nfs: run state manager in privileged mode
[linux-flexiantxendom0-natty.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/sunrpc/clnt.h>
43 #include <linux/nfs.h>
44 #include <linux/nfs4.h>
45 #include <linux/nfs_fs.h>
46 #include <linux/nfs_page.h>
47 #include <linux/namei.h>
48 #include <linux/mount.h>
49 #include <linux/module.h>
50 #include <linux/sunrpc/bc_xprt.h>
51
52 #include "nfs4_fs.h"
53 #include "delegation.h"
54 #include "internal.h"
55 #include "iostat.h"
56 #include "callback.h"
57
58 #define NFSDBG_FACILITY         NFSDBG_PROC
59
60 #define NFS4_POLL_RETRY_MIN     (HZ/10)
61 #define NFS4_POLL_RETRY_MAX     (15*HZ)
62
63 #define NFS4_MAX_LOOP_ON_RECOVER (10)
64
65 struct nfs4_opendata;
66 static int _nfs4_proc_open(struct nfs4_opendata *data);
67 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
68 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
69 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
70 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
71 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
72
73 /* Prevent leaks of NFSv4 errors into userland */
74 static int nfs4_map_errors(int err)
75 {
76         if (err >= -1000)
77                 return err;
78         switch (err) {
79         case -NFS4ERR_RESOURCE:
80                 return -EREMOTEIO;
81         default:
82                 dprintk("%s could not handle NFSv4 error %d\n",
83                                 __func__, -err);
84                 break;
85         }
86         return -EIO;
87 }
88
89 /*
90  * This is our standard bitmap for GETATTR requests.
91  */
92 const u32 nfs4_fattr_bitmap[2] = {
93         FATTR4_WORD0_TYPE
94         | FATTR4_WORD0_CHANGE
95         | FATTR4_WORD0_SIZE
96         | FATTR4_WORD0_FSID
97         | FATTR4_WORD0_FILEID,
98         FATTR4_WORD1_MODE
99         | FATTR4_WORD1_NUMLINKS
100         | FATTR4_WORD1_OWNER
101         | FATTR4_WORD1_OWNER_GROUP
102         | FATTR4_WORD1_RAWDEV
103         | FATTR4_WORD1_SPACE_USED
104         | FATTR4_WORD1_TIME_ACCESS
105         | FATTR4_WORD1_TIME_METADATA
106         | FATTR4_WORD1_TIME_MODIFY
107 };
108
109 const u32 nfs4_statfs_bitmap[2] = {
110         FATTR4_WORD0_FILES_AVAIL
111         | FATTR4_WORD0_FILES_FREE
112         | FATTR4_WORD0_FILES_TOTAL,
113         FATTR4_WORD1_SPACE_AVAIL
114         | FATTR4_WORD1_SPACE_FREE
115         | FATTR4_WORD1_SPACE_TOTAL
116 };
117
118 const u32 nfs4_pathconf_bitmap[2] = {
119         FATTR4_WORD0_MAXLINK
120         | FATTR4_WORD0_MAXNAME,
121         0
122 };
123
124 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
125                         | FATTR4_WORD0_MAXREAD
126                         | FATTR4_WORD0_MAXWRITE
127                         | FATTR4_WORD0_LEASE_TIME,
128                         0
129 };
130
131 const u32 nfs4_fs_locations_bitmap[2] = {
132         FATTR4_WORD0_TYPE
133         | FATTR4_WORD0_CHANGE
134         | FATTR4_WORD0_SIZE
135         | FATTR4_WORD0_FSID
136         | FATTR4_WORD0_FILEID
137         | FATTR4_WORD0_FS_LOCATIONS,
138         FATTR4_WORD1_MODE
139         | FATTR4_WORD1_NUMLINKS
140         | FATTR4_WORD1_OWNER
141         | FATTR4_WORD1_OWNER_GROUP
142         | FATTR4_WORD1_RAWDEV
143         | FATTR4_WORD1_SPACE_USED
144         | FATTR4_WORD1_TIME_ACCESS
145         | FATTR4_WORD1_TIME_METADATA
146         | FATTR4_WORD1_TIME_MODIFY
147         | FATTR4_WORD1_MOUNTED_ON_FILEID
148 };
149
150 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
151                 struct nfs4_readdir_arg *readdir)
152 {
153         __be32 *start, *p;
154
155         BUG_ON(readdir->count < 80);
156         if (cookie > 2) {
157                 readdir->cookie = cookie;
158                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
159                 return;
160         }
161
162         readdir->cookie = 0;
163         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
164         if (cookie == 2)
165                 return;
166         
167         /*
168          * NFSv4 servers do not return entries for '.' and '..'
