]> git.kernelconcepts.de Git - karo-tx-linux.git/blob - fs/nfs/nfs4proc.c
NFSv4.1: Clean up nfs4_free_slot
[karo-tx-linux.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/nfs_idmap.h>
55 #include <linux/sunrpc/bc_xprt.h>
56 #include <linux/xattr.h>
57 #include <linux/utsname.h>
58 #include <linux/freezer.h>
59
60 #include "nfs4_fs.h"
61 #include "delegation.h"
62 #include "internal.h"
63 #include "iostat.h"
64 #include "callback.h"
65 #include "pnfs.h"
66 #include "netns.h"
67
68 #define NFSDBG_FACILITY         NFSDBG_PROC
69
70 #define NFS4_POLL_RETRY_MIN     (HZ/10)
71 #define NFS4_POLL_RETRY_MAX     (15*HZ)
72
73 #define NFS4_MAX_LOOP_ON_RECOVER (10)
74
75 struct nfs4_opendata;
76 static int _nfs4_proc_open(struct nfs4_opendata *data);
77 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
78 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
79 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
80 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
81 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *);
82 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
83 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
84                             struct nfs_fattr *fattr, struct iattr *sattr,
85                             struct nfs4_state *state);
86 #ifdef CONFIG_NFS_V4_1
87 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
88 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
89 #endif
90 /* Prevent leaks of NFSv4 errors into userland */
91 static int nfs4_map_errors(int err)
92 {
93         if (err >= -1000)
94                 return err;
95         switch (err) {
96         case -NFS4ERR_RESOURCE:
97                 return -EREMOTEIO;
98         case -NFS4ERR_WRONGSEC:
99                 return -EPERM;
100         case -NFS4ERR_BADOWNER:
101         case -NFS4ERR_BADNAME:
102                 return -EINVAL;
103         case -NFS4ERR_SHARE_DENIED:
104                 return -EACCES;
105         case -NFS4ERR_MINOR_VERS_MISMATCH:
106                 return -EPROTONOSUPPORT;
107         case -NFS4ERR_ACCESS:
108                 return -EACCES;
109         default:
110                 dprintk("%s could not handle NFSv4 error %d\n",
111                                 __func__, -err);
112                 break;
113         }
114         return -EIO;
115 }
116
117 /*
118  * This is our standard bitmap for GETATTR requests.
119  */
120 const u32 nfs4_fattr_bitmap[3] = {
121         FATTR4_WORD0_TYPE
122         | FATTR4_WORD0_CHANGE
123         | FATTR4_WORD0_SIZE
124         | FATTR4_WORD0_FSID
125         | FATTR4_WORD0_FILEID,
126         FATTR4_WORD1_MODE
127         | FATTR4_WORD1_NUMLINKS
128         | FATTR4_WORD1_OWNER
129         | FATTR4_WORD1_OWNER_GROUP
130         | FATTR4_WORD1_RAWDEV
131         | FATTR4_WORD1_SPACE_USED
132         | FATTR4_WORD1_TIME_ACCESS
133         | FATTR4_WORD1_TIME_METADATA
134         | FATTR4_WORD1_TIME_MODIFY
135 };
136
137 static const u32 nfs4_pnfs_open_bitmap[3] = {
138         FATTR4_WORD0_TYPE
139         | FATTR4_WORD0_CHANGE
140         | FATTR4_WORD0_SIZE
141         | FATTR4_WORD0_FSID
142         | FATTR4_WORD0_FILEID,
143         FATTR4_WORD1_MODE
144         | FATTR4_WORD1_NUMLINKS
145         | FATTR4_WORD1_OWNER
146         | FATTR4_WORD1_OWNER_GROUP
147         | FATTR4_WORD1_RAWDEV
148         | FATTR4_WORD1_SPACE_USED
149         | FATTR4_WORD1_TIME_ACCESS
150         | FATTR4_WORD1_TIME_METADATA
151         | FATTR4_WORD1_TIME_MODIFY,
152         FATTR4_WORD2_MDSTHRESHOLD
153 };
154
155 static const u32 nfs4_open_noattr_bitmap[3] = {
156         FATTR4_WORD0_TYPE
157         | FATTR4_WORD0_CHANGE
158         | FATTR4_WORD0_FILEID,
159 };
160
161 const u32 nfs4_statfs_bitmap[2] = {
162         FATTR4_WORD0_FILES_AVAIL
163         | FATTR4_WORD0_FILES_FREE
164         | FATTR4_WORD0_FILES_TOTAL,
165         FATTR4_WORD1_SPACE_AVAIL
166         | FATTR4_WORD1_SPACE_FREE
167         | FATTR4_WORD1_SPACE_TOTAL
168 };
169
170 const u32 nfs4_pathconf_bitmap[2] = {
171         FATTR4_WORD0_MAXLINK
172         | FATTR4_WORD0_MAXNAME,
173         0
174 };
175
176 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
177                         | FATTR4_WORD0_MAXREAD
178                         | FATTR4_WORD0_MAXWRITE
179                         | FATTR4_WORD0_LEASE_TIME,
180                         FATTR4_WORD1_TIME_DELTA
181                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
182                         FATTR4_WORD2_LAYOUT_BLKSIZE
183 };
184
185 const u32 nfs4_fs_locations_bitmap[2] = {
186         FATTR4_WORD0_TYPE
187         | FATTR4_WORD0_CHANGE
188         | FATTR4_WORD0_SIZE
189         | FATTR4_WORD0_FSID
190         | FATTR4_WORD0_FILEID
191         | FATTR4_WORD0_FS_LOCATIONS,
192         FATTR4_WORD1_MODE
193         | FATTR4_WORD1_NUMLINKS
194         | FATTR4_WORD1_OWNER
195         | FATTR4_WORD1_OWNER_GROUP
196         | FATTR4_WORD1_RAWDEV
197         | FATTR4_WORD1_SPACE_USED
198         | FATTR4_WORD1_TIME_ACCESS
199         | FATTR4_WORD1_TIME_METADATA
200         | FATTR4_WORD1_TIME_MODIFY
201         | FATTR4_WORD1_MOUNTED_ON_FILEID
202 };
203
204 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
205                 struct nfs4_readdir_arg *readdir)
206 {
207         __be32 *start, *p;
208
209         if (cookie > 2) {
210                 readdir->cookie = cookie;
211                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
212                 return;
213         }
214
215         readdir->cookie = 0;
216         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
217         if (cookie == 2)
218                 return;
219         
220         /*
221          * NFSv4 servers do not return entries for '.' and '..'
222          * Therefore, we fake these entries here.  We let '.'
223          * have cookie 0 and '..' have cookie 1.  Note that
224          * when talking to the server, we always send cookie 0
225          * instead of 1 or 2.
226          */
227         start = p = kmap_atomic(*readdir->pages);
228         
229         if (cookie == 0) {
230                 *p++ = xdr_one;                                  /* next */
231                 *p++ = xdr_zero;                   /* cookie, first word */
232                 *p++ = xdr_one;                   /* cookie, second word */
233                 *p++ = xdr_one;                             /* entry len */
234                 memcpy(p, ".\0\0\0", 4);                        /* entry */
235                 p++;
236                 *p++ = xdr_one;                         /* bitmap length */
237                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
238                 *p++ = htonl(8);              /* attribute buffer length */
239                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
240         }
241         
242         *p++ = xdr_one;                                  /* next */
243         *p++ = xdr_zero;                   /* cookie, first word */
244         *p++ = xdr_two;                   /* cookie, second word */
245         *p++ = xdr_two;                             /* entry len */
246         memcpy(p, "..\0\0", 4);                         /* entry */
247         p++;
248         *p++ = xdr_one;                         /* bitmap length */
249         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
250         *p++ = htonl(8);              /* attribute buffer length */
251         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
252
253         readdir->pgbase = (char *)p - (char *)start;
254         readdir->count -= readdir->pgbase;
255         kunmap_atomic(start);
256 }
257
258 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
259 {
260         int res;
261
262         might_sleep();
263
264         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
265                         nfs_wait_bit_killable, TASK_KILLABLE);
266         if (res)
267                 return res;
268
269         if (clp->cl_cons_state < 0)
270                 return clp->cl_cons_state;
271         return 0;
272 }
273
274 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
275 {
276         int res = 0;
277
278         might_sleep();
279
280         if (*timeout <= 0)
281                 *timeout = NFS4_POLL_RETRY_MIN;
282         if (*timeout > NFS4_POLL_RETRY_MAX)
283                 *timeout = NFS4_POLL_RETRY_MAX;
284         freezable_schedule_timeout_killable(*timeout);
285         if (fatal_signal_pending(current))
286                 res = -ERESTARTSYS;
287         *timeout <<= 1;
288         return res;
289 }
290
291 /* This is the error handling routine for processes that are allowed
292  * to sleep.
293  */
294 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
295 {
296         struct nfs_client *clp = server->nfs_client;
297         struct nfs4_state *state = exception->state;
298         struct inode *inode = exception->inode;
299         int ret = errorcode;
300
301         exception->retry = 0;
302         switch(errorcode) {
303                 case 0:
304                         return 0;
305                 case -NFS4ERR_OPENMODE:
306                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
307                                 nfs4_inode_return_delegation(inode);
308                                 exception->retry = 1;
309                                 return 0;
310                         }
311                         if (state == NULL)
312                                 break;
313                         nfs4_schedule_stateid_recovery(server, state);
314                         goto wait_on_recovery;
315                 case -NFS4ERR_DELEG_REVOKED:
316                 case -NFS4ERR_ADMIN_REVOKED:
317                 case -NFS4ERR_BAD_STATEID:
318                         if (state == NULL)
319                                 break;
320                         nfs_remove_bad_delegation(state->inode);
321                         nfs4_schedule_stateid_recovery(server, state);
322                         goto wait_on_recovery;
323                 case -NFS4ERR_EXPIRED:
324                         if (state != NULL)
325                                 nfs4_schedule_stateid_recovery(server, state);
326                 case -NFS4ERR_STALE_STATEID:
327                 case -NFS4ERR_STALE_CLIENTID:
328                         nfs4_schedule_lease_recovery(clp);
329                         goto wait_on_recovery;
330 #if defined(CONFIG_NFS_V4_1)
331                 case -NFS4ERR_BADSESSION:
332                 case -NFS4ERR_BADSLOT:
333                 case -NFS4ERR_BAD_HIGH_SLOT:
334                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
335                 case -NFS4ERR_DEADSESSION:
336                 case -NFS4ERR_SEQ_FALSE_RETRY:
337                 case -NFS4ERR_SEQ_MISORDERED:
338                         dprintk("%s ERROR: %d Reset session\n", __func__,
339                                 errorcode);
340                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
341                         goto wait_on_recovery;
342 #endif /* defined(CONFIG_NFS_V4_1) */
343                 case -NFS4ERR_FILE_OPEN:
344                         if (exception->timeout > HZ) {
345                                 /* We have retried a decent amount, time to
346                                  * fail
347                                  */
348                                 ret = -EBUSY;
349                                 break;
350                         }
351                 case -NFS4ERR_GRACE:
352                 case -NFS4ERR_DELAY:
353                 case -EKEYEXPIRED:
354                         ret = nfs4_delay(server->client, &exception->timeout);
355                         if (ret != 0)
356                                 break;
357                 case -NFS4ERR_RETRY_UNCACHED_REP:
358                 case -NFS4ERR_OLD_STATEID:
359                         exception->retry = 1;
360                         break;
361                 case -NFS4ERR_BADOWNER:
362                         /* The following works around a Linux server bug! */
363                 case -NFS4ERR_BADNAME:
364                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
365                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
366                                 exception->retry = 1;
367                                 printk(KERN_WARNING "NFS: v4 server %s "
368                                                 "does not accept raw "
369                                                 "uid/gids. "
370                                                 "Reenabling the idmapper.\n",
371                                                 server->nfs_client->cl_hostname);
372                         }
373         }
374         /* We failed to handle the error */
375         return nfs4_map_errors(ret);
376 wait_on_recovery:
377         ret = nfs4_wait_clnt_recover(clp);
378         if (ret == 0)
379                 exception->retry = 1;
380         return ret;
381 }
382
383
384 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
385 {
386         spin_lock(&clp->cl_lock);
387         if (time_before(clp->cl_last_renewal,timestamp))
388                 clp->cl_last_renewal = timestamp;
389         spin_unlock(&clp->cl_lock);
390 }
391
392 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
393 {
394         do_renew_lease(server->nfs_client, timestamp);
395 }
396
397 #if defined(CONFIG_NFS_V4_1)
398
399 /*
400  * nfs4_free_slot - free a slot and efficiently update slot table.
401  *
402  * freeing a slot is trivially done by clearing its respective bit
403  * in the bitmap.
404  * If the freed slotid equals highest_used_slotid we want to update it
405  * so that the server would be able to size down the slot table if needed,
406  * otherwise we know that the highest_used_slotid is still in use.
407  * When updating highest_used_slotid there may be "holes" in the bitmap
408  * so we need to scan down from highest_used_slotid to 0 looking for the now
409  * highest slotid in use.
410  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
411  *
412  * Must be called while holding tbl->slot_tbl_lock
413  */
414 static void
415 nfs4_free_slot(struct nfs4_slot_table *tbl, struct nfs4_slot *slot)
416 {
417         u32 slotid = slot->slot_nr;
418
419         /* clear used bit in bitmap */
420         __clear_bit(slotid, tbl->used_slots);
421
422         /* update highest_used_slotid when it is freed */
423         if (slotid == tbl->highest_used_slotid) {
424                 u32 new_max = find_last_bit(tbl->used_slots, slotid);
425                 if (new_max < slotid)
426                         tbl->highest_used_slotid = new_max;
427                 else
428                         tbl->highest_used_slotid = NFS4_NO_SLOT;
429         }
430         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
431                 slotid, tbl->highest_used_slotid);
432 }
433
434 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
435 {
436         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
437         return true;
438 }
439
440 /*
441  * Signal state manager thread if session fore channel is drained
442  */
443 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
444 {
445         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
446                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
447                                 nfs4_set_task_privileged, NULL);
448                 return;
449         }
450
451         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
452                 return;
453
454         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
455         complete(&ses->fc_slot_table.complete);
456 }
457
458 /*
459  * Signal state manager thread if session back channel is drained
460  */
461 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
462 {
463         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
464             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
465                 return;
466         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
467         complete(&ses->bc_slot_table.complete);
468 }
469
470 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
471 {
472         struct nfs4_session *session;
473         struct nfs4_slot_table *tbl;
474
475         if (!res->sr_slot) {
476                 /* just wake up the next guy waiting since
477                  * we may have not consumed a slot after all */
478                 dprintk("%s: No slot\n", __func__);
479                 return;
480         }
481         tbl = res->sr_slot->table;
482         session = tbl->session;
483
484         spin_lock(&tbl->slot_tbl_lock);
485         nfs4_free_slot(tbl, res->sr_slot);
486         nfs4_check_drain_fc_complete(session);
487         spin_unlock(&tbl->slot_tbl_lock);
488         res->sr_slot = NULL;
489 }
490
491 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
492 {
493         struct nfs4_session *session;
494         struct nfs4_slot *slot;
495         unsigned long timestamp;
496         struct nfs_client *clp;
497
498         /*
499          * sr_status remains 1 if an RPC level error occurred. The server
500          * may or may not have processed the sequence operation..
501          * Proceed as if the server received and processed the sequence
502          * operation.
503          */
504         if (res->sr_status == 1)
505                 res->sr_status = NFS_OK;
506
507         /* don't increment the sequence number if the task wasn't sent */
508         if (!RPC_WAS_SENT(task))
509                 goto out;
510
511         slot = res->sr_slot;
512         session = slot->table->session;
513
514         /* Check the SEQUENCE operation status */
515         switch (res->sr_status) {
516         case 0:
517                 /* Update the slot's sequence and clientid lease timer */
518                 ++slot->seq_nr;
519                 timestamp = slot->renewal_time;
520                 clp = session->clp;
521                 do_renew_lease(clp, timestamp);
522                 /* Check sequence flags */
523                 if (res->sr_status_flags != 0)
524                         nfs4_schedule_lease_recovery(clp);
525                 break;
526         case -NFS4ERR_DELAY:
527                 /* The server detected a resend of the RPC call and
528                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
529                  * of RFC5661.
530                  */
531                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
532                         __func__,
533                         slot->slot_nr,
534                         slot->seq_nr);
535                 goto out_retry;
536         default:
537                 /* Just update the slot sequence no. */
538                 ++slot->seq_nr;
539         }
540 out:
541         /* The session may be reset by one of the error handlers. */
542         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
543         nfs41_sequence_free_slot(res);
544         return 1;
545 out_retry:
546         if (!rpc_restart_call(task))
547                 goto out;
548         rpc_delay(task, NFS4_POLL_RETRY_MAX);
549         return 0;
550 }
551
552 static int nfs4_sequence_done(struct rpc_task *task,
553                                struct nfs4_sequence_res *res)
554 {
555         if (res->sr_slot == NULL)
556                 return 1;
557         return nfs41_sequence_done(task, res);
558 }
559
560 /*
561  * nfs4_alloc_slot - efficiently look for a free slot
562  *
563  * nfs4_alloc_slot looks for an unset bit in the used_slots bitmap.
564  * If found, we mark the slot as used, update the highest_used_slotid,
565  * and respectively set up the sequence operation args.
566  *
567  * Note: must be called with under the slot_tbl_lock.
568  */
569 static struct nfs4_slot *nfs4_alloc_slot(struct nfs4_slot_table *tbl)
570 {
571         struct nfs4_slot *ret = NULL;
572         u32 slotid;
573
574         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
575                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
576                 tbl->max_slots);
577         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
578         if (slotid >= tbl->max_slots)
579                 goto out;
580         __set_bit(slotid, tbl->used_slots);
581         if (slotid > tbl->highest_used_slotid ||
582                         tbl->highest_used_slotid == NFS4_NO_SLOT)
583                 tbl->highest_used_slotid = slotid;
584         ret = &tbl->slots[slotid];
585         ret->renewal_time = jiffies;
586
587 out:
588         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
589                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
590                 ret ? ret->slot_nr : -1);
591         return ret;
592 }
593
594 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
595                 struct nfs4_sequence_res *res, int cache_reply)
596 {
597         args->sa_slot = NULL;
598         args->sa_cache_this = 0;
599         if (cache_reply)
600                 args->sa_cache_this = 1;
601         res->sr_slot = NULL;
602 }
603
604 int nfs41_setup_sequence(struct nfs4_session *session,
605                                 struct nfs4_sequence_args *args,
606                                 struct nfs4_sequence_res *res,
607                                 struct rpc_task *task)
608 {
609         struct nfs4_slot *slot;
610         struct nfs4_slot_table *tbl;
611
612         dprintk("--> %s\n", __func__);
613         /* slot already allocated? */
614         if (res->sr_slot != NULL)
615                 return 0;
616
617         tbl = &session->fc_slot_table;
618
619         spin_lock(&tbl->slot_tbl_lock);
620         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
621             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
622                 /* The state manager will wait until the slot table is empty */
623                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
624                 spin_unlock(&tbl->slot_tbl_lock);
625                 dprintk("%s session is draining\n", __func__);
626                 return -EAGAIN;
627         }
628
629         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
630             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
631                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
632                 spin_unlock(&tbl->slot_tbl_lock);
633                 dprintk("%s enforce FIFO order\n", __func__);
634                 return -EAGAIN;
635         }
636
637         slot = nfs4_alloc_slot(tbl);
638         if (slot == NULL) {
639                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
640                 spin_unlock(&tbl->slot_tbl_lock);
641                 dprintk("<-- %s: no free slots\n", __func__);
642                 return -EAGAIN;
643         }
644         spin_unlock(&tbl->slot_tbl_lock);
645
646         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
647
648         args->sa_slot = slot;
649
650         dprintk("<-- %s slotid=%d seqid=%d\n", __func__,
651                         slot->slot_nr, slot->seq_nr);
652
653         res->sr_slot = slot;
654         res->sr_status_flags = 0;
655         /*
656          * sr_status is only set in decode_sequence, and so will remain
657          * set to 1 if an rpc level failure occurs.
658          */
659         res->sr_status = 1;
660         return 0;
661 }
662 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
663
664 int nfs4_setup_sequence(const struct nfs_server *server,
665                         struct nfs4_sequence_args *args,
666                         struct nfs4_sequence_res *res,
667                         struct rpc_task *task)
668 {
669         struct nfs4_session *session = nfs4_get_session(server);
670         int ret = 0;
671
672         if (session == NULL)
673                 goto out;
674
675         dprintk("--> %s clp %p session %p sr_slot %d\n",
676                 __func__, session->clp, session, res->sr_slot ?
