NFSv4.1: LAYOUTGET EDELAY loops timeout to the MDS
[linux-3.10.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/nfs_idmap.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4session.h"
67 #include "fscache.h"
68
69 #define NFSDBG_FACILITY         NFSDBG_PROC
70
71 #define NFS4_POLL_RETRY_MIN     (HZ/10)
72 #define NFS4_POLL_RETRY_MAX     (15*HZ)
73
74 struct nfs4_opendata;
75 static int _nfs4_proc_open(struct nfs4_opendata *data);
76 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
77 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
78 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
79 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
80 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *);
81 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
82 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
83                             struct nfs_fattr *fattr, struct iattr *sattr,
84                             struct nfs4_state *state);
85 #ifdef CONFIG_NFS_V4_1
86 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
87 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
88 #endif
89 /* Prevent leaks of NFSv4 errors into userland */
90 static int nfs4_map_errors(int err)
91 {
92         if (err >= -1000)
93                 return err;
94         switch (err) {
95         case -NFS4ERR_RESOURCE:
96         case -NFS4ERR_LAYOUTTRYLATER:
97         case -NFS4ERR_RECALLCONFLICT:
98                 return -EREMOTEIO;
99         case -NFS4ERR_WRONGSEC:
100                 return -EPERM;
101         case -NFS4ERR_BADOWNER:
102         case -NFS4ERR_BADNAME:
103                 return -EINVAL;
104         case -NFS4ERR_SHARE_DENIED:
105                 return -EACCES;
106         case -NFS4ERR_MINOR_VERS_MISMATCH:
107                 return -EPROTONOSUPPORT;
108         case -NFS4ERR_ACCESS:
109                 return -EACCES;
110         default:
111                 dprintk("%s could not handle NFSv4 error %d\n",
112                                 __func__, -err);
113                 break;
114         }
115         return -EIO;
116 }
117
118 /*
119  * This is our standard bitmap for GETATTR requests.
120  */
121 const u32 nfs4_fattr_bitmap[3] = {
122         FATTR4_WORD0_TYPE
123         | FATTR4_WORD0_CHANGE
124         | FATTR4_WORD0_SIZE
125         | FATTR4_WORD0_FSID
126         | FATTR4_WORD0_FILEID,
127         FATTR4_WORD1_MODE
128         | FATTR4_WORD1_NUMLINKS
129         | FATTR4_WORD1_OWNER
130         | FATTR4_WORD1_OWNER_GROUP
131         | FATTR4_WORD1_RAWDEV
132         | FATTR4_WORD1_SPACE_USED
133         | FATTR4_WORD1_TIME_ACCESS
134         | FATTR4_WORD1_TIME_METADATA
135         | FATTR4_WORD1_TIME_MODIFY
136 };
137
138 static const u32 nfs4_pnfs_open_bitmap[3] = {
139         FATTR4_WORD0_TYPE
140         | FATTR4_WORD0_CHANGE
141         | FATTR4_WORD0_SIZE
142         | FATTR4_WORD0_FSID
143         | FATTR4_WORD0_FILEID,
144         FATTR4_WORD1_MODE
145         | FATTR4_WORD1_NUMLINKS
146         | FATTR4_WORD1_OWNER
147         | FATTR4_WORD1_OWNER_GROUP
148         | FATTR4_WORD1_RAWDEV
149         | FATTR4_WORD1_SPACE_USED
150         | FATTR4_WORD1_TIME_ACCESS
151         | FATTR4_WORD1_TIME_METADATA
152         | FATTR4_WORD1_TIME_MODIFY,
153         FATTR4_WORD2_MDSTHRESHOLD
154 };
155
156 static const u32 nfs4_open_noattr_bitmap[3] = {
157         FATTR4_WORD0_TYPE
158         | FATTR4_WORD0_CHANGE
159         | FATTR4_WORD0_FILEID,
160 };
161
162 const u32 nfs4_statfs_bitmap[2] = {
163         FATTR4_WORD0_FILES_AVAIL
164         | FATTR4_WORD0_FILES_FREE
165         | FATTR4_WORD0_FILES_TOTAL,
166         FATTR4_WORD1_SPACE_AVAIL
167         | FATTR4_WORD1_SPACE_FREE
168         | FATTR4_WORD1_SPACE_TOTAL
169 };
170
171 const u32 nfs4_pathconf_bitmap[2] = {
172         FATTR4_WORD0_MAXLINK
173         | FATTR4_WORD0_MAXNAME,
174         0
175 };
176
177 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
178                         | FATTR4_WORD0_MAXREAD
179                         | FATTR4_WORD0_MAXWRITE
180                         | FATTR4_WORD0_LEASE_TIME,
181                         FATTR4_WORD1_TIME_DELTA
182                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
183                         FATTR4_WORD2_LAYOUT_BLKSIZE
184 };
185
186 const u32 nfs4_fs_locations_bitmap[2] = {
187         FATTR4_WORD0_TYPE
188         | FATTR4_WORD0_CHANGE
189         | FATTR4_WORD0_SIZE
190         | FATTR4_WORD0_FSID
191         | FATTR4_WORD0_FILEID
192         | FATTR4_WORD0_FS_LOCATIONS,
193         FATTR4_WORD1_MODE
194         | FATTR4_WORD1_NUMLINKS
195         | FATTR4_WORD1_OWNER
196         | FATTR4_WORD1_OWNER_GROUP
197         | FATTR4_WORD1_RAWDEV
198         | FATTR4_WORD1_SPACE_USED
199         | FATTR4_WORD1_TIME_ACCESS
200         | FATTR4_WORD1_TIME_METADATA
201         | FATTR4_WORD1_TIME_MODIFY
202         | FATTR4_WORD1_MOUNTED_ON_FILEID
203 };
204
205 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
206                 struct nfs4_readdir_arg *readdir)
207 {
208         __be32 *start, *p;
209
210         if (cookie > 2) {
211                 readdir->cookie = cookie;
212                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
213                 return;
214         }
215
216         readdir->cookie = 0;
217         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
218         if (cookie == 2)
219                 return;
220         
221         /*
222          * NFSv4 servers do not return entries for '.' and '..'
223          * Therefore, we fake these entries here.  We let '.'
224          * have cookie 0 and '..' have cookie 1.  Note that
225          * when talking to the server, we always send cookie 0
226          * instead of 1 or 2.
227          */
228         start = p = kmap_atomic(*readdir->pages);
229         
230         if (cookie == 0) {
231                 *p++ = xdr_one;                                  /* next */
232                 *p++ = xdr_zero;                   /* cookie, first word */
233                 *p++ = xdr_one;                   /* cookie, second word */
234                 *p++ = xdr_one;                             /* entry len */
235                 memcpy(p, ".\0\0\0", 4);                        /* entry */
236                 p++;
237                 *p++ = xdr_one;                         /* bitmap length */
238                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
239                 *p++ = htonl(8);              /* attribute buffer length */
240                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
241         }
242         
243         *p++ = xdr_one;                                  /* next */
244         *p++ = xdr_zero;                   /* cookie, first word */
245         *p++ = xdr_two;                   /* cookie, second word */
246         *p++ = xdr_two;                             /* entry len */
247         memcpy(p, "..\0\0", 4);                         /* entry */
248         p++;
249         *p++ = xdr_one;                         /* bitmap length */
250         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
251         *p++ = htonl(8);              /* attribute buffer length */
252         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
253
254         readdir->pgbase = (char *)p - (char *)start;
255         readdir->count -= readdir->pgbase;
256         kunmap_atomic(start);
257 }
258
259 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
260 {
261         int res = 0;
262
263         might_sleep();
264
265         if (*timeout <= 0)
266                 *timeout = NFS4_POLL_RETRY_MIN;
267         if (*timeout > NFS4_POLL_RETRY_MAX)
268                 *timeout = NFS4_POLL_RETRY_MAX;
269         freezable_schedule_timeout_killable(*timeout);
270         if (fatal_signal_pending(current))
271                 res = -ERESTARTSYS;
272         *timeout <<= 1;
273         return res;
274 }
275
276 /* This is the error handling routine for processes that are allowed
277  * to sleep.
278  */
279 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
280 {
281         struct nfs_client *clp = server->nfs_client;
282         struct nfs4_state *state = exception->state;
283         struct inode *inode = exception->inode;
284         int ret = errorcode;
285
286         exception->retry = 0;
287         switch(errorcode) {
288                 case 0:
289                         return 0;
290                 case -NFS4ERR_OPENMODE:
291                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
292                                 nfs4_inode_return_delegation(inode);
293                                 exception->retry = 1;
294                                 return 0;
295                         }
296                         if (state == NULL)
297                                 break;
298                         nfs4_schedule_stateid_recovery(server, state);
299                         goto wait_on_recovery;
300                 case -NFS4ERR_DELEG_REVOKED:
301                 case -NFS4ERR_ADMIN_REVOKED:
302                 case -NFS4ERR_BAD_STATEID:
303                         if (state == NULL)
304                                 break;
305                         nfs_remove_bad_delegation(state->inode);
306                         nfs4_schedule_stateid_recovery(server, state);
307                         goto wait_on_recovery;
308                 case -NFS4ERR_EXPIRED:
309                         if (state != NULL)
310                                 nfs4_schedule_stateid_recovery(server, state);
311                 case -NFS4ERR_STALE_STATEID:
312                 case -NFS4ERR_STALE_CLIENTID:
313                         nfs4_schedule_lease_recovery(clp);
314                         goto wait_on_recovery;
315 #if defined(CONFIG_NFS_V4_1)
316                 case -NFS4ERR_BADSESSION:
317                 case -NFS4ERR_BADSLOT:
318                 case -NFS4ERR_BAD_HIGH_SLOT:
319                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
320                 case -NFS4ERR_DEADSESSION:
321                 case -NFS4ERR_SEQ_FALSE_RETRY:
322                 case -NFS4ERR_SEQ_MISORDERED:
323                         dprintk("%s ERROR: %d Reset session\n", __func__,
324                                 errorcode);
325                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
326                         goto wait_on_recovery;
327 #endif /* defined(CONFIG_NFS_V4_1) */
328                 case -NFS4ERR_FILE_OPEN:
329                         if (exception->timeout > HZ) {
330                                 /* We have retried a decent amount, time to
331                                  * fail
332                                  */
333                                 ret = -EBUSY;
334                                 break;
335                         }
336                 case -NFS4ERR_GRACE:
337                 case -NFS4ERR_DELAY:
338                         ret = nfs4_delay(server->client, &exception->timeout);
339                         if (ret != 0)
340                                 break;
341                 case -NFS4ERR_RETRY_UNCACHED_REP:
342                 case -NFS4ERR_OLD_STATEID:
343                         exception->retry = 1;
344                         break;
345                 case -NFS4ERR_BADOWNER:
346                         /* The following works around a Linux server bug! */
347                 case -NFS4ERR_BADNAME:
348                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
349                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
350                                 exception->retry = 1;
351                                 printk(KERN_WARNING "NFS: v4 server %s "
352                                                 "does not accept raw "
353                                                 "uid/gids. "
354                                                 "Reenabling the idmapper.\n",
355                                                 server->nfs_client->cl_hostname);
356                         }
357         }
358         /* We failed to handle the error */
359         return nfs4_map_errors(ret);
360 wait_on_recovery:
361         ret = nfs4_wait_clnt_recover(clp);
362         if (ret == 0)
363                 exception->retry = 1;
364         return ret;
365 }
366
367
368 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
369 {
370         spin_lock(&clp->cl_lock);
371         if (time_before(clp->cl_last_renewal,timestamp))
372                 clp->cl_last_renewal = timestamp;
373         spin_unlock(&clp->cl_lock);
374 }
375
376 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
377 {
378         do_renew_lease(server->nfs_client, timestamp);
379 }
380
381 #if defined(CONFIG_NFS_V4_1)
382
383 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
384 {
385         struct nfs4_session *session;
386         struct nfs4_slot_table *tbl;
387         bool send_new_highest_used_slotid = false;
388
389         if (!res->sr_slot) {
390                 /* just wake up the next guy waiting since
391                  * we may have not consumed a slot after all */
392                 dprintk("%s: No slot\n", __func__);
393                 return;
394         }
395         tbl = res->sr_slot->table;
396         session = tbl->session;
397
398         spin_lock(&tbl->slot_tbl_lock);
399         /* Be nice to the server: try to ensure that the last transmitted
400          * value for highest_user_slotid <= target_highest_slotid
401          */
402         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
403                 send_new_highest_used_slotid = true;
404
405         if (nfs41_wake_and_assign_slot(tbl, res->sr_slot)) {
406                 send_new_highest_used_slotid = false;
407                 goto out_unlock;
408         }
409         nfs4_free_slot(tbl, res->sr_slot);
410
411         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
412                 send_new_highest_used_slotid = false;
413 out_unlock:
414         spin_unlock(&tbl->slot_tbl_lock);
415         res->sr_slot = NULL;
416         if (send_new_highest_used_slotid)
417                 nfs41_server_notify_highest_slotid_update(session->clp);
418 }
419
420 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
421 {
422         struct nfs4_session *session;
423         struct nfs4_slot *slot;
424         struct nfs_client *clp;
425         bool interrupted = false;
426         int ret = 1;
427
428         /* don't increment the sequence number if the task wasn't sent */
429         if (!RPC_WAS_SENT(task))
430                 goto out;
431
432         slot = res->sr_slot;
433         session = slot->table->session;
434
435         if (slot->interrupted) {
436                 slot->interrupted = 0;
437                 interrupted = true;
438         }
439
440         /* Check the SEQUENCE operation status */
441         switch (res->sr_status) {
442         case 0:
443                 /* Update the slot's sequence and clientid lease timer */
444                 ++slot->seq_nr;
445                 clp = session->clp;
446                 do_renew_lease(clp, res->sr_timestamp);
447                 /* Check sequence flags */
448                 if (res->sr_status_flags != 0)
449                         nfs4_schedule_lease_recovery(clp);
450                 nfs41_update_target_slotid(slot->table, slot, res);
451                 break;
452         case 1:
453                 /*
454                  * sr_status remains 1 if an RPC level error occurred.
455                  * The server may or may not have processed the sequence
456                  * operation..
457                  * Mark the slot as having hosted an interrupted RPC call.
458                  */
459                 slot->interrupted = 1;
460                 goto out;
461         case -NFS4ERR_DELAY:
462                 /* The server detected a resend of the RPC call and
463                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
464                  * of RFC5661.
465                  */
466                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
467                         __func__,
468                         slot->slot_nr,
469                         slot->seq_nr);
470                 goto out_retry;
471         case -NFS4ERR_BADSLOT:
472                 /*
473                  * The slot id we used was probably retired. Try again
474                  * using a different slot id.
475                  */
476                 goto retry_nowait;
477         case -NFS4ERR_SEQ_MISORDERED:
478                 /*
479                  * Was the last operation on this sequence interrupted?
480                  * If so, retry after bumping the sequence number.
481                  */
482                 if (interrupted) {
483                         ++slot->seq_nr;
484                         goto retry_nowait;
485                 }
486                 /*
487                  * Could this slot have been previously retired?
488                  * If so, then the server may be expecting seq_nr = 1!
489                  */
490                 if (slot->seq_nr != 1) {
491                         slot->seq_nr = 1;
492                         goto retry_nowait;
493                 }
494                 break;
495         case -NFS4ERR_SEQ_FALSE_RETRY:
496                 ++slot->seq_nr;
497                 goto retry_nowait;
498         default:
499                 /* Just update the slot sequence no. */
500                 ++slot->seq_nr;
501         }
502 out:
503         /* The session may be reset by one of the error handlers. */
504         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
505         nfs41_sequence_free_slot(res);
506         return ret;
507 retry_nowait:
508         if (rpc_restart_call_prepare(task)) {
509                 task->tk_status = 0;
510                 ret = 0;
511         }
512         goto out;
513 out_retry:
514         if (!rpc_restart_call(task))
515                 goto out;
516         rpc_delay(task, NFS4_POLL_RETRY_MAX);
517         return 0;
518 }
519
520 static int nfs4_sequence_done(struct rpc_task *task,
521                                struct nfs4_sequence_res *res)
522 {
523         if (res->sr_slot == NULL)
524                 return 1;
525         return nfs41_sequence_done(task, res);
526 }
527
528 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
529                 struct nfs4_sequence_res *res, int cache_reply)
530 {
531         args->sa_slot = NULL;
532         args->sa_cache_this = 0;
533         args->sa_privileged = 0;
534         if (cache_reply)
535                 args->sa_cache_this = 1;
536         res->sr_slot = NULL;
537 }
538
539 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
540 {
541         args->sa_privileged = 1;
542 }
543
544 int nfs41_setup_sequence(struct nfs4_session *session,
545                                 struct nfs4_sequence_args *args,
546                                 struct nfs4_sequence_res *res,
547                                 struct rpc_task *task)
548 {
549         struct nfs4_slot *slot;
550         struct nfs4_slot_table *tbl;
551
552         dprintk("--> %s\n", __func__);
553         /* slot already allocated? */
554         if (res->sr_slot != NULL)
555                 goto out_success;
556
557         tbl = &session->fc_slot_table;
558
559         task->tk_timeout = 0;
560
561         spin_lock(&tbl->slot_tbl_lock);
562         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
563             !args->sa_privileged) {
564                 /* The state manager will wait until the slot table is empty */
565                 dprintk("%s session is draining\n", __func__);
566                 goto out_sleep;
567         }
568
569         slot = nfs4_alloc_slot(tbl);
570         if (IS_ERR(slot)) {
571                 /* If out of memory, try again in 1/4 second */
572                 if (slot == ERR_PTR(-ENOMEM))
573                         task->tk_timeout = HZ >> 2;
574                 dprintk("<-- %s: no free slots\n", __func__);
575                 goto out_sleep;
576         }
577         spin_unlock(&tbl->slot_tbl_lock);
578
579         args->sa_slot = slot;
580
581         dprintk("<-- %s slotid=%d seqid=%d\n", __func__,
582                         slot->slot_nr, slot->seq_nr);
583
584         res->sr_slot = slot;
585         res->sr_timestamp = jiffies;
586         res->sr_status_flags = 0;
587         /*
588          * sr_status is only set in decode_sequence, and so will remain
589          * set to 1 if an rpc level failure occurs.
590          */
591         res->sr_status = 1;
592 out_success:
593         rpc_call_start(task);
594         return 0;
595 out_sleep:
596         /* Privileged tasks are queued with top priority */
597         if (args->sa_privileged)
598                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
599                                 NULL, RPC_PRIORITY_PRIVILEGED);
600         else
601                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
602         spin_unlock(&tbl->slot_tbl_lock);
603         return -EAGAIN;
604 }
605 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
606
607 int nfs4_setup_sequence(const struct nfs_server *server,
608                         struct nfs4_sequence_args *args,
609                         struct nfs4_sequence_res *res,
610                         struct rpc_task *task)
611 {
612         struct nfs4_session *session = nfs4_get_session(server);
613         int ret = 0;
614
615         if (session == NULL) {
616                 rpc_call_start(task);
617                 goto out;
618         }
619
620         dprintk("--> %s clp %p session %p sr_slot %d\n",
621                 __func__, session->clp, session, res->sr_slot ?