169          * Therefore, we fake these entries here.  We let '.'
170          * have cookie 0 and '..' have cookie 1.  Note that
171          * when talking to the server, we always send cookie 0
172          * instead of 1 or 2.
173          */
174         start = p = kmap_atomic(*readdir->pages, KM_USER0);
175         
176         if (cookie == 0) {
177                 *p++ = xdr_one;                                  /* next */
178                 *p++ = xdr_zero;                   /* cookie, first word */
179                 *p++ = xdr_one;                   /* cookie, second word */
180                 *p++ = xdr_one;                             /* entry len */
181                 memcpy(p, ".\0\0\0", 4);                        /* entry */
182                 p++;
183                 *p++ = xdr_one;                         /* bitmap length */
184                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
185                 *p++ = htonl(8);              /* attribute buffer length */
186                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
187         }
188         
189         *p++ = xdr_one;                                  /* next */
190         *p++ = xdr_zero;                   /* cookie, first word */
191         *p++ = xdr_two;                   /* cookie, second word */
192         *p++ = xdr_two;                             /* entry len */
193         memcpy(p, "..\0\0", 4);                         /* entry */
194         p++;
195         *p++ = xdr_one;                         /* bitmap length */
196         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
197         *p++ = htonl(8);              /* attribute buffer length */
198         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
199
200         readdir->pgbase = (char *)p - (char *)start;
201         readdir->count -= readdir->pgbase;
202         kunmap_atomic(start, KM_USER0);
203 }
204
205 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
206 {
207         int res;
208
209         might_sleep();
210
211         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
212                         nfs_wait_bit_killable, TASK_KILLABLE);
213         return res;
214 }
215
216 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
217 {
218         int res = 0;
219
220         might_sleep();
221
222         if (*timeout <= 0)
223                 *timeout = NFS4_POLL_RETRY_MIN;
224         if (*timeout > NFS4_POLL_RETRY_MAX)
225                 *timeout = NFS4_POLL_RETRY_MAX;
226         schedule_timeout_killable(*timeout);
227         if (fatal_signal_pending(current))
228                 res = -ERESTARTSYS;
229         *timeout <<= 1;
230         return res;
231 }
232
233 /* This is the error handling routine for processes that are allowed
234  * to sleep.
235  */
236 static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
237 {
238         struct nfs_client *clp = server->nfs_client;
239         struct nfs4_state *state = exception->state;
240         int ret = errorcode;
241
242         exception->retry = 0;
243         switch(errorcode) {
244                 case 0:
245                         return 0;
246                 case -NFS4ERR_ADMIN_REVOKED:
247                 case -NFS4ERR_BAD_STATEID:
248                 case -NFS4ERR_OPENMODE:
249                         if (state == NULL)
250                                 break;
251                         nfs4_state_mark_reclaim_nograce(clp, state);
252                 case -NFS4ERR_STALE_CLIENTID:
253                 case -NFS4ERR_STALE_STATEID:
254                 case -NFS4ERR_EXPIRED:
255                         nfs4_schedule_state_recovery(clp);
256                         ret = nfs4_wait_clnt_recover(clp);
257                         if (ret == 0)
258                                 exception->retry = 1;
259 #if !defined(CONFIG_NFS_V4_1)
260                         break;
261 #else /* !defined(CONFIG_NFS_V4_1) */
262                         if (!nfs4_has_session(server->nfs_client))
263                                 break;
264                         /* FALLTHROUGH */
265                 case -NFS4ERR_BADSESSION:
266                 case -NFS4ERR_BADSLOT:
267                 case -NFS4ERR_BAD_HIGH_SLOT:
268                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
269                 case -NFS4ERR_DEADSESSION:
270                 case -NFS4ERR_SEQ_FALSE_RETRY:
271                 case -NFS4ERR_SEQ_MISORDERED:
272                         dprintk("%s ERROR: %d Reset session\n", __func__,
273                                 errorcode);
274                         nfs4_schedule_state_recovery(clp);
275                         exception->retry = 1;
276                         break;
277 #endif /* !defined(CONFIG_NFS_V4_1) */
278                 case -NFS4ERR_FILE_OPEN:
279                         if (exception->timeout > HZ) {
280                                 /* We have retried a decent amount, time to
281                                  * fail
282                                  */
283                                 ret = -EBUSY;
284                                 break;
285                         }
286                 case -NFS4ERR_GRACE:
287                 case -NFS4ERR_DELAY:
288                         ret = nfs4_delay(server->client, &exception->timeout);
289                         if (ret != 0)
290                                 break;
291                 case -NFS4ERR_OLD_STATEID:
292                         exception->retry = 1;
293         }
294         /* We failed to handle the error */
295         return nfs4_map_errors(ret);
296 }
297
298
299 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
300 {
301         struct nfs_client *clp = server->nfs_client;
302         spin_lock(&clp->cl_lock);
303         if (time_before(clp->cl_last_renewal,timestamp))
304                 clp->cl_last_renewal = timestamp;
305         spin_unlock(&clp->cl_lock);
306 }
307
308 #if defined(CONFIG_NFS_V4_1)
309
310 /*
311  * nfs4_free_slot - free a slot and efficiently update slot table.
