NFSv4.1: Shrink struct nfs4_sequence_res by moving the session pointer
[linux-2.6-block.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/nfs_idmap.h>
55 #include <linux/sunrpc/bc_xprt.h>
56 #include <linux/xattr.h>
57 #include <linux/utsname.h>
58 #include <linux/freezer.h>
59
60 #include "nfs4_fs.h"
61 #include "delegation.h"
62 #include "internal.h"
63 #include "iostat.h"
64 #include "callback.h"
65 #include "pnfs.h"
66 #include "netns.h"
67
68 #define NFSDBG_FACILITY         NFSDBG_PROC
69
70 #define NFS4_POLL_RETRY_MIN     (HZ/10)
71 #define NFS4_POLL_RETRY_MAX     (15*HZ)
72
73 #define NFS4_MAX_LOOP_ON_RECOVER (10)
74
75 struct nfs4_opendata;
76 static int _nfs4_proc_open(struct nfs4_opendata *data);
77 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
78 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
79 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
80 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
81 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *);
82 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
83 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
84                             struct nfs_fattr *fattr, struct iattr *sattr,
85                             struct nfs4_state *state);
86 #ifdef CONFIG_NFS_V4_1
87 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
88 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
89 #endif
90 /* Prevent leaks of NFSv4 errors into userland */
91 static int nfs4_map_errors(int err)
92 {
93         if (err >= -1000)
94                 return err;
95         switch (err) {
96         case -NFS4ERR_RESOURCE:
97                 return -EREMOTEIO;
98         case -NFS4ERR_WRONGSEC:
99                 return -EPERM;
100         case -NFS4ERR_BADOWNER:
101         case -NFS4ERR_BADNAME:
102                 return -EINVAL;
103         case -NFS4ERR_SHARE_DENIED:
104                 return -EACCES;
105         case -NFS4ERR_MINOR_VERS_MISMATCH:
106                 return -EPROTONOSUPPORT;
107         case -NFS4ERR_ACCESS:
108                 return -EACCES;
109         default:
110                 dprintk("%s could not handle NFSv4 error %d\n",
111                                 __func__, -err);
112                 break;
113         }
114         return -EIO;
115 }
116
117 /*
118  * This is our standard bitmap for GETATTR requests.
119  */
120 const u32 nfs4_fattr_bitmap[3] = {
121         FATTR4_WORD0_TYPE
122         | FATTR4_WORD0_CHANGE
123         | FATTR4_WORD0_SIZE
124         | FATTR4_WORD0_FSID
125         | FATTR4_WORD0_FILEID,
126         FATTR4_WORD1_MODE
127         | FATTR4_WORD1_NUMLINKS
128         | FATTR4_WORD1_OWNER
129         | FATTR4_WORD1_OWNER_GROUP
130         | FATTR4_WORD1_RAWDEV
131         | FATTR4_WORD1_SPACE_USED
132         | FATTR4_WORD1_TIME_ACCESS
133         | FATTR4_WORD1_TIME_METADATA
134         | FATTR4_WORD1_TIME_MODIFY
135 };
136
137 static const u32 nfs4_pnfs_open_bitmap[3] = {
138         FATTR4_WORD0_TYPE
139         | FATTR4_WORD0_CHANGE
140         | FATTR4_WORD0_SIZE
141         | FATTR4_WORD0_FSID
142         | FATTR4_WORD0_FILEID,
143         FATTR4_WORD1_MODE
144         | FATTR4_WORD1_NUMLINKS
145         | FATTR4_WORD1_OWNER
146         | FATTR4_WORD1_OWNER_GROUP
147         | FATTR4_WORD1_RAWDEV
148         | FATTR4_WORD1_SPACE_USED
149         | FATTR4_WORD1_TIME_ACCESS
150         | FATTR4_WORD1_TIME_METADATA
151         | FATTR4_WORD1_TIME_MODIFY,
152         FATTR4_WORD2_MDSTHRESHOLD
153 };
154
155 static const u32 nfs4_open_noattr_bitmap[3] = {
156         FATTR4_WORD0_TYPE
157         | FATTR4_WORD0_CHANGE
158         | FATTR4_WORD0_FILEID,
159 };
160
161 const u32 nfs4_statfs_bitmap[2] = {
162         FATTR4_WORD0_FILES_AVAIL
163         | FATTR4_WORD0_FILES_FREE
164         | FATTR4_WORD0_FILES_TOTAL,
165         FATTR4_WORD1_SPACE_AVAIL
166         | FATTR4_WORD1_SPACE_FREE
167         | FATTR4_WORD1_SPACE_TOTAL
168 };
169
170 const u32 nfs4_pathconf_bitmap[2] = {
171         FATTR4_WORD0_MAXLINK
172         | FATTR4_WORD0_MAXNAME,
173         0
174 };
175
176 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
177                         | FATTR4_WORD0_MAXREAD
178                         | FATTR4_WORD0_MAXWRITE
179                         | FATTR4_WORD0_LEASE_TIME,
180                         FATTR4_WORD1_TIME_DELTA
181                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
182                         FATTR4_WORD2_LAYOUT_BLKSIZE
183 };
184
185 const u32 nfs4_fs_locations_bitmap[2] = {
186         FATTR4_WORD0_TYPE
187         | FATTR4_WORD0_CHANGE
188         | FATTR4_WORD0_SIZE
189         | FATTR4_WORD0_FSID
190         | FATTR4_WORD0_FILEID
191         | FATTR4_WORD0_FS_LOCATIONS,
192         FATTR4_WORD1_MODE
193         | FATTR4_WORD1_NUMLINKS
194         | FATTR4_WORD1_OWNER
195         | FATTR4_WORD1_OWNER_GROUP
196         | FATTR4_WORD1_RAWDEV
197         | FATTR4_WORD1_SPACE_USED
198         | FATTR4_WORD1_TIME_ACCESS
199         | FATTR4_WORD1_TIME_METADATA
200         | FATTR4_WORD1_TIME_MODIFY
201         | FATTR4_WORD1_MOUNTED_ON_FILEID
202 };
203
204 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
205                 struct nfs4_readdir_arg *readdir)
206 {
207         __be32 *start, *p;
208
209         if (cookie > 2) {
210                 readdir->cookie = cookie;
211                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
212                 return;
213         }
214
215         readdir->cookie = 0;
216         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
217         if (cookie == 2)
218                 return;
219         
220         /*
221          * NFSv4 servers do not return entries for '.' and '..'
222          * Therefore, we fake these entries here.  We let '.'
223          * have cookie 0 and '..' have cookie 1.  Note that
224          * when talking to the server, we always send cookie 0
225          * instead of 1 or 2.
226          */
227         start = p = kmap_atomic(*readdir->pages);
228         
229         if (cookie == 0) {
230                 *p++ = xdr_one;                                  /* next */
231                 *p++ = xdr_zero;                   /* cookie, first word */
232                 *p++ = xdr_one;                   /* cookie, second word */
233                 *p++ = xdr_one;                             /* entry len */
234                 memcpy(p, ".\0\0\0", 4);                        /* entry */
235                 p++;
236                 *p++ = xdr_one;                         /* bitmap length */
237                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
238                 *p++ = htonl(8);              /* attribute buffer length */
239                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
240         }
241         
242         *p++ = xdr_one;                                  /* next */
243         *p++ = xdr_zero;                   /* cookie, first word */
244         *p++ = xdr_two;                   /* cookie, second word */
245         *p++ = xdr_two;                             /* entry len */
246         memcpy(p, "..\0\0", 4);                         /* entry */
247         p++;
248         *p++ = xdr_one;                         /* bitmap length */
249         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
250         *p++ = htonl(8);              /* attribute buffer length */
251         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
252
253         readdir->pgbase = (char *)p - (char *)start;
254         readdir->count -= readdir->pgbase;
255         kunmap_atomic(start);
256 }
257
258 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
259 {
260         int res;
261
262         might_sleep();
263
264         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
265                         nfs_wait_bit_killable, TASK_KILLABLE);
266         if (res)
267                 return res;
268
269         if (clp->cl_cons_state < 0)
270                 return clp->cl_cons_state;
271         return 0;
272 }
273
274 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
275 {
276         int res = 0;
277
278         might_sleep();
279
280         if (*timeout <= 0)
281                 *timeout = NFS4_POLL_RETRY_MIN;
282         if (*timeout > NFS4_POLL_RETRY_MAX)
283                 *timeout = NFS4_POLL_RETRY_MAX;
284         freezable_schedule_timeout_killable(*timeout);
285         if (fatal_signal_pending(current))
286                 res = -ERESTARTSYS;
287         *timeout <<= 1;
288         return res;
289 }
290
291 /* This is the error handling routine for processes that are allowed
292  * to sleep.
293  */
294 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
295 {
296         struct nfs_client *clp = server->nfs_client;
297         struct nfs4_state *state = exception->state;
298         struct inode *inode = exception->inode;
299         int ret = errorcode;
300
301         exception->retry = 0;
302         switch(errorcode) {
303                 case 0:
304                         return 0;
305                 case -NFS4ERR_OPENMODE:
306                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
307                                 nfs4_inode_return_delegation(inode);
308                                 exception->retry = 1;
309                                 return 0;
310                         }
311                         if (state == NULL)
312                                 break;
313                         nfs4_schedule_stateid_recovery(server, state);
314                         goto wait_on_recovery;
315                 case -NFS4ERR_DELEG_REVOKED:
316                 case -NFS4ERR_ADMIN_REVOKED:
317                 case -NFS4ERR_BAD_STATEID:
318                         if (state == NULL)
319                                 break;
320                         nfs_remove_bad_delegation(state->inode);
321                         nfs4_schedule_stateid_recovery(server, state);
322                         goto wait_on_recovery;
323                 case -NFS4ERR_EXPIRED:
324                         if (state != NULL)
325                                 nfs4_schedule_stateid_recovery(server, state);
326                 case -NFS4ERR_STALE_STATEID:
327                 case -NFS4ERR_STALE_CLIENTID:
328                         nfs4_schedule_lease_recovery(clp);
329                         goto wait_on_recovery;
330 #if defined(CONFIG_NFS_V4_1)
331                 case -NFS4ERR_BADSESSION:
332                 case -NFS4ERR_BADSLOT:
333                 case -NFS4ERR_BAD_HIGH_SLOT:
334                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
335                 case -NFS4ERR_DEADSESSION:
336                 case -NFS4ERR_SEQ_FALSE_RETRY:
337                 case -NFS4ERR_SEQ_MISORDERED:
338                         dprintk("%s ERROR: %d Reset session\n", __func__,
339                                 errorcode);
340                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
341                         goto wait_on_recovery;
342 #endif /* defined(CONFIG_NFS_V4_1) */
343                 case -NFS4ERR_FILE_OPEN:
344                         if (exception->timeout > HZ) {
345                                 /* We have retried a decent amount, time to
346                                  * fail
347                                  */
348                                 ret = -EBUSY;
349                                 break;
350                         }
351                 case -NFS4ERR_GRACE:
352                 case -NFS4ERR_DELAY:
353                 case -EKEYEXPIRED:
354                         ret = nfs4_delay(server->client, &exception->timeout);
355                         if (ret != 0)
356                                 break;
357                 case -NFS4ERR_RETRY_UNCACHED_REP:
358                 case -NFS4ERR_OLD_STATEID:
359                         exception->retry = 1;
360                         break;
361                 case -NFS4ERR_BADOWNER:
362                         /* The following works around a Linux server bug! */
363                 case -NFS4ERR_BADNAME:
364                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
365                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
366                                 exception->retry = 1;
367                                 printk(KERN_WARNING "NFS: v4 server %s "
368                                                 "does not accept raw "
369                                                 "uid/gids. "
370                                                 "Reenabling the idmapper.\n",
371                                                 server->nfs_client->cl_hostname);
372                         }
373         }
374         /* We failed to handle the error */
375         return nfs4_map_errors(ret);
376 wait_on_recovery:
377         ret = nfs4_wait_clnt_recover(clp);
378         if (ret == 0)
379                 exception->retry = 1;
380         return ret;
381 }
382
383
384 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
385 {
386         spin_lock(&clp->cl_lock);
387         if (time_before(clp->cl_last_renewal,timestamp))
388                 clp->cl_last_renewal = timestamp;
389         spin_unlock(&clp->cl_lock);
390 }
391
392 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
393 {
394         do_renew_lease(server->nfs_client, timestamp);
395 }
396
397 #if defined(CONFIG_NFS_V4_1)
398
399 /*
400  * nfs4_free_slot - free a slot and efficiently update slot table.
401  *
402  * freeing a slot is trivially done by clearing its respective bit
403  * in the bitmap.
404  * If the freed slotid equals highest_used_slotid we want to update it
405  * so that the server would be able to size down the slot table if needed,
406  * otherwise we know that the highest_used_slotid is still in use.
407  * When updating highest_used_slotid there may be "holes" in the bitmap
408  * so we need to scan down from highest_used_slotid to 0 looking for the now
409  * highest slotid in use.
410  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
411  *
412  * Must be called while holding tbl->slot_tbl_lock
413  */
414 static void
415 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
416 {
417         /* clear used bit in bitmap */
418         __clear_bit(slotid, tbl->used_slots);
419
420         /* update highest_used_slotid when it is freed */
421         if (slotid == tbl->highest_used_slotid) {
422                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
423                 if (slotid < tbl->max_slots)
424                         tbl->highest_used_slotid = slotid;
425                 else
426                         tbl->highest_used_slotid = NFS4_NO_SLOT;
427         }
428         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
429                 slotid, tbl->highest_used_slotid);
430 }
431
432 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
433 {
434         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
435         return true;
436 }
437
438 /*
439  * Signal state manager thread if session fore channel is drained
440  */
441 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
442 {
443         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
444                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
445                                 nfs4_set_task_privileged, NULL);
446                 return;
447         }
448
449         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
450                 return;
451
452         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
453         complete(&ses->fc_slot_table.complete);
454 }
455
456 /*
457  * Signal state manager thread if session back channel is drained
458  */
459 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
460 {
461         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
462             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
463                 return;
464         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
465         complete(&ses->bc_slot_table.complete);
466 }
467
468 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
469 {
470         struct nfs4_session *session;
471         struct nfs4_slot_table *tbl;
472
473         if (!res->sr_slot) {
474                 /* just wake up the next guy waiting since
475                  * we may have not consumed a slot after all */
476                 dprintk("%s: No slot\n", __func__);
477                 return;
478         }
479         tbl = res->sr_slot->table;
480         session = tbl->session;
481
482         spin_lock(&tbl->slot_tbl_lock);
483         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
484         nfs4_check_drain_fc_complete(session);
485         spin_unlock(&tbl->slot_tbl_lock);
486         res->sr_slot = NULL;
487 }
488
489 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
490 {
491         struct nfs4_session *session;
492         struct nfs4_slot *slot;
493         unsigned long timestamp;
494         struct nfs_client *clp;
495
496         /*
497          * sr_status remains 1 if an RPC level error occurred. The server
498          * may or may not have processed the sequence operation..
499          * Proceed as if the server received and processed the sequence
500          * operation.
501          */
502         if (res->sr_status == 1)
503                 res->sr_status = NFS_OK;
504
505         /* don't increment the sequence number if the task wasn't sent */
506         if (!RPC_WAS_SENT(task))
507                 goto out;
508
509         slot = res->sr_slot;
510         session = slot->table->session;
511
512         /* Check the SEQUENCE operation status */
513         switch (res->sr_status) {
514         case 0:
515                 /* Update the slot's sequence and clientid lease timer */
516                 ++slot->seq_nr;
517                 timestamp = slot->renewal_time;
518                 clp = session->clp;
519                 do_renew_lease(clp, timestamp);
520                 /* Check sequence flags */
521                 if (res->sr_status_flags != 0)
522                         nfs4_schedule_lease_recovery(clp);
523                 break;
524         case -NFS4ERR_DELAY:
525                 /* The server detected a resend of the RPC call and
526                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
527                  * of RFC5661.
528                  */
529                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
530                         __func__,
531                         slot - session->fc_slot_table.slots,
532                         slot->seq_nr);
533                 goto out_retry;
534         default:
535                 /* Just update the slot sequence no. */
536                 ++slot->seq_nr;
537         }
538 out:
539         /* The session may be reset by one of the error handlers. */
540         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
541         nfs41_sequence_free_slot(res);
542         return 1;
543 out_retry:
544         if (!rpc_restart_call(task))
545                 goto out;
546         rpc_delay(task, NFS4_POLL_RETRY_MAX);
547         return 0;
548 }
549
550 static int nfs4_sequence_done(struct rpc_task *task,
551                                struct nfs4_sequence_res *res)
552 {
553         if (res->sr_slot == NULL)
554                 return 1;
555         return nfs41_sequence_done(task, res);
556 }
557
558 /*
559  * nfs4_find_slot - efficiently look for a free slot
560  *
561  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
562  * If found, we mark the slot as used, update the highest_used_slotid,
563  * and respectively set up the sequence operation args.
564  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
565  *
566  * Note: must be called with under the slot_tbl_lock.
567  */
568 static u32
569 nfs4_find_slot(struct nfs4_slot_table *tbl)
570 {
571         u32 slotid;
572         u32 ret_id = NFS4_NO_SLOT;
573
574         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
575                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
576                 tbl->max_slots);
577         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
578         if (slotid >= tbl->max_slots)
579                 goto out;
580         __set_bit(slotid, tbl->used_slots);
581         if (slotid > tbl->highest_used_slotid ||
582                         tbl->highest_used_slotid == NFS4_NO_SLOT)
583                 tbl->highest_used_slotid = slotid;
584         ret_id = slotid;
585 out:
586         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
587                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
588         return ret_id;
589 }
590
591 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
592                 struct nfs4_sequence_res *res, int cache_reply)
593 {
594         args->sa_session = NULL;
595         args->sa_cache_this = 0;
596         if (cache_reply)
597                 args->sa_cache_this = 1;
598         res->sr_slot = NULL;
599 }
600
601 int nfs41_setup_sequence(struct nfs4_session *session,
602                                 struct nfs4_sequence_args *args,
603                                 struct nfs4_sequence_res *res,
604                                 struct rpc_task *task)
605 {
606         struct nfs4_slot *slot;
607         struct nfs4_slot_table *tbl;
608         u32 slotid;
609
610         dprintk("--> %s\n", __func__);
611         /* slot already allocated? */
612         if (res->sr_slot != NULL)
613                 return 0;
614
615         tbl = &session->fc_slot_table;
616
617         spin_lock(&tbl->slot_tbl_lock);
618         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
619             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
620                 /* The state manager will wait until the slot table is empty */
621                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
622                 spin_unlock(&tbl->slot_tbl_lock);
623                 dprintk("%s session is draining\n", __func__);
624                 return -EAGAIN;
625         }
626
627         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
628             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
629                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
630                 spin_unlock(&tbl->slot_tbl_lock);
631                 dprintk("%s enforce FIFO order\n", __func__);
632                 return -EAGAIN;
633         }
634
635         slotid = nfs4_find_slot(tbl);
636         if (slotid == NFS4_NO_SLOT) {
637                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
638                 spin_unlock(&tbl->slot_tbl_lock);
639                 dprintk("<-- %s: no free slots\n", __func__);
640                 return -EAGAIN;
641         }
642         spin_unlock(&tbl->slot_tbl_lock);
643
644         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
645         slot = tbl->slots + slotid;
646         slot->renewal_time = jiffies;
647         args->sa_session = session;
648         args->sa_slotid = slotid;
649
650         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
651
652         res->sr_slot = slot;
653         res->sr_status_flags = 0;
654         /*
655          * sr_status is only set in decode_sequence, and so will remain
656          * set to 1 if an rpc level failure occurs.
657          */
658         res->sr_status = 1;
659         return 0;
660 }
661 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
662
663 int nfs4_setup_sequence(const struct nfs_server *server,
664                         struct nfs4_sequence_args *args,
665                         struct nfs4_sequence_res *res,
666                         struct rpc_task *task)
667 {
668         struct nfs4_session *session = nfs4_get_session(server);
669         int ret = 0;
670
671         if (session == NULL)
672                 goto out;
673
674         dprintk("--> %s clp %p session %p sr_slot %td\n",
675                 __func__, session->clp, session, res->sr_slot ?