622                         res->sr_slot->slot_nr : -1);
623
624         ret = nfs41_setup_sequence(session, args, res, task);
625 out:
626         dprintk("<-- %s status=%d\n", __func__, ret);
627         return ret;
628 }
629
630 struct nfs41_call_sync_data {
631         const struct nfs_server *seq_server;
632         struct nfs4_sequence_args *seq_args;
633         struct nfs4_sequence_res *seq_res;
634 };
635
636 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
637 {
638         struct nfs41_call_sync_data *data = calldata;
639         struct nfs4_session *session = nfs4_get_session(data->seq_server);
640
641         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
642
643         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
644 }
645
646 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
647 {
648         struct nfs41_call_sync_data *data = calldata;
649
650         nfs41_sequence_done(task, data->seq_res);
651 }
652
653 static const struct rpc_call_ops nfs41_call_sync_ops = {
654         .rpc_call_prepare = nfs41_call_sync_prepare,
655         .rpc_call_done = nfs41_call_sync_done,
656 };
657
658 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
659                                    struct nfs_server *server,
660                                    struct rpc_message *msg,
661                                    struct nfs4_sequence_args *args,
662                                    struct nfs4_sequence_res *res)
663 {
664         int ret;
665         struct rpc_task *task;
666         struct nfs41_call_sync_data data = {
667                 .seq_server = server,
668                 .seq_args = args,
669                 .seq_res = res,
670         };
671         struct rpc_task_setup task_setup = {
672                 .rpc_client = clnt,
673                 .rpc_message = msg,
674                 .callback_ops = &nfs41_call_sync_ops,
675                 .callback_data = &data
676         };
677
678         task = rpc_run_task(&task_setup);
679         if (IS_ERR(task))
680                 ret = PTR_ERR(task);
681         else {
682                 ret = task->tk_status;
683                 rpc_put_task(task);
684         }
685         return ret;
686 }
687
688 #else
689 static
690 void nfs41_init_sequence(struct nfs4_sequence_args *args,
691                 struct nfs4_sequence_res *res, int cache_reply)
692 {
693 }
694
695 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
696 {
697 }
698
699
700 static int nfs4_sequence_done(struct rpc_task *task,
701                                struct nfs4_sequence_res *res)
702 {
703         return 1;
704 }
705 #endif /* CONFIG_NFS_V4_1 */
706
707 static
708 int _nfs4_call_sync(struct rpc_clnt *clnt,
709                     struct nfs_server *server,
710                     struct rpc_message *msg,
711                     struct nfs4_sequence_args *args,
712                     struct nfs4_sequence_res *res)
713 {
714         return rpc_call_sync(clnt, msg, 0);
715 }
716
717 static
718 int nfs4_call_sync(struct rpc_clnt *clnt,
719                    struct nfs_server *server,
720                    struct rpc_message *msg,
721                    struct nfs4_sequence_args *args,
722                    struct nfs4_sequence_res *res,
723                    int cache_reply)
724 {
725         nfs41_init_sequence(args, res, cache_reply);
726         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
727                                                 args, res);
728 }
729
730 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
731 {
732         struct nfs_inode *nfsi = NFS_I(dir);
733
734         spin_lock(&dir->i_lock);
735         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
736         if (!cinfo->atomic || cinfo->before != dir->i_version)
737                 nfs_force_lookup_revalidate(dir);
738         dir->i_version = cinfo->after;
739         nfs_fscache_invalidate(dir);
740         spin_unlock(&dir->i_lock);
741 }
742
743 struct nfs4_opendata {
744         struct kref kref;
745         struct nfs_openargs o_arg;
746         struct nfs_openres o_res;
747         struct nfs_open_confirmargs c_arg;
748         struct nfs_open_confirmres c_res;
749         struct nfs4_string owner_name;
750         struct nfs4_string group_name;
751         struct nfs_fattr f_attr;
752         struct dentry *dir;
753         struct dentry *dentry;
754         struct nfs4_state_owner *owner;
755         struct nfs4_state *state;
756         struct iattr attrs;
757         unsigned long timestamp;
758         unsigned int rpc_done : 1;
759         int rpc_status;
760         int cancelled;
761 };
762
763
764 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
765 {
766         p->o_res.f_attr = &p->f_attr;
767         p->o_res.seqid = p->o_arg.seqid;
768         p->c_res.seqid = p->c_arg.seqid;
769         p->o_res.server = p->o_arg.server;
770         p->o_res.access_request = p->o_arg.access;
771         nfs_fattr_init(&p->f_attr);
772         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
773 }
774
775 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
776                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
777                 const struct iattr *attrs,
778                 gfp_t gfp_mask)
779 {
780         struct dentry *parent = dget_parent(dentry);
781         struct inode *dir = parent->d_inode;
782         struct nfs_server *server = NFS_SERVER(dir);
783         struct nfs4_opendata *p;
784
785         p = kzalloc(sizeof(*p), gfp_mask);
786         if (p == NULL)
787                 goto err;
788         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
789         if (p->o_arg.seqid == NULL)
790                 goto err_free;
791         nfs_sb_active(dentry->d_sb);
792         p->dentry = dget(dentry);
793         p->dir = parent;
794         p->owner = sp;
795         atomic_inc(&sp->so_count);
796         p->o_arg.fh = NFS_FH(dir);
797         p->o_arg.open_flags = flags;
798         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
799         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
800          * will return permission denied for all bits until close */
801         if (!(flags & O_EXCL)) {
802                 /* ask server to check for all possible rights as results
803                  * are cached */
804                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
805                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
806         }
807         p->o_arg.clientid = server->nfs_client->cl_clientid;
808         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
809         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
810         p->o_arg.name = &dentry->d_name;
811         p->o_arg.server = server;
812         p->o_arg.bitmask = server->attr_bitmask;
813         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
814         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
815         if (attrs != NULL && attrs->ia_valid != 0) {
816                 __be32 verf[2];
817
818                 p->o_arg.u.attrs = &p->attrs;
819                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
820
821                 verf[0] = jiffies;
822                 verf[1] = current->pid;
823                 memcpy(p->o_arg.u.verifier.data, verf,
824                                 sizeof(p->o_arg.u.verifier.data));
825         }
826         p->c_arg.fh = &p->o_res.fh;
827         p->c_arg.stateid = &p->o_res.stateid;
828         p->c_arg.seqid = p->o_arg.seqid;
829         nfs4_init_opendata_res(p);
830         kref_init(&p->kref);
831         return p;
832 err_free:
833         kfree(p);
834 err:
835         dput(parent);
836         return NULL;
837 }
838
839 static void nfs4_opendata_free(struct kref *kref)
840 {
841         struct nfs4_opendata *p = container_of(kref,
842                         struct nfs4_opendata, kref);
843         struct super_block *sb = p->dentry->d_sb;
844
845         nfs_free_seqid(p->o_arg.seqid);
846         if (p->state != NULL)
847                 nfs4_put_open_state(p->state);
848         nfs4_put_state_owner(p->owner);
849         dput(p->dir);
850         dput(p->dentry);
851         nfs_sb_deactive(sb);
852         nfs_fattr_free_names(&p->f_attr);
853         kfree(p);
854 }
855
856 static void nfs4_opendata_put(struct nfs4_opendata *p)
857 {
858         if (p != NULL)
859                 kref_put(&p->kref, nfs4_opendata_free);
860 }
861
862 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
863 {
864         int ret;
865
866         ret = rpc_wait_for_completion_task(task);
867         return ret;
868 }
869
870 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
871 {
872         int ret = 0;
873
874         if (open_mode & (O_EXCL|O_TRUNC))
875                 goto out;
876         switch (mode & (FMODE_READ|FMODE_WRITE)) {
877                 case FMODE_READ:
878                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
879                                 && state->n_rdonly != 0;
880                         break;
881                 case FMODE_WRITE:
882                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
883                                 && state->n_wronly != 0;
884                         break;
885                 case FMODE_READ|FMODE_WRITE:
886                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
887                                 && state->n_rdwr != 0;
888         }
889 out:
890         return ret;
891 }
892
893 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
894 {
895         if (delegation == NULL)
896                 return 0;
897         if ((delegation->type & fmode) != fmode)
898                 return 0;
899         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
900                 return 0;
901         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
902                 return 0;
903         nfs_mark_delegation_referenced(delegation);
904         return 1;
905 }
906
907 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
908 {
909         switch (fmode) {
910                 case FMODE_WRITE:
911                         state->n_wronly++;
912                         break;
913                 case FMODE_READ:
914                         state->n_rdonly++;
915                         break;
916                 case FMODE_READ|FMODE_WRITE:
917                         state->n_rdwr++;
918         }
919         nfs4_state_set_mode_locked(state, state->state | fmode);
920 }
921
922 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
923 {
924         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
925                 nfs4_stateid_copy(&state->stateid, stateid);
926         nfs4_stateid_copy(&state->open_stateid, stateid);
927         switch (fmode) {
928                 case FMODE_READ:
929                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
930                         break;
931                 case FMODE_WRITE:
932                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
933                         break;
934                 case FMODE_READ|FMODE_WRITE:
935                         set_bit(NFS_O_RDWR_STATE, &state->flags);
936         }
937 }
938
939 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
940 {
941         write_seqlock(&state->seqlock);
942         nfs_set_open_stateid_locked(state, stateid, fmode);
943         write_sequnlock(&state->seqlock);
944 }
945
946 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
947 {
948         /*
949          * Protect the call to nfs4_state_set_mode_locked and
950          * serialise the stateid update
951          */
952         write_seqlock(&state->seqlock);
953         if (deleg_stateid != NULL) {
954                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
955                 set_bit(NFS_DELEGATED_STATE, &state->flags);
956         }
957         if (open_stateid != NULL)
958                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
959         write_sequnlock(&state->seqlock);
960         spin_lock(&state->owner->so_lock);
961         update_open_stateflags(state, fmode);
962         spin_unlock(&state->owner->so_lock);
963 }
964
965 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
966 {
967         struct nfs_inode *nfsi = NFS_I(state->inode);
968         struct nfs_delegation *deleg_cur;
969         int ret = 0;
970
971         fmode &= (FMODE_READ|FMODE_WRITE);
972
973         rcu_read_lock();
974         deleg_cur = rcu_dereference(nfsi->delegation);
975         if (deleg_cur == NULL)
976                 goto no_delegation;
977
978         spin_lock(&deleg_cur->lock);
979         if (nfsi->delegation != deleg_cur ||
980            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
981             (deleg_cur->type & fmode) != fmode)
982                 goto no_delegation_unlock;
983
984         if (delegation == NULL)
985                 delegation = &deleg_cur->stateid;
986         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
987                 goto no_delegation_unlock;
988
989         nfs_mark_delegation_referenced(deleg_cur);
990         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
991         ret = 1;
992 no_delegation_unlock:
993         spin_unlock(&deleg_cur->lock);
994 no_delegation:
995         rcu_read_unlock();
996
997         if (!ret && open_stateid != NULL) {
998                 __update_open_stateid(state, open_stateid, NULL, fmode);
999                 ret = 1;
1000         }
1001
1002         return ret;
1003 }
1004
1005
1006 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1007 {
1008         struct nfs_delegation *delegation;
1009
1010         rcu_read_lock();
1011         delegation = rcu_dereference(NFS_I(inode)->delegation);
1012         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1013                 rcu_read_unlock();
1014                 return;
1015         }
1016         rcu_read_unlock();
1017         nfs4_inode_return_delegation(inode);
1018 }
1019
1020 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1021 {
1022         struct nfs4_state *state = opendata->state;
1023         struct nfs_inode *nfsi = NFS_I(state->inode);
1024         struct nfs_delegation *delegation;
1025         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1026         fmode_t fmode = opendata->o_arg.fmode;
1027         nfs4_stateid stateid;
1028         int ret = -EAGAIN;
1029
1030         for (;;) {
1031                 if (can_open_cached(state, fmode, open_mode)) {
1032                         spin_lock(&state->owner->so_lock);
1033                         if (can_open_cached(state, fmode, open_mode)) {
1034                                 update_open_stateflags(state, fmode);
1035                                 spin_unlock(&state->owner->so_lock);
1036                                 goto out_return_state;
1037                         }
1038                         spin_unlock(&state->owner->so_lock);
1039                 }
1040                 rcu_read_lock();
1041                 delegation = rcu_dereference(nfsi->delegation);
1042                 if (!can_open_delegated(delegation, fmode)) {
1043                         rcu_read_unlock();
1044                         break;
1045                 }
1046                 /* Save the delegation */
1047                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1048                 rcu_read_unlock();
1049                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1050                 if (ret != 0)
1051                         goto out;
1052                 ret = -EAGAIN;
1053
1054                 /* Try to update the stateid using the delegation */
1055                 if (update_open_stateid(state, NULL, &stateid, fmode))
1056                         goto out_return_state;
1057         }
1058 out:
1059         return ERR_PTR(ret);
1060 out_return_state:
1061         atomic_inc(&state->count);
1062         return state;
1063 }
1064
1065 static void
1066 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1067 {
1068         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1069         struct nfs_delegation *delegation;
1070         int delegation_flags = 0;
1071
1072         rcu_read_lock();
1073         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1074         if (delegation)
1075                 delegation_flags = delegation->flags;
1076         rcu_read_unlock();
1077         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1078                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1079                                    "returning a delegation for "
1080                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1081                                    clp->cl_hostname);
1082         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1083                 nfs_inode_set_delegation(state->inode,
1084                                          data->owner->so_cred,
1085                                          &data->o_res);
1086         else
1087                 nfs_inode_reclaim_delegation(state->inode,
1088                                              data->owner->so_cred,
1089                                              &data->o_res);
1090 }
1091
1092 /*
1093  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1094  * and update the nfs4_state.
1095  */
1096 static struct nfs4_state *
1097 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1098 {
1099         struct inode *inode = data->state->inode;
1100         struct nfs4_state *state = data->state;
1101         int ret;
1102
1103         if (!data->rpc_done) {
1104                 ret = data->rpc_status;
1105                 goto err;
1106         }
1107
1108         ret = -ESTALE;
1109         if (!(data->f_attr.valid & NFS_ATTR_FATTR_TYPE) ||
1110             !(data->f_attr.valid & NFS_ATTR_FATTR_FILEID) ||
1111             !(data->f_attr.valid & NFS_ATTR_FATTR_CHANGE))
1112                 goto err;
1113
1114         ret = -ENOMEM;
1115         state = nfs4_get_open_state(inode, data->owner);
1116         if (state == NULL)
1117                 goto err;
1118
1119         ret = nfs_refresh_inode(inode, &data->f_attr);
1120         if (ret)
1121                 goto err;
1122
1123         if (data->o_res.delegation_type != 0)
1124                 nfs4_opendata_check_deleg(data, state);
1125         update_open_stateid(state, &data->o_res.stateid, NULL,
1126                             data->o_arg.fmode);
1127
1128         return state;
1129 err:
1130         return ERR_PTR(ret);
1131
1132 }
1133
1134 static struct nfs4_state *
1135 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1136 {
1137         struct inode *inode;
1138         struct nfs4_state *state = NULL;
1139         int ret;
1140
1141         if (!data->rpc_done) {
1142                 state = nfs4_try_open_cached(data);
1143                 goto out;
1144         }
1145
1146         ret = -EAGAIN;
1147         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1148                 goto err;
1149         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1150         ret = PTR_ERR(inode);
1151         if (IS_ERR(inode))
1152                 goto err;
1153         ret = -ENOMEM;
1154         state = nfs4_get_open_state(inode, data->owner);
1155         if (state == NULL)
1156                 goto err_put_inode;
1157         if (data->o_res.delegation_type != 0)
1158                 nfs4_opendata_check_deleg(data, state);
1159         update_open_stateid(state, &data->o_res.stateid, NULL,
1160                         data->o_arg.fmode);
1161         iput(inode);
1162 out:
1163         nfs_release_seqid(data->o_arg.seqid);
1164         return state;
1165 err_put_inode:
1166         iput(inode);
1167 err:
1168         return ERR_PTR(ret);
1169 }
1170
1171 static struct nfs4_state *
1172 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1173 {
1174         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1175                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1176         return _nfs4_opendata_to_nfs4_state(data);
1177 }
1178
1179 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1180 {
1181         struct nfs_inode *nfsi = NFS_I(state->inode);
1182         struct nfs_open_context *ctx;
1183
1184         spin_lock(&state->inode->i_lock);
1185         list_for_each_entry(ctx, &nfsi->open_files, list) {
1186                 if (ctx->state != state)
1187                         continue;
1188                 get_nfs_open_context(ctx);
1189                 spin_unlock(&state->inode->i_lock);
1190                 return ctx;
1191         }
1192         spin_unlock(&state->inode->i_lock);
1193         return ERR_PTR(-ENOENT);
1194 }
1195
1196 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1197 {
1198         struct nfs4_opendata *opendata;
1199
1200         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1201         if (opendata == NULL)
1202                 return ERR_PTR(-ENOMEM);
1203         opendata->state = state;
1204         atomic_inc(&state->count);
1205         return opendata;
1206 }
1207
1208 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1209 {
1210         struct nfs4_state *newstate;
1211         int ret;
1212
1213         opendata->o_arg.open_flags = 0;
1214         opendata->o_arg.fmode = fmode;
1215         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1216         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1217         nfs4_init_opendata_res(opendata);
1218         ret = _nfs4_recover_proc_open(opendata);
1219         if (ret != 0)
1220                 return ret; 
1221         newstate = nfs4_opendata_to_nfs4_state(opendata);
1222         if (IS_ERR(newstate))
1223                 return PTR_ERR(newstate);
1224         nfs4_close_state(newstate, fmode);
1225         *res = newstate;
1226         return 0;
1227 }
1228
1229 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1230 {
1231         struct nfs4_state *newstate;
1232         int ret;
1233
1234         /* memory barrier prior to reading state->n_* */
1235         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1236         smp_rmb();
1237         if (state->n_rdwr != 0) {
1238                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1239                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1240                 if (ret != 0)
1241                         return ret;
1242                 if (newstate != state)
1243                         return -ESTALE;
1244         }
1245         if (state->n_wronly != 0) {
1246                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1247                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1248                 if (ret != 0)
1249                         return ret;
1250                 if (newstate != state)
1251                         return -ESTALE;
1252         }
1253         if (state->n_rdonly != 0) {
1254                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1255                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1256                 if (ret != 0)
1257                         return ret;
1258                 if (newstate != state)
1259                         return -ESTALE;
1260         }
1261         /*
1262          * We may have performed cached opens for all three recoveries.
1263          * Check if we need to update the current stateid.
1264          */
1265         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1266             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1267                 write_seqlock(&state->seqlock);
1268                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1269                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1270                 write_sequnlock(&state->seqlock);
1271         }
1272         return 0;
1273 }
1274
1275 /*
1276  * OPEN_RECLAIM:
1277  *      reclaim state on the server after a reboot.
1278  */
1279 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1280 {
1281         struct nfs_delegation *delegation;
1282         struct nfs4_opendata *opendata;
1283         fmode_t delegation_type = 0;
1284         int status;
1285
1286         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1287         if (IS_ERR(opendata))
1288                 return PTR_ERR(opendata);
1289         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1290         opendata->o_arg.fh = NFS_FH(state->inode);
1291         rcu_read_lock();
1292         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1293         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1294                 delegation_type = delegation->type;
1295         rcu_read_unlock();
1296         opendata->o_arg.u.delegation_type = delegation_type;
1297         status = nfs4_open_recover(opendata, state);
1298         nfs4_opendata_put(opendata);
1299         return status;
1300 }
1301
1302 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1303 {
1304         struct nfs_server *server = NFS_SERVER(state->inode);
1305         struct nfs4_exception exception = { };
1306         int err;
1307         do {
1308                 err = _nfs4_do_open_reclaim(ctx, state);
1309                 if (err != -NFS4ERR_DELAY)
1310                         break;
1311                 nfs4_handle_exception(server, err, &exception);
1312         } while (exception.retry);
1313         return err;
1314 }
1315
1316 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1317 {
1318         struct nfs_open_context *ctx;
1319         int ret;
1320
1321         ctx = nfs4_state_find_open_context(state);
1322         if (IS_ERR(ctx))
1323                 return PTR_ERR(ctx);
1324         ret = nfs4_do_open_reclaim(ctx, state);
1325         put_nfs_open_context(ctx);
1326         return ret;
1327 }
1328
1329 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1330 {
1331         struct nfs4_opendata *opendata;
1332         int ret;
1333
1334         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1335         if (IS_ERR(opendata))
1336                 return PTR_ERR(opendata);
1337         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1338         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1339         ret = nfs4_open_recover(opendata, state);
1340         nfs4_opendata_put(opendata);
1341         return ret;
1342 }
1343
1344 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1345 {
1346         struct nfs4_exception exception = { };
1347         struct nfs_server *server = NFS_SERVER(state->inode);
1348         int err;
1349         do {
1350                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1351                 switch (err) {
1352                         case 0:
1353                         case -ENOENT:
1354                         case -ESTALE:
1355                                 goto out;
1356                         case -NFS4ERR_BADSESSION:
1357                         case -NFS4ERR_BADSLOT:
1358                         case -NFS4ERR_BAD_HIGH_SLOT:
1359                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1360                         case -NFS4ERR_DEADSESSION:
1361                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1362                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1363                                 err = -EAGAIN;
1364                                 goto out;
1365                         case -NFS4ERR_STALE_CLIENTID:
1366                         case -NFS4ERR_STALE_STATEID:
1367                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1368                         case -NFS4ERR_EXPIRED:
1369                                 /* Don't recall a delegation if it was lost */
1370                                 nfs4_schedule_lease_recovery(server->nfs_client);
1371                                 err = -EAGAIN;
1372                                 goto out;
1373                         case -NFS4ERR_DELEG_REVOKED:
1374                         case -NFS4ERR_ADMIN_REVOKED:
1375                         case -NFS4ERR_BAD_STATEID:
1376                                 nfs_inode_find_state_and_recover(state->inode,
1377                                                 stateid);
1378                                 nfs4_schedule_stateid_recovery(server, state);
1379                         case -ENOMEM:
1380                                 err = 0;
1381                                 goto out;
1382                 }
1383                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1384                 err = nfs4_handle_exception(server, err, &exception);
1385         } while (exception.retry);
1386 out:
1387         return err;
1388 }
1389
1390 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1391 {
1392         struct nfs4_opendata *data = calldata;
1393
1394         data->rpc_status = task->tk_status;
1395         if (data->rpc_status == 0) {
1396                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1397                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1398                 renew_lease(data->o_res.server, data->timestamp);
1399                 data->rpc_done = 1;
1400         }
1401 }
1402
1403 static void nfs4_open_confirm_release(void *calldata)
1404 {
1405         struct nfs4_opendata *data = calldata;
1406         struct nfs4_state *state = NULL;
1407
1408         /* If this request hasn't been cancelled, do nothing */
1409         if (data->cancelled == 0)
1410                 goto out_free;
1411         /* In case of error, no cleanup! */
1412         if (!data->rpc_done)
1413                 goto out_free;
1414         state = nfs4_opendata_to_nfs4_state(data);
1415         if (!IS_ERR(state))
1416                 nfs4_close_state(state, data->o_arg.fmode);
1417 out_free:
1418         nfs4_opendata_put(data);
1419 }
1420
1421 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1422         .rpc_call_done = nfs4_open_confirm_done,
1423         .rpc_release = nfs4_open_confirm_release,
1424 };
1425
1426 /*
1427  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1428  */
1429 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1430 {
1431         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1432         struct rpc_task *task;
1433         struct  rpc_message msg = {
1434                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1435                 .rpc_argp = &data->c_arg,
1436                 .rpc_resp = &data->c_res,
1437                 .rpc_cred = data->owner->so_cred,
1438         };
1439         struct rpc_task_setup task_setup_data = {
1440                 .rpc_client = server->client,
1441                 .rpc_message = &msg,
1442                 .callback_ops = &nfs4_open_confirm_ops,
1443                 .callback_data = data,
1444                 .workqueue = nfsiod_workqueue,
1445                 .flags = RPC_TASK_ASYNC,
1446         };
1447         int status;
1448
1449         kref_get(&data->kref);
1450         data->rpc_done = 0;
1451         data->rpc_status = 0;
1452         data->timestamp = jiffies;
1453         task = rpc_run_task(&task_setup_data);
1454         if (IS_ERR(task))
1455                 return PTR_ERR(task);
1456         status = nfs4_wait_for_completion_rpc_task(task);
1457         if (status != 0) {
1458                 data->cancelled = 1;
1459                 smp_wmb();
1460         } else
1461                 status = data->rpc_status;
1462         rpc_put_task(task);
1463         return status;
1464 }
1465
1466 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1467 {
1468         struct nfs4_opendata *data = calldata;
1469         struct nfs4_state_owner *sp = data->owner;
1470
1471         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1472                 goto out_wait;
1473         /*
1474          * Check if we still need to send an OPEN call, or if we can use
1475          * a delegation instead.