312  *
313  * freeing a slot is trivially done by clearing its respective bit
314  * in the bitmap.
315  * If the freed slotid equals highest_used_slotid we want to update it
316  * so that the server would be able to size down the slot table if needed,
317  * otherwise we know that the highest_used_slotid is still in use.
318  * When updating highest_used_slotid there may be "holes" in the bitmap
319  * so we need to scan down from highest_used_slotid to 0 looking for the now
320  * highest slotid in use.
321  * If none found, highest_used_slotid is set to -1.
322  *
323  * Must be called while holding tbl->slot_tbl_lock
324  */
325 static void
326 nfs4_free_slot(struct nfs4_slot_table *tbl, u8 free_slotid)
327 {
328         int slotid = free_slotid;
329
330         /* clear used bit in bitmap */
331         __clear_bit(slotid, tbl->used_slots);
332
333         /* update highest_used_slotid when it is freed */
334         if (slotid == tbl->highest_used_slotid) {
335                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
336                 if (slotid < tbl->max_slots)
337                         tbl->highest_used_slotid = slotid;
338                 else
339                         tbl->highest_used_slotid = -1;
340         }
341         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
342                 free_slotid, tbl->highest_used_slotid);
343 }
344
345 /*
346  * Signal state manager thread if session is drained
347  */
348 static void nfs41_check_drain_session_complete(struct nfs4_session *ses)
349 {
350         struct rpc_task *task;
351
352         if (!test_bit(NFS4CLNT_SESSION_DRAINING, &ses->clp->cl_state)) {
353                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
354                 if (task)
355                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
356                 return;
357         }
358
359         if (ses->fc_slot_table.highest_used_slotid != -1)
360                 return;
361
362         dprintk("%s COMPLETE: Session Drained\n", __func__);
363         complete(&ses->complete);
364 }
365
366 static void nfs41_sequence_free_slot(const struct nfs_client *clp,
367                               struct nfs4_sequence_res *res)
368 {
369         struct nfs4_slot_table *tbl;
370
371         tbl = &clp->cl_session->fc_slot_table;
372         if (res->sr_slotid == NFS4_MAX_SLOT_TABLE) {
373                 /* just wake up the next guy waiting since
374                  * we may have not consumed a slot after all */
375                 dprintk("%s: No slot\n", __func__);
376                 return;
377         }
378
379         spin_lock(&tbl->slot_tbl_lock);
380         nfs4_free_slot(tbl, res->sr_slotid);
381         nfs41_check_drain_session_complete(clp->cl_session);
382         spin_unlock(&tbl->slot_tbl_lock);
383         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
384 }
385
386 static void nfs41_sequence_done(struct nfs_client *clp,
387                                 struct nfs4_sequence_res *res,
388                                 int rpc_status)
389 {
390         unsigned long timestamp;
391         struct nfs4_slot_table *tbl;
392         struct nfs4_slot *slot;
393
394         /*
395          * sr_status remains 1 if an RPC level error occurred. The server
396          * may or may not have processed the sequence operation..
397          * Proceed as if the server received and processed the sequence
398          * operation.