677                         res->sr_slot->slot_nr : -1);
678
679         ret = nfs41_setup_sequence(session, args, res, task);
680 out:
681         dprintk("<-- %s status=%d\n", __func__, ret);
682         return ret;
683 }
684
685 struct nfs41_call_sync_data {
686         const struct nfs_server *seq_server;
687         struct nfs4_sequence_args *seq_args;
688         struct nfs4_sequence_res *seq_res;
689 };
690
691 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
692 {
693         struct nfs41_call_sync_data *data = calldata;
694
695         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
696
697         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
698                                 data->seq_res, task))
699                 return;
700         rpc_call_start(task);
701 }
702
703 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
704 {
705         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
706         nfs41_call_sync_prepare(task, calldata);
707 }
708
709 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
710 {
711         struct nfs41_call_sync_data *data = calldata;
712
713         nfs41_sequence_done(task, data->seq_res);
714 }
715
716 static const struct rpc_call_ops nfs41_call_sync_ops = {
717         .rpc_call_prepare = nfs41_call_sync_prepare,
718         .rpc_call_done = nfs41_call_sync_done,
719 };
720
721 static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
722         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
723         .rpc_call_done = nfs41_call_sync_done,
724 };
725
726 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
727                                    struct nfs_server *server,
728                                    struct rpc_message *msg,
729                                    struct nfs4_sequence_args *args,
730                                    struct nfs4_sequence_res *res,
731                                    int privileged)
732 {
733         int ret;
734         struct rpc_task *task;
735         struct nfs41_call_sync_data data = {
736                 .seq_server = server,
737                 .seq_args = args,
738                 .seq_res = res,
739         };
740         struct rpc_task_setup task_setup = {
741                 .rpc_client = clnt,
742                 .rpc_message = msg,
743                 .callback_ops = &nfs41_call_sync_ops,
744                 .callback_data = &data
745         };
746
747         if (privileged)
748                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
749         task = rpc_run_task(&task_setup);
750         if (IS_ERR(task))
751                 ret = PTR_ERR(task);
752         else {
753                 ret = task->tk_status;
754                 rpc_put_task(task);
755         }
756         return ret;
757 }
758
759 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
760                             struct nfs_server *server,
761                             struct rpc_message *msg,
762                             struct nfs4_sequence_args *args,
763                             struct nfs4_sequence_res *res,
764                             int cache_reply)
765 {
766         nfs41_init_sequence(args, res, cache_reply);
767         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
768 }
769
770 #else
771 static inline
772 void nfs41_init_sequence(struct nfs4_sequence_args *args,
773                 struct nfs4_sequence_res *res, int cache_reply)
774 {
775 }
776
777 static int nfs4_sequence_done(struct rpc_task *task,
778                                struct nfs4_sequence_res *res)
779 {
780         return 1;
781 }
782 #endif /* CONFIG_NFS_V4_1 */
783
784 int _nfs4_call_sync(struct rpc_clnt *clnt,
785                     struct nfs_server *server,
786                     struct rpc_message *msg,
787                     struct nfs4_sequence_args *args,
788                     struct nfs4_sequence_res *res,
789                     int cache_reply)
790 {
791         nfs41_init_sequence(args, res, cache_reply);
792         return rpc_call_sync(clnt, msg, 0);
793 }
794
795 static inline
796 int nfs4_call_sync(struct rpc_clnt *clnt,
797                    struct nfs_server *server,
798                    struct rpc_message *msg,
799                    struct nfs4_sequence_args *args,
800                    struct nfs4_sequence_res *res,
801                    int cache_reply)
802 {
803         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
804                                                 args, res, cache_reply);
805 }
806
807 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
808 {
809         struct nfs_inode *nfsi = NFS_I(dir);
810
811         spin_lock(&dir->i_lock);
812         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
813         if (!cinfo->atomic || cinfo->before != dir->i_version)
814                 nfs_force_lookup_revalidate(dir);
815         dir->i_version = cinfo->after;
816         spin_unlock(&dir->i_lock);
817 }
818
819 struct nfs4_opendata {
820         struct kref kref;
821         struct nfs_openargs o_arg;
822         struct nfs_openres o_res;
823         struct nfs_open_confirmargs c_arg;
824         struct nfs_open_confirmres c_res;
825         struct nfs4_string owner_name;
826         struct nfs4_string group_name;
827         struct nfs_fattr f_attr;
828         struct dentry *dir;
829         struct dentry *dentry;
830         struct nfs4_state_owner *owner;
831         struct nfs4_state *state;
832         struct iattr attrs;
833         unsigned long timestamp;
834         unsigned int rpc_done : 1;
835         int rpc_status;
836         int cancelled;
837 };
838
839
840 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
841 {
842         p->o_res.f_attr = &p->f_attr;
843         p->o_res.seqid = p->o_arg.seqid;
844         p->c_res.seqid = p->c_arg.seqid;
845         p->o_res.server = p->o_arg.server;
846         p->o_res.access_request = p->o_arg.access;
847         nfs_fattr_init(&p->f_attr);
848         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
849 }
850
851 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
852                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
853                 const struct iattr *attrs,
854                 gfp_t gfp_mask)
855 {
856         struct dentry *parent = dget_parent(dentry);
857         struct inode *dir = parent->d_inode;
858         struct nfs_server *server = NFS_SERVER(dir);
859         struct nfs4_opendata *p;
860
861         p = kzalloc(sizeof(*p), gfp_mask);
862         if (p == NULL)
863                 goto err;
864         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
865         if (p->o_arg.seqid == NULL)
866                 goto err_free;
867         nfs_sb_active(dentry->d_sb);
868         p->dentry = dget(dentry);
869         p->dir = parent;
870         p->owner = sp;
871         atomic_inc(&sp->so_count);
872         p->o_arg.fh = NFS_FH(dir);
873         p->o_arg.open_flags = flags;
874         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
875         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
876          * will return permission denied for all bits until close */
877         if (!(flags & O_EXCL)) {
878                 /* ask server to check for all possible rights as results
879                  * are cached */
880                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
881                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
882         }
883         p->o_arg.clientid = server->nfs_client->cl_clientid;
884         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
885         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
886         p->o_arg.name = &dentry->d_name;
887         p->o_arg.server = server;
888         p->o_arg.bitmask = server->attr_bitmask;
889         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
890         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
891         if (attrs != NULL && attrs->ia_valid != 0) {
892                 __be32 verf[2];
893
894                 p->o_arg.u.attrs = &p->attrs;
895                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
896
897                 verf[0] = jiffies;
898                 verf[1] = current->pid;
899                 memcpy(p->o_arg.u.verifier.data, verf,
900                                 sizeof(p->o_arg.u.verifier.data));
901         }
902         p->c_arg.fh = &p->o_res.fh;
903         p->c_arg.stateid = &p->o_res.stateid;
904         p->c_arg.seqid = p->o_arg.seqid;
905         nfs4_init_opendata_res(p);
906         kref_init(&p->kref);
907         return p;
908 err_free:
909         kfree(p);
910 err:
911         dput(parent);
912         return NULL;
913 }
914
915 static void nfs4_opendata_free(struct kref *kref)
916 {
917         struct nfs4_opendata *p = container_of(kref,
918                         struct nfs4_opendata, kref);
919         struct super_block *sb = p->dentry->d_sb;
920
921         nfs_free_seqid(p->o_arg.seqid);
922         if (p->state != NULL)
923                 nfs4_put_open_state(p->state);
924         nfs4_put_state_owner(p->owner);
925         dput(p->dir);
926         dput(p->dentry);
927         nfs_sb_deactive(sb);
928         nfs_fattr_free_names(&p->f_attr);
929         kfree(p);
930 }
931
932 static void nfs4_opendata_put(struct nfs4_opendata *p)
933 {
934         if (p != NULL)
935                 kref_put(&p->kref, nfs4_opendata_free);
936 }
937
938 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
939 {
940         int ret;
941
942         ret = rpc_wait_for_completion_task(task);
943         return ret;
944 }
945
946 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
947 {
948         int ret = 0;
949
950         if (open_mode & (O_EXCL|O_TRUNC))
951                 goto out;
952         switch (mode & (FMODE_READ|FMODE_WRITE)) {
953                 case FMODE_READ:
954                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
955                                 && state->n_rdonly != 0;
956                         break;
957                 case FMODE_WRITE:
958                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
959                                 && state->n_wronly != 0;
960                         break;
961                 case FMODE_READ|FMODE_WRITE:
962                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
963                                 && state->n_rdwr != 0;
964         }
965 out:
966         return ret;
967 }
968
969 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
970 {
971         if (delegation == NULL)
972                 return 0;
973         if ((delegation->type & fmode) != fmode)
974                 return 0;
975         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
976                 return 0;
977         nfs_mark_delegation_referenced(delegation);
978         return 1;
979 }
980
981 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
982 {
983         switch (fmode) {
984                 case FMODE_WRITE:
985                         state->n_wronly++;
986                         break;
987                 case FMODE_READ:
988                         state->n_rdonly++;
989                         break;
990                 case FMODE_READ|FMODE_WRITE:
991                         state->n_rdwr++;
992         }
993         nfs4_state_set_mode_locked(state, state->state | fmode);
994 }
995
996 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
997 {
998         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
999                 nfs4_stateid_copy(&state->stateid, stateid);
1000         nfs4_stateid_copy(&state->open_stateid, stateid);
1001         switch (fmode) {
1002                 case FMODE_READ:
1003                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1004                         break;
1005                 case FMODE_WRITE:
1006                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1007                         break;
1008                 case FMODE_READ|FMODE_WRITE:
1009                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1010         }
1011 }
1012
1013 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1014 {
1015         write_seqlock(&state->seqlock);
1016         nfs_set_open_stateid_locked(state, stateid, fmode);
1017         write_sequnlock(&state->seqlock);
1018 }
1019
1020 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1021 {
1022         /*
1023          * Protect the call to nfs4_state_set_mode_locked and
1024          * serialise the stateid update
1025          */
1026         write_seqlock(&state->seqlock);
1027         if (deleg_stateid != NULL) {
1028                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1029                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1030         }
1031         if (open_stateid != NULL)
1032                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1033         write_sequnlock(&state->seqlock);
1034         spin_lock(&state->owner->so_lock);
1035         update_open_stateflags(state, fmode);
1036         spin_unlock(&state->owner->so_lock);
1037 }
1038
1039 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1040 {
1041         struct nfs_inode *nfsi = NFS_I(state->inode);
1042         struct nfs_delegation *deleg_cur;
1043         int ret = 0;
1044
1045         fmode &= (FMODE_READ|FMODE_WRITE);
1046
1047         rcu_read_lock();
1048         deleg_cur = rcu_dereference(nfsi->delegation);
1049         if (deleg_cur == NULL)
1050                 goto no_delegation;
1051
1052         spin_lock(&deleg_cur->lock);
1053         if (nfsi->delegation != deleg_cur ||
1054             (deleg_cur->type & fmode) != fmode)
1055                 goto no_delegation_unlock;
1056
1057         if (delegation == NULL)
1058                 delegation = &deleg_cur->stateid;
1059         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1060                 goto no_delegation_unlock;
1061
1062         nfs_mark_delegation_referenced(deleg_cur);
1063         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1064         ret = 1;
1065 no_delegation_unlock:
1066         spin_unlock(&deleg_cur->lock);
1067 no_delegation:
1068         rcu_read_unlock();
1069
1070         if (!ret && open_stateid != NULL) {
1071                 __update_open_stateid(state, open_stateid, NULL, fmode);
1072                 ret = 1;
1073         }
1074
1075         return ret;
1076 }
1077
1078
1079 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1080 {
1081         struct nfs_delegation *delegation;
1082
1083         rcu_read_lock();
1084         delegation = rcu_dereference(NFS_I(inode)->delegation);
1085         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1086                 rcu_read_unlock();
1087                 return;
1088         }
1089         rcu_read_unlock();
1090         nfs4_inode_return_delegation(inode);
1091 }
1092
1093 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1094 {
1095         struct nfs4_state *state = opendata->state;
1096         struct nfs_inode *nfsi = NFS_I(state->inode);
1097         struct nfs_delegation *delegation;
1098         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1099         fmode_t fmode = opendata->o_arg.fmode;
1100         nfs4_stateid stateid;
1101         int ret = -EAGAIN;
1102
1103         for (;;) {
1104                 if (can_open_cached(state, fmode, open_mode)) {
1105                         spin_lock(&state->owner->so_lock);
1106                         if (can_open_cached(state, fmode, open_mode)) {
1107                                 update_open_stateflags(state, fmode);
1108                                 spin_unlock(&state->owner->so_lock);
1109                                 goto out_return_state;
1110                         }
1111                         spin_unlock(&state->owner->so_lock);
1112                 }
1113                 rcu_read_lock();
1114                 delegation = rcu_dereference(nfsi->delegation);
1115                 if (!can_open_delegated(delegation, fmode)) {
1116                         rcu_read_unlock();
1117                         break;
1118                 }
1119                 /* Save the delegation */
1120                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1121                 rcu_read_unlock();
1122                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1123                 if (ret != 0)
1124                         goto out;
1125                 ret = -EAGAIN;
1126
1127                 /* Try to update the stateid using the delegation */
1128                 if (update_open_stateid(state, NULL, &stateid, fmode))
1129                         goto out_return_state;
1130         }
1131 out:
1132         return ERR_PTR(ret);
1133 out_return_state:
1134         atomic_inc(&state->count);
1135         return state;
1136 }
1137
1138 static void
1139 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1140 {
1141         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1142         struct nfs_delegation *delegation;
1143         int delegation_flags = 0;
1144
1145         rcu_read_lock();
1146         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1147         if (delegation)
1148                 delegation_flags = delegation->flags;
1149         rcu_read_unlock();
1150         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1151                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1152                                    "returning a delegation for "
1153                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1154                                    clp->cl_hostname);
1155         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1156                 nfs_inode_set_delegation(state->inode,
1157                                          data->owner->so_cred,
1158                                          &data->o_res);
1159         else
1160                 nfs_inode_reclaim_delegation(state->inode,
1161                                              data->owner->so_cred,
1162                                              &data->o_res);
1163 }
1164
1165 /*
1166  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1167  * and update the nfs4_state.
1168  */
1169 static struct nfs4_state *
1170 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1171 {
1172         struct inode *inode = data->state->inode;
1173         struct nfs4_state *state = data->state;
1174         int ret;
1175
1176         if (!data->rpc_done) {
1177                 ret = data->rpc_status;
1178                 goto err;
1179         }
1180
1181         ret = -ESTALE;
1182         if (!(data->f_attr.valid & NFS_ATTR_FATTR_TYPE) ||
1183             !(data->f_attr.valid & NFS_ATTR_FATTR_FILEID) ||
1184             !(data->f_attr.valid & NFS_ATTR_FATTR_CHANGE))
1185                 goto err;
1186
1187         ret = -ENOMEM;
1188         state = nfs4_get_open_state(inode, data->owner);
1189         if (state == NULL)
1190                 goto err;
1191
1192         ret = nfs_refresh_inode(inode, &data->f_attr);
1193         if (ret)
1194                 goto err;
1195
1196         if (data->o_res.delegation_type != 0)
1197                 nfs4_opendata_check_deleg(data, state);
1198         update_open_stateid(state, &data->o_res.stateid, NULL,
1199                             data->o_arg.fmode);
1200
1201         return state;
1202 err:
1203         return ERR_PTR(ret);
1204
1205 }
1206
1207 static struct nfs4_state *
1208 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1209 {
1210         struct inode *inode;
1211         struct nfs4_state *state = NULL;
1212         int ret;
1213
1214         if (!data->rpc_done) {
1215                 state = nfs4_try_open_cached(data);
1216                 goto out;
1217         }
1218
1219         ret = -EAGAIN;
1220         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1221                 goto err;
1222         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1223         ret = PTR_ERR(inode);
1224         if (IS_ERR(inode))
1225                 goto err;
1226         ret = -ENOMEM;
1227         state = nfs4_get_open_state(inode, data->owner);
1228         if (state == NULL)
1229                 goto err_put_inode;
1230         if (data->o_res.delegation_type != 0)
1231                 nfs4_opendata_check_deleg(data, state);
1232         update_open_stateid(state, &data->o_res.stateid, NULL,
1233                         data->o_arg.fmode);
1234         iput(inode);
1235 out:
1236         return state;
1237 err_put_inode:
1238         iput(inode);
1239 err:
1240         return ERR_PTR(ret);
1241 }
1242
1243 static struct nfs4_state *
1244 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1245 {
1246         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1247                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1248         return _nfs4_opendata_to_nfs4_state(data);
1249 }
1250
1251 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1252 {
1253         struct nfs_inode *nfsi = NFS_I(state->inode);
1254         struct nfs_open_context *ctx;
1255
1256         spin_lock(&state->inode->i_lock);
1257         list_for_each_entry(ctx, &nfsi->open_files, list) {
1258                 if (ctx->state != state)
1259                         continue;
1260                 get_nfs_open_context(ctx);
1261                 spin_unlock(&state->inode->i_lock);
1262                 return ctx;
1263         }
1264         spin_unlock(&state->inode->i_lock);
1265         return ERR_PTR(-ENOENT);
1266 }
1267
1268 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1269 {
1270         struct nfs4_opendata *opendata;
1271
1272         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1273         if (opendata == NULL)
1274                 return ERR_PTR(-ENOMEM);
1275         opendata->state = state;
1276         atomic_inc(&state->count);
1277         return opendata;
1278 }
1279
1280 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1281 {
1282         struct nfs4_state *newstate;
1283         int ret;
1284
1285         opendata->o_arg.open_flags = 0;
1286         opendata->o_arg.fmode = fmode;
1287         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1288         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1289         nfs4_init_opendata_res(opendata);
1290         ret = _nfs4_recover_proc_open(opendata);
1291         if (ret != 0)
1292                 return ret; 
1293         newstate = nfs4_opendata_to_nfs4_state(opendata);
1294         if (IS_ERR(newstate))
1295                 return PTR_ERR(newstate);
1296         nfs4_close_state(newstate, fmode);
1297         *res = newstate;
1298         return 0;
1299 }
1300
1301 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1302 {
1303         struct nfs4_state *newstate;
1304         int ret;
1305
1306         /* memory barrier prior to reading state->n_* */
1307         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1308         smp_rmb();
1309         if (state->n_rdwr != 0) {
1310                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1311                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1312                 if (ret != 0)
1313                         return ret;
1314                 if (newstate != state)
1315                         return -ESTALE;
1316         }
1317         if (state->n_wronly != 0) {
1318                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1319                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1320                 if (ret != 0)
1321                         return ret;
1322                 if (newstate != state)
1323                         return -ESTALE;
1324         }
1325         if (state->n_rdonly != 0) {
1326                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1327                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1328                 if (ret != 0)
1329                         return ret;
1330                 if (newstate != state)
1331                         return -ESTALE;
1332         }
1333         /*
1334          * We may have performed cached opens for all three recoveries.
1335          * Check if we need to update the current stateid.
1336          */
1337         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1338             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1339                 write_seqlock(&state->seqlock);
1340                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1341                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1342                 write_sequnlock(&state->seqlock);
1343         }
1344         return 0;
1345 }
1346
1347 /*
1348  * OPEN_RECLAIM:
1349  *      reclaim state on the server after a reboot.
1350  */
1351 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1352 {
1353         struct nfs_delegation *delegation;
1354         struct nfs4_opendata *opendata;
1355         fmode_t delegation_type = 0;
1356         int status;
1357
1358         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1359         if (IS_ERR(opendata))
1360                 return PTR_ERR(opendata);
1361         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1362         opendata->o_arg.fh = NFS_FH(state->inode);
1363         rcu_read_lock();
1364         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1365         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1366                 delegation_type = delegation->type;
1367         rcu_read_unlock();
1368         opendata->o_arg.u.delegation_type = delegation_type;
1369         status = nfs4_open_recover(opendata, state);
1370         nfs4_opendata_put(opendata);
1371         return status;
1372 }
1373
1374 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1375 {
1376         struct nfs_server *server = NFS_SERVER(state->inode);
1377         struct nfs4_exception exception = { };
1378         int err;
1379         do {
1380                 err = _nfs4_do_open_reclaim(ctx, state);
1381                 if (err != -NFS4ERR_DELAY)
1382                         break;
1383                 nfs4_handle_exception(server, err, &exception);
1384         } while (exception.retry);
1385         return err;
1386 }
1387
1388 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1389 {
1390         struct nfs_open_context *ctx;
1391         int ret;
1392
1393         ctx = nfs4_state_find_open_context(state);
1394         if (IS_ERR(ctx))
1395                 return PTR_ERR(ctx);
1396         ret = nfs4_do_open_reclaim(ctx, state);
1397         put_nfs_open_context(ctx);
1398         return ret;
1399 }
1400
1401 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1402 {
1403         struct nfs4_opendata *opendata;
1404         int ret;
1405
1406         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1407         if (IS_ERR(opendata))
1408                 return PTR_ERR(opendata);
1409         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1410         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1411         ret = nfs4_open_recover(opendata, state);
1412         nfs4_opendata_put(opendata);
1413         return ret;
1414 }
1415
1416 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1417 {
1418         struct nfs4_exception exception = { };
1419         struct nfs_server *server = NFS_SERVER(state->inode);
1420         int err;
1421         do {
1422                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1423                 switch (err) {
1424                         case 0:
1425                         case -ENOENT:
1426                         case -ESTALE:
1427                                 goto out;
1428                         case -NFS4ERR_BADSESSION:
1429                         case -NFS4ERR_BADSLOT:
1430                         case -NFS4ERR_BAD_HIGH_SLOT:
1431                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1432                         case -NFS4ERR_DEADSESSION:
1433                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1434                                 goto out;
1435                         case -NFS4ERR_STALE_CLIENTID:
1436                         case -NFS4ERR_STALE_STATEID:
1437                         case -NFS4ERR_EXPIRED:
1438                                 /* Don't recall a delegation if it was lost */
1439                                 nfs4_schedule_lease_recovery(server->nfs_client);
1440                                 goto out;
1441                         case -ERESTARTSYS:
1442                                 /*
1443                                  * The show must go on: exit, but mark the
1444                                  * stateid as needing recovery.
1445                                  */
1446                         case -NFS4ERR_DELEG_REVOKED:
1447                         case -NFS4ERR_ADMIN_REVOKED:
1448                         case -NFS4ERR_BAD_STATEID:
1449                                 nfs_inode_find_state_and_recover(state->inode,
1450                                                 stateid);
1451                                 nfs4_schedule_stateid_recovery(server, state);
1452                         case -EKEYEXPIRED:
1453                                 /*
1454                                  * User RPCSEC_GSS context has expired.
1455                                  * We cannot recover this stateid now, so
1456                                  * skip it and allow recovery thread to
1457                                  * proceed.
1458                                  */
1459                         case -ENOMEM:
1460                                 err = 0;
1461                                 goto out;
1462                 }
1463                 err = nfs4_handle_exception(server, err, &exception);
1464         } while (exception.retry);
1465 out:
1466         return err;
1467 }
1468
1469 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1470 {
1471         struct nfs4_opendata *data = calldata;
1472
1473         data->rpc_status = task->tk_status;
1474         if (data->rpc_status == 0) {
1475                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1476                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1477                 renew_lease(data->o_res.server, data->timestamp);
1478                 data->rpc_done = 1;
1479         }
1480 }
1481
1482 static void nfs4_open_confirm_release(void *calldata)
1483 {
1484         struct nfs4_opendata *data = calldata;
1485         struct nfs4_state *state = NULL;
1486
1487         /* If this request hasn't been cancelled, do nothing */
1488         if (data->cancelled == 0)
1489                 goto out_free;
1490         /* In case of error, no cleanup! */
1491         if (!data->rpc_done)
1492                 goto out_free;
1493         state = nfs4_opendata_to_nfs4_state(data);
1494         if (!IS_ERR(state))
1495                 nfs4_close_state(state, data->o_arg.fmode);
1496 out_free:
1497         nfs4_opendata_put(data);
1498 }
1499
1500 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1501         .rpc_call_done = nfs4_open_confirm_done,
1502         .rpc_release = nfs4_open_confirm_release,
1503 };
1504
1505 /*
1506  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1507  */
1508 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1509 {
1510         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1511         struct rpc_task *task;
1512         struct  rpc_message msg = {
1513                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1514                 .rpc_argp = &data->c_arg,
1515                 .rpc_resp = &data->c_res,
1516                 .rpc_cred = data->owner->so_cred,
1517         };
1518         struct rpc_task_setup task_setup_data = {
1519                 .rpc_client = server->client,
1520                 .rpc_message = &msg,
1521                 .callback_ops = &nfs4_open_confirm_ops,
1522                 .callback_data = data,
1523                 .workqueue = nfsiod_workqueue,
1524                 .flags = RPC_TASK_ASYNC,
1525         };
1526         int status;
1527
1528         kref_get(&data->kref);
1529         data->rpc_done = 0;
1530         data->rpc_status = 0;
1531         data->timestamp = jiffies;
1532         task = rpc_run_task(&task_setup_data);
1533         if (IS_ERR(task))
1534                 return PTR_ERR(task);
1535         status = nfs4_wait_for_completion_rpc_task(task);
1536         if (status != 0) {
1537                 data->cancelled = 1;
1538                 smp_wmb();
1539         } else
1540                 status = data->rpc_status;
1541         rpc_put_task(task);
1542         return status;
1543 }
1544
1545 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1546 {
1547         struct nfs4_opendata *data = calldata;
1548         struct nfs4_state_owner *sp = data->owner;
1549
1550         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1551                 return;
1552         /*
1553          * Check if we still need to send an OPEN call, or if we can use
1554          * a delegation instead.
1555          */
1556         if (data->state != NULL) {
1557                 struct nfs_delegation *delegation;
1558
1559                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1560                         goto out_no_action;
1561                 rcu_read_lock();
1562                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1563                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1564                     can_open_delegated(delegation, data->o_arg.fmode))
1565                         goto unlock_no_action;
1566                 rcu_read_unlock();
1567         }
1568         /* Update client id. */
1569         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1570         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1571                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1572                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1573                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1574         }
1575         data->timestamp = jiffies;
1576         if (nfs4_setup_sequence(data->o_arg.server,
1577                                 &data->o_arg.seq_args,
1578                                 &data->o_res.seq_res,
1579                                 task) != 0)
1580                 nfs_release_seqid(data->o_arg.seqid);
1581         else
1582                 rpc_call_start(task);
1583         return;
1584 unlock_no_action:
1585         rcu_read_unlock();
1586 out_no_action:
1587         task->tk_action = NULL;
1588
1589 }
1590
1591 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1592 {
1593         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1594         nfs4_open_prepare(task, calldata);
1595 }
1596
1597 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1598 {
1599         struct nfs4_opendata *data = calldata;
1600
1601         data->rpc_status = task->tk_status;
1602
1603         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1604                 return;
1605
1606         if (task->tk_status == 0) {
1607                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1608                         switch (data->o_res.f_attr->mode & S_IFMT) {
1609                         case S_IFREG:
1610                                 break;
1611                         case S_IFLNK:
1612                                 data->rpc_status = -ELOOP;
1613                                 break;
1614                         case S_IFDIR:
1615                                 data->rpc_status = -EISDIR;
1616                                 break;
1617                         default:
1618                                 data->rpc_status = -ENOTDIR;
1619                         }
1620                 }
1621                 renew_lease(data->o_res.server, data->timestamp);
1622                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1623                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1624         }
1625         data->rpc_done = 1;
1626 }
1627
1628 static void nfs4_open_release(void *calldata)
1629 {
1630         struct nfs4_opendata *data = calldata;
1631         struct nfs4_state *state = NULL;
1632
1633         /* If this request hasn't been cancelled, do nothing */
1634         if (data->cancelled == 0)
1635                 goto out_free;
1636         /* In case of error, no cleanup! */
1637         if (data->rpc_status != 0 || !data->rpc_done)
1638                 goto out_free;
1639         /* In case we need an open_confirm, no cleanup! */
1640         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1641                 goto out_free;
1642         state = nfs4_opendata_to_nfs4_state(data);
1643         if (!IS_ERR(state))
1644                 nfs4_close_state(state, data->o_arg.fmode);
1645 out_free:
1646         nfs4_opendata_put(data);
1647 }
1648
1649 static const struct rpc_call_ops nfs4_open_ops = {
1650         .rpc_call_prepare = nfs4_open_prepare,
1651         .rpc_call_done = nfs4_open_done,
1652         .rpc_release = nfs4_open_release,
1653 };
1654
1655 static const struct rpc_call_ops nfs4_recover_open_ops = {
1656         .rpc_call_prepare = nfs4_recover_open_prepare,
1657         .rpc_call_done = nfs4_open_done,
1658         .rpc_release = nfs4_open_release,
1659 };
1660
1661 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1662 {
1663         struct inode *dir = data->dir->d_inode;
1664         struct nfs_server *server = NFS_SERVER(dir);
1665         struct nfs_openargs *o_arg = &data->o_arg;
1666         struct nfs_openres *o_res = &data->o_res;
1667         struct rpc_task *task;
1668         struct rpc_message msg = {
1669                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1670                 .rpc_argp = o_arg,
1671                 .rpc_resp = o_res,
1672                 .rpc_cred = data->owner->so_cred,
1673         };
1674         struct rpc_task_setup task_setup_data = {
1675                 .rpc_client = server->client,
1676                 .rpc_message = &msg,
1677                 .callback_ops = &nfs4_open_ops,
1678                 .callback_data = data,
1679                 .workqueue = nfsiod_workqueue,
1680                 .flags = RPC_TASK_ASYNC,
1681         };
1682         int status;
1683
1684         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1685         kref_get(&data->kref);
1686         data->rpc_done = 0;
1687         data->rpc_status = 0;
1688         data->cancelled = 0;
1689         if (isrecover)
1690                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1691         task = rpc_run_task(&task_setup_data);
1692         if (IS_ERR(task))
1693                 return PTR_ERR(task);
1694         status = nfs4_wait_for_completion_rpc_task(task);
1695         if (status != 0) {
1696                 data->cancelled = 1;
1697                 smp_wmb();
1698         } else
1699                 status = data->rpc_status;
1700         rpc_put_task(task);
1701
1702         return status;
1703 }
1704
1705 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1706 {
1707         struct inode *dir = data->dir->d_inode;
1708         struct nfs_openres *o_res = &data->o_res;
1709         int status;
1710
1711         status = nfs4_run_open_task(data, 1);
1712         if (status != 0 || !data->rpc_done)
1713                 return status;
1714
1715         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1716
1717         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1718                 status = _nfs4_proc_open_confirm(data);
1719                 if (status != 0)
1720                         return status;
1721         }
1722
1723         return status;
1724 }
1725
1726 static int nfs4_opendata_access(struct rpc_cred *cred,
1727                                 struct nfs4_opendata *opendata,
1728                                 struct nfs4_state *state, fmode_t fmode)
1729 {
1730         struct nfs_access_entry cache;
1731         u32 mask;
1732
1733         /* access call failed or for some reason the server doesn't
1734          * support any access modes -- defer access call until later */
1735         if (opendata->o_res.access_supported == 0)
1736                 return 0;
1737
1738         mask = 0;
1739         /* don't check MAY_WRITE - a newly created file may not have
1740          * write mode bits, but POSIX allows the creating process to write */
1741         if (fmode & FMODE_READ)
1742                 mask |= MAY_READ;
1743         if (fmode & FMODE_EXEC)
1744                 mask |= MAY_EXEC;
1745
1746         cache.cred = cred;
1747         cache.jiffies = jiffies;
1748         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1749         nfs_access_add_cache(state->inode, &cache);
1750
1751         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1752                 return 0;
1753
1754         /* even though OPEN succeeded, access is denied. Close the file */
1755         nfs4_close_state(state, fmode);
1756         return -EACCES;
1757 }
1758
1759 /*
1760  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1761  */
1762 static int _nfs4_proc_open(struct nfs4_opendata *data)
1763 {
1764         struct inode *dir = data->dir->d_inode;
1765         struct nfs_server *server = NFS_SERVER(dir);
1766         struct nfs_openargs *o_arg = &data->o_arg;
1767         struct nfs_openres *o_res = &data->o_res;
1768         int status;
1769
1770         status = nfs4_run_open_task(data, 0);
1771         if (!data->rpc_done)
1772                 return status;
1773         if (status != 0) {
1774                 if (status == -NFS4ERR_BADNAME &&
1775                                 !(o_arg->open_flags & O_CREAT))
1776                         return -ENOENT;
1777                 return status;
1778         }
1779
1780         nfs_fattr_map_and_free_names(server, &data->f_attr);
1781
1782         if (o_arg->open_flags & O_CREAT)
1783                 update_changeattr(dir, &o_res->cinfo);
1784         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1785                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1786         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1787                 status = _nfs4_proc_open_confirm(data);
1788                 if (status != 0)
1789                         return status;
1790         }
1791         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1792                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1793         return 0;
1794 }
1795
1796 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1797 {
1798         unsigned int loop;
1799         int ret;
1800
1801         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1802                 ret = nfs4_wait_clnt_recover(clp);
1803                 if (ret != 0)
1804                         break;
1805                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1806                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1807                         break;
1808                 nfs4_schedule_state_manager(clp);
1809                 ret = -EIO;
1810         }
1811         return ret;
1812 }
1813
1814 static int nfs4_recover_expired_lease(struct nfs_server *server)
1815 {
1816         return nfs4_client_recover_expired_lease(server->nfs_client);
1817 }
1818
1819 /*
1820  * OPEN_EXPIRED:
1821  *      reclaim state on the server after a network partition.