1476          */
1477         if (data->state != NULL) {
1478                 struct nfs_delegation *delegation;
1479
1480                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1481                         goto out_no_action;
1482                 rcu_read_lock();
1483                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1484                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1485                     can_open_delegated(delegation, data->o_arg.fmode))
1486                         goto unlock_no_action;
1487                 rcu_read_unlock();
1488         }
1489         /* Update client id. */
1490         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1491         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1492                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1493                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1494                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1495         }
1496         data->timestamp = jiffies;
1497         if (nfs4_setup_sequence(data->o_arg.server,
1498                                 &data->o_arg.seq_args,
1499                                 &data->o_res.seq_res,
1500                                 task) != 0)
1501                 nfs_release_seqid(data->o_arg.seqid);
1502         return;
1503 unlock_no_action:
1504         rcu_read_unlock();
1505 out_no_action:
1506         task->tk_action = NULL;
1507 out_wait:
1508         nfs4_sequence_done(task, &data->o_res.seq_res);
1509 }
1510
1511 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1512 {
1513         struct nfs4_opendata *data = calldata;
1514
1515         data->rpc_status = task->tk_status;
1516
1517         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1518                 return;
1519
1520         if (task->tk_status == 0) {
1521                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1522                         switch (data->o_res.f_attr->mode & S_IFMT) {
1523                         case S_IFREG:
1524                                 break;
1525                         case S_IFLNK:
1526                                 data->rpc_status = -ELOOP;
1527                                 break;
1528                         case S_IFDIR:
1529                                 data->rpc_status = -EISDIR;
1530                                 break;
1531                         default:
1532                                 data->rpc_status = -ENOTDIR;
1533                         }
1534                 }
1535                 renew_lease(data->o_res.server, data->timestamp);
1536                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1537                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1538         }
1539         data->rpc_done = 1;
1540 }
1541
1542 static void nfs4_open_release(void *calldata)
1543 {
1544         struct nfs4_opendata *data = calldata;
1545         struct nfs4_state *state = NULL;
1546
1547         /* If this request hasn't been cancelled, do nothing */
1548         if (data->cancelled == 0)
1549                 goto out_free;
1550         /* In case of error, no cleanup! */
1551         if (data->rpc_status != 0 || !data->rpc_done)
1552                 goto out_free;
1553         /* In case we need an open_confirm, no cleanup! */
1554         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1555                 goto out_free;
1556         state = nfs4_opendata_to_nfs4_state(data);
1557         if (!IS_ERR(state))
1558                 nfs4_close_state(state, data->o_arg.fmode);
1559 out_free:
1560         nfs4_opendata_put(data);
1561 }
1562
1563 static const struct rpc_call_ops nfs4_open_ops = {
1564         .rpc_call_prepare = nfs4_open_prepare,
1565         .rpc_call_done = nfs4_open_done,
1566         .rpc_release = nfs4_open_release,
1567 };
1568
1569 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1570 {
1571         struct inode *dir = data->dir->d_inode;
1572         struct nfs_server *server = NFS_SERVER(dir);
1573         struct nfs_openargs *o_arg = &data->o_arg;
1574         struct nfs_openres *o_res = &data->o_res;
1575         struct rpc_task *task;
1576         struct rpc_message msg = {
1577                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1578                 .rpc_argp = o_arg,
1579                 .rpc_resp = o_res,
1580                 .rpc_cred = data->owner->so_cred,
1581         };
1582         struct rpc_task_setup task_setup_data = {
1583                 .rpc_client = server->client,
1584                 .rpc_message = &msg,
1585                 .callback_ops = &nfs4_open_ops,
1586                 .callback_data = data,
1587                 .workqueue = nfsiod_workqueue,
1588                 .flags = RPC_TASK_ASYNC,
1589         };
1590         int status;
1591
1592         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1593         kref_get(&data->kref);
1594         data->rpc_done = 0;
1595         data->rpc_status = 0;
1596         data->cancelled = 0;
1597         if (isrecover)
1598                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1599         task = rpc_run_task(&task_setup_data);
1600         if (IS_ERR(task))
1601                 return PTR_ERR(task);
1602         status = nfs4_wait_for_completion_rpc_task(task);
1603         if (status != 0) {
1604                 data->cancelled = 1;
1605                 smp_wmb();
1606         } else
1607                 status = data->rpc_status;
1608         rpc_put_task(task);
1609
1610         return status;
1611 }
1612
1613 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1614 {
1615         struct inode *dir = data->dir->d_inode;
1616         struct nfs_openres *o_res = &data->o_res;
1617         int status;
1618
1619         status = nfs4_run_open_task(data, 1);
1620         if (status != 0 || !data->rpc_done)
1621                 return status;
1622
1623         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1624
1625         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1626                 status = _nfs4_proc_open_confirm(data);
1627                 if (status != 0)
1628                         return status;
1629         }
1630
1631         return status;
1632 }
1633
1634 static int nfs4_opendata_access(struct rpc_cred *cred,
1635                                 struct nfs4_opendata *opendata,
1636                                 struct nfs4_state *state, fmode_t fmode,
1637                                 int openflags)
1638 {
1639         struct nfs_access_entry cache;
1640         u32 mask;
1641
1642         /* access call failed or for some reason the server doesn't
1643          * support any access modes -- defer access call until later */
1644         if (opendata->o_res.access_supported == 0)
1645                 return 0;
1646
1647         mask = 0;
1648         /* don't check MAY_WRITE - a newly created file may not have
1649          * write mode bits, but POSIX allows the creating process to write.
1650          * use openflags to check for exec, because fmode won't
1651          * always have FMODE_EXEC set when file open for exec. */
1652         if (openflags & __FMODE_EXEC) {
1653                 /* ONLY check for exec rights */
1654                 mask = MAY_EXEC;
1655         } else if (fmode & FMODE_READ)
1656                 mask = MAY_READ;
1657
1658         cache.cred = cred;
1659         cache.jiffies = jiffies;
1660         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1661         nfs_access_add_cache(state->inode, &cache);
1662
1663         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1664                 return 0;
1665
1666         /* even though OPEN succeeded, access is denied. Close the file */
1667         nfs4_close_state(state, fmode);
1668         return -EACCES;
1669 }
1670
1671 /*
1672  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1673  */
1674 static int _nfs4_proc_open(struct nfs4_opendata *data)
1675 {
1676         struct inode *dir = data->dir->d_inode;
1677         struct nfs_server *server = NFS_SERVER(dir);
1678         struct nfs_openargs *o_arg = &data->o_arg;
1679         struct nfs_openres *o_res = &data->o_res;
1680         int status;
1681
1682         status = nfs4_run_open_task(data, 0);
1683         if (!data->rpc_done)
1684                 return status;
1685         if (status != 0) {
1686                 if (status == -NFS4ERR_BADNAME &&
1687                                 !(o_arg->open_flags & O_CREAT))
1688                         return -ENOENT;
1689                 return status;
1690         }
1691
1692         nfs_fattr_map_and_free_names(server, &data->f_attr);
1693
1694         if (o_arg->open_flags & O_CREAT)
1695                 update_changeattr(dir, &o_res->cinfo);
1696         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1697                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1698         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1699                 status = _nfs4_proc_open_confirm(data);
1700                 if (status != 0)
1701                         return status;
1702         }
1703         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1704                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1705         return 0;
1706 }
1707
1708 static int nfs4_recover_expired_lease(struct nfs_server *server)
1709 {
1710         return nfs4_client_recover_expired_lease(server->nfs_client);
1711 }
1712
1713 /*
1714  * OPEN_EXPIRED:
1715  *      reclaim state on the server after a network partition.
1716  *      Assumes caller holds the appropriate lock
1717  */
1718 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1719 {
1720         struct nfs4_opendata *opendata;
1721         int ret;
1722
1723         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1724         if (IS_ERR(opendata))
1725                 return PTR_ERR(opendata);
1726         ret = nfs4_open_recover(opendata, state);
1727         if (ret == -ESTALE)
1728                 d_drop(ctx->dentry);
1729         nfs4_opendata_put(opendata);
1730         return ret;
1731 }
1732
1733 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1734 {
1735         struct nfs_server *server = NFS_SERVER(state->inode);
1736         struct nfs4_exception exception = { };
1737         int err;
1738
1739         do {
1740                 err = _nfs4_open_expired(ctx, state);
1741                 switch (err) {
1742                 default:
1743                         goto out;
1744                 case -NFS4ERR_GRACE:
1745                 case -NFS4ERR_DELAY:
1746                         nfs4_handle_exception(server, err, &exception);
1747                         err = 0;
1748                 }
1749         } while (exception.retry);
1750 out:
1751         return err;
1752 }
1753
1754 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1755 {
1756         struct nfs_open_context *ctx;
1757         int ret;
1758
1759         ctx = nfs4_state_find_open_context(state);
1760         if (IS_ERR(ctx))
1761                 return PTR_ERR(ctx);
1762         ret = nfs4_do_open_expired(ctx, state);
1763         put_nfs_open_context(ctx);
1764         return ret;
1765 }
1766
1767 #if defined(CONFIG_NFS_V4_1)
1768 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1769 {
1770         struct nfs_server *server = NFS_SERVER(state->inode);
1771         nfs4_stateid *stateid = &state->stateid;
1772         int status;
1773
1774         /* If a state reset has been done, test_stateid is unneeded */
1775         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1776                 return;
1777
1778         status = nfs41_test_stateid(server, stateid);
1779         if (status != NFS_OK) {
1780                 /* Free the stateid unless the server explicitly
1781                  * informs us the stateid is unrecognized. */
1782                 if (status != -NFS4ERR_BAD_STATEID)
1783                         nfs41_free_stateid(server, stateid);
1784                 nfs_remove_bad_delegation(state->inode);
1785
1786                 write_seqlock(&state->seqlock);
1787                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1788                 write_sequnlock(&state->seqlock);
1789                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1790         }
1791 }
1792
1793 /**
1794  * nfs41_check_open_stateid - possibly free an open stateid
1795  *
1796  * @state: NFSv4 state for an inode
1797  *
1798  * Returns NFS_OK if recovery for this stateid is now finished.
1799  * Otherwise a negative NFS4ERR value is returned.
1800  */
1801 static int nfs41_check_open_stateid(struct nfs4_state *state)
1802 {
1803         struct nfs_server *server = NFS_SERVER(state->inode);
1804         nfs4_stateid *stateid = &state->open_stateid;
1805         int status;
1806
1807         /* If a state reset has been done, test_stateid is unneeded */
1808         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1809             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1810             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1811                 return -NFS4ERR_BAD_STATEID;
1812
1813         status = nfs41_test_stateid(server, stateid);
1814         if (status != NFS_OK) {
1815                 /* Free the stateid unless the server explicitly
1816                  * informs us the stateid is unrecognized. */
1817                 if (status != -NFS4ERR_BAD_STATEID)
1818                         nfs41_free_stateid(server, stateid);
1819
1820                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1821                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1822                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1823         }
1824         return status;
1825 }
1826
1827 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1828 {
1829         int status;
1830
1831         nfs41_clear_delegation_stateid(state);
1832         status = nfs41_check_open_stateid(state);
1833         if (status != NFS_OK)
1834                 status = nfs4_open_expired(sp, state);
1835         return status;
1836 }
1837 #endif
1838
1839 /*
1840  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1841  * fields corresponding to attributes that were used to store the verifier.
1842  * Make sure we clobber those fields in the later setattr call
1843  */
1844 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1845 {
1846         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1847             !(sattr->ia_valid & ATTR_ATIME_SET))
1848                 sattr->ia_valid |= ATTR_ATIME;
1849
1850         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1851             !(sattr->ia_valid & ATTR_MTIME_SET))
1852                 sattr->ia_valid |= ATTR_MTIME;
1853 }
1854
1855 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
1856                 fmode_t fmode,
1857                 int flags,
1858                 struct nfs4_state **res)
1859 {
1860         struct nfs4_state_owner *sp = opendata->owner;
1861         struct nfs_server *server = sp->so_server;
1862         struct nfs4_state *state;
1863         unsigned int seq;
1864         int ret;
1865
1866         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
1867
1868         ret = _nfs4_proc_open(opendata);
1869         if (ret != 0)
1870                 goto out;
1871
1872         state = nfs4_opendata_to_nfs4_state(opendata);
1873         ret = PTR_ERR(state);
1874         if (IS_ERR(state))
1875                 goto out;
1876         if (server->caps & NFS_CAP_POSIX_LOCK)
1877                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1878
1879         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
1880         if (ret != 0)
1881                 goto out;
1882
1883         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
1884                 nfs4_schedule_stateid_recovery(server, state);
1885                 nfs4_wait_clnt_recover(server->nfs_client);
1886         }
1887         *res = state;
1888 out:
1889         return ret;
1890 }
1891
1892 /*
1893  * Returns a referenced nfs4_state
1894  */
1895 static int _nfs4_do_open(struct inode *dir,
1896                         struct dentry *dentry,
1897                         fmode_t fmode,
1898                         int flags,
1899                         struct iattr *sattr,
1900                         struct rpc_cred *cred,
1901                         struct nfs4_state **res,
1902                         struct nfs4_threshold **ctx_th)
1903 {
1904         struct nfs4_state_owner  *sp;
1905         struct nfs4_state     *state = NULL;
1906         struct nfs_server       *server = NFS_SERVER(dir);
1907         struct nfs4_opendata *opendata;
1908         int status;
1909
1910         /* Protect against reboot recovery conflicts */
1911         status = -ENOMEM;
1912         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1913         if (sp == NULL) {
1914                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1915                 goto out_err;
1916         }
1917         status = nfs4_recover_expired_lease(server);
1918         if (status != 0)
1919                 goto err_put_state_owner;
1920         if (dentry->d_inode != NULL)
1921                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1922         status = -ENOMEM;
1923         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1924         if (opendata == NULL)
1925                 goto err_put_state_owner;
1926
1927         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1928                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1929                 if (!opendata->f_attr.mdsthreshold)
1930                         goto err_opendata_put;
1931                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
1932         }
1933         if (dentry->d_inode != NULL)
1934                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1935
1936         status = _nfs4_open_and_get_state(opendata, fmode, flags, &state);
1937         if (status != 0)
1938                 goto err_opendata_put;
1939
1940         if (opendata->o_arg.open_flags & O_EXCL) {
1941                 nfs4_exclusive_attrset(opendata, sattr);
1942
1943                 nfs_fattr_init(opendata->o_res.f_attr);
1944                 status = nfs4_do_setattr(state->inode, cred,
1945                                 opendata->o_res.f_attr, sattr,
1946                                 state);
1947                 if (status == 0)
1948                         nfs_setattr_update_inode(state->inode, sattr);
1949                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1950         }
1951
1952         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
1953                 *ctx_th = opendata->f_attr.mdsthreshold;
1954         else
1955                 kfree(opendata->f_attr.mdsthreshold);
1956         opendata->f_attr.mdsthreshold = NULL;
1957
1958         nfs4_opendata_put(opendata);
1959         nfs4_put_state_owner(sp);
1960         *res = state;
1961         return 0;
1962 err_opendata_put:
1963         kfree(opendata->f_attr.mdsthreshold);
1964         nfs4_opendata_put(opendata);
1965 err_put_state_owner:
1966         nfs4_put_state_owner(sp);
1967 out_err:
1968         *res = NULL;
1969         return status;
1970 }
1971
1972
1973 static struct nfs4_state *nfs4_do_open(struct inode *dir,
1974                                         struct dentry *dentry,
1975                                         fmode_t fmode,
1976                                         int flags,
1977                                         struct iattr *sattr,
1978                                         struct rpc_cred *cred,
1979                                         struct nfs4_threshold **ctx_th)
1980 {
1981         struct nfs4_exception exception = { };
1982         struct nfs4_state *res;
1983         int status;
1984
1985         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
1986         do {
1987                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
1988                                        &res, ctx_th);
1989                 if (status == 0)
1990                         break;
1991                 /* NOTE: BAD_SEQID means the server and client disagree about the
1992                  * book-keeping w.r.t. state-changing operations
1993                  * (OPEN/CLOSE/LOCK/LOCKU...)
1994                  * It is actually a sign of a bug on the client or on the server.
1995                  *
1996                  * If we receive a BAD_SEQID error in the particular case of
1997                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1998                  * have unhashed the old state_owner for us, and that we can
1999                  * therefore safely retry using a new one. We should still warn
2000                  * the user though...
2001                  */
2002                 if (status == -NFS4ERR_BAD_SEQID) {
2003                         pr_warn_ratelimited("NFS: v4 server %s "
2004                                         " returned a bad sequence-id error!\n",
2005                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2006                         exception.retry = 1;
2007                         continue;
2008                 }
2009                 /*
2010                  * BAD_STATEID on OPEN means that the server cancelled our
2011                  * state before it received the OPEN_CONFIRM.
2012                  * Recover by retrying the request as per the discussion
2013                  * on Page 181 of RFC3530.
2014                  */
2015                 if (status == -NFS4ERR_BAD_STATEID) {
2016                         exception.retry = 1;
2017                         continue;
2018                 }
2019                 if (status == -EAGAIN) {
2020                         /* We must have found a delegation */
2021                         exception.retry = 1;
2022                         continue;
2023                 }
2024                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
2025                                         status, &exception));
2026         } while (exception.retry);
2027         return res;
2028 }
2029
2030 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2031                             struct nfs_fattr *fattr, struct iattr *sattr,
2032                             struct nfs4_state *state)
2033 {
2034         struct nfs_server *server = NFS_SERVER(inode);
2035         struct nfs_setattrargs  arg = {
2036                 .fh             = NFS_FH(inode),
2037                 .iap            = sattr,
2038                 .server         = server,
2039                 .bitmask = server->attr_bitmask,
2040         };
2041         struct nfs_setattrres  res = {
2042                 .fattr          = fattr,
2043                 .server         = server,
2044         };
2045         struct rpc_message msg = {
2046                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2047                 .rpc_argp       = &arg,
2048                 .rpc_resp       = &res,
2049                 .rpc_cred       = cred,
2050         };
2051         unsigned long timestamp = jiffies;
2052         int status;
2053
2054         nfs_fattr_init(fattr);
2055
2056         if (state != NULL) {
2057                 struct nfs_lockowner lockowner = {
2058                         .l_owner = current->files,
2059                         .l_pid = current->tgid,
2060                 };
2061                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2062                                 &lockowner);
2063         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
2064                                 FMODE_WRITE)) {
2065                 /* Use that stateid */
2066         } else
2067                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2068
2069         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2070         if (status == 0 && state != NULL)
2071                 renew_lease(server, timestamp);
2072         return status;
2073 }
2074
2075 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2076                            struct nfs_fattr *fattr, struct iattr *sattr,
2077                            struct nfs4_state *state)
2078 {
2079         struct nfs_server *server = NFS_SERVER(inode);
2080         struct nfs4_exception exception = {
2081                 .state = state,
2082                 .inode = inode,
2083         };
2084         int err;
2085         do {
2086                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2087                 switch (err) {
2088                 case -NFS4ERR_OPENMODE:
2089                         if (state && !(state->state & FMODE_WRITE)) {
2090                                 err = -EBADF;
2091                                 if (sattr->ia_valid & ATTR_OPEN)
2092                                         err = -EACCES;
2093                                 goto out;
2094                         }
2095                 }
2096                 err = nfs4_handle_exception(server, err, &exception);
2097         } while (exception.retry);
2098 out:
2099         return err;
2100 }
2101
2102 struct nfs4_closedata {
2103         struct inode *inode;
2104         struct nfs4_state *state;
2105         struct nfs_closeargs arg;
2106         struct nfs_closeres res;
2107         struct nfs_fattr fattr;
2108         unsigned long timestamp;
2109         bool roc;
2110         u32 roc_barrier;
2111 };
2112
2113 static void nfs4_free_closedata(void *data)
2114 {
2115         struct nfs4_closedata *calldata = data;
2116         struct nfs4_state_owner *sp = calldata->state->owner;
2117         struct super_block *sb = calldata->state->inode->i_sb;
2118
2119         if (calldata->roc)
2120                 pnfs_roc_release(calldata->state->inode);
2121         nfs4_put_open_state(calldata->state);
2122         nfs_free_seqid(calldata->arg.seqid);
2123         nfs4_put_state_owner(sp);
2124         nfs_sb_deactive(sb);
2125         kfree(calldata);
2126 }
2127
2128 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2129                 fmode_t fmode)
2130 {
2131         spin_lock(&state->owner->so_lock);
2132         if (!(fmode & FMODE_READ))
2133                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2134         if (!(fmode & FMODE_WRITE))
2135                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2136         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2137         spin_unlock(&state->owner->so_lock);
2138 }
2139
2140 static void nfs4_close_done(struct rpc_task *task, void *data)
2141 {
2142         struct nfs4_closedata *calldata = data;
2143         struct nfs4_state *state = calldata->state;
2144         struct nfs_server *server = NFS_SERVER(calldata->inode);
2145
2146         dprintk("%s: begin!\n", __func__);
2147         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2148                 return;
2149         /* hmm. we are done with the inode, and in the process of freeing
2150          * the state_owner. we keep this around to process errors
2151          */
2152         switch (task->tk_status) {
2153                 case 0:
2154                         if (calldata->roc)
2155                                 pnfs_roc_set_barrier(state->inode,
2156                                                      calldata->roc_barrier);
2157                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2158                         renew_lease(server, calldata->timestamp);
2159                         nfs4_close_clear_stateid_flags(state,
2160                                         calldata->arg.fmode);
2161                         break;
2162                 case -NFS4ERR_STALE_STATEID:
2163                 case -NFS4ERR_OLD_STATEID:
2164                 case -NFS4ERR_BAD_STATEID:
2165                 case -NFS4ERR_EXPIRED:
2166                         if (calldata->arg.fmode == 0)
2167                                 break;
2168                 default:
2169                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2170                                 rpc_restart_call_prepare(task);
2171         }
2172         nfs_release_seqid(calldata->arg.seqid);
2173         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2174         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2175 }
2176
2177 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2178 {
2179         struct nfs4_closedata *calldata = data;
2180         struct nfs4_state *state = calldata->state;
2181         struct inode *inode = calldata->inode;
2182         int call_close = 0;
2183
2184         dprintk("%s: begin!\n", __func__);
2185         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2186                 goto out_wait;
2187
2188         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2189         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2190         spin_lock(&state->owner->so_lock);
2191         /* Calculate the change in open mode */
2192         if (state->n_rdwr == 0) {
2193                 if (state->n_rdonly == 0) {
2194                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2195                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2196                         calldata->arg.fmode &= ~FMODE_READ;
2197                 }
2198                 if (state->n_wronly == 0) {
2199                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2200                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2201                         calldata->arg.fmode &= ~FMODE_WRITE;
2202                 }
2203         }
2204         spin_unlock(&state->owner->so_lock);
2205
2206         if (!call_close) {
2207                 /* Note: exit _without_ calling nfs4_close_done */
2208                 goto out_no_action;
2209         }
2210
2211         if (calldata->arg.fmode == 0) {
2212                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2213                 if (calldata->roc &&
2214                     pnfs_roc_drain(inode, &calldata->roc_barrier, task))
2215                         goto out_wait;
2216         }
2217
2218         nfs_fattr_init(calldata->res.fattr);
2219         calldata->timestamp = jiffies;
2220         if (nfs4_setup_sequence(NFS_SERVER(inode),
2221                                 &calldata->arg.seq_args,
2222                                 &calldata->res.seq_res,
2223                                 task) != 0)
2224                 nfs_release_seqid(calldata->arg.seqid);
2225         dprintk("%s: done!\n", __func__);
2226         return;
2227 out_no_action:
2228         task->tk_action = NULL;
2229 out_wait:
2230         nfs4_sequence_done(task, &calldata->res.seq_res);
2231 }
2232
2233 static const struct rpc_call_ops nfs4_close_ops = {
2234         .rpc_call_prepare = nfs4_close_prepare,
2235         .rpc_call_done = nfs4_close_done,
2236         .rpc_release = nfs4_free_closedata,
2237 };
2238
2239 /* 
2240  * It is possible for data to be read/written from a mem-mapped file 
2241  * after the sys_close call (which hits the vfs layer as a flush).
2242  * This means that we can't safely call nfsv4 close on a file until 
2243  * the inode is cleared. This in turn means that we are not good
2244  * NFSv4 citizens - we do not indicate to the server to update the file's 
2245  * share state even when we are done with one of the three share 
2246  * stateid's in the inode.
2247  *
2248  * NOTE: Caller must be holding the sp->so_owner semaphore!
