99650aaf8937a28a52f9fca9946771f61379a8c7
[linux-2.6.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/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/sunrpc/gss_api.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/nfs_idmap.h>
56 #include <linux/sunrpc/bc_xprt.h>
57 #include <linux/xattr.h>
58 #include <linux/utsname.h>
59 #include <linux/freezer.h>
60
61 #include "nfs4_fs.h"
62 #include "delegation.h"
63 #include "internal.h"
64 #include "iostat.h"
65 #include "callback.h"
66 #include "pnfs.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 static unsigned short max_session_slots = NFS4_DEF_SLOT_TABLE_SIZE;
76
77 struct nfs4_opendata;
78 static int _nfs4_proc_open(struct nfs4_opendata *data);
79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
85                             struct nfs_fattr *fattr, struct iattr *sattr,
86                             struct nfs4_state *state);
87 #ifdef CONFIG_NFS_V4_1
88 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
89 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
90 #endif
91 /* Prevent leaks of NFSv4 errors into userland */
92 static int nfs4_map_errors(int err)
93 {
94         if (err >= -1000)
95                 return err;
96         switch (err) {
97         case -NFS4ERR_RESOURCE:
98                 return -EREMOTEIO;
99         case -NFS4ERR_WRONGSEC:
100                 return -EPERM;
101         case -NFS4ERR_BADOWNER:
102         case -NFS4ERR_BADNAME:
103                 return -EINVAL;
104         default:
105                 dprintk("%s could not handle NFSv4 error %d\n",
106                                 __func__, -err);
107                 break;
108         }
109         return -EIO;
110 }
111
112 /*
113  * This is our standard bitmap for GETATTR requests.
114  */
115 const u32 nfs4_fattr_bitmap[2] = {
116         FATTR4_WORD0_TYPE
117         | FATTR4_WORD0_CHANGE
118         | FATTR4_WORD0_SIZE
119         | FATTR4_WORD0_FSID
120         | FATTR4_WORD0_FILEID,
121         FATTR4_WORD1_MODE
122         | FATTR4_WORD1_NUMLINKS
123         | FATTR4_WORD1_OWNER
124         | FATTR4_WORD1_OWNER_GROUP
125         | FATTR4_WORD1_RAWDEV
126         | FATTR4_WORD1_SPACE_USED
127         | FATTR4_WORD1_TIME_ACCESS
128         | FATTR4_WORD1_TIME_METADATA
129         | FATTR4_WORD1_TIME_MODIFY
130 };
131
132 const u32 nfs4_statfs_bitmap[2] = {
133         FATTR4_WORD0_FILES_AVAIL
134         | FATTR4_WORD0_FILES_FREE
135         | FATTR4_WORD0_FILES_TOTAL,
136         FATTR4_WORD1_SPACE_AVAIL
137         | FATTR4_WORD1_SPACE_FREE
138         | FATTR4_WORD1_SPACE_TOTAL
139 };
140
141 const u32 nfs4_pathconf_bitmap[2] = {
142         FATTR4_WORD0_MAXLINK
143         | FATTR4_WORD0_MAXNAME,
144         0
145 };
146
147 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
148                         | FATTR4_WORD0_MAXREAD
149                         | FATTR4_WORD0_MAXWRITE
150                         | FATTR4_WORD0_LEASE_TIME,
151                         FATTR4_WORD1_TIME_DELTA
152                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
153                         FATTR4_WORD2_LAYOUT_BLKSIZE
154 };
155
156 const u32 nfs4_fs_locations_bitmap[2] = {
157         FATTR4_WORD0_TYPE
158         | FATTR4_WORD0_CHANGE
159         | FATTR4_WORD0_SIZE
160         | FATTR4_WORD0_FSID
161         | FATTR4_WORD0_FILEID
162         | FATTR4_WORD0_FS_LOCATIONS,
163         FATTR4_WORD1_MODE
164         | FATTR4_WORD1_NUMLINKS
165         | FATTR4_WORD1_OWNER
166         | FATTR4_WORD1_OWNER_GROUP
167         | FATTR4_WORD1_RAWDEV
168         | FATTR4_WORD1_SPACE_USED
169         | FATTR4_WORD1_TIME_ACCESS
170         | FATTR4_WORD1_TIME_METADATA
171         | FATTR4_WORD1_TIME_MODIFY
172         | FATTR4_WORD1_MOUNTED_ON_FILEID
173 };
174
175 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
176                 struct nfs4_readdir_arg *readdir)
177 {
178         __be32 *start, *p;
179
180         BUG_ON(readdir->count < 80);
181         if (cookie > 2) {
182                 readdir->cookie = cookie;
183                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
184                 return;
185         }
186
187         readdir->cookie = 0;
188         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
189         if (cookie == 2)
190                 return;
191         
192         /*
193          * NFSv4 servers do not return entries for '.' and '..'
194          * Therefore, we fake these entries here.  We let '.'
195          * have cookie 0 and '..' have cookie 1.  Note that
196          * when talking to the server, we always send cookie 0
197          * instead of 1 or 2.
198          */
199         start = p = kmap_atomic(*readdir->pages);
200         
201         if (cookie == 0) {
202                 *p++ = xdr_one;                                  /* next */
203                 *p++ = xdr_zero;                   /* cookie, first word */
204                 *p++ = xdr_one;                   /* cookie, second word */
205                 *p++ = xdr_one;                             /* entry len */
206                 memcpy(p, ".\0\0\0", 4);                        /* entry */
207                 p++;
208                 *p++ = xdr_one;                         /* bitmap length */
209                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
210                 *p++ = htonl(8);              /* attribute buffer length */
211                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
212         }
213         
214         *p++ = xdr_one;                                  /* next */
215         *p++ = xdr_zero;                   /* cookie, first word */
216         *p++ = xdr_two;                   /* cookie, second word */
217         *p++ = xdr_two;                             /* entry len */
218         memcpy(p, "..\0\0", 4);                         /* entry */
219         p++;
220         *p++ = xdr_one;                         /* bitmap length */
221         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
222         *p++ = htonl(8);              /* attribute buffer length */
223         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
224
225         readdir->pgbase = (char *)p - (char *)start;
226         readdir->count -= readdir->pgbase;
227         kunmap_atomic(start);
228 }
229
230 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
231 {
232         int res;
233
234         might_sleep();
235
236         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
237                         nfs_wait_bit_killable, TASK_KILLABLE);
238         return res;
239 }
240
241 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
242 {
243         int res = 0;
244
245         might_sleep();
246
247         if (*timeout <= 0)
248                 *timeout = NFS4_POLL_RETRY_MIN;
249         if (*timeout > NFS4_POLL_RETRY_MAX)
250                 *timeout = NFS4_POLL_RETRY_MAX;
251         freezable_schedule_timeout_killable(*timeout);
252         if (fatal_signal_pending(current))
253                 res = -ERESTARTSYS;
254         *timeout <<= 1;
255         return res;
256 }
257
258 /* This is the error handling routine for processes that are allowed
259  * to sleep.
260  */
261 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
262 {
263         struct nfs_client *clp = server->nfs_client;
264         struct nfs4_state *state = exception->state;
265         struct inode *inode = exception->inode;
266         int ret = errorcode;
267
268         exception->retry = 0;
269         switch(errorcode) {
270                 case 0:
271                         return 0;
272                 case -NFS4ERR_OPENMODE:
273                         if (inode && nfs_have_delegation(inode, FMODE_READ)) {
274                                 nfs_inode_return_delegation(inode);
275                                 exception->retry = 1;
276                                 return 0;
277                         }
278                         if (state == NULL)
279                                 break;
280                         nfs4_schedule_stateid_recovery(server, state);
281                         goto wait_on_recovery;
282                 case -NFS4ERR_DELEG_REVOKED:
283                 case -NFS4ERR_ADMIN_REVOKED:
284                 case -NFS4ERR_BAD_STATEID:
285                         if (state == NULL)
286                                 break;
287                         nfs_remove_bad_delegation(state->inode);
288                         nfs4_schedule_stateid_recovery(server, state);
289                         goto wait_on_recovery;
290                 case -NFS4ERR_EXPIRED:
291                         if (state != NULL)
292                                 nfs4_schedule_stateid_recovery(server, state);
293                 case -NFS4ERR_STALE_STATEID:
294                 case -NFS4ERR_STALE_CLIENTID:
295                         nfs4_schedule_lease_recovery(clp);
296                         goto wait_on_recovery;
297 #if defined(CONFIG_NFS_V4_1)
298                 case -NFS4ERR_BADSESSION:
299                 case -NFS4ERR_BADSLOT:
300                 case -NFS4ERR_BAD_HIGH_SLOT:
301                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
302                 case -NFS4ERR_DEADSESSION:
303                 case -NFS4ERR_SEQ_FALSE_RETRY:
304                 case -NFS4ERR_SEQ_MISORDERED:
305                         dprintk("%s ERROR: %d Reset session\n", __func__,
306                                 errorcode);
307                         nfs4_schedule_session_recovery(clp->cl_session);
308                         exception->retry = 1;
309                         break;
310 #endif /* defined(CONFIG_NFS_V4_1) */
311                 case -NFS4ERR_FILE_OPEN:
312                         if (exception->timeout > HZ) {
313                                 /* We have retried a decent amount, time to
314                                  * fail
315                                  */
316                                 ret = -EBUSY;
317                                 break;
318                         }
319                 case -NFS4ERR_GRACE:
320                 case -NFS4ERR_DELAY:
321                 case -EKEYEXPIRED:
322                         ret = nfs4_delay(server->client, &exception->timeout);
323                         if (ret != 0)
324                                 break;
325                 case -NFS4ERR_RETRY_UNCACHED_REP:
326                 case -NFS4ERR_OLD_STATEID:
327                         exception->retry = 1;
328                         break;
329                 case -NFS4ERR_BADOWNER:
330                         /* The following works around a Linux server bug! */
331                 case -NFS4ERR_BADNAME:
332                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
333                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
334                                 exception->retry = 1;
335                                 printk(KERN_WARNING "NFS: v4 server %s "
336                                                 "does not accept raw "
337                                                 "uid/gids. "
338                                                 "Reenabling the idmapper.\n",
339                                                 server->nfs_client->cl_hostname);
340                         }
341         }
342         /* We failed to handle the error */
343         return nfs4_map_errors(ret);
344 wait_on_recovery:
345         ret = nfs4_wait_clnt_recover(clp);
346         if (ret == 0)
347                 exception->retry = 1;
348         return ret;
349 }
350
351
352 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
353 {
354         spin_lock(&clp->cl_lock);
355         if (time_before(clp->cl_last_renewal,timestamp))
356                 clp->cl_last_renewal = timestamp;
357         spin_unlock(&clp->cl_lock);
358 }
359
360 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
361 {
362         do_renew_lease(server->nfs_client, timestamp);
363 }
364
365 #if defined(CONFIG_NFS_V4_1)
366
367 /*
368  * nfs4_free_slot - free a slot and efficiently update slot table.
369  *
370  * freeing a slot is trivially done by clearing its respective bit
371  * in the bitmap.
372  * If the freed slotid equals highest_used_slotid we want to update it
373  * so that the server would be able to size down the slot table if needed,
374  * otherwise we know that the highest_used_slotid is still in use.
375  * When updating highest_used_slotid there may be "holes" in the bitmap
376  * so we need to scan down from highest_used_slotid to 0 looking for the now
377  * highest slotid in use.
378  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
379  *
380  * Must be called while holding tbl->slot_tbl_lock
381  */
382 static void
383 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
384 {
385         BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
386         /* clear used bit in bitmap */
387         __clear_bit(slotid, tbl->used_slots);
388
389         /* update highest_used_slotid when it is freed */
390         if (slotid == tbl->highest_used_slotid) {
391                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
392                 if (slotid < tbl->max_slots)
393                         tbl->highest_used_slotid = slotid;
394                 else
395                         tbl->highest_used_slotid = NFS4_NO_SLOT;
396         }
397         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
398                 slotid, tbl->highest_used_slotid);
399 }
400
401 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
402 {
403         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
404         return true;
405 }
406
407 /*
408  * Signal state manager thread if session fore channel is drained
409  */
410 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
411 {
412         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
413                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
414                                 nfs4_set_task_privileged, NULL);
415                 return;
416         }
417
418         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
419                 return;
420
421         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
422         complete(&ses->fc_slot_table.complete);
423 }
424
425 /*
426  * Signal state manager thread if session back channel is drained
427  */
428 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
429 {
430         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
431             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
432                 return;
433         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
434         complete(&ses->bc_slot_table.complete);
435 }
436
437 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
438 {
439         struct nfs4_slot_table *tbl;
440
441         tbl = &res->sr_session->fc_slot_table;
442         if (!res->sr_slot) {
443                 /* just wake up the next guy waiting since
444                  * we may have not consumed a slot after all */
445                 dprintk("%s: No slot\n", __func__);
446                 return;
447         }
448
449         spin_lock(&tbl->slot_tbl_lock);
450         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
451         nfs4_check_drain_fc_complete(res->sr_session);
452         spin_unlock(&tbl->slot_tbl_lock);
453         res->sr_slot = NULL;
454 }
455
456 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
457 {
458         unsigned long timestamp;
459         struct nfs_client *clp;
460
461         /*
462          * sr_status remains 1 if an RPC level error occurred. The server
463          * may or may not have processed the sequence operation..
464          * Proceed as if the server received and processed the sequence
465          * operation.
466          */
467         if (res->sr_status == 1)
468                 res->sr_status = NFS_OK;
469
470         /* don't increment the sequence number if the task wasn't sent */
471         if (!RPC_WAS_SENT(task))
472                 goto out;
473
474         /* Check the SEQUENCE operation status */
475         switch (res->sr_status) {
476         case 0:
477                 /* Update the slot's sequence and clientid lease timer */
478                 ++res->sr_slot->seq_nr;
479                 timestamp = res->sr_renewal_time;
480                 clp = res->sr_session->clp;
481                 do_renew_lease(clp, timestamp);
482                 /* Check sequence flags */
483                 if (res->sr_status_flags != 0)
484                         nfs4_schedule_lease_recovery(clp);
485                 break;
486         case -NFS4ERR_DELAY:
487                 /* The server detected a resend of the RPC call and
488                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
489                  * of RFC5661.
490                  */
491                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
492                         __func__,
493                         res->sr_slot - res->sr_session->fc_slot_table.slots,
494                         res->sr_slot->seq_nr);
495                 goto out_retry;
496         default:
497                 /* Just update the slot sequence no. */
498                 ++res->sr_slot->seq_nr;
499         }
500 out:
501         /* The session may be reset by one of the error handlers. */
502         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
503         nfs41_sequence_free_slot(res);
504         return 1;
505 out_retry:
506         if (!rpc_restart_call(task))
507                 goto out;
508         rpc_delay(task, NFS4_POLL_RETRY_MAX);
509         return 0;
510 }
511
512 static int nfs4_sequence_done(struct rpc_task *task,
513                                struct nfs4_sequence_res *res)
514 {
515         if (res->sr_session == NULL)
516                 return 1;
517         return nfs41_sequence_done(task, res);
518 }
519
520 /*
521  * nfs4_find_slot - efficiently look for a free slot
522  *
523  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
524  * If found, we mark the slot as used, update the highest_used_slotid,
525  * and respectively set up the sequence operation args.
526  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
527  *
528  * Note: must be called with under the slot_tbl_lock.
529  */
530 static u32
531 nfs4_find_slot(struct nfs4_slot_table *tbl)
532 {
533         u32 slotid;
534         u32 ret_id = NFS4_NO_SLOT;
535
536         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
537                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
538                 tbl->max_slots);
539         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
540         if (slotid >= tbl->max_slots)
541                 goto out;
542         __set_bit(slotid, tbl->used_slots);
543         if (slotid > tbl->highest_used_slotid ||
544                         tbl->highest_used_slotid == NFS4_NO_SLOT)
545                 tbl->highest_used_slotid = slotid;
546         ret_id = slotid;
547 out:
548         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
549                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
550         return ret_id;
551 }
552
553 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
554                 struct nfs4_sequence_res *res, int cache_reply)
555 {
556         args->sa_session = NULL;
557         args->sa_cache_this = 0;
558         if (cache_reply)
559                 args->sa_cache_this = 1;
560         res->sr_session = NULL;
561         res->sr_slot = NULL;
562 }
563
564 int nfs41_setup_sequence(struct nfs4_session *session,
565                                 struct nfs4_sequence_args *args,
566                                 struct nfs4_sequence_res *res,
567                                 struct rpc_task *task)
568 {
569         struct nfs4_slot *slot;
570         struct nfs4_slot_table *tbl;
571         u32 slotid;
572
573         dprintk("--> %s\n", __func__);
574         /* slot already allocated? */
575         if (res->sr_slot != NULL)
576                 return 0;
577
578         tbl = &session->fc_slot_table;
579
580         spin_lock(&tbl->slot_tbl_lock);
581         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
582             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
583                 /* The state manager will wait until the slot table is empty */
584                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
585                 spin_unlock(&tbl->slot_tbl_lock);
586                 dprintk("%s session is draining\n", __func__);
587                 return -EAGAIN;
588         }
589
590         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
591             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
592                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
593                 spin_unlock(&tbl->slot_tbl_lock);
594                 dprintk("%s enforce FIFO order\n", __func__);
595                 return -EAGAIN;
596         }
597
598         slotid = nfs4_find_slot(tbl);
599         if (slotid == NFS4_NO_SLOT) {
600                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
601                 spin_unlock(&tbl->slot_tbl_lock);
602                 dprintk("<-- %s: no free slots\n", __func__);
603                 return -EAGAIN;
604         }
605         spin_unlock(&tbl->slot_tbl_lock);
606
607         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
608         slot = tbl->slots + slotid;
609         args->sa_session = session;
610         args->sa_slotid = slotid;
611
612         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
613
614         res->sr_session = session;
615         res->sr_slot = slot;
616         res->sr_renewal_time = jiffies;
617         res->sr_status_flags = 0;
618         /*
619          * sr_status is only set in decode_sequence, and so will remain
620          * set to 1 if an rpc level failure occurs.
621          */
622         res->sr_status = 1;
623         return 0;
624 }
625 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
626
627 int nfs4_setup_sequence(const struct nfs_server *server,
628                         struct nfs4_sequence_args *args,
629                         struct nfs4_sequence_res *res,
630                         struct rpc_task *task)
631 {
632         struct nfs4_session *session = nfs4_get_session(server);
633         int ret = 0;
634
635         if (session == NULL)
636                 goto out;
637
638         dprintk("--> %s clp %p session %p sr_slot %td\n",
639                 __func__, session->clp, session, res->sr_slot ?