1822  *      Assumes caller holds the appropriate lock
1823  */
1824 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1825 {
1826         struct nfs4_opendata *opendata;
1827         int ret;
1828
1829         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1830         if (IS_ERR(opendata))
1831                 return PTR_ERR(opendata);
1832         ret = nfs4_open_recover(opendata, state);
1833         if (ret == -ESTALE)
1834                 d_drop(ctx->dentry);
1835         nfs4_opendata_put(opendata);
1836         return ret;
1837 }
1838
1839 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1840 {
1841         struct nfs_server *server = NFS_SERVER(state->inode);
1842         struct nfs4_exception exception = { };
1843         int err;
1844
1845         do {
1846                 err = _nfs4_open_expired(ctx, state);
1847                 switch (err) {
1848                 default:
1849                         goto out;
1850                 case -NFS4ERR_GRACE:
1851                 case -NFS4ERR_DELAY:
1852                         nfs4_handle_exception(server, err, &exception);
1853                         err = 0;
1854                 }
1855         } while (exception.retry);
1856 out:
1857         return err;
1858 }
1859
1860 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1861 {
1862         struct nfs_open_context *ctx;
1863         int ret;
1864
1865         ctx = nfs4_state_find_open_context(state);
1866         if (IS_ERR(ctx))
1867                 return PTR_ERR(ctx);
1868         ret = nfs4_do_open_expired(ctx, state);
1869         put_nfs_open_context(ctx);
1870         return ret;
1871 }
1872
1873 #if defined(CONFIG_NFS_V4_1)
1874 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1875 {
1876         struct nfs_server *server = NFS_SERVER(state->inode);
1877         nfs4_stateid *stateid = &state->stateid;
1878         int status;
1879
1880         /* If a state reset has been done, test_stateid is unneeded */
1881         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1882                 return;
1883
1884         status = nfs41_test_stateid(server, stateid);
1885         if (status != NFS_OK) {
1886                 /* Free the stateid unless the server explicitly
1887                  * informs us the stateid is unrecognized. */
1888                 if (status != -NFS4ERR_BAD_STATEID)
1889                         nfs41_free_stateid(server, stateid);
1890                 nfs_remove_bad_delegation(state->inode);
1891
1892                 write_seqlock(&state->seqlock);
1893                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1894                 write_sequnlock(&state->seqlock);
1895                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1896         }
1897 }
1898
1899 /**
1900  * nfs41_check_open_stateid - possibly free an open stateid
1901  *
1902  * @state: NFSv4 state for an inode
1903  *
1904  * Returns NFS_OK if recovery for this stateid is now finished.
1905  * Otherwise a negative NFS4ERR value is returned.
1906  */
1907 static int nfs41_check_open_stateid(struct nfs4_state *state)
1908 {
1909         struct nfs_server *server = NFS_SERVER(state->inode);
1910         nfs4_stateid *stateid = &state->open_stateid;
1911         int status;
1912
1913         /* If a state reset has been done, test_stateid is unneeded */
1914         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1915             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1916             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1917                 return -NFS4ERR_BAD_STATEID;
1918
1919         status = nfs41_test_stateid(server, stateid);
1920         if (status != NFS_OK) {
1921                 /* Free the stateid unless the server explicitly
1922                  * informs us the stateid is unrecognized. */
1923                 if (status != -NFS4ERR_BAD_STATEID)
1924                         nfs41_free_stateid(server, stateid);
1925
1926                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1927                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1928                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1929         }
1930         return status;
1931 }
1932
1933 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1934 {
1935         int status;
1936
1937         nfs41_clear_delegation_stateid(state);
1938         status = nfs41_check_open_stateid(state);
1939         if (status != NFS_OK)
1940                 status = nfs4_open_expired(sp, state);
1941         return status;
1942 }
1943 #endif
1944
1945 /*
1946  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1947  * fields corresponding to attributes that were used to store the verifier.
1948  * Make sure we clobber those fields in the later setattr call
1949  */
1950 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1951 {
1952         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1953             !(sattr->ia_valid & ATTR_ATIME_SET))
1954                 sattr->ia_valid |= ATTR_ATIME;
1955
1956         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1957             !(sattr->ia_valid & ATTR_MTIME_SET))
1958                 sattr->ia_valid |= ATTR_MTIME;
1959 }
1960
1961 /*
1962  * Returns a referenced nfs4_state
1963  */
1964 static int _nfs4_do_open(struct inode *dir,
1965                         struct dentry *dentry,
1966                         fmode_t fmode,
1967                         int flags,
1968                         struct iattr *sattr,
1969                         struct rpc_cred *cred,
1970                         struct nfs4_state **res,
1971                         struct nfs4_threshold **ctx_th)
1972 {
1973         struct nfs4_state_owner  *sp;
1974         struct nfs4_state     *state = NULL;
1975         struct nfs_server       *server = NFS_SERVER(dir);
1976         struct nfs4_opendata *opendata;
1977         int status;
1978
1979         /* Protect against reboot recovery conflicts */
1980         status = -ENOMEM;
1981         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1982         if (sp == NULL) {
1983                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1984                 goto out_err;
1985         }
1986         status = nfs4_recover_expired_lease(server);
1987         if (status != 0)
1988                 goto err_put_state_owner;
1989         if (dentry->d_inode != NULL)
1990                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1991         status = -ENOMEM;
1992         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1993         if (opendata == NULL)
1994                 goto err_put_state_owner;
1995
1996         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1997                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1998                 if (!opendata->f_attr.mdsthreshold)
1999                         goto err_opendata_put;
2000                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2001         }
2002         if (dentry->d_inode != NULL)
2003                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2004
2005         status = _nfs4_proc_open(opendata);
2006         if (status != 0)
2007                 goto err_opendata_put;
2008
2009         state = nfs4_opendata_to_nfs4_state(opendata);
2010         status = PTR_ERR(state);
2011         if (IS_ERR(state))
2012                 goto err_opendata_put;
2013         if (server->caps & NFS_CAP_POSIX_LOCK)
2014                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2015
2016         status = nfs4_opendata_access(cred, opendata, state, fmode);
2017         if (status != 0)
2018                 goto err_opendata_put;
2019
2020         if (opendata->o_arg.open_flags & O_EXCL) {
2021                 nfs4_exclusive_attrset(opendata, sattr);
2022
2023                 nfs_fattr_init(opendata->o_res.f_attr);
2024                 status = nfs4_do_setattr(state->inode, cred,
2025                                 opendata->o_res.f_attr, sattr,
2026                                 state);
2027                 if (status == 0)
2028                         nfs_setattr_update_inode(state->inode, sattr);
2029                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2030         }
2031
2032         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2033                 *ctx_th = opendata->f_attr.mdsthreshold;
2034         else
2035                 kfree(opendata->f_attr.mdsthreshold);
2036         opendata->f_attr.mdsthreshold = NULL;
2037
2038         nfs4_opendata_put(opendata);
2039         nfs4_put_state_owner(sp);
2040         *res = state;
2041         return 0;
2042 err_opendata_put:
2043         kfree(opendata->f_attr.mdsthreshold);
2044         nfs4_opendata_put(opendata);
2045 err_put_state_owner:
2046         nfs4_put_state_owner(sp);
2047 out_err:
2048         *res = NULL;
2049         return status;
2050 }
2051
2052
2053 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2054                                         struct dentry *dentry,
2055                                         fmode_t fmode,
2056                                         int flags,
2057                                         struct iattr *sattr,
2058                                         struct rpc_cred *cred,
2059                                         struct nfs4_threshold **ctx_th)
2060 {
2061         struct nfs4_exception exception = { };
2062         struct nfs4_state *res;
2063         int status;
2064
2065         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
2066         do {
2067                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
2068                                        &res, ctx_th);
2069                 if (status == 0)
2070                         break;
2071                 /* NOTE: BAD_SEQID means the server and client disagree about the
2072                  * book-keeping w.r.t. state-changing operations
2073                  * (OPEN/CLOSE/LOCK/LOCKU...)
2074                  * It is actually a sign of a bug on the client or on the server.
2075                  *
2076                  * If we receive a BAD_SEQID error in the particular case of
2077                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2078                  * have unhashed the old state_owner for us, and that we can
2079                  * therefore safely retry using a new one. We should still warn
2080                  * the user though...
2081                  */
2082                 if (status == -NFS4ERR_BAD_SEQID) {
2083                         pr_warn_ratelimited("NFS: v4 server %s "
2084                                         " returned a bad sequence-id error!\n",
2085                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2086                         exception.retry = 1;
2087                         continue;
2088                 }
2089                 /*
2090                  * BAD_STATEID on OPEN means that the server cancelled our
2091                  * state before it received the OPEN_CONFIRM.
2092                  * Recover by retrying the request as per the discussion
2093                  * on Page 181 of RFC3530.
2094                  */
2095                 if (status == -NFS4ERR_BAD_STATEID) {
2096                         exception.retry = 1;
2097                         continue;
2098                 }
2099                 if (status == -EAGAIN) {
2100                         /* We must have found a delegation */
2101                         exception.retry = 1;
2102                         continue;
2103                 }
2104                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
2105                                         status, &exception));
2106         } while (exception.retry);
2107         return res;
2108 }
2109
2110 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2111                             struct nfs_fattr *fattr, struct iattr *sattr,
2112                             struct nfs4_state *state)
2113 {
2114         struct nfs_server *server = NFS_SERVER(inode);
2115         struct nfs_setattrargs  arg = {
2116                 .fh             = NFS_FH(inode),
2117                 .iap            = sattr,
2118                 .server         = server,
2119                 .bitmask = server->attr_bitmask,
2120         };
2121         struct nfs_setattrres  res = {
2122                 .fattr          = fattr,
2123                 .server         = server,
2124         };
2125         struct rpc_message msg = {
2126                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2127                 .rpc_argp       = &arg,
2128                 .rpc_resp       = &res,
2129                 .rpc_cred       = cred,
2130         };
2131         unsigned long timestamp = jiffies;
2132         int status;
2133
2134         nfs_fattr_init(fattr);
2135
2136         if (state != NULL) {
2137                 struct nfs_lockowner lockowner = {
2138                         .l_owner = current->files,
2139                         .l_pid = current->tgid,
2140                 };
2141                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2142                                 &lockowner);
2143         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
2144                                 FMODE_WRITE)) {
2145                 /* Use that stateid */
2146         } else
2147                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2148
2149         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2150         if (status == 0 && state != NULL)
2151                 renew_lease(server, timestamp);
2152         return status;
2153 }
2154
2155 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2156                            struct nfs_fattr *fattr, struct iattr *sattr,
2157                            struct nfs4_state *state)
2158 {
2159         struct nfs_server *server = NFS_SERVER(inode);
2160         struct nfs4_exception exception = {
2161                 .state = state,
2162                 .inode = inode,
2163         };
2164         int err;
2165         do {
2166                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2167                 switch (err) {
2168                 case -NFS4ERR_OPENMODE:
2169                         if (state && !(state->state & FMODE_WRITE)) {
2170                                 err = -EBADF;
2171                                 if (sattr->ia_valid & ATTR_OPEN)
2172                                         err = -EACCES;
2173                                 goto out;
2174                         }
2175                 }
2176                 err = nfs4_handle_exception(server, err, &exception);
2177         } while (exception.retry);
2178 out:
2179         return err;
2180 }
2181
2182 struct nfs4_closedata {
2183         struct inode *inode;
2184         struct nfs4_state *state;
2185         struct nfs_closeargs arg;
2186         struct nfs_closeres res;
2187         struct nfs_fattr fattr;
2188         unsigned long timestamp;
2189         bool roc;
2190         u32 roc_barrier;
2191 };
2192
2193 static void nfs4_free_closedata(void *data)
2194 {
2195         struct nfs4_closedata *calldata = data;
2196         struct nfs4_state_owner *sp = calldata->state->owner;
2197         struct super_block *sb = calldata->state->inode->i_sb;
2198
2199         if (calldata->roc)
2200                 pnfs_roc_release(calldata->state->inode);
2201         nfs4_put_open_state(calldata->state);
2202         nfs_free_seqid(calldata->arg.seqid);
2203         nfs4_put_state_owner(sp);
2204         nfs_sb_deactive_async(sb);
2205         kfree(calldata);
2206 }
2207
2208 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2209                 fmode_t fmode)
2210 {
2211         spin_lock(&state->owner->so_lock);
2212         if (!(fmode & FMODE_READ))
2213                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2214         if (!(fmode & FMODE_WRITE))
2215                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2216         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2217         spin_unlock(&state->owner->so_lock);
2218 }
2219
2220 static void nfs4_close_done(struct rpc_task *task, void *data)
2221 {
2222         struct nfs4_closedata *calldata = data;
2223         struct nfs4_state *state = calldata->state;
2224         struct nfs_server *server = NFS_SERVER(calldata->inode);
2225
2226         dprintk("%s: begin!\n", __func__);
2227         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2228                 return;
2229         /* hmm. we are done with the inode, and in the process of freeing
2230          * the state_owner. we keep this around to process errors
2231          */
2232         switch (task->tk_status) {
2233                 case 0:
2234                         if (calldata->roc)
2235                                 pnfs_roc_set_barrier(state->inode,
2236                                                      calldata->roc_barrier);
2237                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2238                         renew_lease(server, calldata->timestamp);
2239                         nfs4_close_clear_stateid_flags(state,
2240                                         calldata->arg.fmode);
2241                         break;
2242                 case -NFS4ERR_STALE_STATEID:
2243                 case -NFS4ERR_OLD_STATEID:
2244                 case -NFS4ERR_BAD_STATEID:
2245                 case -NFS4ERR_EXPIRED:
2246                         if (calldata->arg.fmode == 0)
2247                                 break;
2248                 default:
2249                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2250                                 rpc_restart_call_prepare(task);
2251         }
2252         nfs_release_seqid(calldata->arg.seqid);
2253         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2254         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2255 }
2256
2257 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2258 {
2259         struct nfs4_closedata *calldata = data;
2260         struct nfs4_state *state = calldata->state;
2261         struct inode *inode = calldata->inode;
2262         int call_close = 0;
2263
2264         dprintk("%s: begin!\n", __func__);
2265         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2266                 return;
2267
2268         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2269         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2270         spin_lock(&state->owner->so_lock);
2271         /* Calculate the change in open mode */
2272         if (state->n_rdwr == 0) {
2273                 if (state->n_rdonly == 0) {
2274                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2275                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2276                         calldata->arg.fmode &= ~FMODE_READ;
2277                 }
2278                 if (state->n_wronly == 0) {
2279                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2280                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2281                         calldata->arg.fmode &= ~FMODE_WRITE;
2282                 }
2283         }
2284         spin_unlock(&state->owner->so_lock);
2285
2286         if (!call_close) {
2287                 /* Note: exit _without_ calling nfs4_close_done */
2288                 task->tk_action = NULL;
2289                 goto out;
2290         }
2291
2292         if (calldata->arg.fmode == 0) {
2293                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2294                 if (calldata->roc &&
2295                     pnfs_roc_drain(inode, &calldata->roc_barrier, task))
2296                         goto out;
2297         }
2298
2299         nfs_fattr_init(calldata->res.fattr);
2300         calldata->timestamp = jiffies;
2301         if (nfs4_setup_sequence(NFS_SERVER(inode),
2302                                 &calldata->arg.seq_args,
2303                                 &calldata->res.seq_res,
2304                                 task) != 0)
2305                 nfs_release_seqid(calldata->arg.seqid);
2306         else
2307                 rpc_call_start(task);
2308 out:
2309         dprintk("%s: done!\n", __func__);
2310 }
2311
2312 static const struct rpc_call_ops nfs4_close_ops = {
2313         .rpc_call_prepare = nfs4_close_prepare,
2314         .rpc_call_done = nfs4_close_done,
2315         .rpc_release = nfs4_free_closedata,
2316 };
2317
2318 /* 
2319  * It is possible for data to be read/written from a mem-mapped file 
2320  * after the sys_close call (which hits the vfs layer as a flush).
2321  * This means that we can't safely call nfsv4 close on a file until 
2322  * the inode is cleared. This in turn means that we are not good
2323  * NFSv4 citizens - we do not indicate to the server to update the file's 
2324  * share state even when we are done with one of the three share 
2325  * stateid's in the inode.
2326  *
2327  * NOTE: Caller must be holding the sp->so_owner semaphore!
2328  */
2329 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2330 {
2331         struct nfs_server *server = NFS_SERVER(state->inode);
2332         struct nfs4_closedata *calldata;
2333         struct nfs4_state_owner *sp = state->owner;
2334         struct rpc_task *task;
2335         struct rpc_message msg = {
2336                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2337                 .rpc_cred = state->owner->so_cred,
2338         };
2339         struct rpc_task_setup task_setup_data = {
2340                 .rpc_client = server->client,
2341                 .rpc_message = &msg,
2342                 .callback_ops = &nfs4_close_ops,
2343                 .workqueue = nfsiod_workqueue,
2344                 .flags = RPC_TASK_ASYNC,
2345         };
2346         int status = -ENOMEM;
2347
2348         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2349         if (calldata == NULL)
2350                 goto out;
2351         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2352         calldata->inode = state->inode;
2353         calldata->state = state;
2354         calldata->arg.fh = NFS_FH(state->inode);
2355         calldata->arg.stateid = &state->open_stateid;
2356         /* Serialization for the sequence id */
2357         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2358         if (calldata->arg.seqid == NULL)
2359                 goto out_free_calldata;
2360         calldata->arg.fmode = 0;
2361         calldata->arg.bitmask = server->cache_consistency_bitmask;
2362         calldata->res.fattr = &calldata->fattr;
2363         calldata->res.seqid = calldata->arg.seqid;
2364         calldata->res.server = server;
2365         calldata->roc = pnfs_roc(state->inode);
2366         nfs_sb_active(calldata->inode->i_sb);
2367
2368         msg.rpc_argp = &calldata->arg;
2369         msg.rpc_resp = &calldata->res;
2370         task_setup_data.callback_data = calldata;
2371         task = rpc_run_task(&task_setup_data);
2372         if (IS_ERR(task))
2373                 return PTR_ERR(task);
2374         status = 0;
2375         if (wait)
2376                 status = rpc_wait_for_completion_task(task);
2377         rpc_put_task(task);
2378         return status;
2379 out_free_calldata:
2380         kfree(calldata);
2381 out:
2382         nfs4_put_open_state(state);
2383         nfs4_put_state_owner(sp);
2384         return status;
2385 }
2386
2387 static struct inode *
2388 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2389 {
2390         struct nfs4_state *state;
2391
2392         /* Protect against concurrent sillydeletes */
2393         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2394                              ctx->cred, &ctx->mdsthreshold);
2395         if (IS_ERR(state))
2396                 return ERR_CAST(state);
2397         ctx->state = state;
2398         return igrab(state->inode);
2399 }
2400
2401 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2402 {
2403         if (ctx->state == NULL)
2404                 return;
2405         if (is_sync)
2406                 nfs4_close_sync(ctx->state, ctx->mode);
2407         else
2408                 nfs4_close_state(ctx->state, ctx->mode);
2409 }
2410
2411 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2412 {
2413         struct nfs4_server_caps_arg args = {
2414                 .fhandle = fhandle,
2415         };
2416         struct nfs4_server_caps_res res = {};
2417         struct rpc_message msg = {
2418                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2419                 .rpc_argp = &args,
2420                 .rpc_resp = &res,
2421         };
2422         int status;
2423
2424         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2425         if (status == 0) {
2426                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2427                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2428                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2429                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2430                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2431                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2432                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2433                         server->caps |= NFS_CAP_ACLS;
2434                 if (res.has_links != 0)
2435                         server->caps |= NFS_CAP_HARDLINKS;
2436                 if (res.has_symlinks != 0)
2437                         server->caps |= NFS_CAP_SYMLINKS;
2438                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2439                         server->caps |= NFS_CAP_FILEID;
2440                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2441                         server->caps |= NFS_CAP_MODE;
2442                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2443                         server->caps |= NFS_CAP_NLINK;
2444                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2445                         server->caps |= NFS_CAP_OWNER;
2446                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2447                         server->caps |= NFS_CAP_OWNER_GROUP;
2448                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2449                         server->caps |= NFS_CAP_ATIME;
2450                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2451                         server->caps |= NFS_CAP_CTIME;
2452                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2453                         server->caps |= NFS_CAP_MTIME;
2454
2455                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2456                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2457                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2458                 server->acl_bitmask = res.acl_bitmask;
2459                 server->fh_expire_type = res.fh_expire_type;
2460         }
2461
2462         return status;
2463 }
2464
2465 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2466 {
2467         struct nfs4_exception exception = { };
2468         int err;
2469         do {
2470                 err = nfs4_handle_exception(server,
2471                                 _nfs4_server_capabilities(server, fhandle),
2472                                 &exception);
2473         } while (exception.retry);
2474         return err;
2475 }
2476
2477 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2478                 struct nfs_fsinfo *info)
2479 {
2480         struct nfs4_lookup_root_arg args = {
2481                 .bitmask = nfs4_fattr_bitmap,
2482         };
2483         struct nfs4_lookup_res res = {
2484                 .server = server,
2485                 .fattr = info->fattr,
2486                 .fh = fhandle,
2487         };
2488         struct rpc_message msg = {
2489                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2490                 .rpc_argp = &args,
2491                 .rpc_resp = &res,
2492         };
2493
2494         nfs_fattr_init(info->fattr);
2495         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2496 }
2497
2498 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2499                 struct nfs_fsinfo *info)
2500 {
2501         struct nfs4_exception exception = { };
2502         int err;
2503         do {
2504                 err = _nfs4_lookup_root(server, fhandle, info);
2505                 switch (err) {
2506                 case 0:
2507                 case -NFS4ERR_WRONGSEC:
2508                         goto out;
2509                 default:
2510                         err = nfs4_handle_exception(server, err, &exception);
2511                 }
2512         } while (exception.retry);
2513 out:
2514         return err;
2515 }
2516
2517 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2518                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2519 {
2520         struct rpc_auth *auth;
2521         int ret;
2522
2523         auth = rpcauth_create(flavor, server->client);
2524         if (IS_ERR(auth)) {
2525                 ret = -EIO;
2526                 goto out;
2527         }
2528         ret = nfs4_lookup_root(server, fhandle, info);
2529 out:
2530         return ret;
2531 }
2532
2533 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2534                               struct nfs_fsinfo *info)
2535 {
2536         int i, len, status = 0;
2537         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2538
2539         len = rpcauth_list_flavors(flav_array, ARRAY_SIZE(flav_array));
2540         if (len < 0)
2541                 return len;
2542
2543         for (i = 0; i < len; i++) {
2544                 /* AUTH_UNIX is the default flavor if none was specified,
2545                  * thus has already been tried. */
2546                 if (flav_array[i] == RPC_AUTH_UNIX)
2547                         continue;
2548
2549                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2550                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2551                         continue;
2552                 break;
2553         }
2554         /*
2555          * -EACCESS could mean that the user doesn't have correct permissions
2556          * to access the mount.  It could also mean that we tried to mount
2557          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2558          * existing mount programs don't handle -EACCES very well so it should
2559          * be mapped to -EPERM instead.
2560          */
2561         if (status == -EACCES)
2562                 status = -EPERM;
2563         return status;
2564 }
2565
2566 /*
2567  * get the file handle for the "/" directory on the server
2568  */
2569 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2570                          struct nfs_fsinfo *info)
2571 {
2572         int minor_version = server->nfs_client->cl_minorversion;
2573         int status = nfs4_lookup_root(server, fhandle, info);
2574         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2575                 /*
2576                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2577                  * by nfs4_map_errors() as this function exits.
2578                  */
2579                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2580         if (status == 0)
2581                 status = nfs4_server_capabilities(server, fhandle);
2582         if (status == 0)
2583                 status = nfs4_do_fsinfo(server, fhandle, info);
2584         return nfs4_map_errors(status);
2585 }
2586
2587 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2588                               struct nfs_fsinfo *info)
2589 {
2590         int error;
2591         struct nfs_fattr *fattr = info->fattr;
2592
2593         error = nfs4_server_capabilities(server, mntfh);
2594         if (error < 0) {
2595                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2596                 return error;
2597         }
2598
2599         error = nfs4_proc_getattr(server, mntfh, fattr);
2600         if (error < 0) {
2601                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2602                 return error;
2603         }
2604
2605         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2606             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2607                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2608
2609         return error;
2610 }
2611
2612 /*
2613  * Get locations and (maybe) other attributes of a referral.
2614  * Note that we'll actually follow the referral later when
2615  * we detect fsid mismatch in inode revalidation
2616  */
2617 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2618                              const struct qstr *name, struct nfs_fattr *fattr,
2619                              struct nfs_fh *fhandle)
2620 {
2621         int status = -ENOMEM;
2622         struct page *page = NULL;
2623         struct nfs4_fs_locations *locations = NULL;
2624
2625         page = alloc_page(GFP_KERNEL);
2626         if (page == NULL)
2627                 goto out;
2628         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2629         if (locations == NULL)
2630                 goto out;
2631
2632         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2633         if (status != 0)
2634                 goto out;
2635         /* Make sure server returned a different fsid for the referral */
2636         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2637                 dprintk("%s: server did not return a different fsid for"
2638                         " a referral at %s\n", __func__, name->name);
2639                 status = -EIO;
2640                 goto out;
2641         }
2642         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2643         nfs_fixup_referral_attributes(&locations->fattr);
2644
2645         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2646         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2647         memset(fhandle, 0, sizeof(struct nfs_fh));
2648 out:
2649         if (page)
2650                 __free_page(page);
2651         kfree(locations);
2652         return status;
2653 }
2654
2655 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2656 {
2657         struct nfs4_getattr_arg args = {
2658                 .fh = fhandle,
2659                 .bitmask = server->attr_bitmask,
2660         };
2661         struct nfs4_getattr_res res = {
2662                 .fattr = fattr,
2663                 .server = server,
2664         };
2665         struct rpc_message msg = {
2666                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2667                 .rpc_argp = &args,
2668                 .rpc_resp = &res,
2669         };
2670         
2671         nfs_fattr_init(fattr);
2672         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2673 }
2674
2675 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2676 {
2677         struct nfs4_exception exception = { };
2678         int err;
2679         do {
2680                 err = nfs4_handle_exception(server,
2681                                 _nfs4_proc_getattr(server, fhandle, fattr),
2682                                 &exception);
2683         } while (exception.retry);
2684         return err;
2685 }
2686
2687 /* 
2688  * The file is not closed if it is opened due to the a request to change
2689  * the size of the file. The open call will not be needed once the
2690  * VFS layer lookup-intents are implemented.
2691  *
2692  * Close is called when the inode is destroyed.