2249  */
2250 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2251 {
2252         struct nfs_server *server = NFS_SERVER(state->inode);
2253         struct nfs4_closedata *calldata;
2254         struct nfs4_state_owner *sp = state->owner;
2255         struct rpc_task *task;
2256         struct rpc_message msg = {
2257                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2258                 .rpc_cred = state->owner->so_cred,
2259         };
2260         struct rpc_task_setup task_setup_data = {
2261                 .rpc_client = server->client,
2262                 .rpc_message = &msg,
2263                 .callback_ops = &nfs4_close_ops,
2264                 .workqueue = nfsiod_workqueue,
2265                 .flags = RPC_TASK_ASYNC,
2266         };
2267         int status = -ENOMEM;
2268
2269         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2270         if (calldata == NULL)
2271                 goto out;
2272         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2273         calldata->inode = state->inode;
2274         calldata->state = state;
2275         calldata->arg.fh = NFS_FH(state->inode);
2276         calldata->arg.stateid = &state->open_stateid;
2277         /* Serialization for the sequence id */
2278         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2279         if (calldata->arg.seqid == NULL)
2280                 goto out_free_calldata;
2281         calldata->arg.fmode = 0;
2282         calldata->arg.bitmask = server->cache_consistency_bitmask;
2283         calldata->res.fattr = &calldata->fattr;
2284         calldata->res.seqid = calldata->arg.seqid;
2285         calldata->res.server = server;
2286         calldata->roc = pnfs_roc(state->inode);
2287         nfs_sb_active(calldata->inode->i_sb);
2288
2289         msg.rpc_argp = &calldata->arg;
2290         msg.rpc_resp = &calldata->res;
2291         task_setup_data.callback_data = calldata;
2292         task = rpc_run_task(&task_setup_data);
2293         if (IS_ERR(task))
2294                 return PTR_ERR(task);
2295         status = 0;
2296         if (wait)
2297                 status = rpc_wait_for_completion_task(task);
2298         rpc_put_task(task);
2299         return status;
2300 out_free_calldata:
2301         kfree(calldata);
2302 out:
2303         nfs4_put_open_state(state);
2304         nfs4_put_state_owner(sp);
2305         return status;
2306 }
2307
2308 static struct inode *
2309 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2310 {
2311         struct nfs4_state *state;
2312
2313         /* Protect against concurrent sillydeletes */
2314         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2315                              ctx->cred, &ctx->mdsthreshold);
2316         if (IS_ERR(state))
2317                 return ERR_CAST(state);
2318         ctx->state = state;
2319         return igrab(state->inode);
2320 }
2321
2322 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2323 {
2324         if (ctx->state == NULL)
2325                 return;
2326         if (is_sync)
2327                 nfs4_close_sync(ctx->state, ctx->mode);
2328         else
2329                 nfs4_close_state(ctx->state, ctx->mode);
2330 }
2331
2332 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2333 {
2334         struct nfs4_server_caps_arg args = {
2335                 .fhandle = fhandle,
2336         };
2337         struct nfs4_server_caps_res res = {};
2338         struct rpc_message msg = {
2339                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2340                 .rpc_argp = &args,
2341                 .rpc_resp = &res,
2342         };
2343         int status;
2344
2345         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2346         if (status == 0) {
2347                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2348                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2349                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2350                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2351                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2352                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2353                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2354                         server->caps |= NFS_CAP_ACLS;
2355                 if (res.has_links != 0)
2356                         server->caps |= NFS_CAP_HARDLINKS;
2357                 if (res.has_symlinks != 0)
2358                         server->caps |= NFS_CAP_SYMLINKS;
2359                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2360                         server->caps |= NFS_CAP_FILEID;
2361                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2362                         server->caps |= NFS_CAP_MODE;
2363                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2364                         server->caps |= NFS_CAP_NLINK;
2365                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2366                         server->caps |= NFS_CAP_OWNER;
2367                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2368                         server->caps |= NFS_CAP_OWNER_GROUP;
2369                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2370                         server->caps |= NFS_CAP_ATIME;
2371                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2372                         server->caps |= NFS_CAP_CTIME;
2373                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2374                         server->caps |= NFS_CAP_MTIME;
2375
2376                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2377                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2378                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2379                 server->acl_bitmask = res.acl_bitmask;
2380                 server->fh_expire_type = res.fh_expire_type;
2381         }
2382
2383         return status;
2384 }
2385
2386 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2387 {
2388         struct nfs4_exception exception = { };
2389         int err;
2390         do {
2391                 err = nfs4_handle_exception(server,
2392                                 _nfs4_server_capabilities(server, fhandle),
2393                                 &exception);
2394         } while (exception.retry);
2395         return err;
2396 }
2397
2398 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2399                 struct nfs_fsinfo *info)
2400 {
2401         struct nfs4_lookup_root_arg args = {
2402                 .bitmask = nfs4_fattr_bitmap,
2403         };
2404         struct nfs4_lookup_res res = {
2405                 .server = server,
2406                 .fattr = info->fattr,
2407                 .fh = fhandle,
2408         };
2409         struct rpc_message msg = {
2410                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2411                 .rpc_argp = &args,
2412                 .rpc_resp = &res,
2413         };
2414
2415         nfs_fattr_init(info->fattr);
2416         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2417 }
2418
2419 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2420                 struct nfs_fsinfo *info)
2421 {
2422         struct nfs4_exception exception = { };
2423         int err;
2424         do {
2425                 err = _nfs4_lookup_root(server, fhandle, info);
2426                 switch (err) {
2427                 case 0:
2428                 case -NFS4ERR_WRONGSEC:
2429                         goto out;
2430                 default:
2431                         err = nfs4_handle_exception(server, err, &exception);
2432                 }
2433         } while (exception.retry);
2434 out:
2435         return err;
2436 }
2437
2438 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2439                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2440 {
2441         struct rpc_auth *auth;
2442         int ret;
2443
2444         auth = rpcauth_create(flavor, server->client);
2445         if (IS_ERR(auth)) {
2446                 ret = -EIO;
2447                 goto out;
2448         }
2449         ret = nfs4_lookup_root(server, fhandle, info);
2450 out:
2451         return ret;
2452 }
2453
2454 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2455                               struct nfs_fsinfo *info)
2456 {
2457         int i, len, status = 0;
2458         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2459
2460         len = rpcauth_list_flavors(flav_array, ARRAY_SIZE(flav_array));
2461         if (len < 0)
2462                 return len;
2463
2464         for (i = 0; i < len; i++) {
2465                 /* AUTH_UNIX is the default flavor if none was specified,
2466                  * thus has already been tried. */
2467                 if (flav_array[i] == RPC_AUTH_UNIX)
2468                         continue;
2469
2470                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2471                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2472                         continue;
2473                 break;
2474         }
2475         /*
2476          * -EACCESS could mean that the user doesn't have correct permissions
2477          * to access the mount.  It could also mean that we tried to mount
2478          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2479          * existing mount programs don't handle -EACCES very well so it should
2480          * be mapped to -EPERM instead.
2481          */
2482         if (status == -EACCES)
2483                 status = -EPERM;
2484         return status;
2485 }
2486
2487 /*
2488  * get the file handle for the "/" directory on the server
2489  */
2490 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2491                          struct nfs_fsinfo *info)
2492 {
2493         int minor_version = server->nfs_client->cl_minorversion;
2494         int status = nfs4_lookup_root(server, fhandle, info);
2495         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2496                 /*
2497                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2498                  * by nfs4_map_errors() as this function exits.
2499                  */
2500                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2501         if (status == 0)
2502                 status = nfs4_server_capabilities(server, fhandle);
2503         if (status == 0)
2504                 status = nfs4_do_fsinfo(server, fhandle, info);
2505         return nfs4_map_errors(status);
2506 }
2507
2508 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2509                               struct nfs_fsinfo *info)
2510 {
2511         int error;
2512         struct nfs_fattr *fattr = info->fattr;
2513
2514         error = nfs4_server_capabilities(server, mntfh);
2515         if (error < 0) {
2516                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2517                 return error;
2518         }
2519
2520         error = nfs4_proc_getattr(server, mntfh, fattr);
2521         if (error < 0) {
2522                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2523                 return error;
2524         }
2525
2526         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2527             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2528                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2529
2530         return error;
2531 }
2532
2533 /*
2534  * Get locations and (maybe) other attributes of a referral.
2535  * Note that we'll actually follow the referral later when
2536  * we detect fsid mismatch in inode revalidation
2537  */
2538 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2539                              const struct qstr *name, struct nfs_fattr *fattr,
2540                              struct nfs_fh *fhandle)
2541 {
2542         int status = -ENOMEM;
2543         struct page *page = NULL;
2544         struct nfs4_fs_locations *locations = NULL;
2545
2546         page = alloc_page(GFP_KERNEL);
2547         if (page == NULL)
2548                 goto out;
2549         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2550         if (locations == NULL)
2551                 goto out;
2552
2553         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2554         if (status != 0)
2555                 goto out;
2556         /* Make sure server returned a different fsid for the referral */
2557         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2558                 dprintk("%s: server did not return a different fsid for"
2559                         " a referral at %s\n", __func__, name->name);
2560                 status = -EIO;
2561                 goto out;
2562         }
2563         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2564         nfs_fixup_referral_attributes(&locations->fattr);
2565
2566         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2567         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2568         memset(fhandle, 0, sizeof(struct nfs_fh));
2569 out:
2570         if (page)
2571                 __free_page(page);
2572         kfree(locations);
2573         return status;
2574 }
2575
2576 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2577 {
2578         struct nfs4_getattr_arg args = {
2579                 .fh = fhandle,
2580                 .bitmask = server->attr_bitmask,
2581         };
2582         struct nfs4_getattr_res res = {
2583                 .fattr = fattr,
2584                 .server = server,
2585         };
2586         struct rpc_message msg = {
2587                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2588                 .rpc_argp = &args,
2589                 .rpc_resp = &res,
2590         };
2591         
2592         nfs_fattr_init(fattr);
2593         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2594 }
2595
2596 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2597 {
2598         struct nfs4_exception exception = { };
2599         int err;
2600         do {
2601                 err = nfs4_handle_exception(server,
2602                                 _nfs4_proc_getattr(server, fhandle, fattr),
2603                                 &exception);
2604         } while (exception.retry);
2605         return err;
2606 }
2607
2608 /* 
2609  * The file is not closed if it is opened due to the a request to change
2610  * the size of the file. The open call will not be needed once the
2611  * VFS layer lookup-intents are implemented.
2612  *
2613  * Close is called when the inode is destroyed.
2614  * If we haven't opened the file for O_WRONLY, we
2615  * need to in the size_change case to obtain a stateid.
2616  *
2617  * Got race?
2618  * Because OPEN is always done by name in nfsv4, it is
2619  * possible that we opened a different file by the same
2620  * name.  We can recognize this race condition, but we
2621  * can't do anything about it besides returning an error.
2622  *
2623  * This will be fixed with VFS changes (lookup-intent).
2624  */
2625 static int
2626 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2627                   struct iattr *sattr)
2628 {
2629         struct inode *inode = dentry->d_inode;
2630         struct rpc_cred *cred = NULL;
2631         struct nfs4_state *state = NULL;
2632         int status;
2633
2634         if (pnfs_ld_layoutret_on_setattr(inode))
2635                 pnfs_return_layout(inode);
2636
2637         nfs_fattr_init(fattr);
2638         
2639         /* Deal with open(O_TRUNC) */
2640         if (sattr->ia_valid & ATTR_OPEN)
2641                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2642
2643         /* Optimization: if the end result is no change, don't RPC */
2644         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2645                 return 0;
2646
2647         /* Search for an existing open(O_WRITE) file */
2648         if (sattr->ia_valid & ATTR_FILE) {
2649                 struct nfs_open_context *ctx;
2650
2651                 ctx = nfs_file_open_context(sattr->ia_file);
2652                 if (ctx) {
2653                         cred = ctx->cred;
2654                         state = ctx->state;
2655                 }
2656         }
2657
2658         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2659         if (status == 0)
2660                 nfs_setattr_update_inode(inode, sattr);
2661         return status;
2662 }
2663
2664 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2665                 const struct qstr *name, struct nfs_fh *fhandle,
2666                 struct nfs_fattr *fattr)
2667 {
2668         struct nfs_server *server = NFS_SERVER(dir);
2669         int                    status;
2670         struct nfs4_lookup_arg args = {
2671                 .bitmask = server->attr_bitmask,
2672                 .dir_fh = NFS_FH(dir),
2673                 .name = name,
2674         };
2675         struct nfs4_lookup_res res = {
2676                 .server = server,
2677                 .fattr = fattr,
2678                 .fh = fhandle,
2679         };
2680         struct rpc_message msg = {
2681                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2682                 .rpc_argp = &args,
2683                 .rpc_resp = &res,
2684         };
2685
2686         nfs_fattr_init(fattr);
2687
2688         dprintk("NFS call  lookup %s\n", name->name);
2689         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2690         dprintk("NFS reply lookup: %d\n", status);
2691         return status;
2692 }
2693
2694 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2695 {
2696         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2697                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2698         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2699         fattr->nlink = 2;
2700 }
2701
2702 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2703                                    struct qstr *name, struct nfs_fh *fhandle,
2704                                    struct nfs_fattr *fattr)
2705 {
2706         struct nfs4_exception exception = { };
2707         struct rpc_clnt *client = *clnt;
2708         int err;
2709         do {
2710                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2711                 switch (err) {
2712                 case -NFS4ERR_BADNAME:
2713                         err = -ENOENT;
2714                         goto out;
2715                 case -NFS4ERR_MOVED:
2716                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2717                         goto out;
2718                 case -NFS4ERR_WRONGSEC:
2719                         err = -EPERM;
2720                         if (client != *clnt)
2721                                 goto out;
2722
2723                         client = nfs4_create_sec_client(client, dir, name);
2724                         if (IS_ERR(client))
2725                                 return PTR_ERR(client);
2726
2727                         exception.retry = 1;
2728                         break;
2729                 default:
2730                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2731                 }
2732         } while (exception.retry);
2733
2734 out:
2735         if (err == 0)
2736                 *clnt = client;
2737         else if (client != *clnt)
2738                 rpc_shutdown_client(client);
2739
2740         return err;
2741 }
2742
2743 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2744                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2745 {
2746         int status;
2747         struct rpc_clnt *client = NFS_CLIENT(dir);
2748
2749         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2750         if (client != NFS_CLIENT(dir)) {
2751                 rpc_shutdown_client(client);
2752                 nfs_fixup_secinfo_attributes(fattr);
2753         }
2754         return status;
2755 }
2756
2757 struct rpc_clnt *
2758 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2759                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2760 {
2761         int status;
2762         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2763
2764         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2765         if (status < 0) {
2766                 rpc_shutdown_client(client);
2767                 return ERR_PTR(status);
2768         }
2769         return client;
2770 }
2771
2772 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2773 {
2774         struct nfs_server *server = NFS_SERVER(inode);
2775         struct nfs4_accessargs args = {
2776                 .fh = NFS_FH(inode),
2777                 .bitmask = server->cache_consistency_bitmask,
2778         };
2779         struct nfs4_accessres res = {
2780                 .server = server,
2781         };
2782         struct rpc_message msg = {
2783                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2784                 .rpc_argp = &args,
2785                 .rpc_resp = &res,
2786                 .rpc_cred = entry->cred,
2787         };
2788         int mode = entry->mask;
2789         int status;
2790
2791         /*
2792          * Determine which access bits we want to ask for...
2793          */
2794         if (mode & MAY_READ)
2795                 args.access |= NFS4_ACCESS_READ;
2796         if (S_ISDIR(inode->i_mode)) {
2797                 if (mode & MAY_WRITE)
2798                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2799                 if (mode & MAY_EXEC)
2800                         args.access |= NFS4_ACCESS_LOOKUP;
2801         } else {
2802                 if (mode & MAY_WRITE)
2803                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2804                 if (mode & MAY_EXEC)
2805                         args.access |= NFS4_ACCESS_EXECUTE;
2806         }
2807
2808         res.fattr = nfs_alloc_fattr();
2809         if (res.fattr == NULL)
2810                 return -ENOMEM;
2811
2812         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2813         if (!status) {
2814                 nfs_access_set_mask(entry, res.access);
2815                 nfs_refresh_inode(inode, res.fattr);
2816         }
2817         nfs_free_fattr(res.fattr);
2818         return status;
2819 }
2820
2821 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2822 {
2823         struct nfs4_exception exception = { };
2824         int err;
2825         do {
2826                 err = nfs4_handle_exception(NFS_SERVER(inode),
2827                                 _nfs4_proc_access(inode, entry),
2828                                 &exception);
2829         } while (exception.retry);
2830         return err;
2831 }
2832
2833 /*
2834  * TODO: For the time being, we don't try to get any attributes
2835  * along with any of the zero-copy operations READ, READDIR,
2836  * READLINK, WRITE.
2837  *
2838  * In the case of the first three, we want to put the GETATTR
2839  * after the read-type operation -- this is because it is hard
2840  * to predict the length of a GETATTR response in v4, and thus
2841  * align the READ data correctly.  This means that the GETATTR
2842  * may end up partially falling into the page cache, and we should
2843  * shift it into the 'tail' of the xdr_buf before processing.
2844  * To do this efficiently, we need to know the total length
2845  * of data received, which doesn't seem to be available outside
2846  * of the RPC layer.
2847  *
2848  * In the case of WRITE, we also want to put the GETATTR after
2849  * the operation -- in this case because we want to make sure
2850  * we get the post-operation mtime and size.
2851  *
2852  * Both of these changes to the XDR layer would in fact be quite
2853  * minor, but I decided to leave them for a subsequent patch.
