9154                 bool privileged)
9155 {
9156         struct rpc_message msg = {
9157                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9158                 .rpc_cred = cred,
9159         };
9160         struct rpc_task_setup task_setup = {
9161                 .rpc_client = server->client,
9162                 .rpc_message = &msg,
9163                 .callback_ops = &nfs41_free_stateid_ops,
9164                 .flags = RPC_TASK_ASYNC,
9165         };
9166         struct nfs_free_stateid_data *data;
9167
9168         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9169                 &task_setup.rpc_client, &msg);
9170
9171         dprintk("NFS call  free_stateid %p\n", stateid);
9172         data = kmalloc(sizeof(*data), GFP_NOFS);
9173         if (!data)
9174                 return ERR_PTR(-ENOMEM);
9175         data->server = server;
9176         nfs4_stateid_copy(&data->args.stateid, stateid);
9177
9178         task_setup.callback_data = data;
9179
9180         msg.rpc_argp = &data->args;
9181         msg.rpc_resp = &data->res;
9182         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9183         if (privileged)
9184                 nfs4_set_sequence_privileged(&data->args.seq_args);
9185
9186         return rpc_run_task(&task_setup);
9187 }
9188
9189 /**
9190  * nfs41_free_stateid - perform a FREE_STATEID operation
9191  *
9192  * @server: server / transport on which to perform the operation
9193  * @stateid: state ID to release
9194  * @cred: credential
9195  * @is_recovery: set to true if this call needs to be privileged
9196  *
9197  * Note: this function is always asynchronous.
9198  */
9199 static int nfs41_free_stateid(struct nfs_server *server,
9200                 const nfs4_stateid *stateid,
9201                 struct rpc_cred *cred,
9202                 bool is_recovery)
9203 {
9204         struct rpc_task *task;
9205
9206         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9207         if (IS_ERR(task))
9208                 return PTR_ERR(task);
9209         rpc_put_task(task);
9210         return 0;
9211 }
9212
9213 static void
9214 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9215 {
9216         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9217
9218         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9219         nfs4_free_lock_state(server, lsp);
9220 }
9221
9222 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9223                 const nfs4_stateid *s2)
9224 {
9225         if (s1->type != s2->type)
9226                 return false;
9227
9228         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9229                 return false;
9230
9231         if (s1->seqid == s2->seqid)
9232                 return true;
9233         if (s1->seqid == 0 || s2->seqid == 0)
9234                 return true;
9235
9236         return false;
9237 }
9238
9239 #endif /* CONFIG_NFS_V4_1 */
9240
9241 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9242                 const nfs4_stateid *s2)
9243 {
9244         return nfs4_stateid_match(s1, s2);
9245 }
9246
9247
9248 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9249         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9250         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9251         .recover_open   = nfs4_open_reclaim,
9252         .recover_lock   = nfs4_lock_reclaim,
9253         .establish_clid = nfs4_init_clientid,
9254         .detect_trunking = nfs40_discover_server_trunking,
9255 };
9256
9257 #if defined(CONFIG_NFS_V4_1)
9258 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9259         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9260         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9261         .recover_open   = nfs4_open_reclaim,
9262         .recover_lock   = nfs4_lock_reclaim,
9263         .establish_clid = nfs41_init_clientid,
9264         .reclaim_complete = nfs41_proc_reclaim_complete,
9265         .detect_trunking = nfs41_discover_server_trunking,
9266 };
9267 #endif /* CONFIG_NFS_V4_1 */
9268
9269 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9270         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9271         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9272         .recover_open   = nfs40_open_expired,
9273         .recover_lock   = nfs4_lock_expired,
9274         .establish_clid = nfs4_init_clientid,
9275 };
9276
9277 #if defined(CONFIG_NFS_V4_1)
9278 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9279         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9280         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9281         .recover_open   = nfs41_open_expired,
9282         .recover_lock   = nfs41_lock_expired,
9283         .establish_clid = nfs41_init_clientid,
9284 };