2693  * If we haven't opened the file for O_WRONLY, we
2694  * need to in the size_change case to obtain a stateid.
2695  *
2696  * Got race?
2697  * Because OPEN is always done by name in nfsv4, it is
2698  * possible that we opened a different file by the same
2699  * name.  We can recognize this race condition, but we
2700  * can't do anything about it besides returning an error.
2701  *
2702  * This will be fixed with VFS changes (lookup-intent).
2703  */
2704 static int
2705 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2706                   struct iattr *sattr)
2707 {
2708         struct inode *inode = dentry->d_inode;
2709         struct rpc_cred *cred = NULL;
2710         struct nfs4_state *state = NULL;
2711         int status;
2712
2713         if (pnfs_ld_layoutret_on_setattr(inode))
2714                 pnfs_return_layout(inode);
2715
2716         nfs_fattr_init(fattr);
2717         
2718         /* Deal with open(O_TRUNC) */
2719         if (sattr->ia_valid & ATTR_OPEN)
2720                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2721
2722         /* Optimization: if the end result is no change, don't RPC */
2723         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2724                 return 0;
2725
2726         /* Search for an existing open(O_WRITE) file */
2727         if (sattr->ia_valid & ATTR_FILE) {
2728                 struct nfs_open_context *ctx;
2729
2730                 ctx = nfs_file_open_context(sattr->ia_file);
2731                 if (ctx) {
2732                         cred = ctx->cred;
2733                         state = ctx->state;
2734                 }
2735         }
2736
2737         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2738         if (status == 0)
2739                 nfs_setattr_update_inode(inode, sattr);
2740         return status;
2741 }
2742
2743 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2744                 const struct qstr *name, struct nfs_fh *fhandle,
2745                 struct nfs_fattr *fattr)
2746 {
2747         struct nfs_server *server = NFS_SERVER(dir);
2748         int                    status;
2749         struct nfs4_lookup_arg args = {
2750                 .bitmask = server->attr_bitmask,
2751                 .dir_fh = NFS_FH(dir),
2752                 .name = name,
2753         };
2754         struct nfs4_lookup_res res = {
2755                 .server = server,
2756                 .fattr = fattr,
2757                 .fh = fhandle,
2758         };
2759         struct rpc_message msg = {
2760                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2761                 .rpc_argp = &args,
2762                 .rpc_resp = &res,
2763         };
2764
2765         nfs_fattr_init(fattr);
2766
2767         dprintk("NFS call  lookup %s\n", name->name);
2768         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2769         dprintk("NFS reply lookup: %d\n", status);
2770         return status;
2771 }
2772
2773 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2774 {
2775         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2776                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2777         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2778         fattr->nlink = 2;
2779 }
2780
2781 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2782                                    struct qstr *name, struct nfs_fh *fhandle,
2783                                    struct nfs_fattr *fattr)
2784 {
2785         struct nfs4_exception exception = { };
2786         struct rpc_clnt *client = *clnt;
2787         int err;
2788         do {
2789                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2790                 switch (err) {
2791                 case -NFS4ERR_BADNAME:
2792                         err = -ENOENT;
2793                         goto out;
2794                 case -NFS4ERR_MOVED:
2795                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2796                         goto out;
2797                 case -NFS4ERR_WRONGSEC:
2798                         err = -EPERM;
2799                         if (client != *clnt)
2800                                 goto out;
2801
2802                         client = nfs4_create_sec_client(client, dir, name);
2803                         if (IS_ERR(client))
2804                                 return PTR_ERR(client);
2805
2806                         exception.retry = 1;
2807                         break;
2808                 default:
2809                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2810                 }
2811         } while (exception.retry);
2812
2813 out:
2814         if (err == 0)
2815                 *clnt = client;
2816         else if (client != *clnt)
2817                 rpc_shutdown_client(client);
2818
2819         return err;
2820 }
2821
2822 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2823                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2824 {
2825         int status;
2826         struct rpc_clnt *client = NFS_CLIENT(dir);
2827
2828         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2829         if (client != NFS_CLIENT(dir)) {
2830                 rpc_shutdown_client(client);
2831                 nfs_fixup_secinfo_attributes(fattr);
2832         }
2833         return status;
2834 }
2835
2836 struct rpc_clnt *
2837 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2838                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2839 {
2840         int status;
2841         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2842
2843         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2844         if (status < 0) {
2845                 rpc_shutdown_client(client);
2846                 return ERR_PTR(status);
2847         }
2848         return client;
2849 }
2850
2851 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2852 {
2853         struct nfs_server *server = NFS_SERVER(inode);
2854         struct nfs4_accessargs args = {
2855                 .fh = NFS_FH(inode),
2856                 .bitmask = server->cache_consistency_bitmask,
2857         };
2858         struct nfs4_accessres res = {
2859                 .server = server,
2860         };
2861         struct rpc_message msg = {
2862                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2863                 .rpc_argp = &args,
2864                 .rpc_resp = &res,
2865                 .rpc_cred = entry->cred,
2866         };
2867         int mode = entry->mask;
2868         int status;
2869
2870         /*
2871          * Determine which access bits we want to ask for...
2872          */
2873         if (mode & MAY_READ)
2874                 args.access |= NFS4_ACCESS_READ;
2875         if (S_ISDIR(inode->i_mode)) {
2876                 if (mode & MAY_WRITE)
2877                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2878                 if (mode & MAY_EXEC)
2879                         args.access |= NFS4_ACCESS_LOOKUP;
2880         } else {
2881                 if (mode & MAY_WRITE)
2882                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2883                 if (mode & MAY_EXEC)
2884                         args.access |= NFS4_ACCESS_EXECUTE;
2885         }
2886
2887         res.fattr = nfs_alloc_fattr();
2888         if (res.fattr == NULL)
2889                 return -ENOMEM;
2890
2891         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2892         if (!status) {
2893                 nfs_access_set_mask(entry, res.access);
2894                 nfs_refresh_inode(inode, res.fattr);
2895         }
2896         nfs_free_fattr(res.fattr);
2897         return status;
2898 }
2899
2900 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2901 {
2902         struct nfs4_exception exception = { };
2903         int err;
2904         do {
2905                 err = nfs4_handle_exception(NFS_SERVER(inode),
2906                                 _nfs4_proc_access(inode, entry),
2907                                 &exception);
2908         } while (exception.retry);
2909         return err;
2910 }
2911
2912 /*
2913  * TODO: For the time being, we don't try to get any attributes
2914  * along with any of the zero-copy operations READ, READDIR,
2915  * READLINK, WRITE.
2916  *
2917  * In the case of the first three, we want to put the GETATTR
2918  * after the read-type operation -- this is because it is hard
2919  * to predict the length of a GETATTR response in v4, and thus
2920  * align the READ data correctly.  This means that the GETATTR
2921  * may end up partially falling into the page cache, and we should
2922  * shift it into the 'tail' of the xdr_buf before processing.
2923  * To do this efficiently, we need to know the total length
2924  * of data received, which doesn't seem to be available outside
2925  * of the RPC layer.
2926  *
2927  * In the case of WRITE, we also want to put the GETATTR after
2928  * the operation -- in this case because we want to make sure
2929  * we get the post-operation mtime and size.
2930  *
2931  * Both of these changes to the XDR layer would in fact be quite
2932  * minor, but I decided to leave them for a subsequent patch.
2933  */
2934 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2935                 unsigned int pgbase, unsigned int pglen)
2936 {
2937         struct nfs4_readlink args = {
2938                 .fh       = NFS_FH(inode),
2939                 .pgbase   = pgbase,
2940                 .pglen    = pglen,
2941                 .pages    = &page,
2942         };
2943         struct nfs4_readlink_res res;
2944         struct rpc_message msg = {
2945                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2946                 .rpc_argp = &args,
2947                 .rpc_resp = &res,
2948         };
2949
2950         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2951 }
2952
2953 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2954                 unsigned int pgbase, unsigned int pglen)
2955 {
2956         struct nfs4_exception exception = { };
2957         int err;
2958         do {
2959                 err = nfs4_handle_exception(NFS_SERVER(inode),
2960                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2961                                 &exception);
2962         } while (exception.retry);
2963         return err;
2964 }
2965
2966 /*
2967  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2968  */
2969 static int
2970 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2971                  int flags)
2972 {
2973         struct nfs_open_context *ctx;
2974         struct nfs4_state *state;
2975         int status = 0;
2976
2977         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2978         if (IS_ERR(ctx))
2979                 return PTR_ERR(ctx);
2980
2981         sattr->ia_mode &= ~current_umask();
2982         state = nfs4_do_open(dir, dentry, ctx->mode,
2983                         flags, sattr, ctx->cred,
2984                         &ctx->mdsthreshold);
2985         d_drop(dentry);
2986         if (IS_ERR(state)) {
2987                 status = PTR_ERR(state);
2988                 goto out;
2989         }
2990         d_add(dentry, igrab(state->inode));
2991         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2992         ctx->state = state;
2993 out:
2994         put_nfs_open_context(ctx);
2995         return status;
2996 }
2997
2998 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2999 {
3000         struct nfs_server *server = NFS_SERVER(dir);
3001         struct nfs_removeargs args = {
3002                 .fh = NFS_FH(dir),
3003                 .name = *name,
3004         };
3005         struct nfs_removeres res = {
3006                 .server = server,
3007         };
3008         struct rpc_message msg = {
3009                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3010                 .rpc_argp = &args,
3011                 .rpc_resp = &res,
3012         };
3013         int status;
3014
3015         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3016         if (status == 0)
3017                 update_changeattr(dir, &res.cinfo);
3018         return status;
3019 }
3020
3021 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3022 {
3023         struct nfs4_exception exception = { };
3024         int err;
3025         do {
3026                 err = nfs4_handle_exception(NFS_SERVER(dir),
3027                                 _nfs4_proc_remove(dir, name),
3028                                 &exception);
3029         } while (exception.retry);
3030         return err;
3031 }
3032
3033 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3034 {
3035         struct nfs_server *server = NFS_SERVER(dir);
3036         struct nfs_removeargs *args = msg->rpc_argp;
3037         struct nfs_removeres *res = msg->rpc_resp;
3038
3039         res->server = server;
3040         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3041         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3042 }
3043
3044 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3045 {
3046         if (nfs4_setup_sequence(NFS_SERVER(data->dir),
3047                                 &data->args.seq_args,
3048                                 &data->res.seq_res,
3049                                 task))
3050                 return;
3051         rpc_call_start(task);
3052 }
3053
3054 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3055 {
3056         struct nfs_removeres *res = task->tk_msg.rpc_resp;
3057
3058         if (!nfs4_sequence_done(task, &res->seq_res))
3059                 return 0;
3060         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3061                 return 0;
3062         update_changeattr(dir, &res->cinfo);
3063         return 1;
3064 }
3065
3066 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3067 {
3068         struct nfs_server *server = NFS_SERVER(dir);
3069         struct nfs_renameargs *arg = msg->rpc_argp;
3070         struct nfs_renameres *res = msg->rpc_resp;
3071
3072         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3073         res->server = server;
3074         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3075 }
3076
3077 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3078 {
3079         if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3080                                 &data->args.seq_args,
3081                                 &data->res.seq_res,
3082                                 task))
3083                 return;
3084         rpc_call_start(task);
3085 }
3086
3087 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3088                                  struct inode *new_dir)
3089 {
3090         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3091
3092         if (!nfs4_sequence_done(task, &res->seq_res))
3093                 return 0;
3094         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3095                 return 0;
3096
3097         update_changeattr(old_dir, &res->old_cinfo);
3098         update_changeattr(new_dir, &res->new_cinfo);
3099         return 1;
3100 }
3101
3102 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3103                 struct inode *new_dir, struct qstr *new_name)
3104 {
3105         struct nfs_server *server = NFS_SERVER(old_dir);
3106         struct nfs_renameargs arg = {
3107                 .old_dir = NFS_FH(old_dir),
3108                 .new_dir = NFS_FH(new_dir),
3109                 .old_name = old_name,
3110                 .new_name = new_name,
3111         };
3112         struct nfs_renameres res = {
3113                 .server = server,
3114         };
3115         struct rpc_message msg = {
3116                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3117                 .rpc_argp = &arg,
3118                 .rpc_resp = &res,
3119         };
3120         int status = -ENOMEM;
3121         
3122         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3123         if (!status) {
3124                 update_changeattr(old_dir, &res.old_cinfo);
3125                 update_changeattr(new_dir, &res.new_cinfo);
3126         }
3127         return status;
3128 }
3129
3130 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3131                 struct inode *new_dir, struct qstr *new_name)
3132 {
3133         struct nfs4_exception exception = { };
3134         int err;
3135         do {
3136                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3137                                 _nfs4_proc_rename(old_dir, old_name,
3138                                         new_dir, new_name),
3139                                 &exception);
3140         } while (exception.retry);
3141         return err;
3142 }
3143
3144 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3145 {
3146         struct nfs_server *server = NFS_SERVER(inode);
3147         struct nfs4_link_arg arg = {
3148                 .fh     = NFS_FH(inode),
3149                 .dir_fh = NFS_FH(dir),
3150                 .name   = name,
3151                 .bitmask = server->attr_bitmask,
3152         };
3153         struct nfs4_link_res res = {
3154                 .server = server,
3155         };
3156         struct rpc_message msg = {
3157                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3158                 .rpc_argp = &arg,
3159                 .rpc_resp = &res,
3160         };
3161         int status = -ENOMEM;
3162
3163         res.fattr = nfs_alloc_fattr();
3164         if (res.fattr == NULL)
3165                 goto out;
3166
3167         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3168         if (!status) {
3169                 update_changeattr(dir, &res.cinfo);
3170                 nfs_post_op_update_inode(inode, res.fattr);
3171         }
3172 out:
3173         nfs_free_fattr(res.fattr);
3174         return status;
3175 }
3176
3177 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3178 {
3179         struct nfs4_exception exception = { };
3180         int err;
3181         do {
3182                 err = nfs4_handle_exception(NFS_SERVER(inode),
3183                                 _nfs4_proc_link(inode, dir, name),
3184                                 &exception);
3185         } while (exception.retry);
3186         return err;
3187 }
3188
3189 struct nfs4_createdata {
3190         struct rpc_message msg;
3191         struct nfs4_create_arg arg;
3192         struct nfs4_create_res res;
3193         struct nfs_fh fh;
3194         struct nfs_fattr fattr;
3195 };
3196
3197 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3198                 struct qstr *name, struct iattr *sattr, u32 ftype)
3199 {
3200         struct nfs4_createdata *data;
3201
3202         data = kzalloc(sizeof(*data), GFP_KERNEL);
3203         if (data != NULL) {
3204                 struct nfs_server *server = NFS_SERVER(dir);
3205
3206                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3207                 data->msg.rpc_argp = &data->arg;
3208                 data->msg.rpc_resp = &data->res;
3209                 data->arg.dir_fh = NFS_FH(dir);
3210                 data->arg.server = server;
3211                 data->arg.name = name;
3212                 data->arg.attrs = sattr;
3213                 data->arg.ftype = ftype;
3214                 data->arg.bitmask = server->attr_bitmask;
3215                 data->res.server = server;
3216                 data->res.fh = &data->fh;
3217                 data->res.fattr = &data->fattr;
3218                 nfs_fattr_init(data->res.fattr);
3219         }
3220         return data;
3221 }
3222
3223 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3224 {
3225         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3226                                     &data->arg.seq_args, &data->res.seq_res, 1);
3227         if (status == 0) {
3228                 update_changeattr(dir, &data->res.dir_cinfo);
3229                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3230         }
3231         return status;
3232 }
3233
3234 static void nfs4_free_createdata(struct nfs4_createdata *data)
3235 {
3236         kfree(data);
3237 }
3238
3239 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3240                 struct page *page, unsigned int len, struct iattr *sattr)
3241 {
3242         struct nfs4_createdata *data;
3243         int status = -ENAMETOOLONG;
3244
3245         if (len > NFS4_MAXPATHLEN)
3246                 goto out;
3247
3248         status = -ENOMEM;
3249         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3250         if (data == NULL)
3251                 goto out;
3252
3253         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3254         data->arg.u.symlink.pages = &page;
3255         data->arg.u.symlink.len = len;
3256         
3257         status = nfs4_do_create(dir, dentry, data);
3258
3259         nfs4_free_createdata(data);
3260 out:
3261         return status;
3262 }
3263
3264 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3265                 struct page *page, unsigned int len, struct iattr *sattr)
3266 {
3267         struct nfs4_exception exception = { };
3268         int err;
3269         do {
3270                 err = nfs4_handle_exception(NFS_SERVER(dir),
3271                                 _nfs4_proc_symlink(dir, dentry, page,
3272                                                         len, sattr),
3273                                 &exception);
3274         } while (exception.retry);
3275         return err;
3276 }
3277
3278 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3279                 struct iattr *sattr)
3280 {
3281         struct nfs4_createdata *data;
3282         int status = -ENOMEM;
3283
3284         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3285         if (data == NULL)
3286                 goto out;
3287
3288         status = nfs4_do_create(dir, dentry, data);
3289
3290         nfs4_free_createdata(data);
3291 out:
3292         return status;
3293 }
3294
3295 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3296                 struct iattr *sattr)
3297 {
3298         struct nfs4_exception exception = { };
3299         int err;
3300
3301         sattr->ia_mode &= ~current_umask();
3302         do {
3303                 err = nfs4_handle_exception(NFS_SERVER(dir),
3304                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3305                                 &exception);
3306         } while (exception.retry);
3307         return err;
3308 }
3309
3310 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3311                 u64 cookie, struct page **pages, unsigned int count, int plus)
3312 {
3313         struct inode            *dir = dentry->d_inode;
3314         struct nfs4_readdir_arg args = {
3315                 .fh = NFS_FH(dir),
3316                 .pages = pages,
3317                 .pgbase = 0,
3318                 .count = count,
3319                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3320                 .plus = plus,
3321         };
3322         struct nfs4_readdir_res res;
3323         struct rpc_message msg = {
3324                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3325                 .rpc_argp = &args,
3326                 .rpc_resp = &res,
3327                 .rpc_cred = cred,
3328         };
3329         int                     status;
3330
3331         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3332                         dentry->d_parent->d_name.name,
3333                         dentry->d_name.name,
3334                         (unsigned long long)cookie);
3335         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3336         res.pgbase = args.pgbase;
3337         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3338         if (status >= 0) {
3339                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3340                 status += args.pgbase;
3341         }
3342
3343         nfs_invalidate_atime(dir);
3344
3345         dprintk("%s: returns %d\n", __func__, status);
3346         return status;
3347 }
3348
3349 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3350                 u64 cookie, struct page **pages, unsigned int count, int plus)
3351 {
3352         struct nfs4_exception exception = { };
3353         int err;
3354         do {
3355                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3356                                 _nfs4_proc_readdir(dentry, cred, cookie,
3357                                         pages, count, plus),
3358                                 &exception);
3359         } while (exception.retry);
3360         return err;
3361 }
3362
3363 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3364                 struct iattr *sattr, dev_t rdev)
3365 {
3366         struct nfs4_createdata *data;
3367         int mode = sattr->ia_mode;
3368         int status = -ENOMEM;
3369
3370         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3371         if (data == NULL)
3372                 goto out;
3373
3374         if (S_ISFIFO(mode))
3375                 data->arg.ftype = NF4FIFO;
3376         else if (S_ISBLK(mode)) {
3377                 data->arg.ftype = NF4BLK;
3378                 data->arg.u.device.specdata1 = MAJOR(rdev);
3379                 data->arg.u.device.specdata2 = MINOR(rdev);
3380         }
3381         else if (S_ISCHR(mode)) {
3382                 data->arg.ftype = NF4CHR;
3383                 data->arg.u.device.specdata1 = MAJOR(rdev);
3384                 data->arg.u.device.specdata2 = MINOR(rdev);
3385         } else if (!S_ISSOCK(mode)) {
3386                 status = -EINVAL;
3387                 goto out_free;
3388         }
3389         
3390         status = nfs4_do_create(dir, dentry, data);
3391 out_free:
3392         nfs4_free_createdata(data);
3393 out:
3394         return status;
3395 }
3396
3397 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3398                 struct iattr *sattr, dev_t rdev)
3399 {
3400         struct nfs4_exception exception = { };
3401         int err;
3402
3403         sattr->ia_mode &= ~current_umask();
3404         do {
3405                 err = nfs4_handle_exception(NFS_SERVER(dir),
3406                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3407                                 &exception);
3408         } while (exception.retry);
3409         return err;
3410 }
3411
3412 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3413                  struct nfs_fsstat *fsstat)
3414 {
3415         struct nfs4_statfs_arg args = {
3416                 .fh = fhandle,
3417                 .bitmask = server->attr_bitmask,
3418         };
3419         struct nfs4_statfs_res res = {
3420                 .fsstat = fsstat,
3421         };
3422         struct rpc_message msg = {
3423                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3424                 .rpc_argp = &args,
3425                 .rpc_resp = &res,
3426         };
3427
3428         nfs_fattr_init(fsstat->fattr);
3429         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3430 }
3431
3432 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3433 {
3434         struct nfs4_exception exception = { };
3435         int err;
3436         do {
3437                 err = nfs4_handle_exception(server,
3438                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3439                                 &exception);
3440         } while (exception.retry);
3441         return err;
3442 }
3443
3444 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3445                 struct nfs_fsinfo *fsinfo)
3446 {
3447         struct nfs4_fsinfo_arg args = {
3448                 .fh = fhandle,
3449                 .bitmask = server->attr_bitmask,
3450         };
3451         struct nfs4_fsinfo_res res = {
3452                 .fsinfo = fsinfo,
3453         };
3454         struct rpc_message msg = {
3455                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3456                 .rpc_argp = &args,
3457                 .rpc_resp = &res,
3458         };
3459
3460         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3461 }
3462
3463 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3464 {
3465         struct nfs4_exception exception = { };
3466         int err;
3467
3468         do {
3469                 err = nfs4_handle_exception(server,
3470                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3471                                 &exception);
3472         } while (exception.retry);
3473         return err;
3474 }
3475
3476 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3477 {
3478         int error;
3479
3480         nfs_fattr_init(fsinfo->fattr);
3481         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3482         if (error == 0) {
3483                 /* block layout checks this! */
3484                 server->pnfs_blksize = fsinfo->blksize;
3485                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3486         }
3487
3488         return error;
3489 }
3490
3491 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3492                 struct nfs_pathconf *pathconf)
3493 {
3494         struct nfs4_pathconf_arg args = {
3495                 .fh = fhandle,
3496                 .bitmask = server->attr_bitmask,
3497         };
3498         struct nfs4_pathconf_res res = {
3499                 .pathconf = pathconf,
3500         };
3501         struct rpc_message msg = {
3502                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3503                 .rpc_argp = &args,
3504                 .rpc_resp = &res,
3505         };
3506
3507         /* None of the pathconf attributes are mandatory to implement */
3508         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3509                 memset(pathconf, 0, sizeof(*pathconf));
3510                 return 0;
3511         }
3512
3513         nfs_fattr_init(pathconf->fattr);
3514         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3515 }
3516
3517 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3518                 struct nfs_pathconf *pathconf)
3519 {
3520         struct nfs4_exception exception = { };
3521         int err;
3522
3523         do {
3524                 err = nfs4_handle_exception(server,
3525                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3526                                 &exception);
3527         } while (exception.retry);
3528         return err;
3529 }
3530
3531 void __nfs4_read_done_cb(struct nfs_read_data *data)
3532 {
3533         nfs_invalidate_atime(data->header->inode);
3534 }
3535
3536 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3537 {
3538         struct nfs_server *server = NFS_SERVER(data->header->inode);
3539
3540         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3541                 rpc_restart_call_prepare(task);
3542                 return -EAGAIN;
3543         }
3544
3545         __nfs4_read_done_cb(data);
3546         if (task->tk_status > 0)
3547                 renew_lease(server, data->timestamp);
3548         return 0;
3549 }
3550
3551 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3552 {
3553
3554         dprintk("--> %s\n", __func__);
3555
3556         if (!nfs4_sequence_done(task, &data->res.seq_res))
3557                 return -EAGAIN;
3558
3559         return data->read_done_cb ? data->read_done_cb(task, data) :
3560                                     nfs4_read_done_cb(task, data);
3561 }
3562
3563 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3564 {
3565         data->timestamp   = jiffies;
3566         data->read_done_cb = nfs4_read_done_cb;
3567         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3568         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3569 }
3570
3571 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3572 {
3573         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3574                                 &data->args.seq_args,
3575                                 &data->res.seq_res,
3576                                 task))
3577                 return;
3578         rpc_call_start(task);
3579 }
3580
3581 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3582 {
3583         struct inode *inode = data->header->inode;
3584         
3585         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3586                 rpc_restart_call_prepare(task);
3587                 return -EAGAIN;
3588         }
3589         if (task->tk_status >= 0) {
3590                 renew_lease(NFS_SERVER(inode), data->timestamp);
3591                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3592         }
3593         return 0;
3594 }
3595
3596 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3597 {
3598         if (!nfs4_sequence_done(task, &data->res.seq_res))
3599                 return -EAGAIN;
3600         return data->write_done_cb ? data->write_done_cb(task, data) :
3601                 nfs4_write_done_cb(task, data);
3602 }
3603
3604 static
3605 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3606 {
3607         const struct nfs_pgio_header *hdr = data->header;
3608
3609         /* Don't request attributes for pNFS or O_DIRECT writes */
3610         if (data->ds_clp != NULL || hdr->dreq != NULL)
3611                 return false;
3612         /* Otherwise, request attributes if and only if we don't hold
3613          * a delegation
3614          */
3615         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3616 }
3617
3618 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3619 {
3620         struct nfs_server *server = NFS_SERVER(data->header->inode);
3621
3622         if (!nfs4_write_need_cache_consistency_data(data)) {
3623                 data->args.bitmask = NULL;
3624                 data->res.fattr = NULL;
3625         } else
3626                 data->args.bitmask = server->cache_consistency_bitmask;
3627
3628         if (!data->write_done_cb)
3629                 data->write_done_cb = nfs4_write_done_cb;
3630         data->res.server = server;
3631         data->timestamp   = jiffies;
3632
3633         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3634         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3635 }
3636
3637 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3638 {
3639         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3640                                 &data->args.seq_args,
3641                                 &data->res.seq_res,
3642                                 task))
3643                 return;
3644         rpc_call_start(task);
3645 }
3646
3647 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3648 {
3649         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3650                                 &data->args.seq_args,
3651                                 &data->res.seq_res,
3652                                 task))
3653                 return;
3654         rpc_call_start(task);
3655 }
3656
3657 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3658 {
3659         struct inode *inode = data->inode;
3660
3661         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3662                 rpc_restart_call_prepare(task);
3663                 return -EAGAIN;
3664         }
3665         return 0;
3666 }
3667
3668 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3669 {
3670         if (!nfs4_sequence_done(task, &data->res.seq_res))
3671                 return -EAGAIN;
3672         return data->commit_done_cb(task, data);
3673 }
3674
3675 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3676 {
3677         struct nfs_server *server = NFS_SERVER(data->inode);
3678
3679         if (data->commit_done_cb == NULL)
3680                 data->commit_done_cb = nfs4_commit_done_cb;
3681         data->res.server = server;
3682         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3683         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3684 }
3685
3686 struct nfs4_renewdata {
3687         struct nfs_client       *client;
3688         unsigned long           timestamp;
3689 };
3690
3691 /*
3692  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3693  * standalone procedure for queueing an asynchronous RENEW.