676                         res->sr_slot - session->fc_slot_table.slots : -1);
677
678         ret = nfs41_setup_sequence(session, args, res, task);
679 out:
680         dprintk("<-- %s status=%d\n", __func__, ret);
681         return ret;
682 }
683
684 struct nfs41_call_sync_data {
685         const struct nfs_server *seq_server;
686         struct nfs4_sequence_args *seq_args;
687         struct nfs4_sequence_res *seq_res;
688 };
689
690 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
691 {
692         struct nfs41_call_sync_data *data = calldata;
693
694         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
695
696         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
697                                 data->seq_res, task))
698                 return;
699         rpc_call_start(task);
700 }
701
702 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
703 {
704         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
705         nfs41_call_sync_prepare(task, calldata);
706 }
707
708 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
709 {
710         struct nfs41_call_sync_data *data = calldata;
711
712         nfs41_sequence_done(task, data->seq_res);
713 }
714
715 static const struct rpc_call_ops nfs41_call_sync_ops = {
716         .rpc_call_prepare = nfs41_call_sync_prepare,
717         .rpc_call_done = nfs41_call_sync_done,
718 };
719
720 static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
721         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
722         .rpc_call_done = nfs41_call_sync_done,
723 };
724
725 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
726                                    struct nfs_server *server,
727                                    struct rpc_message *msg,
728                                    struct nfs4_sequence_args *args,
729                                    struct nfs4_sequence_res *res,
730                                    int privileged)
731 {
732         int ret;
733         struct rpc_task *task;
734         struct nfs41_call_sync_data data = {
735                 .seq_server = server,
736                 .seq_args = args,
737                 .seq_res = res,
738         };
739         struct rpc_task_setup task_setup = {
740                 .rpc_client = clnt,
741                 .rpc_message = msg,
742                 .callback_ops = &nfs41_call_sync_ops,
743                 .callback_data = &data
744         };
745
746         if (privileged)
747                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
748         task = rpc_run_task(&task_setup);
749         if (IS_ERR(task))
750                 ret = PTR_ERR(task);
751         else {
752                 ret = task->tk_status;
753                 rpc_put_task(task);
754         }
755         return ret;
756 }
757
758 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
759                             struct nfs_server *server,
760                             struct rpc_message *msg,
761                             struct nfs4_sequence_args *args,
762                             struct nfs4_sequence_res *res,
763                             int cache_reply)
764 {
765         nfs41_init_sequence(args, res, cache_reply);
766         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
767 }
768
769 #else
770 static inline
771 void nfs41_init_sequence(struct nfs4_sequence_args *args,
772                 struct nfs4_sequence_res *res, int cache_reply)
773 {
774 }
775
776 static int nfs4_sequence_done(struct rpc_task *task,
777                                struct nfs4_sequence_res *res)
778 {
779         return 1;
780 }
781 #endif /* CONFIG_NFS_V4_1 */
782
783 int _nfs4_call_sync(struct rpc_clnt *clnt,
784                     struct nfs_server *server,
785                     struct rpc_message *msg,
786                     struct nfs4_sequence_args *args,
787                     struct nfs4_sequence_res *res,
788                     int cache_reply)
789 {
790         nfs41_init_sequence(args, res, cache_reply);
791         return rpc_call_sync(clnt, msg, 0);
792 }
793
794 static inline
795 int nfs4_call_sync(struct rpc_clnt *clnt,
796                    struct nfs_server *server,
797                    struct rpc_message *msg,
798                    struct nfs4_sequence_args *args,
799                    struct nfs4_sequence_res *res,
800                    int cache_reply)
801 {
802         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
803                                                 args, res, cache_reply);
804 }
805
806 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
807 {
808         struct nfs_inode *nfsi = NFS_I(dir);
809
810         spin_lock(&dir->i_lock);
811         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
812         if (!cinfo->atomic || cinfo->before != dir->i_version)
813                 nfs_force_lookup_revalidate(dir);
814         dir->i_version = cinfo->after;
815         spin_unlock(&dir->i_lock);
816 }
817
818 struct nfs4_opendata {
819         struct kref kref;
820         struct nfs_openargs o_arg;
821         struct nfs_openres o_res;
822         struct nfs_open_confirmargs c_arg;
823         struct nfs_open_confirmres c_res;
824         struct nfs4_string owner_name;
825         struct nfs4_string group_name;
826         struct nfs_fattr f_attr;
827         struct dentry *dir;
828         struct dentry *dentry;
829         struct nfs4_state_owner *owner;
830         struct nfs4_state *state;
831         struct iattr attrs;
832         unsigned long timestamp;
833         unsigned int rpc_done : 1;
834         int rpc_status;
835         int cancelled;
836 };
837
838
839 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
840 {
841         p->o_res.f_attr = &p->f_attr;
842         p->o_res.seqid = p->o_arg.seqid;
843         p->c_res.seqid = p->c_arg.seqid;
844         p->o_res.server = p->o_arg.server;
845         p->o_res.access_request = p->o_arg.access;
846         nfs_fattr_init(&p->f_attr);
847         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
848 }
849
850 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
851                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
852                 const struct iattr *attrs,
853                 gfp_t gfp_mask)
854 {
855         struct dentry *parent = dget_parent(dentry);
856         struct inode *dir = parent->d_inode;
857         struct nfs_server *server = NFS_SERVER(dir);
858         struct nfs4_opendata *p;
859
860         p = kzalloc(sizeof(*p), gfp_mask);
861         if (p == NULL)
862                 goto err;
863         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
864         if (p->o_arg.seqid == NULL)
865                 goto err_free;
866         nfs_sb_active(dentry->d_sb);
867         p->dentry = dget(dentry);
868         p->dir = parent;
869         p->owner = sp;
870         atomic_inc(&sp->so_count);
871         p->o_arg.fh = NFS_FH(dir);
872         p->o_arg.open_flags = flags;
873         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
874         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
875          * will return permission denied for all bits until close */
876         if (!(flags & O_EXCL)) {
877                 /* ask server to check for all possible rights as results
878                  * are cached */
879                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
880                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
881         }
882         p->o_arg.clientid = server->nfs_client->cl_clientid;
883         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
884         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
885         p->o_arg.name = &dentry->d_name;
886         p->o_arg.server = server;
887         p->o_arg.bitmask = server->attr_bitmask;
888         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
889         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
890         if (attrs != NULL && attrs->ia_valid != 0) {
891                 __be32 verf[2];
892
893                 p->o_arg.u.attrs = &p->attrs;
894                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
895
896                 verf[0] = jiffies;
897                 verf[1] = current->pid;
898                 memcpy(p->o_arg.u.verifier.data, verf,
899                                 sizeof(p->o_arg.u.verifier.data));
900         }
901         p->c_arg.fh = &p->o_res.fh;
902         p->c_arg.stateid = &p->o_res.stateid;
903         p->c_arg.seqid = p->o_arg.seqid;
904         nfs4_init_opendata_res(p);
905         kref_init(&p->kref);
906         return p;
907 err_free:
908         kfree(p);
909 err:
910         dput(parent);
911         return NULL;
912 }
913
914 static void nfs4_opendata_free(struct kref *kref)
915 {
916         struct nfs4_opendata *p = container_of(kref,
917                         struct nfs4_opendata, kref);
918         struct super_block *sb = p->dentry->d_sb;
919
920         nfs_free_seqid(p->o_arg.seqid);
921         if (p->state != NULL)
922                 nfs4_put_open_state(p->state);
923         nfs4_put_state_owner(p->owner);
924         dput(p->dir);
925         dput(p->dentry);
926         nfs_sb_deactive(sb);
927         nfs_fattr_free_names(&p->f_attr);
928         kfree(p);
929 }
930
931 static void nfs4_opendata_put(struct nfs4_opendata *p)
932 {
933         if (p != NULL)
934                 kref_put(&p->kref, nfs4_opendata_free);
935 }
936
937 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
938 {
939         int ret;
940
941         ret = rpc_wait_for_completion_task(task);
942         return ret;
943 }
944
945 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
946 {
947         int ret = 0;
948
949         if (open_mode & (O_EXCL|O_TRUNC))
950                 goto out;
951         switch (mode & (FMODE_READ|FMODE_WRITE)) {
952                 case FMODE_READ:
953                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
954                                 && state->n_rdonly != 0;
955                         break;
956                 case FMODE_WRITE:
957                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
958                                 && state->n_wronly != 0;
959                         break;
960                 case FMODE_READ|FMODE_WRITE:
961                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
962                                 && state->n_rdwr != 0;
963         }
964 out:
965         return ret;
966 }
967
968 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
969 {
970         if (delegation == NULL)
971                 return 0;
972         if ((delegation->type & fmode) != fmode)
973                 return 0;
974         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
975                 return 0;
976         nfs_mark_delegation_referenced(delegation);
977         return 1;
978 }
979
980 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
981 {
982         switch (fmode) {
983                 case FMODE_WRITE:
984                         state->n_wronly++;
985                         break;
986                 case FMODE_READ:
987                         state->n_rdonly++;
988                         break;
989                 case FMODE_READ|FMODE_WRITE:
990                         state->n_rdwr++;
991         }
992         nfs4_state_set_mode_locked(state, state->state | fmode);
993 }
994
995 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
996 {
997         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
998                 nfs4_stateid_copy(&state->stateid, stateid);
999         nfs4_stateid_copy(&state->open_stateid, stateid);
1000         switch (fmode) {
1001                 case FMODE_READ:
1002                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1003                         break;
1004                 case FMODE_WRITE:
1005                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1006                         break;
1007                 case FMODE_READ|FMODE_WRITE:
1008                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1009         }
1010 }
1011
1012 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1013 {
1014         write_seqlock(&state->seqlock);
1015         nfs_set_open_stateid_locked(state, stateid, fmode);
1016         write_sequnlock(&state->seqlock);
1017 }
1018
1019 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1020 {
1021         /*
1022          * Protect the call to nfs4_state_set_mode_locked and
1023          * serialise the stateid update
1024          */
1025         write_seqlock(&state->seqlock);
1026         if (deleg_stateid != NULL) {
1027                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1028                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1029         }
1030         if (open_stateid != NULL)
1031                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1032         write_sequnlock(&state->seqlock);
1033         spin_lock(&state->owner->so_lock);
1034         update_open_stateflags(state, fmode);
1035         spin_unlock(&state->owner->so_lock);
1036 }
1037
1038 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1039 {
1040         struct nfs_inode *nfsi = NFS_I(state->inode);
1041         struct nfs_delegation *deleg_cur;
1042         int ret = 0;
1043
1044         fmode &= (FMODE_READ|FMODE_WRITE);
1045
1046         rcu_read_lock();
1047         deleg_cur = rcu_dereference(nfsi->delegation);
1048         if (deleg_cur == NULL)
1049                 goto no_delegation;
1050
1051         spin_lock(&deleg_cur->lock);
1052         if (nfsi->delegation != deleg_cur ||
1053             (deleg_cur->type & fmode) != fmode)
1054                 goto no_delegation_unlock;
1055
1056         if (delegation == NULL)
1057                 delegation = &deleg_cur->stateid;
1058         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1059                 goto no_delegation_unlock;
1060
1061         nfs_mark_delegation_referenced(deleg_cur);
1062         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1063         ret = 1;
1064 no_delegation_unlock:
1065         spin_unlock(&deleg_cur->lock);
1066 no_delegation:
1067         rcu_read_unlock();
1068
1069         if (!ret && open_stateid != NULL) {
1070                 __update_open_stateid(state, open_stateid, NULL, fmode);
1071                 ret = 1;
1072         }
1073
1074         return ret;
1075 }
1076
1077
1078 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1079 {
1080         struct nfs_delegation *delegation;
1081
1082         rcu_read_lock();
1083         delegation = rcu_dereference(NFS_I(inode)->delegation);
1084         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1085                 rcu_read_unlock();
1086                 return;
1087         }
1088         rcu_read_unlock();
1089         nfs4_inode_return_delegation(inode);
1090 }
1091
1092 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1093 {
1094         struct nfs4_state *state = opendata->state;
1095         struct nfs_inode *nfsi = NFS_I(state->inode);
1096         struct nfs_delegation *delegation;
1097         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1098         fmode_t fmode = opendata->o_arg.fmode;
1099         nfs4_stateid stateid;
1100         int ret = -EAGAIN;
1101
1102         for (;;) {
1103                 if (can_open_cached(state, fmode, open_mode)) {
1104                         spin_lock(&state->owner->so_lock);
1105                         if (can_open_cached(state, fmode, open_mode)) {
1106                                 update_open_stateflags(state, fmode);
1107                                 spin_unlock(&state->owner->so_lock);
1108                                 goto out_return_state;
1109                         }
1110                         spin_unlock(&state->owner->so_lock);
1111                 }
1112                 rcu_read_lock();
1113                 delegation = rcu_dereference(nfsi->delegation);
1114                 if (!can_open_delegated(delegation, fmode)) {
1115                         rcu_read_unlock();
1116                         break;
1117                 }
1118                 /* Save the delegation */
1119                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1120                 rcu_read_unlock();
1121                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1122                 if (ret != 0)
1123                         goto out;
1124                 ret = -EAGAIN;
1125
1126                 /* Try to update the stateid using the delegation */
1127                 if (update_open_stateid(state, NULL, &stateid, fmode))
1128                         goto out_return_state;
1129         }
1130 out:
1131         return ERR_PTR(ret);
1132 out_return_state:
1133         atomic_inc(&state->count);
1134         return state;
1135 }
1136
1137 static void
1138 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1139 {
1140         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1141         struct nfs_delegation *delegation;
1142         int delegation_flags = 0;
1143
1144         rcu_read_lock();
1145         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1146         if (delegation)
1147                 delegation_flags = delegation->flags;
1148         rcu_read_unlock();
1149         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1150                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1151                                    "returning a delegation for "
1152                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1153                                    clp->cl_hostname);
1154         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1155                 nfs_inode_set_delegation(state->inode,
1156                                          data->owner->so_cred,
1157                                          &data->o_res);
1158         else
1159                 nfs_inode_reclaim_delegation(state->inode,
1160                                              data->owner->so_cred,
1161                                              &data->o_res);
1162 }
1163
1164 /*
1165  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1166  * and update the nfs4_state.
1167  */
1168 static struct nfs4_state *
1169 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1170 {
1171         struct inode *inode = data->state->inode;
1172         struct nfs4_state *state = data->state;
1173         int ret;
1174
1175         if (!data->rpc_done) {
1176                 ret = data->rpc_status;
1177                 goto err;
1178         }
1179
1180         ret = -ESTALE;
1181         if (!(data->f_attr.valid & NFS_ATTR_FATTR_TYPE) ||
1182             !(data->f_attr.valid & NFS_ATTR_FATTR_FILEID) ||
1183             !(data->f_attr.valid & NFS_ATTR_FATTR_CHANGE))
1184                 goto err;
1185
1186         ret = -ENOMEM;
1187         state = nfs4_get_open_state(inode, data->owner);
1188         if (state == NULL)
1189                 goto err;
1190
1191         ret = nfs_refresh_inode(inode, &data->f_attr);
1192         if (ret)
1193                 goto err;
1194
1195         if (data->o_res.delegation_type != 0)
1196                 nfs4_opendata_check_deleg(data, state);
1197         update_open_stateid(state, &data->o_res.stateid, NULL,
1198                             data->o_arg.fmode);
1199
1200         return state;
1201 err:
1202         return ERR_PTR(ret);
1203
1204 }
1205
1206 static struct nfs4_state *
1207 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1208 {
1209         struct inode *inode;
1210         struct nfs4_state *state = NULL;
1211         int ret;
1212
1213         if (!data->rpc_done) {
1214                 state = nfs4_try_open_cached(data);
1215                 goto out;
1216         }
1217
1218         ret = -EAGAIN;
1219         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1220                 goto err;
1221         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1222         ret = PTR_ERR(inode);
1223         if (IS_ERR(inode))
1224                 goto err;
1225         ret = -ENOMEM;
1226         state = nfs4_get_open_state(inode, data->owner);
1227         if (state == NULL)
1228                 goto err_put_inode;
1229         if (data->o_res.delegation_type != 0)
1230                 nfs4_opendata_check_deleg(data, state);
1231         update_open_stateid(state, &data->o_res.stateid, NULL,
1232                         data->o_arg.fmode);
1233         iput(inode);
1234 out:
1235         return state;
1236 err_put_inode:
1237         iput(inode);
1238 err:
1239         return ERR_PTR(ret);
1240 }
1241
1242 static struct nfs4_state *
1243 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1244 {
1245         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1246                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1247         return _nfs4_opendata_to_nfs4_state(data);
1248 }
1249
1250 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1251 {
1252         struct nfs_inode *nfsi = NFS_I(state->inode);
1253         struct nfs_open_context *ctx;
1254
1255         spin_lock(&state->inode->i_lock);
1256         list_for_each_entry(ctx, &nfsi->open_files, list) {
1257                 if (ctx->state != state)
1258                         continue;
1259                 get_nfs_open_context(ctx);
1260                 spin_unlock(&state->inode->i_lock);
1261                 return ctx;
1262         }
1263         spin_unlock(&state->inode->i_lock);
1264         return ERR_PTR(-ENOENT);
1265 }
1266
1267 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1268 {
1269         struct nfs4_opendata *opendata;
1270
1271         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1272         if (opendata == NULL)
1273                 return ERR_PTR(-ENOMEM);
1274         opendata->state = state;
1275         atomic_inc(&state->count);
1276         return opendata;
1277 }
1278
1279 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1280 {
1281         struct nfs4_state *newstate;
1282         int ret;
1283
1284         opendata->o_arg.open_flags = 0;
1285         opendata->o_arg.fmode = fmode;
1286         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1287         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1288         nfs4_init_opendata_res(opendata);
1289         ret = _nfs4_recover_proc_open(opendata);
1290         if (ret != 0)
1291                 return ret; 
1292         newstate = nfs4_opendata_to_nfs4_state(opendata);
1293         if (IS_ERR(newstate))
1294                 return PTR_ERR(newstate);
1295         nfs4_close_state(newstate, fmode);
1296         *res = newstate;
1297         return 0;
1298 }
1299
1300 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1301 {
1302         struct nfs4_state *newstate;
1303         int ret;
1304
1305         /* memory barrier prior to reading state->n_* */
1306         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1307         smp_rmb();
1308         if (state->n_rdwr != 0) {
1309                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1310                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1311                 if (ret != 0)
1312                         return ret;
1313                 if (newstate != state)
1314                         return -ESTALE;
1315         }
1316         if (state->n_wronly != 0) {
1317                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1318                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1319                 if (ret != 0)
1320                         return ret;
1321                 if (newstate != state)
1322                         return -ESTALE;
1323         }
1324         if (state->n_rdonly != 0) {
1325                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1326                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1327                 if (ret != 0)
1328                         return ret;
1329                 if (newstate != state)
1330                         return -ESTALE;
1331         }
1332         /*
1333          * We may have performed cached opens for all three recoveries.
1334          * Check if we need to update the current stateid.
1335          */
1336         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1337             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1338                 write_seqlock(&state->seqlock);
1339                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1340                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1341                 write_sequnlock(&state->seqlock);
1342         }
1343         return 0;
1344 }
1345
1346 /*
1347  * OPEN_RECLAIM:
1348  *      reclaim state on the server after a reboot.
1349  */
1350 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1351 {
1352         struct nfs_delegation *delegation;
1353         struct nfs4_opendata *opendata;
1354         fmode_t delegation_type = 0;
1355         int status;
1356
1357         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1358         if (IS_ERR(opendata))
1359                 return PTR_ERR(opendata);
1360         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1361         opendata->o_arg.fh = NFS_FH(state->inode);
1362         rcu_read_lock();
1363         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1364         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1365                 delegation_type = delegation->type;
1366         rcu_read_unlock();
1367         opendata->o_arg.u.delegation_type = delegation_type;
1368         status = nfs4_open_recover(opendata, state);
1369         nfs4_opendata_put(opendata);
1370         return status;
1371 }
1372
1373 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1374 {
1375         struct nfs_server *server = NFS_SERVER(state->inode);
1376         struct nfs4_exception exception = { };
1377         int err;
1378         do {
1379                 err = _nfs4_do_open_reclaim(ctx, state);
1380                 if (err != -NFS4ERR_DELAY)
1381                         break;
1382                 nfs4_handle_exception(server, err, &exception);
1383         } while (exception.retry);
1384         return err;
1385 }
1386
1387 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1388 {
1389         struct nfs_open_context *ctx;
1390         int ret;
1391
1392         ctx = nfs4_state_find_open_context(state);
1393         if (IS_ERR(ctx))
1394                 return PTR_ERR(ctx);
1395         ret = nfs4_do_open_reclaim(ctx, state);
1396         put_nfs_open_context(ctx);
1397         return ret;
1398 }
1399
1400 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1401 {
1402         struct nfs4_opendata *opendata;
1403         int ret;
1404
1405         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1406         if (IS_ERR(opendata))
1407                 return PTR_ERR(opendata);
1408         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1409         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1410         ret = nfs4_open_recover(opendata, state);
1411         nfs4_opendata_put(opendata);
1412         return ret;
1413 }
1414
1415 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1416 {
1417         struct nfs4_exception exception = { };
1418         struct nfs_server *server = NFS_SERVER(state->inode);
1419         int err;
1420         do {
1421                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1422                 switch (err) {
1423                         case 0:
1424                         case -ENOENT:
1425                         case -ESTALE:
1426                                 goto out;
1427                         case -NFS4ERR_BADSESSION:
1428                         case -NFS4ERR_BADSLOT:
1429                         case -NFS4ERR_BAD_HIGH_SLOT:
1430                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1431                         case -NFS4ERR_DEADSESSION:
1432                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1433                                 goto out;
1434                         case -NFS4ERR_STALE_CLIENTID:
1435                         case -NFS4ERR_STALE_STATEID:
1436                         case -NFS4ERR_EXPIRED:
1437                                 /* Don't recall a delegation if it was lost */
1438                                 nfs4_schedule_lease_recovery(server->nfs_client);
1439                                 goto out;
1440                         case -ERESTARTSYS:
1441                                 /*
1442                                  * The show must go on: exit, but mark the
1443                                  * stateid as needing recovery.
1444                                  */
1445                         case -NFS4ERR_DELEG_REVOKED:
1446                         case -NFS4ERR_ADMIN_REVOKED:
1447                         case -NFS4ERR_BAD_STATEID:
1448                                 nfs_inode_find_state_and_recover(state->inode,
1449                                                 stateid);
1450                                 nfs4_schedule_stateid_recovery(server, state);
1451                         case -EKEYEXPIRED:
1452                                 /*
1453                                  * User RPCSEC_GSS context has expired.
1454                                  * We cannot recover this stateid now, so
1455                                  * skip it and allow recovery thread to
1456                                  * proceed.
1457                                  */
1458                         case -ENOMEM:
1459                                 err = 0;
1460                                 goto out;
1461                 }
1462                 err = nfs4_handle_exception(server, err, &exception);
1463         } while (exception.retry);
1464 out:
1465         return err;
1466 }
1467
1468 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1469 {
1470         struct nfs4_opendata *data = calldata;
1471
1472         data->rpc_status = task->tk_status;
1473         if (data->rpc_status == 0) {
1474                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1475                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1476                 renew_lease(data->o_res.server, data->timestamp);
1477                 data->rpc_done = 1;
1478         }
1479 }
1480
1481 static void nfs4_open_confirm_release(void *calldata)
1482 {
1483         struct nfs4_opendata *data = calldata;
1484         struct nfs4_state *state = NULL;
1485
1486         /* If this request hasn't been cancelled, do nothing */
1487         if (data->cancelled == 0)
1488                 goto out_free;
1489         /* In case of error, no cleanup! */
1490         if (!data->rpc_done)
1491                 goto out_free;
1492         state = nfs4_opendata_to_nfs4_state(data);
1493         if (!IS_ERR(state))
1494                 nfs4_close_state(state, data->o_arg.fmode);
1495 out_free:
1496         nfs4_opendata_put(data);
1497 }
1498
1499 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1500         .rpc_call_done = nfs4_open_confirm_done,
1501         .rpc_release = nfs4_open_confirm_release,
1502 };
1503
1504 /*
1505  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1506  */
1507 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1508 {
1509         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1510         struct rpc_task *task;
1511         struct  rpc_message msg = {
1512                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1513                 .rpc_argp = &data->c_arg,
1514                 .rpc_resp = &data->c_res,
1515                 .rpc_cred = data->owner->so_cred,
1516         };
1517         struct rpc_task_setup task_setup_data = {
1518                 .rpc_client = server->client,
1519                 .rpc_message = &msg,
1520                 .callback_ops = &nfs4_open_confirm_ops,
1521                 .callback_data = data,
1522                 .workqueue = nfsiod_workqueue,
1523                 .flags = RPC_TASK_ASYNC,
1524         };
1525         int status;
1526
1527         kref_get(&data->kref);
1528         data->rpc_done = 0;
1529         data->rpc_status = 0;
1530         data->timestamp = jiffies;
1531         task = rpc_run_task(&task_setup_data);
1532         if (IS_ERR(task))
1533                 return PTR_ERR(task);
1534         status = nfs4_wait_for_completion_rpc_task(task);
1535         if (status != 0) {
1536                 data->cancelled = 1;
1537                 smp_wmb();
1538         } else
1539                 status = data->rpc_status;
1540         rpc_put_task(task);
1541         return status;
1542 }
1543
1544 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1545 {
1546         struct nfs4_opendata *data = calldata;
1547         struct nfs4_state_owner *sp = data->owner;
1548
1549         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1550                 return;
1551         /*
1552          * Check if we still need to send an OPEN call, or if we can use
1553          * a delegation instead.
1554          */
1555         if (data->state != NULL) {
1556                 struct nfs_delegation *delegation;
1557
1558                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1559                         goto out_no_action;
1560                 rcu_read_lock();
1561                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1562                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1563                     can_open_delegated(delegation, data->o_arg.fmode))
1564                         goto unlock_no_action;
1565                 rcu_read_unlock();
1566         }
1567         /* Update client id. */
1568         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1569         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1570                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1571                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1572                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1573         }
1574         data->timestamp = jiffies;
1575         if (nfs4_setup_sequence(data->o_arg.server,
1576                                 &data->o_arg.seq_args,
1577                                 &data->o_res.seq_res,
1578                                 task) != 0)
1579                 nfs_release_seqid(data->o_arg.seqid);
1580         else
1581                 rpc_call_start(task);
1582         return;
1583 unlock_no_action:
1584         rcu_read_unlock();
1585 out_no_action:
1586         task->tk_action = NULL;
1587
1588 }
1589
1590 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1591 {
1592         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1593         nfs4_open_prepare(task, calldata);
1594 }
1595
1596 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1597 {
1598         struct nfs4_opendata *data = calldata;
1599
1600         data->rpc_status = task->tk_status;
1601
1602         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1603                 return;
1604
1605         if (task->tk_status == 0) {
1606                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1607                         switch (data->o_res.f_attr->mode & S_IFMT) {
1608                         case S_IFREG:
1609                                 break;
1610                         case S_IFLNK:
1611                                 data->rpc_status = -ELOOP;
1612                                 break;
1613                         case S_IFDIR:
1614                                 data->rpc_status = -EISDIR;
1615                                 break;
1616                         default:
1617                                 data->rpc_status = -ENOTDIR;
1618                         }
1619                 }
1620                 renew_lease(data->o_res.server, data->timestamp);
1621                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1622                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1623         }
1624         data->rpc_done = 1;
1625 }
1626
1627 static void nfs4_open_release(void *calldata)
1628 {
1629         struct nfs4_opendata *data = calldata;
1630         struct nfs4_state *state = NULL;
1631
1632         /* If this request hasn't been cancelled, do nothing */
1633         if (data->cancelled == 0)
1634                 goto out_free;
1635         /* In case of error, no cleanup! */
1636         if (data->rpc_status != 0 || !data->rpc_done)
1637                 goto out_free;
1638         /* In case we need an open_confirm, no cleanup! */
1639         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1640                 goto out_free;
1641         state = nfs4_opendata_to_nfs4_state(data);
1642         if (!IS_ERR(state))
1643                 nfs4_close_state(state, data->o_arg.fmode);
1644 out_free:
1645         nfs4_opendata_put(data);
1646 }
1647
1648 static const struct rpc_call_ops nfs4_open_ops = {
1649         .rpc_call_prepare = nfs4_open_prepare,
1650         .rpc_call_done = nfs4_open_done,
1651         .rpc_release = nfs4_open_release,
1652 };
1653
1654 static const struct rpc_call_ops nfs4_recover_open_ops = {
1655         .rpc_call_prepare = nfs4_recover_open_prepare,
1656         .rpc_call_done = nfs4_open_done,
1657         .rpc_release = nfs4_open_release,
1658 };
1659
1660 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1661 {
1662         struct inode *dir = data->dir->d_inode;
1663         struct nfs_server *server = NFS_SERVER(dir);
1664         struct nfs_openargs *o_arg = &data->o_arg;
1665         struct nfs_openres *o_res = &data->o_res;
1666         struct rpc_task *task;
1667         struct rpc_message msg = {
1668                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1669                 .rpc_argp = o_arg,
1670                 .rpc_resp = o_res,
1671                 .rpc_cred = data->owner->so_cred,
1672         };
1673         struct rpc_task_setup task_setup_data = {
1674                 .rpc_client = server->client,
1675                 .rpc_message = &msg,
1676                 .callback_ops = &nfs4_open_ops,
1677                 .callback_data = data,
1678                 .workqueue = nfsiod_workqueue,
1679                 .flags = RPC_TASK_ASYNC,
1680         };
1681         int status;
1682
1683         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1684         kref_get(&data->kref);
1685         data->rpc_done = 0;
1686         data->rpc_status = 0;
1687         data->cancelled = 0;
1688         if (isrecover)
1689                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1690         task = rpc_run_task(&task_setup_data);
1691         if (IS_ERR(task))
1692                 return PTR_ERR(task);
1693         status = nfs4_wait_for_completion_rpc_task(task);
1694         if (status != 0) {
1695                 data->cancelled = 1;
1696                 smp_wmb();
1697         } else
1698                 status = data->rpc_status;
1699         rpc_put_task(task);
1700
1701         return status;
1702 }
1703
1704 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1705 {
1706         struct inode *dir = data->dir->d_inode;
1707         struct nfs_openres *o_res = &data->o_res;
1708         int status;
1709
1710         status = nfs4_run_open_task(data, 1);
1711         if (status != 0 || !data->rpc_done)
1712                 return status;
1713
1714         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1715
1716         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1717                 status = _nfs4_proc_open_confirm(data);
1718                 if (status != 0)
1719                         return status;
1720         }
1721
1722         return status;
1723 }
1724
1725 static int nfs4_opendata_access(struct rpc_cred *cred,
1726                                 struct nfs4_opendata *opendata,
1727                                 struct nfs4_state *state, fmode_t fmode)
1728 {
1729         struct nfs_access_entry cache;
1730         u32 mask;
1731
1732         /* access call failed or for some reason the server doesn't
1733          * support any access modes -- defer access call until later */
1734         if (opendata->o_res.access_supported == 0)
1735                 return 0;
1736
1737         mask = 0;
1738         /* don't check MAY_WRITE - a newly created file may not have
1739          * write mode bits, but POSIX allows the creating process to write */
1740         if (fmode & FMODE_READ)
1741                 mask |= MAY_READ;
1742         if (fmode & FMODE_EXEC)
1743                 mask |= MAY_EXEC;
1744
1745         cache.cred = cred;
1746         cache.jiffies = jiffies;
1747         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1748         nfs_access_add_cache(state->inode, &cache);
1749
1750         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1751                 return 0;
1752
1753         /* even though OPEN succeeded, access is denied. Close the file */
1754         nfs4_close_state(state, fmode);
1755         return -EACCES;
1756 }
1757
1758 /*
1759  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1760  */
1761 static int _nfs4_proc_open(struct nfs4_opendata *data)
1762 {
1763         struct inode *dir = data->dir->d_inode;
1764         struct nfs_server *server = NFS_SERVER(dir);
1765         struct nfs_openargs *o_arg = &data->o_arg;
1766         struct nfs_openres *o_res = &data->o_res;
1767         int status;
1768
1769         status = nfs4_run_open_task(data, 0);
1770         if (!data->rpc_done)
1771                 return status;
1772         if (status != 0) {
1773                 if (status == -NFS4ERR_BADNAME &&
1774                                 !(o_arg->open_flags & O_CREAT))
1775                         return -ENOENT;
1776                 return status;
1777         }
1778
1779         nfs_fattr_map_and_free_names(server, &data->f_attr);
1780
1781         if (o_arg->open_flags & O_CREAT)
1782                 update_changeattr(dir, &o_res->cinfo);
1783         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1784                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1785         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1786                 status = _nfs4_proc_open_confirm(data);
1787                 if (status != 0)
1788                         return status;
1789         }
1790         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1791                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1792         return 0;
1793 }
1794
1795 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1796 {
1797         unsigned int loop;
1798         int ret;
1799
1800         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1801                 ret = nfs4_wait_clnt_recover(clp);
1802                 if (ret != 0)
1803                         break;
1804                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1805                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1806                         break;
1807                 nfs4_schedule_state_manager(clp);
1808                 ret = -EIO;
1809         }
1810         return ret;
1811 }
1812
1813 static int nfs4_recover_expired_lease(struct nfs_server *server)
1814 {
1815         return nfs4_client_recover_expired_lease(server->nfs_client);
1816 }
1817
1818 /*
1819  * OPEN_EXPIRED:
1820  *      reclaim state on the server after a network partition.