399          */
400         if (res->sr_status == 1)
401                 res->sr_status = NFS_OK;
402
403         /* -ERESTARTSYS can result in skipping nfs41_sequence_setup */
404         if (res->sr_slotid == NFS4_MAX_SLOT_TABLE)
405                 goto out;
406
407         /* Check the SEQUENCE operation status */
408         if (res->sr_status == 0) {
409                 tbl = &clp->cl_session->fc_slot_table;
410                 slot = tbl->slots + res->sr_slotid;
411                 /* Update the slot's sequence and clientid lease timer */
412                 ++slot->seq_nr;
413                 timestamp = res->sr_renewal_time;
414                 spin_lock(&clp->cl_lock);
415                 if (time_before(clp->cl_last_renewal, timestamp))
416                         clp->cl_last_renewal = timestamp;
417                 spin_unlock(&clp->cl_lock);
418                 /* Check sequence flags */
419                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
420         }
421 out:
422         /* The session may be reset by one of the error handlers. */
423         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
424         nfs41_sequence_free_slot(clp, res);
425 }
426
427 /*
428  * nfs4_find_slot - efficiently look for a free slot
429  *
430  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
431  * If found, we mark the slot as used, update the highest_used_slotid,
432  * and respectively set up the sequence operation args.
433  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
434  *
435  * Note: must be called with under the slot_tbl_lock.
436  */
437 static u8
438 nfs4_find_slot(struct nfs4_slot_table *tbl)
439 {
440         int slotid;
441         u8 ret_id = NFS4_MAX_SLOT_TABLE;
442         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
443
444         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
445                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
446                 tbl->max_slots);
447         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
448         if (slotid >= tbl->max_slots)
449                 goto out;
450         __set_bit(slotid, tbl->used_slots);
451         if (slotid > tbl->highest_used_slotid)
452                 tbl->highest_used_slotid = slotid;
453         ret_id = slotid;
454 out:
455         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
456                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
457         return ret_id;
458 }
459
460 static int nfs41_setup_sequence(struct nfs4_session *session,
461                                 struct nfs4_sequence_args *args,
462                                 struct nfs4_sequence_res *res,
463                                 int cache_reply,
464                                 struct rpc_task *task)
465 {
466         struct nfs4_slot *slot;
467         struct nfs4_slot_table *tbl;
468         u8 slotid;
469
470         dprintk("--> %s\n", __func__);
471         /* slot already allocated? */
472         if (res->sr_slotid != NFS4_MAX_SLOT_TABLE)
473                 return 0;
474
475         memset(res, 0, sizeof(*res));
476         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
477         tbl = &session->fc_slot_table;
478
479         spin_lock(&tbl->slot_tbl_lock);
480         if (test_bit(NFS4CLNT_SESSION_DRAINING, &session->clp->cl_state) &&
481             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
482                 /*
483                  * The state manager will wait until the slot table is empty.
484                  * Schedule the reset thread
485                  */
486                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
487                 spin_unlock(&tbl->slot_tbl_lock);
488                 dprintk("%s Schedule Session Reset\n", __func__);
489                 return -EAGAIN;
490         }
491
492         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
493             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
494                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
495                 spin_unlock(&tbl->slot_tbl_lock);
496                 dprintk("%s enforce FIFO order\n", __func__);
497                 return -EAGAIN;
498         }
499
500         slotid = nfs4_find_slot(tbl);
501         if (slotid == NFS4_MAX_SLOT_TABLE) {
502                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
503                 spin_unlock(&tbl->slot_tbl_lock);
504                 dprintk("<-- %s: no free slots\n", __func__);
505                 return -EAGAIN;
506         }
507         spin_unlock(&tbl->slot_tbl_lock);
508
509         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
510         slot = tbl->slots + slotid;
511         args->sa_session = session;
512         args->sa_slotid = slotid;
513         args->sa_cache_this = cache_reply;
514
515         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
516
517         res->sr_session = session;
518         res->sr_slotid = slotid;
519         res->sr_renewal_time = jiffies;
520         /*
521          * sr_status is only set in decode_sequence, and so will remain
522          * set to 1 if an rpc level failure occurs.