3694  */
3695 static void nfs4_renew_release(void *calldata)
3696 {
3697         struct nfs4_renewdata *data = calldata;
3698         struct nfs_client *clp = data->client;
3699
3700         if (atomic_read(&clp->cl_count) > 1)
3701                 nfs4_schedule_state_renewal(clp);
3702         nfs_put_client(clp);
3703         kfree(data);
3704 }
3705
3706 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3707 {
3708         struct nfs4_renewdata *data = calldata;
3709         struct nfs_client *clp = data->client;
3710         unsigned long timestamp = data->timestamp;
3711
3712         if (task->tk_status < 0) {
3713                 /* Unless we're shutting down, schedule state recovery! */
3714                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3715                         return;
3716                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3717                         nfs4_schedule_lease_recovery(clp);
3718                         return;
3719                 }
3720                 nfs4_schedule_path_down_recovery(clp);
3721         }
3722         do_renew_lease(clp, timestamp);
3723 }
3724
3725 static const struct rpc_call_ops nfs4_renew_ops = {
3726         .rpc_call_done = nfs4_renew_done,
3727         .rpc_release = nfs4_renew_release,
3728 };
3729
3730 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3731 {
3732         struct rpc_message msg = {
3733                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3734                 .rpc_argp       = clp,
3735                 .rpc_cred       = cred,
3736         };
3737         struct nfs4_renewdata *data;
3738
3739         if (renew_flags == 0)
3740                 return 0;
3741         if (!atomic_inc_not_zero(&clp->cl_count))
3742                 return -EIO;
3743         data = kmalloc(sizeof(*data), GFP_NOFS);
3744         if (data == NULL)
3745                 return -ENOMEM;
3746         data->client = clp;
3747         data->timestamp = jiffies;
3748         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3749                         &nfs4_renew_ops, data);
3750 }
3751
3752 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3753 {
3754         struct rpc_message msg = {
3755                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3756                 .rpc_argp       = clp,
3757                 .rpc_cred       = cred,
3758         };
3759         unsigned long now = jiffies;
3760         int status;
3761
3762         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3763         if (status < 0)
3764                 return status;
3765         do_renew_lease(clp, now);
3766         return 0;
3767 }
3768
3769 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3770 {
3771         return (server->caps & NFS_CAP_ACLS)
3772                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3773                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3774 }
3775
3776 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3777  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3778  * the stack.
3779  */
3780 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3781
3782 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3783                 struct page **pages, unsigned int *pgbase)
3784 {
3785         struct page *newpage, **spages;
3786         int rc = 0;
3787         size_t len;
3788         spages = pages;
3789
3790         do {
3791                 len = min_t(size_t, PAGE_SIZE, buflen);
3792                 newpage = alloc_page(GFP_KERNEL);
3793
3794                 if (newpage == NULL)
3795                         goto unwind;
3796                 memcpy(page_address(newpage), buf, len);
3797                 buf += len;
3798                 buflen -= len;
3799                 *pages++ = newpage;
3800                 rc++;
3801         } while (buflen != 0);
3802
3803         return rc;
3804
3805 unwind:
3806         for(; rc > 0; rc--)
3807                 __free_page(spages[rc-1]);
3808         return -ENOMEM;
3809 }
3810
3811 struct nfs4_cached_acl {
3812         int cached;
3813         size_t len;
3814         char data[0];
3815 };
3816
3817 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3818 {
3819         struct nfs_inode *nfsi = NFS_I(inode);
3820
3821         spin_lock(&inode->i_lock);
3822         kfree(nfsi->nfs4_acl);
3823         nfsi->nfs4_acl = acl;
3824         spin_unlock(&inode->i_lock);
3825 }
3826
3827 static void nfs4_zap_acl_attr(struct inode *inode)
3828 {
3829         nfs4_set_cached_acl(inode, NULL);
3830 }
3831
3832 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3833 {
3834         struct nfs_inode *nfsi = NFS_I(inode);
3835         struct nfs4_cached_acl *acl;
3836         int ret = -ENOENT;
3837
3838         spin_lock(&inode->i_lock);
3839         acl = nfsi->nfs4_acl;
3840         if (acl == NULL)
3841                 goto out;
3842         if (buf == NULL) /* user is just asking for length */
3843                 goto out_len;
3844         if (acl->cached == 0)
3845                 goto out;
3846         ret = -ERANGE; /* see getxattr(2) man page */
3847         if (acl->len > buflen)
3848                 goto out;
3849         memcpy(buf, acl->data, acl->len);
3850 out_len:
3851         ret = acl->len;
3852 out:
3853         spin_unlock(&inode->i_lock);
3854         return ret;
3855 }
3856
3857 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3858 {
3859         struct nfs4_cached_acl *acl;
3860         size_t buflen = sizeof(*acl) + acl_len;
3861
3862         if (buflen <= PAGE_SIZE) {
3863                 acl = kmalloc(buflen, GFP_KERNEL);
3864                 if (acl == NULL)
3865                         goto out;
3866                 acl->cached = 1;
3867                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3868         } else {
3869                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3870                 if (acl == NULL)
3871                         goto out;
3872                 acl->cached = 0;
3873         }
3874         acl->len = acl_len;
3875 out:
3876         nfs4_set_cached_acl(inode, acl);
3877 }
3878
3879 /*
3880  * The getxattr API returns the required buffer length when called with a
3881  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3882  * the required buf.  On a NULL buf, we send a page of data to the server
3883  * guessing that the ACL request can be serviced by a page. If so, we cache
3884  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3885  * the cache. If not so, we throw away the page, and cache the required
3886  * length. The next getxattr call will then produce another round trip to
3887  * the server, this time with the input buf of the required size.
3888  */
3889 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3890 {
3891         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3892         struct nfs_getaclargs args = {
3893                 .fh = NFS_FH(inode),
3894                 .acl_pages = pages,
3895                 .acl_len = buflen,
3896         };
3897         struct nfs_getaclres res = {
3898                 .acl_len = buflen,
3899         };
3900         struct rpc_message msg = {
3901                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3902                 .rpc_argp = &args,
3903                 .rpc_resp = &res,
3904         };
3905         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3906         int ret = -ENOMEM, i;
3907
3908         /* As long as we're doing a round trip to the server anyway,
3909          * let's be prepared for a page of acl data. */
3910         if (npages == 0)
3911                 npages = 1;
3912         if (npages > ARRAY_SIZE(pages))
3913                 return -ERANGE;
3914
3915         for (i = 0; i < npages; i++) {
3916                 pages[i] = alloc_page(GFP_KERNEL);
3917                 if (!pages[i])
3918                         goto out_free;
3919         }
3920
3921         /* for decoding across pages */
3922         res.acl_scratch = alloc_page(GFP_KERNEL);
3923         if (!res.acl_scratch)
3924                 goto out_free;
3925
3926         args.acl_len = npages * PAGE_SIZE;
3927         args.acl_pgbase = 0;
3928
3929         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3930                 __func__, buf, buflen, npages, args.acl_len);
3931         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3932                              &msg, &args.seq_args, &res.seq_res, 0);
3933         if (ret)
3934                 goto out_free;
3935
3936         /* Handle the case where the passed-in buffer is too short */
3937         if (res.acl_flags & NFS4_ACL_TRUNC) {
3938                 /* Did the user only issue a request for the acl length? */
3939                 if (buf == NULL)
3940                         goto out_ok;
3941                 ret = -ERANGE;
3942                 goto out_free;
3943         }
3944         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
3945         if (buf)
3946                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
3947 out_ok:
3948         ret = res.acl_len;
3949 out_free:
3950         for (i = 0; i < npages; i++)
3951                 if (pages[i])
3952                         __free_page(pages[i]);
3953         if (res.acl_scratch)
3954                 __free_page(res.acl_scratch);
3955         return ret;
3956 }
3957
3958 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3959 {
3960         struct nfs4_exception exception = { };
3961         ssize_t ret;
3962         do {
3963                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3964                 if (ret >= 0)
3965                         break;
3966                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3967         } while (exception.retry);
3968         return ret;
3969 }
3970
3971 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3972 {
3973         struct nfs_server *server = NFS_SERVER(inode);
3974         int ret;
3975
3976         if (!nfs4_server_supports_acls(server))
3977                 return -EOPNOTSUPP;
3978         ret = nfs_revalidate_inode(server, inode);
3979         if (ret < 0)
3980                 return ret;
3981         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3982                 nfs_zap_acl_cache(inode);
3983         ret = nfs4_read_cached_acl(inode, buf, buflen);
3984         if (ret != -ENOENT)
3985                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3986                  * but no cached acl data, just the acl length */
3987                 return ret;
3988         return nfs4_get_acl_uncached(inode, buf, buflen);
3989 }
3990
3991 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3992 {
3993         struct nfs_server *server = NFS_SERVER(inode);
3994         struct page *pages[NFS4ACL_MAXPAGES];
3995         struct nfs_setaclargs arg = {
3996                 .fh             = NFS_FH(inode),
3997                 .acl_pages      = pages,
3998                 .acl_len        = buflen,
3999         };
4000         struct nfs_setaclres res;
4001         struct rpc_message msg = {
4002                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4003                 .rpc_argp       = &arg,
4004                 .rpc_resp       = &res,
4005         };
4006         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4007         int ret, i;
4008
4009         if (!nfs4_server_supports_acls(server))
4010                 return -EOPNOTSUPP;
4011         if (npages > ARRAY_SIZE(pages))
4012                 return -ERANGE;
4013         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4014         if (i < 0)
4015                 return i;
4016         nfs4_inode_return_delegation(inode);
4017         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4018
4019         /*
4020          * Free each page after tx, so the only ref left is
4021          * held by the network stack
4022          */
4023         for (; i > 0; i--)
4024                 put_page(pages[i-1]);
4025
4026         /*
4027          * Acl update can result in inode attribute update.
4028          * so mark the attribute cache invalid.
4029          */
4030         spin_lock(&inode->i_lock);
4031         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4032         spin_unlock(&inode->i_lock);
4033         nfs_access_zap_cache(inode);
4034         nfs_zap_acl_cache(inode);
4035         return ret;
4036 }
4037
4038 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4039 {
4040         struct nfs4_exception exception = { };
4041         int err;
4042         do {
4043                 err = nfs4_handle_exception(NFS_SERVER(inode),
4044                                 __nfs4_proc_set_acl(inode, buf, buflen),
4045                                 &exception);
4046         } while (exception.retry);
4047         return err;
4048 }
4049
4050 static int
4051 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4052 {
4053         struct nfs_client *clp = server->nfs_client;
4054
4055         if (task->tk_status >= 0)
4056                 return 0;
4057         switch(task->tk_status) {
4058                 case -NFS4ERR_DELEG_REVOKED:
4059                 case -NFS4ERR_ADMIN_REVOKED:
4060                 case -NFS4ERR_BAD_STATEID:
4061                         if (state == NULL)
4062                                 break;
4063                         nfs_remove_bad_delegation(state->inode);
4064                 case -NFS4ERR_OPENMODE:
4065                         if (state == NULL)
4066                                 break;
4067                         nfs4_schedule_stateid_recovery(server, state);
4068                         goto wait_on_recovery;
4069                 case -NFS4ERR_EXPIRED:
4070                         if (state != NULL)
4071                                 nfs4_schedule_stateid_recovery(server, state);
4072                 case -NFS4ERR_STALE_STATEID:
4073                 case -NFS4ERR_STALE_CLIENTID:
4074                         nfs4_schedule_lease_recovery(clp);
4075                         goto wait_on_recovery;
4076 #if defined(CONFIG_NFS_V4_1)
4077                 case -NFS4ERR_BADSESSION:
4078                 case -NFS4ERR_BADSLOT:
4079                 case -NFS4ERR_BAD_HIGH_SLOT:
4080                 case -NFS4ERR_DEADSESSION:
4081                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4082                 case -NFS4ERR_SEQ_FALSE_RETRY:
4083                 case -NFS4ERR_SEQ_MISORDERED:
4084                         dprintk("%s ERROR %d, Reset session\n", __func__,
4085                                 task->tk_status);
4086                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4087                         task->tk_status = 0;
4088                         return -EAGAIN;
4089 #endif /* CONFIG_NFS_V4_1 */
4090                 case -NFS4ERR_DELAY:
4091                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4092                 case -NFS4ERR_GRACE:
4093                 case -EKEYEXPIRED:
4094                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4095                         task->tk_status = 0;
4096                         return -EAGAIN;
4097                 case -NFS4ERR_RETRY_UNCACHED_REP:
4098                 case -NFS4ERR_OLD_STATEID:
4099                         task->tk_status = 0;
4100                         return -EAGAIN;
4101         }
4102         task->tk_status = nfs4_map_errors(task->tk_status);
4103         return 0;
4104 wait_on_recovery:
4105         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4106         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4107                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4108         task->tk_status = 0;
4109         return -EAGAIN;
4110 }
4111
4112 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4113                                     nfs4_verifier *bootverf)
4114 {
4115         __be32 verf[2];
4116
4117         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4118                 /* An impossible timestamp guarantees this value
4119                  * will never match a generated boot time. */
4120                 verf[0] = 0;
4121                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4122         } else {
4123                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4124                 verf[0] = (__be32)nn->boot_time.tv_sec;
4125                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4126         }
4127         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4128 }
4129
4130 static unsigned int
4131 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4132                                    char *buf, size_t len)
4133 {
4134         unsigned int result;
4135
4136         rcu_read_lock();
4137         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4138                                 clp->cl_ipaddr,
4139                                 rpc_peeraddr2str(clp->cl_rpcclient,
4140                                                         RPC_DISPLAY_ADDR),
4141                                 rpc_peeraddr2str(clp->cl_rpcclient,
4142                                                         RPC_DISPLAY_PROTO));
4143         rcu_read_unlock();
4144         return result;
4145 }
4146
4147 static unsigned int
4148 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4149                                 char *buf, size_t len)
4150 {
4151         char *nodename = clp->cl_rpcclient->cl_nodename;
4152
4153         if (nfs4_client_id_uniquifier[0] != '\0')
4154                 nodename = nfs4_client_id_uniquifier;
4155         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4156                                 clp->rpc_ops->version, clp->cl_minorversion,
4157                                 nodename);
4158 }
4159
4160 /**
4161  * nfs4_proc_setclientid - Negotiate client ID
4162  * @clp: state data structure
4163  * @program: RPC program for NFSv4 callback service
4164  * @port: IP port number for NFS4 callback service
4165  * @cred: RPC credential to use for this call
4166  * @res: where to place the result
4167  *
4168  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4169  */
4170 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4171                 unsigned short port, struct rpc_cred *cred,
4172                 struct nfs4_setclientid_res *res)
4173 {
4174         nfs4_verifier sc_verifier;
4175         struct nfs4_setclientid setclientid = {
4176                 .sc_verifier = &sc_verifier,
4177                 .sc_prog = program,
4178                 .sc_cb_ident = clp->cl_cb_ident,
4179         };
4180         struct rpc_message msg = {
4181                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4182                 .rpc_argp = &setclientid,
4183                 .rpc_resp = res,
4184                 .rpc_cred = cred,
4185         };
4186         int status;
4187
4188         /* nfs_client_id4 */
4189         nfs4_init_boot_verifier(clp, &sc_verifier);
4190         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4191                 setclientid.sc_name_len =
4192                                 nfs4_init_uniform_client_string(clp,
4193                                                 setclientid.sc_name,
4194                                                 sizeof(setclientid.sc_name));
4195         else
4196                 setclientid.sc_name_len =
4197                                 nfs4_init_nonuniform_client_string(clp,
4198                                                 setclientid.sc_name,
4199                                                 sizeof(setclientid.sc_name));
4200         /* cb_client4 */
4201         rcu_read_lock();
4202         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4203                                 sizeof(setclientid.sc_netid),
4204                                 rpc_peeraddr2str(clp->cl_rpcclient,
4205                                                         RPC_DISPLAY_NETID));
4206         rcu_read_unlock();
4207         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4208                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4209                                 clp->cl_ipaddr, port >> 8, port & 255);
4210
4211         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4212                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4213                 setclientid.sc_name_len, setclientid.sc_name);
4214         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4215         dprintk("NFS reply setclientid: %d\n", status);
4216         return status;
4217 }
4218
4219 /**
4220  * nfs4_proc_setclientid_confirm - Confirm client ID
4221  * @clp: state data structure
4222  * @res: result of a previous SETCLIENTID
4223  * @cred: RPC credential to use for this call
4224  *
4225  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4226  */
4227 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4228                 struct nfs4_setclientid_res *arg,
4229                 struct rpc_cred *cred)
4230 {
4231         struct nfs_fsinfo fsinfo;
4232         struct rpc_message msg = {
4233                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4234                 .rpc_argp = arg,
4235                 .rpc_resp = &fsinfo,
4236                 .rpc_cred = cred,
4237         };
4238         unsigned long now;
4239         int status;
4240
4241         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4242                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4243                 clp->cl_clientid);
4244         now = jiffies;
4245         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4246         if (status == 0) {
4247                 spin_lock(&clp->cl_lock);
4248                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4249                 clp->cl_last_renewal = now;
4250                 spin_unlock(&clp->cl_lock);
4251         }
4252         dprintk("NFS reply setclientid_confirm: %d\n", status);
4253         return status;
4254 }
4255
4256 struct nfs4_delegreturndata {
4257         struct nfs4_delegreturnargs args;
4258         struct nfs4_delegreturnres res;
4259         struct nfs_fh fh;
4260         nfs4_stateid stateid;
4261         unsigned long timestamp;
4262         struct nfs_fattr fattr;
4263         int rpc_status;
4264 };
4265
4266 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4267 {
4268         struct nfs4_delegreturndata *data = calldata;
4269
4270         if (!nfs4_sequence_done(task, &data->res.seq_res))
4271                 return;
4272
4273         switch (task->tk_status) {
4274         case -NFS4ERR_STALE_STATEID:
4275         case -NFS4ERR_EXPIRED:
4276         case 0:
4277                 renew_lease(data->res.server, data->timestamp);
4278                 break;
4279         default:
4280                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4281                                 -EAGAIN) {
4282                         rpc_restart_call_prepare(task);
4283                         return;
4284                 }
4285         }
4286         data->rpc_status = task->tk_status;
4287 }
4288
4289 static void nfs4_delegreturn_release(void *calldata)
4290 {
4291         kfree(calldata);
4292 }
4293
4294 #if defined(CONFIG_NFS_V4_1)
4295 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4296 {
4297         struct nfs4_delegreturndata *d_data;
4298
4299         d_data = (struct nfs4_delegreturndata *)data;
4300
4301         if (nfs4_setup_sequence(d_data->res.server,
4302                                 &d_data->args.seq_args,
4303                                 &d_data->res.seq_res, task))
4304                 return;
4305         rpc_call_start(task);
4306 }
4307 #endif /* CONFIG_NFS_V4_1 */
4308
4309 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4310 #if defined(CONFIG_NFS_V4_1)
4311         .rpc_call_prepare = nfs4_delegreturn_prepare,
4312 #endif /* CONFIG_NFS_V4_1 */
4313         .rpc_call_done = nfs4_delegreturn_done,
4314         .rpc_release = nfs4_delegreturn_release,
4315 };
4316
4317 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4318 {
4319         struct nfs4_delegreturndata *data;
4320         struct nfs_server *server = NFS_SERVER(inode);
4321         struct rpc_task *task;
4322         struct rpc_message msg = {
4323                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4324                 .rpc_cred = cred,
4325         };
4326         struct rpc_task_setup task_setup_data = {
4327                 .rpc_client = server->client,
4328                 .rpc_message = &msg,
4329                 .callback_ops = &nfs4_delegreturn_ops,
4330                 .flags = RPC_TASK_ASYNC,
4331         };
4332         int status = 0;
4333
4334         data = kzalloc(sizeof(*data), GFP_NOFS);
4335         if (data == NULL)
4336                 return -ENOMEM;
4337         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4338         data->args.fhandle = &data->fh;
4339         data->args.stateid = &data->stateid;
4340         data->args.bitmask = server->cache_consistency_bitmask;
4341         nfs_copy_fh(&data->fh, NFS_FH(inode));
4342         nfs4_stateid_copy(&data->stateid, stateid);
4343         data->res.fattr = &data->fattr;
4344         data->res.server = server;
4345         nfs_fattr_init(data->res.fattr);
4346         data->timestamp = jiffies;
4347         data->rpc_status = 0;
4348
4349         task_setup_data.callback_data = data;
4350         msg.rpc_argp = &data->args;
4351         msg.rpc_resp = &data->res;
4352         task = rpc_run_task(&task_setup_data);
4353         if (IS_ERR(task))
4354                 return PTR_ERR(task);
4355         if (!issync)
4356                 goto out;
4357         status = nfs4_wait_for_completion_rpc_task(task);
4358         if (status != 0)
4359                 goto out;
4360         status = data->rpc_status;
4361         if (status == 0)
4362                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4363         else
4364                 nfs_refresh_inode(inode, &data->fattr);
4365 out:
4366         rpc_put_task(task);
4367         return status;
4368 }
4369
4370 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4371 {
4372         struct nfs_server *server = NFS_SERVER(inode);
4373         struct nfs4_exception exception = { };
4374         int err;
4375         do {
4376                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4377                 switch (err) {
4378                         case -NFS4ERR_STALE_STATEID:
4379                         case -NFS4ERR_EXPIRED:
4380                         case 0:
4381                                 return 0;
4382                 }
4383                 err = nfs4_handle_exception(server, err, &exception);
4384         } while (exception.retry);
4385         return err;
4386 }
4387
4388 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4389 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4390
4391 /* 
4392  * sleep, with exponential backoff, and retry the LOCK operation. 
4393  */
4394 static unsigned long
4395 nfs4_set_lock_task_retry(unsigned long timeout)
4396 {
4397         freezable_schedule_timeout_killable(timeout);
4398         timeout <<= 1;
4399         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4400                 return NFS4_LOCK_MAXTIMEOUT;
4401         return timeout;
4402 }
4403
4404 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4405 {
4406         struct inode *inode = state->inode;
4407         struct nfs_server *server = NFS_SERVER(inode);
4408         struct nfs_client *clp = server->nfs_client;
4409         struct nfs_lockt_args arg = {
4410                 .fh = NFS_FH(inode),
4411                 .fl = request,
4412         };
4413         struct nfs_lockt_res res = {
4414                 .denied = request,
4415         };
4416         struct rpc_message msg = {
4417                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4418                 .rpc_argp       = &arg,
4419                 .rpc_resp       = &res,
4420                 .rpc_cred       = state->owner->so_cred,
4421         };
4422         struct nfs4_lock_state *lsp;
4423         int status;
4424
4425         arg.lock_owner.clientid = clp->cl_clientid;
4426         status = nfs4_set_lock_state(state, request);
4427         if (status != 0)
4428                 goto out;
4429         lsp = request->fl_u.nfs4_fl.owner;
4430         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4431         arg.lock_owner.s_dev = server->s_dev;
4432         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4433         switch (status) {
4434                 case 0:
4435                         request->fl_type = F_UNLCK;
4436                         break;
4437                 case -NFS4ERR_DENIED:
4438                         status = 0;
4439         }
4440         request->fl_ops->fl_release_private(request);
4441 out:
4442         return status;
4443 }
4444
4445 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4446 {
4447         struct nfs4_exception exception = { };
4448         int err;
4449
4450         do {
4451                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4452                                 _nfs4_proc_getlk(state, cmd, request),
4453                                 &exception);
4454         } while (exception.retry);
4455         return err;
4456 }
4457
4458 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4459 {
4460         int res = 0;
4461         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4462                 case FL_POSIX:
4463                         res = posix_lock_file_wait(file, fl);
4464                         break;
4465                 case FL_FLOCK:
4466                         res = flock_lock_file_wait(file, fl);
4467                         break;
4468                 default:
4469                         BUG();
4470         }
4471         return res;
4472 }
4473
4474 struct nfs4_unlockdata {
4475         struct nfs_locku_args arg;
4476         struct nfs_locku_res res;
4477         struct nfs4_lock_state *lsp;
4478         struct nfs_open_context *ctx;
4479         struct file_lock fl;
4480         const struct nfs_server *server;
4481         unsigned long timestamp;
4482 };
4483
4484 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4485                 struct nfs_open_context *ctx,
4486                 struct nfs4_lock_state *lsp,
4487                 struct nfs_seqid *seqid)
4488 {
4489         struct nfs4_unlockdata *p;
4490         struct inode *inode = lsp->ls_state->inode;
4491
4492         p = kzalloc(sizeof(*p), GFP_NOFS);
4493         if (p == NULL)
4494                 return NULL;
4495         p->arg.fh = NFS_FH(inode);
4496         p->arg.fl = &p->fl;
4497         p->arg.seqid = seqid;
4498         p->res.seqid = seqid;
4499         p->arg.stateid = &lsp->ls_stateid;
4500         p->lsp = lsp;
4501         atomic_inc(&lsp->ls_count);
4502         /* Ensure we don't close file until we're done freeing locks! */
4503         p->ctx = get_nfs_open_context(ctx);
4504         memcpy(&p->fl, fl, sizeof(p->fl));
4505         p->server = NFS_SERVER(inode);
4506         return p;
4507 }
4508
4509 static void nfs4_locku_release_calldata(void *data)
4510 {
4511         struct nfs4_unlockdata *calldata = data;
4512         nfs_free_seqid(calldata->arg.seqid);
4513         nfs4_put_lock_state(calldata->lsp);
4514         put_nfs_open_context(calldata->ctx);
4515         kfree(calldata);
4516 }
4517
4518 static void nfs4_locku_done(struct rpc_task *task, void *data)
4519 {
4520         struct nfs4_unlockdata *calldata = data;
4521
4522         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4523                 return;
4524         switch (task->tk_status) {
4525                 case 0:
4526                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4527                                         &calldata->res.stateid);
4528                         renew_lease(calldata->server, calldata->timestamp);
4529                         break;
4530                 case -NFS4ERR_BAD_STATEID:
4531                 case -NFS4ERR_OLD_STATEID:
4532                 case -NFS4ERR_STALE_STATEID:
4533                 case -NFS4ERR_EXPIRED:
4534                         break;
4535                 default:
4536                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4537                                 rpc_restart_call_prepare(task);
4538         }
4539         nfs_release_seqid(calldata->arg.seqid);
4540 }
4541
4542 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4543 {
4544         struct nfs4_unlockdata *calldata = data;
4545
4546         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4547                 return;
4548         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4549                 /* Note: exit _without_ running nfs4_locku_done */
4550                 task->tk_action = NULL;
4551                 return;
4552         }
4553         calldata->timestamp = jiffies;
4554         if (nfs4_setup_sequence(calldata->server,
4555                                 &calldata->arg.seq_args,
4556                                 &calldata->res.seq_res,
4557                                 task) != 0)
4558                 nfs_release_seqid(calldata->arg.seqid);
4559         else
4560                 rpc_call_start(task);
4561 }
4562
4563 static const struct rpc_call_ops nfs4_locku_ops = {
4564         .rpc_call_prepare = nfs4_locku_prepare,
4565         .rpc_call_done = nfs4_locku_done,
4566         .rpc_release = nfs4_locku_release_calldata,
4567 };
4568
4569 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4570                 struct nfs_open_context *ctx,
4571                 struct nfs4_lock_state *lsp,
4572                 struct nfs_seqid *seqid)
4573 {
4574         struct nfs4_unlockdata *data;
4575         struct rpc_message msg = {
4576                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4577                 .rpc_cred = ctx->cred,
4578         };
4579         struct rpc_task_setup task_setup_data = {
4580                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4581                 .rpc_message = &msg,
4582                 .callback_ops = &nfs4_locku_ops,
4583                 .workqueue = nfsiod_workqueue,
4584                 .flags = RPC_TASK_ASYNC,
4585         };
4586
4587         /* Ensure this is an unlock - when canceling a lock, the
4588          * canceled lock is passed in, and it won't be an unlock.