1821  *      Assumes caller holds the appropriate lock
1822  */
1823 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1824 {
1825         struct nfs4_opendata *opendata;
1826         int ret;
1827
1828         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1829         if (IS_ERR(opendata))
1830                 return PTR_ERR(opendata);
1831         ret = nfs4_open_recover(opendata, state);
1832         if (ret == -ESTALE)
1833                 d_drop(ctx->dentry);
1834         nfs4_opendata_put(opendata);
1835         return ret;
1836 }
1837
1838 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1839 {
1840         struct nfs_server *server = NFS_SERVER(state->inode);
1841         struct nfs4_exception exception = { };
1842         int err;
1843
1844         do {
1845                 err = _nfs4_open_expired(ctx, state);
1846                 switch (err) {
1847                 default:
1848                         goto out;
1849                 case -NFS4ERR_GRACE:
1850                 case -NFS4ERR_DELAY:
1851                         nfs4_handle_exception(server, err, &exception);
1852                         err = 0;
1853                 }
1854         } while (exception.retry);
1855 out:
1856         return err;
1857 }
1858
1859 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1860 {
1861         struct nfs_open_context *ctx;
1862         int ret;
1863
1864         ctx = nfs4_state_find_open_context(state);
1865         if (IS_ERR(ctx))
1866                 return PTR_ERR(ctx);
1867         ret = nfs4_do_open_expired(ctx, state);
1868         put_nfs_open_context(ctx);
1869         return ret;
1870 }
1871
1872 #if defined(CONFIG_NFS_V4_1)
1873 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1874 {
1875         struct nfs_server *server = NFS_SERVER(state->inode);
1876         nfs4_stateid *stateid = &state->stateid;
1877         int status;
1878
1879         /* If a state reset has been done, test_stateid is unneeded */
1880         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1881                 return;
1882
1883         status = nfs41_test_stateid(server, stateid);
1884         if (status != NFS_OK) {
1885                 /* Free the stateid unless the server explicitly
1886                  * informs us the stateid is unrecognized. */
1887                 if (status != -NFS4ERR_BAD_STATEID)
1888                         nfs41_free_stateid(server, stateid);
1889                 nfs_remove_bad_delegation(state->inode);
1890
1891                 write_seqlock(&state->seqlock);
1892                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1893                 write_sequnlock(&state->seqlock);
1894                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1895         }
1896 }
1897
1898 /**
1899  * nfs41_check_open_stateid - possibly free an open stateid
1900  *
1901  * @state: NFSv4 state for an inode
1902  *
1903  * Returns NFS_OK if recovery for this stateid is now finished.
1904  * Otherwise a negative NFS4ERR value is returned.
1905  */
1906 static int nfs41_check_open_stateid(struct nfs4_state *state)
1907 {
1908         struct nfs_server *server = NFS_SERVER(state->inode);
1909         nfs4_stateid *stateid = &state->open_stateid;
1910         int status;
1911
1912         /* If a state reset has been done, test_stateid is unneeded */
1913         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1914             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1915             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1916                 return -NFS4ERR_BAD_STATEID;
1917
1918         status = nfs41_test_stateid(server, stateid);
1919         if (status != NFS_OK) {
1920                 /* Free the stateid unless the server explicitly
1921                  * informs us the stateid is unrecognized. */
1922                 if (status != -NFS4ERR_BAD_STATEID)
1923                         nfs41_free_stateid(server, stateid);
1924
1925                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1926                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1927                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1928         }
1929         return status;
1930 }
1931
1932 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1933 {
1934         int status;
1935
1936         nfs41_clear_delegation_stateid(state);
1937         status = nfs41_check_open_stateid(state);
1938         if (status != NFS_OK)
1939                 status = nfs4_open_expired(sp, state);
1940         return status;
1941 }
1942 #endif
1943
1944 /*
1945  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1946  * fields corresponding to attributes that were used to store the verifier.
1947  * Make sure we clobber those fields in the later setattr call
1948  */
1949 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1950 {
1951         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1952             !(sattr->ia_valid & ATTR_ATIME_SET))
1953                 sattr->ia_valid |= ATTR_ATIME;
1954
1955         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1956             !(sattr->ia_valid & ATTR_MTIME_SET))
1957                 sattr->ia_valid |= ATTR_MTIME;
1958 }
1959
1960 /*
1961  * Returns a referenced nfs4_state
1962  */
1963 static int _nfs4_do_open(struct inode *dir,
1964                         struct dentry *dentry,
1965                         fmode_t fmode,
1966                         int flags,
1967                         struct iattr *sattr,
1968                         struct rpc_cred *cred,
1969                         struct nfs4_state **res,
1970                         struct nfs4_threshold **ctx_th)
1971 {
1972         struct nfs4_state_owner  *sp;
1973         struct nfs4_state     *state = NULL;
1974         struct nfs_server       *server = NFS_SERVER(dir);
1975         struct nfs4_opendata *opendata;
1976         int status;
1977
1978         /* Protect against reboot recovery conflicts */
1979         status = -ENOMEM;
1980         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1981         if (sp == NULL) {
1982                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1983                 goto out_err;
1984         }
1985         status = nfs4_recover_expired_lease(server);
1986         if (status != 0)
1987                 goto err_put_state_owner;
1988         if (dentry->d_inode != NULL)
1989                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1990         status = -ENOMEM;
1991         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1992         if (opendata == NULL)
1993                 goto err_put_state_owner;
1994
1995         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1996                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1997                 if (!opendata->f_attr.mdsthreshold)
1998                         goto err_opendata_put;
1999                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2000         }
2001         if (dentry->d_inode != NULL)
2002                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2003
2004         status = _nfs4_proc_open(opendata);
2005         if (status != 0)
2006                 goto err_opendata_put;
2007
2008         state = nfs4_opendata_to_nfs4_state(opendata);
2009         status = PTR_ERR(state);
2010         if (IS_ERR(state))
2011                 goto err_opendata_put;
2012         if (server->caps & NFS_CAP_POSIX_LOCK)
2013                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2014
2015         status = nfs4_opendata_access(cred, opendata, state, fmode);
2016         if (status != 0)
2017                 goto err_opendata_put;
2018
2019         if (opendata->o_arg.open_flags & O_EXCL) {
2020                 nfs4_exclusive_attrset(opendata, sattr);
2021
2022                 nfs_fattr_init(opendata->o_res.f_attr);
2023                 status = nfs4_do_setattr(state->inode, cred,
2024                                 opendata->o_res.f_attr, sattr,
2025                                 state);
2026                 if (status == 0)
2027                         nfs_setattr_update_inode(state->inode, sattr);
2028                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2029         }
2030
2031         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2032                 *ctx_th = opendata->f_attr.mdsthreshold;
2033         else
2034                 kfree(opendata->f_attr.mdsthreshold);
2035         opendata->f_attr.mdsthreshold = NULL;
2036
2037         nfs4_opendata_put(opendata);
2038         nfs4_put_state_owner(sp);
2039         *res = state;
2040         return 0;
2041 err_opendata_put:
2042         kfree(opendata->f_attr.mdsthreshold);
2043         nfs4_opendata_put(opendata);
2044 err_put_state_owner:
2045         nfs4_put_state_owner(sp);
2046 out_err:
2047         *res = NULL;
2048         return status;
2049 }
2050
2051
2052 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2053                                         struct dentry *dentry,
2054                                         fmode_t fmode,
2055                                         int flags,
2056                                         struct iattr *sattr,
2057                                         struct rpc_cred *cred,
2058                                         struct nfs4_threshold **ctx_th)
2059 {
2060         struct nfs4_exception exception = { };
2061         struct nfs4_state *res;
2062         int status;
2063
2064         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
2065         do {
2066                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
2067                                        &res, ctx_th);
2068                 if (status == 0)
2069                         break;
2070                 /* NOTE: BAD_SEQID means the server and client disagree about the
2071                  * book-keeping w.r.t. state-changing operations
2072                  * (OPEN/CLOSE/LOCK/LOCKU...)
2073                  * It is actually a sign of a bug on the client or on the server.
2074                  *
2075                  * If we receive a BAD_SEQID error in the particular case of
2076                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2077                  * have unhashed the old state_owner for us, and that we can
2078                  * therefore safely retry using a new one. We should still warn
2079                  * the user though...
2080                  */
2081                 if (status == -NFS4ERR_BAD_SEQID) {
2082                         pr_warn_ratelimited("NFS: v4 server %s "
2083                                         " returned a bad sequence-id error!\n",
2084                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2085                         exception.retry = 1;
2086                         continue;
2087                 }
2088                 /*
2089                  * BAD_STATEID on OPEN means that the server cancelled our
2090                  * state before it received the OPEN_CONFIRM.
2091                  * Recover by retrying the request as per the discussion
2092                  * on Page 181 of RFC3530.
2093                  */
2094                 if (status == -NFS4ERR_BAD_STATEID) {
2095                         exception.retry = 1;
2096                         continue;
2097                 }
2098                 if (status == -EAGAIN) {
2099                         /* We must have found a delegation */
2100                         exception.retry = 1;
2101                         continue;
2102                 }
2103                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
2104                                         status, &exception));
2105         } while (exception.retry);
2106         return res;
2107 }
2108
2109 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2110                             struct nfs_fattr *fattr, struct iattr *sattr,
2111                             struct nfs4_state *state)
2112 {
2113         struct nfs_server *server = NFS_SERVER(inode);
2114         struct nfs_setattrargs  arg = {
2115                 .fh             = NFS_FH(inode),
2116                 .iap            = sattr,
2117                 .server         = server,
2118                 .bitmask = server->attr_bitmask,
2119         };
2120         struct nfs_setattrres  res = {
2121                 .fattr          = fattr,
2122                 .server         = server,
2123         };
2124         struct rpc_message msg = {
2125                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2126                 .rpc_argp       = &arg,
2127                 .rpc_resp       = &res,
2128                 .rpc_cred       = cred,
2129         };
2130         unsigned long timestamp = jiffies;
2131         int status;
2132
2133         nfs_fattr_init(fattr);
2134
2135         if (state != NULL) {
2136                 struct nfs_lockowner lockowner = {
2137                         .l_owner = current->files,
2138                         .l_pid = current->tgid,
2139                 };
2140                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2141                                 &lockowner);
2142         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
2143                                 FMODE_WRITE)) {
2144                 /* Use that stateid */
2145         } else
2146                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2147
2148         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2149         if (status == 0 && state != NULL)
2150                 renew_lease(server, timestamp);
2151         return status;
2152 }
2153
2154 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2155                            struct nfs_fattr *fattr, struct iattr *sattr,
2156                            struct nfs4_state *state)
2157 {
2158         struct nfs_server *server = NFS_SERVER(inode);
2159         struct nfs4_exception exception = {
2160                 .state = state,
2161                 .inode = inode,
2162         };
2163         int err;
2164         do {
2165                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2166                 switch (err) {
2167                 case -NFS4ERR_OPENMODE:
2168                         if (state && !(state->state & FMODE_WRITE)) {
2169                                 err = -EBADF;
2170                                 if (sattr->ia_valid & ATTR_OPEN)
2171                                         err = -EACCES;
2172                                 goto out;
2173                         }
2174                 }
2175                 err = nfs4_handle_exception(server, err, &exception);
2176         } while (exception.retry);
2177 out:
2178         return err;
2179 }
2180
2181 struct nfs4_closedata {
2182         struct inode *inode;
2183         struct nfs4_state *state;
2184         struct nfs_closeargs arg;
2185         struct nfs_closeres res;
2186         struct nfs_fattr fattr;
2187         unsigned long timestamp;
2188         bool roc;
2189         u32 roc_barrier;
2190 };
2191
2192 static void nfs4_free_closedata(void *data)
2193 {
2194         struct nfs4_closedata *calldata = data;
2195         struct nfs4_state_owner *sp = calldata->state->owner;
2196         struct super_block *sb = calldata->state->inode->i_sb;
2197
2198         if (calldata->roc)
2199                 pnfs_roc_release(calldata->state->inode);
2200         nfs4_put_open_state(calldata->state);
2201         nfs_free_seqid(calldata->arg.seqid);
2202         nfs4_put_state_owner(sp);
2203         nfs_sb_deactive_async(sb);
2204         kfree(calldata);
2205 }
2206
2207 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2208                 fmode_t fmode)
2209 {
2210         spin_lock(&state->owner->so_lock);
2211         if (!(fmode & FMODE_READ))
2212                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2213         if (!(fmode & FMODE_WRITE))
2214                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2215         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2216         spin_unlock(&state->owner->so_lock);
2217 }
2218
2219 static void nfs4_close_done(struct rpc_task *task, void *data)
2220 {
2221         struct nfs4_closedata *calldata = data;
2222         struct nfs4_state *state = calldata->state;
2223         struct nfs_server *server = NFS_SERVER(calldata->inode);
2224
2225         dprintk("%s: begin!\n", __func__);
2226         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2227                 return;
2228         /* hmm. we are done with the inode, and in the process of freeing
2229          * the state_owner. we keep this around to process errors
2230          */
2231         switch (task->tk_status) {
2232                 case 0:
2233                         if (calldata->roc)
2234                                 pnfs_roc_set_barrier(state->inode,
2235                                                      calldata->roc_barrier);
2236                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2237                         renew_lease(server, calldata->timestamp);
2238                         nfs4_close_clear_stateid_flags(state,
2239                                         calldata->arg.fmode);
2240                         break;
2241                 case -NFS4ERR_STALE_STATEID:
2242                 case -NFS4ERR_OLD_STATEID:
2243                 case -NFS4ERR_BAD_STATEID:
2244                 case -NFS4ERR_EXPIRED:
2245                         if (calldata->arg.fmode == 0)
2246                                 break;
2247                 default:
2248                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2249                                 rpc_restart_call_prepare(task);
2250         }
2251         nfs_release_seqid(calldata->arg.seqid);
2252         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2253         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2254 }
2255
2256 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2257 {
2258         struct nfs4_closedata *calldata = data;
2259         struct nfs4_state *state = calldata->state;
2260         struct inode *inode = calldata->inode;
2261         int call_close = 0;
2262
2263         dprintk("%s: begin!\n", __func__);
2264         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2265                 return;
2266
2267         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2268         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2269         spin_lock(&state->owner->so_lock);
2270         /* Calculate the change in open mode */
2271         if (state->n_rdwr == 0) {
2272                 if (state->n_rdonly == 0) {
2273                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2274                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2275                         calldata->arg.fmode &= ~FMODE_READ;
2276                 }
2277                 if (state->n_wronly == 0) {
2278                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2279                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2280                         calldata->arg.fmode &= ~FMODE_WRITE;
2281                 }
2282         }
2283         spin_unlock(&state->owner->so_lock);
2284
2285         if (!call_close) {
2286                 /* Note: exit _without_ calling nfs4_close_done */
2287                 task->tk_action = NULL;
2288                 goto out;
2289         }
2290
2291         if (calldata->arg.fmode == 0) {
2292                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2293                 if (calldata->roc &&
2294                     pnfs_roc_drain(inode, &calldata->roc_barrier, task))
2295                         goto out;
2296         }
2297
2298         nfs_fattr_init(calldata->res.fattr);
2299         calldata->timestamp = jiffies;
2300         if (nfs4_setup_sequence(NFS_SERVER(inode),
2301                                 &calldata->arg.seq_args,
2302                                 &calldata->res.seq_res,
2303                                 task) != 0)
2304                 nfs_release_seqid(calldata->arg.seqid);
2305         else
2306                 rpc_call_start(task);
2307 out:
2308         dprintk("%s: done!\n", __func__);
2309 }
2310
2311 static const struct rpc_call_ops nfs4_close_ops = {
2312         .rpc_call_prepare = nfs4_close_prepare,
2313         .rpc_call_done = nfs4_close_done,
2314         .rpc_release = nfs4_free_closedata,
2315 };
2316
2317 /* 
2318  * It is possible for data to be read/written from a mem-mapped file 
2319  * after the sys_close call (which hits the vfs layer as a flush).
2320  * This means that we can't safely call nfsv4 close on a file until 
2321  * the inode is cleared. This in turn means that we are not good
2322  * NFSv4 citizens - we do not indicate to the server to update the file's 
2323  * share state even when we are done with one of the three share 
2324  * stateid's in the inode.
2325  *
2326  * NOTE: Caller must be holding the sp->so_owner semaphore!
2327  */
2328 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2329 {
2330         struct nfs_server *server = NFS_SERVER(state->inode);
2331         struct nfs4_closedata *calldata;
2332         struct nfs4_state_owner *sp = state->owner;
2333         struct rpc_task *task;
2334         struct rpc_message msg = {
2335                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2336                 .rpc_cred = state->owner->so_cred,
2337         };
2338         struct rpc_task_setup task_setup_data = {
2339                 .rpc_client = server->client,
2340                 .rpc_message = &msg,
2341                 .callback_ops = &nfs4_close_ops,
2342                 .workqueue = nfsiod_workqueue,
2343                 .flags = RPC_TASK_ASYNC,
2344         };
2345         int status = -ENOMEM;
2346
2347         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2348         if (calldata == NULL)
2349                 goto out;
2350         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2351         calldata->inode = state->inode;
2352         calldata->state = state;
2353         calldata->arg.fh = NFS_FH(state->inode);
2354         calldata->arg.stateid = &state->open_stateid;
2355         /* Serialization for the sequence id */
2356         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2357         if (calldata->arg.seqid == NULL)
2358                 goto out_free_calldata;
2359         calldata->arg.fmode = 0;
2360         calldata->arg.bitmask = server->cache_consistency_bitmask;
2361         calldata->res.fattr = &calldata->fattr;
2362         calldata->res.seqid = calldata->arg.seqid;
2363         calldata->res.server = server;
2364         calldata->roc = pnfs_roc(state->inode);
2365         nfs_sb_active(calldata->inode->i_sb);
2366
2367         msg.rpc_argp = &calldata->arg;
2368         msg.rpc_resp = &calldata->res;
2369         task_setup_data.callback_data = calldata;
2370         task = rpc_run_task(&task_setup_data);
2371         if (IS_ERR(task))
2372                 return PTR_ERR(task);
2373         status = 0;
2374         if (wait)
2375                 status = rpc_wait_for_completion_task(task);
2376         rpc_put_task(task);
2377         return status;
2378 out_free_calldata:
2379         kfree(calldata);
2380 out:
2381         nfs4_put_open_state(state);
2382         nfs4_put_state_owner(sp);
2383         return status;
2384 }
2385
2386 static struct inode *
2387 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2388 {
2389         struct nfs4_state *state;
2390
2391         /* Protect against concurrent sillydeletes */
2392         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2393                              ctx->cred, &ctx->mdsthreshold);
2394         if (IS_ERR(state))
2395                 return ERR_CAST(state);
2396         ctx->state = state;
2397         return igrab(state->inode);
2398 }
2399
2400 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2401 {
2402         if (ctx->state == NULL)
2403                 return;
2404         if (is_sync)
2405                 nfs4_close_sync(ctx->state, ctx->mode);
2406         else
2407                 nfs4_close_state(ctx->state, ctx->mode);
2408 }
2409
2410 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2411 {
2412         struct nfs4_server_caps_arg args = {
2413                 .fhandle = fhandle,
2414         };
2415         struct nfs4_server_caps_res res = {};
2416         struct rpc_message msg = {
2417                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2418                 .rpc_argp = &args,
2419                 .rpc_resp = &res,
2420         };
2421         int status;
2422
2423         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2424         if (status == 0) {
2425                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2426                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2427                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2428                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2429                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2430                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2431                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2432                         server->caps |= NFS_CAP_ACLS;
2433                 if (res.has_links != 0)
2434                         server->caps |= NFS_CAP_HARDLINKS;
2435                 if (res.has_symlinks != 0)
2436                         server->caps |= NFS_CAP_SYMLINKS;
2437                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2438                         server->caps |= NFS_CAP_FILEID;
2439                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2440                         server->caps |= NFS_CAP_MODE;
2441                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2442                         server->caps |= NFS_CAP_NLINK;
2443                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2444                         server->caps |= NFS_CAP_OWNER;
2445                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2446                         server->caps |= NFS_CAP_OWNER_GROUP;
2447                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2448                         server->caps |= NFS_CAP_ATIME;
2449                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2450                         server->caps |= NFS_CAP_CTIME;
2451                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2452                         server->caps |= NFS_CAP_MTIME;
2453
2454                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2455                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2456                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2457                 server->acl_bitmask = res.acl_bitmask;
2458                 server->fh_expire_type = res.fh_expire_type;
2459         }
2460
2461         return status;
2462 }
2463
2464 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2465 {
2466         struct nfs4_exception exception = { };
2467         int err;
2468         do {
2469                 err = nfs4_handle_exception(server,
2470                                 _nfs4_server_capabilities(server, fhandle),
2471                                 &exception);
2472         } while (exception.retry);
2473         return err;
2474 }
2475
2476 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2477                 struct nfs_fsinfo *info)
2478 {
2479         struct nfs4_lookup_root_arg args = {
2480                 .bitmask = nfs4_fattr_bitmap,
2481         };
2482         struct nfs4_lookup_res res = {
2483                 .server = server,
2484                 .fattr = info->fattr,
2485                 .fh = fhandle,
2486         };
2487         struct rpc_message msg = {
2488                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2489                 .rpc_argp = &args,
2490                 .rpc_resp = &res,
2491         };
2492
2493         nfs_fattr_init(info->fattr);
2494         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2495 }
2496
2497 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2498                 struct nfs_fsinfo *info)
2499 {
2500         struct nfs4_exception exception = { };
2501         int err;
2502         do {
2503                 err = _nfs4_lookup_root(server, fhandle, info);
2504                 switch (err) {
2505                 case 0:
2506                 case -NFS4ERR_WRONGSEC:
2507                         goto out;
2508                 default:
2509                         err = nfs4_handle_exception(server, err, &exception);
2510                 }
2511         } while (exception.retry);
2512 out:
2513         return err;
2514 }
2515
2516 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2517                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2518 {
2519         struct rpc_auth *auth;
2520         int ret;
2521
2522         auth = rpcauth_create(flavor, server->client);
2523         if (IS_ERR(auth)) {
2524                 ret = -EIO;
2525                 goto out;
2526         }
2527         ret = nfs4_lookup_root(server, fhandle, info);
2528 out:
2529         return ret;
2530 }
2531
2532 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2533                               struct nfs_fsinfo *info)
2534 {
2535         int i, len, status = 0;
2536         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2537
2538         len = rpcauth_list_flavors(flav_array, ARRAY_SIZE(flav_array));
2539         if (len < 0)
2540                 return len;
2541
2542         for (i = 0; i < len; i++) {
2543                 /* AUTH_UNIX is the default flavor if none was specified,
2544                  * thus has already been tried. */
2545                 if (flav_array[i] == RPC_AUTH_UNIX)
2546                         continue;
2547
2548                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2549                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2550                         continue;
2551                 break;
2552         }
2553         /*
2554          * -EACCESS could mean that the user doesn't have correct permissions
2555          * to access the mount.  It could also mean that we tried to mount
2556          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2557          * existing mount programs don't handle -EACCES very well so it should
2558          * be mapped to -EPERM instead.
2559          */
2560         if (status == -EACCES)
2561                 status = -EPERM;
2562         return status;
2563 }
2564
2565 /*
2566  * get the file handle for the "/" directory on the server
2567  */
2568 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2569                          struct nfs_fsinfo *info)
2570 {
2571         int minor_version = server->nfs_client->cl_minorversion;
2572         int status = nfs4_lookup_root(server, fhandle, info);
2573         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2574                 /*
2575                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2576                  * by nfs4_map_errors() as this function exits.
2577                  */
2578                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2579         if (status == 0)
2580                 status = nfs4_server_capabilities(server, fhandle);
2581         if (status == 0)
2582                 status = nfs4_do_fsinfo(server, fhandle, info);
2583         return nfs4_map_errors(status);
2584 }
2585
2586 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2587                               struct nfs_fsinfo *info)
2588 {
2589         int error;
2590         struct nfs_fattr *fattr = info->fattr;
2591
2592         error = nfs4_server_capabilities(server, mntfh);
2593         if (error < 0) {
2594                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2595                 return error;
2596         }
2597
2598         error = nfs4_proc_getattr(server, mntfh, fattr);
2599         if (error < 0) {
2600                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2601                 return error;
2602         }
2603
2604         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2605             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2606                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2607
2608         return error;
2609 }
2610
2611 /*
2612  * Get locations and (maybe) other attributes of a referral.
2613  * Note that we'll actually follow the referral later when
2614  * we detect fsid mismatch in inode revalidation
2615  */
2616 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2617                              const struct qstr *name, struct nfs_fattr *fattr,
2618                              struct nfs_fh *fhandle)
2619 {
2620         int status = -ENOMEM;
2621         struct page *page = NULL;
2622         struct nfs4_fs_locations *locations = NULL;
2623
2624         page = alloc_page(GFP_KERNEL);
2625         if (page == NULL)
2626                 goto out;
2627         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2628         if (locations == NULL)
2629                 goto out;
2630
2631         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2632         if (status != 0)
2633                 goto out;
2634         /* Make sure server returned a different fsid for the referral */
2635         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2636                 dprintk("%s: server did not return a different fsid for"
2637                         " a referral at %s\n", __func__, name->name);
2638                 status = -EIO;
2639                 goto out;
2640         }
2641         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2642         nfs_fixup_referral_attributes(&locations->fattr);
2643
2644         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2645         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2646         memset(fhandle, 0, sizeof(struct nfs_fh));
2647 out:
2648         if (page)
2649                 __free_page(page);
2650         kfree(locations);
2651         return status;
2652 }
2653
2654 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2655 {
2656         struct nfs4_getattr_arg args = {
2657                 .fh = fhandle,
2658                 .bitmask = server->attr_bitmask,
2659         };
2660         struct nfs4_getattr_res res = {
2661                 .fattr = fattr,
2662                 .server = server,
2663         };
2664         struct rpc_message msg = {
2665                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2666                 .rpc_argp = &args,
2667                 .rpc_resp = &res,
2668         };
2669         
2670         nfs_fattr_init(fattr);
2671         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2672 }
2673
2674 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2675 {
2676         struct nfs4_exception exception = { };
2677         int err;
2678         do {
2679                 err = nfs4_handle_exception(server,
2680                                 _nfs4_proc_getattr(server, fhandle, fattr),
2681                                 &exception);
2682         } while (exception.retry);
2683         return err;
2684 }
2685
2686 /* 
2687  * The file is not closed if it is opened due to the a request to change
2688  * the size of the file. The open call will not be needed once the
2689  * VFS layer lookup-intents are implemented.
2690  *
2691  * Close is called when the inode is destroyed.
2692  * If we haven't opened the file for O_WRONLY, we
2693  * need to in the size_change case to obtain a stateid.
2694  *
2695  * Got race?
2696  * Because OPEN is always done by name in nfsv4, it is
2697  * possible that we opened a different file by the same
2698  * name.  We can recognize this race condition, but we
2699  * can't do anything about it besides returning an error.