640                         res->sr_slot - session->fc_slot_table.slots : -1);
641
642         ret = nfs41_setup_sequence(session, args, res, task);
643 out:
644         dprintk("<-- %s status=%d\n", __func__, ret);
645         return ret;
646 }
647
648 struct nfs41_call_sync_data {
649         const struct nfs_server *seq_server;
650         struct nfs4_sequence_args *seq_args;
651         struct nfs4_sequence_res *seq_res;
652 };
653
654 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
655 {
656         struct nfs41_call_sync_data *data = calldata;
657
658         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
659
660         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
661                                 data->seq_res, task))
662                 return;
663         rpc_call_start(task);
664 }
665
666 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
667 {
668         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
669         nfs41_call_sync_prepare(task, calldata);
670 }
671
672 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
673 {
674         struct nfs41_call_sync_data *data = calldata;
675
676         nfs41_sequence_done(task, data->seq_res);
677 }
678
679 static const struct rpc_call_ops nfs41_call_sync_ops = {
680         .rpc_call_prepare = nfs41_call_sync_prepare,
681         .rpc_call_done = nfs41_call_sync_done,
682 };
683
684 static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
685         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
686         .rpc_call_done = nfs41_call_sync_done,
687 };
688
689 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
690                                    struct nfs_server *server,
691                                    struct rpc_message *msg,
692                                    struct nfs4_sequence_args *args,
693                                    struct nfs4_sequence_res *res,
694                                    int privileged)
695 {
696         int ret;
697         struct rpc_task *task;
698         struct nfs41_call_sync_data data = {
699                 .seq_server = server,
700                 .seq_args = args,
701                 .seq_res = res,
702         };
703         struct rpc_task_setup task_setup = {
704                 .rpc_client = clnt,
705                 .rpc_message = msg,
706                 .callback_ops = &nfs41_call_sync_ops,
707                 .callback_data = &data
708         };
709
710         if (privileged)
711                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
712         task = rpc_run_task(&task_setup);
713         if (IS_ERR(task))
714                 ret = PTR_ERR(task);
715         else {
716                 ret = task->tk_status;
717                 rpc_put_task(task);
718         }
719         return ret;
720 }
721
722 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
723                             struct nfs_server *server,
724                             struct rpc_message *msg,
725                             struct nfs4_sequence_args *args,
726                             struct nfs4_sequence_res *res,
727                             int cache_reply)
728 {
729         nfs41_init_sequence(args, res, cache_reply);
730         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
731 }
732
733 #else
734 static inline
735 void nfs41_init_sequence(struct nfs4_sequence_args *args,
736                 struct nfs4_sequence_res *res, int cache_reply)
737 {
738 }
739
740 static int nfs4_sequence_done(struct rpc_task *task,
741                                struct nfs4_sequence_res *res)
742 {
743         return 1;
744 }
745 #endif /* CONFIG_NFS_V4_1 */
746
747 int _nfs4_call_sync(struct rpc_clnt *clnt,
748                     struct nfs_server *server,
749                     struct rpc_message *msg,
750                     struct nfs4_sequence_args *args,
751                     struct nfs4_sequence_res *res,
752                     int cache_reply)
753 {
754         nfs41_init_sequence(args, res, cache_reply);
755         return rpc_call_sync(clnt, msg, 0);
756 }
757
758 static inline
759 int nfs4_call_sync(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         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
767                                                 args, res, cache_reply);
768 }
769
770 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
771 {
772         struct nfs_inode *nfsi = NFS_I(dir);
773
774         spin_lock(&dir->i_lock);
775         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
776         if (!cinfo->atomic || cinfo->before != dir->i_version)
777                 nfs_force_lookup_revalidate(dir);
778         dir->i_version = cinfo->after;
779         spin_unlock(&dir->i_lock);
780 }
781
782 struct nfs4_opendata {
783         struct kref kref;
784         struct nfs_openargs o_arg;
785         struct nfs_openres o_res;
786         struct nfs_open_confirmargs c_arg;
787         struct nfs_open_confirmres c_res;
788         struct nfs4_string owner_name;
789         struct nfs4_string group_name;
790         struct nfs_fattr f_attr;
791         struct nfs_fattr dir_attr;
792         struct dentry *dir;
793         struct dentry *dentry;
794         struct nfs4_state_owner *owner;
795         struct nfs4_state *state;
796         struct iattr attrs;
797         unsigned long timestamp;
798         unsigned int rpc_done : 1;
799         int rpc_status;
800         int cancelled;
801 };
802
803
804 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
805 {
806         p->o_res.f_attr = &p->f_attr;
807         p->o_res.dir_attr = &p->dir_attr;
808         p->o_res.seqid = p->o_arg.seqid;
809         p->c_res.seqid = p->c_arg.seqid;
810         p->o_res.server = p->o_arg.server;
811         nfs_fattr_init(&p->f_attr);
812         nfs_fattr_init(&p->dir_attr);
813         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
814 }
815
816 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
817                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
818                 const struct iattr *attrs,
819                 gfp_t gfp_mask)
820 {
821         struct dentry *parent = dget_parent(dentry);
822         struct inode *dir = parent->d_inode;
823         struct nfs_server *server = NFS_SERVER(dir);
824         struct nfs4_opendata *p;
825
826         p = kzalloc(sizeof(*p), gfp_mask);
827         if (p == NULL)
828                 goto err;
829         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
830         if (p->o_arg.seqid == NULL)
831                 goto err_free;
832         nfs_sb_active(dentry->d_sb);
833         p->dentry = dget(dentry);
834         p->dir = parent;
835         p->owner = sp;
836         atomic_inc(&sp->so_count);
837         p->o_arg.fh = NFS_FH(dir);
838         p->o_arg.open_flags = flags;
839         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
840         p->o_arg.clientid = server->nfs_client->cl_clientid;
841         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
842         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
843         p->o_arg.name = &dentry->d_name;
844         p->o_arg.server = server;
845         p->o_arg.bitmask = server->attr_bitmask;
846         p->o_arg.dir_bitmask = server->cache_consistency_bitmask;
847         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
848         if (attrs != NULL && attrs->ia_valid != 0) {
849                 __be32 verf[2];
850
851                 p->o_arg.u.attrs = &p->attrs;
852                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
853
854                 verf[0] = jiffies;
855                 verf[1] = current->pid;
856                 memcpy(p->o_arg.u.verifier.data, verf,
857                                 sizeof(p->o_arg.u.verifier.data));
858         }
859         p->c_arg.fh = &p->o_res.fh;
860         p->c_arg.stateid = &p->o_res.stateid;
861         p->c_arg.seqid = p->o_arg.seqid;
862         nfs4_init_opendata_res(p);
863         kref_init(&p->kref);
864         return p;
865 err_free:
866         kfree(p);
867 err:
868         dput(parent);
869         return NULL;
870 }
871
872 static void nfs4_opendata_free(struct kref *kref)
873 {
874         struct nfs4_opendata *p = container_of(kref,
875                         struct nfs4_opendata, kref);
876         struct super_block *sb = p->dentry->d_sb;
877
878         nfs_free_seqid(p->o_arg.seqid);
879         if (p->state != NULL)
880                 nfs4_put_open_state(p->state);
881         nfs4_put_state_owner(p->owner);
882         dput(p->dir);
883         dput(p->dentry);
884         nfs_sb_deactive(sb);
885         nfs_fattr_free_names(&p->f_attr);
886         kfree(p);
887 }
888
889 static void nfs4_opendata_put(struct nfs4_opendata *p)
890 {
891         if (p != NULL)
892                 kref_put(&p->kref, nfs4_opendata_free);
893 }
894
895 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
896 {
897         int ret;
898
899         ret = rpc_wait_for_completion_task(task);
900         return ret;
901 }
902
903 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
904 {
905         int ret = 0;
906
907         if (open_mode & (O_EXCL|O_TRUNC))
908                 goto out;
909         switch (mode & (FMODE_READ|FMODE_WRITE)) {
910                 case FMODE_READ:
911                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
912                                 && state->n_rdonly != 0;
913                         break;
914                 case FMODE_WRITE:
915                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
916                                 && state->n_wronly != 0;
917                         break;
918                 case FMODE_READ|FMODE_WRITE:
919                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
920                                 && state->n_rdwr != 0;
921         }
922 out:
923         return ret;
924 }
925
926 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
927 {
928         if (delegation == NULL)
929                 return 0;
930         if ((delegation->type & fmode) != fmode)
931                 return 0;
932         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
933                 return 0;
934         nfs_mark_delegation_referenced(delegation);
935         return 1;
936 }
937
938 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
939 {
940         switch (fmode) {
941                 case FMODE_WRITE:
942                         state->n_wronly++;
943                         break;
944                 case FMODE_READ:
945                         state->n_rdonly++;
946                         break;
947                 case FMODE_READ|FMODE_WRITE:
948                         state->n_rdwr++;
949         }
950         nfs4_state_set_mode_locked(state, state->state | fmode);
951 }
952
953 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
954 {
955         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
956                 nfs4_stateid_copy(&state->stateid, stateid);
957         nfs4_stateid_copy(&state->open_stateid, stateid);
958         switch (fmode) {
959                 case FMODE_READ:
960                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
961                         break;
962                 case FMODE_WRITE:
963                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
964                         break;
965                 case FMODE_READ|FMODE_WRITE:
966                         set_bit(NFS_O_RDWR_STATE, &state->flags);
967         }
968 }
969
970 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
971 {
972         write_seqlock(&state->seqlock);
973         nfs_set_open_stateid_locked(state, stateid, fmode);
974         write_sequnlock(&state->seqlock);
975 }
976
977 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
978 {
979         /*
980          * Protect the call to nfs4_state_set_mode_locked and
981          * serialise the stateid update
982          */
983         write_seqlock(&state->seqlock);
984         if (deleg_stateid != NULL) {
985                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
986                 set_bit(NFS_DELEGATED_STATE, &state->flags);
987         }
988         if (open_stateid != NULL)
989                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
990         write_sequnlock(&state->seqlock);
991         spin_lock(&state->owner->so_lock);
992         update_open_stateflags(state, fmode);
993         spin_unlock(&state->owner->so_lock);
994 }
995
996 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
997 {
998         struct nfs_inode *nfsi = NFS_I(state->inode);
999         struct nfs_delegation *deleg_cur;
1000         int ret = 0;
1001
1002         fmode &= (FMODE_READ|FMODE_WRITE);
1003
1004         rcu_read_lock();
1005         deleg_cur = rcu_dereference(nfsi->delegation);
1006         if (deleg_cur == NULL)
1007                 goto no_delegation;
1008
1009         spin_lock(&deleg_cur->lock);
1010         if (nfsi->delegation != deleg_cur ||
1011             (deleg_cur->type & fmode) != fmode)
1012                 goto no_delegation_unlock;
1013
1014         if (delegation == NULL)
1015                 delegation = &deleg_cur->stateid;
1016         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1017                 goto no_delegation_unlock;
1018
1019         nfs_mark_delegation_referenced(deleg_cur);
1020         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1021         ret = 1;
1022 no_delegation_unlock:
1023         spin_unlock(&deleg_cur->lock);
1024 no_delegation:
1025         rcu_read_unlock();
1026
1027         if (!ret && open_stateid != NULL) {
1028                 __update_open_stateid(state, open_stateid, NULL, fmode);
1029                 ret = 1;
1030         }
1031
1032         return ret;
1033 }
1034
1035
1036 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1037 {
1038         struct nfs_delegation *delegation;
1039
1040         rcu_read_lock();
1041         delegation = rcu_dereference(NFS_I(inode)->delegation);
1042         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1043                 rcu_read_unlock();
1044                 return;
1045         }
1046         rcu_read_unlock();
1047         nfs_inode_return_delegation(inode);
1048 }
1049
1050 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1051 {
1052         struct nfs4_state *state = opendata->state;
1053         struct nfs_inode *nfsi = NFS_I(state->inode);
1054         struct nfs_delegation *delegation;
1055         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1056         fmode_t fmode = opendata->o_arg.fmode;
1057         nfs4_stateid stateid;
1058         int ret = -EAGAIN;
1059
1060         for (;;) {
1061                 if (can_open_cached(state, fmode, open_mode)) {
1062                         spin_lock(&state->owner->so_lock);
1063                         if (can_open_cached(state, fmode, open_mode)) {
1064                                 update_open_stateflags(state, fmode);
1065                                 spin_unlock(&state->owner->so_lock);
1066                                 goto out_return_state;
1067                         }
1068                         spin_unlock(&state->owner->so_lock);
1069                 }
1070                 rcu_read_lock();
1071                 delegation = rcu_dereference(nfsi->delegation);
1072                 if (!can_open_delegated(delegation, fmode)) {
1073                         rcu_read_unlock();
1074                         break;
1075                 }
1076                 /* Save the delegation */
1077                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1078                 rcu_read_unlock();
1079                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1080                 if (ret != 0)
1081                         goto out;
1082                 ret = -EAGAIN;
1083
1084                 /* Try to update the stateid using the delegation */
1085                 if (update_open_stateid(state, NULL, &stateid, fmode))
1086                         goto out_return_state;
1087         }
1088 out:
1089         return ERR_PTR(ret);
1090 out_return_state:
1091         atomic_inc(&state->count);
1092         return state;
1093 }
1094
1095 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1096 {
1097         struct inode *inode;
1098         struct nfs4_state *state = NULL;
1099         struct nfs_delegation *delegation;
1100         int ret;
1101
1102         if (!data->rpc_done) {
1103                 state = nfs4_try_open_cached(data);
1104                 goto out;
1105         }
1106
1107         ret = -EAGAIN;
1108         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1109                 goto err;
1110         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1111         ret = PTR_ERR(inode);
1112         if (IS_ERR(inode))
1113                 goto err;
1114         ret = -ENOMEM;
1115         state = nfs4_get_open_state(inode, data->owner);
1116         if (state == NULL)
1117                 goto err_put_inode;
1118         if (data->o_res.delegation_type != 0) {
1119                 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1120                 int delegation_flags = 0;
1121
1122                 rcu_read_lock();
1123                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1124                 if (delegation)
1125                         delegation_flags = delegation->flags;
1126                 rcu_read_unlock();
1127                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1128                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1129                                         "returning a delegation for "
1130                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1131                                         clp->cl_hostname);
1132                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1133                         nfs_inode_set_delegation(state->inode,
1134                                         data->owner->so_cred,
1135                                         &data->o_res);
1136                 else
1137                         nfs_inode_reclaim_delegation(state->inode,
1138                                         data->owner->so_cred,
1139                                         &data->o_res);
1140         }
1141
1142         update_open_stateid(state, &data->o_res.stateid, NULL,
1143                         data->o_arg.fmode);
1144         iput(inode);
1145 out:
1146         return state;
1147 err_put_inode:
1148         iput(inode);
1149 err:
1150         return ERR_PTR(ret);
1151 }
1152
1153 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1154 {
1155         struct nfs_inode *nfsi = NFS_I(state->inode);
1156         struct nfs_open_context *ctx;
1157
1158         spin_lock(&state->inode->i_lock);
1159         list_for_each_entry(ctx, &nfsi->open_files, list) {
1160                 if (ctx->state != state)
1161                         continue;
1162                 get_nfs_open_context(ctx);
1163                 spin_unlock(&state->inode->i_lock);
1164                 return ctx;
1165         }
1166         spin_unlock(&state->inode->i_lock);
1167         return ERR_PTR(-ENOENT);
1168 }
1169
1170 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1171 {
1172         struct nfs4_opendata *opendata;
1173
1174         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1175         if (opendata == NULL)
1176                 return ERR_PTR(-ENOMEM);
1177         opendata->state = state;
1178         atomic_inc(&state->count);
1179         return opendata;
1180 }
1181
1182 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1183 {
1184         struct nfs4_state *newstate;
1185         int ret;
1186
1187         opendata->o_arg.open_flags = 0;
1188         opendata->o_arg.fmode = fmode;
1189         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1190         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1191         nfs4_init_opendata_res(opendata);
1192         ret = _nfs4_recover_proc_open(opendata);
1193         if (ret != 0)
1194                 return ret; 
1195         newstate = nfs4_opendata_to_nfs4_state(opendata);
1196         if (IS_ERR(newstate))
1197                 return PTR_ERR(newstate);
1198         nfs4_close_state(newstate, fmode);
1199         *res = newstate;
1200         return 0;
1201 }
1202
1203 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1204 {
1205         struct nfs4_state *newstate;
1206         int ret;
1207
1208         /* memory barrier prior to reading state->n_* */
1209         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1210         smp_rmb();
1211         if (state->n_rdwr != 0) {
1212                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1213                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1214                 if (ret != 0)
1215                         return ret;
1216                 if (newstate != state)
1217                         return -ESTALE;
1218         }
1219         if (state->n_wronly != 0) {
1220                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1221                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1222                 if (ret != 0)
1223                         return ret;
1224                 if (newstate != state)
1225                         return -ESTALE;
1226         }
1227         if (state->n_rdonly != 0) {
1228                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1229                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1230                 if (ret != 0)
1231                         return ret;
1232                 if (newstate != state)
1233                         return -ESTALE;
1234         }
1235         /*
1236          * We may have performed cached opens for all three recoveries.
1237          * Check if we need to update the current stateid.
1238          */
1239         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1240             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1241                 write_seqlock(&state->seqlock);
1242                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1243                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1244                 write_sequnlock(&state->seqlock);
1245         }
1246         return 0;
1247 }
1248
1249 /*
1250  * OPEN_RECLAIM:
1251  *      reclaim state on the server after a reboot.
1252  */
1253 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1254 {
1255         struct nfs_delegation *delegation;
1256         struct nfs4_opendata *opendata;
1257         fmode_t delegation_type = 0;
1258         int status;
1259
1260         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1261         if (IS_ERR(opendata))
1262                 return PTR_ERR(opendata);
1263         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1264         opendata->o_arg.fh = NFS_FH(state->inode);
1265         rcu_read_lock();
1266         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1267         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1268                 delegation_type = delegation->type;
1269         rcu_read_unlock();
1270         opendata->o_arg.u.delegation_type = delegation_type;
1271         status = nfs4_open_recover(opendata, state);
1272         nfs4_opendata_put(opendata);
1273         return status;
1274 }
1275
1276 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1277 {
1278         struct nfs_server *server = NFS_SERVER(state->inode);
1279         struct nfs4_exception exception = { };
1280         int err;
1281         do {
1282                 err = _nfs4_do_open_reclaim(ctx, state);
1283                 if (err != -NFS4ERR_DELAY)
1284                         break;
1285                 nfs4_handle_exception(server, err, &exception);
1286         } while (exception.retry);
1287         return err;
1288 }
1289
1290 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1291 {
1292         struct nfs_open_context *ctx;
1293         int ret;
1294
1295         ctx = nfs4_state_find_open_context(state);
1296         if (IS_ERR(ctx))
1297                 return PTR_ERR(ctx);
1298         ret = nfs4_do_open_reclaim(ctx, state);
1299         put_nfs_open_context(ctx);
1300         return ret;
1301 }
1302
1303 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1304 {
1305         struct nfs4_opendata *opendata;
1306         int ret;
1307
1308         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1309         if (IS_ERR(opendata))
1310                 return PTR_ERR(opendata);
1311         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1312         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1313         ret = nfs4_open_recover(opendata, state);
1314         nfs4_opendata_put(opendata);
1315         return ret;
1316 }
1317
1318 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1319 {
1320         struct nfs4_exception exception = { };
1321         struct nfs_server *server = NFS_SERVER(state->inode);
1322         int err;
1323         do {
1324                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1325                 switch (err) {
1326                         case 0:
1327                         case -ENOENT:
1328                         case -ESTALE:
1329                                 goto out;
1330                         case -NFS4ERR_BADSESSION:
1331                         case -NFS4ERR_BADSLOT:
1332                         case -NFS4ERR_BAD_HIGH_SLOT:
1333                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1334                         case -NFS4ERR_DEADSESSION:
1335                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1336                                 goto out;
1337                         case -NFS4ERR_STALE_CLIENTID:
1338                         case -NFS4ERR_STALE_STATEID:
1339                         case -NFS4ERR_EXPIRED:
1340                                 /* Don't recall a delegation if it was lost */
1341                                 nfs4_schedule_lease_recovery(server->nfs_client);
1342                                 goto out;
1343                         case -ERESTARTSYS:
1344                                 /*
1345                                  * The show must go on: exit, but mark the
1346                                  * stateid as needing recovery.
1347                                  */
1348                         case -NFS4ERR_DELEG_REVOKED:
1349                         case -NFS4ERR_ADMIN_REVOKED:
1350                         case -NFS4ERR_BAD_STATEID:
1351                                 nfs_inode_find_state_and_recover(state->inode,
1352                                                 stateid);
1353                                 nfs4_schedule_stateid_recovery(server, state);
1354                         case -EKEYEXPIRED:
1355                                 /*
1356                                  * User RPCSEC_GSS context has expired.
1357                                  * We cannot recover this stateid now, so
1358                                  * skip it and allow recovery thread to
1359                                  * proceed.
1360                                  */
1361                         case -ENOMEM:
1362                                 err = 0;
1363                                 goto out;
1364                 }
1365                 err = nfs4_handle_exception(server, err, &exception);
1366         } while (exception.retry);
1367 out:
1368         return err;
1369 }
1370
1371 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1372 {
1373         struct nfs4_opendata *data = calldata;
1374
1375         data->rpc_status = task->tk_status;
1376         if (data->rpc_status == 0) {
1377                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1378                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1379                 renew_lease(data->o_res.server, data->timestamp);
1380                 data->rpc_done = 1;
1381         }
1382 }
1383
1384 static void nfs4_open_confirm_release(void *calldata)
1385 {
1386         struct nfs4_opendata *data = calldata;
1387         struct nfs4_state *state = NULL;
1388
1389         /* If this request hasn't been cancelled, do nothing */
1390         if (data->cancelled == 0)
1391                 goto out_free;
1392         /* In case of error, no cleanup! */
1393         if (!data->rpc_done)
1394                 goto out_free;
1395         state = nfs4_opendata_to_nfs4_state(data);
1396         if (!IS_ERR(state))
1397                 nfs4_close_state(state, data->o_arg.fmode);
1398 out_free:
1399         nfs4_opendata_put(data);
1400 }
1401
1402 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1403         .rpc_call_done = nfs4_open_confirm_done,
1404         .rpc_release = nfs4_open_confirm_release,
1405 };
1406
1407 /*
1408  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1409  */
1410 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1411 {
1412         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1413         struct rpc_task *task;
1414         struct  rpc_message msg = {
1415                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1416                 .rpc_argp = &data->c_arg,
1417                 .rpc_resp = &data->c_res,
1418                 .rpc_cred = data->owner->so_cred,
1419         };
1420         struct rpc_task_setup task_setup_data = {
1421                 .rpc_client = server->client,
1422                 .rpc_message = &msg,
1423                 .callback_ops = &nfs4_open_confirm_ops,
1424                 .callback_data = data,
1425                 .workqueue = nfsiod_workqueue,
1426                 .flags = RPC_TASK_ASYNC,
1427         };
1428         int status;
1429
1430         kref_get(&data->kref);
1431         data->rpc_done = 0;
1432         data->rpc_status = 0;
1433         data->timestamp = jiffies;
1434         task = rpc_run_task(&task_setup_data);
1435         if (IS_ERR(task))
1436                 return PTR_ERR(task);
1437         status = nfs4_wait_for_completion_rpc_task(task);
1438         if (status != 0) {
1439                 data->cancelled = 1;
1440                 smp_wmb();
1441         } else
1442                 status = data->rpc_status;
1443         rpc_put_task(task);
1444         return status;
1445 }
1446
1447 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1448 {
1449         struct nfs4_opendata *data = calldata;
1450         struct nfs4_state_owner *sp = data->owner;
1451
1452         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1453                 return;
1454         /*
1455          * Check if we still need to send an OPEN call, or if we can use
1456          * a delegation instead.
1457          */
1458         if (data->state != NULL) {
1459                 struct nfs_delegation *delegation;
1460
1461                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1462                         goto out_no_action;
1463                 rcu_read_lock();
1464                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1465                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1466                     can_open_delegated(delegation, data->o_arg.fmode))
1467                         goto unlock_no_action;
1468                 rcu_read_unlock();
1469         }
1470         /* Update client id. */
1471         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1472         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1473                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1474                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1475         }
1476         data->timestamp = jiffies;
1477         if (nfs4_setup_sequence(data->o_arg.server,
1478                                 &data->o_arg.seq_args,
1479                                 &data->o_res.seq_res, task))
1480                 return;
1481         rpc_call_start(task);
1482         return;
1483 unlock_no_action:
1484         rcu_read_unlock();
1485 out_no_action:
1486         task->tk_action = NULL;
1487
1488 }
1489
1490 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1491 {
1492         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1493         nfs4_open_prepare(task, calldata);
1494 }
1495
1496 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1497 {
1498         struct nfs4_opendata *data = calldata;
1499
1500         data->rpc_status = task->tk_status;
1501
1502         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1503                 return;
1504
1505         if (task->tk_status == 0) {
1506                 switch (data->o_res.f_attr->mode & S_IFMT) {
1507                         case S_IFREG:
1508                                 break;
1509                         case S_IFLNK:
1510                                 data->rpc_status = -ELOOP;
1511                                 break;
1512                         case S_IFDIR:
1513                                 data->rpc_status = -EISDIR;
1514                                 break;
1515                         default:
1516                                 data->rpc_status = -ENOTDIR;
1517                 }
1518                 renew_lease(data->o_res.server, data->timestamp);
1519                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1520                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1521         }
1522         data->rpc_done = 1;
1523 }
1524
1525 static void nfs4_open_release(void *calldata)
1526 {
1527         struct nfs4_opendata *data = calldata;
1528         struct nfs4_state *state = NULL;
1529
1530         /* If this request hasn't been cancelled, do nothing */
1531         if (data->cancelled == 0)
1532                 goto out_free;
1533         /* In case of error, no cleanup! */
1534         if (data->rpc_status != 0 || !data->rpc_done)
1535                 goto out_free;
1536         /* In case we need an open_confirm, no cleanup! */
1537         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1538                 goto out_free;
1539         state = nfs4_opendata_to_nfs4_state(data);
1540         if (!IS_ERR(state))
1541                 nfs4_close_state(state, data->o_arg.fmode);
1542 out_free:
1543         nfs4_opendata_put(data);
1544 }
1545
1546 static const struct rpc_call_ops nfs4_open_ops = {
1547         .rpc_call_prepare = nfs4_open_prepare,
1548         .rpc_call_done = nfs4_open_done,
1549         .rpc_release = nfs4_open_release,
1550 };
1551
1552 static const struct rpc_call_ops nfs4_recover_open_ops = {
1553         .rpc_call_prepare = nfs4_recover_open_prepare,
1554         .rpc_call_done = nfs4_open_done,
1555         .rpc_release = nfs4_open_release,
1556 };
1557
1558 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1559 {
1560         struct inode *dir = data->dir->d_inode;
1561         struct nfs_server *server = NFS_SERVER(dir);
1562         struct nfs_openargs *o_arg = &data->o_arg;
1563         struct nfs_openres *o_res = &data->o_res;
1564         struct rpc_task *task;
1565         struct rpc_message msg = {
1566                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1567                 .rpc_argp = o_arg,
1568                 .rpc_resp = o_res,
1569                 .rpc_cred = data->owner->so_cred,
1570         };
1571         struct rpc_task_setup task_setup_data = {
1572                 .rpc_client = server->client,
1573                 .rpc_message = &msg,
1574                 .callback_ops = &nfs4_open_ops,
1575                 .callback_data = data,
1576                 .workqueue = nfsiod_workqueue,
1577                 .flags = RPC_TASK_ASYNC,
1578         };
1579         int status;
1580
1581         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1582         kref_get(&data->kref);
1583         data->rpc_done = 0;
1584         data->rpc_status = 0;
1585         data->cancelled = 0;
1586         if (isrecover)
1587                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1588         task = rpc_run_task(&task_setup_data);
1589         if (IS_ERR(task))
1590                 return PTR_ERR(task);
1591         status = nfs4_wait_for_completion_rpc_task(task);
1592         if (status != 0) {
1593                 data->cancelled = 1;
1594                 smp_wmb();
1595         } else
1596                 status = data->rpc_status;
1597         rpc_put_task(task);
1598
1599         return status;
1600 }
1601
1602 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1603 {
1604         struct inode *dir = data->dir->d_inode;
1605         struct nfs_openres *o_res = &data->o_res;
1606         int status;
1607
1608         status = nfs4_run_open_task(data, 1);
1609         if (status != 0 || !data->rpc_done)
1610                 return status;
1611
1612         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1613
1614         nfs_refresh_inode(dir, o_res->dir_attr);
1615
1616         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1617                 status = _nfs4_proc_open_confirm(data);
1618                 if (status != 0)
1619                         return status;
1620         }
1621
1622         return status;
1623 }
1624
1625 /*
1626  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1627  */
1628 static int _nfs4_proc_open(struct nfs4_opendata *data)
1629 {
1630         struct inode *dir = data->dir->d_inode;
1631         struct nfs_server *server = NFS_SERVER(dir);
1632         struct nfs_openargs *o_arg = &data->o_arg;
1633         struct nfs_openres *o_res = &data->o_res;
1634         int status;
1635
1636         status = nfs4_run_open_task(data, 0);
1637         if (!data->rpc_done)
1638                 return status;
1639         if (status != 0) {
1640                 if (status == -NFS4ERR_BADNAME &&
1641                                 !(o_arg->open_flags & O_CREAT))
1642                         return -ENOENT;
1643                 return status;
1644         }
1645
1646         nfs_fattr_map_and_free_names(server, &data->f_attr);
1647
1648         if (o_arg->open_flags & O_CREAT) {
1649                 update_changeattr(dir, &o_res->cinfo);
1650                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1651         } else
1652                 nfs_refresh_inode(dir, o_res->dir_attr);
1653         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1654                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1655         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1656                 status = _nfs4_proc_open_confirm(data);
1657                 if (status != 0)
1658                         return status;
1659         }
1660         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1661                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1662         return 0;
1663 }
1664
1665 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1666 {
1667         unsigned int loop;
1668         int ret;
1669
1670         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1671                 ret = nfs4_wait_clnt_recover(clp);
1672                 if (ret != 0)
1673                         break;
1674                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1675                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1676                         break;
1677                 nfs4_schedule_state_manager(clp);
1678                 ret = -EIO;
1679         }
1680         return ret;
1681 }
1682
1683 static int nfs4_recover_expired_lease(struct nfs_server *server)
1684 {
1685         return nfs4_client_recover_expired_lease(server->nfs_client);
1686 }
1687
1688 /*
1689  * OPEN_EXPIRED:
1690  *      reclaim state on the server after a network partition.