4589          */
4590         fl->fl_type = F_UNLCK;
4591
4592         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4593         if (data == NULL) {
4594                 nfs_free_seqid(seqid);
4595                 return ERR_PTR(-ENOMEM);
4596         }
4597
4598         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4599         msg.rpc_argp = &data->arg;
4600         msg.rpc_resp = &data->res;
4601         task_setup_data.callback_data = data;
4602         return rpc_run_task(&task_setup_data);
4603 }
4604
4605 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4606 {
4607         struct nfs_inode *nfsi = NFS_I(state->inode);
4608         struct nfs_seqid *seqid;
4609         struct nfs4_lock_state *lsp;
4610         struct rpc_task *task;
4611         int status = 0;
4612         unsigned char fl_flags = request->fl_flags;
4613
4614         status = nfs4_set_lock_state(state, request);
4615         /* Unlock _before_ we do the RPC call */
4616         request->fl_flags |= FL_EXISTS;
4617         down_read(&nfsi->rwsem);
4618         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4619                 up_read(&nfsi->rwsem);
4620                 goto out;
4621         }
4622         up_read(&nfsi->rwsem);
4623         if (status != 0)
4624                 goto out;
4625         /* Is this a delegated lock? */
4626         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4627                 goto out;
4628         lsp = request->fl_u.nfs4_fl.owner;
4629         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4630         status = -ENOMEM;
4631         if (seqid == NULL)
4632                 goto out;
4633         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4634         status = PTR_ERR(task);
4635         if (IS_ERR(task))
4636                 goto out;
4637         status = nfs4_wait_for_completion_rpc_task(task);
4638         rpc_put_task(task);
4639 out:
4640         request->fl_flags = fl_flags;
4641         return status;
4642 }
4643
4644 struct nfs4_lockdata {
4645         struct nfs_lock_args arg;
4646         struct nfs_lock_res res;
4647         struct nfs4_lock_state *lsp;
4648         struct nfs_open_context *ctx;
4649         struct file_lock fl;
4650         unsigned long timestamp;
4651         int rpc_status;
4652         int cancelled;
4653         struct nfs_server *server;
4654 };
4655
4656 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4657                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4658                 gfp_t gfp_mask)
4659 {
4660         struct nfs4_lockdata *p;
4661         struct inode *inode = lsp->ls_state->inode;
4662         struct nfs_server *server = NFS_SERVER(inode);
4663
4664         p = kzalloc(sizeof(*p), gfp_mask);
4665         if (p == NULL)
4666                 return NULL;
4667
4668         p->arg.fh = NFS_FH(inode);
4669         p->arg.fl = &p->fl;
4670         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4671         if (p->arg.open_seqid == NULL)
4672                 goto out_free;
4673         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4674         if (p->arg.lock_seqid == NULL)
4675                 goto out_free_seqid;
4676         p->arg.lock_stateid = &lsp->ls_stateid;
4677         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4678         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4679         p->arg.lock_owner.s_dev = server->s_dev;
4680         p->res.lock_seqid = p->arg.lock_seqid;
4681         p->lsp = lsp;
4682         p->server = server;
4683         atomic_inc(&lsp->ls_count);
4684         p->ctx = get_nfs_open_context(ctx);
4685         memcpy(&p->fl, fl, sizeof(p->fl));
4686         return p;
4687 out_free_seqid:
4688         nfs_free_seqid(p->arg.open_seqid);
4689 out_free:
4690         kfree(p);
4691         return NULL;
4692 }
4693
4694 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4695 {
4696         struct nfs4_lockdata *data = calldata;
4697         struct nfs4_state *state = data->lsp->ls_state;
4698
4699         dprintk("%s: begin!\n", __func__);
4700         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4701                 return;
4702         /* Do we need to do an open_to_lock_owner? */
4703         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4704                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4705                         goto out_release_lock_seqid;
4706                 data->arg.open_stateid = &state->stateid;
4707                 data->arg.new_lock_owner = 1;
4708                 data->res.open_seqid = data->arg.open_seqid;
4709         } else
4710                 data->arg.new_lock_owner = 0;
4711         data->timestamp = jiffies;
4712         if (nfs4_setup_sequence(data->server,
4713                                 &data->arg.seq_args,
4714                                 &data->res.seq_res,
4715                                 task) == 0) {
4716                 rpc_call_start(task);
4717                 return;
4718         }
4719         nfs_release_seqid(data->arg.open_seqid);
4720 out_release_lock_seqid:
4721         nfs_release_seqid(data->arg.lock_seqid);
4722         dprintk("%s: done!, ret = %d\n", __func__, task->tk_status);
4723 }
4724
4725 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4726 {
4727         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4728         nfs4_lock_prepare(task, calldata);
4729 }
4730
4731 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4732 {
4733         struct nfs4_lockdata *data = calldata;
4734
4735         dprintk("%s: begin!\n", __func__);
4736
4737         if (!nfs4_sequence_done(task, &data->res.seq_res))
4738                 return;
4739
4740         data->rpc_status = task->tk_status;
4741         if (data->arg.new_lock_owner != 0) {
4742                 if (data->rpc_status == 0)
4743                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4744                 else
4745                         goto out;
4746         }
4747         if (data->rpc_status == 0) {
4748                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4749                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4750                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4751         }
4752 out:
4753         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4754 }
4755
4756 static void nfs4_lock_release(void *calldata)
4757 {
4758         struct nfs4_lockdata *data = calldata;
4759
4760         dprintk("%s: begin!\n", __func__);
4761         nfs_free_seqid(data->arg.open_seqid);
4762         if (data->cancelled != 0) {
4763                 struct rpc_task *task;
4764                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4765                                 data->arg.lock_seqid);
4766                 if (!IS_ERR(task))
4767                         rpc_put_task_async(task);
4768                 dprintk("%s: cancelling lock!\n", __func__);
4769         } else
4770                 nfs_free_seqid(data->arg.lock_seqid);
4771         nfs4_put_lock_state(data->lsp);
4772         put_nfs_open_context(data->ctx);
4773         kfree(data);
4774         dprintk("%s: done!\n", __func__);
4775 }
4776
4777 static const struct rpc_call_ops nfs4_lock_ops = {
4778         .rpc_call_prepare = nfs4_lock_prepare,
4779         .rpc_call_done = nfs4_lock_done,
4780         .rpc_release = nfs4_lock_release,
4781 };
4782
4783 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4784         .rpc_call_prepare = nfs4_recover_lock_prepare,
4785         .rpc_call_done = nfs4_lock_done,
4786         .rpc_release = nfs4_lock_release,
4787 };
4788
4789 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4790 {
4791         switch (error) {
4792         case -NFS4ERR_ADMIN_REVOKED:
4793         case -NFS4ERR_BAD_STATEID:
4794                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4795                 if (new_lock_owner != 0 ||
4796                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4797                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4798                 break;
4799         case -NFS4ERR_STALE_STATEID:
4800                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4801         case -NFS4ERR_EXPIRED:
4802                 nfs4_schedule_lease_recovery(server->nfs_client);
4803         };
4804 }
4805
4806 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4807 {
4808         struct nfs4_lockdata *data;
4809         struct rpc_task *task;
4810         struct rpc_message msg = {
4811                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4812                 .rpc_cred = state->owner->so_cred,
4813         };
4814         struct rpc_task_setup task_setup_data = {
4815                 .rpc_client = NFS_CLIENT(state->inode),
4816                 .rpc_message = &msg,
4817                 .callback_ops = &nfs4_lock_ops,
4818                 .workqueue = nfsiod_workqueue,
4819                 .flags = RPC_TASK_ASYNC,
4820         };
4821         int ret;
4822
4823         dprintk("%s: begin!\n", __func__);
4824         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4825                         fl->fl_u.nfs4_fl.owner,
4826                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4827         if (data == NULL)
4828                 return -ENOMEM;
4829         if (IS_SETLKW(cmd))
4830                 data->arg.block = 1;
4831         if (recovery_type > NFS_LOCK_NEW) {
4832                 if (recovery_type == NFS_LOCK_RECLAIM)
4833                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4834                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4835         }
4836         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4837         msg.rpc_argp = &data->arg;
4838         msg.rpc_resp = &data->res;
4839         task_setup_data.callback_data = data;
4840         task = rpc_run_task(&task_setup_data);
4841         if (IS_ERR(task))
4842                 return PTR_ERR(task);
4843         ret = nfs4_wait_for_completion_rpc_task(task);
4844         if (ret == 0) {
4845                 ret = data->rpc_status;
4846                 if (ret)
4847                         nfs4_handle_setlk_error(data->server, data->lsp,
4848                                         data->arg.new_lock_owner, ret);
4849         } else
4850                 data->cancelled = 1;
4851         rpc_put_task(task);
4852         dprintk("%s: done, ret = %d!\n", __func__, ret);
4853         return ret;
4854 }
4855
4856 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4857 {
4858         struct nfs_server *server = NFS_SERVER(state->inode);
4859         struct nfs4_exception exception = {
4860                 .inode = state->inode,
4861         };
4862         int err;
4863
4864         do {
4865                 /* Cache the lock if possible... */
4866                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4867                         return 0;
4868                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4869                 if (err != -NFS4ERR_DELAY)
4870                         break;
4871                 nfs4_handle_exception(server, err, &exception);
4872         } while (exception.retry);
4873         return err;
4874 }
4875
4876 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4877 {
4878         struct nfs_server *server = NFS_SERVER(state->inode);
4879         struct nfs4_exception exception = {
4880                 .inode = state->inode,
4881         };
4882         int err;
4883
4884         err = nfs4_set_lock_state(state, request);
4885         if (err != 0)
4886                 return err;
4887         do {
4888                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4889                         return 0;
4890                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4891                 switch (err) {
4892                 default:
4893                         goto out;
4894                 case -NFS4ERR_GRACE:
4895                 case -NFS4ERR_DELAY:
4896                         nfs4_handle_exception(server, err, &exception);
4897                         err = 0;
4898                 }
4899         } while (exception.retry);
4900 out:
4901         return err;
4902 }
4903
4904 #if defined(CONFIG_NFS_V4_1)
4905 /**
4906  * nfs41_check_expired_locks - possibly free a lock stateid
4907  *
4908  * @state: NFSv4 state for an inode
4909  *
4910  * Returns NFS_OK if recovery for this stateid is now finished.
4911  * Otherwise a negative NFS4ERR value is returned.
4912  */
4913 static int nfs41_check_expired_locks(struct nfs4_state *state)
4914 {
4915         int status, ret = -NFS4ERR_BAD_STATEID;
4916         struct nfs4_lock_state *lsp;
4917         struct nfs_server *server = NFS_SERVER(state->inode);
4918
4919         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4920                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
4921                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4922                         if (status != NFS_OK) {
4923                                 /* Free the stateid unless the server
4924                                  * informs us the stateid is unrecognized. */
4925                                 if (status != -NFS4ERR_BAD_STATEID)
4926                                         nfs41_free_stateid(server,
4927                                                         &lsp->ls_stateid);
4928                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
4929                                 ret = status;
4930                         }
4931                 }
4932         };
4933
4934         return ret;
4935 }
4936
4937 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4938 {
4939         int status = NFS_OK;
4940
4941         if (test_bit(LK_STATE_IN_USE, &state->flags))
4942                 status = nfs41_check_expired_locks(state);
4943         if (status != NFS_OK)
4944                 status = nfs4_lock_expired(state, request);
4945         return status;
4946 }
4947 #endif
4948
4949 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4950 {
4951         struct nfs_inode *nfsi = NFS_I(state->inode);
4952         unsigned char fl_flags = request->fl_flags;
4953         int status = -ENOLCK;
4954
4955         if ((fl_flags & FL_POSIX) &&
4956                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4957                 goto out;
4958         /* Is this a delegated open? */
4959         status = nfs4_set_lock_state(state, request);
4960         if (status != 0)
4961                 goto out;
4962         request->fl_flags |= FL_ACCESS;
4963         status = do_vfs_lock(request->fl_file, request);
4964         if (status < 0)
4965                 goto out;
4966         down_read(&nfsi->rwsem);
4967         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4968                 /* Yes: cache locks! */
4969                 /* ...but avoid races with delegation recall... */
4970                 request->fl_flags = fl_flags & ~FL_SLEEP;
4971                 status = do_vfs_lock(request->fl_file, request);
4972                 goto out_unlock;
4973         }
4974         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4975         if (status != 0)
4976                 goto out_unlock;
4977         /* Note: we always want to sleep here! */
4978         request->fl_flags = fl_flags | FL_SLEEP;
4979         if (do_vfs_lock(request->fl_file, request) < 0)
4980                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4981                         "manager!\n", __func__);
4982 out_unlock:
4983         up_read(&nfsi->rwsem);
4984 out:
4985         request->fl_flags = fl_flags;
4986         return status;
4987 }
4988
4989 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4990 {
4991         struct nfs4_exception exception = {
4992                 .state = state,
4993                 .inode = state->inode,
4994         };
4995         int err;
4996
4997         do {
4998                 err = _nfs4_proc_setlk(state, cmd, request);
4999                 if (err == -NFS4ERR_DENIED)
5000                         err = -EAGAIN;
5001                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5002                                 err, &exception);
5003         } while (exception.retry);
5004         return err;
5005 }
5006
5007 static int
5008 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5009 {
5010         struct nfs_open_context *ctx;
5011         struct nfs4_state *state;
5012         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5013         int status;
5014
5015         /* verify open state */
5016         ctx = nfs_file_open_context(filp);
5017         state = ctx->state;
5018
5019         if (request->fl_start < 0 || request->fl_end < 0)
5020                 return -EINVAL;
5021
5022         if (IS_GETLK(cmd)) {
5023                 if (state != NULL)
5024                         return nfs4_proc_getlk(state, F_GETLK, request);
5025                 return 0;
5026         }
5027
5028         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5029                 return -EINVAL;
5030
5031         if (request->fl_type == F_UNLCK) {
5032                 if (state != NULL)
5033                         return nfs4_proc_unlck(state, cmd, request);
5034                 return 0;
5035         }
5036
5037         if (state == NULL)
5038                 return -ENOLCK;
5039         /*
5040          * Don't rely on the VFS having checked the file open mode,
5041          * since it won't do this for flock() locks.
5042          */
5043         switch (request->fl_type) {
5044         case F_RDLCK:
5045                 if (!(filp->f_mode & FMODE_READ))
5046                         return -EBADF;
5047                 break;
5048         case F_WRLCK:
5049                 if (!(filp->f_mode & FMODE_WRITE))
5050                         return -EBADF;
5051         }
5052
5053         do {
5054                 status = nfs4_proc_setlk(state, cmd, request);
5055                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5056                         break;
5057                 timeout = nfs4_set_lock_task_retry(timeout);
5058                 status = -ERESTARTSYS;
5059                 if (signalled())
5060                         break;
5061         } while(status < 0);
5062         return status;
5063 }
5064
5065 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
5066 {
5067         struct nfs_server *server = NFS_SERVER(state->inode);
5068         struct nfs4_exception exception = { };
5069         int err;
5070
5071         err = nfs4_set_lock_state(state, fl);
5072         if (err != 0)
5073                 goto out;
5074         do {
5075                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5076                 switch (err) {
5077                         default:
5078                                 printk(KERN_ERR "NFS: %s: unhandled error "
5079                                         "%d.\n", __func__, err);
5080                         case 0:
5081                         case -ESTALE:
5082                                 goto out;
5083                         case -NFS4ERR_EXPIRED:
5084                                 nfs4_schedule_stateid_recovery(server, state);
5085                         case -NFS4ERR_STALE_CLIENTID:
5086                         case -NFS4ERR_STALE_STATEID:
5087                                 nfs4_schedule_lease_recovery(server->nfs_client);
5088                                 goto out;
5089                         case -NFS4ERR_BADSESSION:
5090                         case -NFS4ERR_BADSLOT:
5091                         case -NFS4ERR_BAD_HIGH_SLOT:
5092                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
5093                         case -NFS4ERR_DEADSESSION:
5094                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
5095                                 goto out;
5096                         case -ERESTARTSYS:
5097                                 /*
5098                                  * The show must go on: exit, but mark the
5099                                  * stateid as needing recovery.
5100                                  */
5101                         case -NFS4ERR_DELEG_REVOKED:
5102                         case -NFS4ERR_ADMIN_REVOKED:
5103                         case -NFS4ERR_BAD_STATEID:
5104                         case -NFS4ERR_OPENMODE:
5105                                 nfs4_schedule_stateid_recovery(server, state);
5106                                 err = 0;
5107                                 goto out;
5108                         case -EKEYEXPIRED:
5109                                 /*
5110                                  * User RPCSEC_GSS context has expired.
5111                                  * We cannot recover this stateid now, so
5112                                  * skip it and allow recovery thread to
5113                                  * proceed.
5114                                  */
5115                                 err = 0;
5116                                 goto out;
5117                         case -ENOMEM:
5118                         case -NFS4ERR_DENIED:
5119                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
5120                                 err = 0;
5121                                 goto out;
5122                         case -NFS4ERR_DELAY:
5123                                 break;
5124                 }
5125                 err = nfs4_handle_exception(server, err, &exception);
5126         } while (exception.retry);
5127 out:
5128         return err;
5129 }
5130
5131 struct nfs_release_lockowner_data {
5132         struct nfs4_lock_state *lsp;
5133         struct nfs_server *server;
5134         struct nfs_release_lockowner_args args;
5135 };
5136
5137 static void nfs4_release_lockowner_release(void *calldata)
5138 {
5139         struct nfs_release_lockowner_data *data = calldata;
5140         nfs4_free_lock_state(data->server, data->lsp);
5141         kfree(calldata);
5142 }
5143
5144 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5145         .rpc_release = nfs4_release_lockowner_release,
5146 };
5147
5148 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
5149 {
5150         struct nfs_server *server = lsp->ls_state->owner->so_server;
5151         struct nfs_release_lockowner_data *data;
5152         struct rpc_message msg = {
5153                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5154         };
5155
5156         if (server->nfs_client->cl_mvops->minor_version != 0)
5157                 return -EINVAL;
5158         data = kmalloc(sizeof(*data), GFP_NOFS);
5159         if (!data)
5160                 return -ENOMEM;
5161         data->lsp = lsp;
5162         data->server = server;
5163         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5164         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5165         data->args.lock_owner.s_dev = server->s_dev;
5166         msg.rpc_argp = &data->args;
5167         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5168         return 0;
5169 }
5170
5171 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5172
5173 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5174                                    const void *buf, size_t buflen,
5175                                    int flags, int type)
5176 {
5177         if (strcmp(key, "") != 0)
5178                 return -EINVAL;
5179
5180         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5181 }
5182
5183 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5184                                    void *buf, size_t buflen, int type)
5185 {
5186         if (strcmp(key, "") != 0)
5187                 return -EINVAL;
5188
5189         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5190 }
5191
5192 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5193                                        size_t list_len, const char *name,
5194                                        size_t name_len, int type)
5195 {
5196         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5197
5198         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5199                 return 0;
5200
5201         if (list && len <= list_len)
5202                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5203         return len;
5204 }
5205
5206 /*
5207  * nfs_fhget will use either the mounted_on_fileid or the fileid
5208  */
5209 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5210 {
5211         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5212                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5213               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5214               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5215                 return;
5216
5217         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5218                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5219         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5220         fattr->nlink = 2;
5221 }
5222
5223 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5224                                    const struct qstr *name,
5225                                    struct nfs4_fs_locations *fs_locations,
5226                                    struct page *page)
5227 {
5228         struct nfs_server *server = NFS_SERVER(dir);
5229         u32 bitmask[2] = {
5230                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5231         };
5232         struct nfs4_fs_locations_arg args = {
5233                 .dir_fh = NFS_FH(dir),
5234                 .name = name,
5235                 .page = page,
5236                 .bitmask = bitmask,
5237         };
5238         struct nfs4_fs_locations_res res = {
5239                 .fs_locations = fs_locations,
5240         };
5241         struct rpc_message msg = {
5242                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5243                 .rpc_argp = &args,
5244                 .rpc_resp = &res,
5245         };
5246         int status;
5247
5248         dprintk("%s: start\n", __func__);
5249
5250         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5251          * is not supported */
5252         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5253                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5254         else
5255                 bitmask[0] |= FATTR4_WORD0_FILEID;
5256
5257         nfs_fattr_init(&fs_locations->fattr);
5258         fs_locations->server = server;
5259         fs_locations->nlocations = 0;
5260         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5261         dprintk("%s: returned status = %d\n", __func__, status);
5262         return status;
5263 }
5264
5265 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5266                            const struct qstr *name,
5267                            struct nfs4_fs_locations *fs_locations,
5268                            struct page *page)
5269 {
5270         struct nfs4_exception exception = { };
5271         int err;
5272         do {
5273                 err = nfs4_handle_exception(NFS_SERVER(dir),
5274                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5275                                 &exception);
5276         } while (exception.retry);
5277         return err;
5278 }
5279
5280 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5281 {
5282         int status;
5283         struct nfs4_secinfo_arg args = {
5284                 .dir_fh = NFS_FH(dir),
5285                 .name   = name,
5286         };
5287         struct nfs4_secinfo_res res = {
5288                 .flavors     = flavors,
5289         };
5290         struct rpc_message msg = {
5291                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5292                 .rpc_argp = &args,
5293                 .rpc_resp = &res,
5294         };
5295
5296         dprintk("NFS call  secinfo %s\n", name->name);
5297         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5298         dprintk("NFS reply  secinfo: %d\n", status);
5299         return status;
5300 }
5301
5302 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5303                       struct nfs4_secinfo_flavors *flavors)
5304 {
5305         struct nfs4_exception exception = { };
5306         int err;
5307         do {
5308                 err = nfs4_handle_exception(NFS_SERVER(dir),
5309                                 _nfs4_proc_secinfo(dir, name, flavors),
5310                                 &exception);
5311         } while (exception.retry);
5312         return err;
5313 }
5314
5315 #ifdef CONFIG_NFS_V4_1
5316 /*
5317  * Check the exchange flags returned by the server for invalid flags, having
5318  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5319  * DS flags set.
5320  */
5321 static int nfs4_check_cl_exchange_flags(u32 flags)
5322 {
5323         if (flags & ~EXCHGID4_FLAG_MASK_R)
5324                 goto out_inval;
5325         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5326             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5327                 goto out_inval;
5328         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5329                 goto out_inval;
5330         return NFS_OK;
5331 out_inval:
5332         return -NFS4ERR_INVAL;
5333 }
5334
5335 static bool
5336 nfs41_same_server_scope(struct nfs41_server_scope *a,
5337                         struct nfs41_server_scope *b)
5338 {
5339         if (a->server_scope_sz == b->server_scope_sz &&
5340             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5341                 return true;
5342
5343         return false;
5344 }
5345
5346 /*
5347  * nfs4_proc_bind_conn_to_session()
5348  *
5349  * The 4.1 client currently uses the same TCP connection for the
5350  * fore and backchannel.
5351  */
5352 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5353 {
5354         int status;
5355         struct nfs41_bind_conn_to_session_res res;
5356         struct rpc_message msg = {
5357                 .rpc_proc =
5358                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5359                 .rpc_argp = clp,
5360                 .rpc_resp = &res,
5361                 .rpc_cred = cred,
5362         };
5363
5364         dprintk("--> %s\n", __func__);
5365
5366         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5367         if (unlikely(res.session == NULL)) {
5368                 status = -ENOMEM;
5369                 goto out;
5370         }
5371
5372         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5373         if (status == 0) {
5374                 if (memcmp(res.session->sess_id.data,
5375                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5376                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5377                         status = -EIO;
5378                         goto out_session;
5379                 }
5380                 if (res.dir != NFS4_CDFS4_BOTH) {
5381                         dprintk("NFS: %s: Unexpected direction from server\n",
5382                                 __func__);
5383                         status = -EIO;
5384                         goto out_session;
5385                 }
5386                 if (res.use_conn_in_rdma_mode) {
5387                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5388                                 __func__);
5389                         status = -EIO;
5390                         goto out_session;
5391                 }
5392         }
5393 out_session:
5394         kfree(res.session);
5395 out:
5396         dprintk("<-- %s status= %d\n", __func__, status);
5397         return status;
5398 }
5399
5400 /*
5401  * nfs4_proc_exchange_id()
5402  *
5403  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5404  *
5405  * Since the clientid has expired, all compounds using sessions
5406  * associated with the stale clientid will be returning
5407  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5408  * be in some phase of session reset.