2700  *
2701  * This will be fixed with VFS changes (lookup-intent).
2702  */
2703 static int
2704 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2705                   struct iattr *sattr)
2706 {
2707         struct inode *inode = dentry->d_inode;
2708         struct rpc_cred *cred = NULL;
2709         struct nfs4_state *state = NULL;
2710         int status;
2711
2712         if (pnfs_ld_layoutret_on_setattr(inode))
2713                 pnfs_return_layout(inode);
2714
2715         nfs_fattr_init(fattr);
2716         
2717         /* Deal with open(O_TRUNC) */
2718         if (sattr->ia_valid & ATTR_OPEN)
2719                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2720
2721         /* Optimization: if the end result is no change, don't RPC */
2722         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2723                 return 0;
2724
2725         /* Search for an existing open(O_WRITE) file */
2726         if (sattr->ia_valid & ATTR_FILE) {
2727                 struct nfs_open_context *ctx;
2728
2729                 ctx = nfs_file_open_context(sattr->ia_file);
2730                 if (ctx) {
2731                         cred = ctx->cred;
2732                         state = ctx->state;
2733                 }
2734         }
2735
2736         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2737         if (status == 0)
2738                 nfs_setattr_update_inode(inode, sattr);
2739         return status;
2740 }
2741
2742 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2743                 const struct qstr *name, struct nfs_fh *fhandle,
2744                 struct nfs_fattr *fattr)
2745 {
2746         struct nfs_server *server = NFS_SERVER(dir);
2747         int                    status;
2748         struct nfs4_lookup_arg args = {
2749                 .bitmask = server->attr_bitmask,
2750                 .dir_fh = NFS_FH(dir),
2751                 .name = name,
2752         };
2753         struct nfs4_lookup_res res = {
2754                 .server = server,
2755                 .fattr = fattr,
2756                 .fh = fhandle,
2757         };
2758         struct rpc_message msg = {
2759                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2760                 .rpc_argp = &args,
2761                 .rpc_resp = &res,
2762         };
2763
2764         nfs_fattr_init(fattr);
2765
2766         dprintk("NFS call  lookup %s\n", name->name);
2767         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2768         dprintk("NFS reply lookup: %d\n", status);
2769         return status;
2770 }
2771
2772 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2773 {
2774         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2775                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2776         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2777         fattr->nlink = 2;
2778 }
2779
2780 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2781                                    struct qstr *name, struct nfs_fh *fhandle,
2782                                    struct nfs_fattr *fattr)
2783 {
2784         struct nfs4_exception exception = { };
2785         struct rpc_clnt *client = *clnt;
2786         int err;
2787         do {
2788                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2789                 switch (err) {
2790                 case -NFS4ERR_BADNAME:
2791                         err = -ENOENT;
2792                         goto out;
2793                 case -NFS4ERR_MOVED:
2794                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2795                         goto out;
2796                 case -NFS4ERR_WRONGSEC:
2797                         err = -EPERM;
2798                         if (client != *clnt)
2799                                 goto out;
2800
2801                         client = nfs4_create_sec_client(client, dir, name);
2802                         if (IS_ERR(client))
2803                                 return PTR_ERR(client);
2804
2805                         exception.retry = 1;
2806                         break;
2807                 default:
2808                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2809                 }
2810         } while (exception.retry);
2811
2812 out:
2813         if (err == 0)
2814                 *clnt = client;
2815         else if (client != *clnt)
2816                 rpc_shutdown_client(client);
2817
2818         return err;
2819 }
2820
2821 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2822                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2823 {
2824         int status;
2825         struct rpc_clnt *client = NFS_CLIENT(dir);
2826
2827         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2828         if (client != NFS_CLIENT(dir)) {
2829                 rpc_shutdown_client(client);
2830                 nfs_fixup_secinfo_attributes(fattr);
2831         }
2832         return status;
2833 }
2834
2835 struct rpc_clnt *
2836 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2837                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2838 {
2839         int status;
2840         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2841
2842         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2843         if (status < 0) {
2844                 rpc_shutdown_client(client);
2845                 return ERR_PTR(status);
2846         }
2847         return client;
2848 }
2849
2850 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2851 {
2852         struct nfs_server *server = NFS_SERVER(inode);
2853         struct nfs4_accessargs args = {
2854                 .fh = NFS_FH(inode),
2855                 .bitmask = server->cache_consistency_bitmask,
2856         };
2857         struct nfs4_accessres res = {
2858                 .server = server,
2859         };
2860         struct rpc_message msg = {
2861                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2862                 .rpc_argp = &args,
2863                 .rpc_resp = &res,
2864                 .rpc_cred = entry->cred,
2865         };
2866         int mode = entry->mask;
2867         int status;
2868
2869         /*
2870          * Determine which access bits we want to ask for...
2871          */
2872         if (mode & MAY_READ)
2873                 args.access |= NFS4_ACCESS_READ;
2874         if (S_ISDIR(inode->i_mode)) {
2875                 if (mode & MAY_WRITE)
2876                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2877                 if (mode & MAY_EXEC)
2878                         args.access |= NFS4_ACCESS_LOOKUP;
2879         } else {
2880                 if (mode & MAY_WRITE)
2881                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2882                 if (mode & MAY_EXEC)
2883                         args.access |= NFS4_ACCESS_EXECUTE;
2884         }
2885
2886         res.fattr = nfs_alloc_fattr();
2887         if (res.fattr == NULL)
2888                 return -ENOMEM;
2889
2890         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2891         if (!status) {
2892                 nfs_access_set_mask(entry, res.access);
2893                 nfs_refresh_inode(inode, res.fattr);
2894         }
2895         nfs_free_fattr(res.fattr);
2896         return status;
2897 }
2898
2899 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2900 {
2901         struct nfs4_exception exception = { };
2902         int err;
2903         do {
2904                 err = nfs4_handle_exception(NFS_SERVER(inode),
2905                                 _nfs4_proc_access(inode, entry),
2906                                 &exception);
2907         } while (exception.retry);
2908         return err;
2909 }
2910
2911 /*
2912  * TODO: For the time being, we don't try to get any attributes
2913  * along with any of the zero-copy operations READ, READDIR,
2914  * READLINK, WRITE.
2915  *
2916  * In the case of the first three, we want to put the GETATTR
2917  * after the read-type operation -- this is because it is hard
2918  * to predict the length of a GETATTR response in v4, and thus
2919  * align the READ data correctly.  This means that the GETATTR
2920  * may end up partially falling into the page cache, and we should
2921  * shift it into the 'tail' of the xdr_buf before processing.
2922  * To do this efficiently, we need to know the total length
2923  * of data received, which doesn't seem to be available outside
2924  * of the RPC layer.
2925  *
2926  * In the case of WRITE, we also want to put the GETATTR after
2927  * the operation -- in this case because we want to make sure
2928  * we get the post-operation mtime and size.
2929  *
2930  * Both of these changes to the XDR layer would in fact be quite
2931  * minor, but I decided to leave them for a subsequent patch.
2932  */
2933 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2934                 unsigned int pgbase, unsigned int pglen)
2935 {
2936         struct nfs4_readlink args = {
2937                 .fh       = NFS_FH(inode),
2938                 .pgbase   = pgbase,
2939                 .pglen    = pglen,
2940                 .pages    = &page,
2941         };
2942         struct nfs4_readlink_res res;
2943         struct rpc_message msg = {
2944                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2945                 .rpc_argp = &args,
2946                 .rpc_resp = &res,
2947         };
2948
2949         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2950 }
2951
2952 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2953                 unsigned int pgbase, unsigned int pglen)
2954 {
2955         struct nfs4_exception exception = { };
2956         int err;
2957         do {
2958                 err = nfs4_handle_exception(NFS_SERVER(inode),
2959                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2960                                 &exception);
2961         } while (exception.retry);
2962         return err;
2963 }
2964
2965 /*
2966  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2967  */
2968 static int
2969 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2970                  int flags)
2971 {
2972         struct nfs_open_context *ctx;
2973         struct nfs4_state *state;
2974         int status = 0;
2975
2976         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2977         if (IS_ERR(ctx))
2978                 return PTR_ERR(ctx);
2979
2980         sattr->ia_mode &= ~current_umask();
2981         state = nfs4_do_open(dir, dentry, ctx->mode,
2982                         flags, sattr, ctx->cred,
2983                         &ctx->mdsthreshold);
2984         d_drop(dentry);
2985         if (IS_ERR(state)) {
2986                 status = PTR_ERR(state);
2987                 goto out;
2988         }
2989         d_add(dentry, igrab(state->inode));
2990         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2991         ctx->state = state;
2992 out:
2993         put_nfs_open_context(ctx);
2994         return status;
2995 }
2996
2997 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2998 {
2999         struct nfs_server *server = NFS_SERVER(dir);
3000         struct nfs_removeargs args = {
3001                 .fh = NFS_FH(dir),
3002                 .name = *name,
3003         };
3004         struct nfs_removeres res = {
3005                 .server = server,
3006         };
3007         struct rpc_message msg = {
3008                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3009                 .rpc_argp = &args,
3010                 .rpc_resp = &res,
3011         };
3012         int status;
3013
3014         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3015         if (status == 0)
3016                 update_changeattr(dir, &res.cinfo);
3017         return status;
3018 }
3019
3020 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3021 {
3022         struct nfs4_exception exception = { };
3023         int err;
3024         do {
3025                 err = nfs4_handle_exception(NFS_SERVER(dir),
3026                                 _nfs4_proc_remove(dir, name),
3027                                 &exception);
3028         } while (exception.retry);
3029         return err;
3030 }
3031
3032 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3033 {
3034         struct nfs_server *server = NFS_SERVER(dir);
3035         struct nfs_removeargs *args = msg->rpc_argp;
3036         struct nfs_removeres *res = msg->rpc_resp;
3037
3038         res->server = server;
3039         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3040         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3041 }
3042
3043 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3044 {
3045         if (nfs4_setup_sequence(NFS_SERVER(data->dir),
3046                                 &data->args.seq_args,
3047                                 &data->res.seq_res,
3048                                 task))
3049                 return;
3050         rpc_call_start(task);
3051 }
3052
3053 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3054 {
3055         struct nfs_removeres *res = task->tk_msg.rpc_resp;
3056
3057         if (!nfs4_sequence_done(task, &res->seq_res))
3058                 return 0;
3059         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3060                 return 0;
3061         update_changeattr(dir, &res->cinfo);
3062         return 1;
3063 }
3064
3065 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3066 {
3067         struct nfs_server *server = NFS_SERVER(dir);
3068         struct nfs_renameargs *arg = msg->rpc_argp;
3069         struct nfs_renameres *res = msg->rpc_resp;
3070
3071         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3072         res->server = server;
3073         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3074 }
3075
3076 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3077 {
3078         if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3079                                 &data->args.seq_args,
3080                                 &data->res.seq_res,
3081                                 task))
3082                 return;
3083         rpc_call_start(task);
3084 }
3085
3086 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3087                                  struct inode *new_dir)
3088 {
3089         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3090
3091         if (!nfs4_sequence_done(task, &res->seq_res))
3092                 return 0;
3093         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3094                 return 0;
3095
3096         update_changeattr(old_dir, &res->old_cinfo);
3097         update_changeattr(new_dir, &res->new_cinfo);
3098         return 1;
3099 }
3100
3101 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3102                 struct inode *new_dir, struct qstr *new_name)
3103 {
3104         struct nfs_server *server = NFS_SERVER(old_dir);
3105         struct nfs_renameargs arg = {
3106                 .old_dir = NFS_FH(old_dir),
3107                 .new_dir = NFS_FH(new_dir),
3108                 .old_name = old_name,
3109                 .new_name = new_name,
3110         };
3111         struct nfs_renameres res = {
3112                 .server = server,
3113         };
3114         struct rpc_message msg = {
3115                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3116                 .rpc_argp = &arg,
3117                 .rpc_resp = &res,
3118         };
3119         int status = -ENOMEM;
3120         
3121         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3122         if (!status) {
3123                 update_changeattr(old_dir, &res.old_cinfo);
3124                 update_changeattr(new_dir, &res.new_cinfo);
3125         }
3126         return status;
3127 }
3128
3129 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3130                 struct inode *new_dir, struct qstr *new_name)
3131 {
3132         struct nfs4_exception exception = { };
3133         int err;
3134         do {
3135                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3136                                 _nfs4_proc_rename(old_dir, old_name,
3137                                         new_dir, new_name),
3138                                 &exception);
3139         } while (exception.retry);
3140         return err;
3141 }
3142
3143 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3144 {
3145         struct nfs_server *server = NFS_SERVER(inode);
3146         struct nfs4_link_arg arg = {
3147                 .fh     = NFS_FH(inode),
3148                 .dir_fh = NFS_FH(dir),
3149                 .name   = name,
3150                 .bitmask = server->attr_bitmask,
3151         };
3152         struct nfs4_link_res res = {
3153                 .server = server,
3154         };
3155         struct rpc_message msg = {
3156                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3157                 .rpc_argp = &arg,
3158                 .rpc_resp = &res,
3159         };
3160         int status = -ENOMEM;
3161
3162         res.fattr = nfs_alloc_fattr();
3163         if (res.fattr == NULL)
3164                 goto out;
3165
3166         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3167         if (!status) {
3168                 update_changeattr(dir, &res.cinfo);
3169                 nfs_post_op_update_inode(inode, res.fattr);
3170         }
3171 out:
3172         nfs_free_fattr(res.fattr);
3173         return status;
3174 }
3175
3176 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3177 {
3178         struct nfs4_exception exception = { };
3179         int err;
3180         do {
3181                 err = nfs4_handle_exception(NFS_SERVER(inode),
3182                                 _nfs4_proc_link(inode, dir, name),
3183                                 &exception);
3184         } while (exception.retry);
3185         return err;
3186 }
3187
3188 struct nfs4_createdata {
3189         struct rpc_message msg;
3190         struct nfs4_create_arg arg;
3191         struct nfs4_create_res res;
3192         struct nfs_fh fh;
3193         struct nfs_fattr fattr;
3194 };
3195
3196 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3197                 struct qstr *name, struct iattr *sattr, u32 ftype)
3198 {
3199         struct nfs4_createdata *data;
3200
3201         data = kzalloc(sizeof(*data), GFP_KERNEL);
3202         if (data != NULL) {
3203                 struct nfs_server *server = NFS_SERVER(dir);
3204
3205                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3206                 data->msg.rpc_argp = &data->arg;
3207                 data->msg.rpc_resp = &data->res;
3208                 data->arg.dir_fh = NFS_FH(dir);
3209                 data->arg.server = server;
3210                 data->arg.name = name;
3211                 data->arg.attrs = sattr;
3212                 data->arg.ftype = ftype;
3213                 data->arg.bitmask = server->attr_bitmask;
3214                 data->res.server = server;
3215                 data->res.fh = &data->fh;
3216                 data->res.fattr = &data->fattr;
3217                 nfs_fattr_init(data->res.fattr);
3218         }
3219         return data;
3220 }
3221
3222 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3223 {
3224         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3225                                     &data->arg.seq_args, &data->res.seq_res, 1);
3226         if (status == 0) {
3227                 update_changeattr(dir, &data->res.dir_cinfo);
3228                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3229         }
3230         return status;
3231 }
3232
3233 static void nfs4_free_createdata(struct nfs4_createdata *data)
3234 {
3235         kfree(data);
3236 }
3237
3238 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3239                 struct page *page, unsigned int len, struct iattr *sattr)
3240 {
3241         struct nfs4_createdata *data;
3242         int status = -ENAMETOOLONG;
3243
3244         if (len > NFS4_MAXPATHLEN)
3245                 goto out;
3246
3247         status = -ENOMEM;
3248         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3249         if (data == NULL)
3250                 goto out;
3251
3252         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3253         data->arg.u.symlink.pages = &page;
3254         data->arg.u.symlink.len = len;
3255         
3256         status = nfs4_do_create(dir, dentry, data);
3257
3258         nfs4_free_createdata(data);
3259 out:
3260         return status;
3261 }
3262
3263 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3264                 struct page *page, unsigned int len, struct iattr *sattr)
3265 {
3266         struct nfs4_exception exception = { };
3267         int err;
3268         do {
3269                 err = nfs4_handle_exception(NFS_SERVER(dir),
3270                                 _nfs4_proc_symlink(dir, dentry, page,
3271                                                         len, sattr),
3272                                 &exception);
3273         } while (exception.retry);
3274         return err;
3275 }
3276
3277 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3278                 struct iattr *sattr)
3279 {
3280         struct nfs4_createdata *data;
3281         int status = -ENOMEM;
3282
3283         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3284         if (data == NULL)
3285                 goto out;
3286
3287         status = nfs4_do_create(dir, dentry, data);
3288
3289         nfs4_free_createdata(data);
3290 out:
3291         return status;
3292 }
3293
3294 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3295                 struct iattr *sattr)
3296 {
3297         struct nfs4_exception exception = { };
3298         int err;
3299
3300         sattr->ia_mode &= ~current_umask();
3301         do {
3302                 err = nfs4_handle_exception(NFS_SERVER(dir),
3303                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3304                                 &exception);
3305         } while (exception.retry);
3306         return err;
3307 }
3308
3309 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3310                 u64 cookie, struct page **pages, unsigned int count, int plus)
3311 {
3312         struct inode            *dir = dentry->d_inode;
3313         struct nfs4_readdir_arg args = {
3314                 .fh = NFS_FH(dir),
3315                 .pages = pages,
3316                 .pgbase = 0,
3317                 .count = count,
3318                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3319                 .plus = plus,
3320         };
3321         struct nfs4_readdir_res res;
3322         struct rpc_message msg = {
3323                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3324                 .rpc_argp = &args,
3325                 .rpc_resp = &res,
3326                 .rpc_cred = cred,
3327         };
3328         int                     status;
3329
3330         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3331                         dentry->d_parent->d_name.name,
3332                         dentry->d_name.name,
3333                         (unsigned long long)cookie);
3334         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3335         res.pgbase = args.pgbase;
3336         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3337         if (status >= 0) {
3338                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3339                 status += args.pgbase;
3340         }
3341
3342         nfs_invalidate_atime(dir);
3343
3344         dprintk("%s: returns %d\n", __func__, status);
3345         return status;
3346 }
3347
3348 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3349                 u64 cookie, struct page **pages, unsigned int count, int plus)
3350 {
3351         struct nfs4_exception exception = { };
3352         int err;
3353         do {
3354                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3355                                 _nfs4_proc_readdir(dentry, cred, cookie,
3356                                         pages, count, plus),
3357                                 &exception);
3358         } while (exception.retry);
3359         return err;
3360 }
3361
3362 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3363                 struct iattr *sattr, dev_t rdev)
3364 {
3365         struct nfs4_createdata *data;
3366         int mode = sattr->ia_mode;
3367         int status = -ENOMEM;
3368
3369         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3370         if (data == NULL)
3371                 goto out;
3372
3373         if (S_ISFIFO(mode))
3374                 data->arg.ftype = NF4FIFO;
3375         else if (S_ISBLK(mode)) {
3376                 data->arg.ftype = NF4BLK;
3377                 data->arg.u.device.specdata1 = MAJOR(rdev);
3378                 data->arg.u.device.specdata2 = MINOR(rdev);
3379         }
3380         else if (S_ISCHR(mode)) {
3381                 data->arg.ftype = NF4CHR;
3382                 data->arg.u.device.specdata1 = MAJOR(rdev);
3383                 data->arg.u.device.specdata2 = MINOR(rdev);
3384         } else if (!S_ISSOCK(mode)) {
3385                 status = -EINVAL;
3386                 goto out_free;
3387         }
3388         
3389         status = nfs4_do_create(dir, dentry, data);
3390 out_free:
3391         nfs4_free_createdata(data);
3392 out:
3393         return status;
3394 }
3395
3396 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3397                 struct iattr *sattr, dev_t rdev)
3398 {
3399         struct nfs4_exception exception = { };
3400         int err;
3401
3402         sattr->ia_mode &= ~current_umask();
3403         do {
3404                 err = nfs4_handle_exception(NFS_SERVER(dir),
3405                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3406                                 &exception);
3407         } while (exception.retry);
3408         return err;
3409 }
3410
3411 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3412                  struct nfs_fsstat *fsstat)
3413 {
3414         struct nfs4_statfs_arg args = {
3415                 .fh = fhandle,
3416                 .bitmask = server->attr_bitmask,
3417         };
3418         struct nfs4_statfs_res res = {
3419                 .fsstat = fsstat,
3420         };
3421         struct rpc_message msg = {
3422                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3423                 .rpc_argp = &args,
3424                 .rpc_resp = &res,
3425         };
3426
3427         nfs_fattr_init(fsstat->fattr);
3428         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3429 }
3430
3431 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3432 {
3433         struct nfs4_exception exception = { };
3434         int err;
3435         do {
3436                 err = nfs4_handle_exception(server,
3437                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3438                                 &exception);
3439         } while (exception.retry);
3440         return err;
3441 }
3442
3443 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3444                 struct nfs_fsinfo *fsinfo)
3445 {
3446         struct nfs4_fsinfo_arg args = {
3447                 .fh = fhandle,
3448                 .bitmask = server->attr_bitmask,
3449         };
3450         struct nfs4_fsinfo_res res = {
3451                 .fsinfo = fsinfo,
3452         };
3453         struct rpc_message msg = {
3454                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3455                 .rpc_argp = &args,
3456                 .rpc_resp = &res,
3457         };
3458
3459         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3460 }
3461
3462 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3463 {
3464         struct nfs4_exception exception = { };
3465         int err;
3466
3467         do {
3468                 err = nfs4_handle_exception(server,
3469                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3470                                 &exception);
3471         } while (exception.retry);
3472         return err;
3473 }
3474
3475 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3476 {
3477         int error;
3478
3479         nfs_fattr_init(fsinfo->fattr);
3480         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3481         if (error == 0) {
3482                 /* block layout checks this! */
3483                 server->pnfs_blksize = fsinfo->blksize;
3484                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3485         }
3486
3487         return error;
3488 }
3489
3490 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3491                 struct nfs_pathconf *pathconf)
3492 {
3493         struct nfs4_pathconf_arg args = {
3494                 .fh = fhandle,
3495                 .bitmask = server->attr_bitmask,
3496         };
3497         struct nfs4_pathconf_res res = {
3498                 .pathconf = pathconf,
3499         };
3500         struct rpc_message msg = {
3501                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3502                 .rpc_argp = &args,
3503                 .rpc_resp = &res,
3504         };
3505
3506         /* None of the pathconf attributes are mandatory to implement */
3507         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3508                 memset(pathconf, 0, sizeof(*pathconf));
3509                 return 0;
3510         }
3511
3512         nfs_fattr_init(pathconf->fattr);
3513         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3514 }
3515
3516 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3517                 struct nfs_pathconf *pathconf)
3518 {
3519         struct nfs4_exception exception = { };
3520         int err;
3521
3522         do {
3523                 err = nfs4_handle_exception(server,
3524                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3525                                 &exception);
3526         } while (exception.retry);
3527         return err;
3528 }
3529
3530 void __nfs4_read_done_cb(struct nfs_read_data *data)
3531 {
3532         nfs_invalidate_atime(data->header->inode);
3533 }
3534
3535 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3536 {
3537         struct nfs_server *server = NFS_SERVER(data->header->inode);
3538
3539         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3540                 rpc_restart_call_prepare(task);
3541                 return -EAGAIN;
3542         }
3543
3544         __nfs4_read_done_cb(data);
3545         if (task->tk_status > 0)
3546                 renew_lease(server, data->timestamp);
3547         return 0;
3548 }
3549
3550 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3551 {
3552
3553         dprintk("--> %s\n", __func__);
3554
3555         if (!nfs4_sequence_done(task, &data->res.seq_res))
3556                 return -EAGAIN;
3557
3558         return data->read_done_cb ? data->read_done_cb(task, data) :
3559                                     nfs4_read_done_cb(task, data);
3560 }
3561
3562 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3563 {
3564         data->timestamp   = jiffies;
3565         data->read_done_cb = nfs4_read_done_cb;
3566         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3567         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3568 }
3569
3570 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3571 {
3572         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3573                                 &data->args.seq_args,
3574                                 &data->res.seq_res,
3575                                 task))
3576                 return;
3577         rpc_call_start(task);
3578 }
3579
3580 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3581 {
3582         struct inode *inode = data->header->inode;
3583         
3584         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3585                 rpc_restart_call_prepare(task);
3586                 return -EAGAIN;
3587         }
3588         if (task->tk_status >= 0) {
3589                 renew_lease(NFS_SERVER(inode), data->timestamp);
3590                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3591         }
3592         return 0;
3593 }
3594
3595 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3596 {
3597         if (!nfs4_sequence_done(task, &data->res.seq_res))
3598                 return -EAGAIN;
3599         return data->write_done_cb ? data->write_done_cb(task, data) :
3600                 nfs4_write_done_cb(task, data);
3601 }
3602
3603 static
3604 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3605 {
3606         const struct nfs_pgio_header *hdr = data->header;
3607
3608         /* Don't request attributes for pNFS or O_DIRECT writes */
3609         if (data->ds_clp != NULL || hdr->dreq != NULL)
3610                 return false;
3611         /* Otherwise, request attributes if and only if we don't hold
3612          * a delegation
3613          */
3614         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3615 }
3616
3617 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3618 {
3619         struct nfs_server *server = NFS_SERVER(data->header->inode);
3620
3621         if (!nfs4_write_need_cache_consistency_data(data)) {
3622                 data->args.bitmask = NULL;
3623                 data->res.fattr = NULL;
3624         } else
3625                 data->args.bitmask = server->cache_consistency_bitmask;
3626
3627         if (!data->write_done_cb)
3628                 data->write_done_cb = nfs4_write_done_cb;
3629         data->res.server = server;
3630         data->timestamp   = jiffies;
3631
3632         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3633         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3634 }
3635
3636 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3637 {
3638         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3639                                 &data->args.seq_args,
3640                                 &data->res.seq_res,
3641                                 task))
3642                 return;
3643         rpc_call_start(task);
3644 }
3645
3646 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3647 {
3648         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3649                                 &data->args.seq_args,
3650                                 &data->res.seq_res,
3651                                 task))
3652                 return;
3653         rpc_call_start(task);
3654 }
3655
3656 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3657 {
3658         struct inode *inode = data->inode;
3659
3660         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3661                 rpc_restart_call_prepare(task);
3662                 return -EAGAIN;
3663         }
3664         return 0;
3665 }
3666
3667 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3668 {
3669         if (!nfs4_sequence_done(task, &data->res.seq_res))
3670                 return -EAGAIN;
3671         return data->commit_done_cb(task, data);
3672 }
3673
3674 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3675 {
3676         struct nfs_server *server = NFS_SERVER(data->inode);
3677
3678         if (data->commit_done_cb == NULL)
3679                 data->commit_done_cb = nfs4_commit_done_cb;
3680         data->res.server = server;
3681         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3682         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3683 }
3684
3685 struct nfs4_renewdata {
3686         struct nfs_client       *client;
3687         unsigned long           timestamp;
3688 };
3689
3690 /*
3691  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3692  * standalone procedure for queueing an asynchronous RENEW.