523          */
524         res->sr_status = 1;
525         return 0;
526 }
527
528 int nfs4_setup_sequence(struct nfs_client *clp,
529                         struct nfs4_sequence_args *args,
530                         struct nfs4_sequence_res *res,
531                         int cache_reply,
532                         struct rpc_task *task)
533 {
534         int ret = 0;
535
536         dprintk("--> %s clp %p session %p sr_slotid %d\n",
537                 __func__, clp, clp->cl_session, res->sr_slotid);
538
539         if (!nfs4_has_session(clp))
540                 goto out;
541         ret = nfs41_setup_sequence(clp->cl_session, args, res, cache_reply,
542                                    task);
543         if (ret && ret != -EAGAIN) {
544                 /* terminate rpc task */
545                 task->tk_status = ret;
546                 task->tk_action = NULL;
547         }
548 out:
549         dprintk("<-- %s status=%d\n", __func__, ret);
550         return ret;
551 }
552
553 struct nfs41_call_sync_data {
554         struct nfs_client *clp;
555         struct nfs4_sequence_args *seq_args;
556         struct nfs4_sequence_res *seq_res;
557         int cache_reply;
558 };
559
560 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
561 {
562         struct nfs41_call_sync_data *data = calldata;
563
564         dprintk("--> %s data->clp->cl_session %p\n", __func__,
565                 data->clp->cl_session);
566         if (nfs4_setup_sequence(data->clp, data->seq_args,
567                                 data->seq_res, data->cache_reply, task))
568                 return;
569         rpc_call_start(task);
570 }
571
572 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
573 {
574         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
575         nfs41_call_sync_prepare(task, calldata);
576 }
577
578 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
579 {
580         struct nfs41_call_sync_data *data = calldata;
581
582         nfs41_sequence_done(data->clp, data->seq_res, task->tk_status);
583 }
584
585 struct rpc_call_ops nfs41_call_sync_ops = {
586         .rpc_call_prepare = nfs41_call_sync_prepare,
587         .rpc_call_done = nfs41_call_sync_done,
588 };
589
590 struct rpc_call_ops nfs41_call_priv_sync_ops = {
591         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
592         .rpc_call_done = nfs41_call_sync_done,
593 };
594
595 static int nfs4_call_sync_sequence(struct nfs_client *clp,
596                                    struct rpc_clnt *clnt,
597                                    struct rpc_message *msg,
598                                    struct nfs4_sequence_args *args,
599                                    struct nfs4_sequence_res *res,
600                                    int cache_reply,
601                                    int privileged)
602 {
603         int ret;
604         struct rpc_task *task;
605         struct nfs41_call_sync_data data = {
606                 .clp = clp,
607                 .seq_args = args,
608                 .seq_res = res,
609                 .cache_reply = cache_reply,
610         };
611         struct rpc_task_setup task_setup = {
612                 .rpc_client = clnt,
613                 .rpc_message = msg,
614                 .callback_ops = &nfs41_call_sync_ops,
615                 .callback_data = &data
616         };
617
618         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
619         if (privileged)
620                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
621         task = rpc_run_task(&task_setup);
622         if (IS_ERR(task))
623                 ret = PTR_ERR(task);
624         else {
625                 ret = task->tk_status;
626                 rpc_put_task(task);
627         }
628         return ret;
629 }
630
631 int _nfs4_call_sync_session(struct nfs_server *server,
632                             struct rpc_message *msg,
633                             struct nfs4_sequence_args *args,
634                             struct nfs4_sequence_res *res,
635                             int cache_reply)
636 {
637         return nfs4_call_sync_sequence(server->nfs_client, server->client,
638                                        msg, args, res, cache_reply, 0);
639 }
640
641 #endif /* CONFIG_NFS_V4_1 */
642
643 int _nfs4_call_sync(struct nfs_server *server,
644                     struct rpc_message *msg,
645                     struct nfs4_sequence_args *args,
646                     struct nfs4_sequence_res *res,
647                     int cache_reply)
648 {
649         args->sa_session = res->sr_session = NULL;
650         return rpc_call_sync(server->client, msg, 0);
651 }
652
653 #define nfs4_call_sync(server, msg, args, res, cache_reply) \
654         (server)->nfs_client->cl_call_sync((server), (msg), &(args)->seq_args, \
655                         &(res)->seq_res, (cache_reply))
656
657 static void nfs4_sequence_done(const struct nfs_server *server,
658                                struct nfs4_sequence_res *res, int rpc_status)
659 {
660 #ifdef CONFIG_NFS_V4_1
661         if (nfs4_has_session(server->nfs_client))