5409  */
5410 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5411 {
5412         nfs4_verifier verifier;
5413         struct nfs41_exchange_id_args args = {
5414                 .verifier = &verifier,
5415                 .client = clp,
5416                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5417         };
5418         struct nfs41_exchange_id_res res = {
5419                 0
5420         };
5421         int status;
5422         struct rpc_message msg = {
5423                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5424                 .rpc_argp = &args,
5425                 .rpc_resp = &res,
5426                 .rpc_cred = cred,
5427         };
5428
5429         nfs4_init_boot_verifier(clp, &verifier);
5430         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5431                                                         sizeof(args.id));
5432         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5433                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5434                 args.id_len, args.id);
5435
5436         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5437                                         GFP_NOFS);
5438         if (unlikely(res.server_owner == NULL)) {
5439                 status = -ENOMEM;
5440                 goto out;
5441         }
5442
5443         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5444                                         GFP_NOFS);
5445         if (unlikely(res.server_scope == NULL)) {
5446                 status = -ENOMEM;
5447                 goto out_server_owner;
5448         }
5449
5450         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5451         if (unlikely(res.impl_id == NULL)) {
5452                 status = -ENOMEM;
5453                 goto out_server_scope;
5454         }
5455
5456         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5457         if (status == 0)
5458                 status = nfs4_check_cl_exchange_flags(res.flags);
5459
5460         if (status == 0) {
5461                 clp->cl_clientid = res.clientid;
5462                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5463                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5464                         clp->cl_seqid = res.seqid;
5465
5466                 kfree(clp->cl_serverowner);
5467                 clp->cl_serverowner = res.server_owner;
5468                 res.server_owner = NULL;
5469
5470                 /* use the most recent implementation id */
5471                 kfree(clp->cl_implid);
5472                 clp->cl_implid = res.impl_id;
5473
5474                 if (clp->cl_serverscope != NULL &&
5475                     !nfs41_same_server_scope(clp->cl_serverscope,
5476                                              res.server_scope)) {
5477                         dprintk("%s: server_scope mismatch detected\n",
5478                                 __func__);
5479                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5480                         kfree(clp->cl_serverscope);
5481                         clp->cl_serverscope = NULL;
5482                 }
5483
5484                 if (clp->cl_serverscope == NULL) {
5485                         clp->cl_serverscope = res.server_scope;
5486                         goto out;
5487                 }
5488         } else
5489                 kfree(res.impl_id);
5490
5491 out_server_owner:
5492         kfree(res.server_owner);
5493 out_server_scope:
5494         kfree(res.server_scope);
5495 out:
5496         if (clp->cl_implid != NULL)
5497                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5498                         "domain: %s, name: %s, date: %llu,%u\n",
5499                         clp->cl_implid->domain, clp->cl_implid->name,
5500                         clp->cl_implid->date.seconds,
5501                         clp->cl_implid->date.nseconds);
5502         dprintk("NFS reply exchange_id: %d\n", status);
5503         return status;
5504 }
5505
5506 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5507                 struct rpc_cred *cred)
5508 {
5509         struct rpc_message msg = {
5510                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5511                 .rpc_argp = clp,
5512                 .rpc_cred = cred,
5513         };
5514         int status;
5515
5516         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5517         if (status)
5518                 dprintk("NFS: Got error %d from the server %s on "
5519                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5520         return status;
5521 }
5522
5523 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5524                 struct rpc_cred *cred)
5525 {
5526         unsigned int loop;
5527         int ret;
5528
5529         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5530                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5531                 switch (ret) {
5532                 case -NFS4ERR_DELAY:
5533                 case -NFS4ERR_CLIENTID_BUSY:
5534                         ssleep(1);
5535                         break;
5536                 default:
5537                         return ret;
5538                 }
5539         }
5540         return 0;
5541 }
5542
5543 int nfs4_destroy_clientid(struct nfs_client *clp)
5544 {
5545         struct rpc_cred *cred;
5546         int ret = 0;
5547
5548         if (clp->cl_mvops->minor_version < 1)
5549                 goto out;
5550         if (clp->cl_exchange_flags == 0)
5551                 goto out;
5552         if (clp->cl_preserve_clid)
5553                 goto out;
5554         cred = nfs4_get_exchange_id_cred(clp);
5555         ret = nfs4_proc_destroy_clientid(clp, cred);
5556         if (cred)
5557                 put_rpccred(cred);
5558         switch (ret) {
5559         case 0:
5560         case -NFS4ERR_STALE_CLIENTID:
5561                 clp->cl_exchange_flags = 0;
5562         }
5563 out:
5564         return ret;
5565 }
5566
5567 struct nfs4_get_lease_time_data {
5568         struct nfs4_get_lease_time_args *args;
5569         struct nfs4_get_lease_time_res *res;
5570         struct nfs_client *clp;
5571 };
5572
5573 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5574                                         void *calldata)
5575 {
5576         int ret;
5577         struct nfs4_get_lease_time_data *data =
5578                         (struct nfs4_get_lease_time_data *)calldata;
5579
5580         dprintk("--> %s\n", __func__);
5581         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5582         /* just setup sequence, do not trigger session recovery
5583            since we're invoked within one */
5584         ret = nfs41_setup_sequence(data->clp->cl_session,
5585                                    &data->args->la_seq_args,
5586                                    &data->res->lr_seq_res, task);
5587
5588         if (ret != -EAGAIN)
5589                 rpc_call_start(task);
5590         dprintk("<-- %s\n", __func__);
5591 }
5592
5593 /*
5594  * Called from nfs4_state_manager thread for session setup, so don't recover
5595  * from sequence operation or clientid errors.
5596  */
5597 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5598 {
5599         struct nfs4_get_lease_time_data *data =
5600                         (struct nfs4_get_lease_time_data *)calldata;
5601
5602         dprintk("--> %s\n", __func__);
5603         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5604                 return;
5605         switch (task->tk_status) {
5606         case -NFS4ERR_DELAY:
5607         case -NFS4ERR_GRACE:
5608                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5609                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5610                 task->tk_status = 0;
5611                 /* fall through */
5612         case -NFS4ERR_RETRY_UNCACHED_REP:
5613                 rpc_restart_call_prepare(task);
5614                 return;
5615         }
5616         dprintk("<-- %s\n", __func__);
5617 }
5618
5619 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5620         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5621         .rpc_call_done = nfs4_get_lease_time_done,
5622 };
5623
5624 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5625 {
5626         struct rpc_task *task;
5627         struct nfs4_get_lease_time_args args;
5628         struct nfs4_get_lease_time_res res = {
5629                 .lr_fsinfo = fsinfo,
5630         };
5631         struct nfs4_get_lease_time_data data = {
5632                 .args = &args,
5633                 .res = &res,
5634                 .clp = clp,
5635         };
5636         struct rpc_message msg = {
5637                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5638                 .rpc_argp = &args,
5639                 .rpc_resp = &res,
5640         };
5641         struct rpc_task_setup task_setup = {
5642                 .rpc_client = clp->cl_rpcclient,
5643                 .rpc_message = &msg,
5644                 .callback_ops = &nfs4_get_lease_time_ops,
5645                 .callback_data = &data,
5646                 .flags = RPC_TASK_TIMEOUT,
5647         };
5648         int status;
5649
5650         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5651         dprintk("--> %s\n", __func__);
5652         task = rpc_run_task(&task_setup);
5653
5654         if (IS_ERR(task))
5655                 status = PTR_ERR(task);
5656         else {
5657                 status = task->tk_status;
5658                 rpc_put_task(task);
5659         }
5660         dprintk("<-- %s return %d\n", __func__, status);
5661
5662         return status;
5663 }
5664
5665 struct nfs4_slot *nfs4_alloc_slots(struct nfs4_slot_table *table,
5666                 u32 max_slots, gfp_t gfp_flags)
5667 {
5668         struct nfs4_slot *tbl;
5669         u32 i;
5670
5671         tbl = kmalloc_array(max_slots, sizeof(*tbl), gfp_flags);
5672         if (tbl != NULL) {
5673                 for (i = 0; i < max_slots; i++) {
5674                         tbl[i].table = table;
5675                         tbl[i].slot_nr = i;
5676                 }
5677         }
5678         return tbl;
5679 }
5680
5681 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5682                 struct nfs4_slot *new,
5683                 u32 max_slots,
5684                 u32 ivalue)
5685 {
5686         struct nfs4_slot *old = NULL;
5687         u32 i;
5688
5689         spin_lock(&tbl->slot_tbl_lock);
5690         if (new) {
5691                 old = tbl->slots;
5692                 tbl->slots = new;
5693                 tbl->max_slots = max_slots;
5694         }
5695         tbl->highest_used_slotid = NFS4_NO_SLOT;
5696         for (i = 0; i < tbl->max_slots; i++)
5697                 tbl->slots[i].seq_nr = ivalue;
5698         spin_unlock(&tbl->slot_tbl_lock);
5699         kfree(old);
5700 }
5701
5702 /*
5703  * (re)Initialise a slot table
5704  */
5705 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5706                                  u32 ivalue)
5707 {
5708         struct nfs4_slot *new = NULL;
5709         int ret = -ENOMEM;
5710
5711         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5712                 max_reqs, tbl->max_slots);
5713
5714         /* Does the newly negotiated max_reqs match the existing slot table? */
5715         if (max_reqs != tbl->max_slots) {
5716                 new = nfs4_alloc_slots(tbl, max_reqs, GFP_NOFS);
5717                 if (!new)
5718                         goto out;
5719         }
5720         ret = 0;
5721
5722         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5723         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5724                 tbl, tbl->slots, tbl->max_slots);
5725 out:
5726         dprintk("<-- %s: return %d\n", __func__, ret);
5727         return ret;
5728 }
5729
5730 /* Destroy the slot table */
5731 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5732 {
5733         if (session->fc_slot_table.slots != NULL) {
5734                 kfree(session->fc_slot_table.slots);
5735                 session->fc_slot_table.slots = NULL;
5736         }
5737         if (session->bc_slot_table.slots != NULL) {
5738                 kfree(session->bc_slot_table.slots);
5739                 session->bc_slot_table.slots = NULL;
5740         }
5741         return;
5742 }
5743
5744 /*
5745  * Initialize or reset the forechannel and backchannel tables
5746  */
5747 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5748 {
5749         struct nfs4_slot_table *tbl;
5750         int status;
5751
5752         dprintk("--> %s\n", __func__);
5753         /* Fore channel */
5754         tbl = &ses->fc_slot_table;
5755         tbl->session = ses;
5756         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5757         if (status) /* -ENOMEM */
5758                 return status;
5759         /* Back channel */
5760         tbl = &ses->bc_slot_table;
5761         tbl->session = ses;
5762         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5763         if (status && tbl->slots == NULL)
5764                 /* Fore and back channel share a connection so get
5765                  * both slot tables or neither */
5766                 nfs4_destroy_slot_tables(ses);
5767         return status;
5768 }
5769
5770 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5771 {
5772         struct nfs4_session *session;
5773         struct nfs4_slot_table *tbl;
5774
5775         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5776         if (!session)
5777                 return NULL;
5778
5779         tbl = &session->fc_slot_table;
5780         tbl->highest_used_slotid = NFS4_NO_SLOT;
5781         spin_lock_init(&tbl->slot_tbl_lock);
5782         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5783         init_completion(&tbl->complete);
5784
5785         tbl = &session->bc_slot_table;
5786         tbl->highest_used_slotid = NFS4_NO_SLOT;
5787         spin_lock_init(&tbl->slot_tbl_lock);
5788         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5789         init_completion(&tbl->complete);
5790
5791         session->session_state = 1<<NFS4_SESSION_INITING;
5792
5793         session->clp = clp;
5794         return session;
5795 }
5796
5797 void nfs4_destroy_session(struct nfs4_session *session)
5798 {
5799         struct rpc_xprt *xprt;
5800         struct rpc_cred *cred;
5801
5802         cred = nfs4_get_exchange_id_cred(session->clp);
5803         nfs4_proc_destroy_session(session, cred);
5804         if (cred)
5805                 put_rpccred(cred);
5806
5807         rcu_read_lock();
5808         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5809         rcu_read_unlock();
5810         dprintk("%s Destroy backchannel for xprt %p\n",
5811                 __func__, xprt);
5812         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5813         nfs4_destroy_slot_tables(session);
5814         kfree(session);
5815 }
5816
5817 /*
5818  * Initialize the values to be used by the client in CREATE_SESSION
5819  * If nfs4_init_session set the fore channel request and response sizes,
5820  * use them.
5821  *
5822  * Set the back channel max_resp_sz_cached to zero to force the client to
5823  * always set csa_cachethis to FALSE because the current implementation
5824  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5825  */
5826 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5827 {
5828         struct nfs4_session *session = args->client->cl_session;
5829         unsigned int mxrqst_sz = session->fc_target_max_rqst_sz,
5830                      mxresp_sz = session->fc_target_max_resp_sz;
5831
5832         if (mxrqst_sz == 0)
5833                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5834         if (mxresp_sz == 0)
5835                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5836         /* Fore channel attributes */
5837         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5838         args->fc_attrs.max_resp_sz = mxresp_sz;
5839         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5840         args->fc_attrs.max_reqs = max_session_slots;
5841
5842         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5843                 "max_ops=%u max_reqs=%u\n",
5844                 __func__,
5845                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5846                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5847
5848         /* Back channel attributes */
5849         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5850         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5851         args->bc_attrs.max_resp_sz_cached = 0;
5852         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5853         args->bc_attrs.max_reqs = 1;
5854
5855         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5856                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5857                 __func__,
5858                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5859                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5860                 args->bc_attrs.max_reqs);
5861 }
5862
5863 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5864 {
5865         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5866         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5867
5868         if (rcvd->max_resp_sz > sent->max_resp_sz)
5869                 return -EINVAL;
5870         /*
5871          * Our requested max_ops is the minimum we need; we're not
5872          * prepared to break up compounds into smaller pieces than that.
5873          * So, no point even trying to continue if the server won't
5874          * cooperate:
5875          */
5876         if (rcvd->max_ops < sent->max_ops)
5877                 return -EINVAL;
5878         if (rcvd->max_reqs == 0)
5879                 return -EINVAL;
5880         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5881                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5882         return 0;
5883 }
5884
5885 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5886 {
5887         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5888         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5889
5890         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5891                 return -EINVAL;
5892         if (rcvd->max_resp_sz < sent->max_resp_sz)
5893                 return -EINVAL;
5894         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5895                 return -EINVAL;
5896         /* These would render the backchannel useless: */
5897         if (rcvd->max_ops != sent->max_ops)
5898                 return -EINVAL;
5899         if (rcvd->max_reqs != sent->max_reqs)
5900                 return -EINVAL;
5901         return 0;
5902 }
5903
5904 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5905                                      struct nfs4_session *session)
5906 {
5907         int ret;
5908
5909         ret = nfs4_verify_fore_channel_attrs(args, session);
5910         if (ret)
5911                 return ret;
5912         return nfs4_verify_back_channel_attrs(args, session);
5913 }
5914
5915 static int _nfs4_proc_create_session(struct nfs_client *clp,
5916                 struct rpc_cred *cred)
5917 {
5918         struct nfs4_session *session = clp->cl_session;
5919         struct nfs41_create_session_args args = {
5920                 .client = clp,
5921                 .cb_program = NFS4_CALLBACK,
5922         };
5923         struct nfs41_create_session_res res = {
5924                 .client = clp,
5925         };
5926         struct rpc_message msg = {
5927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5928                 .rpc_argp = &args,
5929                 .rpc_resp = &res,
5930                 .rpc_cred = cred,
5931         };
5932         int status;
5933
5934         nfs4_init_channel_attrs(&args);
5935         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5936
5937         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5938
5939         if (!status) {
5940                 /* Verify the session's negotiated channel_attrs values */
5941                 status = nfs4_verify_channel_attrs(&args, session);
5942                 /* Increment the clientid slot sequence id */
5943                 clp->cl_seqid++;
5944         }
5945
5946         return status;
5947 }
5948
5949 /*
5950  * Issues a CREATE_SESSION operation to the server.
5951  * It is the responsibility of the caller to verify the session is
5952  * expired before calling this routine.
5953  */
5954 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5955 {
5956         int status;
5957         unsigned *ptr;
5958         struct nfs4_session *session = clp->cl_session;
5959
5960         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5961
5962         status = _nfs4_proc_create_session(clp, cred);
5963         if (status)
5964                 goto out;
5965
5966         /* Init or reset the session slot tables */
5967         status = nfs4_setup_session_slot_tables(session);
5968         dprintk("slot table setup returned %d\n", status);
5969         if (status)
5970                 goto out;
5971
5972         ptr = (unsigned *)&session->sess_id.data[0];
5973         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5974                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5975 out:
5976         dprintk("<-- %s\n", __func__);
5977         return status;
5978 }
5979
5980 /*
5981  * Issue the over-the-wire RPC DESTROY_SESSION.
5982  * The caller must serialize access to this routine.
5983  */
5984 int nfs4_proc_destroy_session(struct nfs4_session *session,
5985                 struct rpc_cred *cred)
5986 {
5987         struct rpc_message msg = {
5988                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5989                 .rpc_argp = session,
5990                 .rpc_cred = cred,
5991         };
5992         int status = 0;
5993
5994         dprintk("--> nfs4_proc_destroy_session\n");
5995
5996         /* session is still being setup */
5997         if (session->clp->cl_cons_state != NFS_CS_READY)
5998                 return status;
5999
6000         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6001
6002         if (status)
6003                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
6004                         "Session has been destroyed regardless...\n", status);
6005
6006         dprintk("<-- nfs4_proc_destroy_session\n");
6007         return status;
6008 }
6009
6010 /*
6011  * With sessions, the client is not marked ready until after a
6012  * successful EXCHANGE_ID and CREATE_SESSION.
6013  *
6014  * Map errors cl_cons_state errors to EPROTONOSUPPORT to indicate
6015  * other versions of NFS can be tried.
6016  */
6017 static int nfs41_check_session_ready(struct nfs_client *clp)
6018 {
6019         int ret;
6020         
6021         if (clp->cl_cons_state == NFS_CS_SESSION_INITING) {
6022                 ret = nfs4_client_recover_expired_lease(clp);
6023                 if (ret)
6024                         return ret;
6025         }
6026         if (clp->cl_cons_state < NFS_CS_READY)
6027                 return -EPROTONOSUPPORT;
6028         smp_rmb();
6029         return 0;
6030 }
6031
6032 int nfs4_init_session(struct nfs_server *server)
6033 {
6034         struct nfs_client *clp = server->nfs_client;
6035         struct nfs4_session *session;
6036         unsigned int target_max_rqst_sz = NFS_MAX_FILE_IO_SIZE;
6037         unsigned int target_max_resp_sz = NFS_MAX_FILE_IO_SIZE;
6038
6039         if (!nfs4_has_session(clp))
6040                 return 0;
6041
6042         if (server->rsize != 0)
6043                 target_max_resp_sz = server->rsize;
6044         target_max_resp_sz += nfs41_maxread_overhead;
6045
6046         if (server->wsize != 0)
6047                 target_max_rqst_sz = server->wsize;
6048         target_max_rqst_sz += nfs41_maxwrite_overhead;
6049
6050         session = clp->cl_session;
6051         spin_lock(&clp->cl_lock);
6052         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
6053                 /* Initialise targets and channel attributes */
6054                 session->fc_target_max_rqst_sz = target_max_rqst_sz;
6055                 session->fc_attrs.max_rqst_sz = target_max_rqst_sz;
6056                 session->fc_target_max_resp_sz = target_max_resp_sz;
6057                 session->fc_attrs.max_resp_sz = target_max_resp_sz;
6058         } else {
6059                 /* Just adjust the targets */
6060                 if (target_max_rqst_sz > session->fc_target_max_rqst_sz) {
6061                         session->fc_target_max_rqst_sz = target_max_rqst_sz;
6062                         set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
6063                 }
6064                 if (target_max_resp_sz > session->fc_target_max_resp_sz) {
6065                         session->fc_target_max_resp_sz = target_max_resp_sz;
6066                         set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
6067                 }
6068         }
6069         spin_unlock(&clp->cl_lock);
6070
6071         if (test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
6072                 nfs4_schedule_lease_recovery(clp);
6073
6074         return nfs41_check_session_ready(clp);
6075 }
6076
6077 int nfs4_init_ds_session(struct nfs_client *clp, unsigned long lease_time)
6078 {
6079         struct nfs4_session *session = clp->cl_session;
6080         int ret;
6081
6082         spin_lock(&clp->cl_lock);
6083         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
6084                 /*
6085                  * Do not set NFS_CS_CHECK_LEASE_TIME instead set the
6086                  * DS lease to be equal to the MDS lease.
6087                  */
6088                 clp->cl_lease_time = lease_time;
6089                 clp->cl_last_renewal = jiffies;
6090         }
6091         spin_unlock(&clp->cl_lock);
6092
6093         ret = nfs41_check_session_ready(clp);
6094         if (ret)
6095                 return ret;
6096         /* Test for the DS role */
6097         if (!is_ds_client(clp))
6098                 return -ENODEV;
6099         return 0;
6100 }
6101 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
6102
6103
6104 /*
6105  * Renew the cl_session lease.
6106  */
6107 struct nfs4_sequence_data {
6108         struct nfs_client *clp;
6109         struct nfs4_sequence_args args;
6110         struct nfs4_sequence_res res;
6111 };
6112
6113 static void nfs41_sequence_release(void *data)
6114 {
6115         struct nfs4_sequence_data *calldata = data;
6116         struct nfs_client *clp = calldata->clp;
6117
6118         if (atomic_read(&clp->cl_count) > 1)
6119                 nfs4_schedule_state_renewal(clp);
6120         nfs_put_client(clp);
6121         kfree(calldata);
6122 }
6123
6124 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6125 {
6126         switch(task->tk_status) {
6127         case -NFS4ERR_DELAY:
6128                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6129                 return -EAGAIN;
6130         default:
6131                 nfs4_schedule_lease_recovery(clp);
6132         }
6133         return 0;
6134 }
6135
6136 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
6137 {
6138         struct nfs4_sequence_data *calldata = data;
6139         struct nfs_client *clp = calldata->clp;
6140
6141         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
6142                 return;
6143
6144         if (task->tk_status < 0) {
6145                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
6146                 if (atomic_read(&clp->cl_count) == 1)
6147                         goto out;
6148
6149                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
6150                         rpc_restart_call_prepare(task);
6151                         return;
6152                 }
6153         }
6154         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
6155 out:
6156         dprintk("<-- %s\n", __func__);
6157 }
6158
6159 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
6160 {
6161         struct nfs4_sequence_data *calldata = data;
6162         struct nfs_client *clp = calldata->clp;
6163         struct nfs4_sequence_args *args;
6164         struct nfs4_sequence_res *res;
6165
6166         args = task->tk_msg.rpc_argp;
6167         res = task->tk_msg.rpc_resp;
6168
6169         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
6170                 return;
6171         rpc_call_start(task);
6172 }
6173
6174 static void nfs41_sequence_prepare_privileged(struct rpc_task *task, void *data)
6175 {
6176         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
6177         nfs41_sequence_prepare(task, data);
6178 }
6179
6180 static const struct rpc_call_ops nfs41_sequence_ops = {
6181         .rpc_call_done = nfs41_sequence_call_done,
6182         .rpc_call_prepare = nfs41_sequence_prepare,
6183         .rpc_release = nfs41_sequence_release,
6184 };
6185
6186 static const struct rpc_call_ops nfs41_sequence_privileged_ops = {
6187         .rpc_call_done = nfs41_sequence_call_done,
6188         .rpc_call_prepare = nfs41_sequence_prepare_privileged,
6189         .rpc_release = nfs41_sequence_release,
6190 };
6191
6192 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred,
6193                                              const struct rpc_call_ops *seq_ops)
6194 {
6195         struct nfs4_sequence_data *calldata;
6196         struct rpc_message msg = {
6197                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6198                 .rpc_cred = cred,
6199         };
6200         struct rpc_task_setup task_setup_data = {
6201                 .rpc_client = clp->cl_rpcclient,
6202                 .rpc_message = &msg,
6203                 .callback_ops = seq_ops,
6204                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
6205         };
6206
6207         if (!atomic_inc_not_zero(&clp->cl_count))
6208                 return ERR_PTR(-EIO);
6209         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6210         if (calldata == NULL) {
6211                 nfs_put_client(clp);
6212                 return ERR_PTR(-ENOMEM);
6213         }
6214         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6215         msg.rpc_argp = &calldata->args;
6216         msg.rpc_resp = &calldata->res;
6217         calldata->clp = clp;
6218         task_setup_data.callback_data = calldata;
6219
6220         return rpc_run_task(&task_setup_data);
6221 }
6222
6223 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6224 {
6225         struct rpc_task *task;
6226         int ret = 0;
6227
6228         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6229                 return 0;
6230         task = _nfs41_proc_sequence(clp, cred, &nfs41_sequence_ops);
6231         if (IS_ERR(task))
6232                 ret = PTR_ERR(task);
6233         else
6234                 rpc_put_task_async(task);
6235         dprintk("<-- %s status=%d\n", __func__, ret);
6236         return ret;
6237 }
6238
6239 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6240 {
6241         struct rpc_task *task;
6242         int ret;
6243
6244         task = _nfs41_proc_sequence(clp, cred, &nfs41_sequence_privileged_ops);
6245         if (IS_ERR(task)) {
6246                 ret = PTR_ERR(task);
6247                 goto out;
6248         }
6249         ret = rpc_wait_for_completion_task(task);
6250         if (!ret) {
6251                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6252
6253                 if (task->tk_status == 0)
6254                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6255                 ret = task->tk_status;
6256         }
6257         rpc_put_task(task);
6258 out:
6259         dprintk("<-- %s status=%d\n", __func__, ret);
6260         return ret;
6261 }
6262
6263 struct nfs4_reclaim_complete_data {
6264         struct nfs_client *clp;
6265         struct nfs41_reclaim_complete_args arg;
6266         struct nfs41_reclaim_complete_res res;
6267 };
6268
6269 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6270 {
6271         struct nfs4_reclaim_complete_data *calldata = data;
6272
6273         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
6274         if (nfs41_setup_sequence(calldata->clp->cl_session,
6275                                 &calldata->arg.seq_args,
6276                                 &calldata->res.seq_res, task))
6277                 return;
6278
6279         rpc_call_start(task);
6280 }
6281
6282 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6283 {
6284         switch(task->tk_status) {
6285         case 0:
6286         case -NFS4ERR_COMPLETE_ALREADY:
6287         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6288                 break;
6289         case -NFS4ERR_DELAY:
6290                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6291                 /* fall through */
6292         case -NFS4ERR_RETRY_UNCACHED_REP:
6293                 return -EAGAIN;
6294         default:
6295                 nfs4_schedule_lease_recovery(clp);
6296         }
6297         return 0;
6298 }
6299
6300 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6301 {
6302         struct nfs4_reclaim_complete_data *calldata = data;
6303         struct nfs_client *clp = calldata->clp;
6304         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6305
6306         dprintk("--> %s\n", __func__);
6307         if (!nfs41_sequence_done(task, res))
6308                 return;
6309
6310         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6311                 rpc_restart_call_prepare(task);
6312                 return;
6313         }
6314         dprintk("<-- %s\n", __func__);
6315 }
6316
6317 static void nfs4_free_reclaim_complete_data(void *data)
6318 {
6319         struct nfs4_reclaim_complete_data *calldata = data;
6320
6321         kfree(calldata);
6322 }
6323
6324 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6325         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6326         .rpc_call_done = nfs4_reclaim_complete_done,
6327         .rpc_release = nfs4_free_reclaim_complete_data,
6328 };
6329
6330 /*
6331  * Issue a global reclaim complete.