3693  */
3694 static void nfs4_renew_release(void *calldata)
3695 {
3696         struct nfs4_renewdata *data = calldata;
3697         struct nfs_client *clp = data->client;
3698
3699         if (atomic_read(&clp->cl_count) > 1)
3700                 nfs4_schedule_state_renewal(clp);
3701         nfs_put_client(clp);
3702         kfree(data);
3703 }
3704
3705 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3706 {
3707         struct nfs4_renewdata *data = calldata;
3708         struct nfs_client *clp = data->client;
3709         unsigned long timestamp = data->timestamp;
3710
3711         if (task->tk_status < 0) {
3712                 /* Unless we're shutting down, schedule state recovery! */
3713                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3714                         return;
3715                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3716                         nfs4_schedule_lease_recovery(clp);
3717                         return;
3718                 }
3719                 nfs4_schedule_path_down_recovery(clp);
3720         }
3721         do_renew_lease(clp, timestamp);
3722 }
3723
3724 static const struct rpc_call_ops nfs4_renew_ops = {
3725         .rpc_call_done = nfs4_renew_done,
3726         .rpc_release = nfs4_renew_release,
3727 };
3728
3729 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3730 {
3731         struct rpc_message msg = {
3732                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3733                 .rpc_argp       = clp,
3734                 .rpc_cred       = cred,
3735         };
3736         struct nfs4_renewdata *data;
3737
3738         if (renew_flags == 0)
3739                 return 0;
3740         if (!atomic_inc_not_zero(&clp->cl_count))
3741                 return -EIO;
3742         data = kmalloc(sizeof(*data), GFP_NOFS);
3743         if (data == NULL)
3744                 return -ENOMEM;
3745         data->client = clp;
3746         data->timestamp = jiffies;
3747         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3748                         &nfs4_renew_ops, data);
3749 }
3750
3751 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3752 {
3753         struct rpc_message msg = {
3754                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3755                 .rpc_argp       = clp,
3756                 .rpc_cred       = cred,
3757         };
3758         unsigned long now = jiffies;
3759         int status;
3760
3761         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3762         if (status < 0)
3763                 return status;
3764         do_renew_lease(clp, now);
3765         return 0;
3766 }
3767
3768 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3769 {
3770         return (server->caps & NFS_CAP_ACLS)
3771                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3772                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3773 }
3774
3775 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3776  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3777  * the stack.
3778  */
3779 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3780
3781 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3782                 struct page **pages, unsigned int *pgbase)
3783 {
3784         struct page *newpage, **spages;
3785         int rc = 0;
3786         size_t len;
3787         spages = pages;
3788
3789         do {
3790                 len = min_t(size_t, PAGE_SIZE, buflen);
3791                 newpage = alloc_page(GFP_KERNEL);
3792
3793                 if (newpage == NULL)
3794                         goto unwind;
3795                 memcpy(page_address(newpage), buf, len);
3796                 buf += len;
3797                 buflen -= len;
3798                 *pages++ = newpage;
3799                 rc++;
3800         } while (buflen != 0);
3801
3802         return rc;
3803
3804 unwind:
3805         for(; rc > 0; rc--)
3806                 __free_page(spages[rc-1]);
3807         return -ENOMEM;
3808 }
3809
3810 struct nfs4_cached_acl {
3811         int cached;
3812         size_t len;
3813         char data[0];
3814 };
3815
3816 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3817 {
3818         struct nfs_inode *nfsi = NFS_I(inode);
3819
3820         spin_lock(&inode->i_lock);
3821         kfree(nfsi->nfs4_acl);
3822         nfsi->nfs4_acl = acl;
3823         spin_unlock(&inode->i_lock);
3824 }
3825
3826 static void nfs4_zap_acl_attr(struct inode *inode)
3827 {
3828         nfs4_set_cached_acl(inode, NULL);
3829 }
3830
3831 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3832 {
3833         struct nfs_inode *nfsi = NFS_I(inode);
3834         struct nfs4_cached_acl *acl;
3835         int ret = -ENOENT;
3836
3837         spin_lock(&inode->i_lock);
3838         acl = nfsi->nfs4_acl;
3839         if (acl == NULL)
3840                 goto out;
3841         if (buf == NULL) /* user is just asking for length */
3842                 goto out_len;
3843         if (acl->cached == 0)
3844                 goto out;
3845         ret = -ERANGE; /* see getxattr(2) man page */
3846         if (acl->len > buflen)
3847                 goto out;
3848         memcpy(buf, acl->data, acl->len);
3849 out_len:
3850         ret = acl->len;
3851 out:
3852         spin_unlock(&inode->i_lock);
3853         return ret;
3854 }
3855
3856 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3857 {
3858         struct nfs4_cached_acl *acl;
3859         size_t buflen = sizeof(*acl) + acl_len;
3860
3861         if (buflen <= PAGE_SIZE) {
3862                 acl = kmalloc(buflen, GFP_KERNEL);
3863                 if (acl == NULL)
3864                         goto out;
3865                 acl->cached = 1;
3866                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3867         } else {
3868                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3869                 if (acl == NULL)
3870                         goto out;
3871                 acl->cached = 0;
3872         }
3873         acl->len = acl_len;
3874 out:
3875         nfs4_set_cached_acl(inode, acl);
3876 }
3877
3878 /*
3879  * The getxattr API returns the required buffer length when called with a
3880  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3881  * the required buf.  On a NULL buf, we send a page of data to the server
3882  * guessing that the ACL request can be serviced by a page. If so, we cache
3883  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3884  * the cache. If not so, we throw away the page, and cache the required
3885  * length. The next getxattr call will then produce another round trip to
3886  * the server, this time with the input buf of the required size.
3887  */
3888 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3889 {
3890         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3891         struct nfs_getaclargs args = {
3892                 .fh = NFS_FH(inode),
3893                 .acl_pages = pages,
3894                 .acl_len = buflen,
3895         };
3896         struct nfs_getaclres res = {
3897                 .acl_len = buflen,
3898         };
3899         struct rpc_message msg = {
3900                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3901                 .rpc_argp = &args,
3902                 .rpc_resp = &res,
3903         };
3904         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3905         int ret = -ENOMEM, i;
3906
3907         /* As long as we're doing a round trip to the server anyway,
3908          * let's be prepared for a page of acl data. */
3909         if (npages == 0)
3910                 npages = 1;
3911         if (npages > ARRAY_SIZE(pages))
3912                 return -ERANGE;
3913
3914         for (i = 0; i < npages; i++) {
3915                 pages[i] = alloc_page(GFP_KERNEL);
3916                 if (!pages[i])
3917                         goto out_free;
3918         }
3919
3920         /* for decoding across pages */
3921         res.acl_scratch = alloc_page(GFP_KERNEL);
3922         if (!res.acl_scratch)
3923                 goto out_free;
3924
3925         args.acl_len = npages * PAGE_SIZE;
3926         args.acl_pgbase = 0;
3927
3928         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3929                 __func__, buf, buflen, npages, args.acl_len);
3930         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3931                              &msg, &args.seq_args, &res.seq_res, 0);
3932         if (ret)
3933                 goto out_free;
3934
3935         /* Handle the case where the passed-in buffer is too short */
3936         if (res.acl_flags & NFS4_ACL_TRUNC) {
3937                 /* Did the user only issue a request for the acl length? */
3938                 if (buf == NULL)
3939                         goto out_ok;
3940                 ret = -ERANGE;
3941                 goto out_free;
3942         }
3943         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
3944         if (buf)
3945                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
3946 out_ok:
3947         ret = res.acl_len;
3948 out_free:
3949         for (i = 0; i < npages; i++)
3950                 if (pages[i])
3951                         __free_page(pages[i]);
3952         if (res.acl_scratch)
3953                 __free_page(res.acl_scratch);
3954         return ret;
3955 }
3956
3957 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3958 {
3959         struct nfs4_exception exception = { };
3960         ssize_t ret;
3961         do {
3962                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3963                 if (ret >= 0)
3964                         break;
3965                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3966         } while (exception.retry);
3967         return ret;
3968 }
3969
3970 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3971 {
3972         struct nfs_server *server = NFS_SERVER(inode);
3973         int ret;
3974
3975         if (!nfs4_server_supports_acls(server))
3976                 return -EOPNOTSUPP;
3977         ret = nfs_revalidate_inode(server, inode);
3978         if (ret < 0)
3979                 return ret;
3980         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3981                 nfs_zap_acl_cache(inode);
3982         ret = nfs4_read_cached_acl(inode, buf, buflen);
3983         if (ret != -ENOENT)
3984                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3985                  * but no cached acl data, just the acl length */
3986                 return ret;
3987         return nfs4_get_acl_uncached(inode, buf, buflen);
3988 }
3989
3990 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3991 {
3992         struct nfs_server *server = NFS_SERVER(inode);
3993         struct page *pages[NFS4ACL_MAXPAGES];
3994         struct nfs_setaclargs arg = {
3995                 .fh             = NFS_FH(inode),
3996                 .acl_pages      = pages,
3997                 .acl_len        = buflen,
3998         };
3999         struct nfs_setaclres res;
4000         struct rpc_message msg = {
4001                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4002                 .rpc_argp       = &arg,
4003                 .rpc_resp       = &res,
4004         };
4005         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4006         int ret, i;
4007
4008         if (!nfs4_server_supports_acls(server))
4009                 return -EOPNOTSUPP;
4010         if (npages > ARRAY_SIZE(pages))
4011                 return -ERANGE;
4012         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4013         if (i < 0)
4014                 return i;
4015         nfs4_inode_return_delegation(inode);
4016         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4017
4018         /*
4019          * Free each page after tx, so the only ref left is
4020          * held by the network stack
4021          */
4022         for (; i > 0; i--)
4023                 put_page(pages[i-1]);
4024
4025         /*
4026          * Acl update can result in inode attribute update.
4027          * so mark the attribute cache invalid.
4028          */
4029         spin_lock(&inode->i_lock);
4030         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4031         spin_unlock(&inode->i_lock);
4032         nfs_access_zap_cache(inode);
4033         nfs_zap_acl_cache(inode);
4034         return ret;
4035 }
4036
4037 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4038 {
4039         struct nfs4_exception exception = { };
4040         int err;
4041         do {
4042                 err = nfs4_handle_exception(NFS_SERVER(inode),
4043                                 __nfs4_proc_set_acl(inode, buf, buflen),
4044                                 &exception);
4045         } while (exception.retry);
4046         return err;
4047 }
4048
4049 static int
4050 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4051 {
4052         struct nfs_client *clp = server->nfs_client;
4053
4054         if (task->tk_status >= 0)
4055                 return 0;
4056         switch(task->tk_status) {
4057                 case -NFS4ERR_DELEG_REVOKED:
4058                 case -NFS4ERR_ADMIN_REVOKED:
4059                 case -NFS4ERR_BAD_STATEID:
4060                         if (state == NULL)
4061                                 break;
4062                         nfs_remove_bad_delegation(state->inode);
4063                 case -NFS4ERR_OPENMODE:
4064                         if (state == NULL)
4065                                 break;
4066                         nfs4_schedule_stateid_recovery(server, state);
4067                         goto wait_on_recovery;
4068                 case -NFS4ERR_EXPIRED:
4069                         if (state != NULL)
4070                                 nfs4_schedule_stateid_recovery(server, state);
4071                 case -NFS4ERR_STALE_STATEID:
4072                 case -NFS4ERR_STALE_CLIENTID:
4073                         nfs4_schedule_lease_recovery(clp);
4074                         goto wait_on_recovery;
4075 #if defined(CONFIG_NFS_V4_1)
4076                 case -NFS4ERR_BADSESSION:
4077                 case -NFS4ERR_BADSLOT:
4078                 case -NFS4ERR_BAD_HIGH_SLOT:
4079                 case -NFS4ERR_DEADSESSION:
4080                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4081                 case -NFS4ERR_SEQ_FALSE_RETRY:
4082                 case -NFS4ERR_SEQ_MISORDERED:
4083                         dprintk("%s ERROR %d, Reset session\n", __func__,
4084                                 task->tk_status);
4085                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4086                         task->tk_status = 0;
4087                         return -EAGAIN;
4088 #endif /* CONFIG_NFS_V4_1 */
4089                 case -NFS4ERR_DELAY:
4090                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4091                 case -NFS4ERR_GRACE:
4092                 case -EKEYEXPIRED:
4093                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4094                         task->tk_status = 0;
4095                         return -EAGAIN;
4096                 case -NFS4ERR_RETRY_UNCACHED_REP:
4097                 case -NFS4ERR_OLD_STATEID:
4098                         task->tk_status = 0;
4099                         return -EAGAIN;
4100         }
4101         task->tk_status = nfs4_map_errors(task->tk_status);
4102         return 0;
4103 wait_on_recovery:
4104         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4105         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4106                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4107         task->tk_status = 0;
4108         return -EAGAIN;
4109 }
4110
4111 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4112                                     nfs4_verifier *bootverf)
4113 {
4114         __be32 verf[2];
4115
4116         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4117                 /* An impossible timestamp guarantees this value
4118                  * will never match a generated boot time. */
4119                 verf[0] = 0;
4120                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4121         } else {
4122                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4123                 verf[0] = (__be32)nn->boot_time.tv_sec;
4124                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4125         }
4126         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4127 }
4128
4129 static unsigned int
4130 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4131                                    char *buf, size_t len)
4132 {
4133         unsigned int result;
4134
4135         rcu_read_lock();
4136         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4137                                 clp->cl_ipaddr,
4138                                 rpc_peeraddr2str(clp->cl_rpcclient,
4139                                                         RPC_DISPLAY_ADDR),
4140                                 rpc_peeraddr2str(clp->cl_rpcclient,
4141                                                         RPC_DISPLAY_PROTO));
4142         rcu_read_unlock();
4143         return result;
4144 }
4145
4146 static unsigned int
4147 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4148                                 char *buf, size_t len)
4149 {
4150         char *nodename = clp->cl_rpcclient->cl_nodename;
4151
4152         if (nfs4_client_id_uniquifier[0] != '\0')
4153                 nodename = nfs4_client_id_uniquifier;
4154         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4155                                 clp->rpc_ops->version, clp->cl_minorversion,
4156                                 nodename);
4157 }
4158
4159 /**
4160  * nfs4_proc_setclientid - Negotiate client ID
4161  * @clp: state data structure
4162  * @program: RPC program for NFSv4 callback service
4163  * @port: IP port number for NFS4 callback service
4164  * @cred: RPC credential to use for this call
4165  * @res: where to place the result
4166  *
4167  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4168  */
4169 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4170                 unsigned short port, struct rpc_cred *cred,
4171                 struct nfs4_setclientid_res *res)
4172 {
4173         nfs4_verifier sc_verifier;
4174         struct nfs4_setclientid setclientid = {
4175                 .sc_verifier = &sc_verifier,
4176                 .sc_prog = program,
4177                 .sc_cb_ident = clp->cl_cb_ident,
4178         };
4179         struct rpc_message msg = {
4180                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4181                 .rpc_argp = &setclientid,
4182                 .rpc_resp = res,
4183                 .rpc_cred = cred,
4184         };
4185         int status;
4186
4187         /* nfs_client_id4 */
4188         nfs4_init_boot_verifier(clp, &sc_verifier);
4189         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4190                 setclientid.sc_name_len =
4191                                 nfs4_init_uniform_client_string(clp,
4192                                                 setclientid.sc_name,
4193                                                 sizeof(setclientid.sc_name));
4194         else
4195                 setclientid.sc_name_len =
4196                                 nfs4_init_nonuniform_client_string(clp,
4197                                                 setclientid.sc_name,
4198                                                 sizeof(setclientid.sc_name));
4199         /* cb_client4 */
4200         rcu_read_lock();
4201         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4202                                 sizeof(setclientid.sc_netid),
4203                                 rpc_peeraddr2str(clp->cl_rpcclient,
4204                                                         RPC_DISPLAY_NETID));
4205         rcu_read_unlock();
4206         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4207                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4208                                 clp->cl_ipaddr, port >> 8, port & 255);
4209
4210         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4211                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4212                 setclientid.sc_name_len, setclientid.sc_name);
4213         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4214         dprintk("NFS reply setclientid: %d\n", status);
4215         return status;
4216 }
4217
4218 /**
4219  * nfs4_proc_setclientid_confirm - Confirm client ID
4220  * @clp: state data structure
4221  * @res: result of a previous SETCLIENTID
4222  * @cred: RPC credential to use for this call
4223  *
4224  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4225  */
4226 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4227                 struct nfs4_setclientid_res *arg,
4228                 struct rpc_cred *cred)
4229 {
4230         struct nfs_fsinfo fsinfo;
4231         struct rpc_message msg = {
4232                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4233                 .rpc_argp = arg,
4234                 .rpc_resp = &fsinfo,
4235                 .rpc_cred = cred,
4236         };
4237         unsigned long now;
4238         int status;
4239
4240         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4241                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4242                 clp->cl_clientid);
4243         now = jiffies;
4244         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4245         if (status == 0) {
4246                 spin_lock(&clp->cl_lock);
4247                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4248                 clp->cl_last_renewal = now;
4249                 spin_unlock(&clp->cl_lock);
4250         }
4251         dprintk("NFS reply setclientid_confirm: %d\n", status);
4252         return status;
4253 }
4254
4255 struct nfs4_delegreturndata {
4256         struct nfs4_delegreturnargs args;
4257         struct nfs4_delegreturnres res;
4258         struct nfs_fh fh;
4259         nfs4_stateid stateid;
4260         unsigned long timestamp;
4261         struct nfs_fattr fattr;
4262         int rpc_status;
4263 };
4264
4265 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4266 {
4267         struct nfs4_delegreturndata *data = calldata;
4268
4269         if (!nfs4_sequence_done(task, &data->res.seq_res))
4270                 return;
4271
4272         switch (task->tk_status) {
4273         case -NFS4ERR_STALE_STATEID:
4274         case -NFS4ERR_EXPIRED:
4275         case 0:
4276                 renew_lease(data->res.server, data->timestamp);
4277                 break;
4278         default:
4279                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4280                                 -EAGAIN) {
4281                         rpc_restart_call_prepare(task);
4282                         return;
4283                 }
4284         }
4285         data->rpc_status = task->tk_status;
4286 }
4287
4288 static void nfs4_delegreturn_release(void *calldata)
4289 {
4290         kfree(calldata);
4291 }
4292
4293 #if defined(CONFIG_NFS_V4_1)
4294 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4295 {
4296         struct nfs4_delegreturndata *d_data;
4297
4298         d_data = (struct nfs4_delegreturndata *)data;
4299
4300         if (nfs4_setup_sequence(d_data->res.server,
4301                                 &d_data->args.seq_args,
4302                                 &d_data->res.seq_res, task))
4303                 return;
4304         rpc_call_start(task);
4305 }
4306 #endif /* CONFIG_NFS_V4_1 */
4307
4308 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4309 #if defined(CONFIG_NFS_V4_1)
4310         .rpc_call_prepare = nfs4_delegreturn_prepare,
4311 #endif /* CONFIG_NFS_V4_1 */
4312         .rpc_call_done = nfs4_delegreturn_done,
4313         .rpc_release = nfs4_delegreturn_release,
4314 };
4315
4316 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4317 {
4318         struct nfs4_delegreturndata *data;
4319         struct nfs_server *server = NFS_SERVER(inode);
4320         struct rpc_task *task;
4321         struct rpc_message msg = {
4322                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4323                 .rpc_cred = cred,
4324         };
4325         struct rpc_task_setup task_setup_data = {
4326                 .rpc_client = server->client,
4327                 .rpc_message = &msg,
4328                 .callback_ops = &nfs4_delegreturn_ops,
4329                 .flags = RPC_TASK_ASYNC,
4330         };
4331         int status = 0;
4332
4333         data = kzalloc(sizeof(*data), GFP_NOFS);
4334         if (data == NULL)
4335                 return -ENOMEM;
4336         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4337         data->args.fhandle = &data->fh;
4338         data->args.stateid = &data->stateid;
4339         data->args.bitmask = server->cache_consistency_bitmask;
4340         nfs_copy_fh(&data->fh, NFS_FH(inode));
4341         nfs4_stateid_copy(&data->stateid, stateid);
4342         data->res.fattr = &data->fattr;
4343         data->res.server = server;
4344         nfs_fattr_init(data->res.fattr);
4345         data->timestamp = jiffies;
4346         data->rpc_status = 0;
4347
4348         task_setup_data.callback_data = data;
4349         msg.rpc_argp = &data->args;
4350         msg.rpc_resp = &data->res;
4351         task = rpc_run_task(&task_setup_data);
4352         if (IS_ERR(task))
4353                 return PTR_ERR(task);
4354         if (!issync)
4355                 goto out;
4356         status = nfs4_wait_for_completion_rpc_task(task);
4357         if (status != 0)
4358                 goto out;
4359         status = data->rpc_status;
4360         if (status == 0)
4361                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4362         else
4363                 nfs_refresh_inode(inode, &data->fattr);
4364 out:
4365         rpc_put_task(task);
4366         return status;
4367 }
4368
4369 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4370 {
4371         struct nfs_server *server = NFS_SERVER(inode);
4372         struct nfs4_exception exception = { };
4373         int err;
4374         do {
4375                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4376                 switch (err) {
4377                         case -NFS4ERR_STALE_STATEID:
4378                         case -NFS4ERR_EXPIRED:
4379                         case 0:
4380                                 return 0;
4381                 }
4382                 err = nfs4_handle_exception(server, err, &exception);
4383         } while (exception.retry);
4384         return err;
4385 }
4386
4387 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4388 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4389
4390 /* 
4391  * sleep, with exponential backoff, and retry the LOCK operation. 
4392  */
4393 static unsigned long
4394 nfs4_set_lock_task_retry(unsigned long timeout)
4395 {
4396         freezable_schedule_timeout_killable(timeout);
4397         timeout <<= 1;
4398         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4399                 return NFS4_LOCK_MAXTIMEOUT;
4400         return timeout;
4401 }
4402
4403 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4404 {
4405         struct inode *inode = state->inode;
4406         struct nfs_server *server = NFS_SERVER(inode);
4407         struct nfs_client *clp = server->nfs_client;
4408         struct nfs_lockt_args arg = {
4409                 .fh = NFS_FH(inode),
4410                 .fl = request,
4411         };
4412         struct nfs_lockt_res res = {
4413                 .denied = request,
4414         };
4415         struct rpc_message msg = {
4416                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4417                 .rpc_argp       = &arg,
4418                 .rpc_resp       = &res,
4419                 .rpc_cred       = state->owner->so_cred,
4420         };
4421         struct nfs4_lock_state *lsp;
4422         int status;
4423
4424         arg.lock_owner.clientid = clp->cl_clientid;
4425         status = nfs4_set_lock_state(state, request);
4426         if (status != 0)
4427                 goto out;
4428         lsp = request->fl_u.nfs4_fl.owner;
4429         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4430         arg.lock_owner.s_dev = server->s_dev;
4431         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4432         switch (status) {
4433                 case 0:
4434                         request->fl_type = F_UNLCK;
4435                         break;
4436                 case -NFS4ERR_DENIED:
4437                         status = 0;
4438         }
4439         request->fl_ops->fl_release_private(request);
4440 out:
4441         return status;
4442 }
4443
4444 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4445 {
4446         struct nfs4_exception exception = { };
4447         int err;
4448
4449         do {
4450                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4451                                 _nfs4_proc_getlk(state, cmd, request),
4452                                 &exception);
4453         } while (exception.retry);
4454         return err;
4455 }
4456
4457 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4458 {
4459         int res = 0;
4460         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4461                 case FL_POSIX:
4462                         res = posix_lock_file_wait(file, fl);
4463                         break;
4464                 case FL_FLOCK:
4465                         res = flock_lock_file_wait(file, fl);
4466                         break;
4467                 default:
4468                         BUG();
4469         }
4470         return res;
4471 }
4472
4473 struct nfs4_unlockdata {
4474         struct nfs_locku_args arg;
4475         struct nfs_locku_res res;
4476         struct nfs4_lock_state *lsp;
4477         struct nfs_open_context *ctx;
4478         struct file_lock fl;
4479         const struct nfs_server *server;
4480         unsigned long timestamp;
4481 };
4482
4483 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4484                 struct nfs_open_context *ctx,
4485                 struct nfs4_lock_state *lsp,
4486                 struct nfs_seqid *seqid)
4487 {
4488         struct nfs4_unlockdata *p;
4489         struct inode *inode = lsp->ls_state->inode;
4490
4491         p = kzalloc(sizeof(*p), GFP_NOFS);
4492         if (p == NULL)
4493                 return NULL;
4494         p->arg.fh = NFS_FH(inode);
4495         p->arg.fl = &p->fl;
4496         p->arg.seqid = seqid;
4497         p->res.seqid = seqid;
4498         p->arg.stateid = &lsp->ls_stateid;
4499         p->lsp = lsp;
4500         atomic_inc(&lsp->ls_count);
4501         /* Ensure we don't close file until we're done freeing locks! */
4502         p->ctx = get_nfs_open_context(ctx);
4503         memcpy(&p->fl, fl, sizeof(p->fl));
4504         p->server = NFS_SERVER(inode);
4505         return p;
4506 }
4507
4508 static void nfs4_locku_release_calldata(void *data)
4509 {
4510         struct nfs4_unlockdata *calldata = data;
4511         nfs_free_seqid(calldata->arg.seqid);
4512         nfs4_put_lock_state(calldata->lsp);
4513         put_nfs_open_context(calldata->ctx);
4514         kfree(calldata);
4515 }
4516
4517 static void nfs4_locku_done(struct rpc_task *task, void *data)
4518 {
4519         struct nfs4_unlockdata *calldata = data;
4520
4521         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4522                 return;
4523         switch (task->tk_status) {
4524                 case 0:
4525                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4526                                         &calldata->res.stateid);
4527                         renew_lease(calldata->server, calldata->timestamp);
4528                         break;
4529                 case -NFS4ERR_BAD_STATEID:
4530                 case -NFS4ERR_OLD_STATEID:
4531                 case -NFS4ERR_STALE_STATEID:
4532                 case -NFS4ERR_EXPIRED:
4533                         break;
4534                 default:
4535                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4536                                 rpc_restart_call_prepare(task);
4537         }
4538         nfs_release_seqid(calldata->arg.seqid);
4539 }
4540
4541 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4542 {
4543         struct nfs4_unlockdata *calldata = data;
4544
4545         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4546                 return;
4547         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4548                 /* Note: exit _without_ running nfs4_locku_done */
4549                 task->tk_action = NULL;
4550                 return;
4551         }
4552         calldata->timestamp = jiffies;
4553         if (nfs4_setup_sequence(calldata->server,
4554                                 &calldata->arg.seq_args,
4555                                 &calldata->res.seq_res,
4556                                 task) != 0)
4557                 nfs_release_seqid(calldata->arg.seqid);
4558         else
4559                 rpc_call_start(task);
4560 }
4561
4562 static const struct rpc_call_ops nfs4_locku_ops = {
4563         .rpc_call_prepare = nfs4_locku_prepare,
4564         .rpc_call_done = nfs4_locku_done,
4565         .rpc_release = nfs4_locku_release_calldata,
4566 };
4567
4568 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4569                 struct nfs_open_context *ctx,
4570                 struct nfs4_lock_state *lsp,
4571                 struct nfs_seqid *seqid)
4572 {
4573         struct nfs4_unlockdata *data;
4574         struct rpc_message msg = {
4575                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4576                 .rpc_cred = ctx->cred,
4577         };
4578         struct rpc_task_setup task_setup_data = {
4579                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4580                 .rpc_message = &msg,
4581                 .callback_ops = &nfs4_locku_ops,
4582                 .workqueue = nfsiod_workqueue,
4583                 .flags = RPC_TASK_ASYNC,
4584         };
4585
4586         /* Ensure this is an unlock - when canceling a lock, the
4587          * canceled lock is passed in, and it won't be an unlock.