2854  */
2855 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2856                 unsigned int pgbase, unsigned int pglen)
2857 {
2858         struct nfs4_readlink args = {
2859                 .fh       = NFS_FH(inode),
2860                 .pgbase   = pgbase,
2861                 .pglen    = pglen,
2862                 .pages    = &page,
2863         };
2864         struct nfs4_readlink_res res;
2865         struct rpc_message msg = {
2866                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2867                 .rpc_argp = &args,
2868                 .rpc_resp = &res,
2869         };
2870
2871         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2872 }
2873
2874 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2875                 unsigned int pgbase, unsigned int pglen)
2876 {
2877         struct nfs4_exception exception = { };
2878         int err;
2879         do {
2880                 err = nfs4_handle_exception(NFS_SERVER(inode),
2881                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2882                                 &exception);
2883         } while (exception.retry);
2884         return err;
2885 }
2886
2887 /*
2888  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2889  */
2890 static int
2891 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2892                  int flags)
2893 {
2894         struct nfs_open_context *ctx;
2895         struct nfs4_state *state;
2896         int status = 0;
2897
2898         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2899         if (IS_ERR(ctx))
2900                 return PTR_ERR(ctx);
2901
2902         sattr->ia_mode &= ~current_umask();
2903         state = nfs4_do_open(dir, dentry, ctx->mode,
2904                         flags, sattr, ctx->cred,
2905                         &ctx->mdsthreshold);
2906         d_drop(dentry);
2907         if (IS_ERR(state)) {
2908                 status = PTR_ERR(state);
2909                 goto out;
2910         }
2911         d_add(dentry, igrab(state->inode));
2912         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2913         ctx->state = state;
2914 out:
2915         put_nfs_open_context(ctx);
2916         return status;
2917 }
2918
2919 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2920 {
2921         struct nfs_server *server = NFS_SERVER(dir);
2922         struct nfs_removeargs args = {
2923                 .fh = NFS_FH(dir),
2924                 .name = *name,
2925         };
2926         struct nfs_removeres res = {
2927                 .server = server,
2928         };
2929         struct rpc_message msg = {
2930                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2931                 .rpc_argp = &args,
2932                 .rpc_resp = &res,
2933         };
2934         int status;
2935
2936         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2937         if (status == 0)
2938                 update_changeattr(dir, &res.cinfo);
2939         return status;
2940 }
2941
2942 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2943 {
2944         struct nfs4_exception exception = { };
2945         int err;
2946         do {
2947                 err = nfs4_handle_exception(NFS_SERVER(dir),
2948                                 _nfs4_proc_remove(dir, name),
2949                                 &exception);
2950         } while (exception.retry);
2951         return err;
2952 }
2953
2954 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2955 {
2956         struct nfs_server *server = NFS_SERVER(dir);
2957         struct nfs_removeargs *args = msg->rpc_argp;
2958         struct nfs_removeres *res = msg->rpc_resp;
2959
2960         res->server = server;
2961         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2962         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2963 }
2964
2965 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
2966 {
2967         nfs4_setup_sequence(NFS_SERVER(data->dir),
2968                         &data->args.seq_args,
2969                         &data->res.seq_res,
2970                         task);
2971 }
2972
2973 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2974 {
2975         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2976
2977         if (!nfs4_sequence_done(task, &res->seq_res))
2978                 return 0;
2979         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2980                 return 0;
2981         update_changeattr(dir, &res->cinfo);
2982         return 1;
2983 }
2984
2985 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2986 {
2987         struct nfs_server *server = NFS_SERVER(dir);
2988         struct nfs_renameargs *arg = msg->rpc_argp;
2989         struct nfs_renameres *res = msg->rpc_resp;
2990
2991         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2992         res->server = server;
2993         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2994 }
2995
2996 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
2997 {
2998         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
2999                         &data->args.seq_args,
3000                         &data->res.seq_res,
3001                         task);
3002 }
3003
3004 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3005                                  struct inode *new_dir)
3006 {
3007         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3008
3009         if (!nfs4_sequence_done(task, &res->seq_res))
3010                 return 0;
3011         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3012                 return 0;
3013
3014         update_changeattr(old_dir, &res->old_cinfo);
3015         update_changeattr(new_dir, &res->new_cinfo);
3016         return 1;
3017 }
3018
3019 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3020                 struct inode *new_dir, struct qstr *new_name)
3021 {
3022         struct nfs_server *server = NFS_SERVER(old_dir);
3023         struct nfs_renameargs arg = {
3024                 .old_dir = NFS_FH(old_dir),
3025                 .new_dir = NFS_FH(new_dir),
3026                 .old_name = old_name,
3027                 .new_name = new_name,
3028         };
3029         struct nfs_renameres res = {
3030                 .server = server,
3031         };
3032         struct rpc_message msg = {
3033                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3034                 .rpc_argp = &arg,
3035                 .rpc_resp = &res,
3036         };
3037         int status = -ENOMEM;
3038         
3039         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3040         if (!status) {
3041                 update_changeattr(old_dir, &res.old_cinfo);
3042                 update_changeattr(new_dir, &res.new_cinfo);
3043         }
3044         return status;
3045 }
3046
3047 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3048                 struct inode *new_dir, struct qstr *new_name)
3049 {
3050         struct nfs4_exception exception = { };
3051         int err;
3052         do {
3053                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3054                                 _nfs4_proc_rename(old_dir, old_name,
3055                                         new_dir, new_name),
3056                                 &exception);
3057         } while (exception.retry);
3058         return err;
3059 }
3060
3061 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3062 {
3063         struct nfs_server *server = NFS_SERVER(inode);
3064         struct nfs4_link_arg arg = {
3065                 .fh     = NFS_FH(inode),
3066                 .dir_fh = NFS_FH(dir),
3067                 .name   = name,
3068                 .bitmask = server->attr_bitmask,
3069         };
3070         struct nfs4_link_res res = {
3071                 .server = server,
3072         };
3073         struct rpc_message msg = {
3074                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3075                 .rpc_argp = &arg,
3076                 .rpc_resp = &res,
3077         };
3078         int status = -ENOMEM;
3079
3080         res.fattr = nfs_alloc_fattr();
3081         if (res.fattr == NULL)
3082                 goto out;
3083
3084         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3085         if (!status) {
3086                 update_changeattr(dir, &res.cinfo);
3087                 nfs_post_op_update_inode(inode, res.fattr);
3088         }
3089 out:
3090         nfs_free_fattr(res.fattr);
3091         return status;
3092 }
3093
3094 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3095 {
3096         struct nfs4_exception exception = { };
3097         int err;
3098         do {
3099                 err = nfs4_handle_exception(NFS_SERVER(inode),
3100                                 _nfs4_proc_link(inode, dir, name),
3101                                 &exception);
3102         } while (exception.retry);
3103         return err;
3104 }
3105
3106 struct nfs4_createdata {
3107         struct rpc_message msg;
3108         struct nfs4_create_arg arg;
3109         struct nfs4_create_res res;
3110         struct nfs_fh fh;
3111         struct nfs_fattr fattr;
3112 };
3113
3114 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3115                 struct qstr *name, struct iattr *sattr, u32 ftype)
3116 {
3117         struct nfs4_createdata *data;
3118
3119         data = kzalloc(sizeof(*data), GFP_KERNEL);
3120         if (data != NULL) {
3121                 struct nfs_server *server = NFS_SERVER(dir);
3122
3123                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3124                 data->msg.rpc_argp = &data->arg;
3125                 data->msg.rpc_resp = &data->res;
3126                 data->arg.dir_fh = NFS_FH(dir);
3127                 data->arg.server = server;
3128                 data->arg.name = name;
3129                 data->arg.attrs = sattr;
3130                 data->arg.ftype = ftype;
3131                 data->arg.bitmask = server->attr_bitmask;
3132                 data->res.server = server;
3133                 data->res.fh = &data->fh;
3134                 data->res.fattr = &data->fattr;
3135                 nfs_fattr_init(data->res.fattr);
3136         }
3137         return data;
3138 }
3139
3140 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3141 {
3142         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3143                                     &data->arg.seq_args, &data->res.seq_res, 1);
3144         if (status == 0) {
3145                 update_changeattr(dir, &data->res.dir_cinfo);
3146                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3147         }
3148         return status;
3149 }
3150
3151 static void nfs4_free_createdata(struct nfs4_createdata *data)
3152 {
3153         kfree(data);
3154 }
3155
3156 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3157                 struct page *page, unsigned int len, struct iattr *sattr)
3158 {
3159         struct nfs4_createdata *data;
3160         int status = -ENAMETOOLONG;
3161
3162         if (len > NFS4_MAXPATHLEN)
3163                 goto out;
3164
3165         status = -ENOMEM;
3166         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3167         if (data == NULL)
3168                 goto out;
3169
3170         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3171         data->arg.u.symlink.pages = &page;
3172         data->arg.u.symlink.len = len;
3173         
3174         status = nfs4_do_create(dir, dentry, data);
3175
3176         nfs4_free_createdata(data);
3177 out:
3178         return status;
3179 }
3180
3181 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3182                 struct page *page, unsigned int len, struct iattr *sattr)
3183 {
3184         struct nfs4_exception exception = { };
3185         int err;
3186         do {
3187                 err = nfs4_handle_exception(NFS_SERVER(dir),
3188                                 _nfs4_proc_symlink(dir, dentry, page,
3189                                                         len, sattr),
3190                                 &exception);
3191         } while (exception.retry);
3192         return err;
3193 }
3194
3195 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3196                 struct iattr *sattr)
3197 {
3198         struct nfs4_createdata *data;
3199         int status = -ENOMEM;
3200
3201         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3202         if (data == NULL)
3203                 goto out;
3204
3205         status = nfs4_do_create(dir, dentry, data);
3206
3207         nfs4_free_createdata(data);
3208 out:
3209         return status;
3210 }
3211
3212 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3213                 struct iattr *sattr)
3214 {
3215         struct nfs4_exception exception = { };
3216         int err;
3217
3218         sattr->ia_mode &= ~current_umask();
3219         do {
3220                 err = nfs4_handle_exception(NFS_SERVER(dir),
3221                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3222                                 &exception);
3223         } while (exception.retry);
3224         return err;
3225 }
3226
3227 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3228                 u64 cookie, struct page **pages, unsigned int count, int plus)
3229 {
3230         struct inode            *dir = dentry->d_inode;
3231         struct nfs4_readdir_arg args = {
3232                 .fh = NFS_FH(dir),
3233                 .pages = pages,
3234                 .pgbase = 0,
3235                 .count = count,
3236                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3237                 .plus = plus,
3238         };
3239         struct nfs4_readdir_res res;
3240         struct rpc_message msg = {
3241                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3242                 .rpc_argp = &args,
3243                 .rpc_resp = &res,
3244                 .rpc_cred = cred,
3245         };
3246         int                     status;
3247
3248         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3249                         dentry->d_parent->d_name.name,
3250                         dentry->d_name.name,
3251                         (unsigned long long)cookie);
3252         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3253         res.pgbase = args.pgbase;
3254         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3255         if (status >= 0) {
3256                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3257                 status += args.pgbase;
3258         }
3259
3260         nfs_invalidate_atime(dir);
3261
3262         dprintk("%s: returns %d\n", __func__, status);
3263         return status;
3264 }
3265
3266 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3267                 u64 cookie, struct page **pages, unsigned int count, int plus)
3268 {
3269         struct nfs4_exception exception = { };
3270         int err;
3271         do {
3272                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3273                                 _nfs4_proc_readdir(dentry, cred, cookie,
3274                                         pages, count, plus),
3275                                 &exception);
3276         } while (exception.retry);
3277         return err;
3278 }
3279
3280 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3281                 struct iattr *sattr, dev_t rdev)
3282 {
3283         struct nfs4_createdata *data;
3284         int mode = sattr->ia_mode;
3285         int status = -ENOMEM;
3286
3287         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3288         if (data == NULL)
3289                 goto out;
3290
3291         if (S_ISFIFO(mode))
3292                 data->arg.ftype = NF4FIFO;
3293         else if (S_ISBLK(mode)) {
3294                 data->arg.ftype = NF4BLK;
3295                 data->arg.u.device.specdata1 = MAJOR(rdev);
3296                 data->arg.u.device.specdata2 = MINOR(rdev);
3297         }
3298         else if (S_ISCHR(mode)) {
3299                 data->arg.ftype = NF4CHR;
3300                 data->arg.u.device.specdata1 = MAJOR(rdev);
3301                 data->arg.u.device.specdata2 = MINOR(rdev);
3302         } else if (!S_ISSOCK(mode)) {
3303                 status = -EINVAL;
3304                 goto out_free;
3305         }
3306         
3307         status = nfs4_do_create(dir, dentry, data);
3308 out_free:
3309         nfs4_free_createdata(data);
3310 out:
3311         return status;
3312 }
3313
3314 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3315                 struct iattr *sattr, dev_t rdev)
3316 {
3317         struct nfs4_exception exception = { };
3318         int err;
3319
3320         sattr->ia_mode &= ~current_umask();
3321         do {
3322                 err = nfs4_handle_exception(NFS_SERVER(dir),
3323                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3324                                 &exception);
3325         } while (exception.retry);
3326         return err;
3327 }
3328
3329 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3330                  struct nfs_fsstat *fsstat)
3331 {
3332         struct nfs4_statfs_arg args = {
3333                 .fh = fhandle,
3334                 .bitmask = server->attr_bitmask,
3335         };
3336         struct nfs4_statfs_res res = {
3337                 .fsstat = fsstat,
3338         };
3339         struct rpc_message msg = {
3340                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3341                 .rpc_argp = &args,
3342                 .rpc_resp = &res,
3343         };
3344
3345         nfs_fattr_init(fsstat->fattr);
3346         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3347 }
3348
3349 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3350 {
3351         struct nfs4_exception exception = { };
3352         int err;
3353         do {
3354                 err = nfs4_handle_exception(server,
3355                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3356                                 &exception);
3357         } while (exception.retry);
3358         return err;
3359 }
3360
3361 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3362                 struct nfs_fsinfo *fsinfo)
3363 {
3364         struct nfs4_fsinfo_arg args = {
3365                 .fh = fhandle,
3366                 .bitmask = server->attr_bitmask,
3367         };
3368         struct nfs4_fsinfo_res res = {
3369                 .fsinfo = fsinfo,
3370         };
3371         struct rpc_message msg = {
3372                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3373                 .rpc_argp = &args,
3374                 .rpc_resp = &res,
3375         };
3376
3377         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3378 }
3379
3380 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3381 {
3382         struct nfs4_exception exception = { };
3383         int err;
3384
3385         do {
3386                 err = nfs4_handle_exception(server,
3387                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3388                                 &exception);
3389         } while (exception.retry);
3390         return err;
3391 }
3392
3393 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3394 {
3395         int error;
3396
3397         nfs_fattr_init(fsinfo->fattr);
3398         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3399         if (error == 0) {
3400                 /* block layout checks this! */
3401                 server->pnfs_blksize = fsinfo->blksize;
3402                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3403         }
3404
3405         return error;
3406 }
3407
3408 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3409                 struct nfs_pathconf *pathconf)
3410 {
3411         struct nfs4_pathconf_arg args = {
3412                 .fh = fhandle,
3413                 .bitmask = server->attr_bitmask,
3414         };
3415         struct nfs4_pathconf_res res = {
3416                 .pathconf = pathconf,
3417         };
3418         struct rpc_message msg = {
3419                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3420                 .rpc_argp = &args,
3421                 .rpc_resp = &res,
3422         };
3423
3424         /* None of the pathconf attributes are mandatory to implement */
3425         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3426                 memset(pathconf, 0, sizeof(*pathconf));
3427                 return 0;
3428         }
3429
3430         nfs_fattr_init(pathconf->fattr);
3431         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3432 }
3433
3434 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3435                 struct nfs_pathconf *pathconf)
3436 {
3437         struct nfs4_exception exception = { };
3438         int err;
3439
3440         do {
3441                 err = nfs4_handle_exception(server,
3442                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3443                                 &exception);
3444         } while (exception.retry);
3445         return err;
3446 }
3447
3448 void __nfs4_read_done_cb(struct nfs_read_data *data)
3449 {
3450         nfs_invalidate_atime(data->header->inode);
3451 }
3452
3453 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3454 {
3455         struct nfs_server *server = NFS_SERVER(data->header->inode);
3456
3457         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3458                 rpc_restart_call_prepare(task);
3459                 return -EAGAIN;
3460         }
3461
3462         __nfs4_read_done_cb(data);
3463         if (task->tk_status > 0)
3464                 renew_lease(server, data->timestamp);
3465         return 0;
3466 }
3467
3468 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3469 {
3470
3471         dprintk("--> %s\n", __func__);
3472
3473         if (!nfs4_sequence_done(task, &data->res.seq_res))
3474                 return -EAGAIN;
3475
3476         return data->read_done_cb ? data->read_done_cb(task, data) :
3477                                     nfs4_read_done_cb(task, data);
3478 }
3479
3480 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3481 {
3482         data->timestamp   = jiffies;
3483         data->read_done_cb = nfs4_read_done_cb;
3484         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3485         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3486 }
3487
3488 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3489 {
3490         nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3491                         &data->args.seq_args,
3492                         &data->res.seq_res,
3493                         task);
3494 }
3495
3496 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3497 {
3498         struct inode *inode = data->header->inode;
3499         
3500         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3501                 rpc_restart_call_prepare(task);
3502                 return -EAGAIN;
3503         }
3504         if (task->tk_status >= 0) {
3505                 renew_lease(NFS_SERVER(inode), data->timestamp);
3506                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3507         }
3508         return 0;
3509 }
3510
3511 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3512 {
3513         if (!nfs4_sequence_done(task, &data->res.seq_res))
3514                 return -EAGAIN;
3515         return data->write_done_cb ? data->write_done_cb(task, data) :
3516                 nfs4_write_done_cb(task, data);
3517 }
3518
3519 static
3520 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3521 {
3522         const struct nfs_pgio_header *hdr = data->header;
3523
3524         /* Don't request attributes for pNFS or O_DIRECT writes */
3525         if (data->ds_clp != NULL || hdr->dreq != NULL)
3526                 return false;
3527         /* Otherwise, request attributes if and only if we don't hold
3528          * a delegation
3529          */
3530         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3531 }
3532
3533 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3534 {
3535         struct nfs_server *server = NFS_SERVER(data->header->inode);
3536
3537         if (!nfs4_write_need_cache_consistency_data(data)) {
3538                 data->args.bitmask = NULL;
3539                 data->res.fattr = NULL;
3540         } else
3541                 data->args.bitmask = server->cache_consistency_bitmask;
3542
3543         if (!data->write_done_cb)
3544                 data->write_done_cb = nfs4_write_done_cb;
3545         data->res.server = server;
3546         data->timestamp   = jiffies;
3547
3548         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3549         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3550 }
3551
3552 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3553 {
3554         nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3555                         &data->args.seq_args,
3556                         &data->res.seq_res,
3557                         task);
3558 }
3559
3560 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3561 {
3562         nfs4_setup_sequence(NFS_SERVER(data->inode),
3563                         &data->args.seq_args,
3564                         &data->res.seq_res,
3565                         task);
3566 }
3567
3568 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3569 {
3570         struct inode *inode = data->inode;
3571
3572         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3573                 rpc_restart_call_prepare(task);
3574                 return -EAGAIN;
3575         }
3576         return 0;
3577 }
3578
3579 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3580 {
3581         if (!nfs4_sequence_done(task, &data->res.seq_res))
3582                 return -EAGAIN;
3583         return data->commit_done_cb(task, data);
3584 }
3585
3586 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3587 {
3588         struct nfs_server *server = NFS_SERVER(data->inode);
3589
3590         if (data->commit_done_cb == NULL)
3591                 data->commit_done_cb = nfs4_commit_done_cb;
3592         data->res.server = server;
3593         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3594         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3595 }
3596
3597 struct nfs4_renewdata {
3598         struct nfs_client       *client;
3599         unsigned long           timestamp;
3600 };
3601
3602 /*
3603  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3604  * standalone procedure for queueing an asynchronous RENEW.
3605  */
3606 static void nfs4_renew_release(void *calldata)
3607 {
3608         struct nfs4_renewdata *data = calldata;
3609         struct nfs_client *clp = data->client;
3610
3611         if (atomic_read(&clp->cl_count) > 1)
3612                 nfs4_schedule_state_renewal(clp);
3613         nfs_put_client(clp);
3614         kfree(data);
3615 }
3616
3617 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3618 {
3619         struct nfs4_renewdata *data = calldata;
3620         struct nfs_client *clp = data->client;
3621         unsigned long timestamp = data->timestamp;
3622
3623         if (task->tk_status < 0) {
3624                 /* Unless we're shutting down, schedule state recovery! */
3625                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3626                         return;
3627                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3628                         nfs4_schedule_lease_recovery(clp);
3629                         return;
3630                 }
3631                 nfs4_schedule_path_down_recovery(clp);
3632         }
3633         do_renew_lease(clp, timestamp);
3634 }
3635
3636 static const struct rpc_call_ops nfs4_renew_ops = {
3637         .rpc_call_done = nfs4_renew_done,
3638         .rpc_release = nfs4_renew_release,
3639 };
3640
3641 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3642 {
3643         struct rpc_message msg = {
3644                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3645                 .rpc_argp       = clp,
3646                 .rpc_cred       = cred,
3647         };
3648         struct nfs4_renewdata *data;
3649
3650         if (renew_flags == 0)
3651                 return 0;
3652         if (!atomic_inc_not_zero(&clp->cl_count))
3653                 return -EIO;
3654         data = kmalloc(sizeof(*data), GFP_NOFS);
3655         if (data == NULL)
3656                 return -ENOMEM;
3657         data->client = clp;
3658         data->timestamp = jiffies;
3659         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3660                         &nfs4_renew_ops, data);
3661 }
3662
3663 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3664 {
3665         struct rpc_message msg = {
3666                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3667                 .rpc_argp       = clp,
3668                 .rpc_cred       = cred,
3669         };
3670         unsigned long now = jiffies;
3671         int status;
3672
3673         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3674         if (status < 0)
3675                 return status;
3676         do_renew_lease(clp, now);
3677         return 0;
3678 }
3679
3680 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3681 {
3682         return (server->caps & NFS_CAP_ACLS)
3683                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3684                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3685 }
3686
3687 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3688  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3689  * the stack.
3690  */
3691 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3692
3693 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3694                 struct page **pages, unsigned int *pgbase)
3695 {
3696         struct page *newpage, **spages;
3697         int rc = 0;
3698         size_t len;
3699         spages = pages;
3700
3701         do {
3702                 len = min_t(size_t, PAGE_SIZE, buflen);
3703                 newpage = alloc_page(GFP_KERNEL);
3704
3705                 if (newpage == NULL)
3706                         goto unwind;
3707                 memcpy(page_address(newpage), buf, len);
3708                 buf += len;
3709                 buflen -= len;
3710                 *pages++ = newpage;
3711                 rc++;
3712         } while (buflen != 0);
3713
3714         return rc;
3715
3716 unwind:
3717         for(; rc > 0; rc--)
3718                 __free_page(spages[rc-1]);
3719         return -ENOMEM;
3720 }
3721
3722 struct nfs4_cached_acl {
3723         int cached;
3724         size_t len;
3725         char data[0];
3726 };
3727
3728 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3729 {
3730         struct nfs_inode *nfsi = NFS_I(inode);
3731
3732         spin_lock(&inode->i_lock);
3733         kfree(nfsi->nfs4_acl);
3734         nfsi->nfs4_acl = acl;
3735         spin_unlock(&inode->i_lock);
3736 }
3737
3738 static void nfs4_zap_acl_attr(struct inode *inode)
3739 {
3740         nfs4_set_cached_acl(inode, NULL);
3741 }
3742
3743 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3744 {
3745         struct nfs_inode *nfsi = NFS_I(inode);
3746         struct nfs4_cached_acl *acl;
3747         int ret = -ENOENT;
3748
3749         spin_lock(&inode->i_lock);
3750         acl = nfsi->nfs4_acl;
3751         if (acl == NULL)
3752                 goto out;
3753         if (buf == NULL) /* user is just asking for length */
3754                 goto out_len;
3755         if (acl->cached == 0)
3756                 goto out;
3757         ret = -ERANGE; /* see getxattr(2) man page */
3758         if (acl->len > buflen)
3759                 goto out;
3760         memcpy(buf, acl->data, acl->len);
3761 out_len:
3762         ret = acl->len;
3763 out:
3764         spin_unlock(&inode->i_lock);
3765         return ret;
3766 }
3767
3768 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3769 {
3770         struct nfs4_cached_acl *acl;
3771         size_t buflen = sizeof(*acl) + acl_len;
3772
3773         if (buflen <= PAGE_SIZE) {
3774                 acl = kmalloc(buflen, GFP_KERNEL);
3775                 if (acl == NULL)
3776                         goto out;
3777                 acl->cached = 1;
3778                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3779         } else {
3780                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3781                 if (acl == NULL)
3782                         goto out;
3783                 acl->cached = 0;
3784         }
3785         acl->len = acl_len;
3786 out:
3787         nfs4_set_cached_acl(inode, acl);
3788 }
3789
3790 /*
3791  * The getxattr API returns the required buffer length when called with a
3792  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3793  * the required buf.  On a NULL buf, we send a page of data to the server
3794  * guessing that the ACL request can be serviced by a page. If so, we cache
3795  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3796  * the cache. If not so, we throw away the page, and cache the required
3797  * length. The next getxattr call will then produce another round trip to
3798  * the server, this time with the input buf of the required size.
3799  */
3800 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3801 {
3802         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3803         struct nfs_getaclargs args = {
3804                 .fh = NFS_FH(inode),
3805                 .acl_pages = pages,
3806                 .acl_len = buflen,
3807         };
3808         struct nfs_getaclres res = {
3809                 .acl_len = buflen,
3810         };
3811         struct rpc_message msg = {
3812                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3813                 .rpc_argp = &args,
3814                 .rpc_resp = &res,
3815         };
3816         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3817         int ret = -ENOMEM, i;
3818
3819         /* As long as we're doing a round trip to the server anyway,
3820          * let's be prepared for a page of acl data. */
3821         if (npages == 0)
3822                 npages = 1;
3823         if (npages > ARRAY_SIZE(pages))
3824                 return -ERANGE;
3825
3826         for (i = 0; i < npages; i++) {
3827                 pages[i] = alloc_page(GFP_KERNEL);
3828                 if (!pages[i])
3829                         goto out_free;
3830         }
3831
3832         /* for decoding across pages */
3833         res.acl_scratch = alloc_page(GFP_KERNEL);
3834         if (!res.acl_scratch)
3835                 goto out_free;
3836
3837         args.acl_len = npages * PAGE_SIZE;
3838         args.acl_pgbase = 0;
3839
3840         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3841                 __func__, buf, buflen, npages, args.acl_len);
3842         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3843                              &msg, &args.seq_args, &res.seq_res, 0);
3844         if (ret)
3845                 goto out_free;
3846
3847         /* Handle the case where the passed-in buffer is too short */
3848         if (res.acl_flags & NFS4_ACL_TRUNC) {
3849                 /* Did the user only issue a request for the acl length? */
3850                 if (buf == NULL)
3851                         goto out_ok;
3852                 ret = -ERANGE;
3853                 goto out_free;
3854         }
3855         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
3856         if (buf) {
3857                 if (res.acl_len > buflen) {
3858                         ret = -ERANGE;
3859                         goto out_free;
3860                 }
3861                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
3862         }
3863 out_ok:
3864         ret = res.acl_len;
3865 out_free:
3866         for (i = 0; i < npages; i++)
3867                 if (pages[i])
3868                         __free_page(pages[i]);
3869         if (res.acl_scratch)
3870                 __free_page(res.acl_scratch);
3871         return ret;
3872 }
3873
3874 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3875 {
3876         struct nfs4_exception exception = { };
3877         ssize_t ret;
3878         do {
3879                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3880                 if (ret >= 0)
3881                         break;
3882                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3883         } while (exception.retry);
3884         return ret;
3885 }
3886
3887 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3888 {
3889         struct nfs_server *server = NFS_SERVER(inode);
3890         int ret;
3891
3892         if (!nfs4_server_supports_acls(server))
3893                 return -EOPNOTSUPP;
3894         ret = nfs_revalidate_inode(server, inode);
3895         if (ret < 0)
3896                 return ret;
3897         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3898                 nfs_zap_acl_cache(inode);
3899         ret = nfs4_read_cached_acl(inode, buf, buflen);
3900         if (ret != -ENOENT)
3901                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3902                  * but no cached acl data, just the acl length */
3903                 return ret;
3904         return nfs4_get_acl_uncached(inode, buf, buflen);
3905 }
3906
3907 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3908 {
3909         struct nfs_server *server = NFS_SERVER(inode);
3910         struct page *pages[NFS4ACL_MAXPAGES];
3911         struct nfs_setaclargs arg = {
3912                 .fh             = NFS_FH(inode),
3913                 .acl_pages      = pages,
3914                 .acl_len        = buflen,
3915         };
3916         struct nfs_setaclres res;
3917         struct rpc_message msg = {
3918                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3919                 .rpc_argp       = &arg,
3920                 .rpc_resp       = &res,
3921         };
3922         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3923         int ret, i;
3924
3925         if (!nfs4_server_supports_acls(server))
3926                 return -EOPNOTSUPP;
3927         if (npages > ARRAY_SIZE(pages))
3928                 return -ERANGE;
3929         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3930         if (i < 0)
3931                 return i;
3932         nfs4_inode_return_delegation(inode);
3933         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3934
3935         /*
3936          * Free each page after tx, so the only ref left is
3937          * held by the network stack
3938          */
3939         for (; i > 0; i--)
3940                 put_page(pages[i-1]);
3941
3942         /*
3943          * Acl update can result in inode attribute update.