9285 #endif /* CONFIG_NFS_V4_1 */
9286
9287 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9288         .sched_state_renewal = nfs4_proc_async_renew,
9289         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9290         .renew_lease = nfs4_proc_renew,
9291 };
9292
9293 #if defined(CONFIG_NFS_V4_1)
9294 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9295         .sched_state_renewal = nfs41_proc_async_sequence,
9296         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9297         .renew_lease = nfs4_proc_sequence,
9298 };
9299 #endif
9300
9301 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9302         .get_locations = _nfs40_proc_get_locations,
9303         .fsid_present = _nfs40_proc_fsid_present,
9304 };
9305
9306 #if defined(CONFIG_NFS_V4_1)
9307 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9308         .get_locations = _nfs41_proc_get_locations,
9309         .fsid_present = _nfs41_proc_fsid_present,
9310 };
9311 #endif  /* CONFIG_NFS_V4_1 */
9312
9313 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9314         .minor_version = 0,
9315         .init_caps = NFS_CAP_READDIRPLUS
9316                 | NFS_CAP_ATOMIC_OPEN
9317                 | NFS_CAP_POSIX_LOCK,
9318         .init_client = nfs40_init_client,
9319         .shutdown_client = nfs40_shutdown_client,
9320         .match_stateid = nfs4_match_stateid,
9321         .find_root_sec = nfs4_find_root_sec,
9322         .free_lock_state = nfs4_release_lockowner,
9323         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9324         .alloc_seqid = nfs_alloc_seqid,
9325         .call_sync_ops = &nfs40_call_sync_ops,
9326         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9327         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9328         .state_renewal_ops = &nfs40_state_renewal_ops,
9329         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9330 };
9331
9332 #if defined(CONFIG_NFS_V4_1)
9333 static struct nfs_seqid *
9334 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9335 {
9336         return NULL;
9337 }
9338
9339 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9340         .minor_version = 1,
9341         .init_caps = NFS_CAP_READDIRPLUS
9342                 | NFS_CAP_ATOMIC_OPEN
9343                 | NFS_CAP_POSIX_LOCK
9344                 | NFS_CAP_STATEID_NFSV41
9345                 | NFS_CAP_ATOMIC_OPEN_V1,
9346         .init_client = nfs41_init_client,
9347         .shutdown_client = nfs41_shutdown_client,
9348         .match_stateid = nfs41_match_stateid,
9349         .find_root_sec = nfs41_find_root_sec,
9350         .free_lock_state = nfs41_free_lock_state,
9351         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9352         .alloc_seqid = nfs_alloc_no_seqid,
9353         .session_trunk = nfs4_test_session_trunk,
9354         .call_sync_ops = &nfs41_call_sync_ops,
9355         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9356         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9357         .state_renewal_ops = &nfs41_state_renewal_ops,
9358         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9359 };
9360 #endif
9361
9362 #if defined(CONFIG_NFS_V4_2)
9363 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9364         .minor_version = 2,
9365         .init_caps = NFS_CAP_READDIRPLUS
9366                 | NFS_CAP_ATOMIC_OPEN
9367                 | NFS_CAP_POSIX_LOCK
9368                 | NFS_CAP_STATEID_NFSV41
9369                 | NFS_CAP_ATOMIC_OPEN_V1
9370                 | NFS_CAP_ALLOCATE
9371                 | NFS_CAP_COPY
9372                 | NFS_CAP_DEALLOCATE
9373                 | NFS_CAP_SEEK
9374                 | NFS_CAP_LAYOUTSTATS
9375                 | NFS_CAP_CLONE,
9376         .init_client = nfs41_init_client,
9377         .shutdown_client = nfs41_shutdown_client,
9378         .match_stateid = nfs41_match_stateid,
9379         .find_root_sec = nfs41_find_root_sec,
9380         .free_lock_state = nfs41_free_lock_state,
9381         .call_sync_ops = &nfs41_call_sync_ops,
9382         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9383         .alloc_seqid = nfs_alloc_no_seqid,
9384         .session_trunk = nfs4_test_session_trunk,
9385         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9386         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9387         .state_renewal_ops = &nfs41_state_renewal_ops,
9388         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9389 };
9390 #endif
9391
9392 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9393         [0] = &nfs_v4_0_minor_ops,
9394 #if defined(CONFIG_NFS_V4_1)
9395         [1] = &nfs_v4_1_minor_ops,
9396 #endif
9397 #if defined(CONFIG_NFS_V4_2)
9398         [2] = &nfs_v4_2_minor_ops,
9399 #endif
9400 };
9401