1691  *      Assumes caller holds the appropriate lock
1692  */
1693 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1694 {
1695         struct nfs4_opendata *opendata;
1696         int ret;
1697
1698         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1699         if (IS_ERR(opendata))
1700                 return PTR_ERR(opendata);
1701         ret = nfs4_open_recover(opendata, state);
1702         if (ret == -ESTALE)
1703                 d_drop(ctx->dentry);
1704         nfs4_opendata_put(opendata);
1705         return ret;
1706 }
1707
1708 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1709 {
1710         struct nfs_server *server = NFS_SERVER(state->inode);
1711         struct nfs4_exception exception = { };
1712         int err;
1713
1714         do {
1715                 err = _nfs4_open_expired(ctx, state);
1716                 switch (err) {
1717                 default:
1718                         goto out;
1719                 case -NFS4ERR_GRACE:
1720                 case -NFS4ERR_DELAY:
1721                         nfs4_handle_exception(server, err, &exception);
1722                         err = 0;
1723                 }
1724         } while (exception.retry);
1725 out:
1726         return err;
1727 }
1728
1729 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1730 {
1731         struct nfs_open_context *ctx;
1732         int ret;
1733
1734         ctx = nfs4_state_find_open_context(state);
1735         if (IS_ERR(ctx))
1736                 return PTR_ERR(ctx);
1737         ret = nfs4_do_open_expired(ctx, state);
1738         put_nfs_open_context(ctx);
1739         return ret;
1740 }
1741
1742 #if defined(CONFIG_NFS_V4_1)
1743 static int nfs41_check_expired_stateid(struct nfs4_state *state, nfs4_stateid *stateid, unsigned int flags)
1744 {
1745         int status = NFS_OK;
1746         struct nfs_server *server = NFS_SERVER(state->inode);
1747
1748         if (state->flags & flags) {
1749                 status = nfs41_test_stateid(server, stateid);
1750                 if (status != NFS_OK) {
1751                         nfs41_free_stateid(server, stateid);
1752                         state->flags &= ~flags;
1753                 }
1754         }
1755         return status;
1756 }
1757
1758 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1759 {
1760         int deleg_status, open_status;
1761         int deleg_flags = 1 << NFS_DELEGATED_STATE;
1762         int open_flags = (1 << NFS_O_RDONLY_STATE) | (1 << NFS_O_WRONLY_STATE) | (1 << NFS_O_RDWR_STATE);
1763
1764         deleg_status = nfs41_check_expired_stateid(state, &state->stateid, deleg_flags);
1765         open_status = nfs41_check_expired_stateid(state,  &state->open_stateid, open_flags);
1766
1767         if ((deleg_status == NFS_OK) && (open_status == NFS_OK))
1768                 return NFS_OK;
1769         return nfs4_open_expired(sp, state);
1770 }
1771 #endif
1772
1773 /*
1774  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1775  * fields corresponding to attributes that were used to store the verifier.
1776  * Make sure we clobber those fields in the later setattr call
1777  */
1778 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1779 {
1780         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1781             !(sattr->ia_valid & ATTR_ATIME_SET))
1782                 sattr->ia_valid |= ATTR_ATIME;
1783
1784         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1785             !(sattr->ia_valid & ATTR_MTIME_SET))
1786                 sattr->ia_valid |= ATTR_MTIME;
1787 }
1788
1789 /*
1790  * Returns a referenced nfs4_state
1791  */
1792 static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1793 {
1794         struct nfs4_state_owner  *sp;
1795         struct nfs4_state     *state = NULL;
1796         struct nfs_server       *server = NFS_SERVER(dir);
1797         struct nfs4_opendata *opendata;
1798         int status;
1799
1800         /* Protect against reboot recovery conflicts */
1801         status = -ENOMEM;
1802         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1803         if (sp == NULL) {
1804                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1805                 goto out_err;
1806         }
1807         status = nfs4_recover_expired_lease(server);
1808         if (status != 0)
1809                 goto err_put_state_owner;
1810         if (dentry->d_inode != NULL)
1811                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1812         status = -ENOMEM;
1813         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1814         if (opendata == NULL)
1815                 goto err_put_state_owner;
1816
1817         if (dentry->d_inode != NULL)
1818                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1819
1820         status = _nfs4_proc_open(opendata);
1821         if (status != 0)
1822                 goto err_opendata_put;
1823
1824         state = nfs4_opendata_to_nfs4_state(opendata);
1825         status = PTR_ERR(state);
1826         if (IS_ERR(state))
1827                 goto err_opendata_put;
1828         if (server->caps & NFS_CAP_POSIX_LOCK)
1829                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1830
1831         if (opendata->o_arg.open_flags & O_EXCL) {
1832                 nfs4_exclusive_attrset(opendata, sattr);
1833
1834                 nfs_fattr_init(opendata->o_res.f_attr);
1835                 status = nfs4_do_setattr(state->inode, cred,
1836                                 opendata->o_res.f_attr, sattr,
1837                                 state);
1838                 if (status == 0)
1839                         nfs_setattr_update_inode(state->inode, sattr);
1840                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1841         }
1842         nfs4_opendata_put(opendata);
1843         nfs4_put_state_owner(sp);
1844         *res = state;
1845         return 0;
1846 err_opendata_put:
1847         nfs4_opendata_put(opendata);
1848 err_put_state_owner:
1849         nfs4_put_state_owner(sp);
1850 out_err:
1851         *res = NULL;
1852         return status;
1853 }
1854
1855
1856 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1857 {
1858         struct nfs4_exception exception = { };
1859         struct nfs4_state *res;
1860         int status;
1861
1862         do {
1863                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred, &res);
1864                 if (status == 0)
1865                         break;
1866                 /* NOTE: BAD_SEQID means the server and client disagree about the
1867                  * book-keeping w.r.t. state-changing operations
1868                  * (OPEN/CLOSE/LOCK/LOCKU...)
1869                  * It is actually a sign of a bug on the client or on the server.
1870                  *
1871                  * If we receive a BAD_SEQID error in the particular case of
1872                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1873                  * have unhashed the old state_owner for us, and that we can
1874                  * therefore safely retry using a new one. We should still warn
1875                  * the user though...
1876                  */
1877                 if (status == -NFS4ERR_BAD_SEQID) {
1878                         pr_warn_ratelimited("NFS: v4 server %s "
1879                                         " returned a bad sequence-id error!\n",
1880                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1881                         exception.retry = 1;
1882                         continue;
1883                 }
1884                 /*
1885                  * BAD_STATEID on OPEN means that the server cancelled our
1886                  * state before it received the OPEN_CONFIRM.
1887                  * Recover by retrying the request as per the discussion
1888                  * on Page 181 of RFC3530.
1889                  */
1890                 if (status == -NFS4ERR_BAD_STATEID) {
1891                         exception.retry = 1;
1892                         continue;
1893                 }
1894                 if (status == -EAGAIN) {
1895                         /* We must have found a delegation */
1896                         exception.retry = 1;
1897                         continue;
1898                 }
1899                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1900                                         status, &exception));
1901         } while (exception.retry);
1902         return res;
1903 }
1904
1905 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1906                             struct nfs_fattr *fattr, struct iattr *sattr,
1907                             struct nfs4_state *state)
1908 {
1909         struct nfs_server *server = NFS_SERVER(inode);
1910         struct nfs_setattrargs  arg = {
1911                 .fh             = NFS_FH(inode),
1912                 .iap            = sattr,
1913                 .server         = server,
1914                 .bitmask = server->attr_bitmask,
1915         };
1916         struct nfs_setattrres  res = {
1917                 .fattr          = fattr,
1918                 .server         = server,
1919         };
1920         struct rpc_message msg = {
1921                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1922                 .rpc_argp       = &arg,
1923                 .rpc_resp       = &res,
1924                 .rpc_cred       = cred,
1925         };
1926         unsigned long timestamp = jiffies;
1927         int status;
1928
1929         nfs_fattr_init(fattr);
1930
1931         if (state != NULL) {
1932                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
1933                                 current->files, current->tgid);
1934         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
1935                                 FMODE_WRITE)) {
1936                 /* Use that stateid */
1937         } else
1938                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
1939
1940         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1941         if (status == 0 && state != NULL)
1942                 renew_lease(server, timestamp);
1943         return status;
1944 }
1945
1946 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1947                            struct nfs_fattr *fattr, struct iattr *sattr,
1948                            struct nfs4_state *state)
1949 {
1950         struct nfs_server *server = NFS_SERVER(inode);
1951         struct nfs4_exception exception = {
1952                 .state = state,
1953                 .inode = inode,
1954         };
1955         int err;
1956         do {
1957                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
1958                 switch (err) {
1959                 case -NFS4ERR_OPENMODE:
1960                         if (state && !(state->state & FMODE_WRITE)) {
1961                                 err = -EBADF;
1962                                 if (sattr->ia_valid & ATTR_OPEN)
1963                                         err = -EACCES;
1964                                 goto out;
1965                         }
1966                 }
1967                 err = nfs4_handle_exception(server, err, &exception);
1968         } while (exception.retry);
1969 out:
1970         return err;
1971 }
1972
1973 struct nfs4_closedata {
1974         struct inode *inode;
1975         struct nfs4_state *state;
1976         struct nfs_closeargs arg;
1977         struct nfs_closeres res;
1978         struct nfs_fattr fattr;
1979         unsigned long timestamp;
1980         bool roc;
1981         u32 roc_barrier;
1982 };
1983
1984 static void nfs4_free_closedata(void *data)
1985 {
1986         struct nfs4_closedata *calldata = data;
1987         struct nfs4_state_owner *sp = calldata->state->owner;
1988         struct super_block *sb = calldata->state->inode->i_sb;
1989
1990         if (calldata->roc)
1991                 pnfs_roc_release(calldata->state->inode);
1992         nfs4_put_open_state(calldata->state);
1993         nfs_free_seqid(calldata->arg.seqid);
1994         nfs4_put_state_owner(sp);
1995         nfs_sb_deactive(sb);
1996         kfree(calldata);
1997 }
1998
1999 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2000                 fmode_t fmode)
2001 {
2002         spin_lock(&state->owner->so_lock);
2003         if (!(fmode & FMODE_READ))
2004                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2005         if (!(fmode & FMODE_WRITE))
2006                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2007         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2008         spin_unlock(&state->owner->so_lock);
2009 }
2010
2011 static void nfs4_close_done(struct rpc_task *task, void *data)
2012 {
2013         struct nfs4_closedata *calldata = data;
2014         struct nfs4_state *state = calldata->state;
2015         struct nfs_server *server = NFS_SERVER(calldata->inode);
2016
2017         dprintk("%s: begin!\n", __func__);
2018         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2019                 return;
2020         /* hmm. we are done with the inode, and in the process of freeing
2021          * the state_owner. we keep this around to process errors
2022          */
2023         switch (task->tk_status) {
2024                 case 0:
2025                         if (calldata->roc)
2026                                 pnfs_roc_set_barrier(state->inode,
2027                                                      calldata->roc_barrier);
2028                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2029                         renew_lease(server, calldata->timestamp);
2030                         nfs4_close_clear_stateid_flags(state,
2031                                         calldata->arg.fmode);
2032                         break;
2033                 case -NFS4ERR_STALE_STATEID:
2034                 case -NFS4ERR_OLD_STATEID:
2035                 case -NFS4ERR_BAD_STATEID:
2036                 case -NFS4ERR_EXPIRED:
2037                         if (calldata->arg.fmode == 0)
2038                                 break;
2039                 default:
2040                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2041                                 rpc_restart_call_prepare(task);
2042         }
2043         nfs_release_seqid(calldata->arg.seqid);
2044         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2045         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2046 }
2047
2048 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2049 {
2050         struct nfs4_closedata *calldata = data;
2051         struct nfs4_state *state = calldata->state;
2052         int call_close = 0;
2053
2054         dprintk("%s: begin!\n", __func__);
2055         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2056                 return;
2057
2058         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2059         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2060         spin_lock(&state->owner->so_lock);
2061         /* Calculate the change in open mode */
2062         if (state->n_rdwr == 0) {
2063                 if (state->n_rdonly == 0) {
2064                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2065                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2066                         calldata->arg.fmode &= ~FMODE_READ;
2067                 }
2068                 if (state->n_wronly == 0) {
2069                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2070                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2071                         calldata->arg.fmode &= ~FMODE_WRITE;
2072                 }
2073         }
2074         spin_unlock(&state->owner->so_lock);
2075
2076         if (!call_close) {
2077                 /* Note: exit _without_ calling nfs4_close_done */
2078                 task->tk_action = NULL;
2079                 goto out;
2080         }
2081
2082         if (calldata->arg.fmode == 0) {
2083                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2084                 if (calldata->roc &&
2085                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2086                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2087                                      task, NULL);
2088                         goto out;
2089                 }
2090         }
2091
2092         nfs_fattr_init(calldata->res.fattr);
2093         calldata->timestamp = jiffies;
2094         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2095                                 &calldata->arg.seq_args,
2096                                 &calldata->res.seq_res,
2097                                 task))
2098                 goto out;
2099         rpc_call_start(task);
2100 out:
2101         dprintk("%s: done!\n", __func__);
2102 }
2103
2104 static const struct rpc_call_ops nfs4_close_ops = {
2105         .rpc_call_prepare = nfs4_close_prepare,
2106         .rpc_call_done = nfs4_close_done,
2107         .rpc_release = nfs4_free_closedata,
2108 };
2109
2110 /* 
2111  * It is possible for data to be read/written from a mem-mapped file 
2112  * after the sys_close call (which hits the vfs layer as a flush).
2113  * This means that we can't safely call nfsv4 close on a file until 
2114  * the inode is cleared. This in turn means that we are not good
2115  * NFSv4 citizens - we do not indicate to the server to update the file's 
2116  * share state even when we are done with one of the three share 
2117  * stateid's in the inode.
2118  *
2119  * NOTE: Caller must be holding the sp->so_owner semaphore!
2120  */
2121 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2122 {
2123         struct nfs_server *server = NFS_SERVER(state->inode);
2124         struct nfs4_closedata *calldata;
2125         struct nfs4_state_owner *sp = state->owner;
2126         struct rpc_task *task;
2127         struct rpc_message msg = {
2128                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2129                 .rpc_cred = state->owner->so_cred,
2130         };
2131         struct rpc_task_setup task_setup_data = {
2132                 .rpc_client = server->client,
2133                 .rpc_message = &msg,
2134                 .callback_ops = &nfs4_close_ops,
2135                 .workqueue = nfsiod_workqueue,
2136                 .flags = RPC_TASK_ASYNC,
2137         };
2138         int status = -ENOMEM;
2139
2140         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2141         if (calldata == NULL)
2142                 goto out;
2143         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2144         calldata->inode = state->inode;
2145         calldata->state = state;
2146         calldata->arg.fh = NFS_FH(state->inode);
2147         calldata->arg.stateid = &state->open_stateid;
2148         /* Serialization for the sequence id */
2149         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2150         if (calldata->arg.seqid == NULL)
2151                 goto out_free_calldata;
2152         calldata->arg.fmode = 0;
2153         calldata->arg.bitmask = server->cache_consistency_bitmask;
2154         calldata->res.fattr = &calldata->fattr;
2155         calldata->res.seqid = calldata->arg.seqid;
2156         calldata->res.server = server;
2157         calldata->roc = roc;
2158         nfs_sb_active(calldata->inode->i_sb);
2159
2160         msg.rpc_argp = &calldata->arg;
2161         msg.rpc_resp = &calldata->res;
2162         task_setup_data.callback_data = calldata;
2163         task = rpc_run_task(&task_setup_data);
2164         if (IS_ERR(task))
2165                 return PTR_ERR(task);
2166         status = 0;
2167         if (wait)
2168                 status = rpc_wait_for_completion_task(task);
2169         rpc_put_task(task);
2170         return status;
2171 out_free_calldata:
2172         kfree(calldata);
2173 out:
2174         if (roc)
2175                 pnfs_roc_release(state->inode);
2176         nfs4_put_open_state(state);
2177         nfs4_put_state_owner(sp);
2178         return status;
2179 }
2180
2181 static struct inode *
2182 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2183 {
2184         struct nfs4_state *state;
2185
2186         /* Protect against concurrent sillydeletes */
2187         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2188         if (IS_ERR(state))
2189                 return ERR_CAST(state);
2190         ctx->state = state;
2191         return igrab(state->inode);
2192 }
2193
2194 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2195 {
2196         if (ctx->state == NULL)
2197                 return;
2198         if (is_sync)
2199                 nfs4_close_sync(ctx->state, ctx->mode);
2200         else
2201                 nfs4_close_state(ctx->state, ctx->mode);
2202 }
2203
2204 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2205 {
2206         struct nfs4_server_caps_arg args = {
2207                 .fhandle = fhandle,
2208         };
2209         struct nfs4_server_caps_res res = {};
2210         struct rpc_message msg = {
2211                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2212                 .rpc_argp = &args,
2213                 .rpc_resp = &res,
2214         };
2215         int status;
2216
2217         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2218         if (status == 0) {
2219                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2220                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2221                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2222                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2223                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2224                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2225                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2226                         server->caps |= NFS_CAP_ACLS;
2227                 if (res.has_links != 0)
2228                         server->caps |= NFS_CAP_HARDLINKS;
2229                 if (res.has_symlinks != 0)
2230                         server->caps |= NFS_CAP_SYMLINKS;
2231                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2232                         server->caps |= NFS_CAP_FILEID;
2233                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2234                         server->caps |= NFS_CAP_MODE;
2235                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2236                         server->caps |= NFS_CAP_NLINK;
2237                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2238                         server->caps |= NFS_CAP_OWNER;
2239                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2240                         server->caps |= NFS_CAP_OWNER_GROUP;
2241                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2242                         server->caps |= NFS_CAP_ATIME;
2243                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2244                         server->caps |= NFS_CAP_CTIME;
2245                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2246                         server->caps |= NFS_CAP_MTIME;
2247
2248                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2249                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2250                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2251                 server->acl_bitmask = res.acl_bitmask;
2252                 server->fh_expire_type = res.fh_expire_type;
2253         }
2254
2255         return status;
2256 }
2257
2258 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2259 {
2260         struct nfs4_exception exception = { };
2261         int err;
2262         do {
2263                 err = nfs4_handle_exception(server,
2264                                 _nfs4_server_capabilities(server, fhandle),
2265                                 &exception);
2266         } while (exception.retry);
2267         return err;
2268 }
2269
2270 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2271                 struct nfs_fsinfo *info)
2272 {
2273         struct nfs4_lookup_root_arg args = {
2274                 .bitmask = nfs4_fattr_bitmap,
2275         };
2276         struct nfs4_lookup_res res = {
2277                 .server = server,
2278                 .fattr = info->fattr,
2279                 .fh = fhandle,
2280         };
2281         struct rpc_message msg = {
2282                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2283                 .rpc_argp = &args,
2284                 .rpc_resp = &res,
2285         };
2286
2287         nfs_fattr_init(info->fattr);
2288         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2289 }
2290
2291 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2292                 struct nfs_fsinfo *info)
2293 {
2294         struct nfs4_exception exception = { };
2295         int err;
2296         do {
2297                 err = _nfs4_lookup_root(server, fhandle, info);
2298                 switch (err) {
2299                 case 0:
2300                 case -NFS4ERR_WRONGSEC:
2301                         goto out;
2302                 default:
2303                         err = nfs4_handle_exception(server, err, &exception);
2304                 }
2305         } while (exception.retry);
2306 out:
2307         return err;
2308 }
2309
2310 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2311                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2312 {
2313         struct rpc_auth *auth;
2314         int ret;
2315
2316         auth = rpcauth_create(flavor, server->client);
2317         if (!auth) {
2318                 ret = -EIO;
2319                 goto out;
2320         }
2321         ret = nfs4_lookup_root(server, fhandle, info);
2322 out:
2323         return ret;
2324 }
2325
2326 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2327                               struct nfs_fsinfo *info)
2328 {
2329         int i, len, status = 0;
2330         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2331
2332         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2333         flav_array[len] = RPC_AUTH_NULL;
2334         len += 1;
2335
2336         for (i = 0; i < len; i++) {
2337                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2338                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2339                         continue;
2340                 break;
2341         }
2342         /*
2343          * -EACCESS could mean that the user doesn't have correct permissions
2344          * to access the mount.  It could also mean that we tried to mount
2345          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2346          * existing mount programs don't handle -EACCES very well so it should
2347          * be mapped to -EPERM instead.