3944          * so mark the attribute cache invalid.
3945          */
3946         spin_lock(&inode->i_lock);
3947         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3948         spin_unlock(&inode->i_lock);
3949         nfs_access_zap_cache(inode);
3950         nfs_zap_acl_cache(inode);
3951         return ret;
3952 }
3953
3954 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3955 {
3956         struct nfs4_exception exception = { };
3957         int err;
3958         do {
3959                 err = nfs4_handle_exception(NFS_SERVER(inode),
3960                                 __nfs4_proc_set_acl(inode, buf, buflen),
3961                                 &exception);
3962         } while (exception.retry);
3963         return err;
3964 }
3965
3966 static int
3967 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3968 {
3969         struct nfs_client *clp = server->nfs_client;
3970
3971         if (task->tk_status >= 0)
3972                 return 0;
3973         switch(task->tk_status) {
3974                 case -NFS4ERR_DELEG_REVOKED:
3975                 case -NFS4ERR_ADMIN_REVOKED:
3976                 case -NFS4ERR_BAD_STATEID:
3977                         if (state == NULL)
3978                                 break;
3979                         nfs_remove_bad_delegation(state->inode);
3980                 case -NFS4ERR_OPENMODE:
3981                         if (state == NULL)
3982                                 break;
3983                         nfs4_schedule_stateid_recovery(server, state);
3984                         goto wait_on_recovery;
3985                 case -NFS4ERR_EXPIRED:
3986                         if (state != NULL)
3987                                 nfs4_schedule_stateid_recovery(server, state);
3988                 case -NFS4ERR_STALE_STATEID:
3989                 case -NFS4ERR_STALE_CLIENTID:
3990                         nfs4_schedule_lease_recovery(clp);
3991                         goto wait_on_recovery;
3992 #if defined(CONFIG_NFS_V4_1)
3993                 case -NFS4ERR_BADSESSION:
3994                 case -NFS4ERR_BADSLOT:
3995                 case -NFS4ERR_BAD_HIGH_SLOT:
3996                 case -NFS4ERR_DEADSESSION:
3997                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3998                 case -NFS4ERR_SEQ_FALSE_RETRY:
3999                 case -NFS4ERR_SEQ_MISORDERED:
4000                         dprintk("%s ERROR %d, Reset session\n", __func__,
4001                                 task->tk_status);
4002                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4003                         task->tk_status = 0;
4004                         return -EAGAIN;
4005 #endif /* CONFIG_NFS_V4_1 */
4006                 case -NFS4ERR_DELAY:
4007                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4008                 case -NFS4ERR_GRACE:
4009                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4010                         task->tk_status = 0;
4011                         return -EAGAIN;
4012                 case -NFS4ERR_RETRY_UNCACHED_REP:
4013                 case -NFS4ERR_OLD_STATEID:
4014                         task->tk_status = 0;
4015                         return -EAGAIN;
4016         }
4017         task->tk_status = nfs4_map_errors(task->tk_status);
4018         return 0;
4019 wait_on_recovery:
4020         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4021         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4022                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4023         task->tk_status = 0;
4024         return -EAGAIN;
4025 }
4026
4027 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4028                                     nfs4_verifier *bootverf)
4029 {
4030         __be32 verf[2];
4031
4032         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4033                 /* An impossible timestamp guarantees this value
4034                  * will never match a generated boot time. */
4035                 verf[0] = 0;
4036                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4037         } else {
4038                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4039                 verf[0] = (__be32)nn->boot_time.tv_sec;
4040                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4041         }
4042         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4043 }
4044
4045 static unsigned int
4046 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4047                                    char *buf, size_t len)
4048 {
4049         unsigned int result;
4050
4051         rcu_read_lock();
4052         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4053                                 clp->cl_ipaddr,
4054                                 rpc_peeraddr2str(clp->cl_rpcclient,
4055                                                         RPC_DISPLAY_ADDR),
4056                                 rpc_peeraddr2str(clp->cl_rpcclient,
4057                                                         RPC_DISPLAY_PROTO));
4058         rcu_read_unlock();
4059         return result;
4060 }
4061
4062 static unsigned int
4063 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4064                                 char *buf, size_t len)
4065 {
4066         char *nodename = clp->cl_rpcclient->cl_nodename;
4067
4068         if (nfs4_client_id_uniquifier[0] != '\0')
4069                 nodename = nfs4_client_id_uniquifier;
4070         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4071                                 clp->rpc_ops->version, clp->cl_minorversion,
4072                                 nodename);
4073 }
4074
4075 /**
4076  * nfs4_proc_setclientid - Negotiate client ID
4077  * @clp: state data structure
4078  * @program: RPC program for NFSv4 callback service
4079  * @port: IP port number for NFS4 callback service
4080  * @cred: RPC credential to use for this call
4081  * @res: where to place the result
4082  *
4083  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4084  */
4085 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4086                 unsigned short port, struct rpc_cred *cred,
4087                 struct nfs4_setclientid_res *res)
4088 {
4089         nfs4_verifier sc_verifier;
4090         struct nfs4_setclientid setclientid = {
4091                 .sc_verifier = &sc_verifier,
4092                 .sc_prog = program,
4093                 .sc_cb_ident = clp->cl_cb_ident,
4094         };
4095         struct rpc_message msg = {
4096                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4097                 .rpc_argp = &setclientid,
4098                 .rpc_resp = res,
4099                 .rpc_cred = cred,
4100         };
4101         int status;
4102
4103         /* nfs_client_id4 */
4104         nfs4_init_boot_verifier(clp, &sc_verifier);
4105         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4106                 setclientid.sc_name_len =
4107                                 nfs4_init_uniform_client_string(clp,
4108                                                 setclientid.sc_name,
4109                                                 sizeof(setclientid.sc_name));
4110         else
4111                 setclientid.sc_name_len =
4112                                 nfs4_init_nonuniform_client_string(clp,
4113                                                 setclientid.sc_name,
4114                                                 sizeof(setclientid.sc_name));
4115         /* cb_client4 */
4116         rcu_read_lock();
4117         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4118                                 sizeof(setclientid.sc_netid),
4119                                 rpc_peeraddr2str(clp->cl_rpcclient,
4120                                                         RPC_DISPLAY_NETID));
4121         rcu_read_unlock();
4122         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4123                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4124                                 clp->cl_ipaddr, port >> 8, port & 255);
4125
4126         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4127                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4128                 setclientid.sc_name_len, setclientid.sc_name);
4129         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4130         dprintk("NFS reply setclientid: %d\n", status);
4131         return status;
4132 }
4133
4134 /**
4135  * nfs4_proc_setclientid_confirm - Confirm client ID
4136  * @clp: state data structure
4137  * @res: result of a previous SETCLIENTID
4138  * @cred: RPC credential to use for this call
4139  *
4140  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4141  */
4142 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4143                 struct nfs4_setclientid_res *arg,
4144                 struct rpc_cred *cred)
4145 {
4146         struct nfs_fsinfo fsinfo;
4147         struct rpc_message msg = {
4148                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4149                 .rpc_argp = arg,
4150                 .rpc_resp = &fsinfo,
4151                 .rpc_cred = cred,
4152         };
4153         unsigned long now;
4154         int status;
4155
4156         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4157                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4158                 clp->cl_clientid);
4159         now = jiffies;
4160         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4161         if (status == 0) {
4162                 spin_lock(&clp->cl_lock);
4163                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4164                 clp->cl_last_renewal = now;
4165                 spin_unlock(&clp->cl_lock);
4166         }
4167         dprintk("NFS reply setclientid_confirm: %d\n", status);
4168         return status;
4169 }
4170
4171 struct nfs4_delegreturndata {
4172         struct nfs4_delegreturnargs args;
4173         struct nfs4_delegreturnres res;
4174         struct nfs_fh fh;
4175         nfs4_stateid stateid;
4176         unsigned long timestamp;
4177         struct nfs_fattr fattr;
4178         int rpc_status;
4179 };
4180
4181 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4182 {
4183         struct nfs4_delegreturndata *data = calldata;
4184
4185         if (!nfs4_sequence_done(task, &data->res.seq_res))
4186                 return;
4187
4188         switch (task->tk_status) {
4189         case -NFS4ERR_STALE_STATEID:
4190         case -NFS4ERR_EXPIRED:
4191         case 0:
4192                 renew_lease(data->res.server, data->timestamp);
4193                 break;
4194         default:
4195                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4196                                 -EAGAIN) {
4197                         rpc_restart_call_prepare(task);
4198                         return;
4199                 }
4200         }
4201         data->rpc_status = task->tk_status;
4202 }
4203
4204 static void nfs4_delegreturn_release(void *calldata)
4205 {
4206         kfree(calldata);
4207 }
4208
4209 #if defined(CONFIG_NFS_V4_1)
4210 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4211 {
4212         struct nfs4_delegreturndata *d_data;
4213
4214         d_data = (struct nfs4_delegreturndata *)data;
4215
4216         nfs4_setup_sequence(d_data->res.server,
4217                         &d_data->args.seq_args,
4218                         &d_data->res.seq_res,
4219                         task);
4220 }
4221 #endif /* CONFIG_NFS_V4_1 */
4222
4223 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4224 #if defined(CONFIG_NFS_V4_1)
4225         .rpc_call_prepare = nfs4_delegreturn_prepare,
4226 #endif /* CONFIG_NFS_V4_1 */
4227         .rpc_call_done = nfs4_delegreturn_done,
4228         .rpc_release = nfs4_delegreturn_release,
4229 };
4230
4231 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4232 {
4233         struct nfs4_delegreturndata *data;
4234         struct nfs_server *server = NFS_SERVER(inode);
4235         struct rpc_task *task;
4236         struct rpc_message msg = {
4237                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4238                 .rpc_cred = cred,
4239         };
4240         struct rpc_task_setup task_setup_data = {
4241                 .rpc_client = server->client,
4242                 .rpc_message = &msg,
4243                 .callback_ops = &nfs4_delegreturn_ops,
4244                 .flags = RPC_TASK_ASYNC,
4245         };
4246         int status = 0;
4247
4248         data = kzalloc(sizeof(*data), GFP_NOFS);
4249         if (data == NULL)
4250                 return -ENOMEM;
4251         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4252         data->args.fhandle = &data->fh;
4253         data->args.stateid = &data->stateid;
4254         data->args.bitmask = server->cache_consistency_bitmask;
4255         nfs_copy_fh(&data->fh, NFS_FH(inode));
4256         nfs4_stateid_copy(&data->stateid, stateid);
4257         data->res.fattr = &data->fattr;
4258         data->res.server = server;
4259         nfs_fattr_init(data->res.fattr);
4260         data->timestamp = jiffies;
4261         data->rpc_status = 0;
4262
4263         task_setup_data.callback_data = data;
4264         msg.rpc_argp = &data->args;
4265         msg.rpc_resp = &data->res;
4266         task = rpc_run_task(&task_setup_data);
4267         if (IS_ERR(task))
4268                 return PTR_ERR(task);
4269         if (!issync)
4270                 goto out;
4271         status = nfs4_wait_for_completion_rpc_task(task);
4272         if (status != 0)
4273                 goto out;
4274         status = data->rpc_status;
4275         if (status == 0)
4276                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4277         else
4278                 nfs_refresh_inode(inode, &data->fattr);
4279 out:
4280         rpc_put_task(task);
4281         return status;
4282 }
4283
4284 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4285 {
4286         struct nfs_server *server = NFS_SERVER(inode);
4287         struct nfs4_exception exception = { };
4288         int err;
4289         do {
4290                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4291                 switch (err) {
4292                         case -NFS4ERR_STALE_STATEID:
4293                         case -NFS4ERR_EXPIRED:
4294                         case 0:
4295                                 return 0;
4296                 }
4297                 err = nfs4_handle_exception(server, err, &exception);
4298         } while (exception.retry);
4299         return err;
4300 }
4301
4302 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4303 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4304
4305 /* 
4306  * sleep, with exponential backoff, and retry the LOCK operation. 
4307  */
4308 static unsigned long
4309 nfs4_set_lock_task_retry(unsigned long timeout)
4310 {
4311         freezable_schedule_timeout_killable(timeout);
4312         timeout <<= 1;
4313         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4314                 return NFS4_LOCK_MAXTIMEOUT;
4315         return timeout;
4316 }
4317
4318 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4319 {
4320         struct inode *inode = state->inode;
4321         struct nfs_server *server = NFS_SERVER(inode);
4322         struct nfs_client *clp = server->nfs_client;
4323         struct nfs_lockt_args arg = {
4324                 .fh = NFS_FH(inode),
4325                 .fl = request,
4326         };
4327         struct nfs_lockt_res res = {
4328                 .denied = request,
4329         };
4330         struct rpc_message msg = {
4331                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4332                 .rpc_argp       = &arg,
4333                 .rpc_resp       = &res,
4334                 .rpc_cred       = state->owner->so_cred,
4335         };
4336         struct nfs4_lock_state *lsp;
4337         int status;
4338
4339         arg.lock_owner.clientid = clp->cl_clientid;
4340         status = nfs4_set_lock_state(state, request);
4341         if (status != 0)
4342                 goto out;
4343         lsp = request->fl_u.nfs4_fl.owner;
4344         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4345         arg.lock_owner.s_dev = server->s_dev;
4346         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4347         switch (status) {
4348                 case 0:
4349                         request->fl_type = F_UNLCK;
4350                         break;
4351                 case -NFS4ERR_DENIED:
4352                         status = 0;
4353         }
4354         request->fl_ops->fl_release_private(request);
4355 out:
4356         return status;
4357 }
4358
4359 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4360 {
4361         struct nfs4_exception exception = { };
4362         int err;
4363
4364         do {
4365                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4366                                 _nfs4_proc_getlk(state, cmd, request),
4367                                 &exception);
4368         } while (exception.retry);
4369         return err;
4370 }
4371
4372 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4373 {
4374         int res = 0;
4375         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4376                 case FL_POSIX:
4377                         res = posix_lock_file_wait(file, fl);
4378                         break;
4379                 case FL_FLOCK:
4380                         res = flock_lock_file_wait(file, fl);
4381                         break;
4382                 default:
4383                         BUG();
4384         }
4385         return res;
4386 }
4387
4388 struct nfs4_unlockdata {
4389         struct nfs_locku_args arg;
4390         struct nfs_locku_res res;
4391         struct nfs4_lock_state *lsp;
4392         struct nfs_open_context *ctx;
4393         struct file_lock fl;
4394         const struct nfs_server *server;
4395         unsigned long timestamp;
4396 };
4397
4398 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4399                 struct nfs_open_context *ctx,
4400                 struct nfs4_lock_state *lsp,
4401                 struct nfs_seqid *seqid)
4402 {
4403         struct nfs4_unlockdata *p;
4404         struct inode *inode = lsp->ls_state->inode;
4405
4406         p = kzalloc(sizeof(*p), GFP_NOFS);
4407         if (p == NULL)
4408                 return NULL;
4409         p->arg.fh = NFS_FH(inode);
4410         p->arg.fl = &p->fl;
4411         p->arg.seqid = seqid;
4412         p->res.seqid = seqid;
4413         p->arg.stateid = &lsp->ls_stateid;
4414         p->lsp = lsp;
4415         atomic_inc(&lsp->ls_count);
4416         /* Ensure we don't close file until we're done freeing locks! */
4417         p->ctx = get_nfs_open_context(ctx);
4418         memcpy(&p->fl, fl, sizeof(p->fl));
4419         p->server = NFS_SERVER(inode);
4420         return p;
4421 }
4422
4423 static void nfs4_locku_release_calldata(void *data)
4424 {
4425         struct nfs4_unlockdata *calldata = data;
4426         nfs_free_seqid(calldata->arg.seqid);
4427         nfs4_put_lock_state(calldata->lsp);
4428         put_nfs_open_context(calldata->ctx);
4429         kfree(calldata);
4430 }
4431
4432 static void nfs4_locku_done(struct rpc_task *task, void *data)
4433 {
4434         struct nfs4_unlockdata *calldata = data;
4435
4436         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4437                 return;
4438         switch (task->tk_status) {
4439                 case 0:
4440                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4441                                         &calldata->res.stateid);
4442                         renew_lease(calldata->server, calldata->timestamp);
4443                         break;
4444                 case -NFS4ERR_BAD_STATEID:
4445                 case -NFS4ERR_OLD_STATEID:
4446                 case -NFS4ERR_STALE_STATEID:
4447                 case -NFS4ERR_EXPIRED:
4448                         break;
4449                 default:
4450                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4451                                 rpc_restart_call_prepare(task);
4452         }
4453         nfs_release_seqid(calldata->arg.seqid);
4454 }
4455
4456 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4457 {
4458         struct nfs4_unlockdata *calldata = data;
4459
4460         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4461                 goto out_wait;
4462         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4463                 /* Note: exit _without_ running nfs4_locku_done */
4464                 goto out_no_action;
4465         }
4466         calldata->timestamp = jiffies;
4467         if (nfs4_setup_sequence(calldata->server,
4468                                 &calldata->arg.seq_args,
4469                                 &calldata->res.seq_res,
4470                                 task) != 0)
4471                 nfs_release_seqid(calldata->arg.seqid);
4472         return;
4473 out_no_action:
4474         task->tk_action = NULL;
4475 out_wait:
4476         nfs4_sequence_done(task, &calldata->res.seq_res);
4477 }
4478
4479 static const struct rpc_call_ops nfs4_locku_ops = {
4480         .rpc_call_prepare = nfs4_locku_prepare,
4481         .rpc_call_done = nfs4_locku_done,
4482         .rpc_release = nfs4_locku_release_calldata,
4483 };
4484
4485 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4486                 struct nfs_open_context *ctx,
4487                 struct nfs4_lock_state *lsp,
4488                 struct nfs_seqid *seqid)
4489 {
4490         struct nfs4_unlockdata *data;
4491         struct rpc_message msg = {
4492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4493                 .rpc_cred = ctx->cred,
4494         };
4495         struct rpc_task_setup task_setup_data = {
4496                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4497                 .rpc_message = &msg,
4498                 .callback_ops = &nfs4_locku_ops,
4499                 .workqueue = nfsiod_workqueue,
4500                 .flags = RPC_TASK_ASYNC,
4501         };
4502
4503         /* Ensure this is an unlock - when canceling a lock, the
4504          * canceled lock is passed in, and it won't be an unlock.