4588          */
4589         fl->fl_type = F_UNLCK;
4590
4591         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4592         if (data == NULL) {
4593                 nfs_free_seqid(seqid);
4594                 return ERR_PTR(-ENOMEM);
4595         }
4596
4597         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4598         msg.rpc_argp = &data->arg;
4599         msg.rpc_resp = &data->res;
4600         task_setup_data.callback_data = data;
4601         return rpc_run_task(&task_setup_data);
4602 }
4603
4604 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4605 {
4606         struct nfs_inode *nfsi = NFS_I(state->inode);
4607         struct nfs_seqid *seqid;
4608         struct nfs4_lock_state *lsp;
4609         struct rpc_task *task;
4610         int status = 0;
4611         unsigned char fl_flags = request->fl_flags;
4612
4613         status = nfs4_set_lock_state(state, request);
4614         /* Unlock _before_ we do the RPC call */
4615         request->fl_flags |= FL_EXISTS;
4616         down_read(&nfsi->rwsem);
4617         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4618                 up_read(&nfsi->rwsem);
4619                 goto out;
4620         }
4621         up_read(&nfsi->rwsem);
4622         if (status != 0)
4623                 goto out;
4624         /* Is this a delegated lock? */
4625         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4626                 goto out;
4627         lsp = request->fl_u.nfs4_fl.owner;
4628         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4629         status = -ENOMEM;
4630         if (seqid == NULL)
4631                 goto out;
4632         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4633         status = PTR_ERR(task);
4634         if (IS_ERR(task))
4635                 goto out;
4636         status = nfs4_wait_for_completion_rpc_task(task);
4637         rpc_put_task(task);
4638 out:
4639         request->fl_flags = fl_flags;
4640         return status;
4641 }
4642
4643 struct nfs4_lockdata {
4644         struct nfs_lock_args arg;
4645         struct nfs_lock_res res;
4646         struct nfs4_lock_state *lsp;
4647         struct nfs_open_context *ctx;
4648         struct file_lock fl;
4649         unsigned long timestamp;
4650         int rpc_status;
4651         int cancelled;
4652         struct nfs_server *server;
4653 };
4654
4655 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4656                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4657                 gfp_t gfp_mask)
4658 {
4659         struct nfs4_lockdata *p;
4660         struct inode *inode = lsp->ls_state->inode;
4661         struct nfs_server *server = NFS_SERVER(inode);
4662
4663         p = kzalloc(sizeof(*p), gfp_mask);
4664         if (p == NULL)
4665                 return NULL;
4666
4667         p->arg.fh = NFS_FH(inode);
4668         p->arg.fl = &p->fl;
4669         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4670         if (p->arg.open_seqid == NULL)
4671                 goto out_free;
4672         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4673         if (p->arg.lock_seqid == NULL)
4674                 goto out_free_seqid;
4675         p->arg.lock_stateid = &lsp->ls_stateid;
4676         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4677         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4678         p->arg.lock_owner.s_dev = server->s_dev;
4679         p->res.lock_seqid = p->arg.lock_seqid;
4680         p->lsp = lsp;
4681         p->server = server;
4682         atomic_inc(&lsp->ls_count);
4683         p->ctx = get_nfs_open_context(ctx);
4684         memcpy(&p->fl, fl, sizeof(p->fl));
4685         return p;
4686 out_free_seqid:
4687         nfs_free_seqid(p->arg.open_seqid);
4688 out_free:
4689         kfree(p);
4690         return NULL;
4691 }
4692
4693 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4694 {
4695         struct nfs4_lockdata *data = calldata;
4696         struct nfs4_state *state = data->lsp->ls_state;
4697
4698         dprintk("%s: begin!\n", __func__);
4699         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4700                 return;
4701         /* Do we need to do an open_to_lock_owner? */
4702         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4703                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4704                         goto out_release_lock_seqid;
4705                 data->arg.open_stateid = &state->stateid;
4706                 data->arg.new_lock_owner = 1;
4707                 data->res.open_seqid = data->arg.open_seqid;
4708         } else
4709                 data->arg.new_lock_owner = 0;
4710         data->timestamp = jiffies;
4711         if (nfs4_setup_sequence(data->server,
4712                                 &data->arg.seq_args,
4713                                 &data->res.seq_res,
4714                                 task) == 0) {
4715                 rpc_call_start(task);
4716                 return;
4717         }
4718         nfs_release_seqid(data->arg.open_seqid);
4719 out_release_lock_seqid:
4720         nfs_release_seqid(data->arg.lock_seqid);
4721         dprintk("%s: done!, ret = %d\n", __func__, task->tk_status);
4722 }
4723
4724 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4725 {
4726         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4727         nfs4_lock_prepare(task, calldata);
4728 }
4729
4730 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4731 {
4732         struct nfs4_lockdata *data = calldata;
4733
4734         dprintk("%s: begin!\n", __func__);
4735
4736         if (!nfs4_sequence_done(task, &data->res.seq_res))
4737                 return;
4738
4739         data->rpc_status = task->tk_status;
4740         if (data->arg.new_lock_owner != 0) {
4741                 if (data->rpc_status == 0)
4742                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4743                 else
4744                         goto out;
4745         }
4746         if (data->rpc_status == 0) {
4747                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4748                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4749                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4750         }
4751 out:
4752         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4753 }
4754
4755 static void nfs4_lock_release(void *calldata)
4756 {
4757         struct nfs4_lockdata *data = calldata;
4758
4759         dprintk("%s: begin!\n", __func__);
4760         nfs_free_seqid(data->arg.open_seqid);
4761         if (data->cancelled != 0) {
4762                 struct rpc_task *task;
4763                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4764                                 data->arg.lock_seqid);
4765                 if (!IS_ERR(task))
4766                         rpc_put_task_async(task);
4767                 dprintk("%s: cancelling lock!\n", __func__);
4768         } else
4769                 nfs_free_seqid(data->arg.lock_seqid);
4770         nfs4_put_lock_state(data->lsp);
4771         put_nfs_open_context(data->ctx);
4772         kfree(data);
4773         dprintk("%s: done!\n", __func__);
4774 }
4775
4776 static const struct rpc_call_ops nfs4_lock_ops = {
4777         .rpc_call_prepare = nfs4_lock_prepare,
4778         .rpc_call_done = nfs4_lock_done,
4779         .rpc_release = nfs4_lock_release,
4780 };
4781
4782 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4783         .rpc_call_prepare = nfs4_recover_lock_prepare,
4784         .rpc_call_done = nfs4_lock_done,
4785         .rpc_release = nfs4_lock_release,
4786 };
4787
4788 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4789 {
4790         switch (error) {
4791         case -NFS4ERR_ADMIN_REVOKED:
4792         case -NFS4ERR_BAD_STATEID:
4793                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4794                 if (new_lock_owner != 0 ||
4795                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4796                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4797                 break;
4798         case -NFS4ERR_STALE_STATEID:
4799                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4800         case -NFS4ERR_EXPIRED:
4801                 nfs4_schedule_lease_recovery(server->nfs_client);
4802         };
4803 }
4804
4805 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4806 {
4807         struct nfs4_lockdata *data;
4808         struct rpc_task *task;
4809         struct rpc_message msg = {
4810                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4811                 .rpc_cred = state->owner->so_cred,
4812         };
4813         struct rpc_task_setup task_setup_data = {
4814                 .rpc_client = NFS_CLIENT(state->inode),
4815                 .rpc_message = &msg,
4816                 .callback_ops = &nfs4_lock_ops,
4817                 .workqueue = nfsiod_workqueue,
4818                 .flags = RPC_TASK_ASYNC,
4819         };
4820         int ret;
4821
4822         dprintk("%s: begin!\n", __func__);
4823         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4824                         fl->fl_u.nfs4_fl.owner,
4825                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4826         if (data == NULL)
4827                 return -ENOMEM;
4828         if (IS_SETLKW(cmd))
4829                 data->arg.block = 1;
4830         if (recovery_type > NFS_LOCK_NEW) {
4831                 if (recovery_type == NFS_LOCK_RECLAIM)
4832                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4833                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4834         }
4835         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4836         msg.rpc_argp = &data->arg;
4837         msg.rpc_resp = &data->res;
4838         task_setup_data.callback_data = data;
4839         task = rpc_run_task(&task_setup_data);
4840         if (IS_ERR(task))
4841                 return PTR_ERR(task);
4842         ret = nfs4_wait_for_completion_rpc_task(task);
4843         if (ret == 0) {
4844                 ret = data->rpc_status;
4845                 if (ret)
4846                         nfs4_handle_setlk_error(data->server, data->lsp,
4847                                         data->arg.new_lock_owner, ret);
4848         } else
4849                 data->cancelled = 1;
4850         rpc_put_task(task);
4851         dprintk("%s: done, ret = %d!\n", __func__, ret);
4852         return ret;
4853 }
4854
4855 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4856 {
4857         struct nfs_server *server = NFS_SERVER(state->inode);
4858         struct nfs4_exception exception = {
4859                 .inode = state->inode,
4860         };
4861         int err;
4862
4863         do {
4864                 /* Cache the lock if possible... */
4865                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4866                         return 0;
4867                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4868                 if (err != -NFS4ERR_DELAY)
4869                         break;
4870                 nfs4_handle_exception(server, err, &exception);
4871         } while (exception.retry);
4872         return err;
4873 }
4874
4875 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4876 {
4877         struct nfs_server *server = NFS_SERVER(state->inode);
4878         struct nfs4_exception exception = {
4879                 .inode = state->inode,
4880         };
4881         int err;
4882
4883         err = nfs4_set_lock_state(state, request);
4884         if (err != 0)
4885                 return err;
4886         do {
4887                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4888                         return 0;
4889                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4890                 switch (err) {
4891                 default:
4892                         goto out;
4893                 case -NFS4ERR_GRACE:
4894                 case -NFS4ERR_DELAY:
4895                         nfs4_handle_exception(server, err, &exception);
4896                         err = 0;
4897                 }
4898         } while (exception.retry);
4899 out:
4900         return err;
4901 }
4902
4903 #if defined(CONFIG_NFS_V4_1)
4904 /**
4905  * nfs41_check_expired_locks - possibly free a lock stateid
4906  *
4907  * @state: NFSv4 state for an inode
4908  *
4909  * Returns NFS_OK if recovery for this stateid is now finished.
4910  * Otherwise a negative NFS4ERR value is returned.
4911  */
4912 static int nfs41_check_expired_locks(struct nfs4_state *state)
4913 {
4914         int status, ret = -NFS4ERR_BAD_STATEID;
4915         struct nfs4_lock_state *lsp;
4916         struct nfs_server *server = NFS_SERVER(state->inode);
4917
4918         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4919                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
4920                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4921                         if (status != NFS_OK) {
4922                                 /* Free the stateid unless the server
4923                                  * informs us the stateid is unrecognized. */
4924                                 if (status != -NFS4ERR_BAD_STATEID)
4925                                         nfs41_free_stateid(server,
4926                                                         &lsp->ls_stateid);
4927                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
4928                                 ret = status;
4929                         }
4930                 }
4931         };
4932
4933         return ret;
4934 }
4935
4936 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4937 {
4938         int status = NFS_OK;
4939
4940         if (test_bit(LK_STATE_IN_USE, &state->flags))
4941                 status = nfs41_check_expired_locks(state);
4942         if (status != NFS_OK)
4943                 status = nfs4_lock_expired(state, request);
4944         return status;
4945 }
4946 #endif
4947
4948 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4949 {
4950         struct nfs_inode *nfsi = NFS_I(state->inode);
4951         unsigned char fl_flags = request->fl_flags;
4952         int status = -ENOLCK;
4953
4954         if ((fl_flags & FL_POSIX) &&
4955                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4956                 goto out;
4957         /* Is this a delegated open? */
4958         status = nfs4_set_lock_state(state, request);
4959         if (status != 0)
4960                 goto out;
4961         request->fl_flags |= FL_ACCESS;
4962         status = do_vfs_lock(request->fl_file, request);
4963         if (status < 0)
4964                 goto out;
4965         down_read(&nfsi->rwsem);
4966         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4967                 /* Yes: cache locks! */
4968                 /* ...but avoid races with delegation recall... */
4969                 request->fl_flags = fl_flags & ~FL_SLEEP;
4970                 status = do_vfs_lock(request->fl_file, request);
4971                 goto out_unlock;
4972         }
4973         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4974         if (status != 0)
4975                 goto out_unlock;
4976         /* Note: we always want to sleep here! */
4977         request->fl_flags = fl_flags | FL_SLEEP;
4978         if (do_vfs_lock(request->fl_file, request) < 0)
4979                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4980                         "manager!\n", __func__);
4981 out_unlock:
4982         up_read(&nfsi->rwsem);
4983 out:
4984         request->fl_flags = fl_flags;
4985         return status;
4986 }
4987
4988 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4989 {
4990         struct nfs4_exception exception = {
4991                 .state = state,
4992                 .inode = state->inode,
4993         };
4994         int err;
4995
4996         do {
4997                 err = _nfs4_proc_setlk(state, cmd, request);
4998                 if (err == -NFS4ERR_DENIED)
4999                         err = -EAGAIN;
5000                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5001                                 err, &exception);
5002         } while (exception.retry);
5003         return err;
5004 }
5005
5006 static int
5007 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5008 {
5009         struct nfs_open_context *ctx;
5010         struct nfs4_state *state;
5011         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5012         int status;
5013
5014         /* verify open state */
5015         ctx = nfs_file_open_context(filp);
5016         state = ctx->state;
5017
5018         if (request->fl_start < 0 || request->fl_end < 0)
5019                 return -EINVAL;
5020
5021         if (IS_GETLK(cmd)) {
5022                 if (state != NULL)
5023                         return nfs4_proc_getlk(state, F_GETLK, request);
5024                 return 0;
5025         }
5026
5027         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5028                 return -EINVAL;
5029
5030         if (request->fl_type == F_UNLCK) {
5031                 if (state != NULL)
5032                         return nfs4_proc_unlck(state, cmd, request);
5033                 return 0;
5034         }
5035
5036         if (state == NULL)
5037                 return -ENOLCK;
5038         /*
5039          * Don't rely on the VFS having checked the file open mode,
5040          * since it won't do this for flock() locks.
5041          */
5042         switch (request->fl_type) {
5043         case F_RDLCK:
5044                 if (!(filp->f_mode & FMODE_READ))
5045                         return -EBADF;
5046                 break;
5047         case F_WRLCK:
5048                 if (!(filp->f_mode & FMODE_WRITE))
5049                         return -EBADF;
5050         }
5051
5052         do {
5053                 status = nfs4_proc_setlk(state, cmd, request);
5054                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5055                         break;
5056                 timeout = nfs4_set_lock_task_retry(timeout);
5057                 status = -ERESTARTSYS;
5058                 if (signalled())
5059                         break;
5060         } while(status < 0);
5061         return status;
5062 }
5063
5064 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
5065 {
5066         struct nfs_server *server = NFS_SERVER(state->inode);
5067         struct nfs4_exception exception = { };
5068         int err;
5069
5070         err = nfs4_set_lock_state(state, fl);
5071         if (err != 0)
5072                 goto out;
5073         do {
5074                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5075                 switch (err) {
5076                         default:
5077                                 printk(KERN_ERR "NFS: %s: unhandled error "
5078                                         "%d.\n", __func__, err);
5079                         case 0:
5080                         case -ESTALE:
5081                                 goto out;
5082                         case -NFS4ERR_EXPIRED:
5083                                 nfs4_schedule_stateid_recovery(server, state);
5084                         case -NFS4ERR_STALE_CLIENTID:
5085                         case -NFS4ERR_STALE_STATEID:
5086                                 nfs4_schedule_lease_recovery(server->nfs_client);
5087                                 goto out;
5088                         case -NFS4ERR_BADSESSION:
5089                         case -NFS4ERR_BADSLOT:
5090                         case -NFS4ERR_BAD_HIGH_SLOT:
5091                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
5092                         case -NFS4ERR_DEADSESSION:
5093                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
5094                                 goto out;
5095                         case -ERESTARTSYS:
5096                                 /*
5097                                  * The show must go on: exit, but mark the
5098                                  * stateid as needing recovery.
5099                                  */
5100                         case -NFS4ERR_DELEG_REVOKED:
5101                         case -NFS4ERR_ADMIN_REVOKED:
5102                         case -NFS4ERR_BAD_STATEID:
5103                         case -NFS4ERR_OPENMODE:
5104                                 nfs4_schedule_stateid_recovery(server, state);
5105                                 err = 0;
5106                                 goto out;
5107                         case -EKEYEXPIRED:
5108                                 /*
5109                                  * User RPCSEC_GSS context has expired.
5110                                  * We cannot recover this stateid now, so
5111                                  * skip it and allow recovery thread to
5112                                  * proceed.
5113                                  */
5114                                 err = 0;
5115                                 goto out;
5116                         case -ENOMEM:
5117                         case -NFS4ERR_DENIED:
5118                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
5119                                 err = 0;
5120                                 goto out;
5121                         case -NFS4ERR_DELAY:
5122                                 break;
5123                 }
5124                 err = nfs4_handle_exception(server, err, &exception);
5125         } while (exception.retry);
5126 out:
5127         return err;
5128 }
5129
5130 struct nfs_release_lockowner_data {
5131         struct nfs4_lock_state *lsp;
5132         struct nfs_server *server;
5133         struct nfs_release_lockowner_args args;
5134 };
5135
5136 static void nfs4_release_lockowner_release(void *calldata)
5137 {
5138         struct nfs_release_lockowner_data *data = calldata;
5139         nfs4_free_lock_state(data->server, data->lsp);
5140         kfree(calldata);
5141 }
5142
5143 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5144         .rpc_release = nfs4_release_lockowner_release,
5145 };
5146
5147 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
5148 {
5149         struct nfs_server *server = lsp->ls_state->owner->so_server;
5150         struct nfs_release_lockowner_data *data;
5151         struct rpc_message msg = {
5152                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5153         };
5154
5155         if (server->nfs_client->cl_mvops->minor_version != 0)
5156                 return -EINVAL;
5157         data = kmalloc(sizeof(*data), GFP_NOFS);
5158         if (!data)
5159                 return -ENOMEM;
5160         data->lsp = lsp;
5161         data->server = server;
5162         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5163         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5164         data->args.lock_owner.s_dev = server->s_dev;
5165         msg.rpc_argp = &data->args;
5166         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5167         return 0;
5168 }
5169
5170 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5171
5172 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5173                                    const void *buf, size_t buflen,
5174                                    int flags, int type)
5175 {
5176         if (strcmp(key, "") != 0)
5177                 return -EINVAL;
5178
5179         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5180 }
5181
5182 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5183                                    void *buf, size_t buflen, int type)
5184 {
5185         if (strcmp(key, "") != 0)
5186                 return -EINVAL;
5187
5188         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5189 }
5190
5191 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5192                                        size_t list_len, const char *name,
5193                                        size_t name_len, int type)
5194 {
5195         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5196
5197         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5198                 return 0;
5199
5200         if (list && len <= list_len)
5201                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5202         return len;
5203 }
5204
5205 /*
5206  * nfs_fhget will use either the mounted_on_fileid or the fileid
5207  */
5208 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5209 {
5210         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5211                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5212               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5213               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5214                 return;
5215
5216         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5217                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5218         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5219         fattr->nlink = 2;
5220 }
5221
5222 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5223                                    const struct qstr *name,
5224                                    struct nfs4_fs_locations *fs_locations,
5225                                    struct page *page)
5226 {
5227         struct nfs_server *server = NFS_SERVER(dir);
5228         u32 bitmask[2] = {
5229                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5230         };
5231         struct nfs4_fs_locations_arg args = {
5232                 .dir_fh = NFS_FH(dir),
5233                 .name = name,
5234                 .page = page,
5235                 .bitmask = bitmask,
5236         };
5237         struct nfs4_fs_locations_res res = {
5238                 .fs_locations = fs_locations,
5239         };
5240         struct rpc_message msg = {
5241                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5242                 .rpc_argp = &args,
5243                 .rpc_resp = &res,
5244         };
5245         int status;
5246
5247         dprintk("%s: start\n", __func__);
5248
5249         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5250          * is not supported */
5251         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5252                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5253         else
5254                 bitmask[0] |= FATTR4_WORD0_FILEID;
5255
5256         nfs_fattr_init(&fs_locations->fattr);
5257         fs_locations->server = server;
5258         fs_locations->nlocations = 0;
5259         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5260         dprintk("%s: returned status = %d\n", __func__, status);
5261         return status;
5262 }
5263
5264 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5265                            const struct qstr *name,
5266                            struct nfs4_fs_locations *fs_locations,
5267                            struct page *page)
5268 {
5269         struct nfs4_exception exception = { };
5270         int err;
5271         do {
5272                 err = nfs4_handle_exception(NFS_SERVER(dir),
5273                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5274                                 &exception);
5275         } while (exception.retry);
5276         return err;
5277 }
5278
5279 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5280 {
5281         int status;
5282         struct nfs4_secinfo_arg args = {
5283                 .dir_fh = NFS_FH(dir),
5284                 .name   = name,
5285         };
5286         struct nfs4_secinfo_res res = {
5287                 .flavors     = flavors,
5288         };
5289         struct rpc_message msg = {
5290                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5291                 .rpc_argp = &args,
5292                 .rpc_resp = &res,
5293         };
5294
5295         dprintk("NFS call  secinfo %s\n", name->name);
5296         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5297         dprintk("NFS reply  secinfo: %d\n", status);
5298         return status;
5299 }
5300
5301 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5302                       struct nfs4_secinfo_flavors *flavors)
5303 {
5304         struct nfs4_exception exception = { };
5305         int err;
5306         do {
5307                 err = nfs4_handle_exception(NFS_SERVER(dir),
5308                                 _nfs4_proc_secinfo(dir, name, flavors),
5309                                 &exception);
5310         } while (exception.retry);
5311         return err;
5312 }
5313
5314 #ifdef CONFIG_NFS_V4_1
5315 /*
5316  * Check the exchange flags returned by the server for invalid flags, having
5317  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5318  * DS flags set.
5319  */
5320 static int nfs4_check_cl_exchange_flags(u32 flags)
5321 {
5322         if (flags & ~EXCHGID4_FLAG_MASK_R)
5323                 goto out_inval;
5324         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5325             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5326                 goto out_inval;
5327         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5328                 goto out_inval;
5329         return NFS_OK;
5330 out_inval:
5331         return -NFS4ERR_INVAL;
5332 }
5333
5334 static bool
5335 nfs41_same_server_scope(struct nfs41_server_scope *a,
5336                         struct nfs41_server_scope *b)
5337 {
5338         if (a->server_scope_sz == b->server_scope_sz &&
5339             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5340                 return true;
5341
5342         return false;
5343 }
5344
5345 /*
5346  * nfs4_proc_bind_conn_to_session()
5347  *
5348  * The 4.1 client currently uses the same TCP connection for the
5349  * fore and backchannel.
5350  */
5351 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5352 {
5353         int status;
5354         struct nfs41_bind_conn_to_session_res res;
5355         struct rpc_message msg = {
5356                 .rpc_proc =
5357                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5358                 .rpc_argp = clp,
5359                 .rpc_resp = &res,
5360                 .rpc_cred = cred,
5361         };
5362
5363         dprintk("--> %s\n", __func__);
5364
5365         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5366         if (unlikely(res.session == NULL)) {
5367                 status = -ENOMEM;
5368                 goto out;
5369         }
5370
5371         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5372         if (status == 0) {
5373                 if (memcmp(res.session->sess_id.data,
5374                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5375                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5376                         status = -EIO;
5377                         goto out_session;
5378                 }
5379                 if (res.dir != NFS4_CDFS4_BOTH) {
5380                         dprintk("NFS: %s: Unexpected direction from server\n",
5381                                 __func__);
5382                         status = -EIO;
5383                         goto out_session;
5384                 }
5385                 if (res.use_conn_in_rdma_mode) {
5386                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5387                                 __func__);
5388                         status = -EIO;
5389                         goto out_session;
5390                 }
5391         }
5392 out_session:
5393         kfree(res.session);
5394 out:
5395         dprintk("<-- %s status= %d\n", __func__, status);
5396         return status;
5397 }
5398
5399 /*
5400  * nfs4_proc_exchange_id()
5401  *
5402  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5403  *
5404  * Since the clientid has expired, all compounds using sessions
5405  * associated with the stale clientid will be returning
5406  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5407  * be in some phase of session reset.