2348          */
2349         if (status == -EACCES)
2350                 status = -EPERM;
2351         return status;
2352 }
2353
2354 /*
2355  * get the file handle for the "/" directory on the server
2356  */
2357 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2358                               struct nfs_fsinfo *info)
2359 {
2360         int minor_version = server->nfs_client->cl_minorversion;
2361         int status = nfs4_lookup_root(server, fhandle, info);
2362         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2363                 /*
2364                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2365                  * by nfs4_map_errors() as this function exits.
2366                  */
2367                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2368         if (status == 0)
2369                 status = nfs4_server_capabilities(server, fhandle);
2370         if (status == 0)
2371                 status = nfs4_do_fsinfo(server, fhandle, info);
2372         return nfs4_map_errors(status);
2373 }
2374
2375 /*
2376  * Get locations and (maybe) other attributes of a referral.
2377  * Note that we'll actually follow the referral later when
2378  * we detect fsid mismatch in inode revalidation
2379  */
2380 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2381                              const struct qstr *name, struct nfs_fattr *fattr,
2382                              struct nfs_fh *fhandle)
2383 {
2384         int status = -ENOMEM;
2385         struct page *page = NULL;
2386         struct nfs4_fs_locations *locations = NULL;
2387
2388         page = alloc_page(GFP_KERNEL);
2389         if (page == NULL)
2390                 goto out;
2391         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2392         if (locations == NULL)
2393                 goto out;
2394
2395         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2396         if (status != 0)
2397                 goto out;
2398         /* Make sure server returned a different fsid for the referral */
2399         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2400                 dprintk("%s: server did not return a different fsid for"
2401                         " a referral at %s\n", __func__, name->name);
2402                 status = -EIO;
2403                 goto out;
2404         }
2405         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2406         nfs_fixup_referral_attributes(&locations->fattr);
2407
2408         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2409         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2410         memset(fhandle, 0, sizeof(struct nfs_fh));
2411 out:
2412         if (page)
2413                 __free_page(page);
2414         kfree(locations);
2415         return status;
2416 }
2417
2418 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2419 {
2420         struct nfs4_getattr_arg args = {
2421                 .fh = fhandle,
2422                 .bitmask = server->attr_bitmask,
2423         };
2424         struct nfs4_getattr_res res = {
2425                 .fattr = fattr,
2426                 .server = server,
2427         };
2428         struct rpc_message msg = {
2429                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2430                 .rpc_argp = &args,
2431                 .rpc_resp = &res,
2432         };
2433         
2434         nfs_fattr_init(fattr);
2435         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2436 }
2437
2438 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2439 {
2440         struct nfs4_exception exception = { };
2441         int err;
2442         do {
2443                 err = nfs4_handle_exception(server,
2444                                 _nfs4_proc_getattr(server, fhandle, fattr),
2445                                 &exception);
2446         } while (exception.retry);
2447         return err;
2448 }
2449
2450 /* 
2451  * The file is not closed if it is opened due to the a request to change
2452  * the size of the file. The open call will not be needed once the
2453  * VFS layer lookup-intents are implemented.
2454  *
2455  * Close is called when the inode is destroyed.
2456  * If we haven't opened the file for O_WRONLY, we
2457  * need to in the size_change case to obtain a stateid.
2458  *
2459  * Got race?
2460  * Because OPEN is always done by name in nfsv4, it is
2461  * possible that we opened a different file by the same
2462  * name.  We can recognize this race condition, but we
2463  * can't do anything about it besides returning an error.
2464  *
2465  * This will be fixed with VFS changes (lookup-intent).
2466  */
2467 static int
2468 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2469                   struct iattr *sattr)
2470 {
2471         struct inode *inode = dentry->d_inode;
2472         struct rpc_cred *cred = NULL;
2473         struct nfs4_state *state = NULL;
2474         int status;
2475
2476         if (pnfs_ld_layoutret_on_setattr(inode))
2477                 pnfs_return_layout(inode);
2478
2479         nfs_fattr_init(fattr);
2480         
2481         /* Search for an existing open(O_WRITE) file */
2482         if (sattr->ia_valid & ATTR_FILE) {
2483                 struct nfs_open_context *ctx;
2484
2485                 ctx = nfs_file_open_context(sattr->ia_file);
2486                 if (ctx) {
2487                         cred = ctx->cred;
2488                         state = ctx->state;
2489                 }
2490         }
2491
2492         /* Deal with open(O_TRUNC) */
2493         if (sattr->ia_valid & ATTR_OPEN)
2494                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2495
2496         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2497         if (status == 0)
2498                 nfs_setattr_update_inode(inode, sattr);
2499         return status;
2500 }
2501
2502 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2503                 const struct qstr *name, struct nfs_fh *fhandle,
2504                 struct nfs_fattr *fattr)
2505 {
2506         struct nfs_server *server = NFS_SERVER(dir);
2507         int                    status;
2508         struct nfs4_lookup_arg args = {
2509                 .bitmask = server->attr_bitmask,
2510                 .dir_fh = NFS_FH(dir),
2511                 .name = name,
2512         };
2513         struct nfs4_lookup_res res = {
2514                 .server = server,
2515                 .fattr = fattr,
2516                 .fh = fhandle,
2517         };
2518         struct rpc_message msg = {
2519                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2520                 .rpc_argp = &args,
2521                 .rpc_resp = &res,
2522         };
2523
2524         nfs_fattr_init(fattr);
2525
2526         dprintk("NFS call  lookup %s\n", name->name);
2527         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2528         dprintk("NFS reply lookup: %d\n", status);
2529         return status;
2530 }
2531
2532 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2533 {
2534         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2535                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2536         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2537         fattr->nlink = 2;
2538 }
2539
2540 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2541                                    struct qstr *name, struct nfs_fh *fhandle,
2542                                    struct nfs_fattr *fattr)
2543 {
2544         struct nfs4_exception exception = { };
2545         struct rpc_clnt *client = *clnt;
2546         int err;
2547         do {
2548                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2549                 switch (err) {
2550                 case -NFS4ERR_BADNAME:
2551                         err = -ENOENT;
2552                         goto out;
2553                 case -NFS4ERR_MOVED:
2554                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2555                         goto out;
2556                 case -NFS4ERR_WRONGSEC:
2557                         err = -EPERM;
2558                         if (client != *clnt)
2559                                 goto out;
2560
2561                         client = nfs4_create_sec_client(client, dir, name);
2562                         if (IS_ERR(client))
2563                                 return PTR_ERR(client);
2564
2565                         exception.retry = 1;
2566                         break;
2567                 default:
2568                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2569                 }
2570         } while (exception.retry);
2571
2572 out:
2573         if (err == 0)
2574                 *clnt = client;
2575         else if (client != *clnt)
2576                 rpc_shutdown_client(client);
2577
2578         return err;
2579 }
2580
2581 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2582                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2583 {
2584         int status;
2585         struct rpc_clnt *client = NFS_CLIENT(dir);
2586
2587         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2588         if (client != NFS_CLIENT(dir)) {
2589                 rpc_shutdown_client(client);
2590                 nfs_fixup_secinfo_attributes(fattr);
2591         }
2592         return status;
2593 }
2594
2595 struct rpc_clnt *
2596 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2597                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2598 {
2599         int status;
2600         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2601
2602         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2603         if (status < 0) {
2604                 rpc_shutdown_client(client);
2605                 return ERR_PTR(status);
2606         }
2607         return client;
2608 }
2609
2610 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2611 {
2612         struct nfs_server *server = NFS_SERVER(inode);
2613         struct nfs4_accessargs args = {
2614                 .fh = NFS_FH(inode),
2615                 .bitmask = server->cache_consistency_bitmask,
2616         };
2617         struct nfs4_accessres res = {
2618                 .server = server,
2619         };
2620         struct rpc_message msg = {
2621                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2622                 .rpc_argp = &args,
2623                 .rpc_resp = &res,
2624                 .rpc_cred = entry->cred,
2625         };
2626         int mode = entry->mask;
2627         int status;
2628
2629         /*
2630          * Determine which access bits we want to ask for...
2631          */
2632         if (mode & MAY_READ)
2633                 args.access |= NFS4_ACCESS_READ;
2634         if (S_ISDIR(inode->i_mode)) {
2635                 if (mode & MAY_WRITE)
2636                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2637                 if (mode & MAY_EXEC)
2638                         args.access |= NFS4_ACCESS_LOOKUP;
2639         } else {
2640                 if (mode & MAY_WRITE)
2641                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2642                 if (mode & MAY_EXEC)
2643                         args.access |= NFS4_ACCESS_EXECUTE;
2644         }
2645
2646         res.fattr = nfs_alloc_fattr();
2647         if (res.fattr == NULL)
2648                 return -ENOMEM;
2649
2650         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2651         if (!status) {
2652                 entry->mask = 0;
2653                 if (res.access & NFS4_ACCESS_READ)
2654                         entry->mask |= MAY_READ;
2655                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2656                         entry->mask |= MAY_WRITE;
2657                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2658                         entry->mask |= MAY_EXEC;
2659                 nfs_refresh_inode(inode, res.fattr);
2660         }
2661         nfs_free_fattr(res.fattr);
2662         return status;
2663 }
2664
2665 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2666 {
2667         struct nfs4_exception exception = { };
2668         int err;
2669         do {
2670                 err = nfs4_handle_exception(NFS_SERVER(inode),
2671                                 _nfs4_proc_access(inode, entry),
2672                                 &exception);
2673         } while (exception.retry);
2674         return err;
2675 }
2676
2677 /*
2678  * TODO: For the time being, we don't try to get any attributes
2679  * along with any of the zero-copy operations READ, READDIR,
2680  * READLINK, WRITE.
2681  *
2682  * In the case of the first three, we want to put the GETATTR
2683  * after the read-type operation -- this is because it is hard
2684  * to predict the length of a GETATTR response in v4, and thus
2685  * align the READ data correctly.  This means that the GETATTR
2686  * may end up partially falling into the page cache, and we should
2687  * shift it into the 'tail' of the xdr_buf before processing.
2688  * To do this efficiently, we need to know the total length
2689  * of data received, which doesn't seem to be available outside
2690  * of the RPC layer.
2691  *
2692  * In the case of WRITE, we also want to put the GETATTR after
2693  * the operation -- in this case because we want to make sure
2694  * we get the post-operation mtime and size.  This means that
2695  * we can't use xdr_encode_pages() as written: we need a variant
2696  * of it which would leave room in the 'tail' iovec.
2697  *
2698  * Both of these changes to the XDR layer would in fact be quite
2699  * minor, but I decided to leave them for a subsequent patch.
2700  */
2701 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2702                 unsigned int pgbase, unsigned int pglen)
2703 {
2704         struct nfs4_readlink args = {
2705                 .fh       = NFS_FH(inode),
2706                 .pgbase   = pgbase,
2707                 .pglen    = pglen,
2708                 .pages    = &page,
2709         };
2710         struct nfs4_readlink_res res;
2711         struct rpc_message msg = {
2712                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2713                 .rpc_argp = &args,
2714                 .rpc_resp = &res,
2715         };
2716
2717         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2718 }
2719
2720 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2721                 unsigned int pgbase, unsigned int pglen)
2722 {
2723         struct nfs4_exception exception = { };
2724         int err;
2725         do {
2726                 err = nfs4_handle_exception(NFS_SERVER(inode),
2727                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2728                                 &exception);
2729         } while (exception.retry);
2730         return err;
2731 }
2732
2733 /*
2734  * Got race?
2735  * We will need to arrange for the VFS layer to provide an atomic open.
2736  * Until then, this create/open method is prone to inefficiency and race
2737  * conditions due to the lookup, create, and open VFS calls from sys_open()
2738  * placed on the wire.
2739  *
2740  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2741  * The file will be opened again in the subsequent VFS open call
2742  * (nfs4_proc_file_open).
2743  *
2744  * The open for read will just hang around to be used by any process that
2745  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2746  */
2747
2748 static int
2749 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2750                  int flags, struct nfs_open_context *ctx)
2751 {
2752         struct dentry *de = dentry;
2753         struct nfs4_state *state;
2754         struct rpc_cred *cred = NULL;
2755         fmode_t fmode = 0;
2756         int status = 0;
2757
2758         if (ctx != NULL) {
2759                 cred = ctx->cred;
2760                 de = ctx->dentry;
2761                 fmode = ctx->mode;
2762         }
2763         sattr->ia_mode &= ~current_umask();
2764         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2765         d_drop(dentry);
2766         if (IS_ERR(state)) {
2767                 status = PTR_ERR(state);
2768                 goto out;
2769         }
2770         d_add(dentry, igrab(state->inode));
2771         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2772         if (ctx != NULL)
2773                 ctx->state = state;
2774         else
2775                 nfs4_close_sync(state, fmode);
2776 out:
2777         return status;
2778 }
2779
2780 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2781 {
2782         struct nfs_server *server = NFS_SERVER(dir);
2783         struct nfs_removeargs args = {
2784                 .fh = NFS_FH(dir),
2785                 .name.len = name->len,
2786                 .name.name = name->name,
2787                 .bitmask = server->attr_bitmask,
2788         };
2789         struct nfs_removeres res = {
2790                 .server = server,
2791         };
2792         struct rpc_message msg = {
2793                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2794                 .rpc_argp = &args,
2795                 .rpc_resp = &res,
2796         };
2797         int status = -ENOMEM;
2798
2799         res.dir_attr = nfs_alloc_fattr();
2800         if (res.dir_attr == NULL)
2801                 goto out;
2802
2803         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2804         if (status == 0) {
2805                 update_changeattr(dir, &res.cinfo);
2806                 nfs_post_op_update_inode(dir, res.dir_attr);
2807         }
2808         nfs_free_fattr(res.dir_attr);
2809 out:
2810         return status;
2811 }
2812
2813 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2814 {
2815         struct nfs4_exception exception = { };
2816         int err;
2817         do {
2818                 err = nfs4_handle_exception(NFS_SERVER(dir),
2819                                 _nfs4_proc_remove(dir, name),
2820                                 &exception);
2821         } while (exception.retry);
2822         return err;
2823 }
2824
2825 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2826 {
2827         struct nfs_server *server = NFS_SERVER(dir);
2828         struct nfs_removeargs *args = msg->rpc_argp;
2829         struct nfs_removeres *res = msg->rpc_resp;
2830
2831         args->bitmask = server->cache_consistency_bitmask;
2832         res->server = server;
2833         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2834         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2835 }
2836
2837 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
2838 {
2839         if (nfs4_setup_sequence(NFS_SERVER(data->dir),
2840                                 &data->args.seq_args,
2841                                 &data->res.seq_res,
2842                                 task))
2843                 return;
2844         rpc_call_start(task);
2845 }
2846
2847 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2848 {
2849         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2850
2851         if (!nfs4_sequence_done(task, &res->seq_res))
2852                 return 0;
2853         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2854                 return 0;
2855         update_changeattr(dir, &res->cinfo);
2856         nfs_post_op_update_inode(dir, res->dir_attr);
2857         return 1;
2858 }
2859
2860 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2861 {
2862         struct nfs_server *server = NFS_SERVER(dir);
2863         struct nfs_renameargs *arg = msg->rpc_argp;
2864         struct nfs_renameres *res = msg->rpc_resp;
2865
2866         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2867         arg->bitmask = server->attr_bitmask;
2868         res->server = server;
2869         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2870 }
2871
2872 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
2873 {
2874         if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
2875                                 &data->args.seq_args,
2876                                 &data->res.seq_res,
2877                                 task))
2878                 return;
2879         rpc_call_start(task);
2880 }
2881
2882 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2883                                  struct inode *new_dir)
2884 {
2885         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2886
2887         if (!nfs4_sequence_done(task, &res->seq_res))
2888                 return 0;
2889         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2890                 return 0;
2891
2892         update_changeattr(old_dir, &res->old_cinfo);
2893         nfs_post_op_update_inode(old_dir, res->old_fattr);
2894         update_changeattr(new_dir, &res->new_cinfo);
2895         nfs_post_op_update_inode(new_dir, res->new_fattr);
2896         return 1;
2897 }
2898
2899 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2900                 struct inode *new_dir, struct qstr *new_name)
2901 {
2902         struct nfs_server *server = NFS_SERVER(old_dir);
2903         struct nfs_renameargs arg = {
2904                 .old_dir = NFS_FH(old_dir),
2905                 .new_dir = NFS_FH(new_dir),
2906                 .old_name = old_name,
2907                 .new_name = new_name,
2908                 .bitmask = server->attr_bitmask,
2909         };
2910         struct nfs_renameres res = {
2911                 .server = server,
2912         };
2913         struct rpc_message msg = {
2914                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2915                 .rpc_argp = &arg,
2916                 .rpc_resp = &res,
2917         };
2918         int status = -ENOMEM;
2919         
2920         res.old_fattr = nfs_alloc_fattr();
2921         res.new_fattr = nfs_alloc_fattr();
2922         if (res.old_fattr == NULL || res.new_fattr == NULL)
2923                 goto out;
2924
2925         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2926         if (!status) {
2927                 update_changeattr(old_dir, &res.old_cinfo);
2928                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2929                 update_changeattr(new_dir, &res.new_cinfo);
2930                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2931         }
2932 out:
2933         nfs_free_fattr(res.new_fattr);
2934         nfs_free_fattr(res.old_fattr);
2935         return status;
2936 }
2937
2938 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2939                 struct inode *new_dir, struct qstr *new_name)
2940 {
2941         struct nfs4_exception exception = { };
2942         int err;
2943         do {
2944                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2945                                 _nfs4_proc_rename(old_dir, old_name,
2946                                         new_dir, new_name),
2947                                 &exception);
2948         } while (exception.retry);
2949         return err;
2950 }
2951
2952 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2953 {
2954         struct nfs_server *server = NFS_SERVER(inode);
2955         struct nfs4_link_arg arg = {
2956                 .fh     = NFS_FH(inode),
2957                 .dir_fh = NFS_FH(dir),
2958                 .name   = name,
2959                 .bitmask = server->attr_bitmask,
2960         };
2961         struct nfs4_link_res res = {
2962                 .server = server,
2963         };
2964         struct rpc_message msg = {
2965                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2966                 .rpc_argp = &arg,
2967                 .rpc_resp = &res,
2968         };
2969         int status = -ENOMEM;
2970
2971         res.fattr = nfs_alloc_fattr();
2972         res.dir_attr = nfs_alloc_fattr();
2973         if (res.fattr == NULL || res.dir_attr == NULL)
2974                 goto out;
2975
2976         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2977         if (!status) {
2978                 update_changeattr(dir, &res.cinfo);
2979                 nfs_post_op_update_inode(dir, res.dir_attr);
2980                 nfs_post_op_update_inode(inode, res.fattr);
2981         }
2982 out:
2983         nfs_free_fattr(res.dir_attr);
2984         nfs_free_fattr(res.fattr);
2985         return status;
2986 }
2987
2988 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2989 {
2990         struct nfs4_exception exception = { };
2991         int err;
2992         do {
2993                 err = nfs4_handle_exception(NFS_SERVER(inode),
2994                                 _nfs4_proc_link(inode, dir, name),
2995                                 &exception);
2996         } while (exception.retry);
2997         return err;
2998 }
2999
3000 struct nfs4_createdata {
3001         struct rpc_message msg;
3002         struct nfs4_create_arg arg;
3003         struct nfs4_create_res res;
3004         struct nfs_fh fh;
3005         struct nfs_fattr fattr;
3006         struct nfs_fattr dir_fattr;
3007 };
3008
3009 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3010                 struct qstr *name, struct iattr *sattr, u32 ftype)
3011 {
3012         struct nfs4_createdata *data;
3013
3014         data = kzalloc(sizeof(*data), GFP_KERNEL);
3015         if (data != NULL) {
3016                 struct nfs_server *server = NFS_SERVER(dir);
3017
3018                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3019                 data->msg.rpc_argp = &data->arg;
3020                 data->msg.rpc_resp = &data->res;
3021                 data->arg.dir_fh = NFS_FH(dir);
3022                 data->arg.server = server;
3023                 data->arg.name = name;
3024                 data->arg.attrs = sattr;
3025                 data->arg.ftype = ftype;
3026                 data->arg.bitmask = server->attr_bitmask;
3027                 data->res.server = server;
3028                 data->res.fh = &data->fh;
3029                 data->res.fattr = &data->fattr;
3030                 data->res.dir_fattr = &data->dir_fattr;
3031                 nfs_fattr_init(data->res.fattr);
3032                 nfs_fattr_init(data->res.dir_fattr);
3033         }
3034         return data;
3035 }
3036
3037 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3038 {
3039         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3040                                     &data->arg.seq_args, &data->res.seq_res, 1);
3041         if (status == 0) {
3042                 update_changeattr(dir, &data->res.dir_cinfo);
3043                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
3044                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3045         }
3046         return status;
3047 }
3048
3049 static void nfs4_free_createdata(struct nfs4_createdata *data)
3050 {
3051         kfree(data);
3052 }
3053
3054 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3055                 struct page *page, unsigned int len, struct iattr *sattr)
3056 {
3057         struct nfs4_createdata *data;
3058         int status = -ENAMETOOLONG;
3059
3060         if (len > NFS4_MAXPATHLEN)
3061                 goto out;
3062
3063         status = -ENOMEM;
3064         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3065         if (data == NULL)
3066                 goto out;
3067
3068         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3069         data->arg.u.symlink.pages = &page;
3070         data->arg.u.symlink.len = len;
3071         
3072         status = nfs4_do_create(dir, dentry, data);
3073
3074         nfs4_free_createdata(data);
3075 out:
3076         return status;
3077 }
3078
3079 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3080                 struct page *page, unsigned int len, struct iattr *sattr)
3081 {
3082         struct nfs4_exception exception = { };