4505          */
4506         fl->fl_type = F_UNLCK;
4507
4508         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4509         if (data == NULL) {
4510                 nfs_free_seqid(seqid);
4511                 return ERR_PTR(-ENOMEM);
4512         }
4513
4514         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4515         msg.rpc_argp = &data->arg;
4516         msg.rpc_resp = &data->res;
4517         task_setup_data.callback_data = data;
4518         return rpc_run_task(&task_setup_data);
4519 }
4520
4521 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4522 {
4523         struct inode *inode = state->inode;
4524         struct nfs4_state_owner *sp = state->owner;
4525         struct nfs_inode *nfsi = NFS_I(inode);
4526         struct nfs_seqid *seqid;
4527         struct nfs4_lock_state *lsp;
4528         struct rpc_task *task;
4529         int status = 0;
4530         unsigned char fl_flags = request->fl_flags;
4531
4532         status = nfs4_set_lock_state(state, request);
4533         /* Unlock _before_ we do the RPC call */
4534         request->fl_flags |= FL_EXISTS;
4535         /* Exclude nfs_delegation_claim_locks() */
4536         mutex_lock(&sp->so_delegreturn_mutex);
4537         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
4538         down_read(&nfsi->rwsem);
4539         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4540                 up_read(&nfsi->rwsem);
4541                 mutex_unlock(&sp->so_delegreturn_mutex);
4542                 goto out;
4543         }
4544         up_read(&nfsi->rwsem);
4545         mutex_unlock(&sp->so_delegreturn_mutex);
4546         if (status != 0)
4547                 goto out;
4548         /* Is this a delegated lock? */
4549         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4550                 goto out;
4551         lsp = request->fl_u.nfs4_fl.owner;
4552         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4553         status = -ENOMEM;
4554         if (seqid == NULL)
4555                 goto out;
4556         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4557         status = PTR_ERR(task);
4558         if (IS_ERR(task))
4559                 goto out;
4560         status = nfs4_wait_for_completion_rpc_task(task);
4561         rpc_put_task(task);
4562 out:
4563         request->fl_flags = fl_flags;
4564         return status;
4565 }
4566
4567 struct nfs4_lockdata {
4568         struct nfs_lock_args arg;
4569         struct nfs_lock_res res;
4570         struct nfs4_lock_state *lsp;
4571         struct nfs_open_context *ctx;
4572         struct file_lock fl;
4573         unsigned long timestamp;
4574         int rpc_status;
4575         int cancelled;
4576         struct nfs_server *server;
4577 };
4578
4579 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4580                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4581                 gfp_t gfp_mask)
4582 {
4583         struct nfs4_lockdata *p;
4584         struct inode *inode = lsp->ls_state->inode;
4585         struct nfs_server *server = NFS_SERVER(inode);
4586
4587         p = kzalloc(sizeof(*p), gfp_mask);
4588         if (p == NULL)
4589                 return NULL;
4590
4591         p->arg.fh = NFS_FH(inode);
4592         p->arg.fl = &p->fl;
4593         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4594         if (p->arg.open_seqid == NULL)
4595                 goto out_free;
4596         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4597         if (p->arg.lock_seqid == NULL)
4598                 goto out_free_seqid;
4599         p->arg.lock_stateid = &lsp->ls_stateid;
4600         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4601         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4602         p->arg.lock_owner.s_dev = server->s_dev;
4603         p->res.lock_seqid = p->arg.lock_seqid;
4604         p->lsp = lsp;
4605         p->server = server;
4606         atomic_inc(&lsp->ls_count);
4607         p->ctx = get_nfs_open_context(ctx);
4608         memcpy(&p->fl, fl, sizeof(p->fl));
4609         return p;
4610 out_free_seqid:
4611         nfs_free_seqid(p->arg.open_seqid);
4612 out_free:
4613         kfree(p);
4614         return NULL;
4615 }
4616
4617 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4618 {
4619         struct nfs4_lockdata *data = calldata;
4620         struct nfs4_state *state = data->lsp->ls_state;
4621
4622         dprintk("%s: begin!\n", __func__);
4623         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4624                 goto out_wait;
4625         /* Do we need to do an open_to_lock_owner? */
4626         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4627                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
4628                         goto out_release_lock_seqid;
4629                 }
4630                 data->arg.open_stateid = &state->stateid;
4631                 data->arg.new_lock_owner = 1;
4632                 data->res.open_seqid = data->arg.open_seqid;
4633         } else
4634                 data->arg.new_lock_owner = 0;
4635         data->timestamp = jiffies;
4636         if (nfs4_setup_sequence(data->server,
4637                                 &data->arg.seq_args,
4638                                 &data->res.seq_res,
4639                                 task) == 0)
4640                 return;
4641         nfs_release_seqid(data->arg.open_seqid);
4642 out_release_lock_seqid:
4643         nfs_release_seqid(data->arg.lock_seqid);
4644 out_wait:
4645         nfs4_sequence_done(task, &data->res.seq_res);
4646         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4647 }
4648
4649 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4650 {
4651         struct nfs4_lockdata *data = calldata;
4652
4653         dprintk("%s: begin!\n", __func__);
4654
4655         if (!nfs4_sequence_done(task, &data->res.seq_res))
4656                 return;
4657
4658         data->rpc_status = task->tk_status;
4659         if (data->arg.new_lock_owner != 0) {
4660                 if (data->rpc_status == 0)
4661                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4662                 else
4663                         goto out;
4664         }
4665         if (data->rpc_status == 0) {
4666                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4667                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4668                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4669         }
4670 out:
4671         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4672 }
4673
4674 static void nfs4_lock_release(void *calldata)
4675 {
4676         struct nfs4_lockdata *data = calldata;
4677
4678         dprintk("%s: begin!\n", __func__);
4679         nfs_free_seqid(data->arg.open_seqid);
4680         if (data->cancelled != 0) {
4681                 struct rpc_task *task;
4682                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4683                                 data->arg.lock_seqid);
4684                 if (!IS_ERR(task))
4685                         rpc_put_task_async(task);
4686                 dprintk("%s: cancelling lock!\n", __func__);
4687         } else
4688                 nfs_free_seqid(data->arg.lock_seqid);
4689         nfs4_put_lock_state(data->lsp);
4690         put_nfs_open_context(data->ctx);
4691         kfree(data);
4692         dprintk("%s: done!\n", __func__);
4693 }
4694
4695 static const struct rpc_call_ops nfs4_lock_ops = {
4696         .rpc_call_prepare = nfs4_lock_prepare,
4697         .rpc_call_done = nfs4_lock_done,
4698         .rpc_release = nfs4_lock_release,
4699 };
4700
4701 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4702 {
4703         switch (error) {
4704         case -NFS4ERR_ADMIN_REVOKED:
4705         case -NFS4ERR_BAD_STATEID:
4706                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4707                 if (new_lock_owner != 0 ||
4708                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4709                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4710                 break;
4711         case -NFS4ERR_STALE_STATEID:
4712                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4713         case -NFS4ERR_EXPIRED:
4714                 nfs4_schedule_lease_recovery(server->nfs_client);
4715         };
4716 }
4717
4718 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4719 {
4720         struct nfs4_lockdata *data;
4721         struct rpc_task *task;
4722         struct rpc_message msg = {
4723                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4724                 .rpc_cred = state->owner->so_cred,
4725         };
4726         struct rpc_task_setup task_setup_data = {
4727                 .rpc_client = NFS_CLIENT(state->inode),
4728                 .rpc_message = &msg,
4729                 .callback_ops = &nfs4_lock_ops,
4730                 .workqueue = nfsiod_workqueue,
4731                 .flags = RPC_TASK_ASYNC,
4732         };
4733         int ret;
4734
4735         dprintk("%s: begin!\n", __func__);
4736         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4737                         fl->fl_u.nfs4_fl.owner,
4738                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4739         if (data == NULL)
4740                 return -ENOMEM;
4741         if (IS_SETLKW(cmd))
4742                 data->arg.block = 1;
4743         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4744         msg.rpc_argp = &data->arg;
4745         msg.rpc_resp = &data->res;
4746         task_setup_data.callback_data = data;
4747         if (recovery_type > NFS_LOCK_NEW) {
4748                 if (recovery_type == NFS_LOCK_RECLAIM)
4749                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4750                 nfs4_set_sequence_privileged(&data->arg.seq_args);
4751         }
4752         task = rpc_run_task(&task_setup_data);
4753         if (IS_ERR(task))
4754                 return PTR_ERR(task);
4755         ret = nfs4_wait_for_completion_rpc_task(task);
4756         if (ret == 0) {
4757                 ret = data->rpc_status;
4758                 if (ret)
4759                         nfs4_handle_setlk_error(data->server, data->lsp,
4760                                         data->arg.new_lock_owner, ret);
4761         } else
4762                 data->cancelled = 1;
4763         rpc_put_task(task);
4764         dprintk("%s: done, ret = %d!\n", __func__, ret);
4765         return ret;
4766 }
4767
4768 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4769 {
4770         struct nfs_server *server = NFS_SERVER(state->inode);
4771         struct nfs4_exception exception = {
4772                 .inode = state->inode,
4773         };
4774         int err;
4775
4776         do {
4777                 /* Cache the lock if possible... */
4778                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4779                         return 0;
4780                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4781                 if (err != -NFS4ERR_DELAY)
4782                         break;
4783                 nfs4_handle_exception(server, err, &exception);
4784         } while (exception.retry);
4785         return err;
4786 }
4787
4788 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4789 {
4790         struct nfs_server *server = NFS_SERVER(state->inode);
4791         struct nfs4_exception exception = {
4792                 .inode = state->inode,
4793         };
4794         int err;
4795
4796         err = nfs4_set_lock_state(state, request);
4797         if (err != 0)
4798                 return err;
4799         do {
4800                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4801                         return 0;
4802                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4803                 switch (err) {
4804                 default:
4805                         goto out;
4806                 case -NFS4ERR_GRACE:
4807                 case -NFS4ERR_DELAY:
4808                         nfs4_handle_exception(server, err, &exception);
4809                         err = 0;
4810                 }
4811         } while (exception.retry);
4812 out:
4813         return err;
4814 }
4815
4816 #if defined(CONFIG_NFS_V4_1)
4817 /**
4818  * nfs41_check_expired_locks - possibly free a lock stateid
4819  *
4820  * @state: NFSv4 state for an inode
4821  *
4822  * Returns NFS_OK if recovery for this stateid is now finished.
4823  * Otherwise a negative NFS4ERR value is returned.
4824  */
4825 static int nfs41_check_expired_locks(struct nfs4_state *state)
4826 {
4827         int status, ret = -NFS4ERR_BAD_STATEID;
4828         struct nfs4_lock_state *lsp;
4829         struct nfs_server *server = NFS_SERVER(state->inode);
4830
4831         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4832                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
4833                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4834                         if (status != NFS_OK) {
4835                                 /* Free the stateid unless the server
4836                                  * informs us the stateid is unrecognized. */
4837                                 if (status != -NFS4ERR_BAD_STATEID)
4838                                         nfs41_free_stateid(server,
4839                                                         &lsp->ls_stateid);
4840                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
4841                                 ret = status;
4842                         }
4843                 }
4844         };
4845
4846         return ret;
4847 }
4848
4849 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4850 {
4851         int status = NFS_OK;
4852
4853         if (test_bit(LK_STATE_IN_USE, &state->flags))
4854                 status = nfs41_check_expired_locks(state);
4855         if (status != NFS_OK)
4856                 status = nfs4_lock_expired(state, request);
4857         return status;
4858 }
4859 #endif
4860
4861 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4862 {
4863         struct nfs4_state_owner *sp = state->owner;
4864         struct nfs_inode *nfsi = NFS_I(state->inode);
4865         unsigned char fl_flags = request->fl_flags;
4866         unsigned int seq;
4867         int status = -ENOLCK;
4868
4869         if ((fl_flags & FL_POSIX) &&
4870                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4871                 goto out;
4872         /* Is this a delegated open? */
4873         status = nfs4_set_lock_state(state, request);
4874         if (status != 0)
4875                 goto out;
4876         request->fl_flags |= FL_ACCESS;
4877         status = do_vfs_lock(request->fl_file, request);
4878         if (status < 0)
4879                 goto out;
4880         down_read(&nfsi->rwsem);
4881         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4882                 /* Yes: cache locks! */
4883                 /* ...but avoid races with delegation recall... */
4884                 request->fl_flags = fl_flags & ~FL_SLEEP;
4885                 status = do_vfs_lock(request->fl_file, request);
4886                 goto out_unlock;
4887         }
4888         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
4889         up_read(&nfsi->rwsem);
4890         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4891         if (status != 0)
4892                 goto out;
4893         down_read(&nfsi->rwsem);
4894         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
4895                 status = -NFS4ERR_DELAY;
4896                 goto out_unlock;
4897         }
4898         /* Note: we always want to sleep here! */
4899         request->fl_flags = fl_flags | FL_SLEEP;
4900         if (do_vfs_lock(request->fl_file, request) < 0)
4901                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4902                         "manager!\n", __func__);
4903 out_unlock:
4904         up_read(&nfsi->rwsem);
4905 out:
4906         request->fl_flags = fl_flags;
4907         return status;
4908 }
4909
4910 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4911 {
4912         struct nfs4_exception exception = {
4913                 .state = state,
4914                 .inode = state->inode,
4915         };
4916         int err;
4917
4918         do {
4919                 err = _nfs4_proc_setlk(state, cmd, request);
4920                 if (err == -NFS4ERR_DENIED)
4921                         err = -EAGAIN;
4922                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4923                                 err, &exception);
4924         } while (exception.retry);
4925         return err;
4926 }
4927
4928 static int
4929 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4930 {
4931         struct nfs_open_context *ctx;
4932         struct nfs4_state *state;
4933         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4934         int status;
4935
4936         /* verify open state */
4937         ctx = nfs_file_open_context(filp);
4938         state = ctx->state;
4939
4940         if (request->fl_start < 0 || request->fl_end < 0)
4941                 return -EINVAL;
4942
4943         if (IS_GETLK(cmd)) {
4944                 if (state != NULL)
4945                         return nfs4_proc_getlk(state, F_GETLK, request);
4946                 return 0;
4947         }
4948
4949         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4950                 return -EINVAL;
4951
4952         if (request->fl_type == F_UNLCK) {
4953                 if (state != NULL)
4954                         return nfs4_proc_unlck(state, cmd, request);
4955                 return 0;
4956         }
4957
4958         if (state == NULL)
4959                 return -ENOLCK;
4960         /*
4961          * Don't rely on the VFS having checked the file open mode,
4962          * since it won't do this for flock() locks.
4963          */
4964         switch (request->fl_type) {
4965         case F_RDLCK:
4966                 if (!(filp->f_mode & FMODE_READ))
4967                         return -EBADF;
4968                 break;
4969         case F_WRLCK:
4970                 if (!(filp->f_mode & FMODE_WRITE))
4971                         return -EBADF;
4972         }
4973
4974         do {
4975                 status = nfs4_proc_setlk(state, cmd, request);
4976                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4977                         break;
4978                 timeout = nfs4_set_lock_task_retry(timeout);
4979                 status = -ERESTARTSYS;
4980                 if (signalled())
4981                         break;
4982         } while(status < 0);
4983         return status;
4984 }
4985
4986 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4987 {
4988         struct nfs_server *server = NFS_SERVER(state->inode);
4989         struct nfs4_exception exception = { };
4990         int err;
4991
4992         err = nfs4_set_lock_state(state, fl);
4993         if (err != 0)
4994                 goto out;
4995         do {
4996                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4997                 switch (err) {
4998                         default:
4999                                 printk(KERN_ERR "NFS: %s: unhandled error "
5000                                         "%d.\n", __func__, err);
5001                         case 0:
5002                         case -ESTALE:
5003                                 goto out;
5004                         case -NFS4ERR_STALE_CLIENTID:
5005                         case -NFS4ERR_STALE_STATEID:
5006                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
5007                         case -NFS4ERR_EXPIRED:
5008                                 nfs4_schedule_lease_recovery(server->nfs_client);
5009                                 err = -EAGAIN;
5010                                 goto out;
5011                         case -NFS4ERR_BADSESSION:
5012                         case -NFS4ERR_BADSLOT:
5013                         case -NFS4ERR_BAD_HIGH_SLOT:
5014                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
5015                         case -NFS4ERR_DEADSESSION:
5016                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
5017                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
5018                                 err = -EAGAIN;
5019                                 goto out;
5020                         case -NFS4ERR_DELEG_REVOKED:
5021                         case -NFS4ERR_ADMIN_REVOKED:
5022                         case -NFS4ERR_BAD_STATEID:
5023                         case -NFS4ERR_OPENMODE:
5024                                 nfs4_schedule_stateid_recovery(server, state);
5025                                 err = 0;
5026                                 goto out;
5027                         case -ENOMEM:
5028                         case -NFS4ERR_DENIED:
5029                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
5030                                 err = 0;
5031                                 goto out;
5032                 }
5033                 set_bit(NFS_DELEGATED_STATE, &state->flags);
5034                 err = nfs4_handle_exception(server, err, &exception);
5035         } while (exception.retry);
5036 out:
5037         return err;
5038 }
5039
5040 struct nfs_release_lockowner_data {
5041         struct nfs4_lock_state *lsp;
5042         struct nfs_server *server;
5043         struct nfs_release_lockowner_args args;
5044 };
5045
5046 static void nfs4_release_lockowner_release(void *calldata)
5047 {
5048         struct nfs_release_lockowner_data *data = calldata;
5049         nfs4_free_lock_state(data->server, data->lsp);
5050         kfree(calldata);
5051 }
5052
5053 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5054         .rpc_release = nfs4_release_lockowner_release,
5055 };
5056
5057 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
5058 {
5059         struct nfs_server *server = lsp->ls_state->owner->so_server;
5060         struct nfs_release_lockowner_data *data;
5061         struct rpc_message msg = {
5062                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5063         };
5064
5065         if (server->nfs_client->cl_mvops->minor_version != 0)
5066                 return -EINVAL;
5067         data = kmalloc(sizeof(*data), GFP_NOFS);
5068         if (!data)
5069                 return -ENOMEM;
5070         data->lsp = lsp;
5071         data->server = server;
5072         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5073         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5074         data->args.lock_owner.s_dev = server->s_dev;
5075         msg.rpc_argp = &data->args;
5076         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5077         return 0;
5078 }
5079
5080 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5081
5082 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5083                                    const void *buf, size_t buflen,
5084                                    int flags, int type)
5085 {
5086         if (strcmp(key, "") != 0)
5087                 return -EINVAL;
5088
5089         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5090 }
5091
5092 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5093                                    void *buf, size_t buflen, int type)
5094 {
5095         if (strcmp(key, "") != 0)
5096                 return -EINVAL;
5097
5098         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5099 }
5100
5101 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5102                                        size_t list_len, const char *name,
5103                                        size_t name_len, int type)
5104 {
5105         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5106
5107         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5108                 return 0;
5109
5110         if (list && len <= list_len)
5111                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5112         return len;
5113 }
5114
5115 /*
5116  * nfs_fhget will use either the mounted_on_fileid or the fileid
5117  */
5118 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5119 {
5120         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5121                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5122               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5123               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5124                 return;
5125
5126         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5127                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5128         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5129         fattr->nlink = 2;
5130 }
5131
5132 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5133                                    const struct qstr *name,
5134                                    struct nfs4_fs_locations *fs_locations,
5135                                    struct page *page)
5136 {
5137         struct nfs_server *server = NFS_SERVER(dir);
5138         u32 bitmask[2] = {
5139                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5140         };
5141         struct nfs4_fs_locations_arg args = {
5142                 .dir_fh = NFS_FH(dir),
5143                 .name = name,
5144                 .page = page,
5145                 .bitmask = bitmask,
5146         };
5147         struct nfs4_fs_locations_res res = {
5148                 .fs_locations = fs_locations,
5149         };
5150         struct rpc_message msg = {
5151                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5152                 .rpc_argp = &args,
5153                 .rpc_resp = &res,
5154         };
5155         int status;
5156
5157         dprintk("%s: start\n", __func__);
5158
5159         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5160          * is not supported */
5161         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5162                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5163         else
5164                 bitmask[0] |= FATTR4_WORD0_FILEID;
5165
5166         nfs_fattr_init(&fs_locations->fattr);
5167         fs_locations->server = server;
5168         fs_locations->nlocations = 0;
5169         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5170         dprintk("%s: returned status = %d\n", __func__, status);
5171         return status;
5172 }
5173
5174 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5175                            const struct qstr *name,
5176                            struct nfs4_fs_locations *fs_locations,
5177                            struct page *page)
5178 {
5179         struct nfs4_exception exception = { };
5180         int err;
5181         do {
5182                 err = nfs4_handle_exception(NFS_SERVER(dir),
5183                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5184                                 &exception);
5185         } while (exception.retry);
5186         return err;
5187 }
5188
5189 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5190 {
5191         int status;
5192         struct nfs4_secinfo_arg args = {
5193                 .dir_fh = NFS_FH(dir),
5194                 .name   = name,
5195         };
5196         struct nfs4_secinfo_res res = {
5197                 .flavors     = flavors,
5198         };
5199         struct rpc_message msg = {
5200                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5201                 .rpc_argp = &args,
5202                 .rpc_resp = &res,
5203         };
5204
5205         dprintk("NFS call  secinfo %s\n", name->name);
5206         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5207         dprintk("NFS reply  secinfo: %d\n", status);
5208         return status;
5209 }
5210
5211 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5212                       struct nfs4_secinfo_flavors *flavors)
5213 {
5214         struct nfs4_exception exception = { };
5215         int err;
5216         do {
5217                 err = nfs4_handle_exception(NFS_SERVER(dir),
5218                                 _nfs4_proc_secinfo(dir, name, flavors),
5219                                 &exception);
5220         } while (exception.retry);
5221         return err;
5222 }
5223
5224 #ifdef CONFIG_NFS_V4_1
5225 /*
5226  * Check the exchange flags returned by the server for invalid flags, having
5227  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5228  * DS flags set.
5229  */
5230 static int nfs4_check_cl_exchange_flags(u32 flags)
5231 {
5232         if (flags & ~EXCHGID4_FLAG_MASK_R)
5233                 goto out_inval;
5234         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5235             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5236                 goto out_inval;
5237         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5238                 goto out_inval;
5239         return NFS_OK;
5240 out_inval:
5241         return -NFS4ERR_INVAL;
5242 }
5243
5244 static bool
5245 nfs41_same_server_scope(struct nfs41_server_scope *a,
5246                         struct nfs41_server_scope *b)
5247 {
5248         if (a->server_scope_sz == b->server_scope_sz &&
5249             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5250                 return true;
5251
5252         return false;
5253 }
5254
5255 /*
5256  * nfs4_proc_bind_conn_to_session()
5257  *
5258  * The 4.1 client currently uses the same TCP connection for the
5259  * fore and backchannel.
5260  */
5261 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5262 {
5263         int status;
5264         struct nfs41_bind_conn_to_session_res res;
5265         struct rpc_message msg = {
5266                 .rpc_proc =
5267                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5268                 .rpc_argp = clp,
5269                 .rpc_resp = &res,
5270                 .rpc_cred = cred,
5271         };
5272
5273         dprintk("--> %s\n", __func__);
5274
5275         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5276         if (unlikely(res.session == NULL)) {
5277                 status = -ENOMEM;
5278                 goto out;
5279         }
5280
5281         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5282         if (status == 0) {
5283                 if (memcmp(res.session->sess_id.data,
5284                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5285                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5286                         status = -EIO;
5287                         goto out_session;
5288                 }
5289                 if (res.dir != NFS4_CDFS4_BOTH) {
5290                         dprintk("NFS: %s: Unexpected direction from server\n",
5291                                 __func__);
5292                         status = -EIO;
5293                         goto out_session;
5294                 }
5295                 if (res.use_conn_in_rdma_mode) {
5296                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5297                                 __func__);
5298                         status = -EIO;
5299                         goto out_session;
5300                 }
5301         }
5302 out_session:
5303         kfree(res.session);
5304 out:
5305         dprintk("<-- %s status= %d\n", __func__, status);
5306         return status;
5307 }
5308
5309 /*
5310  * nfs4_proc_exchange_id()
5311  *
5312  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5313  *
5314  * Since the clientid has expired, all compounds using sessions
5315  * associated with the stale clientid will be returning
5316  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5317  * be in some phase of session reset.
5318  */
5319 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5320 {
5321         nfs4_verifier verifier;
5322         struct nfs41_exchange_id_args args = {
5323                 .verifier = &verifier,
5324                 .client = clp,
5325                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5326         };
5327         struct nfs41_exchange_id_res res = {
5328                 0
5329         };
5330         int status;
5331         struct rpc_message msg = {
5332                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5333                 .rpc_argp = &args,
5334                 .rpc_resp = &res,
5335                 .rpc_cred = cred,
5336         };
5337
5338         nfs4_init_boot_verifier(clp, &verifier);
5339         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5340                                                         sizeof(args.id));
5341         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5342                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5343                 args.id_len, args.id);
5344
5345         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5346                                         GFP_NOFS);
5347         if (unlikely(res.server_owner == NULL)) {
5348                 status = -ENOMEM;
5349                 goto out;
5350         }
5351
5352         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5353                                         GFP_NOFS);
5354         if (unlikely(res.server_scope == NULL)) {
5355                 status = -ENOMEM;
5356                 goto out_server_owner;
5357         }
5358
5359         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5360         if (unlikely(res.impl_id == NULL)) {
5361                 status = -ENOMEM;
5362                 goto out_server_scope;
5363         }
5364
5365         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5366         if (status == 0)
5367                 status = nfs4_check_cl_exchange_flags(res.flags);
5368
5369         if (status == 0) {
5370                 clp->cl_clientid = res.clientid;
5371                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5372                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5373                         clp->cl_seqid = res.seqid;
5374
5375                 kfree(clp->cl_serverowner);
5376                 clp->cl_serverowner = res.server_owner;
5377                 res.server_owner = NULL;
5378
5379                 /* use the most recent implementation id */
5380                 kfree(clp->cl_implid);
5381                 clp->cl_implid = res.impl_id;
5382
5383                 if (clp->cl_serverscope != NULL &&
5384                     !nfs41_same_server_scope(clp->cl_serverscope,
5385                                              res.server_scope)) {
5386                         dprintk("%s: server_scope mismatch detected\n",
5387                                 __func__);
5388                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5389                         kfree(clp->cl_serverscope);
5390                         clp->cl_serverscope = NULL;
5391                 }
5392
5393                 if (clp->cl_serverscope == NULL) {
5394                         clp->cl_serverscope = res.server_scope;
5395                         goto out;
5396                 }
5397         } else
5398                 kfree(res.impl_id);
5399
5400 out_server_owner:
5401         kfree(res.server_owner);
5402 out_server_scope:
5403         kfree(res.server_scope);
5404 out:
5405         if (clp->cl_implid != NULL)
5406                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5407                         "domain: %s, name: %s, date: %llu,%u\n",
5408                         clp->cl_implid->domain, clp->cl_implid->name,
5409                         clp->cl_implid->date.seconds,
5410                         clp->cl_implid->date.nseconds);
5411         dprintk("NFS reply exchange_id: %d\n", status);
5412         return status;
5413 }
5414
5415 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5416                 struct rpc_cred *cred)
5417 {
5418         struct rpc_message msg = {
5419                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5420                 .rpc_argp = clp,
5421                 .rpc_cred = cred,
5422         };
5423         int status;
5424
5425         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5426         if (status)
5427                 dprintk("NFS: Got error %d from the server %s on "
5428                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5429         return status;
5430 }
5431
5432 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5433                 struct rpc_cred *cred)
5434 {
5435         unsigned int loop;
5436         int ret;
5437
5438         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5439                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5440                 switch (ret) {
5441                 case -NFS4ERR_DELAY:
5442                 case -NFS4ERR_CLIENTID_BUSY:
5443                         ssleep(1);
5444                         break;
5445                 default:
5446                         return ret;
5447                 }
5448         }
5449         return 0;
5450 }
5451
5452 int nfs4_destroy_clientid(struct nfs_client *clp)
5453 {
5454         struct rpc_cred *cred;
5455         int ret = 0;
5456
5457         if (clp->cl_mvops->minor_version < 1)
5458                 goto out;
5459         if (clp->cl_exchange_flags == 0)
5460                 goto out;
5461         if (clp->cl_preserve_clid)
5462                 goto out;
5463         cred = nfs4_get_exchange_id_cred(clp);
5464         ret = nfs4_proc_destroy_clientid(clp, cred);
5465         if (cred)
5466                 put_rpccred(cred);
5467         switch (ret) {
5468         case 0:
5469         case -NFS4ERR_STALE_CLIENTID:
5470                 clp->cl_exchange_flags = 0;
5471         }
5472 out:
5473         return ret;
5474 }
5475
5476 struct nfs4_get_lease_time_data {
5477         struct nfs4_get_lease_time_args *args;
5478         struct nfs4_get_lease_time_res *res;
5479         struct nfs_client *clp;
5480 };
5481
5482 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5483                                         void *calldata)
5484 {
5485         struct nfs4_get_lease_time_data *data =
5486                         (struct nfs4_get_lease_time_data *)calldata;
5487
5488         dprintk("--> %s\n", __func__);
5489         /* just setup sequence, do not trigger session recovery
5490            since we're invoked within one */
5491         nfs41_setup_sequence(data->clp->cl_session,
5492                         &data->args->la_seq_args,
5493                         &data->res->lr_seq_res,
5494                         task);
5495         dprintk("<-- %s\n", __func__);
5496 }
5497
5498 /*
5499  * Called from nfs4_state_manager thread for session setup, so don't recover
5500  * from sequence operation or clientid errors.