6332  */
6333 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6334 {
6335         struct nfs4_reclaim_complete_data *calldata;
6336         struct rpc_task *task;
6337         struct rpc_message msg = {
6338                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6339         };
6340         struct rpc_task_setup task_setup_data = {
6341                 .rpc_client = clp->cl_rpcclient,
6342                 .rpc_message = &msg,
6343                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6344                 .flags = RPC_TASK_ASYNC,
6345         };
6346         int status = -ENOMEM;
6347
6348         dprintk("--> %s\n", __func__);
6349         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6350         if (calldata == NULL)
6351                 goto out;
6352         calldata->clp = clp;
6353         calldata->arg.one_fs = 0;
6354
6355         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6356         msg.rpc_argp = &calldata->arg;
6357         msg.rpc_resp = &calldata->res;
6358         task_setup_data.callback_data = calldata;
6359         task = rpc_run_task(&task_setup_data);
6360         if (IS_ERR(task)) {
6361                 status = PTR_ERR(task);
6362                 goto out;
6363         }
6364         status = nfs4_wait_for_completion_rpc_task(task);
6365         if (status == 0)
6366                 status = task->tk_status;
6367         rpc_put_task(task);
6368         return 0;
6369 out:
6370         dprintk("<-- %s status=%d\n", __func__, status);
6371         return status;
6372 }
6373
6374 static void
6375 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6376 {
6377         struct nfs4_layoutget *lgp = calldata;
6378         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6379
6380         dprintk("--> %s\n", __func__);
6381         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6382          * right now covering the LAYOUTGET we are about to send.
6383          * However, that is not so catastrophic, and there seems
6384          * to be no way to prevent it completely.
6385          */
6386         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
6387                                 &lgp->res.seq_res, task))
6388                 return;
6389         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6390                                           NFS_I(lgp->args.inode)->layout,
6391                                           lgp->args.ctx->state)) {
6392                 rpc_exit(task, NFS4_OK);
6393                 return;
6394         }
6395         rpc_call_start(task);
6396 }
6397
6398 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6399 {
6400         struct nfs4_layoutget *lgp = calldata;
6401         struct inode *inode = lgp->args.inode;
6402         struct nfs_server *server = NFS_SERVER(inode);
6403         struct pnfs_layout_hdr *lo;
6404         struct nfs4_state *state = NULL;
6405
6406         dprintk("--> %s\n", __func__);
6407
6408         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
6409                 goto out;
6410
6411         switch (task->tk_status) {
6412         case 0:
6413                 goto out;
6414         case -NFS4ERR_LAYOUTTRYLATER:
6415         case -NFS4ERR_RECALLCONFLICT:
6416                 task->tk_status = -NFS4ERR_DELAY;
6417                 break;
6418         case -NFS4ERR_EXPIRED:
6419         case -NFS4ERR_BAD_STATEID:
6420                 spin_lock(&inode->i_lock);
6421                 lo = NFS_I(inode)->layout;
6422                 if (!lo || list_empty(&lo->plh_segs)) {
6423                         spin_unlock(&inode->i_lock);
6424                         /* If the open stateid was bad, then recover it. */
6425                         state = lgp->args.ctx->state;
6426                 } else {
6427                         LIST_HEAD(head);
6428
6429                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6430                         spin_unlock(&inode->i_lock);
6431                         /* Mark the bad layout state as invalid, then
6432                          * retry using the open stateid. */
6433                         pnfs_free_lseg_list(&head);
6434                 }
6435         }
6436         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6437                 rpc_restart_call_prepare(task);
6438 out:
6439         dprintk("<-- %s\n", __func__);
6440 }
6441
6442 static size_t max_response_pages(struct nfs_server *server)
6443 {
6444         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6445         return nfs_page_array_len(0, max_resp_sz);
6446 }
6447
6448 static void nfs4_free_pages(struct page **pages, size_t size)
6449 {
6450         int i;
6451
6452         if (!pages)
6453                 return;
6454
6455         for (i = 0; i < size; i++) {
6456                 if (!pages[i])
6457                         break;
6458                 __free_page(pages[i]);
6459         }
6460         kfree(pages);
6461 }
6462
6463 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6464 {
6465         struct page **pages;
6466         int i;
6467
6468         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6469         if (!pages) {
6470                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6471                 return NULL;
6472         }
6473
6474         for (i = 0; i < size; i++) {
6475                 pages[i] = alloc_page(gfp_flags);
6476                 if (!pages[i]) {
6477                         dprintk("%s: failed to allocate page\n", __func__);
6478                         nfs4_free_pages(pages, size);
6479                         return NULL;
6480                 }
6481         }
6482
6483         return pages;
6484 }
6485
6486 static void nfs4_layoutget_release(void *calldata)
6487 {
6488         struct nfs4_layoutget *lgp = calldata;
6489         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6490         size_t max_pages = max_response_pages(server);
6491
6492         dprintk("--> %s\n", __func__);
6493         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6494         put_nfs_open_context(lgp->args.ctx);
6495         kfree(calldata);
6496         dprintk("<-- %s\n", __func__);
6497 }
6498
6499 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6500         .rpc_call_prepare = nfs4_layoutget_prepare,
6501         .rpc_call_done = nfs4_layoutget_done,
6502         .rpc_release = nfs4_layoutget_release,
6503 };
6504
6505 struct pnfs_layout_segment *
6506 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6507 {
6508         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6509         size_t max_pages = max_response_pages(server);
6510         struct rpc_task *task;
6511         struct rpc_message msg = {
6512                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6513                 .rpc_argp = &lgp->args,
6514                 .rpc_resp = &lgp->res,
6515         };
6516         struct rpc_task_setup task_setup_data = {
6517                 .rpc_client = server->client,
6518                 .rpc_message = &msg,
6519                 .callback_ops = &nfs4_layoutget_call_ops,
6520                 .callback_data = lgp,
6521                 .flags = RPC_TASK_ASYNC,
6522         };
6523         struct pnfs_layout_segment *lseg = NULL;
6524         int status = 0;
6525
6526         dprintk("--> %s\n", __func__);
6527
6528         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6529         if (!lgp->args.layout.pages) {
6530                 nfs4_layoutget_release(lgp);
6531                 return ERR_PTR(-ENOMEM);
6532         }
6533         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6534
6535         lgp->res.layoutp = &lgp->args.layout;
6536         lgp->res.seq_res.sr_slot = NULL;
6537         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6538         task = rpc_run_task(&task_setup_data);
6539         if (IS_ERR(task))
6540                 return ERR_CAST(task);
6541         status = nfs4_wait_for_completion_rpc_task(task);
6542         if (status == 0)
6543                 status = task->tk_status;
6544         if (status == 0)
6545                 lseg = pnfs_layout_process(lgp);
6546         rpc_put_task(task);
6547         dprintk("<-- %s status=%d\n", __func__, status);
6548         if (status)
6549                 return ERR_PTR(status);
6550         return lseg;
6551 }
6552
6553 static void
6554 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6555 {
6556         struct nfs4_layoutreturn *lrp = calldata;
6557
6558         dprintk("--> %s\n", __func__);
6559         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
6560                                 &lrp->res.seq_res, task))
6561                 return;
6562         rpc_call_start(task);
6563 }
6564
6565 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6566 {
6567         struct nfs4_layoutreturn *lrp = calldata;
6568         struct nfs_server *server;
6569
6570         dprintk("--> %s\n", __func__);
6571
6572         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
6573                 return;
6574
6575         server = NFS_SERVER(lrp->args.inode);
6576         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6577                 rpc_restart_call_prepare(task);
6578                 return;
6579         }
6580         dprintk("<-- %s\n", __func__);
6581 }
6582
6583 static void nfs4_layoutreturn_release(void *calldata)
6584 {
6585         struct nfs4_layoutreturn *lrp = calldata;
6586         struct pnfs_layout_hdr *lo = lrp->args.layout;
6587
6588         dprintk("--> %s\n", __func__);
6589         spin_lock(&lo->plh_inode->i_lock);
6590         if (lrp->res.lrs_present)
6591                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6592         lo->plh_block_lgets--;
6593         spin_unlock(&lo->plh_inode->i_lock);
6594         pnfs_put_layout_hdr(lrp->args.layout);
6595         kfree(calldata);
6596         dprintk("<-- %s\n", __func__);
6597 }
6598
6599 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6600         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6601         .rpc_call_done = nfs4_layoutreturn_done,
6602         .rpc_release = nfs4_layoutreturn_release,
6603 };
6604
6605 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6606 {
6607         struct rpc_task *task;
6608         struct rpc_message msg = {
6609                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6610                 .rpc_argp = &lrp->args,
6611                 .rpc_resp = &lrp->res,
6612         };
6613         struct rpc_task_setup task_setup_data = {
6614                 .rpc_client = lrp->clp->cl_rpcclient,
6615                 .rpc_message = &msg,
6616                 .callback_ops = &nfs4_layoutreturn_call_ops,
6617                 .callback_data = lrp,
6618         };
6619         int status;
6620
6621         dprintk("--> %s\n", __func__);
6622         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6623         task = rpc_run_task(&task_setup_data);
6624         if (IS_ERR(task))
6625                 return PTR_ERR(task);
6626         status = task->tk_status;
6627         dprintk("<-- %s status=%d\n", __func__, status);
6628         rpc_put_task(task);
6629         return status;
6630 }
6631
6632 /*
6633  * Retrieve the list of Data Server devices from the MDS.
6634  */
6635 static int _nfs4_getdevicelist(struct nfs_server *server,
6636                                     const struct nfs_fh *fh,
6637                                     struct pnfs_devicelist *devlist)
6638 {
6639         struct nfs4_getdevicelist_args args = {
6640                 .fh = fh,
6641                 .layoutclass = server->pnfs_curr_ld->id,
6642         };
6643         struct nfs4_getdevicelist_res res = {
6644                 .devlist = devlist,
6645         };
6646         struct rpc_message msg = {
6647                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6648                 .rpc_argp = &args,
6649                 .rpc_resp = &res,
6650         };
6651         int status;
6652
6653         dprintk("--> %s\n", __func__);
6654         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6655                                 &res.seq_res, 0);
6656         dprintk("<-- %s status=%d\n", __func__, status);
6657         return status;
6658 }
6659
6660 int nfs4_proc_getdevicelist(struct nfs_server *server,
6661                             const struct nfs_fh *fh,
6662                             struct pnfs_devicelist *devlist)
6663 {
6664         struct nfs4_exception exception = { };
6665         int err;
6666
6667         do {
6668                 err = nfs4_handle_exception(server,
6669                                 _nfs4_getdevicelist(server, fh, devlist),
6670                                 &exception);
6671         } while (exception.retry);
6672
6673         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6674                 err, devlist->num_devs);
6675
6676         return err;
6677 }
6678 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6679
6680 static int
6681 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6682 {
6683         struct nfs4_getdeviceinfo_args args = {
6684                 .pdev = pdev,
6685         };
6686         struct nfs4_getdeviceinfo_res res = {
6687                 .pdev = pdev,
6688         };
6689         struct rpc_message msg = {
6690                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6691                 .rpc_argp = &args,
6692                 .rpc_resp = &res,
6693         };
6694         int status;
6695
6696         dprintk("--> %s\n", __func__);
6697         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6698         dprintk("<-- %s status=%d\n", __func__, status);
6699
6700         return status;
6701 }
6702
6703 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6704 {
6705         struct nfs4_exception exception = { };
6706         int err;
6707
6708         do {
6709                 err = nfs4_handle_exception(server,
6710                                         _nfs4_proc_getdeviceinfo(server, pdev),
6711                                         &exception);
6712         } while (exception.retry);
6713         return err;
6714 }
6715 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6716
6717 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6718 {
6719         struct nfs4_layoutcommit_data *data = calldata;
6720         struct nfs_server *server = NFS_SERVER(data->args.inode);
6721
6722         if (nfs4_setup_sequence(server, &data->args.seq_args,
6723                                 &data->res.seq_res, task))
6724                 return;
6725         rpc_call_start(task);
6726 }
6727
6728 static void
6729 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6730 {
6731         struct nfs4_layoutcommit_data *data = calldata;
6732         struct nfs_server *server = NFS_SERVER(data->args.inode);
6733
6734         if (!nfs4_sequence_done(task, &data->res.seq_res))
6735                 return;
6736
6737         switch (task->tk_status) { /* Just ignore these failures */
6738         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6739         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6740         case -NFS4ERR_BADLAYOUT:     /* no layout */
6741         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6742                 task->tk_status = 0;
6743                 break;
6744         case 0:
6745                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6746                                                    data->res.fattr);
6747                 break;
6748         default:
6749                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6750                         rpc_restart_call_prepare(task);
6751                         return;
6752                 }
6753         }
6754 }
6755
6756 static void nfs4_layoutcommit_release(void *calldata)
6757 {
6758         struct nfs4_layoutcommit_data *data = calldata;
6759         struct pnfs_layout_segment *lseg, *tmp;
6760         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6761
6762         pnfs_cleanup_layoutcommit(data);
6763         /* Matched by references in pnfs_set_layoutcommit */
6764         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6765                 list_del_init(&lseg->pls_lc_list);
6766                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6767                                        &lseg->pls_flags))
6768                         pnfs_put_lseg(lseg);
6769         }
6770
6771         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6772         smp_mb__after_clear_bit();
6773         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6774
6775         put_rpccred(data->cred);
6776         kfree(data);
6777 }
6778
6779 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6780         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6781         .rpc_call_done = nfs4_layoutcommit_done,
6782         .rpc_release = nfs4_layoutcommit_release,
6783 };
6784
6785 int
6786 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6787 {
6788         struct rpc_message msg = {
6789                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6790                 .rpc_argp = &data->args,
6791                 .rpc_resp = &data->res,
6792                 .rpc_cred = data->cred,
6793         };
6794         struct rpc_task_setup task_setup_data = {
6795                 .task = &data->task,
6796                 .rpc_client = NFS_CLIENT(data->args.inode),
6797                 .rpc_message = &msg,
6798                 .callback_ops = &nfs4_layoutcommit_ops,
6799                 .callback_data = data,
6800                 .flags = RPC_TASK_ASYNC,
6801         };
6802         struct rpc_task *task;
6803         int status = 0;
6804
6805         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6806                 "lbw: %llu inode %lu\n",
6807                 data->task.tk_pid, sync,
6808                 data->args.lastbytewritten,
6809                 data->args.inode->i_ino);
6810
6811         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6812         task = rpc_run_task(&task_setup_data);
6813         if (IS_ERR(task))
6814                 return PTR_ERR(task);
6815         if (sync == false)
6816                 goto out;
6817         status = nfs4_wait_for_completion_rpc_task(task);
6818         if (status != 0)
6819                 goto out;
6820         status = task->tk_status;
6821 out:
6822         dprintk("%s: status %d\n", __func__, status);
6823         rpc_put_task(task);
6824         return status;
6825 }
6826
6827 static int
6828 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6829                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6830 {
6831         struct nfs41_secinfo_no_name_args args = {
6832                 .style = SECINFO_STYLE_CURRENT_FH,
6833         };
6834         struct nfs4_secinfo_res res = {
6835                 .flavors = flavors,
6836         };
6837         struct rpc_message msg = {
6838                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6839                 .rpc_argp = &args,
6840                 .rpc_resp = &res,
6841         };
6842         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6843 }
6844
6845 static int
6846 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6847                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6848 {
6849         struct nfs4_exception exception = { };
6850         int err;
6851         do {
6852                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6853                 switch (err) {
6854                 case 0:
6855                 case -NFS4ERR_WRONGSEC:
6856                 case -NFS4ERR_NOTSUPP:
6857                         goto out;
6858                 default:
6859                         err = nfs4_handle_exception(server, err, &exception);
6860                 }
6861         } while (exception.retry);
6862 out:
6863         return err;
6864 }
6865
6866 static int
6867 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6868                     struct nfs_fsinfo *info)
6869 {
6870         int err;
6871         struct page *page;
6872         rpc_authflavor_t flavor;
6873         struct nfs4_secinfo_flavors *flavors;
6874
6875         page = alloc_page(GFP_KERNEL);
6876         if (!page) {
6877                 err = -ENOMEM;
6878                 goto out;
6879         }
6880
6881         flavors = page_address(page);
6882         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6883
6884         /*
6885          * Fall back on "guess and check" method if
6886          * the server doesn't support SECINFO_NO_NAME
6887          */
6888         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6889                 err = nfs4_find_root_sec(server, fhandle, info);
6890                 goto out_freepage;
6891         }
6892         if (err)
6893                 goto out_freepage;
6894
6895         flavor = nfs_find_best_sec(flavors);
6896         if (err == 0)
6897                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6898
6899 out_freepage:
6900         put_page(page);
6901         if (err == -EACCES)
6902                 return -EPERM;
6903 out:
6904         return err;
6905 }
6906
6907 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6908 {
6909         int status;
6910         struct nfs41_test_stateid_args args = {
6911                 .stateid = stateid,
6912         };
6913         struct nfs41_test_stateid_res res;
6914         struct rpc_message msg = {
6915                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6916                 .rpc_argp = &args,
6917                 .rpc_resp = &res,
6918         };
6919
6920         dprintk("NFS call  test_stateid %p\n", stateid);
6921         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6922         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6923         if (status != NFS_OK) {
6924                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6925                 return status;
6926         }
6927         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6928         return -res.status;
6929 }
6930
6931 /**
6932  * nfs41_test_stateid - perform a TEST_STATEID operation
6933  *
6934  * @server: server / transport on which to perform the operation
6935  * @stateid: state ID to test
6936  *
6937  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6938  * Otherwise a negative NFS4ERR value is returned if the operation
6939  * failed or the state ID is not currently valid.
6940  */
6941 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6942 {
6943         struct nfs4_exception exception = { };
6944         int err;
6945         do {
6946                 err = _nfs41_test_stateid(server, stateid);
6947                 if (err != -NFS4ERR_DELAY)
6948                         break;
6949                 nfs4_handle_exception(server, err, &exception);
6950         } while (exception.retry);
6951         return err;
6952 }
6953
6954 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6955 {
6956         struct nfs41_free_stateid_args args = {
6957                 .stateid = stateid,
6958         };
6959         struct nfs41_free_stateid_res res;
6960         struct rpc_message msg = {
6961                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6962                 .rpc_argp = &args,
6963                 .rpc_resp = &res,
6964         };
6965         int status;
6966
6967         dprintk("NFS call  free_stateid %p\n", stateid);
6968         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6969         status = nfs4_call_sync_sequence(server->client, server, &msg,
6970                                          &args.seq_args, &res.seq_res, 1);
6971         dprintk("NFS reply free_stateid: %d\n", status);
6972         return status;
6973 }
6974
6975 /**
6976  * nfs41_free_stateid - perform a FREE_STATEID operation
6977  *
6978  * @server: server / transport on which to perform the operation
6979  * @stateid: state ID to release
6980  *
6981  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6982  * negative NFS4ERR value is returned.
6983  */
6984 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6985 {
6986         struct nfs4_exception exception = { };
6987         int err;
6988         do {
6989                 err = _nfs4_free_stateid(server, stateid);
6990                 if (err != -NFS4ERR_DELAY)
6991                         break;
6992                 nfs4_handle_exception(server, err, &exception);
6993         } while (exception.retry);
6994         return err;
6995 }
6996
6997 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6998                 const nfs4_stateid *s2)
6999 {
7000         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
7001                 return false;
7002
7003         if (s1->seqid == s2->seqid)
7004                 return true;
7005         if (s1->seqid == 0 || s2->seqid == 0)
7006                 return true;
7007
7008         return false;
7009 }
7010
7011 #endif /* CONFIG_NFS_V4_1 */
7012
7013 static bool nfs4_match_stateid(const nfs4_stateid *s1,
7014                 const nfs4_stateid *s2)
7015 {
7016         return nfs4_stateid_match(s1, s2);
7017 }
7018
7019
7020 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
7021         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
7022         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
7023         .recover_open   = nfs4_open_reclaim,
7024         .recover_lock   = nfs4_lock_reclaim,
7025         .establish_clid = nfs4_init_clientid,
7026         .get_clid_cred  = nfs4_get_setclientid_cred,
7027         .detect_trunking = nfs40_discover_server_trunking,
7028 };
7029
7030 #if defined(CONFIG_NFS_V4_1)
7031 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
7032         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
7033         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
7034         .recover_open   = nfs4_open_reclaim,
7035         .recover_lock   = nfs4_lock_reclaim,
7036         .establish_clid = nfs41_init_clientid,
7037         .get_clid_cred  = nfs4_get_exchange_id_cred,
7038         .reclaim_complete = nfs41_proc_reclaim_complete,
7039         .detect_trunking = nfs41_discover_server_trunking,
7040 };
7041 #endif /* CONFIG_NFS_V4_1 */
7042
7043 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
7044         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7045         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
7046         .recover_open   = nfs4_open_expired,
7047         .recover_lock   = nfs4_lock_expired,
7048         .establish_clid = nfs4_init_clientid,
7049         .get_clid_cred  = nfs4_get_setclientid_cred,
7050 };
7051
7052 #if defined(CONFIG_NFS_V4_1)
7053 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
7054         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7055         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
7056         .recover_open   = nfs41_open_expired,
7057         .recover_lock   = nfs41_lock_expired,
7058         .establish_clid = nfs41_init_clientid,
7059         .get_clid_cred  = nfs4_get_exchange_id_cred,
7060 };
7061 #endif /* CONFIG_NFS_V4_1 */
7062
7063 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
7064         .sched_state_renewal = nfs4_proc_async_renew,
7065         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
7066         .renew_lease = nfs4_proc_renew,
7067 };
7068
7069 #if defined(CONFIG_NFS_V4_1)
7070 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
7071         .sched_state_renewal = nfs41_proc_async_sequence,
7072         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
7073         .renew_lease = nfs4_proc_sequence,
7074 };
7075 #endif
7076
7077 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
7078         .minor_version = 0,
7079         .call_sync = _nfs4_call_sync,
7080         .match_stateid = nfs4_match_stateid,
7081         .find_root_sec = nfs4_find_root_sec,
7082         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
7083         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
7084         .state_renewal_ops = &nfs40_state_renewal_ops,
7085 };
7086
7087 #if defined(CONFIG_NFS_V4_1)
7088 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
7089         .minor_version = 1,
7090         .call_sync = _nfs4_call_sync_session,
7091         .match_stateid = nfs41_match_stateid,
7092         .find_root_sec = nfs41_find_root_sec,
7093         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
7094         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
7095         .state_renewal_ops = &nfs41_state_renewal_ops,
7096 };
7097 #endif
7098
7099 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
7100         [0] = &nfs_v4_0_minor_ops,
7101 #if defined(CONFIG_NFS_V4_1)
7102         [1] = &nfs_v4_1_minor_ops,
7103 #endif
7104 };
7105
7106 const struct inode_operations nfs4_dir_inode_operations = {
7107         .create         = nfs_create,
7108         .lookup         = nfs_lookup,
7109         .atomic_open    = nfs_atomic_open,
7110         .link           = nfs_link,
7111         .unlink         = nfs_unlink,
7112         .symlink        = nfs_symlink,
7113         .mkdir          = nfs_mkdir,
7114         .rmdir          = nfs_rmdir,
7115         .mknod          = nfs_mknod,
7116         .rename         = nfs_rename,
7117         .permission     = nfs_permission,
7118         .getattr        = nfs_getattr,
7119         .setattr        = nfs_setattr,
7120         .getxattr       = generic_getxattr,
7121         .setxattr       = generic_setxattr,
7122         .listxattr      = generic_listxattr,
7123         .removexattr    = generic_removexattr,
7124 };
7125
7126 static const struct inode_operations nfs4_file_inode_operations = {
7127         .permission     = nfs_permission,
7128         .getattr        = nfs_getattr,
7129         .setattr        = nfs_setattr,
7130         .getxattr       = generic_getxattr,
7131         .setxattr       = generic_setxattr,
7132         .listxattr      = generic_listxattr,
7133         .removexattr    = generic_removexattr,
7134 };
7135
7136 const struct nfs_rpc_ops nfs_v4_clientops = {
7137         .version        = 4,                    /* protocol version */
7138         .dentry_ops     = &nfs4_dentry_operations,
7139         .dir_inode_ops  = &nfs4_dir_inode_operations,
7140         .file_inode_ops = &nfs4_file_inode_operations,
7141         .file_ops       = &nfs4_file_operations,
7142         .getroot        = nfs4_proc_get_root,
7143         .submount       = nfs4_submount,
7144         .try_mount      = nfs4_try_mount,
7145         .getattr        = nfs4_proc_getattr,
7146         .setattr        = nfs4_proc_setattr,
7147         .lookup         = nfs4_proc_lookup,
7148         .access         = nfs4_proc_access,
7149         .readlink       = nfs4_proc_readlink,
7150         .create         = nfs4_proc_create,
7151         .remove         = nfs4_proc_remove,
7152         .unlink_setup   = nfs4_proc_unlink_setup,
7153         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
7154         .unlink_done    = nfs4_proc_unlink_done,
7155         .rename         = nfs4_proc_rename,
7156         .rename_setup   = nfs4_proc_rename_setup,
7157         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
7158         .rename_done    = nfs4_proc_rename_done,
7159         .link           = nfs4_proc_link,
7160         .symlink        = nfs4_proc_symlink,
7161         .mkdir          = nfs4_proc_mkdir,
7162         .rmdir          = nfs4_proc_remove,
7163         .readdir        = nfs4_proc_readdir,
7164         .mknod          = nfs4_proc_mknod,
7165         .statfs         = nfs4_proc_statfs,
7166         .fsinfo         = nfs4_proc_fsinfo,
7167         .pathconf       = nfs4_proc_pathconf,
7168         .set_capabilities = nfs4_server_capabilities,
7169         .decode_dirent  = nfs4_decode_dirent,
7170         .read_setup     = nfs4_proc_read_setup,
7171         .read_pageio_init = pnfs_pageio_init_read,
7172         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
7173         .read_done      = nfs4_read_done,
7174         .write_setup    = nfs4_proc_write_setup,
7175         .write_pageio_init = pnfs_pageio_init_write,
7176         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
7177         .write_done     = nfs4_write_done,
7178         .commit_setup   = nfs4_proc_commit_setup,
7179         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
7180         .commit_done    = nfs4_commit_done,
7181         .lock           = nfs4_proc_lock,
7182         .clear_acl_cache = nfs4_zap_acl_attr,
7183         .close_context  = nfs4_close_context,
7184         .open_context   = nfs4_atomic_open,
7185         .have_delegation = nfs4_have_delegation,
7186         .return_delegation = nfs4_inode_return_delegation,
7187         .alloc_client   = nfs4_alloc_client,
7188         .init_client    = nfs4_init_client,
7189         .free_client    = nfs4_free_client,
7190         .create_server  = nfs4_create_server,
7191         .clone_server   = nfs_clone_server,
7192 };
7193
7194 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7195         .prefix = XATTR_NAME_NFSV4_ACL,
7196         .list   = nfs4_xattr_list_nfs4_acl,
7197         .get    = nfs4_xattr_get_nfs4_acl,
7198         .set    = nfs4_xattr_set_nfs4_acl,
7199 };
7200
7201 const struct xattr_handler *nfs4_xattr_handlers[] = {
7202         &nfs4_xattr_nfs4_acl_handler,
7203         NULL
7204 };
7205
7206 /*
7207  * Local variables:
7208  *  c-basic-offset: 8
7209  * End:
7210  */