3083         int err;
3084         do {
3085                 err = nfs4_handle_exception(NFS_SERVER(dir),
3086                                 _nfs4_proc_symlink(dir, dentry, page,
3087                                                         len, sattr),
3088                                 &exception);
3089         } while (exception.retry);
3090         return err;
3091 }
3092
3093 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3094                 struct iattr *sattr)
3095 {
3096         struct nfs4_createdata *data;
3097         int status = -ENOMEM;
3098
3099         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3100         if (data == NULL)
3101                 goto out;
3102
3103         status = nfs4_do_create(dir, dentry, data);
3104
3105         nfs4_free_createdata(data);
3106 out:
3107         return status;
3108 }
3109
3110 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3111                 struct iattr *sattr)
3112 {
3113         struct nfs4_exception exception = { };
3114         int err;
3115
3116         sattr->ia_mode &= ~current_umask();
3117         do {
3118                 err = nfs4_handle_exception(NFS_SERVER(dir),
3119                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3120                                 &exception);
3121         } while (exception.retry);
3122         return err;
3123 }
3124
3125 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3126                 u64 cookie, struct page **pages, unsigned int count, int plus)
3127 {
3128         struct inode            *dir = dentry->d_inode;
3129         struct nfs4_readdir_arg args = {
3130                 .fh = NFS_FH(dir),
3131                 .pages = pages,
3132                 .pgbase = 0,
3133                 .count = count,
3134                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3135                 .plus = plus,
3136         };
3137         struct nfs4_readdir_res res;
3138         struct rpc_message msg = {
3139                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3140                 .rpc_argp = &args,
3141                 .rpc_resp = &res,
3142                 .rpc_cred = cred,
3143         };
3144         int                     status;
3145
3146         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3147                         dentry->d_parent->d_name.name,
3148                         dentry->d_name.name,
3149                         (unsigned long long)cookie);
3150         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3151         res.pgbase = args.pgbase;
3152         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3153         if (status >= 0) {
3154                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3155                 status += args.pgbase;
3156         }
3157
3158         nfs_invalidate_atime(dir);
3159
3160         dprintk("%s: returns %d\n", __func__, status);
3161         return status;
3162 }
3163
3164 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3165                 u64 cookie, struct page **pages, unsigned int count, int plus)
3166 {
3167         struct nfs4_exception exception = { };
3168         int err;
3169         do {
3170                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3171                                 _nfs4_proc_readdir(dentry, cred, cookie,
3172                                         pages, count, plus),
3173                                 &exception);
3174         } while (exception.retry);
3175         return err;
3176 }
3177
3178 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3179                 struct iattr *sattr, dev_t rdev)
3180 {
3181         struct nfs4_createdata *data;
3182         int mode = sattr->ia_mode;
3183         int status = -ENOMEM;
3184
3185         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3186         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3187
3188         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3189         if (data == NULL)
3190                 goto out;
3191
3192         if (S_ISFIFO(mode))
3193                 data->arg.ftype = NF4FIFO;
3194         else if (S_ISBLK(mode)) {
3195                 data->arg.ftype = NF4BLK;
3196                 data->arg.u.device.specdata1 = MAJOR(rdev);
3197                 data->arg.u.device.specdata2 = MINOR(rdev);
3198         }
3199         else if (S_ISCHR(mode)) {
3200                 data->arg.ftype = NF4CHR;
3201                 data->arg.u.device.specdata1 = MAJOR(rdev);
3202                 data->arg.u.device.specdata2 = MINOR(rdev);
3203         }
3204         
3205         status = nfs4_do_create(dir, dentry, data);
3206
3207         nfs4_free_createdata(data);
3208 out:
3209         return status;
3210 }
3211
3212 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3213                 struct iattr *sattr, dev_t rdev)
3214 {
3215         struct nfs4_exception exception = { };
3216         int err;
3217
3218         sattr->ia_mode &= ~current_umask();
3219         do {
3220                 err = nfs4_handle_exception(NFS_SERVER(dir),
3221                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3222                                 &exception);
3223         } while (exception.retry);
3224         return err;
3225 }
3226
3227 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3228                  struct nfs_fsstat *fsstat)
3229 {
3230         struct nfs4_statfs_arg args = {
3231                 .fh = fhandle,
3232                 .bitmask = server->attr_bitmask,
3233         };
3234         struct nfs4_statfs_res res = {
3235                 .fsstat = fsstat,
3236         };
3237         struct rpc_message msg = {
3238                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3239                 .rpc_argp = &args,
3240                 .rpc_resp = &res,
3241         };
3242
3243         nfs_fattr_init(fsstat->fattr);
3244         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3245 }
3246
3247 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3248 {
3249         struct nfs4_exception exception = { };
3250         int err;
3251         do {
3252                 err = nfs4_handle_exception(server,
3253                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3254                                 &exception);
3255         } while (exception.retry);
3256         return err;
3257 }
3258
3259 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3260                 struct nfs_fsinfo *fsinfo)
3261 {
3262         struct nfs4_fsinfo_arg args = {
3263                 .fh = fhandle,
3264                 .bitmask = server->attr_bitmask,
3265         };
3266         struct nfs4_fsinfo_res res = {
3267                 .fsinfo = fsinfo,
3268         };
3269         struct rpc_message msg = {
3270                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3271                 .rpc_argp = &args,
3272                 .rpc_resp = &res,
3273         };
3274
3275         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3276 }
3277
3278 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3279 {
3280         struct nfs4_exception exception = { };
3281         int err;
3282
3283         do {
3284                 err = nfs4_handle_exception(server,
3285                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3286                                 &exception);
3287         } while (exception.retry);
3288         return err;
3289 }
3290
3291 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3292 {
3293         nfs_fattr_init(fsinfo->fattr);
3294         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3295 }
3296
3297 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3298                 struct nfs_pathconf *pathconf)
3299 {
3300         struct nfs4_pathconf_arg args = {
3301                 .fh = fhandle,
3302                 .bitmask = server->attr_bitmask,
3303         };
3304         struct nfs4_pathconf_res res = {
3305                 .pathconf = pathconf,
3306         };
3307         struct rpc_message msg = {
3308                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3309                 .rpc_argp = &args,
3310                 .rpc_resp = &res,
3311         };
3312
3313         /* None of the pathconf attributes are mandatory to implement */
3314         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3315                 memset(pathconf, 0, sizeof(*pathconf));
3316                 return 0;
3317         }
3318
3319         nfs_fattr_init(pathconf->fattr);
3320         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3321 }
3322
3323 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3324                 struct nfs_pathconf *pathconf)
3325 {
3326         struct nfs4_exception exception = { };
3327         int err;
3328
3329         do {
3330                 err = nfs4_handle_exception(server,
3331                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3332                                 &exception);
3333         } while (exception.retry);
3334         return err;
3335 }
3336
3337 void __nfs4_read_done_cb(struct nfs_read_data *data)
3338 {
3339         nfs_invalidate_atime(data->inode);
3340 }
3341
3342 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3343 {
3344         struct nfs_server *server = NFS_SERVER(data->inode);
3345
3346         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3347                 rpc_restart_call_prepare(task);
3348                 return -EAGAIN;
3349         }
3350
3351         __nfs4_read_done_cb(data);
3352         if (task->tk_status > 0)
3353                 renew_lease(server, data->timestamp);
3354         return 0;
3355 }
3356
3357 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3358 {
3359
3360         dprintk("--> %s\n", __func__);
3361
3362         if (!nfs4_sequence_done(task, &data->res.seq_res))
3363                 return -EAGAIN;
3364
3365         return data->read_done_cb ? data->read_done_cb(task, data) :
3366                                     nfs4_read_done_cb(task, data);
3367 }
3368
3369 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3370 {
3371         data->timestamp   = jiffies;
3372         data->read_done_cb = nfs4_read_done_cb;
3373         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3374         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3375 }
3376
3377 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3378 {
3379         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3380                                 &data->args.seq_args,
3381                                 &data->res.seq_res,
3382                                 task))
3383                 return;
3384         rpc_call_start(task);
3385 }
3386
3387 /* Reset the the nfs_read_data to send the read to the MDS. */
3388 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3389 {
3390         dprintk("%s Reset task for i/o through\n", __func__);
3391         put_lseg(data->lseg);
3392         data->lseg = NULL;
3393         /* offsets will differ in the dense stripe case */
3394         data->args.offset = data->mds_offset;
3395         data->ds_clp = NULL;
3396         data->args.fh     = NFS_FH(data->inode);
3397         data->read_done_cb = nfs4_read_done_cb;
3398         task->tk_ops = data->mds_ops;
3399         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3400 }
3401 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3402
3403 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3404 {
3405         struct inode *inode = data->inode;
3406         
3407         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3408                 rpc_restart_call_prepare(task);
3409                 return -EAGAIN;
3410         }
3411         if (task->tk_status >= 0) {
3412                 renew_lease(NFS_SERVER(inode), data->timestamp);
3413                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3414         }
3415         return 0;
3416 }
3417
3418 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3419 {
3420         if (!nfs4_sequence_done(task, &data->res.seq_res))
3421                 return -EAGAIN;
3422         return data->write_done_cb ? data->write_done_cb(task, data) :
3423                 nfs4_write_done_cb(task, data);
3424 }
3425
3426 /* Reset the the nfs_write_data to send the write to the MDS. */
3427 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3428 {
3429         dprintk("%s Reset task for i/o through\n", __func__);
3430         put_lseg(data->lseg);
3431         data->lseg          = NULL;
3432         data->ds_clp        = NULL;
3433         data->write_done_cb = nfs4_write_done_cb;
3434         data->args.fh       = NFS_FH(data->inode);
3435         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3436         data->args.offset   = data->mds_offset;
3437         data->res.fattr     = &data->fattr;
3438         task->tk_ops        = data->mds_ops;
3439         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3440 }
3441 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3442
3443 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3444 {
3445         struct nfs_server *server = NFS_SERVER(data->inode);
3446
3447         if (data->lseg) {
3448                 data->args.bitmask = NULL;
3449                 data->res.fattr = NULL;
3450         } else
3451                 data->args.bitmask = server->cache_consistency_bitmask;
3452         if (!data->write_done_cb)
3453                 data->write_done_cb = nfs4_write_done_cb;
3454         data->res.server = server;
3455         data->timestamp   = jiffies;
3456
3457         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3458         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3459 }
3460
3461 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3462 {
3463         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3464                                 &data->args.seq_args,
3465                                 &data->res.seq_res,
3466                                 task))
3467                 return;
3468         rpc_call_start(task);
3469 }
3470
3471 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3472 {
3473         struct inode *inode = data->inode;
3474
3475         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3476                 rpc_restart_call_prepare(task);
3477                 return -EAGAIN;
3478         }
3479         nfs_refresh_inode(inode, data->res.fattr);
3480         return 0;
3481 }
3482
3483 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3484 {
3485         if (!nfs4_sequence_done(task, &data->res.seq_res))
3486                 return -EAGAIN;
3487         return data->write_done_cb(task, data);
3488 }
3489
3490 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3491 {
3492         struct nfs_server *server = NFS_SERVER(data->inode);
3493
3494         if (data->lseg) {
3495                 data->args.bitmask = NULL;
3496                 data->res.fattr = NULL;
3497         } else
3498                 data->args.bitmask = server->cache_consistency_bitmask;
3499         if (!data->write_done_cb)
3500                 data->write_done_cb = nfs4_commit_done_cb;
3501         data->res.server = server;
3502         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3503         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3504 }
3505
3506 struct nfs4_renewdata {
3507         struct nfs_client       *client;
3508         unsigned long           timestamp;
3509 };
3510
3511 /*
3512  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3513  * standalone procedure for queueing an asynchronous RENEW.
3514  */
3515 static void nfs4_renew_release(void *calldata)
3516 {
3517         struct nfs4_renewdata *data = calldata;
3518         struct nfs_client *clp = data->client;
3519
3520         if (atomic_read(&clp->cl_count) > 1)
3521                 nfs4_schedule_state_renewal(clp);
3522         nfs_put_client(clp);
3523         kfree(data);
3524 }
3525
3526 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3527 {
3528         struct nfs4_renewdata *data = calldata;
3529         struct nfs_client *clp = data->client;
3530         unsigned long timestamp = data->timestamp;
3531
3532         if (task->tk_status < 0) {
3533                 /* Unless we're shutting down, schedule state recovery! */
3534                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3535                         return;
3536                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3537                         nfs4_schedule_lease_recovery(clp);
3538                         return;
3539                 }
3540                 nfs4_schedule_path_down_recovery(clp);
3541         }
3542         do_renew_lease(clp, timestamp);
3543 }
3544
3545 static const struct rpc_call_ops nfs4_renew_ops = {
3546         .rpc_call_done = nfs4_renew_done,
3547         .rpc_release = nfs4_renew_release,
3548 };
3549
3550 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3551 {
3552         struct rpc_message msg = {
3553                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3554                 .rpc_argp       = clp,
3555                 .rpc_cred       = cred,
3556         };
3557         struct nfs4_renewdata *data;
3558
3559         if (renew_flags == 0)
3560                 return 0;
3561         if (!atomic_inc_not_zero(&clp->cl_count))
3562                 return -EIO;
3563         data = kmalloc(sizeof(*data), GFP_NOFS);
3564         if (data == NULL)
3565                 return -ENOMEM;
3566         data->client = clp;
3567         data->timestamp = jiffies;
3568         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3569                         &nfs4_renew_ops, data);
3570 }
3571
3572 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3573 {
3574         struct rpc_message msg = {
3575                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3576                 .rpc_argp       = clp,
3577                 .rpc_cred       = cred,
3578         };
3579         unsigned long now = jiffies;
3580         int status;
3581
3582         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3583         if (status < 0)
3584                 return status;
3585         do_renew_lease(clp, now);
3586         return 0;
3587 }
3588
3589 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3590 {
3591         return (server->caps & NFS_CAP_ACLS)
3592                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3593                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3594 }
3595
3596 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3597  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3598  * the stack.
3599  */
3600 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3601
3602 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3603                 struct page **pages, unsigned int *pgbase)
3604 {
3605         struct page *newpage, **spages;
3606         int rc = 0;
3607         size_t len;
3608         spages = pages;
3609
3610         do {
3611                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3612                 newpage = alloc_page(GFP_KERNEL);
3613
3614                 if (newpage == NULL)
3615                         goto unwind;
3616                 memcpy(page_address(newpage), buf, len);
3617                 buf += len;
3618                 buflen -= len;
3619                 *pages++ = newpage;
3620                 rc++;
3621         } while (buflen != 0);
3622
3623         return rc;
3624
3625 unwind:
3626         for(; rc > 0; rc--)
3627                 __free_page(spages[rc-1]);
3628         return -ENOMEM;
3629 }
3630
3631 struct nfs4_cached_acl {
3632         int cached;
3633         size_t len;
3634         char data[0];
3635 };
3636
3637 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3638 {
3639         struct nfs_inode *nfsi = NFS_I(inode);
3640
3641         spin_lock(&inode->i_lock);
3642         kfree(nfsi->nfs4_acl);
3643         nfsi->nfs4_acl = acl;
3644         spin_unlock(&inode->i_lock);
3645 }
3646
3647 static void nfs4_zap_acl_attr(struct inode *inode)
3648 {
3649         nfs4_set_cached_acl(inode, NULL);
3650 }
3651
3652 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3653 {
3654         struct nfs_inode *nfsi = NFS_I(inode);
3655         struct nfs4_cached_acl *acl;
3656         int ret = -ENOENT;
3657
3658         spin_lock(&inode->i_lock);
3659         acl = nfsi->nfs4_acl;
3660         if (acl == NULL)
3661                 goto out;
3662         if (buf == NULL) /* user is just asking for length */
3663                 goto out_len;
3664         if (acl->cached == 0)
3665                 goto out;
3666         ret = -ERANGE; /* see getxattr(2) man page */
3667         if (acl->len > buflen)
3668                 goto out;
3669         memcpy(buf, acl->data, acl->len);
3670 out_len:
3671         ret = acl->len;
3672 out:
3673         spin_unlock(&inode->i_lock);
3674         return ret;
3675 }
3676
3677 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3678 {
3679         struct nfs4_cached_acl *acl;
3680
3681         if (pages && acl_len <= PAGE_SIZE) {
3682                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3683                 if (acl == NULL)
3684                         goto out;
3685                 acl->cached = 1;
3686                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3687         } else {
3688                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3689                 if (acl == NULL)
3690                         goto out;
3691                 acl->cached = 0;
3692         }
3693         acl->len = acl_len;
3694 out:
3695         nfs4_set_cached_acl(inode, acl);
3696 }
3697
3698 /*
3699  * The getxattr API returns the required buffer length when called with a
3700  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3701  * the required buf.  On a NULL buf, we send a page of data to the server
3702  * guessing that the ACL request can be serviced by a page. If so, we cache
3703  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3704  * the cache. If not so, we throw away the page, and cache the required
3705  * length. The next getxattr call will then produce another round trip to
3706  * the server, this time with the input buf of the required size.
3707  */
3708 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3709 {
3710         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3711         struct nfs_getaclargs args = {
3712                 .fh = NFS_FH(inode),
3713                 .acl_pages = pages,
3714                 .acl_len = buflen,
3715         };
3716         struct nfs_getaclres res = {
3717                 .acl_len = buflen,
3718         };
3719         struct rpc_message msg = {
3720                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3721                 .rpc_argp = &args,
3722                 .rpc_resp = &res,
3723         };
3724         int ret = -ENOMEM, npages, i, acl_len = 0;
3725
3726         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3727         /* As long as we're doing a round trip to the server anyway,
3728          * let's be prepared for a page of acl data. */
3729         if (npages == 0)
3730                 npages = 1;
3731
3732         /* Add an extra page to handle the bitmap returned */
3733         npages++;
3734
3735         for (i = 0; i < npages; i++) {
3736                 pages[i] = alloc_page(GFP_KERNEL);
3737                 if (!pages[i])
3738                         goto out_free;
3739         }
3740
3741         /* for decoding across pages */
3742         res.acl_scratch = alloc_page(GFP_KERNEL);
3743         if (!res.acl_scratch)
3744                 goto out_free;
3745
3746         args.acl_len = npages * PAGE_SIZE;
3747         args.acl_pgbase = 0;
3748
3749         /* Let decode_getfacl know not to fail if the ACL data is larger than
3750          * the page we send as a guess */
3751         if (buf == NULL)
3752                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3753
3754         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3755                 __func__, buf, buflen, npages, args.acl_len);
3756         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3757                              &msg, &args.seq_args, &res.seq_res, 0);
3758         if (ret)
3759                 goto out_free;
3760
3761         acl_len = res.acl_len - res.acl_data_offset;
3762         if (acl_len > args.acl_len)
3763                 nfs4_write_cached_acl(inode, NULL, 0, acl_len);
3764         else
3765                 nfs4_write_cached_acl(inode, pages, res.acl_data_offset,
3766                                       acl_len);
3767         if (buf) {
3768                 ret = -ERANGE;
3769                 if (acl_len > buflen)
3770                         goto out_free;
3771                 _copy_from_pages(buf, pages, res.acl_data_offset,
3772                                 acl_len);
3773         }
3774         ret = acl_len;
3775 out_free:
3776         for (i = 0; i < npages; i++)
3777                 if (pages[i])
3778                         __free_page(pages[i]);
3779         if (res.acl_scratch)
3780                 __free_page(res.acl_scratch);
3781         return ret;
3782 }
3783
3784 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3785 {
3786         struct nfs4_exception exception = { };
3787         ssize_t ret;
3788         do {
3789                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3790                 if (ret >= 0)
3791                         break;
3792                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3793         } while (exception.retry);
3794         return ret;
3795 }
3796
3797 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3798 {
3799         struct nfs_server *server = NFS_SERVER(inode);
3800         int ret;
3801
3802         if (!nfs4_server_supports_acls(server))
3803                 return -EOPNOTSUPP;
3804         ret = nfs_revalidate_inode(server, inode);
3805         if (ret < 0)
3806                 return ret;
3807         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3808                 nfs_zap_acl_cache(inode);
3809         ret = nfs4_read_cached_acl(inode, buf, buflen);
3810         if (ret != -ENOENT)
3811                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3812                  * but no cached acl data, just the acl length */
3813                 return ret;
3814         return nfs4_get_acl_uncached(inode, buf, buflen);
3815 }
3816
3817 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3818 {
3819         struct nfs_server *server = NFS_SERVER(inode);
3820         struct page *pages[NFS4ACL_MAXPAGES];
3821         struct nfs_setaclargs arg = {
3822                 .fh             = NFS_FH(inode),
3823                 .acl_pages      = pages,
3824                 .acl_len        = buflen,
3825         };
3826         struct nfs_setaclres res;
3827         struct rpc_message msg = {
3828                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3829                 .rpc_argp       = &arg,
3830                 .rpc_resp       = &res,
3831         };
3832         int ret, i;
3833
3834         if (!nfs4_server_supports_acls(server))
3835                 return -EOPNOTSUPP;
3836         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3837         if (i < 0)
3838                 return i;
3839         nfs_inode_return_delegation(inode);
3840         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3841
3842         /*
3843          * Free each page after tx, so the only ref left is
3844          * held by the network stack
3845          */
3846         for (; i > 0; i--)
3847                 put_page(pages[i-1]);
3848
3849         /*
3850          * Acl update can result in inode attribute update.
3851          * so mark the attribute cache invalid.