9402 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9403 {
9404         ssize_t error, error2;
9405
9406         error = generic_listxattr(dentry, list, size);
9407         if (error < 0)
9408                 return error;
9409         if (list) {
9410                 list += error;
9411                 size -= error;
9412         }
9413
9414         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9415         if (error2 < 0)
9416                 return error2;
9417         return error + error2;
9418 }
9419
9420 static const struct inode_operations nfs4_dir_inode_operations = {
9421         .create         = nfs_create,
9422         .lookup         = nfs_lookup,
9423         .atomic_open    = nfs_atomic_open,
9424         .link           = nfs_link,
9425         .unlink         = nfs_unlink,
9426         .symlink        = nfs_symlink,
9427         .mkdir          = nfs_mkdir,
9428         .rmdir          = nfs_rmdir,
9429         .mknod          = nfs_mknod,
9430         .rename         = nfs_rename,
9431         .permission     = nfs_permission,
9432         .getattr        = nfs_getattr,
9433         .setattr        = nfs_setattr,
9434         .listxattr      = nfs4_listxattr,
9435 };
9436
9437 static const struct inode_operations nfs4_file_inode_operations = {
9438         .permission     = nfs_permission,
9439         .getattr        = nfs_getattr,
9440         .setattr        = nfs_setattr,
9441         .listxattr      = nfs4_listxattr,
9442 };
9443
9444 const struct nfs_rpc_ops nfs_v4_clientops = {
9445         .version        = 4,                    /* protocol version */
9446         .dentry_ops     = &nfs4_dentry_operations,
9447         .dir_inode_ops  = &nfs4_dir_inode_operations,
9448         .file_inode_ops = &nfs4_file_inode_operations,
9449         .file_ops       = &nfs4_file_operations,
9450         .getroot        = nfs4_proc_get_root,
9451         .submount       = nfs4_submount,
9452         .try_mount      = nfs4_try_mount,
9453         .getattr        = nfs4_proc_getattr,
9454         .setattr        = nfs4_proc_setattr,
9455         .lookup         = nfs4_proc_lookup,
9456         .access         = nfs4_proc_access,
9457         .readlink       = nfs4_proc_readlink,
9458         .create         = nfs4_proc_create,
9459         .remove         = nfs4_proc_remove,
9460         .unlink_setup   = nfs4_proc_unlink_setup,
9461         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9462         .unlink_done    = nfs4_proc_unlink_done,
9463         .rename_setup   = nfs4_proc_rename_setup,
9464         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9465         .rename_done    = nfs4_proc_rename_done,
9466         .link           = nfs4_proc_link,
9467         .symlink        = nfs4_proc_symlink,
9468         .mkdir          = nfs4_proc_mkdir,
9469         .rmdir          = nfs4_proc_remove,
9470         .readdir        = nfs4_proc_readdir,
9471         .mknod          = nfs4_proc_mknod,
9472         .statfs         = nfs4_proc_statfs,
9473         .fsinfo         = nfs4_proc_fsinfo,
9474         .pathconf       = nfs4_proc_pathconf,
9475         .set_capabilities = nfs4_server_capabilities,
9476         .decode_dirent  = nfs4_decode_dirent,
9477         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9478         .read_setup     = nfs4_proc_read_setup,
9479         .read_done      = nfs4_read_done,
9480         .write_setup    = nfs4_proc_write_setup,
9481         .write_done     = nfs4_write_done,
9482         .commit_setup   = nfs4_proc_commit_setup,
9483         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9484         .commit_done    = nfs4_commit_done,
9485         .lock           = nfs4_proc_lock,
9486         .clear_acl_cache = nfs4_zap_acl_attr,
9487         .close_context  = nfs4_close_context,
9488         .open_context   = nfs4_atomic_open,
9489         .have_delegation = nfs4_have_delegation,
9490         .return_delegation = nfs4_inode_return_delegation,
9491         .alloc_client   = nfs4_alloc_client,
9492         .init_client    = nfs4_init_client,
9493         .free_client    = nfs4_free_client,
9494         .create_server  = nfs4_create_server,
9495         .clone_server   = nfs_clone_server,
9496 };
9497
9498 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9499         .name   = XATTR_NAME_NFSV4_ACL,
9500         .list   = nfs4_xattr_list_nfs4_acl,
9501         .get    = nfs4_xattr_get_nfs4_acl,
9502         .set    = nfs4_xattr_set_nfs4_acl,
9503 };
9504
9505 const struct xattr_handler *nfs4_xattr_handlers[] = {
9506         &nfs4_xattr_nfs4_acl_handler,
9507 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9508         &nfs4_xattr_nfs4_label_handler,
9509 #endif
9510         NULL
9511 };
9512
9513 /*
9514  * Local variables:
9515  *  c-basic-offset: 8
9516  * End:
9517  */