5501  */
5502 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5503 {
5504         struct nfs4_get_lease_time_data *data =
5505                         (struct nfs4_get_lease_time_data *)calldata;
5506
5507         dprintk("--> %s\n", __func__);
5508         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5509                 return;
5510         switch (task->tk_status) {
5511         case -NFS4ERR_DELAY:
5512         case -NFS4ERR_GRACE:
5513                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5514                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5515                 task->tk_status = 0;
5516                 /* fall through */
5517         case -NFS4ERR_RETRY_UNCACHED_REP:
5518                 rpc_restart_call_prepare(task);
5519                 return;
5520         }
5521         dprintk("<-- %s\n", __func__);
5522 }
5523
5524 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5525         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5526         .rpc_call_done = nfs4_get_lease_time_done,
5527 };
5528
5529 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5530 {
5531         struct rpc_task *task;
5532         struct nfs4_get_lease_time_args args;
5533         struct nfs4_get_lease_time_res res = {
5534                 .lr_fsinfo = fsinfo,
5535         };
5536         struct nfs4_get_lease_time_data data = {
5537                 .args = &args,
5538                 .res = &res,
5539                 .clp = clp,
5540         };
5541         struct rpc_message msg = {
5542                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5543                 .rpc_argp = &args,
5544                 .rpc_resp = &res,
5545         };
5546         struct rpc_task_setup task_setup = {
5547                 .rpc_client = clp->cl_rpcclient,
5548                 .rpc_message = &msg,
5549                 .callback_ops = &nfs4_get_lease_time_ops,
5550                 .callback_data = &data,
5551                 .flags = RPC_TASK_TIMEOUT,
5552         };
5553         int status;
5554
5555         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5556         nfs4_set_sequence_privileged(&args.la_seq_args);
5557         dprintk("--> %s\n", __func__);
5558         task = rpc_run_task(&task_setup);
5559
5560         if (IS_ERR(task))
5561                 status = PTR_ERR(task);
5562         else {
5563                 status = task->tk_status;
5564                 rpc_put_task(task);
5565         }
5566         dprintk("<-- %s return %d\n", __func__, status);
5567
5568         return status;
5569 }
5570
5571 /*
5572  * Initialize the values to be used by the client in CREATE_SESSION
5573  * If nfs4_init_session set the fore channel request and response sizes,
5574  * use them.
5575  *
5576  * Set the back channel max_resp_sz_cached to zero to force the client to
5577  * always set csa_cachethis to FALSE because the current implementation
5578  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5579  */
5580 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5581 {
5582         struct nfs4_session *session = args->client->cl_session;
5583         unsigned int mxrqst_sz = session->fc_target_max_rqst_sz,
5584                      mxresp_sz = session->fc_target_max_resp_sz;
5585
5586         if (mxrqst_sz == 0)
5587                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5588         if (mxresp_sz == 0)
5589                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5590         /* Fore channel attributes */
5591         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5592         args->fc_attrs.max_resp_sz = mxresp_sz;
5593         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5594         args->fc_attrs.max_reqs = max_session_slots;
5595
5596         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5597                 "max_ops=%u max_reqs=%u\n",
5598                 __func__,
5599                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5600                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5601
5602         /* Back channel attributes */
5603         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5604         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5605         args->bc_attrs.max_resp_sz_cached = 0;
5606         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5607         args->bc_attrs.max_reqs = 1;
5608
5609         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5610                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5611                 __func__,
5612                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5613                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5614                 args->bc_attrs.max_reqs);
5615 }
5616
5617 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5618 {
5619         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5620         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5621
5622         if (rcvd->max_resp_sz > sent->max_resp_sz)
5623                 return -EINVAL;
5624         /*
5625          * Our requested max_ops is the minimum we need; we're not
5626          * prepared to break up compounds into smaller pieces than that.
5627          * So, no point even trying to continue if the server won't
5628          * cooperate:
5629          */
5630         if (rcvd->max_ops < sent->max_ops)
5631                 return -EINVAL;
5632         if (rcvd->max_reqs == 0)
5633                 return -EINVAL;
5634         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5635                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5636         return 0;
5637 }
5638
5639 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5640 {
5641         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5642         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5643
5644         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5645                 return -EINVAL;
5646         if (rcvd->max_resp_sz < sent->max_resp_sz)
5647                 return -EINVAL;
5648         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5649                 return -EINVAL;
5650         /* These would render the backchannel useless: */
5651         if (rcvd->max_ops != sent->max_ops)
5652                 return -EINVAL;
5653         if (rcvd->max_reqs != sent->max_reqs)
5654                 return -EINVAL;
5655         return 0;
5656 }
5657
5658 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5659                                      struct nfs4_session *session)
5660 {
5661         int ret;
5662
5663         ret = nfs4_verify_fore_channel_attrs(args, session);
5664         if (ret)
5665                 return ret;
5666         return nfs4_verify_back_channel_attrs(args, session);
5667 }
5668
5669 static int _nfs4_proc_create_session(struct nfs_client *clp,
5670                 struct rpc_cred *cred)
5671 {
5672         struct nfs4_session *session = clp->cl_session;
5673         struct nfs41_create_session_args args = {
5674                 .client = clp,
5675                 .cb_program = NFS4_CALLBACK,
5676         };
5677         struct nfs41_create_session_res res = {
5678                 .client = clp,
5679         };
5680         struct rpc_message msg = {
5681                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5682                 .rpc_argp = &args,
5683                 .rpc_resp = &res,
5684                 .rpc_cred = cred,
5685         };
5686         int status;
5687
5688         nfs4_init_channel_attrs(&args);
5689         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5690
5691         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5692
5693         if (!status) {
5694                 /* Verify the session's negotiated channel_attrs values */
5695                 status = nfs4_verify_channel_attrs(&args, session);
5696                 /* Increment the clientid slot sequence id */
5697                 clp->cl_seqid++;
5698         }
5699
5700         return status;
5701 }
5702
5703 /*
5704  * Issues a CREATE_SESSION operation to the server.
5705  * It is the responsibility of the caller to verify the session is
5706  * expired before calling this routine.
5707  */
5708 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5709 {
5710         int status;
5711         unsigned *ptr;
5712         struct nfs4_session *session = clp->cl_session;
5713
5714         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5715
5716         status = _nfs4_proc_create_session(clp, cred);
5717         if (status)
5718                 goto out;
5719
5720         /* Init or reset the session slot tables */
5721         status = nfs4_setup_session_slot_tables(session);
5722         dprintk("slot table setup returned %d\n", status);
5723         if (status)
5724                 goto out;
5725
5726         ptr = (unsigned *)&session->sess_id.data[0];
5727         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5728                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5729 out:
5730         dprintk("<-- %s\n", __func__);
5731         return status;
5732 }
5733
5734 /*
5735  * Issue the over-the-wire RPC DESTROY_SESSION.
5736  * The caller must serialize access to this routine.
5737  */
5738 int nfs4_proc_destroy_session(struct nfs4_session *session,
5739                 struct rpc_cred *cred)
5740 {
5741         struct rpc_message msg = {
5742                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5743                 .rpc_argp = session,
5744                 .rpc_cred = cred,
5745         };
5746         int status = 0;
5747
5748         dprintk("--> nfs4_proc_destroy_session\n");
5749
5750         /* session is still being setup */
5751         if (session->clp->cl_cons_state != NFS_CS_READY)
5752                 return status;
5753
5754         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5755
5756         if (status)
5757                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5758                         "Session has been destroyed regardless...\n", status);
5759
5760         dprintk("<-- nfs4_proc_destroy_session\n");
5761         return status;
5762 }
5763
5764 /*
5765  * Renew the cl_session lease.
5766  */
5767 struct nfs4_sequence_data {
5768         struct nfs_client *clp;
5769         struct nfs4_sequence_args args;
5770         struct nfs4_sequence_res res;
5771 };
5772
5773 static void nfs41_sequence_release(void *data)
5774 {
5775         struct nfs4_sequence_data *calldata = data;
5776         struct nfs_client *clp = calldata->clp;
5777
5778         if (atomic_read(&clp->cl_count) > 1)
5779                 nfs4_schedule_state_renewal(clp);
5780         nfs_put_client(clp);
5781         kfree(calldata);
5782 }
5783
5784 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5785 {
5786         switch(task->tk_status) {
5787         case -NFS4ERR_DELAY:
5788                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5789                 return -EAGAIN;
5790         default:
5791                 nfs4_schedule_lease_recovery(clp);
5792         }
5793         return 0;
5794 }
5795
5796 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5797 {
5798         struct nfs4_sequence_data *calldata = data;
5799         struct nfs_client *clp = calldata->clp;
5800
5801         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5802                 return;
5803
5804         if (task->tk_status < 0) {
5805                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5806                 if (atomic_read(&clp->cl_count) == 1)
5807                         goto out;
5808
5809                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5810                         rpc_restart_call_prepare(task);
5811                         return;
5812                 }
5813         }
5814         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5815 out:
5816         dprintk("<-- %s\n", __func__);
5817 }
5818
5819 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5820 {
5821         struct nfs4_sequence_data *calldata = data;
5822         struct nfs_client *clp = calldata->clp;
5823         struct nfs4_sequence_args *args;
5824         struct nfs4_sequence_res *res;
5825
5826         args = task->tk_msg.rpc_argp;
5827         res = task->tk_msg.rpc_resp;
5828
5829         nfs41_setup_sequence(clp->cl_session, args, res, task);
5830 }
5831
5832 static const struct rpc_call_ops nfs41_sequence_ops = {
5833         .rpc_call_done = nfs41_sequence_call_done,
5834         .rpc_call_prepare = nfs41_sequence_prepare,
5835         .rpc_release = nfs41_sequence_release,
5836 };
5837
5838 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
5839                 struct rpc_cred *cred,
5840                 bool is_privileged)
5841 {
5842         struct nfs4_sequence_data *calldata;
5843         struct rpc_message msg = {
5844                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5845                 .rpc_cred = cred,
5846         };
5847         struct rpc_task_setup task_setup_data = {
5848                 .rpc_client = clp->cl_rpcclient,
5849                 .rpc_message = &msg,
5850                 .callback_ops = &nfs41_sequence_ops,
5851                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5852         };
5853
5854         if (!atomic_inc_not_zero(&clp->cl_count))
5855                 return ERR_PTR(-EIO);
5856         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5857         if (calldata == NULL) {
5858                 nfs_put_client(clp);
5859                 return ERR_PTR(-ENOMEM);
5860         }
5861         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5862         if (is_privileged)
5863                 nfs4_set_sequence_privileged(&calldata->args);
5864         msg.rpc_argp = &calldata->args;
5865         msg.rpc_resp = &calldata->res;
5866         calldata->clp = clp;
5867         task_setup_data.callback_data = calldata;
5868
5869         return rpc_run_task(&task_setup_data);
5870 }
5871
5872 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5873 {
5874         struct rpc_task *task;
5875         int ret = 0;
5876
5877         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5878                 return 0;
5879         task = _nfs41_proc_sequence(clp, cred, false);
5880         if (IS_ERR(task))
5881                 ret = PTR_ERR(task);
5882         else
5883                 rpc_put_task_async(task);
5884         dprintk("<-- %s status=%d\n", __func__, ret);
5885         return ret;
5886 }
5887
5888 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5889 {
5890         struct rpc_task *task;
5891         int ret;
5892
5893         task = _nfs41_proc_sequence(clp, cred, true);
5894         if (IS_ERR(task)) {
5895                 ret = PTR_ERR(task);
5896                 goto out;
5897         }
5898         ret = rpc_wait_for_completion_task(task);
5899         if (!ret) {
5900                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5901
5902                 if (task->tk_status == 0)
5903                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5904                 ret = task->tk_status;
5905         }
5906         rpc_put_task(task);
5907 out:
5908         dprintk("<-- %s status=%d\n", __func__, ret);
5909         return ret;
5910 }
5911
5912 struct nfs4_reclaim_complete_data {
5913         struct nfs_client *clp;
5914         struct nfs41_reclaim_complete_args arg;
5915         struct nfs41_reclaim_complete_res res;
5916 };
5917
5918 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5919 {
5920         struct nfs4_reclaim_complete_data *calldata = data;
5921
5922         nfs41_setup_sequence(calldata->clp->cl_session,
5923                         &calldata->arg.seq_args,
5924                         &calldata->res.seq_res,
5925                         task);
5926 }
5927
5928 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5929 {
5930         switch(task->tk_status) {
5931         case 0:
5932         case -NFS4ERR_COMPLETE_ALREADY:
5933         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5934                 break;
5935         case -NFS4ERR_DELAY:
5936                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5937                 /* fall through */
5938         case -NFS4ERR_RETRY_UNCACHED_REP:
5939                 return -EAGAIN;
5940         default:
5941                 nfs4_schedule_lease_recovery(clp);
5942         }
5943         return 0;
5944 }
5945
5946 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5947 {
5948         struct nfs4_reclaim_complete_data *calldata = data;
5949         struct nfs_client *clp = calldata->clp;
5950         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5951
5952         dprintk("--> %s\n", __func__);
5953         if (!nfs41_sequence_done(task, res))
5954                 return;
5955
5956         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5957                 rpc_restart_call_prepare(task);
5958                 return;
5959         }
5960         dprintk("<-- %s\n", __func__);
5961 }
5962
5963 static void nfs4_free_reclaim_complete_data(void *data)
5964 {
5965         struct nfs4_reclaim_complete_data *calldata = data;
5966
5967         kfree(calldata);
5968 }
5969
5970 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5971         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5972         .rpc_call_done = nfs4_reclaim_complete_done,
5973         .rpc_release = nfs4_free_reclaim_complete_data,
5974 };
5975
5976 /*
5977  * Issue a global reclaim complete.
5978  */
5979 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5980 {
5981         struct nfs4_reclaim_complete_data *calldata;
5982         struct rpc_task *task;
5983         struct rpc_message msg = {
5984                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5985         };
5986         struct rpc_task_setup task_setup_data = {
5987                 .rpc_client = clp->cl_rpcclient,
5988                 .rpc_message = &msg,
5989                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5990                 .flags = RPC_TASK_ASYNC,
5991         };
5992         int status = -ENOMEM;
5993
5994         dprintk("--> %s\n", __func__);
5995         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5996         if (calldata == NULL)
5997                 goto out;
5998         calldata->clp = clp;
5999         calldata->arg.one_fs = 0;
6000
6001         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6002         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
6003         msg.rpc_argp = &calldata->arg;
6004         msg.rpc_resp = &calldata->res;
6005         task_setup_data.callback_data = calldata;
6006         task = rpc_run_task(&task_setup_data);
6007         if (IS_ERR(task)) {
6008                 status = PTR_ERR(task);
6009                 goto out;
6010         }
6011         status = nfs4_wait_for_completion_rpc_task(task);
6012         if (status == 0)
6013                 status = task->tk_status;
6014         rpc_put_task(task);
6015         return 0;
6016 out:
6017         dprintk("<-- %s status=%d\n", __func__, status);
6018         return status;
6019 }
6020
6021 static void
6022 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6023 {
6024         struct nfs4_layoutget *lgp = calldata;
6025         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6026         struct nfs4_session *session = nfs4_get_session(server);
6027
6028         dprintk("--> %s\n", __func__);
6029         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6030          * right now covering the LAYOUTGET we are about to send.
6031          * However, that is not so catastrophic, and there seems
6032          * to be no way to prevent it completely.
6033          */
6034         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
6035                                 &lgp->res.seq_res, task))
6036                 return;
6037         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6038                                           NFS_I(lgp->args.inode)->layout,
6039                                           lgp->args.ctx->state)) {
6040                 rpc_exit(task, NFS4_OK);
6041         }
6042 }
6043
6044 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6045 {
6046         struct nfs4_layoutget *lgp = calldata;
6047         struct inode *inode = lgp->args.inode;
6048         struct nfs_server *server = NFS_SERVER(inode);
6049         struct pnfs_layout_hdr *lo;
6050         struct nfs4_state *state = NULL;
6051         unsigned long timeo, giveup;
6052
6053         dprintk("--> %s\n", __func__);
6054
6055         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
6056                 goto out;
6057
6058         switch (task->tk_status) {
6059         case 0:
6060                 goto out;
6061         case -NFS4ERR_LAYOUTTRYLATER:
6062         case -NFS4ERR_RECALLCONFLICT:
6063                 timeo = rpc_get_timeout(task->tk_client);
6064                 giveup = lgp->args.timestamp + timeo;
6065                 if (time_after(giveup, jiffies))
6066                         task->tk_status = -NFS4ERR_DELAY;
6067                 break;
6068         case -NFS4ERR_EXPIRED:
6069         case -NFS4ERR_BAD_STATEID:
6070                 spin_lock(&inode->i_lock);
6071                 lo = NFS_I(inode)->layout;
6072                 if (!lo || list_empty(&lo->plh_segs)) {
6073                         spin_unlock(&inode->i_lock);
6074                         /* If the open stateid was bad, then recover it. */
6075                         state = lgp->args.ctx->state;
6076                 } else {
6077                         LIST_HEAD(head);
6078
6079                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6080                         spin_unlock(&inode->i_lock);
6081                         /* Mark the bad layout state as invalid, then
6082                          * retry using the open stateid. */
6083                         pnfs_free_lseg_list(&head);
6084                 }
6085         }
6086         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6087                 rpc_restart_call_prepare(task);
6088 out:
6089         dprintk("<-- %s\n", __func__);
6090 }
6091
6092 static size_t max_response_pages(struct nfs_server *server)
6093 {
6094         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6095         return nfs_page_array_len(0, max_resp_sz);
6096 }
6097
6098 static void nfs4_free_pages(struct page **pages, size_t size)
6099 {
6100         int i;
6101
6102         if (!pages)
6103                 return;
6104
6105         for (i = 0; i < size; i++) {
6106                 if (!pages[i])
6107                         break;
6108                 __free_page(pages[i]);
6109         }
6110         kfree(pages);
6111 }
6112
6113 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6114 {
6115         struct page **pages;
6116         int i;
6117
6118         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6119         if (!pages) {
6120                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6121                 return NULL;
6122         }
6123
6124         for (i = 0; i < size; i++) {
6125                 pages[i] = alloc_page(gfp_flags);
6126                 if (!pages[i]) {
6127                         dprintk("%s: failed to allocate page\n", __func__);
6128                         nfs4_free_pages(pages, size);
6129                         return NULL;
6130                 }
6131         }
6132
6133         return pages;
6134 }
6135
6136 static void nfs4_layoutget_release(void *calldata)
6137 {
6138         struct nfs4_layoutget *lgp = calldata;
6139         struct inode *inode = lgp->args.inode;
6140         struct nfs_server *server = NFS_SERVER(inode);
6141         size_t max_pages = max_response_pages(server);
6142
6143         dprintk("--> %s\n", __func__);
6144         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6145         pnfs_put_layout_hdr(NFS_I(inode)->layout);
6146         put_nfs_open_context(lgp->args.ctx);
6147         kfree(calldata);
6148         dprintk("<-- %s\n", __func__);
6149 }
6150
6151 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6152         .rpc_call_prepare = nfs4_layoutget_prepare,
6153         .rpc_call_done = nfs4_layoutget_done,
6154         .rpc_release = nfs4_layoutget_release,
6155 };
6156
6157 struct pnfs_layout_segment *
6158 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6159 {
6160         struct inode *inode = lgp->args.inode;
6161         struct nfs_server *server = NFS_SERVER(inode);
6162         size_t max_pages = max_response_pages(server);
6163         struct rpc_task *task;
6164         struct rpc_message msg = {
6165                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6166                 .rpc_argp = &lgp->args,
6167                 .rpc_resp = &lgp->res,
6168         };
6169         struct rpc_task_setup task_setup_data = {
6170                 .rpc_client = server->client,
6171                 .rpc_message = &msg,
6172                 .callback_ops = &nfs4_layoutget_call_ops,
6173                 .callback_data = lgp,
6174                 .flags = RPC_TASK_ASYNC,
6175         };
6176         struct pnfs_layout_segment *lseg = NULL;
6177         int status = 0;
6178
6179         dprintk("--> %s\n", __func__);
6180
6181         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6182         if (!lgp->args.layout.pages) {
6183                 nfs4_layoutget_release(lgp);
6184                 return ERR_PTR(-ENOMEM);
6185         }
6186         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6187         lgp->args.timestamp = jiffies;
6188
6189         lgp->res.layoutp = &lgp->args.layout;
6190       &n