5408  */
5409 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5410 {
5411         nfs4_verifier verifier;
5412         struct nfs41_exchange_id_args args = {
5413                 .verifier = &verifier,
5414                 .client = clp,
5415                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5416         };
5417         struct nfs41_exchange_id_res res = {
5418                 0
5419         };
5420         int status;
5421         struct rpc_message msg = {
5422                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5423                 .rpc_argp = &args,
5424                 .rpc_resp = &res,
5425                 .rpc_cred = cred,
5426         };
5427
5428         nfs4_init_boot_verifier(clp, &verifier);
5429         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5430                                                         sizeof(args.id));
5431         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5432                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5433                 args.id_len, args.id);
5434
5435         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5436                                         GFP_NOFS);
5437         if (unlikely(res.server_owner == NULL)) {
5438                 status = -ENOMEM;
5439                 goto out;
5440         }
5441
5442         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5443                                         GFP_NOFS);
5444         if (unlikely(res.server_scope == NULL)) {
5445                 status = -ENOMEM;
5446                 goto out_server_owner;
5447         }
5448
5449         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5450         if (unlikely(res.impl_id == NULL)) {
5451                 status = -ENOMEM;
5452                 goto out_server_scope;
5453         }
5454
5455         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5456         if (status == 0)
5457                 status = nfs4_check_cl_exchange_flags(res.flags);
5458
5459         if (status == 0) {
5460                 clp->cl_clientid = res.clientid;
5461                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5462                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5463                         clp->cl_seqid = res.seqid;
5464
5465                 kfree(clp->cl_serverowner);
5466                 clp->cl_serverowner = res.server_owner;
5467                 res.server_owner = NULL;
5468
5469                 /* use the most recent implementation id */
5470                 kfree(clp->cl_implid);
5471                 clp->cl_implid = res.impl_id;
5472
5473                 if (clp->cl_serverscope != NULL &&
5474                     !nfs41_same_server_scope(clp->cl_serverscope,
5475                                              res.server_scope)) {
5476                         dprintk("%s: server_scope mismatch detected\n",
5477                                 __func__);
5478                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5479                         kfree(clp->cl_serverscope);
5480                         clp->cl_serverscope = NULL;
5481                 }
5482
5483                 if (clp->cl_serverscope == NULL) {
5484                         clp->cl_serverscope = res.server_scope;
5485                         goto out;
5486                 }
5487         } else
5488                 kfree(res.impl_id);
5489
5490 out_server_owner:
5491         kfree(res.server_owner);
5492 out_server_scope:
5493         kfree(res.server_scope);
5494 out:
5495         if (clp->cl_implid != NULL)
5496                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5497                         "domain: %s, name: %s, date: %llu,%u\n",
5498                         clp->cl_implid->domain, clp->cl_implid->name,
5499                         clp->cl_implid->date.seconds,
5500                         clp->cl_implid->date.nseconds);
5501         dprintk("NFS reply exchange_id: %d\n", status);
5502         return status;
5503 }
5504
5505 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5506                 struct rpc_cred *cred)
5507 {
5508         struct rpc_message msg = {
5509                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5510                 .rpc_argp = clp,
5511                 .rpc_cred = cred,
5512         };
5513         int status;
5514
5515         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5516         if (status)
5517                 dprintk("NFS: Got error %d from the server %s on "
5518                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5519         return status;
5520 }
5521
5522 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5523                 struct rpc_cred *cred)
5524 {
5525         unsigned int loop;
5526         int ret;
5527
5528         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5529                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5530                 switch (ret) {
5531                 case -NFS4ERR_DELAY:
5532                 case -NFS4ERR_CLIENTID_BUSY:
5533                         ssleep(1);
5534                         break;
5535                 default:
5536                         return ret;
5537                 }
5538         }
5539         return 0;
5540 }
5541
5542 int nfs4_destroy_clientid(struct nfs_client *clp)
5543 {
5544         struct rpc_cred *cred;
5545         int ret = 0;
5546
5547         if (clp->cl_mvops->minor_version < 1)
5548                 goto out;
5549         if (clp->cl_exchange_flags == 0)
5550                 goto out;
5551         if (clp->cl_preserve_clid)
5552                 goto out;
5553         cred = nfs4_get_exchange_id_cred(clp);
5554         ret = nfs4_proc_destroy_clientid(clp, cred);
5555         if (cred)
5556                 put_rpccred(cred);
5557         switch (ret) {
5558         case 0:
5559         case -NFS4ERR_STALE_CLIENTID:
5560                 clp->cl_exchange_flags = 0;
5561         }
5562 out:
5563         return ret;
5564 }
5565
5566 struct nfs4_get_lease_time_data {
5567         struct nfs4_get_lease_time_args *args;
5568         struct nfs4_get_lease_time_res *res;
5569         struct nfs_client *clp;
5570 };
5571
5572 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5573                                         void *calldata)
5574 {
5575         int ret;
5576         struct nfs4_get_lease_time_data *data =
5577                         (struct nfs4_get_lease_time_data *)calldata;
5578
5579         dprintk("--> %s\n", __func__);
5580         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5581         /* just setup sequence, do not trigger session recovery
5582            since we're invoked within one */
5583         ret = nfs41_setup_sequence(data->clp->cl_session,
5584                                    &data->args->la_seq_args,
5585                                    &data->res->lr_seq_res, task);
5586
5587         if (ret != -EAGAIN)
5588                 rpc_call_start(task);
5589         dprintk("<-- %s\n", __func__);
5590 }
5591
5592 /*
5593  * Called from nfs4_state_manager thread for session setup, so don't recover
5594  * from sequence operation or clientid errors.
5595  */
5596 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5597 {
5598         struct nfs4_get_lease_time_data *data =
5599                         (struct nfs4_get_lease_time_data *)calldata;
5600
5601         dprintk("--> %s\n", __func__);
5602         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5603                 return;
5604         switch (task->tk_status) {
5605         case -NFS4ERR_DELAY:
5606         case -NFS4ERR_GRACE:
5607                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5608                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5609                 task->tk_status = 0;
5610                 /* fall through */
5611         case -NFS4ERR_RETRY_UNCACHED_REP:
5612                 rpc_restart_call_prepare(task);
5613                 return;
5614         }
5615         dprintk("<-- %s\n", __func__);
5616 }
5617
5618 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5619         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5620         .rpc_call_done = nfs4_get_lease_time_done,
5621 };
5622
5623 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5624 {
5625         struct rpc_task *task;
5626         struct nfs4_get_lease_time_args args;
5627         struct nfs4_get_lease_time_res res = {
5628                 .lr_fsinfo = fsinfo,
5629         };
5630         struct nfs4_get_lease_time_data data = {
5631                 .args = &args,
5632                 .res = &res,
5633                 .clp = clp,
5634         };
5635         struct rpc_message msg = {
5636                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5637                 .rpc_argp = &args,
5638                 .rpc_resp = &res,
5639         };
5640         struct rpc_task_setup task_setup = {
5641                 .rpc_client = clp->cl_rpcclient,
5642                 .rpc_message = &msg,
5643                 .callback_ops = &nfs4_get_lease_time_ops,
5644                 .callback_data = &data,
5645                 .flags = RPC_TASK_TIMEOUT,
5646         };
5647         int status;
5648
5649         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5650         dprintk("--> %s\n", __func__);
5651         task = rpc_run_task(&task_setup);
5652
5653         if (IS_ERR(task))
5654                 status = PTR_ERR(task);
5655         else {
5656                 status = task->tk_status;
5657                 rpc_put_task(task);
5658         }
5659         dprintk("<-- %s return %d\n", __func__, status);
5660
5661         return status;
5662 }
5663
5664 struct nfs4_slot *nfs4_alloc_slots(struct nfs4_slot_table *table,
5665                 u32 max_slots, gfp_t gfp_flags)
5666 {
5667         struct nfs4_slot *tbl;
5668         u32 i;
5669
5670         tbl = kmalloc_array(max_slots, sizeof(*tbl), gfp_flags);
5671         if (tbl != NULL) {
5672                 for (i = 0; i < max_slots; i++)
5673                         tbl[i].table = table;
5674         }
5675         return tbl;
5676 }
5677
5678 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5679                 struct nfs4_slot *new,
5680                 u32 max_slots,
5681                 u32 ivalue)
5682 {
5683         struct nfs4_slot *old = NULL;
5684         u32 i;
5685
5686         spin_lock(&tbl->slot_tbl_lock);
5687         if (new) {
5688                 old = tbl->slots;
5689                 tbl->slots = new;
5690                 tbl->max_slots = max_slots;
5691         }
5692         tbl->highest_used_slotid = NFS4_NO_SLOT;
5693         for (i = 0; i < tbl->max_slots; i++)
5694                 tbl->slots[i].seq_nr = ivalue;
5695         spin_unlock(&tbl->slot_tbl_lock);
5696         kfree(old);
5697 }
5698
5699 /*
5700  * (re)Initialise a slot table
5701  */
5702 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5703                                  u32 ivalue)
5704 {
5705         struct nfs4_slot *new = NULL;
5706         int ret = -ENOMEM;
5707
5708         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5709                 max_reqs, tbl->max_slots);
5710
5711         /* Does the newly negotiated max_reqs match the existing slot table? */
5712         if (max_reqs != tbl->max_slots) {
5713                 new = nfs4_alloc_slots(tbl, max_reqs, GFP_NOFS);
5714                 if (!new)
5715                         goto out;
5716         }
5717         ret = 0;
5718
5719         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5720         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5721                 tbl, tbl->slots, tbl->max_slots);
5722 out:
5723         dprintk("<-- %s: return %d\n", __func__, ret);
5724         return ret;
5725 }
5726
5727 /* Destroy the slot table */
5728 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5729 {
5730         if (session->fc_slot_table.slots != NULL) {
5731                 kfree(session->fc_slot_table.slots);
5732                 session->fc_slot_table.slots = NULL;
5733         }
5734         if (session->bc_slot_table.slots != NULL) {
5735                 kfree(session->bc_slot_table.slots);
5736                 session->bc_slot_table.slots = NULL;
5737         }
5738         return;
5739 }
5740
5741 /*
5742  * Initialize or reset the forechannel and backchannel tables
5743  */
5744 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5745 {
5746         struct nfs4_slot_table *tbl;
5747         int status;
5748
5749         dprintk("--> %s\n", __func__);
5750         /* Fore channel */
5751         tbl = &ses->fc_slot_table;
5752         tbl->session = ses;
5753         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5754         if (status) /* -ENOMEM */
5755                 return status;
5756         /* Back channel */
5757         tbl = &ses->bc_slot_table;
5758         tbl->session = ses;
5759         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5760         if (status && tbl->slots == NULL)
5761                 /* Fore and back channel share a connection so get
5762                  * both slot tables or neither */
5763                 nfs4_destroy_slot_tables(ses);
5764         return status;
5765 }
5766
5767 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5768 {
5769         struct nfs4_session *session;
5770         struct nfs4_slot_table *tbl;
5771
5772         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5773         if (!session)
5774                 return NULL;
5775
5776         tbl = &session->fc_slot_table;
5777         tbl->highest_used_slotid = NFS4_NO_SLOT;
5778         spin_lock_init(&tbl->slot_tbl_lock);
5779         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5780         init_completion(&tbl->complete);
5781
5782         tbl = &session->bc_slot_table;
5783         tbl->highest_used_slotid = NFS4_NO_SLOT;
5784         spin_lock_init(&tbl->slot_tbl_lock);
5785         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5786         init_completion(&tbl->complete);
5787
5788         session->session_state = 1<<NFS4_SESSION_INITING;
5789
5790         session->clp = clp;
5791         return session;
5792 }
5793
5794 void nfs4_destroy_session(struct nfs4_session *session)
5795 {
5796         struct rpc_xprt *xprt;
5797         struct rpc_cred *cred;
5798
5799         cred = nfs4_get_exchange_id_cred(session->clp);
5800         nfs4_proc_destroy_session(session, cred);
5801         if (cred)
5802                 put_rpccred(cred);
5803
5804         rcu_read_lock();
5805         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5806         rcu_read_unlock();
5807         dprintk("%s Destroy backchannel for xprt %p\n",
5808                 __func__, xprt);
5809         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5810         nfs4_destroy_slot_tables(session);
5811         kfree(session);
5812 }
5813
5814 /*
5815  * Initialize the values to be used by the client in CREATE_SESSION
5816  * If nfs4_init_session set the fore channel request and response sizes,
5817  * use them.
5818  *
5819  * Set the back channel max_resp_sz_cached to zero to force the client to
5820  * always set csa_cachethis to FALSE because the current implementation
5821  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5822  */
5823 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5824 {
5825         struct nfs4_session *session = args->client->cl_session;
5826         unsigned int mxrqst_sz = session->fc_target_max_rqst_sz,
5827                      mxresp_sz = session->fc_target_max_resp_sz;
5828
5829         if (mxrqst_sz == 0)
5830                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5831         if (mxresp_sz == 0)
5832                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5833         /* Fore channel attributes */
5834         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5835         args->fc_attrs.max_resp_sz = mxresp_sz;
5836         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5837         args->fc_attrs.max_reqs = max_session_slots;
5838
5839         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5840                 "max_ops=%u max_reqs=%u\n",
5841                 __func__,
5842                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5843                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5844
5845         /* Back channel attributes */
5846         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5847         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5848         args->bc_attrs.max_resp_sz_cached = 0;
5849         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5850         args->bc_attrs.max_reqs = 1;
5851
5852         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5853                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5854                 __func__,
5855                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5856                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5857                 args->bc_attrs.max_reqs);
5858 }
5859
5860 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5861 {
5862         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5863         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5864
5865         if (rcvd->max_resp_sz > sent->max_resp_sz)
5866                 return -EINVAL;
5867         /*
5868          * Our requested max_ops is the minimum we need; we're not
5869          * prepared to break up compounds into smaller pieces than that.
5870          * So, no point even trying to continue if the server won't
5871          * cooperate:
5872          */
5873         if (rcvd->max_ops < sent->max_ops)
5874                 return -EINVAL;
5875         if (rcvd->max_reqs == 0)
5876                 return -EINVAL;
5877         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5878                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5879         return 0;
5880 }
5881
5882 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5883 {
5884         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5885         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5886
5887         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5888                 return -EINVAL;
5889         if (rcvd->max_resp_sz < sent->max_resp_sz)
5890                 return -EINVAL;
5891         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5892                 return -EINVAL;
5893         /* These would render the backchannel useless: */
5894         if (rcvd->max_ops != sent->max_ops)
5895                 return -EINVAL;
5896         if (rcvd->max_reqs != sent->max_reqs)
5897                 return -EINVAL;
5898         return 0;
5899 }
5900
5901 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5902                                      struct nfs4_session *session)
5903 {
5904         int ret;
5905
5906         ret = nfs4_verify_fore_channel_attrs(args, session);
5907         if (ret)
5908                 return ret;
5909         return nfs4_verify_back_channel_attrs(args, session);
5910 }
5911
5912 static int _nfs4_proc_create_session(struct nfs_client *clp,
5913                 struct rpc_cred *cred)
5914 {
5915         struct nfs4_session *session = clp->cl_session;
5916         struct nfs41_create_session_args args = {
5917                 .client = clp,
5918                 .cb_program = NFS4_CALLBACK,
5919         };
5920         struct nfs41_create_session_res res = {
5921                 .client = clp,
5922         };
5923         struct rpc_message msg = {
5924                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5925                 .rpc_argp = &args,
5926                 .rpc_resp = &res,
5927                 .rpc_cred = cred,
5928         };
5929         int status;
5930
5931         nfs4_init_channel_attrs(&args);
5932         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5933
5934         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5935
5936         if (!status) {
5937                 /* Verify the session's negotiated channel_attrs values */
5938                 status = nfs4_verify_channel_attrs(&args, session);
5939                 /* Increment the clientid slot sequence id */
5940                 clp->cl_seqid++;
5941         }
5942
5943         return status;
5944 }
5945
5946 /*
5947  * Issues a CREATE_SESSION operation to the server.
5948  * It is the responsibility of the caller to verify the session is
5949  * expired before calling this routine.
5950  */
5951 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5952 {
5953         int status;
5954         unsigned *ptr;
5955         struct nfs4_session *session = clp->cl_session;
5956
5957         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5958
5959         status = _nfs4_proc_create_session(clp, cred);
5960         if (status)
5961                 goto out;
5962
5963         /* Init or reset the session slot tables */
5964         status = nfs4_setup_session_slot_tables(session);
5965         dprintk("slot table setup returned %d\n", status);
5966         if (status)
5967                 goto out;
5968
5969         ptr = (unsigned *)&session->sess_id.data[0];
5970         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5971                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5972 out:
5973         dprintk("<-- %s\n", __func__);
5974         return status;
5975 }
5976
5977 /*
5978  * Issue the over-the-wire RPC DESTROY_SESSION.
5979  * The caller must serialize access to this routine.
5980  */
5981 int nfs4_proc_destroy_session(struct nfs4_session *session,
5982                 struct rpc_cred *cred)
5983 {
5984         struct rpc_message msg = {
5985                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5986                 .rpc_argp = session,
5987                 .rpc_cred = cred,
5988         };
5989         int status = 0;
5990
5991         dprintk("--> nfs4_proc_destroy_session\n");
5992
5993         /* session is still being setup */
5994         if (session->clp->cl_cons_state != NFS_CS_READY)
5995                 return status;
5996
5997         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5998
5999         if (status)
6000                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
6001                         "Session has been destroyed regardless...\n", status);
6002
6003         dprintk("<-- nfs4_proc_destroy_session\n");
6004         return status;
6005 }
6006
6007 /*
6008  * With sessions, the client is not marked ready until after a
6009  * successful EXCHANGE_ID and CREATE_SESSION.
6010  *
6011  * Map errors cl_cons_state errors to EPROTONOSUPPORT to indicate
6012  * other versions of NFS can be tried.
6013  */
6014 static int nfs41_check_session_ready(struct nfs_client *clp)
6015 {
6016         int ret;
6017         
6018         if (clp->cl_cons_state == NFS_CS_SESSION_INITING) {
6019                 ret = nfs4_client_recover_expired_lease(clp);
6020                 if (ret)
6021                         return ret;
6022         }
6023         if (clp->cl_cons_state < NFS_CS_READY)
6024                 return -EPROTONOSUPPORT;
6025         smp_rmb();
6026         return 0;
6027 }
6028
6029 int nfs4_init_session(struct nfs_server *server)
6030 {
6031         struct nfs_client *clp = server->nfs_client;
6032         struct nfs4_session *session;
6033         unsigned int target_max_rqst_sz = NFS_MAX_FILE_IO_SIZE;
6034         unsigned int target_max_resp_sz = NFS_MAX_FILE_IO_SIZE;
6035
6036         if (!nfs4_has_session(clp))
6037                 return 0;
6038
6039         if (server->rsize != 0)
6040                 target_max_resp_sz = server->rsize;
6041         target_max_resp_sz += nfs41_maxread_overhead;
6042
6043         if (server->wsize != 0)
6044                 target_max_rqst_sz = server->wsize;
6045         target_max_rqst_sz += nfs41_maxwrite_overhead;
6046
6047         session = clp->cl_session;
6048         spin_lock(&clp->cl_lock);
6049         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
6050                 /* Initialise targets and channel attributes */
6051                 session->fc_target_max_rqst_sz = target_max_rqst_sz;
6052                 session->fc_attrs.max_rqst_sz = target_max_rqst_sz;
6053                 session->fc_target_max_resp_sz = target_max_resp_sz;
6054                 session->fc_attrs.max_resp_sz = target_max_resp_sz;
6055         } else {
6056                 /* Just adjust the targets */
6057                 if (target_max_rqst_sz > session->fc_target_max_rqst_sz) {
6058                         session->fc_target_max_rqst_sz = target_max_rqst_sz;
6059                         set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
6060                 }
6061                 if (target_max_resp_sz > session->fc_target_max_resp_sz) {
6062                         session->fc_target_max_resp_sz = target_max_resp_sz;
6063                         set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
6064                 }
6065         }
6066         spin_unlock(&clp->cl_lock);
6067
6068         if (test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
6069                 nfs4_schedule_lease_recovery(clp);
6070
6071         return nfs41_check_session_ready(clp);
6072 }
6073
6074 int nfs4_init_ds_session(struct nfs_client *clp, unsigned long lease_time)
6075 {
6076         struct nfs4_session *session = clp->cl_session;
6077         int ret;
6078
6079         spin_lock(&clp->cl_lock);
6080         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
6081                 /*
6082                  * Do not set NFS_CS_CHECK_LEASE_TIME instead set the
6083                  * DS lease to be equal to the MDS lease.
6084                  */
6085                 clp->cl_lease_time = lease_time;
6086                 clp->cl_last_renewal = jiffies;
6087         }
6088         spin_unlock(&clp->cl_lock);
6089
6090         ret = nfs41_check_session_ready(clp);
6091         if (ret)
6092                 return ret;
6093         /* Test for the DS role */
6094         if (!is_ds_client(clp))
6095                 return -ENODEV;
6096         return 0;
6097 }
6098 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
6099
6100
6101 /*
6102  * Renew the cl_session lease.
6103  */
6104 struct nfs4_sequence_data {
6105         struct nfs_client *clp;
6106         struct nfs4_sequence_args args;
6107         struct nfs4_sequence_res res;
6108 };
6109
6110 static void nfs41_sequence_release(void *data)
6111 {
6112         struct nfs4_sequence_data *calldata = data;
6113         struct nfs_client *clp = calldata->clp;
6114
6115         if (atomic_read(&clp->cl_count) > 1)
6116                 nfs4_schedule_state_renewal(clp);
6117         nfs_put_client(clp);
6118         kfree(calldata);
6119 }
6120
6121 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6122 {
6123         switch(task->tk_status) {
6124         case -NFS4ERR_DELAY:
6125                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6126                 return -EAGAIN;
6127         default:
6128                 nfs4_schedule_lease_recovery(clp);
6129         }
6130         return 0;
6131 }
6132
6133 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
6134 {
6135         struct nfs4_sequence_data *calldata = data;
6136         struct nfs_client *clp = calldata->clp;
6137
6138         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
6139                 return;
6140
6141         if (task->tk_status < 0) {
6142                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
6143                 if (atomic_read(&clp->cl_count) == 1)
6144                         goto out;
6145
6146                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
6147                         rpc_restart_call_prepare(task);
6148                         return;
6149                 }
6150         }
6151         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
6152 out:
6153         dprintk("<-- %s\n", __func__);
6154 }
6155
6156 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
6157 {
6158         struct nfs4_sequence_data *calldata = data;
6159         struct nfs_client *clp = calldata->clp;
6160         struct nfs4_sequence_args *args;
6161         struct nfs4_sequence_res *res;
6162
6163         args = task->tk_msg.rpc_argp;
6164         res = task->tk_msg.rpc_resp;
6165
6166         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
6167                 return;
6168         rpc_call_start(task);
6169 }
6170
6171 static void nfs41_sequence_prepare_privileged(struct rpc_task *task, void *data)
6172 {
6173         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
6174         nfs41_sequence_prepare(task, data);
6175 }
6176
6177 static const struct rpc_call_ops nfs41_sequence_ops = {
6178         .rpc_call_done = nfs41_sequence_call_done,
6179         .rpc_call_prepare = nfs41_sequence_prepare,
6180         .rpc_release = nfs41_sequence_release,
6181 };
6182
6183 static const struct rpc_call_ops nfs41_sequence_privileged_ops = {
6184         .rpc_call_done = nfs41_sequence_call_done,
6185         .rpc_call_prepare = nfs41_sequence_prepare_privileged,
6186         .rpc_release = nfs41_sequence_release,
6187 };
6188
6189 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred,
6190                                              const struct rpc_call_ops *seq_ops)
6191 {
6192         struct nfs4_sequence_data *calldata;
6193         struct rpc_message msg = {
6194                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6195                 .rpc_cred = cred,
6196         };
6197         struct rpc_task_setup task_setup_data = {
6198                 .rpc_client = clp->cl_rpcclient,
6199                 .rpc_message = &msg,
6200                 .callback_ops = seq_ops,
6201                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
6202         };
6203
6204         if (!atomic_inc_not_zero(&clp->cl_count))
6205                 return ERR_PTR(-EIO);
6206         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6207         if (calldata == NULL) {
6208                 nfs_put_client(clp);
6209                 return ERR_PTR(-ENOMEM);
6210         }
6211         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6212         msg.rpc_argp = &calldata->args;
6213         msg.rpc_resp = &calldata->res;
6214         calldata->clp = clp;
6215         task_setup_data.callback_data = calldata;
6216
6217         return rpc_run_task(&task_setup_data);
6218 }
6219
6220 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6221 {
6222         struct rpc_task *task;
6223         int ret = 0;
6224
6225         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6226                 return 0;
6227         task = _nfs41_proc_sequence(clp, cred, &nfs41_sequence_ops);
6228         if (IS_ERR(task))
6229                 ret = PTR_ERR(task);
6230         else
6231                 rpc_put_task_async(task);
6232         dprintk("<-- %s status=%d\n", __func__, ret);
6233         return ret;
6234 }
6235
6236 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6237 {
6238         struct rpc_task *task;
6239         int ret;
6240
6241         task = _nfs41_proc_sequence(clp, cred, &nfs41_sequence_privileged_ops);
6242         if (IS_ERR(task)) {
6243                 ret = PTR_ERR(task);
6244                 goto out;
6245         }
6246         ret = rpc_wait_for_completion_task(task);
6247         if (!ret) {
6248                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6249
6250                 if (task->tk_status == 0)
6251                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6252                 ret = task->tk_status;
6253         }
6254         rpc_put_task(task);
6255 out:
6256         dprintk("<-- %s status=%d\n", __func__, ret);
6257         return ret;
6258 }
6259
6260 struct nfs4_reclaim_complete_data {
6261         struct nfs_client *clp;
6262         struct nfs41_reclaim_complete_args arg;
6263         struct nfs41_reclaim_complete_res res;
6264 };
6265
6266 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6267 {
6268         struct nfs4_reclaim_complete_data *calldata = data;
6269
6270         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
6271         if (nfs41_setup_sequence(calldata->clp->cl_session,
6272                                 &calldata->arg.seq_args,
6273                                 &calldata->res.seq_res, task))
6274                 return;
6275
6276         rpc_call_start(task);
6277 }
6278
6279 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6280 {
6281         switch(task->tk_status) {
6282         case 0:
6283         case -NFS4ERR_COMPLETE_ALREADY:
6284         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6285                 break;
6286         case -NFS4ERR_DELAY:
6287                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6288                 /* fall through */
6289         case -NFS4ERR_RETRY_UNCACHED_REP:
6290                 return -EAGAIN;
6291         default:
6292                 nfs4_schedule_lease_recovery(clp);
6293         }
6294         return 0;
6295 }
6296
6297 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6298 {
6299         struct nfs4_reclaim_complete_data *calldata = data;
6300         struct nfs_client *clp = calldata->clp;
6301         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6302
6303         dprintk("--> %s\n", __func__);
6304         if (!nfs41_sequence_done(task, res))
6305                 return;
6306
6307         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6308                 rpc_restart_call_prepare(task);
6309                 return;
6310         }
6311         dprintk("<-- %s\n", __func__);
6312 }
6313
6314 static void nfs4_free_reclaim_complete_data(void *data)
6315 {
6316         struct nfs4_reclaim_complete_data *calldata = data;
6317
6318         kfree(calldata);
6319 }
6320
6321 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6322         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6323         .rpc_call_done = nfs4_reclaim_complete_done,
6324         .rpc_release = nfs4_free_reclaim_complete_data,
6325 };
6326
6327 /*
6328  * Issue a global reclaim complete.