3852          */
3853         spin_lock(&inode->i_lock);
3854         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3855         spin_unlock(&inode->i_lock);
3856         nfs_access_zap_cache(inode);
3857         nfs_zap_acl_cache(inode);
3858         return ret;
3859 }
3860
3861 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3862 {
3863         struct nfs4_exception exception = { };
3864         int err;
3865         do {
3866                 err = nfs4_handle_exception(NFS_SERVER(inode),
3867                                 __nfs4_proc_set_acl(inode, buf, buflen),
3868                                 &exception);
3869         } while (exception.retry);
3870         return err;
3871 }
3872
3873 static int
3874 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3875 {
3876         struct nfs_client *clp = server->nfs_client;
3877
3878         if (task->tk_status >= 0)
3879                 return 0;
3880         switch(task->tk_status) {
3881                 case -NFS4ERR_DELEG_REVOKED:
3882                 case -NFS4ERR_ADMIN_REVOKED:
3883                 case -NFS4ERR_BAD_STATEID:
3884                         if (state == NULL)
3885                                 break;
3886                         nfs_remove_bad_delegation(state->inode);
3887                 case -NFS4ERR_OPENMODE:
3888                         if (state == NULL)
3889                                 break;
3890                         nfs4_schedule_stateid_recovery(server, state);
3891                         goto wait_on_recovery;
3892                 case -NFS4ERR_EXPIRED:
3893                         if (state != NULL)
3894                                 nfs4_schedule_stateid_recovery(server, state);
3895                 case -NFS4ERR_STALE_STATEID:
3896                 case -NFS4ERR_STALE_CLIENTID:
3897                         nfs4_schedule_lease_recovery(clp);
3898                         goto wait_on_recovery;
3899 #if defined(CONFIG_NFS_V4_1)
3900                 case -NFS4ERR_BADSESSION:
3901                 case -NFS4ERR_BADSLOT:
3902                 case -NFS4ERR_BAD_HIGH_SLOT:
3903                 case -NFS4ERR_DEADSESSION:
3904                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3905                 case -NFS4ERR_SEQ_FALSE_RETRY:
3906                 case -NFS4ERR_SEQ_MISORDERED:
3907                         dprintk("%s ERROR %d, Reset session\n", __func__,
3908                                 task->tk_status);
3909                         nfs4_schedule_session_recovery(clp->cl_session);
3910                         task->tk_status = 0;
3911                         return -EAGAIN;
3912 #endif /* CONFIG_NFS_V4_1 */
3913                 case -NFS4ERR_DELAY:
3914                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3915                 case -NFS4ERR_GRACE:
3916                 case -EKEYEXPIRED:
3917                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3918                         task->tk_status = 0;
3919                         return -EAGAIN;
3920                 case -NFS4ERR_RETRY_UNCACHED_REP:
3921                 case -NFS4ERR_OLD_STATEID:
3922                         task->tk_status = 0;
3923                         return -EAGAIN;
3924         }
3925         task->tk_status = nfs4_map_errors(task->tk_status);
3926         return 0;
3927 wait_on_recovery:
3928         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3929         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3930                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3931         task->tk_status = 0;
3932         return -EAGAIN;
3933 }
3934
3935 static void nfs4_construct_boot_verifier(struct nfs_client *clp,
3936                                          nfs4_verifier *bootverf)
3937 {
3938         __be32 verf[2];
3939
3940         verf[0] = htonl((u32)clp->cl_boot_time.tv_sec);
3941         verf[1] = htonl((u32)clp->cl_boot_time.tv_nsec);
3942         memcpy(bootverf->data, verf, sizeof(bootverf->data));
3943 }
3944
3945 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3946                 unsigned short port, struct rpc_cred *cred,
3947                 struct nfs4_setclientid_res *res)
3948 {
3949         nfs4_verifier sc_verifier;
3950         struct nfs4_setclientid setclientid = {
3951                 .sc_verifier = &sc_verifier,
3952                 .sc_prog = program,
3953                 .sc_cb_ident = clp->cl_cb_ident,
3954         };
3955         struct rpc_message msg = {
3956                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3957                 .rpc_argp = &setclientid,
3958                 .rpc_resp = res,
3959                 .rpc_cred = cred,
3960         };
3961         int loop = 0;
3962         int status;
3963
3964         nfs4_construct_boot_verifier(clp, &sc_verifier);
3965
3966         for(;;) {
3967                 rcu_read_lock();
3968                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3969                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3970                                 clp->cl_ipaddr,
3971                                 rpc_peeraddr2str(clp->cl_rpcclient,
3972                                                         RPC_DISPLAY_ADDR),
3973                                 rpc_peeraddr2str(clp->cl_rpcclient,
3974                                                         RPC_DISPLAY_PROTO),
3975                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3976                                 clp->cl_id_uniquifier);
3977                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3978                                 sizeof(setclientid.sc_netid),
3979                                 rpc_peeraddr2str(clp->cl_rpcclient,
3980                                                         RPC_DISPLAY_NETID));
3981                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3982                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3983                                 clp->cl_ipaddr, port >> 8, port & 255);
3984                 rcu_read_unlock();
3985
3986                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3987                 if (status != -NFS4ERR_CLID_INUSE)
3988                         break;
3989                 if (loop != 0) {
3990                         ++clp->cl_id_uniquifier;
3991                         break;
3992                 }
3993                 ++loop;
3994                 ssleep(clp->cl_lease_time / HZ + 1);
3995         }
3996         return status;
3997 }
3998
3999 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4000                 struct nfs4_setclientid_res *arg,
4001                 struct rpc_cred *cred)
4002 {
4003         struct nfs_fsinfo fsinfo;
4004         struct rpc_message msg = {
4005                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4006                 .rpc_argp = arg,
4007                 .rpc_resp = &fsinfo,
4008                 .rpc_cred = cred,
4009         };
4010         unsigned long now;
4011         int status;
4012
4013         now = jiffies;
4014         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4015         if (status == 0) {
4016                 spin_lock(&clp->cl_lock);
4017                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4018                 clp->cl_last_renewal = now;
4019                 spin_unlock(&clp->cl_lock);
4020         }
4021         return status;
4022 }
4023
4024 struct nfs4_delegreturndata {
4025         struct nfs4_delegreturnargs args;
4026         struct nfs4_delegreturnres res;
4027         struct nfs_fh fh;
4028         nfs4_stateid stateid;
4029         unsigned long timestamp;
4030         struct nfs_fattr fattr;
4031         int rpc_status;
4032 };
4033
4034 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4035 {
4036         struct nfs4_delegreturndata *data = calldata;
4037
4038         if (!nfs4_sequence_done(task, &data->res.seq_res))
4039                 return;
4040
4041         switch (task->tk_status) {
4042         case -NFS4ERR_STALE_STATEID:
4043         case -NFS4ERR_EXPIRED:
4044         case 0:
4045                 renew_lease(data->res.server, data->timestamp);
4046                 break;
4047         default:
4048                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4049                                 -EAGAIN) {
4050                         rpc_restart_call_prepare(task);
4051                         return;
4052                 }
4053         }
4054         data->rpc_status = task->tk_status;
4055 }
4056
4057 static void nfs4_delegreturn_release(void *calldata)
4058 {
4059         kfree(calldata);
4060 }
4061
4062 #if defined(CONFIG_NFS_V4_1)
4063 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4064 {
4065         struct nfs4_delegreturndata *d_data;
4066
4067         d_data = (struct nfs4_delegreturndata *)data;
4068
4069         if (nfs4_setup_sequence(d_data->res.server,
4070                                 &d_data->args.seq_args,
4071                                 &d_data->res.seq_res, task))
4072                 return;
4073         rpc_call_start(task);
4074 }
4075 #endif /* CONFIG_NFS_V4_1 */
4076
4077 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4078 #if defined(CONFIG_NFS_V4_1)
4079         .rpc_call_prepare = nfs4_delegreturn_prepare,
4080 #endif /* CONFIG_NFS_V4_1 */
4081         .rpc_call_done = nfs4_delegreturn_done,
4082         .rpc_release = nfs4_delegreturn_release,
4083 };
4084
4085 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4086 {
4087         struct nfs4_delegreturndata *data;
4088         struct nfs_server *server = NFS_SERVER(inode);
4089         struct rpc_task *task;
4090         struct rpc_message msg = {
4091                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4092                 .rpc_cred = cred,
4093         };
4094         struct rpc_task_setup task_setup_data = {
4095                 .rpc_client = server->client,
4096                 .rpc_message = &msg,
4097                 .callback_ops = &nfs4_delegreturn_ops,
4098                 .flags = RPC_TASK_ASYNC,
4099         };
4100         int status = 0;
4101
4102         data = kzalloc(sizeof(*data), GFP_NOFS);
4103         if (data == NULL)
4104                 return -ENOMEM;
4105         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4106         data->args.fhandle = &data->fh;
4107         data->args.stateid = &data->stateid;
4108         data->args.bitmask = server->attr_bitmask;
4109         nfs_copy_fh(&data->fh, NFS_FH(inode));
4110         nfs4_stateid_copy(&data->stateid, stateid);
4111         data->res.fattr = &data->fattr;
4112         data->res.server = server;
4113         nfs_fattr_init(data->res.fattr);
4114         data->timestamp = jiffies;
4115         data->rpc_status = 0;
4116
4117         task_setup_data.callback_data = data;
4118         msg.rpc_argp = &data->args;
4119         msg.rpc_resp = &data->res;
4120         task = rpc_run_task(&task_setup_data);
4121         if (IS_ERR(task))
4122                 return PTR_ERR(task);
4123         if (!issync)
4124                 goto out;
4125         status = nfs4_wait_for_completion_rpc_task(task);
4126         if (status != 0)
4127                 goto out;
4128         status = data->rpc_status;
4129         if (status != 0)
4130                 goto out;
4131         nfs_refresh_inode(inode, &data->fattr);
4132 out:
4133         rpc_put_task(task);
4134         return status;
4135 }
4136
4137 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4138 {
4139         struct nfs_server *server = NFS_SERVER(inode);
4140         struct nfs4_exception exception = { };
4141         int err;
4142         do {
4143                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4144                 switch (err) {
4145                         case -NFS4ERR_STALE_STATEID:
4146                         case -NFS4ERR_EXPIRED:
4147                         case 0:
4148                                 return 0;
4149                 }
4150                 err = nfs4_handle_exception(server, err, &exception);
4151         } while (exception.retry);
4152         return err;
4153 }
4154
4155 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4156 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4157
4158 /* 
4159  * sleep, with exponential backoff, and retry the LOCK operation. 
4160  */
4161 static unsigned long
4162 nfs4_set_lock_task_retry(unsigned long timeout)
4163 {
4164         freezable_schedule_timeout_killable(timeout);
4165         timeout <<= 1;
4166         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4167                 return NFS4_LOCK_MAXTIMEOUT;
4168         return timeout;
4169 }
4170
4171 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4172 {
4173         struct inode *inode = state->inode;
4174         struct nfs_server *server = NFS_SERVER(inode);
4175         struct nfs_client *clp = server->nfs_client;
4176         struct nfs_lockt_args arg = {
4177                 .fh = NFS_FH(inode),
4178                 .fl = request,
4179         };
4180         struct nfs_lockt_res res = {
4181                 .denied = request,
4182         };
4183         struct rpc_message msg = {
4184                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4185                 .rpc_argp       = &arg,
4186                 .rpc_resp       = &res,
4187                 .rpc_cred       = state->owner->so_cred,
4188         };
4189         struct nfs4_lock_state *lsp;
4190         int status;
4191
4192         arg.lock_owner.clientid = clp->cl_clientid;
4193         status = nfs4_set_lock_state(state, request);
4194         if (status != 0)
4195                 goto out;
4196         lsp = request->fl_u.nfs4_fl.owner;
4197         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4198         arg.lock_owner.s_dev = server->s_dev;
4199         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4200         switch (status) {
4201                 case 0:
4202                         request->fl_type = F_UNLCK;
4203                         break;
4204                 case -NFS4ERR_DENIED:
4205                         status = 0;
4206         }
4207         request->fl_ops->fl_release_private(request);
4208 out:
4209         return status;
4210 }
4211
4212 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4213 {
4214         struct nfs4_exception exception = { };
4215         int err;
4216
4217         do {
4218                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4219                                 _nfs4_proc_getlk(state, cmd, request),
4220                                 &exception);
4221         } while (exception.retry);
4222         return err;
4223 }
4224
4225 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4226 {
4227         int res = 0;
4228         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4229                 case FL_POSIX:
4230                         res = posix_lock_file_wait(file, fl);
4231                         break;
4232                 case FL_FLOCK:
4233                         res = flock_lock_file_wait(file, fl);
4234                         break;
4235                 default:
4236                         BUG();
4237         }
4238         return res;
4239 }
4240
4241 struct nfs4_unlockdata {
4242         struct nfs_locku_args arg;
4243         struct nfs_locku_res res;
4244         struct nfs4_lock_state *lsp;
4245         struct nfs_open_context *ctx;
4246         struct file_lock fl;
4247         const struct nfs_server *server;
4248         unsigned long timestamp;
4249 };
4250
4251 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4252                 struct nfs_open_context *ctx,
4253                 struct nfs4_lock_state *lsp,
4254                 struct nfs_seqid *seqid)
4255 {
4256         struct nfs4_unlockdata *p;
4257         struct inode *inode = lsp->ls_state->inode;
4258
4259         p = kzalloc(sizeof(*p), GFP_NOFS);
4260         if (p == NULL)
4261                 return NULL;
4262         p->arg.fh = NFS_FH(inode);
4263         p->arg.fl = &p->fl;
4264         p->arg.seqid = seqid;
4265         p->res.seqid = seqid;
4266         p->arg.stateid = &lsp->ls_stateid;
4267         p->lsp = lsp;
4268         atomic_inc(&lsp->ls_count);
4269         /* Ensure we don't close file until we're done freeing locks! */
4270         p->ctx = get_nfs_open_context(ctx);
4271         memcpy(&p->fl, fl, sizeof(p->fl));
4272         p->server = NFS_SERVER(inode);
4273         return p;
4274 }
4275
4276 static void nfs4_locku_release_calldata(void *data)
4277 {
4278         struct nfs4_unlockdata *calldata = data;
4279         nfs_free_seqid(calldata->arg.seqid);
4280         nfs4_put_lock_state(calldata->lsp);
4281         put_nfs_open_context(calldata->ctx);
4282         kfree(calldata);
4283 }
4284
4285 static void nfs4_locku_done(struct rpc_task *task, void *data)
4286 {
4287         struct nfs4_unlockdata *calldata = data;
4288
4289         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4290                 return;
4291         switch (task->tk_status) {
4292                 case 0:
4293                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4294                                         &calldata->res.stateid);
4295                         renew_lease(calldata->server, calldata->timestamp);
4296                         break;
4297                 case -NFS4ERR_BAD_STATEID:
4298                 case -NFS4ERR_OLD_STATEID:
4299                 case -NFS4ERR_STALE_STATEID:
4300                 case -NFS4ERR_EXPIRED:
4301                         break;
4302                 default:
4303                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4304                                 rpc_restart_call_prepare(task);
4305         }
4306 }
4307
4308 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4309 {
4310         struct nfs4_unlockdata *calldata = data;
4311
4312         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4313                 return;
4314         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4315                 /* Note: exit _without_ running nfs4_locku_done */
4316                 task->tk_action = NULL;
4317                 return;
4318         }
4319         calldata->timestamp = jiffies;
4320         if (nfs4_setup_sequence(calldata->server,
4321                                 &calldata->arg.seq_args,
4322                                 &calldata->res.seq_res, task))
4323                 return;
4324         rpc_call_start(task);
4325 }
4326
4327 static const struct rpc_call_ops nfs4_locku_ops = {
4328         .rpc_call_prepare = nfs4_locku_prepare,
4329         .rpc_call_done = nfs4_locku_done,
4330         .rpc_release = nfs4_locku_release_calldata,
4331 };
4332
4333 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4334                 struct nfs_open_context *ctx,
4335                 struct nfs4_lock_state *lsp,
4336                 struct nfs_seqid *seqid)
4337 {
4338         struct nfs4_unlockdata *data;
4339         struct rpc_message msg = {
4340                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4341                 .rpc_cred = ctx->cred,
4342         };
4343         struct rpc_task_setup task_setup_data = {
4344                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4345                 .rpc_message = &msg,
4346                 .callback_ops = &nfs4_locku_ops,
4347                 .workqueue = nfsiod_workqueue,
4348                 .flags = RPC_TASK_ASYNC,
4349         };
4350
4351         /* Ensure this is an unlock - when canceling a lock, the
4352          * canceled lock is passed in, and it won't be an unlock.
4353          */
4354         fl->fl_type = F_UNLCK;
4355
4356         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4357         if (data == NULL) {
4358                 nfs_free_seqid(seqid);
4359                 return ERR_PTR(-ENOMEM);
4360         }
4361
4362         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4363         msg.rpc_argp = &data->arg;
4364         msg.rpc_resp = &data->res;
4365         task_setup_data.callback_data = data;
4366         return rpc_run_task(&task_setup_data);
4367 }
4368
4369 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4370 {
4371         struct nfs_inode *nfsi = NFS_I(state->inode);
4372         struct nfs_seqid *seqid;
4373         struct nfs4_lock_state *lsp;
4374         struct rpc_task *task;
4375         int status = 0;
4376         unsigned char fl_flags = request->fl_flags;
4377
4378         status = nfs4_set_lock_state(state, request);
4379         /* Unlock _before_ we do the RPC call */
4380         request->fl_flags |= FL_EXISTS;
4381         down_read(&nfsi->rwsem);
4382         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4383                 up_read(&nfsi->rwsem);
4384                 goto out;
4385         }
4386         up_read(&nfsi->rwsem);
4387         if (status != 0)
4388                 goto out;
4389         /* Is this a delegated lock? */
4390         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4391                 goto out;
4392         lsp = request->fl_u.nfs4_fl.owner;
4393         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4394         status = -ENOMEM;
4395         if (seqid == NULL)
4396                 goto out;
4397         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4398         status = PTR_ERR(task);
4399         if (IS_ERR(task))
4400                 goto out;
4401         status = nfs4_wait_for_completion_rpc_task(task);
4402         rpc_put_task(task);
4403 out:
4404         request->fl_flags = fl_flags;
4405         return status;
4406 }
4407
4408 struct nfs4_lockdata {
4409         struct nfs_lock_args arg;
4410         struct nfs_lock_res res;
4411         struct nfs4_lock_state *lsp;
4412         struct nfs_open_context *ctx;
4413         struct file_lock fl;
4414         unsigned long timestamp;
4415         int rpc_status;
4416         int cancelled;
4417         struct nfs_server *server;
4418 };
4419
4420 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4421                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4422                 gfp_t gfp_mask)
4423 {
4424         struct nfs4_lockdata *p;
4425         struct inode *inode = lsp->ls_state->inode;
4426         struct nfs_server *server = NFS_SERVER(inode);
4427
4428         p = kzalloc(sizeof(*p), gfp_mask);
4429         if (p == NULL)
4430                 return NULL;
4431
4432         p->arg.fh = NFS_FH(inode);
4433         p->arg.fl = &p->fl;
4434         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4435         if (p->arg.open_seqid == NULL)
4436                 goto out_free;
4437         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4438         if (p->arg.lock_seqid == NULL)
4439                 goto out_free_seqid;
4440         p->arg.lock_stateid = &lsp->ls_stateid;
4441         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4442         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4443         p->arg.lock_owner.s_dev = server->s_dev;
4444         p->res.lock_seqid = p->arg.lock_seqid;
4445         p->lsp = lsp;
4446         p->server = server;
4447         atomic_inc(&lsp->ls_count);
4448         p->ctx = get_nfs_open_context(ctx);
4449         memcpy(&p->fl, fl, sizeof(p->fl));
4450         return p;
4451 out_free_seqid:
4452         nfs_free_seqid(p->arg.open_seqid);
4453 out_free:
4454         kfree(p);
4455         return NULL;
4456 }
4457
4458 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4459 {
4460         struct nfs4_lockdata *data = calldata;
4461         struct nfs4_state *state = data->lsp->ls_state;
4462
4463         dprintk("%s: begin!\n", __func__);
4464         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4465                 return;
4466         /* Do we need to do an open_to_lock_owner? */
4467         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4468                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4469                         return;
4470                 data->arg.open_stateid = &state->stateid;
4471                 data->arg.new_lock_owner = 1;
4472                 data->res.open_seqid = data->arg.open_seqid;
4473         } else
4474                 data->arg.new_lock_owner = 0;
4475         data->timestamp = jiffies;
4476         if (nfs4_setup_sequence(data->server,
4477                                 &data->arg.seq_args,
4478                                 &data->res.seq_res, task))
4479                 return;
4480         rpc_call_start(task);
4481         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4482 }
4483
4484 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4485 {
4486         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4487         nfs4_lock_prepare(task, calldata);
4488 }
4489
4490 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4491 {
4492         struct nfs4_lockdata *data = calldata;
4493
4494         dprintk("%s: begin!\n", __func__);
4495
4496         if (!nfs4_sequence_done(task, &data->res.seq_res))
4497                 return;
4498
4499         data->rpc_status = task->tk_status;
4500         if (data->arg.new_lock_owner != 0) {
4501                 if (data->rpc_status == 0)
4502                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4503                 else
4504                         goto out;
4505         }
4506         if (data->rpc_status == 0) {
4507                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4508                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4509                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4510         }
4511 out:
4512         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4513 }
4514
4515 static void nfs4_lock_release(void *calldata)
4516 {
4517         struct nfs4_lockdata *data = calldata;
4518
4519         dprintk("%s: begin!\n", __func__);
4520         nfs_free_seqid(data->arg.open_seqid);
4521         if (data->cancelled != 0) {
4522                 struct rpc_task *task;
4523                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4524                                 data->arg.lock_seqid);
4525                 if (!IS_ERR(task))
4526                         rpc_put_task_async(task);
4527                 dprintk("%s: cancelling lock!\n", __func__);
4528         } else
4529                 nfs_free_seqid(data->arg.lock_seqid);
4530         nfs4_put_lock_state(data->lsp);
4531         put_nfs_open_context(data->ctx);
4532         kfree(data);
4533         dprintk("%s: done!\n", __func__);
4534 }
4535
4536 static const struct rpc_call_ops nfs4_lock_ops = {
4537         .rpc_call_prepare = nfs4_lock_prepare,
4538         .rpc_call_done = nfs4_lock_done,
4539         .rpc_release = nfs4_lock_release,
4540 };
4541
4542 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4543         .rpc_call_prepare = nfs4_recover_lock_prepare,
4544         .rpc_call_done = nfs4_lock_done,
4545         .rpc_release = nfs4_lock_release,
4546 };
4547
4548 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4549 {
4550         switch (error) {
4551         case -NFS4ERR_ADMIN_REVOKED:
4552         case -NFS4ERR_BAD_STATEID:
4553                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4554                 if (new_lock_owner != 0 ||
4555                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4556                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4557                 break;
4558         case -NFS4ERR_STALE_STATEID:
4559                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4560         case -NFS4ERR_EXPIRED:
4561                 nfs4_schedule_lease_recovery(server->nfs_client);
4562         };
4563 }
4564
4565 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4566 {
4567         struct nfs4_lockdata *data;
4568         struct rpc_task *task;
4569         struct rpc_message msg = {
4570                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4571                 .rpc_cred = state->owner->so_cred,
4572         };
4573         struct rpc_task_setup task_setup_data = {
4574                 .rpc_client = NFS_CLIENT(state->inode),
4575                 .rpc_message = &msg,
4576                 .callback_ops = &nfs4_lock_ops,
4577                 .workqueue = nfsiod_workqueue,
4578                 .flags = RPC_TASK_ASYNC,
4579         };
4580         int ret;
4581
4582         dprintk("%s: begin!\n", __func__);
4583         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4584                         fl->fl_u.nfs4_fl.owner,
4585                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4586         if (data == NULL)
4587                 return -ENOMEM;
4588         if (IS_SETLKW(cmd))
4589                 data->arg.block = 1;
4590         if (recovery_type > NFS_LOCK_NEW) {
4591                 if (recovery_type == NFS_LOCK_RECLAIM)
4592                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4593                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4594         }
4595         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4596         msg.rpc_argp = &data->arg;
4597         msg.rpc_resp = &data->res;
4598         task_setup_data.callback_data = data;
4599         task = rpc_run_task(&task_setup_data);
4600         if (IS_ERR(task))
4601                 return PTR_ERR(task);
4602         ret = nfs4_wait_for_completion_rpc_task(task);
4603         if (ret == 0) {
4604                 ret = data->rpc_status;
4605                 if (ret)
4606                         nfs4_handle_setlk_error(data->server, data->lsp,
4607                                         data->arg.new_lock_owner, ret);
4608         } else
4609                 data->cancelled = 1;
4610         rpc_put_task(task);
4611         dprintk("%s: done, ret = %d!\n", __func__, ret);
4612         return ret;
4613 }
4614
4615 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4616 {
4617         struct nfs_server *server = NFS_SERVER(state->inode);
4618         struct nfs4_exception exception = {
4619                 .inode = state->inode,
4620         };
4621         int err;
4622
4623         do {
4624                 /* Cache the lock if possible... */
4625                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4626                         return 0;
4627                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4628                 if (err != -NFS4ERR_DELAY)
4629                         break;
4630                 nfs4_handle_exception(server, err, &exception);
4631         } while (exception.retry);
4632         return err;
4633 }
4634
4635 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4636 {
4637         struct nfs_server *server = NFS_SERVER(state->inode);
4638         struct nfs4_exception exception = {
4639                 .inode = state->inode,
4640         };
4641         int err;
4642
4643         err = nfs4_set_lock_state(state, request);
4644         if (err != 0)
4645                 return err;
4646         do {
4647                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4648                         return 0;
4649                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4650                 switch (err) {
4651                 default:
4652                         goto out;
4653                 case -NFS4ERR_GRACE:
4654                 case -NFS4ERR_DELAY:
4655                         nfs4_handle_exception(server, err, &exception);
4656                         err = 0;
4657                 }
4658         } while (exception.retry);
4659 out:
4660         return err;
4661 }
4662
4663 #if defined(CONFIG_NFS_V4_1)
4664 static int nfs41_check_expired_locks(struct nfs4_state *state)
4665 {
4666         int status, ret = NFS_OK;
4667         struct nfs4_lock_state *lsp;
4668         struct nfs_server *server = NFS_SERVER(state->inode);
4669
4670         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4671                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4672                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4673                         if (status != NFS_OK) {
4674                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4675                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4676                                 ret = status;
4677                         }
4678                 }
4679         };
4680
4681         return ret;
4682 }
4683
4684 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4685 {
4686         int status = NFS_OK;
4687
4688         if (test_bit(LK_STATE_IN_USE, &state->flags))
4689                 status = nfs41_check_expired_locks(state);
4690         if (status == NFS_OK)
4691                 return status;
4692         return nfs4_lock_expired(state, request);
4693 }
4694 #endif
4695
4696 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4697 {
4698         struct nfs_inode *nfsi = NFS_I(state->inode);
4699         unsigned char fl_flags = request->fl_flags;
4700         int status = -ENOLCK;
4701
4702         if ((fl_flags & FL_POSIX) &&
4703                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4704                 goto out;
4705         /* Is this a delegated open? */
4706         status = nfs4_set_lock_state(state, request);
4707         if (status != 0)
4708                 goto out;
4709         request->fl_flags |= FL_ACCESS;
4710         status = do_vfs_lock(request->fl_file, request);
4711         if (status < 0)
4712                 goto out;
4713         down_read(&nfsi->rwsem);
4714         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4715                 /* Yes: cache locks! */
4716                 /* ...but avoid races with delegation recall... */
4717                 request->fl_flags = fl_flags & ~FL_SLEEP;
4718                 status = do_vfs_lock(request->fl_file, request);
4719                 goto out_unlock;
4720         }
4721         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4722         if (status != 0)
4723                 goto out_unlock;
4724         /* Note: we always want to sleep here! */
4725         request->fl_flags = fl_flags | FL_SLEEP;
4726         if (do_vfs_lock(request->fl_file, request) < 0)
4727                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4728                         "manager!\n", __func__);
4729 out_unlock:
4730         up_read(&nfsi->rwsem);
4731 out:
4732         request->fl_flags = fl_flags;
4733         return status;
4734 }
4735
4736 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4737 {
4738         struct nfs4_exception exception = {
4739                 .state = state,
4740                 .inode = state->inode,
4741         };
4742         int err;
4743
4744         do {
4745                 err = _nfs4_proc_setlk(state, cmd, request);
4746                 if (err == -NFS4ERR_DENIED)
4747                         err = -EAGAIN;
4748                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4749                                 err, &exception);
4750         } while (exception.retry);
4751         return err;
4752 }
4753
4754 static int
4755 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4756 {
4757         struct nfs_open_context *ctx;
4758         struct nfs4_state *state;
4759         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4760         int status;
4761
4762         /* verify open state */
4763         ctx = nfs_file_open_context(filp);
4764         state = ctx->state;
4765
4766         if (request->fl_start < 0 || request->fl_end < 0)
4767                 return -EINVAL;
4768
4769         if (IS_GETLK(cmd)) {
4770                 if (state != NULL)
4771                         return nfs4_proc_getlk(state, F_GETLK, request);
4772                 return 0;
4773         }
4774
4775         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4776                 return -EINVAL;
4777
4778         if (request->fl_type == F_UNLCK) {
4779                 if (state != NULL)
4780                         return nfs4_proc_unlck(state, cmd, request);
4781                 return 0;
4782         }
4783
4784         if (state == NULL)
4785                 return -ENOLCK;
4786         /*
4787          * Don't rely on the VFS having checked the file open mode,
4788          * since it won't do this for flock() locks.