6329  */
6330 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6331 {
6332         struct nfs4_reclaim_complete_data *calldata;
6333         struct rpc_task *task;
6334         struct rpc_message msg = {
6335                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6336         };
6337         struct rpc_task_setup task_setup_data = {
6338                 .rpc_client = clp->cl_rpcclient,
6339                 .rpc_message = &msg,
6340                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6341                 .flags = RPC_TASK_ASYNC,
6342         };
6343         int status = -ENOMEM;
6344
6345         dprintk("--> %s\n", __func__);
6346         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6347         if (calldata == NULL)
6348                 goto out;
6349         calldata->clp = clp;
6350         calldata->arg.one_fs = 0;
6351
6352         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6353         msg.rpc_argp = &calldata->arg;
6354         msg.rpc_resp = &calldata->res;
6355         task_setup_data.callback_data = calldata;
6356         task = rpc_run_task(&task_setup_data);
6357         if (IS_ERR(task)) {
6358                 status = PTR_ERR(task);
6359                 goto out;
6360         }
6361         status = nfs4_wait_for_completion_rpc_task(task);
6362         if (status == 0)
6363                 status = task->tk_status;
6364         rpc_put_task(task);
6365         return 0;
6366 out:
6367         dprintk("<-- %s status=%d\n", __func__, status);
6368         return status;
6369 }
6370
6371 static void
6372 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6373 {
6374         struct nfs4_layoutget *lgp = calldata;
6375         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6376
6377         dprintk("--> %s\n", __func__);
6378         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6379          * right now covering the LAYOUTGET we are about to send.
6380          * However, that is not so catastrophic, and there seems
6381          * to be no way to prevent it completely.
6382          */
6383         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
6384                                 &lgp->res.seq_res, task))
6385                 return;
6386         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6387                                           NFS_I(lgp->args.inode)->layout,
6388                                           lgp->args.ctx->state)) {
6389                 rpc_exit(task, NFS4_OK);
6390                 return;
6391         }
6392         rpc_call_start(task);
6393 }
6394
6395 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6396 {
6397         struct nfs4_layoutget *lgp = calldata;
6398         struct inode *inode = lgp->args.inode;
6399         struct nfs_server *server = NFS_SERVER(inode);
6400         struct pnfs_layout_hdr *lo;
6401         struct nfs4_state *state = NULL;
6402
6403         dprintk("--> %s\n", __func__);
6404
6405         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
6406                 goto out;
6407
6408         switch (task->tk_status) {
6409         case 0:
6410                 goto out;
6411         case -NFS4ERR_LAYOUTTRYLATER:
6412         case -NFS4ERR_RECALLCONFLICT:
6413                 task->tk_status = -NFS4ERR_DELAY;
6414                 break;
6415         case -NFS4ERR_EXPIRED:
6416         case -NFS4ERR_BAD_STATEID:
6417                 spin_lock(&inode->i_lock);
6418                 lo = NFS_I(inode)->layout;
6419                 if (!lo || list_empty(&lo->plh_segs)) {
6420                         spin_unlock(&inode->i_lock);
6421                         /* If the open stateid was bad, then recover it. */
6422                         state = lgp->args.ctx->state;
6423                 } else {
6424                         LIST_HEAD(head);
6425
6426                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6427                         spin_unlock(&inode->i_lock);
6428                         /* Mark the bad layout state as invalid, then
6429                          * retry using the open stateid. */
6430                         pnfs_free_lseg_list(&head);
6431                 }
6432         }
6433         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6434                 rpc_restart_call_prepare(task);
6435 out:
6436         dprintk("<-- %s\n", __func__);
6437 }
6438
6439 static size_t max_response_pages(struct nfs_server *server)
6440 {
6441         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6442         return nfs_page_array_len(0, max_resp_sz);
6443 }
6444
6445 static void nfs4_free_pages(struct page **pages, size_t size)
6446 {
6447         int i;
6448
6449         if (!pages)
6450                 return;
6451
6452         for (i = 0; i < size; i++) {
6453                 if (!pages[i])
6454                         break;
6455                 __free_page(pages[i]);
6456         }
6457         kfree(pages);
6458 }
6459
6460 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6461 {
6462         struct page **pages;
6463         int i;
6464
6465         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6466         if (!pages) {
6467                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6468                 return NULL;
6469         }
6470
6471         for (i = 0; i < size; i++) {
6472                 pages[i] = alloc_page(gfp_flags);
6473                 if (!pages[i]) {
6474                         dprintk("%s: failed to allocate page\n", __func__);
6475                         nfs4_free_pages(pages, size);
6476                         return NULL;
6477                 }
6478         }
6479
6480         return pages;
6481 }
6482
6483 static void nfs4_layoutget_release(void *calldata)
6484 {
6485         struct nfs4_layoutget *lgp = calldata;
6486         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6487         size_t max_pages = max_response_pages(server);
6488
6489         dprintk("--> %s\n", __func__);
6490         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6491         put_nfs_open_context(lgp->args.ctx);
6492         kfree(calldata);
6493         dprintk("<-- %s\n", __func__);
6494 }
6495
6496 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6497         .rpc_call_prepare = nfs4_layoutget_prepare,
6498         .rpc_call_done = nfs4_layoutget_done,
6499         .rpc_release = nfs4_layoutget_release,
6500 };
6501
6502 struct pnfs_layout_segment *
6503 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6504 {
6505         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6506         size_t max_pages = max_response_pages(server);
6507         struct rpc_task *task;
6508         struct rpc_message msg = {
6509                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6510                 .rpc_argp = &lgp->args,
6511                 .rpc_resp = &lgp->res,
6512         };
6513         struct rpc_task_setup task_setup_data = {
6514                 .rpc_client = server->client,
6515                 .rpc_message = &msg,
6516                 .callback_ops = &nfs4_layoutget_call_ops,
6517                 .callback_data = lgp,
6518                 .flags = RPC_TASK_ASYNC,
6519         };
6520         struct pnfs_layout_segment *lseg = NULL;
6521         int status = 0;
6522
6523         dprintk("--> %s\n", __func__);
6524
6525         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6526         if (!lgp->args.layout.pages) {
6527                 nfs4_layoutget_release(lgp);
6528                 return ERR_PTR(-ENOMEM);
6529         }
6530         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6531
6532         lgp->res.layoutp = &lgp->args.layout;
6533         lgp->res.seq_res.sr_slot = NULL;
6534         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6535         task = rpc_run_task(&task_setup_data);
6536         if (IS_ERR(task))
6537                 return ERR_CAST(task);
6538         status = nfs4_wait_for_completion_rpc_task(task);
6539         if (status == 0)
6540                 status = task->tk_status;
6541         if (status == 0)
6542                 lseg = pnfs_layout_process(lgp);
6543         rpc_put_task(task);
6544         dprintk("<-- %s status=%d\n", __func__, status);
6545         if (status)
6546                 return ERR_PTR(status);
6547         return lseg;
6548 }
6549
6550 static void
6551 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6552 {
6553         struct nfs4_layoutreturn *lrp = calldata;
6554
6555         dprintk("--> %s\n", __func__);
6556         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
6557                                 &lrp->res.seq_res, task))
6558                 return;
6559         rpc_call_start(task);
6560 }
6561
6562 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6563 {
6564         struct nfs4_layoutreturn *lrp = calldata;
6565         struct nfs_server *server;
6566
6567         dprintk("--> %s\n", __func__);
6568
6569         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
6570                 return;
6571
6572         server = NFS_SERVER(lrp->args.inode);
6573         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6574                 rpc_restart_call_prepare(task);
6575                 return;
6576         }
6577         dprintk("<-- %s\n", __func__);
6578 }
6579
6580 static void nfs4_layoutreturn_release(void *calldata)
6581 {
6582         struct nfs4_layoutreturn *lrp = calldata;
6583         struct pnfs_layout_hdr *lo = lrp->args.layout;
6584
6585         dprintk("--> %s\n", __func__);
6586         spin_lock(&lo->plh_inode->i_lock);
6587         if (lrp->res.lrs_present)
6588                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6589         lo->plh_block_lgets--;
6590         spin_unlock(&lo->plh_inode->i_lock);
6591         pnfs_put_layout_hdr(lrp->args.layout);
6592         kfree(calldata);
6593         dprintk("<-- %s\n", __func__);
6594 }
6595
6596 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6597         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6598         .rpc_call_done = nfs4_layoutreturn_done,
6599         .rpc_release = nfs4_layoutreturn_release,
6600 };
6601
6602 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6603 {
6604         struct rpc_task *task;
6605         struct rpc_message msg = {
6606                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6607                 .rpc_argp = &lrp->args,
6608                 .rpc_resp = &lrp->res,
6609         };
6610         struct rpc_task_setup task_setup_data = {
6611                 .rpc_client = lrp->clp->cl_rpcclient,
6612                 .rpc_message = &msg,
6613                 .callback_ops = &nfs4_layoutreturn_call_ops,
6614                 .callback_data = lrp,
6615         };
6616         int status;
6617
6618         dprintk("--> %s\n", __func__);
6619         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6620         task = rpc_run_task(&task_setup_data);
6621         if (IS_ERR(task))
6622                 return PTR_ERR(task);
6623         status = task->tk_status;
6624         dprintk("<-- %s status=%d\n", __func__, status);
6625         rpc_put_task(task);
6626         return status;
6627 }
6628
6629 /*
6630  * Retrieve the list of Data Server devices from the MDS.
6631  */
6632 static int _nfs4_getdevicelist(struct nfs_server *server,
6633                                     const struct nfs_fh *fh,
6634                                     struct pnfs_devicelist *devlist)
6635 {
6636         struct nfs4_getdevicelist_args args = {
6637                 .fh = fh,
6638                 .layoutclass = server->pnfs_curr_ld->id,
6639         };
6640         struct nfs4_getdevicelist_res res = {
6641                 .devlist = devlist,
6642         };
6643         struct rpc_message msg = {
6644                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6645                 .rpc_argp = &args,
6646                 .rpc_resp = &res,
6647         };
6648         int status;
6649
6650         dprintk("--> %s\n", __func__);
6651         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6652                                 &res.seq_res, 0);
6653         dprintk("<-- %s status=%d\n", __func__, status);
6654         return status;
6655 }
6656
6657 int nfs4_proc_getdevicelist(struct nfs_server *server,
6658                             const struct nfs_fh *fh,
6659                             struct pnfs_devicelist *devlist)
6660 {
6661         struct nfs4_exception exception = { };
6662         int err;
6663
6664         do {
6665                 err = nfs4_handle_exception(server,
6666                                 _nfs4_getdevicelist(server, fh, devlist),
6667                                 &exception);
6668         } while (exception.retry);
6669
6670         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6671                 err, devlist->num_devs);
6672
6673         return err;
6674 }
6675 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6676
6677 static int
6678 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6679 {
6680         struct nfs4_getdeviceinfo_args args = {
6681                 .pdev = pdev,
6682         };
6683         struct nfs4_getdeviceinfo_res res = {
6684                 .pdev = pdev,
6685         };
6686         struct rpc_message msg = {
6687                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6688                 .rpc_argp = &args,
6689                 .rpc_resp = &res,
6690         };
6691         int status;
6692
6693         dprintk("--> %s\n", __func__);
6694         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6695         dprintk("<-- %s status=%d\n", __func__, status);
6696
6697         return status;
6698 }
6699
6700 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6701 {
6702         struct nfs4_exception exception = { };
6703         int err;
6704
6705         do {
6706                 err = nfs4_handle_exception(server,
6707                                         _nfs4_proc_getdeviceinfo(server, pdev),
6708                                         &exception);
6709         } while (exception.retry);
6710         return err;
6711 }
6712 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6713
6714 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6715 {
6716         struct nfs4_layoutcommit_data *data = calldata;
6717         struct nfs_server *server = NFS_SERVER(data->args.inode);
6718
6719         if (nfs4_setup_sequence(server, &data->args.seq_args,
6720                                 &data->res.seq_res, task))
6721                 return;
6722         rpc_call_start(task);
6723 }
6724
6725 static void
6726 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6727 {
6728         struct nfs4_layoutcommit_data *data = calldata;
6729         struct nfs_server *server = NFS_SERVER(data->args.inode);
6730
6731         if (!nfs4_sequence_done(task, &data->res.seq_res))
6732                 return;
6733
6734         switch (task->tk_status) { /* Just ignore these failures */
6735         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6736         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6737         case -NFS4ERR_BADLAYOUT:     /* no layout */
6738         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6739                 task->tk_status = 0;
6740                 break;
6741         case 0:
6742                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6743                                                    data->res.fattr);
6744                 break;
6745         default:
6746                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6747                         rpc_restart_call_prepare(task);
6748                         return;
6749                 }
6750         }
6751 }
6752
6753 static void nfs4_layoutcommit_release(void *calldata)
6754 {
6755         struct nfs4_layoutcommit_data *data = calldata;
6756         struct pnfs_layout_segment *lseg, *tmp;
6757         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6758
6759         pnfs_cleanup_layoutcommit(data);
6760         /* Matched by references in pnfs_set_layoutcommit */
6761         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6762                 list_del_init(&lseg->pls_lc_list);
6763                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6764                                        &lseg->pls_flags))
6765                         pnfs_put_lseg(lseg);
6766         }
6767
6768         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6769         smp_mb__after_clear_bit();
6770         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6771
6772         put_rpccred(data->cred);
6773         kfree(data);
6774 }
6775
6776 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6777         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6778         .rpc_call_done = nfs4_layoutcommit_done,
6779         .rpc_release = nfs4_layoutcommit_release,
6780 };
6781
6782 int
6783 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6784 {
6785         struct rpc_message msg = {
6786                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6787                 .rpc_argp = &data->args,
6788                 .rpc_resp = &data->res,
6789                 .rpc_cred = data->cred,
6790         };
6791         struct rpc_task_setup task_setup_data = {
6792                 .task = &data->task,
6793                 .rpc_client = NFS_CLIENT(data->args.inode),
6794                 .rpc_message = &msg,
6795                 .callback_ops = &nfs4_layoutcommit_ops,
6796                 .callback_data = data,
6797                 .flags = RPC_TASK_ASYNC,
6798         };
6799         struct rpc_task *task;
6800         int status = 0;
6801
6802         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6803                 "lbw: %llu inode %lu\n",
6804                 data->task.tk_pid, sync,
6805                 data->args.lastbytewritten,
6806                 data->args.inode->i_ino);
6807
6808         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6809         task = rpc_run_task(&task_setup_data);
6810         if (IS_ERR(task))
6811                 return PTR_ERR(task);
6812         if (sync == false)
6813                 goto out;
6814         status = nfs4_wait_for_completion_rpc_task(task);
6815         if (status != 0)
6816                 goto out;
6817         status = task->tk_status;
6818 out:
6819         dprintk("%s: status %d\n", __func__, status);
6820         rpc_put_task(task);
6821         return status;
6822 }
6823
6824 static int
6825 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6826                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6827 {
6828         struct nfs41_secinfo_no_name_args args = {
6829                 .style = SECINFO_STYLE_CURRENT_FH,
6830         };
6831         struct nfs4_secinfo_res res = {
6832                 .flavors = flavors,
6833         };
6834         struct rpc_message msg = {
6835                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6836                 .rpc_argp = &args,
6837                 .rpc_resp = &res,
6838         };
6839         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6840 }
6841
6842 static int
6843 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6844                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6845 {
6846         struct nfs4_exception exception = { };
6847         int err;
6848         do {
6849                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6850                 switch (err) {
6851                 case 0:
6852                 case -NFS4ERR_WRONGSEC:
6853                 case -NFS4ERR_NOTSUPP:
6854                         goto out;
6855                 default:
6856                         err = nfs4_handle_exception(server, err, &exception);
6857                 }
6858         } while (exception.retry);
6859 out:
6860         return err;
6861 }
6862
6863 static int
6864 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6865                     struct nfs_fsinfo *info)
6866 {
6867         int err;
6868         struct page *page;
6869         rpc_authflavor_t flavor;
6870         struct nfs4_secinfo_flavors *flavors;
6871
6872         page = alloc_page(GFP_KERNEL);
6873         if (!page) {
6874                 err = -ENOMEM;
6875                 goto out;
6876         }
6877
6878         flavors = page_address(page);
6879         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6880
6881         /*
6882          * Fall back on "guess and check" method if
6883          * the server doesn't support SECINFO_NO_NAME
6884          */
6885         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6886                 err = nfs4_find_root_sec(server, fhandle, info);
6887                 goto out_freepage;
6888         }
6889         if (err)
6890                 goto out_freepage;
6891
6892         flavor = nfs_find_best_sec(flavors);
6893         if (err == 0)
6894                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6895
6896 out_freepage:
6897         put_page(page);
6898         if (err == -EACCES)
6899                 return -EPERM;
6900 out:
6901         return err;
6902 }
6903
6904 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6905 {
6906         int status;
6907         struct nfs41_test_stateid_args args = {
6908                 .stateid = stateid,
6909         };
6910         struct nfs41_test_stateid_res res;
6911         struct rpc_message msg = {
6912                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6913                 .rpc_argp = &args,
6914                 .rpc_resp = &res,
6915         };
6916
6917         dprintk("NFS call  test_stateid %p\n", stateid);
6918         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6919         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6920         if (status != NFS_OK) {
6921                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6922                 return status;
6923         }
6924         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6925         return -res.status;
6926 }
6927
6928 /**
6929  * nfs41_test_stateid - perform a TEST_STATEID operation
6930  *
6931  * @server: server / transport on which to perform the operation
6932  * @stateid: state ID to test
6933  *
6934  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6935  * Otherwise a negative NFS4ERR value is returned if the operation
6936  * failed or the state ID is not currently valid.
6937  */
6938 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6939 {
6940         struct nfs4_exception exception = { };
6941         int err;
6942         do {
6943                 err = _nfs41_test_stateid(server, stateid);
6944                 if (err != -NFS4ERR_DELAY)
6945                         break;
6946                 nfs4_handle_exception(server, err, &exception);
6947         } while (exception.retry);
6948         return err;
6949 }
6950
6951 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6952 {
6953         struct nfs41_free_stateid_args args = {
6954                 .stateid = stateid,
6955         };
6956         struct nfs41_free_stateid_res res;
6957         struct rpc_message msg = {
6958                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6959                 .rpc_argp = &args,
6960                 .rpc_resp = &res,
6961         };
6962         int status;
6963
6964         dprintk("NFS call  free_stateid %p\n", stateid);
6965         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6966         status = nfs4_call_sync_sequence(server->client, server, &msg,
6967                                          &args.seq_args, &res.seq_res, 1);
6968         dprintk("NFS reply free_stateid: %d\n", status);
6969         return status;
6970 }
6971
6972 /**
6973  * nfs41_free_stateid - perform a FREE_STATEID operation
6974  *
6975  * @server: server / transport on which to perform the operation
6976  * @stateid: state ID to release
6977  *
6978  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6979  * negative NFS4ERR value is returned.
6980  */
6981 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6982 {
6983         struct nfs4_exception exception = { };
6984         int err;
6985         do {
6986                 err = _nfs4_free_stateid(server, stateid);
6987                 if (err != -NFS4ERR_DELAY)
6988                         break;
6989                 nfs4_handle_exception(server, err, &exception);
6990         } while (exception.retry);
6991         return err;
6992 }
6993
6994 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6995                 const nfs4_stateid *s2)
6996 {
6997         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6998                 return false;
6999
7000         if (s1->seqid == s2->seqid)
7001                 return true;
7002         if (s1->seqid == 0 || s2->seqid == 0)
7003                 return true;
7004
7005         return false;
7006 }
7007
7008 #endif /* CONFIG_NFS_V4_1 */
7009
7010 static bool nfs4_match_stateid(const nfs4_stateid *s1,
7011                 const nfs4_stateid *s2)
7012 {
7013         return nfs4_stateid_match(s1, s2);
7014 }
7015
7016
7017 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
7018         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
7019         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
7020         .recover_open   = nfs4_open_reclaim,
7021         .recover_lock   = nfs4_lock_reclaim,
7022         .establish_clid = nfs4_init_clientid,
7023         .get_clid_cred  = nfs4_get_setclientid_cred,
7024         .detect_trunking = nfs40_discover_server_trunking,
7025 };
7026
7027 #if defined(CONFIG_NFS_V4_1)
7028 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
7029         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
7030         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
7031         .recover_open   = nfs4_open_reclaim,
7032         .recover_lock   = nfs4_lock_reclaim,
7033         .establish_clid = nfs41_init_clientid,
7034         .get_clid_cred  = nfs4_get_exchange_id_cred,
7035         .reclaim_complete = nfs41_proc_reclaim_complete,
7036         .detect_trunking = nfs41_discover_server_trunking,
7037 };
7038 #endif /* CONFIG_NFS_V4_1 */
7039
7040 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
7041         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7042         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
7043         .recover_open   = nfs4_open_expired,
7044         .recover_lock   = nfs4_lock_expired,
7045         .establish_clid = nfs4_init_clientid,
7046         .get_clid_cred  = nfs4_get_setclientid_cred,
7047 };
7048
7049 #if defined(CONFIG_NFS_V4_1)
7050 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
7051         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7052         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
7053         .recover_open   = nfs41_open_expired,
7054         .recover_lock   = nfs41_lock_expired,
7055         .establish_clid = nfs41_init_clientid,
7056         .get_clid_cred  = nfs4_get_exchange_id_cred,
7057 };
7058 #endif /* CONFIG_NFS_V4_1 */
7059
7060 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
7061         .sched_state_renewal = nfs4_proc_async_renew,
7062         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
7063         .renew_lease = nfs4_proc_renew,
7064 };
7065
7066 #if defined(CONFIG_NFS_V4_1)
7067 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
7068         .sched_state_renewal = nfs41_proc_async_sequence,
7069         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
7070         .renew_lease = nfs4_proc_sequence,
7071 };
7072 #endif
7073
7074 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
7075         .minor_version = 0,
7076         .call_sync = _nfs4_call_sync,
7077         .match_stateid = nfs4_match_stateid,
7078         .find_root_sec = nfs4_find_root_sec,
7079         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
7080         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
7081         .state_renewal_ops = &nfs40_state_renewal_ops,
7082 };
7083
7084 #if defined(CONFIG_NFS_V4_1)
7085 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
7086         .minor_version = 1,
7087         .call_sync = _nfs4_call_sync_session,
7088         .match_stateid = nfs41_match_stateid,
7089         .find_root_sec = nfs41_find_root_sec,
7090         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
7091         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
7092         .state_renewal_ops = &nfs41_state_renewal_ops,
7093 };
7094 #endif
7095
7096 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
7097         [0] = &nfs_v4_0_minor_ops,
7098 #if defined(CONFIG_NFS_V4_1)
7099         [1] = &nfs_v4_1_minor_ops,
7100 #endif
7101 };
7102
7103 const struct inode_operations nfs4_dir_inode_operations = {
7104         .create         = nfs_create,
7105         .lookup         = nfs_lookup,
7106         .atomic_open    = nfs_atomic_open,
7107         .link           = nfs_link,
7108         .unlink         = nfs_unlink,
7109         .symlink        = nfs_symlink,
7110         .mkdir          = nfs_mkdir,
7111         .rmdir          = nfs_rmdir,
7112         .mknod          = nfs_mknod,
7113         .rename         = nfs_rename,
7114         .permission     = nfs_permission,
7115         .getattr        = nfs_getattr,
7116         .setattr        = nfs_setattr,
7117         .getxattr       = generic_getxattr,
7118         .setxattr       = generic_setxattr,
7119         .listxattr      = generic_listxattr,
7120         .removexattr    = generic_removexattr,
7121 };
7122
7123 static const struct inode_operations nfs4_file_inode_operations = {
7124         .permission     = nfs_permission,
7125         .getattr        = nfs_getattr,
7126         .setattr        = nfs_setattr,
7127         .getxattr       = generic_getxattr,
7128         .setxattr       = generic_setxattr,
7129         .listxattr      = generic_listxattr,
7130         .removexattr    = generic_removexattr,
7131 };
7132
7133 const struct nfs_rpc_ops nfs_v4_clientops = {
7134         .version        = 4,                    /* protocol version */
7135         .dentry_ops     = &nfs4_dentry_operations,
7136         .dir_inode_ops  = &nfs4_dir_inode_operations,
7137         .file_inode_ops = &nfs4_file_inode_operations,
7138         .file_ops       = &nfs4_file_operations,
7139         .getroot        = nfs4_proc_get_root,
7140         .submount       = nfs4_submount,
7141         .try_mount      = nfs4_try_mount,
7142         .getattr        = nfs4_proc_getattr,
7143         .setattr        = nfs4_proc_setattr,
7144         .lookup         = nfs4_proc_lookup,
7145         .access         = nfs4_proc_access,
7146         .readlink       = nfs4_proc_readlink,
7147         .create         = nfs4_proc_create,
7148         .remove         = nfs4_proc_remove,
7149         .unlink_setup   = nfs4_proc_unlink_setup,
7150         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
7151         .unlink_done    = nfs4_proc_unlink_done,
7152         .rename         = nfs4_proc_rename,
7153         .rename_setup   = nfs4_proc_rename_setup,
7154         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
7155         .rename_done    = nfs4_proc_rename_done,
7156         .link           = nfs4_proc_link,
7157         .symlink        = nfs4_proc_symlink,
7158         .mkdir          = nfs4_proc_mkdir,
7159         .rmdir          = nfs4_proc_remove,
7160         .readdir        = nfs4_proc_readdir,
7161         .mknod          = nfs4_proc_mknod,
7162         .statfs         = nfs4_proc_statfs,
7163         .fsinfo         = nfs4_proc_fsinfo,
7164         .pathconf       = nfs4_proc_pathconf,
7165         .set_capabilities = nfs4_server_capabilities,
7166         .decode_dirent  = nfs4_decode_dirent,
7167         .read_setup     = nfs4_proc_read_setup,
7168         .read_pageio_init = pnfs_pageio_init_read,
7169         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
7170         .read_done      = nfs4_read_done,
7171         .write_setup    = nfs4_proc_write_setup,
7172         .write_pageio_init = pnfs_pageio_init_write,
7173         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
7174         .write_done     = nfs4_write_done,
7175         .commit_setup   = nfs4_proc_commit_setup,
7176         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
7177         .commit_done    = nfs4_commit_done,
7178         .lock           = nfs4_proc_lock,
7179         .clear_acl_cache = nfs4_zap_acl_attr,
7180         .close_context  = nfs4_close_context,
7181         .open_context   = nfs4_atomic_open,
7182         .have_delegation = nfs4_have_delegation,
7183         .return_delegation = nfs4_inode_return_delegation,
7184         .alloc_client   = nfs4_alloc_client,
7185         .init_client    = nfs4_init_client,
7186         .free_client    = nfs4_free_client,
7187         .create_server  = nfs4_create_server,
7188         .clone_server   = nfs_clone_server,
7189 };
7190
7191 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7192         .prefix = XATTR_NAME_NFSV4_ACL,
7193         .list   = nfs4_xattr_list_nfs4_acl,
7194         .get    = nfs4_xattr_get_nfs4_acl,
7195         .set    = nfs4_xattr_set_nfs4_acl,
7196 };
7197
7198 const struct xattr_handler *nfs4_xattr_handlers[] = {
7199         &nfs4_xattr_nfs4_acl_handler,
7200         NULL
7201 };
7202
7203 /*
7204  * Local variables:
7205  *  c-basic-offset: 8
7206  * End:
7207  */