4789          */
4790         switch (request->fl_type & (F_RDLCK|F_WRLCK|F_UNLCK)) {
4791         case F_RDLCK:
4792                 if (!(filp->f_mode & FMODE_READ))
4793                         return -EBADF;
4794                 break;
4795         case F_WRLCK:
4796                 if (!(filp->f_mode & FMODE_WRITE))
4797                         return -EBADF;
4798         }
4799
4800         do {
4801                 status = nfs4_proc_setlk(state, cmd, request);
4802                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4803                         break;
4804                 timeout = nfs4_set_lock_task_retry(timeout);
4805                 status = -ERESTARTSYS;
4806                 if (signalled())
4807                         break;
4808         } while(status < 0);
4809         return status;
4810 }
4811
4812 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4813 {
4814         struct nfs_server *server = NFS_SERVER(state->inode);
4815         struct nfs4_exception exception = { };
4816         int err;
4817
4818         err = nfs4_set_lock_state(state, fl);
4819         if (err != 0)
4820                 goto out;
4821         do {
4822                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4823                 switch (err) {
4824                         default:
4825                                 printk(KERN_ERR "NFS: %s: unhandled error "
4826                                         "%d.\n", __func__, err);
4827                         case 0:
4828                         case -ESTALE:
4829                                 goto out;
4830                         case -NFS4ERR_EXPIRED:
4831                                 nfs4_schedule_stateid_recovery(server, state);
4832                         case -NFS4ERR_STALE_CLIENTID:
4833                         case -NFS4ERR_STALE_STATEID:
4834                                 nfs4_schedule_lease_recovery(server->nfs_client);
4835                                 goto out;
4836                         case -NFS4ERR_BADSESSION:
4837                         case -NFS4ERR_BADSLOT:
4838                         case -NFS4ERR_BAD_HIGH_SLOT:
4839                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4840                         case -NFS4ERR_DEADSESSION:
4841                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4842                                 goto out;
4843                         case -ERESTARTSYS:
4844                                 /*
4845                                  * The show must go on: exit, but mark the
4846                                  * stateid as needing recovery.
4847                                  */
4848                         case -NFS4ERR_DELEG_REVOKED:
4849                         case -NFS4ERR_ADMIN_REVOKED:
4850                         case -NFS4ERR_BAD_STATEID:
4851                         case -NFS4ERR_OPENMODE:
4852                                 nfs4_schedule_stateid_recovery(server, state);
4853                                 err = 0;
4854                                 goto out;
4855                         case -EKEYEXPIRED:
4856                                 /*
4857                                  * User RPCSEC_GSS context has expired.
4858                                  * We cannot recover this stateid now, so
4859                                  * skip it and allow recovery thread to
4860                                  * proceed.
4861                                  */
4862                                 err = 0;
4863                                 goto out;
4864                         case -ENOMEM:
4865                         case -NFS4ERR_DENIED:
4866                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4867                                 err = 0;
4868                                 goto out;
4869                         case -NFS4ERR_DELAY:
4870                                 break;
4871                 }
4872                 err = nfs4_handle_exception(server, err, &exception);
4873         } while (exception.retry);
4874 out:
4875         return err;
4876 }
4877
4878 struct nfs_release_lockowner_data {
4879         struct nfs4_lock_state *lsp;
4880         struct nfs_server *server;
4881         struct nfs_release_lockowner_args args;
4882 };
4883
4884 static void nfs4_release_lockowner_release(void *calldata)
4885 {
4886         struct nfs_release_lockowner_data *data = calldata;
4887         nfs4_free_lock_state(data->server, data->lsp);
4888         kfree(calldata);
4889 }
4890
4891 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
4892         .rpc_release = nfs4_release_lockowner_release,
4893 };
4894
4895 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
4896 {
4897         struct nfs_server *server = lsp->ls_state->owner->so_server;
4898         struct nfs_release_lockowner_data *data;
4899         struct rpc_message msg = {
4900                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4901         };
4902
4903         if (server->nfs_client->cl_mvops->minor_version != 0)
4904                 return -EINVAL;
4905         data = kmalloc(sizeof(*data), GFP_NOFS);
4906         if (!data)
4907                 return -ENOMEM;
4908         data->lsp = lsp;
4909         data->server = server;
4910         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
4911         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
4912         data->args.lock_owner.s_dev = server->s_dev;
4913         msg.rpc_argp = &data->args;
4914         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
4915         return 0;
4916 }
4917
4918 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4919
4920 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4921                                    const void *buf, size_t buflen,
4922                                    int flags, int type)
4923 {
4924         if (strcmp(key, "") != 0)
4925                 return -EINVAL;
4926
4927         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4928 }
4929
4930 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4931                                    void *buf, size_t buflen, int type)
4932 {
4933         if (strcmp(key, "") != 0)
4934                 return -EINVAL;
4935
4936         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4937 }
4938
4939 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4940                                        size_t list_len, const char *name,
4941                                        size_t name_len, int type)
4942 {
4943         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4944
4945         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4946                 return 0;
4947
4948         if (list && len <= list_len)
4949                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4950         return len;
4951 }
4952
4953 /*
4954  * nfs_fhget will use either the mounted_on_fileid or the fileid
4955  */
4956 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4957 {
4958         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4959                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4960               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4961               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
4962                 return;
4963
4964         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4965                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
4966         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4967         fattr->nlink = 2;
4968 }
4969
4970 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
4971                                    const struct qstr *name,
4972                                    struct nfs4_fs_locations *fs_locations,
4973                                    struct page *page)
4974 {
4975         struct nfs_server *server = NFS_SERVER(dir);
4976         u32 bitmask[2] = {
4977                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4978         };
4979         struct nfs4_fs_locations_arg args = {
4980                 .dir_fh = NFS_FH(dir),
4981                 .name = name,
4982                 .page = page,
4983                 .bitmask = bitmask,
4984         };
4985         struct nfs4_fs_locations_res res = {
4986                 .fs_locations = fs_locations,
4987         };
4988         struct rpc_message msg = {
4989                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4990                 .rpc_argp = &args,
4991                 .rpc_resp = &res,
4992         };
4993         int status;
4994
4995         dprintk("%s: start\n", __func__);
4996
4997         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4998          * is not supported */
4999         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5000                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5001         else
5002                 bitmask[0] |= FATTR4_WORD0_FILEID;
5003
5004         nfs_fattr_init(&fs_locations->fattr);
5005         fs_locations->server = server;
5006         fs_locations->nlocations = 0;
5007         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5008         dprintk("%s: returned status = %d\n", __func__, status);
5009         return status;
5010 }
5011
5012 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5013                            const struct qstr *name,
5014                            struct nfs4_fs_locations *fs_locations,
5015                            struct page *page)
5016 {
5017         struct nfs4_exception exception = { };
5018         int err;
5019         do {
5020                 err = nfs4_handle_exception(NFS_SERVER(dir),
5021                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5022                                 &exception);
5023         } while (exception.retry);
5024         return err;
5025 }
5026
5027 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5028 {
5029         int status;
5030         struct nfs4_secinfo_arg args = {
5031                 .dir_fh = NFS_FH(dir),
5032                 .name   = name,
5033         };
5034         struct nfs4_secinfo_res res = {
5035                 .flavors     = flavors,
5036         };
5037         struct rpc_message msg = {
5038                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5039                 .rpc_argp = &args,
5040                 .rpc_resp = &res,
5041         };
5042
5043         dprintk("NFS call  secinfo %s\n", name->name);
5044         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5045         dprintk("NFS reply  secinfo: %d\n", status);
5046         return status;
5047 }
5048
5049 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5050                       struct nfs4_secinfo_flavors *flavors)
5051 {
5052         struct nfs4_exception exception = { };
5053         int err;
5054         do {
5055                 err = nfs4_handle_exception(NFS_SERVER(dir),
5056                                 _nfs4_proc_secinfo(dir, name, flavors),
5057                                 &exception);
5058         } while (exception.retry);
5059         return err;
5060 }
5061
5062 #ifdef CONFIG_NFS_V4_1
5063 /*
5064  * Check the exchange flags returned by the server for invalid flags, having
5065  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5066  * DS flags set.
5067  */
5068 static int nfs4_check_cl_exchange_flags(u32 flags)
5069 {
5070         if (flags & ~EXCHGID4_FLAG_MASK_R)
5071                 goto out_inval;
5072         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5073             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5074                 goto out_inval;
5075         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5076                 goto out_inval;
5077         return NFS_OK;
5078 out_inval:
5079         return -NFS4ERR_INVAL;
5080 }
5081
5082 static bool
5083 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
5084 {
5085         if (a->server_scope_sz == b->server_scope_sz &&
5086             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5087                 return true;
5088
5089         return false;
5090 }
5091
5092 /*
5093  * nfs4_proc_exchange_id()
5094  *
5095  * Since the clientid has expired, all compounds using sessions
5096  * associated with the stale clientid will be returning
5097  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5098  * be in some phase of session reset.
5099  */
5100 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5101 {
5102         nfs4_verifier verifier;
5103         struct nfs41_exchange_id_args args = {
5104                 .verifier = &verifier,
5105                 .client = clp,
5106                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5107         };
5108         struct nfs41_exchange_id_res res = {
5109                 .client = clp,
5110         };
5111         int status;
5112         struct rpc_message msg = {
5113                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5114                 .rpc_argp = &args,
5115                 .rpc_resp = &res,
5116                 .rpc_cred = cred,
5117         };
5118
5119         dprintk("--> %s\n", __func__);
5120         BUG_ON(clp == NULL);
5121
5122         nfs4_construct_boot_verifier(clp, &verifier);
5123
5124         args.id_len = scnprintf(args.id, sizeof(args.id),
5125                                 "%s/%s/%u",
5126                                 clp->cl_ipaddr,
5127                                 clp->cl_rpcclient->cl_nodename,
5128                                 clp->cl_rpcclient->cl_auth->au_flavor);
5129
5130         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
5131         if (unlikely(!res.server_scope)) {
5132                 status = -ENOMEM;
5133                 goto out;
5134         }
5135
5136         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_KERNEL);
5137         if (unlikely(!res.impl_id)) {
5138                 status = -ENOMEM;
5139                 goto out_server_scope;
5140         }
5141
5142         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5143         if (!status)
5144                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
5145
5146         if (!status) {
5147                 /* use the most recent implementation id */
5148                 kfree(clp->impl_id);
5149                 clp->impl_id = res.impl_id;
5150         } else
5151                 kfree(res.impl_id);
5152
5153         if (!status) {
5154                 if (clp->server_scope &&
5155                     !nfs41_same_server_scope(clp->server_scope,
5156                                              res.server_scope)) {
5157                         dprintk("%s: server_scope mismatch detected\n",
5158                                 __func__);
5159                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5160                         kfree(clp->server_scope);
5161                         clp->server_scope = NULL;
5162                 }
5163
5164                 if (!clp->server_scope) {
5165                         clp->server_scope = res.server_scope;
5166                         goto out;
5167                 }
5168         }
5169
5170 out_server_scope:
5171         kfree(res.server_scope);
5172 out:
5173         if (clp->impl_id)
5174                 dprintk("%s: Server Implementation ID: "
5175                         "domain: %s, name: %s, date: %llu,%u\n",
5176                         __func__, clp->impl_id->domain, clp->impl_id->name,
5177                         clp->impl_id->date.seconds,
5178                         clp->impl_id->date.nseconds);
5179         dprintk("<-- %s status= %d\n", __func__, status);
5180         return status;
5181 }
5182
5183 struct nfs4_get_lease_time_data {
5184         struct nfs4_get_lease_time_args *args;
5185         struct nfs4_get_lease_time_res *res;
5186         struct nfs_client *clp;
5187 };
5188
5189 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5190                                         void *calldata)
5191 {
5192         int ret;
5193         struct nfs4_get_lease_time_data *data =
5194                         (struct nfs4_get_lease_time_data *)calldata;
5195
5196         dprintk("--> %s\n", __func__);
5197         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5198         /* just setup sequence, do not trigger session recovery
5199            since we're invoked within one */
5200         ret = nfs41_setup_sequence(data->clp->cl_session,
5201                                    &data->args->la_seq_args,
5202                                    &data->res->lr_seq_res, task);
5203
5204         BUG_ON(ret == -EAGAIN);
5205         rpc_call_start(task);
5206         dprintk("<-- %s\n", __func__);
5207 }
5208
5209 /*
5210  * Called from nfs4_state_manager thread for session setup, so don't recover
5211  * from sequence operation or clientid errors.
5212  */
5213 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5214 {
5215         struct nfs4_get_lease_time_data *data =
5216                         (struct nfs4_get_lease_time_data *)calldata;
5217
5218         dprintk("--> %s\n", __func__);
5219         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5220                 return;
5221         switch (task->tk_status) {
5222         case -NFS4ERR_DELAY:
5223         case -NFS4ERR_GRACE:
5224                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5225                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5226                 task->tk_status = 0;
5227                 /* fall through */
5228         case -NFS4ERR_RETRY_UNCACHED_REP:
5229                 rpc_restart_call_prepare(task);
5230                 return;
5231         }
5232         dprintk("<-- %s\n", __func__);
5233 }
5234
5235 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5236         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5237         .rpc_call_done = nfs4_get_lease_time_done,
5238 };
5239
5240 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5241 {
5242         struct rpc_task *task;
5243         struct nfs4_get_lease_time_args args;
5244         struct nfs4_get_lease_time_res res = {
5245                 .lr_fsinfo = fsinfo,
5246         };
5247         struct nfs4_get_lease_time_data data = {
5248                 .args = &args,
5249                 .res = &res,
5250                 .clp = clp,
5251         };
5252         struct rpc_message msg = {
5253                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5254                 .rpc_argp = &args,
5255                 .rpc_resp = &res,
5256         };
5257         struct rpc_task_setup task_setup = {
5258                 .rpc_client = clp->cl_rpcclient,
5259                 .rpc_message = &msg,
5260                 .callback_ops = &nfs4_get_lease_time_ops,
5261                 .callback_data = &data,
5262                 .flags = RPC_TASK_TIMEOUT,
5263         };
5264         int status;
5265
5266         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5267         dprintk("--> %s\n", __func__);
5268         task = rpc_run_task(&task_setup);
5269
5270         if (IS_ERR(task))
5271                 status = PTR_ERR(task);
5272         else {
5273                 status = task->tk_status;
5274                 rpc_put_task(task);
5275         }
5276         dprintk("<-- %s return %d\n", __func__, status);
5277
5278         return status;
5279 }
5280
5281 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5282 {
5283         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5284 }
5285
5286 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5287                 struct nfs4_slot *new,
5288                 u32 max_slots,
5289                 u32 ivalue)
5290 {
5291         struct nfs4_slot *old = NULL;
5292         u32 i;
5293
5294         spin_lock(&tbl->slot_tbl_lock);
5295         if (new) {
5296                 old = tbl->slots;
5297                 tbl->slots = new;
5298                 tbl->max_slots = max_slots;
5299         }
5300         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5301         for (i = 0; i < tbl->max_slots; i++)
5302                 tbl->slots[i].seq_nr = ivalue;
5303         spin_unlock(&tbl->slot_tbl_lock);
5304         kfree(old);
5305 }
5306
5307 /*
5308  * (re)Initialise a slot table
5309  */
5310 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5311                                  u32 ivalue)
5312 {
5313         struct nfs4_slot *new = NULL;
5314         int ret = -ENOMEM;
5315
5316         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5317                 max_reqs, tbl->max_slots);
5318
5319         /* Does the newly negotiated max_reqs match the existing slot table? */
5320         if (max_reqs != tbl->max_slots) {
5321                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5322                 if (!new)
5323                         goto out;
5324         }
5325         ret = 0;
5326
5327         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5328         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5329                 tbl, tbl->slots, tbl->max_slots);
5330 out:
5331         dprintk("<-- %s: return %d\n", __func__, ret);
5332         return ret;
5333 }
5334
5335 /* Destroy the slot table */
5336 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5337 {
5338         if (session->fc_slot_table.slots != NULL) {
5339                 kfree(session->fc_slot_table.slots);
5340                 session->fc_slot_table.slots = NULL;
5341         }
5342         if (session->bc_slot_table.slots != NULL) {
5343                 kfree(session->bc_slot_table.slots);
5344                 session->bc_slot_table.slots = NULL;
5345         }
5346         return;
5347 }
5348
5349 /*
5350  * Initialize or reset the forechannel and backchannel tables
5351  */
5352 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5353 {
5354         struct nfs4_slot_table *tbl;
5355         int status;
5356
5357         dprintk("--> %s\n", __func__);
5358         /* Fore channel */
5359         tbl = &ses->fc_slot_table;
5360         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5361         if (status) /* -ENOMEM */
5362                 return status;
5363         /* Back channel */
5364         tbl = &ses->bc_slot_table;
5365         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5366         if (status && tbl->slots == NULL)
5367                 /* Fore and back channel share a connection so get
5368                  * both slot tables or neither */
5369                 nfs4_destroy_slot_tables(ses);
5370         return status;
5371 }
5372
5373 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5374 {
5375         struct nfs4_session *session;
5376         struct nfs4_slot_table *tbl;
5377
5378         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5379         if (!session)
5380                 return NULL;
5381
5382         tbl = &session->fc_slot_table;
5383         tbl->highest_used_slotid = NFS4_NO_SLOT;
5384         spin_lock_init(&tbl->slot_tbl_lock);
5385         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5386         init_completion(&tbl->complete);
5387
5388         tbl = &session->bc_slot_table;
5389         tbl->highest_used_slotid = NFS4_NO_SLOT;
5390         spin_lock_init(&tbl->slot_tbl_lock);
5391         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5392         init_completion(&tbl->complete);
5393
5394         session->session_state = 1<<NFS4_SESSION_INITING;
5395
5396         session->clp = clp;
5397         return session;
5398 }
5399
5400 void nfs4_destroy_session(struct nfs4_session *session)
5401 {
5402         struct rpc_xprt *xprt;
5403
5404         nfs4_proc_destroy_session(session);
5405
5406         rcu_read_lock();
5407         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5408         rcu_read_unlock();
5409         dprintk("%s Destroy backchannel for xprt %p\n",
5410                 __func__, xprt);
5411         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5412         nfs4_destroy_slot_tables(session);
5413         kfree(session);
5414 }
5415
5416 /*
5417  * Initialize the values to be used by the client in CREATE_SESSION
5418  * If nfs4_init_session set the fore channel request and response sizes,
5419  * use them.
5420  *
5421  * Set the back channel max_resp_sz_cached to zero to force the client to
5422  * always set csa_cachethis to FALSE because the current implementation
5423  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5424  */
5425 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5426 {
5427         struct nfs4_session *session = args->client->cl_session;
5428         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5429                      mxresp_sz = session->fc_attrs.max_resp_sz;
5430
5431         if (mxrqst_sz == 0)
5432                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5433         if (mxresp_sz == 0)
5434                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5435         /* Fore channel attributes */
5436         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5437         args->fc_attrs.max_resp_sz = mxresp_sz;
5438         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5439         args->fc_attrs.max_reqs = max_session_slots;
5440
5441         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5442                 "max_ops=%u max_reqs=%u\n",
5443                 __func__,
5444                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5445                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5446
5447         /* Back channel attributes */
5448         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5449         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5450         args->bc_attrs.max_resp_sz_cached = 0;
5451         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5452         args->bc_attrs.max_reqs = 1;
5453
5454         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5455                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5456                 __func__,
5457                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5458                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5459                 args->bc_attrs.max_reqs);
5460 }
5461
5462 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5463 {
5464         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5465         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5466
5467         if (rcvd->max_resp_sz > sent->max_resp_sz)
5468                 return -EINVAL;
5469         /*
5470          * Our requested max_ops is the minimum we need; we're not
5471          * prepared to break up compounds into smaller pieces than that.
5472          * So, no point even trying to continue if the server won't
5473          * cooperate:
5474          */
5475         if (rcvd->max_ops < sent->max_ops)
5476                 return -EINVAL;
5477         if (rcvd->max_reqs == 0)
5478                 return -EINVAL;
5479         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5480                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5481         return 0;
5482 }
5483
5484 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5485 {
5486         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5487         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5488
5489         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5490                 return -EINVAL;
5491         if (rcvd->max_resp_sz < sent->max_resp_sz)
5492                 return -EINVAL;
5493         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5494                 return -EINVAL;
5495         /* These would render the backchannel useless: */
5496         if (rcvd->max_ops != sent->max_ops)
5497                 return -EINVAL;
5498         if (rcvd->max_reqs != sent->max_reqs)
5499                 return -EINVAL;
5500         return 0;
5501 }
5502
5503 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5504                                      struct nfs4_session *session)
5505 {
5506         int ret;
5507
5508         ret = nfs4_verify_fore_channel_attrs(args, session);
5509         if (ret)
5510                 return ret;
5511         return nfs4_verify_back_channel_attrs(args, session);
5512 }
5513
5514 static int _nfs4_proc_create_session(struct nfs_client *clp)
5515 {
5516         struct nfs4_session *session = clp->cl_session;
5517         struct nfs41_create_session_args args = {
5518                 .client = clp,
5519                 .cb_program = NFS4_CALLBACK,
5520         };
5521         struct nfs41_create_session_res res = {
5522                 .client = clp,
5523         };
5524         struct rpc_message msg = {
5525                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5526                 .rpc_argp = &args,
5527                 .rpc_resp = &res,
5528         };
5529         int status;
5530
5531         nfs4_init_channel_attrs(&args);
5532         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5533
5534         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5535
5536         if (!status)
5537                 /* Verify the session's negotiated channel_attrs values */
5538                 status = nfs4_verify_channel_attrs(&args, session);
5539         if (!status) {
5540                 /* Increment the clientid slot sequence id */
5541                 clp->cl_seqid++;
5542         }
5543
5544         return status;
5545 }
5546
5547 /*
5548  * Issues a CREATE_SESSION operation to the server.
5549  * It is the responsibility of the caller to verify the session is
5550  * expired before calling this routine.
5551  */
5552 int nfs4_proc_create_session(struct nfs_client *clp)
5553 {
5554         int status;
5555         unsigned *ptr;
5556         struct nfs4_session *session = clp->cl_session;
5557
5558         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5559
5560         status = _nfs4_proc_create_session(clp);
5561         if (status)
5562                 goto out;
5563
5564         /* Init or reset the session slot tables */
5565         status = nfs4_setup_session_slot_tables(session);
5566         dprintk("slot table setup returned %d\n", status);
5567         if (status)
5568                 goto out;
5569
5570         ptr = (unsigned *)&session->sess_id.data[0];
5571         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5572                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5573 out:
5574         dprintk("<-- %s\n", __func__);
5575         return status;
5576 }
5577
5578 /*
5579  * Issue the over-the-wire RPC DESTROY_SESSION.
5580  * The caller must serialize access to this routine.
5581  */
5582 int nfs4_proc_destroy_session(struct nfs4_session *session)
5583 {
5584         int status = 0;
5585         struct rpc_message msg;
5586
5587         dprintk("--> nfs4_proc_destroy_session\n");
5588
5589         /* session is still being setup */
5590         if (session->clp->cl_cons_state != NFS_CS_READY)
5591                 return status;
5592
5593         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5594         msg.rpc_argp = session;
5595         msg.rpc_resp = NULL;
5596         msg.rpc_cred = NULL;
5597         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5598
5599         if (status)
5600                 printk(KERN_WARNING
5601                         "NFS: Got error %d from the server on DESTROY_SESSION. "
5602                         "Session has been destroyed regardless...\n", status);
5603
5604         dprintk("<-- nfs4_proc_destroy_session\n");
5605         return status;
5606 }
5607
5608 int nfs4_init_session(struct nfs_server *server)
5609 {
5610         struct nfs_client *clp = server->nfs_client;
5611         struct nfs4_session *session;
5612         unsigned int rsize, wsize;
5613         int ret;
5614
5615         if (!nfs4_has_session(clp))
5616                 return 0;
5617
5618         session = clp->cl_session;
5619         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5620                 return 0;
5621
5622         rsize = server->rsize;
5623         if (rsize == 0)
5624                 rsize = NFS_MAX_FILE_IO_SIZE;
5625         wsize = server->wsize;
5626         if (wsize == 0)
5627                 wsize = NFS_MAX_FILE_IO_SIZE;
5628
5629         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5630         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5631
5632         ret = nfs4_recover_expired_lease(server);
5633         if (!ret)
5634                 ret = nfs4_check_client_ready(clp);
5635         return ret;
5636 }
5637
5638 int nfs4_init_ds_session(struct nfs_client *clp)
5639 {
5640         struct nfs4_session *session = clp->cl_session;
5641         int ret;
5642
5643         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5644                 return 0;
5645
5646         ret = nfs4_client_recover_expired_lease(clp);
5647         if (!ret)
5648                 /* Test for the DS role */
5649                 if (!is_ds_client(clp))
5650                         ret = -ENODEV;
5651         if (!ret)
5652                 ret = nfs4_check_client_ready(clp);
5653         return ret;
5654
5655 }
5656 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5657
5658
5659 /*
5660  * Renew the cl_session lease.
5661  */
5662 struct nfs4_sequence_data {
5663         struct nfs_client *clp;
5664         struct nfs4_sequence_args args;
5665         struct nfs4_sequence_res res;
5666 };
5667
5668 static void nfs41_sequence_release(void *data)
5669 {
5670         struct nfs4_sequence_data *calldata = data;
5671         struct nfs_client *clp = calldata->clp;
5672
5673         if (atomic_read(&clp->cl_count) > 1)
5674                 nfs4_schedule_state_renewal(clp);
5675         nfs_put_client(clp);
5676         kfree(calldata);
5677 }
5678
5679 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5680 {
5681         switch(task->tk_status) {
5682         case -NFS4ERR_DELAY:
5683                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5684                 return -EAGAIN;
5685         default:
5686                 nfs4_schedule_lease_recovery(clp);
5687         }
5688         return 0;
5689 }
5690
5691 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5692 {
5693         struct nfs4_sequence_data *calldata = data;
5694         struct nfs_client *clp = calldata->clp;
5695
5696         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5697                 return;
5698
5699         if (task->tk_status < 0) {
5700                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5701                 if (atomic_read(&clp->cl_count) == 1)
5702                         goto out;
5703
5704                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5705                         rpc_restart_call_prepare(task);
5706                         return;
5707                 }
5708         }
5709         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5710 out:
5711         dprintk("<-- %s\n", __func__);
5712 }
5713
5714 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5715 {
5716         struct nfs4_sequence_data *calldata = data;
5717         struct nfs_client *clp = calldata->clp;
5718         struct nfs4_sequence_args *args;
5719         struct nfs4_sequence_res *res;
5720
5721         args = task->tk_msg.rpc_argp;
5722         res = task->tk_msg.rpc_resp;
5723
5724         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5725                 return;
5726         rpc_call_start(task);
5727 }
5728
5729 static const struct rpc_call_ops nfs41_sequence_ops = {
5730         .rpc_call_done = nfs41_sequence_call_done,
5731         .rpc_call_prepare = nfs41_sequence_prepare,
5732         .rpc_release = nfs41_sequence_release,
5733 };
5734
5735 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5736 {
5737         struct nfs4_sequence_data *calldata;
5738         struct rpc_message msg = {
5739                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5740                 .rpc_cred = cred,
5741         };
5742         struct rpc_task_setup task_setup_data = {
5743                 .rpc_client = clp->cl_rpcclient,
5744                 .rpc_message = &msg,
5745                 .callback_ops = &nfs41_sequence_ops,
5746                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5747         };
5748
5749         if (!atomic_inc_not_zero(&clp->cl_count))
5750                 return ERR_PTR(-EIO);
5751         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5752         if (calldata == NULL) {
5753                 nfs_put_client(clp);
5754                 return ERR_PTR(-ENOMEM);
5755         }
5756         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5757         msg.rpc_argp = &calldata->args;
5758         msg.rpc_resp = &calldata->res;
5759         calldata->clp = clp;
5760         task_setup_data.callback_data = calldata;
5761
5762         return rpc_run_task(&task_setup_data);
5763 }
5764
5765 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5766 {
5767         struct rpc_task *task;
5768         int ret = 0;
5769
5770         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5771                 return 0;
5772         task = _nfs41_proc_sequence(clp, cred);
5773         if (IS_ERR(task))
5774                 ret = PTR_ERR(task);
5775         else
5776                 rpc_put_task_async(task);
5777         dprintk("<-- %s status=%d\n", __func__, ret);
5778         return ret;
5779 }
5780
5781 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5782 {
5783         struct rpc_task *task;
5784         int ret;
5785
5786         task = _nfs41_proc_sequence(clp, cred);
5787         if (IS_ERR(task)) {
5788                 ret = PTR_ERR(task);
5789                 goto out;
5790         }
5791         ret = rpc_wait_for_completion_task(task);
5792         if (!ret) {
5793                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5794
5795                 if (task->tk_status == 0)
5796                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5797                 ret = task->tk_status;
5798         }
5799         rpc_put_task(task);
5800 out:
5801         dprintk("<-- %s status=%d\n", __func__, ret);
5802         return ret;
5803 }
5804
5805 struct nfs4_reclaim_complete_data {
5806         struct nfs_client *clp;
5807         struct nfs41_reclaim_complete_args arg;
5808         struct nfs41_reclaim_complete_res res;
5809 };
5810
5811 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5812 {
5813         struct nfs4_reclaim_complete_data *calldata = data;
5814
5815         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5816         if (nfs41_setup_sequence(calldata->clp->cl_session,
5817                                 &calldata->arg.seq_args,
5818                                 &calldata->res.seq_res, task))
5819                 return;
5820
5821         rpc_call_start(task);
5822 }
5823
5824 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5825 {
5826         switch(task->tk_status) {
5827         case 0:
5828         case -NFS4ERR_COMPLETE_ALREADY:
5829         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5830                 break;
5831         case -NFS4ERR_DELAY:
5832                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5833                 /* fall through */
5834         case -NFS4ERR_RETRY_UNCACHED_REP:
5835                 return -EAGAIN;
5836         default:
5837                 nfs4_schedule_lease_recovery(clp);
5838         }
5839         return 0;
5840 }
5841
5842 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5843 {
5844         struct nfs4_reclaim_complete_data *calldata = data;
5845         struct nfs_client *clp = calldata->clp;
5846         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5847
5848         dprintk("--> %s\n", __func__);
5849         if (!nfs41_sequence_done(task, res))
5850                 return;
5851
5852         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5853                 rpc_restart_call_prepare(task);
5854                 return;
5855         }
5856         dprintk("<-- %s\n", __func__);
5857 }
5858
5859 static void nfs4_free_reclaim_complete_data(void *data)
5860 {
5861         struct nfs4_reclaim_complete_data *calldata = data;
5862
5863         kfree(calldata);
5864 }
5865
5866 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5867         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5868         .rpc_call_done = nfs4_reclaim_complete_done,
5869         .rpc_release = nfs4_free_reclaim_complete_data,
5870 };
5871
5872 /*
5873  * Issue a global reclaim complete.
5874  */
5875 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5876 {
5877         struct nfs4_reclaim_complete_data *calldata;
5878         struct rpc_task *task;
5879         struct rpc_message msg = {
5880                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5881         };
5882         struct rpc_task_setup task_setup_data = {
5883                 .rpc_client = clp->cl_rpcclient,
5884                 .rpc_message = &msg,
5885                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5886                 .flags = RPC_TASK_ASYNC,
5887         };
5888         int status = -ENOMEM;
5889
5890         dprintk("--> %s\n", __func__);
5891         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5892         if (calldata == NULL)
5893                 goto out;
5894         calldata->clp = clp;
5895         calldata->arg.one_fs = 0;
5896
5897         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
5898         msg.rpc_argp = &calldata->arg;
5899         msg.rpc_resp = &calldata->res;
5900         task_setup_data.callback_data = calldata;
5901         task = rpc_run_task(&task_setup_data);
5902         if (IS_ERR(task)) {
5903                 status = PTR_ERR(task);
5904                 goto out;
5905         }
5906         status = nfs4_wait_for_completion_rpc_task(task);
5907         if (status == 0)
5908                 status = task->tk_status;
5909         rpc_put_task(task);
5910         return 0;
5911 out:
5912         dprintk("<-- %s status=%d\n", __func__, status);
5913         return status;
5914 }
5915
5916 static void
5917 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5918 {
5919         struct nfs4_layoutget *lgp = calldata;
5920         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5921
5922         dprintk("--> %s\n", __func__);
5923         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5924          * right now covering the LAYOUTGET we are about to send.
5925          * However, that is not so catastrophic, and there seems
5926          * to be no way to prevent it completely.
5927          */
5928         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5929                                 &lgp->res.seq_res, task))
5930                 return;
5931         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5932                                           NFS_I(lgp->args.inode)->layout,
5933                                           lgp->args.ctx->state)) {
5934                 rpc_exit(task, NFS4_OK);
5935                 return;
5936         }
5937         rpc_call_start(task);
5938 }
5939
5940 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5941 {
5942         struct nfs4_layoutget *lgp = calldata;
5943         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5944
5945         dprintk("--> %s\n", __func__);
5946
5947         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5948                 return;
5949
5950         switch (task->tk_status) {
5951         case 0:
5952                 break;
5953         case -NFS4ERR_LAYOUTTRYLATER:
5954         case -NFS4ERR_RECALLCONFLICT:
5955                 task->tk_status = -NFS4ERR_DELAY;
5956                 /* Fall through */
5957         default:
5958                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5959                         rpc_restart_call_prepare(task);
5960                         return;
5961                 }
5962         }
5963         dprintk("<-- %s\n", __func__);
5964 }
5965
5966 static void nfs4_layoutget_release(void *calldata)
5967 {
5968         struct nfs4_layoutget *lgp = calldata;
5969
5970         dprintk("--> %s\n", __func__);
5971         put_nfs_open_context(lgp->args.ctx);
5972         kfree(calldata);
5973         dprintk("<-- %s\n", __func__);
5974 }
5975
5976 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5977         .rpc_call_prepare = nfs4_layoutget_prepare,
5978         .rpc_call_done = nfs4_layoutget_done,
5979         .rpc_release = nfs4_layoutget_release,
5980 };
5981
5982 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5983 {
5984         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5985         struct rpc_task *task;
5986         struct rpc_message msg = {
5987                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5988                 .rpc_argp = &lgp->args,
5989                 .rpc_resp = &lgp->res,
5990         };
5991         struct rpc_task_setup task_setup_data = {
5992                 .rpc_client = server->client,
5993                 .rpc_message = &msg,
5994                 .callback_ops = &nfs4_layoutget_call_ops,
5995                 .callback_data = lgp,
5996                 .flags = RPC_TASK_ASYNC,
5997         };
5998         int status = 0;
5999
6000         dprintk("--> %s\n", __func__);
6001
6002         lgp->res.layoutp = &lgp->args.layout;
6003         lgp->res.seq_res.sr_slot = NULL;
6004         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6005         task = rpc_run_task(&task_setup_data);
6006         if (IS_ERR(task))
6007                 return PTR_ERR(task);
6008         status = nfs4_wait_for_completion_rpc_task(task);
6009         if (status == 0)
6010                 status = task->tk_status;
6011         if (status == 0)
6012                 status = pnfs_layout_process(lgp);
6013         rpc_put_task(task);
6014         dprintk("<-- %s status=%d\n", __func__, status);
6015         return status;
6016 }
6017
6018 static void
6019 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6020 {
6021         struct nfs4_layoutreturn *lrp = calldata;
6022
6023         dprintk("--> %s\n", __func__);
6024         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
6025                                 &lrp->res.seq_res, task))
6026                 return;
6027         rpc_call_start(task);
6028 }
6029
6030 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6031 {
6032         struct nfs4_layoutreturn *lrp = calldata;
6033         struct nfs_server *server;
6034         struct pnfs_layout_hdr *lo = lrp->args.layout;
6035
6036         dprintk("--> %s\n", __func__);
6037
6038         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
6039                 return;
6040
6041         server = NFS_SERVER(lrp->args.inode);
6042         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6043                 rpc_restart_call_prepare(task);
6044                 return;
6045         }
6046         spin_lock(&lo->plh_inode->i_lock);
6047         if (task->tk_status == 0) {
6048                 if (lrp->res.lrs_present) {
6049                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6050                 } else
6051                         BUG_ON(!list_empty(&lo->plh_segs));
6052         }
6053         lo->plh_block_lgets--;
6054         spin_unlock(&lo->plh_inode->i_lock);
6055         dprintk("<-- %s\n", __func__);
6056 }
6057
6058 static void nfs4_layoutreturn_release(void *calldata)
6059 {
6060         struct nfs4_layoutreturn *lrp = calldata;
6061
6062         dprintk("--> %s\n", __func__);
6063         put_layout_hdr(lrp->args.layout);
6064         kfree(calldata);
6065         dprintk("<-- %s\n", __func__);
6066 }
6067
6068 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6069         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6070         .rpc_call_done = nfs4_layoutreturn_done,
6071         .rpc_release = nfs4_layoutreturn_release,
6072 };
6073
6074 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6075 {
6076         struct rpc_task *task;
6077         struct rpc_message msg = {
6078                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6079                 .rpc_argp = &lrp->args,
6080                 .rpc_resp = &lrp->res,
6081         };
6082         struct rpc_task_setup task_setup_data = {
6083                 .rpc_client = lrp->clp->cl_rpcclient,
6084                 .rpc_message = &msg,
6085                 .callback_ops = &nfs4_layoutreturn_call_ops,
6086                 .callback_data = lrp,
6087         };
6088         int status;
6089
6090         dprintk("--> %s\n", __func__);
6091         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6092         task = rpc_run_task(&task_setup_data);
6093         if (IS_ERR(task))
6094                 return PTR_ERR(task);
6095         status = task->tk_status;
6096         dprintk("<-- %s status=%d\n", __func__, status);
6097         rpc_put_task(task);
6098         return status;
6099 }
6100
6101 /*
6102  * Retrieve the list of Data Server devices from the MDS.
6103  */
6104 static int _nfs4_getdevicelist(struct nfs_server *server,
6105                                     const struct nfs_fh *fh,
6106                                     struct pnfs_devicelist *devlist)
6107 {
6108         struct nfs4_getdevicelist_args args = {
6109                 .fh = fh,
6110                 .layoutclass = server->pnfs_curr_ld->id,
6111         };
6112         struct nfs4_getdevicelist_res res = {
6113                 .devlist = devlist,
6114         };
6115         struct rpc_message msg = {
6116                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6117                 .rpc_argp = &args,
6118                 .rpc_resp = &res,
6119         };
6120         int status;
6121
6122         dprintk("--> %s\n", __func__);
6123         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6124                                 &res.seq_res, 0);
6125         dprintk("<-- %s status=%d\n", __func__, status);
6126         return status;
6127 }
6128
6129 int nfs4_proc_getdevicelist(struct nfs_server *server,
6130                             const struct nfs_fh *fh,
6131                             struct pnfs_devicelist *devlist)
6132 {
6133         struct nfs4_exception exception = { };
6134         int err;
6135
6136         do {
6137                 err = nfs4_handle_exception(server,
6138                                 _nfs4_getdevicelist(server, fh, devlist),
6139                                 &exception);
6140         } while (exception.retry);
6141
6142         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6143                 err, devlist->num_devs);
6144
6145         return err;
6146 }
6147 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6148
6149 static int
6150 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6151 {
6152         struct nfs4_getdeviceinfo_args args = {
6153                 .pdev = pdev,
6154         };
6155         struct nfs4_getdeviceinfo_res res = {
6156                 .pdev = pdev,
6157         };
6158         struct rpc_message msg = {
6159                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6160                 .rpc_argp = &args,
6161                 .rpc_resp = &res,
6162         };
6163         int status;
6164
6165         dprintk("--> %s\n", __func__);
6166         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6167         dprintk("<-- %s status=%d\n", __func__, status);
6168
6169         return status;
6170 }
6171
6172 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6173 {
6174         struct nfs4_exception exception = { };
6175         int err;
6176
6177         do {
6178                 err = nfs4_handle_exception(server,
6179                                         _nfs4_proc_getdeviceinfo(server, pdev),
6180                                         &exception);
6181         } while (exception.retry);
6182         return err;
6183 }
6184 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6185
6186 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6187 {
6188         struct nfs4_layoutcommit_data *data = calldata;
6189         struct nfs_server *server = NFS_SERVER(data->args.inode);
6190
6191         if (nfs4_setup_sequence(server, &data->args.seq_args,
6192                                 &data->res.seq_res, task))
6193                 return;
6194         rpc_call_start(task);
6195 }
6196
6197 static void
6198 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6199 {
6200         struct nfs4_layoutcommit_data *data = calldata;
6201         struct nfs_server *server = NFS_SERVER(data->args.inode);
6202
6203         if (!nfs4_sequence_done(task, &data->res.seq_res))
6204                 return;
6205
6206         switch (task->tk_status) { /* Just ignore these failures */
6207         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6208         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6209         case -NFS4ERR_BADLAYOUT:     /* no layout */
6210         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6211                 task->tk_status = 0;
6212                 break;
6213         case 0:
6214                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6215                                                    data->res.fattr);
6216                 break;
6217         default:
6218                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6219                         rpc_restart_call_prepare(task);
6220                         return;
6221                 }
6222         }
6223 }
6224
6225 static void nfs4_layoutcommit_release(void *calldata)
6226 {
6227         struct nfs4_layoutcommit_data *data = calldata;
6228