662                 nfs41_sequence_done(server->nfs_client, res, rpc_status);
663 #endif /* CONFIG_NFS_V4_1 */
664 }
665
666 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
667 {
668         struct nfs_inode *nfsi = NFS_I(dir);
669
670         spin_lock(&dir->i_lock);
671         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
672         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
673                 nfs_force_lookup_revalidate(dir);
674         nfsi->change_attr = cinfo->after;
675         spin_unlock(&dir->i_lock);
676 }
677
678 struct nfs4_opendata {
679         struct kref kref;
680         struct nfs_openargs o_arg;
681         struct nfs_openres o_res;
682         struct nfs_open_confirmargs c_arg;
683         struct nfs_open_confirmres c_res;
684         struct nfs_fattr f_attr;
685         struct nfs_fattr dir_attr;
686         struct path path;
687         struct dentry *dir;
688         struct nfs4_state_owner *owner;
689         struct nfs4_state *state;
690         struct iattr attrs;
691         unsigned long timestamp;
692         unsigned int rpc_done : 1;
693         int rpc_status;
694         int cancelled;
695 };
696
697
698 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
699 {
700         p->o_res.f_attr = &p->f_attr;
701         p->o_res.dir_attr = &p->dir_attr;
702         p->o_res.seqid = p->o_arg.seqid;
703         p->c_res.seqid = p->c_arg.seqid;
704         p->o_res.server = p->o_arg.server;
705         nfs_fattr_init(&p->f_attr);
706         nfs_fattr_init(&p->dir_attr);
707         p->o_res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
708 }
709
710 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
711                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
712                 const struct iattr *attrs)
713 {
714         struct dentry *parent = dget_parent(path->dentry);
715         struct inode *dir = parent->d_inode;
716         struct nfs_server *server = NFS_SERVER(dir);
717         struct nfs4_opendata *p;
718
719         p = kzalloc(sizeof(*p), GFP_KERNEL);
720         if (p == NULL)
721                 goto err;
722         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
723         if (p->o_arg.seqid == NULL)
724                 goto err_free;
725         p->path.mnt = mntget(path->mnt);
726         p->path.dentry = dget(path->dentry);
727         p->dir = parent;
728         p->owner = sp;
729         atomic_inc(&sp->so_count);
730         p->o_arg.fh = NFS_FH(dir);
731         p->o_arg.open_flags = flags;
732         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
733         p->o_arg.clientid = server->nfs_client->cl_clientid;
734         p->o_arg.id = sp->so_owner_id.id;
735         p->o_arg.name = &p->path.dentry->d_name;
736         p->o_arg.server = server;
737         p->o_arg.bitmask = server->attr_bitmask;
738         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
739         if (flags & O_EXCL) {
740                 if (nfs4_has_persistent_session(server->nfs_client)) {
741                         /* GUARDED */
742                         p->o_arg.u.attrs = &p->attrs;
743                         memcpy(&p->attrs, attrs, sizeof(p->attrs));
744                 } else { /* EXCLUSIVE4_1 */
745                         u32 *s = (u32 *) p->o_arg.u.verifier.data;
746                         s[0] = jiffies;
747                         s[1] = current->pid;
748                 }
749         } else if (flags & O_CREAT) {
750                 p->o_arg.u.attrs = &p->attrs;
751                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
752         }
753         p->c_arg.fh = &p->o_res.fh;
754         p->c_arg.stateid = &p->o_res.stateid;
755         p->c_arg.seqid = p->o_arg.seqid;
756         nfs4_init_opendata_res(p);
757         kref_init(&p->kref);
758         return p;
759 err_free:
760         kfree(p);
761 err:
762         dput(parent);
763         return NULL;
764 }
765
766 static void nfs4_opendata_free(struct kref *kref)
767 {
768         struct nfs4_opendata *p = container_of(kref,
769                         struct nfs4_opendata, kref);
770
771         nfs_free_seqid(p->o_arg.seqid);
772         if (p->state != NULL)
773                 nfs4_put_open_state(p->state);
774         nfs4_put_state_owner(p->owner);
775         dput(p->dir);
776         path_put(&p->path);
777         kfree(p);
778 }
779
780 static void nfs4_opendata_put(struct nfs4_opendata *p)
781 {
782         if (p != NULL)
783                 kref_put(&p->kref, nfs4_opendata_free);
784 }
785
786 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
787 {
788         int ret;
789
790         ret = rpc_wait_for_completion_task(task);
791         return ret;
792 }
793
794 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
795 {
796         int ret = 0;
797
798         if (open_mode & O_EXCL)
799                 goto out;
800         switch (mode & (FMODE_READ|FMODE_WRITE)) {
801                 case FMODE_READ:
802                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
803                                 && state->n_rdonly != 0;
804                         break;
805                 case FMODE_WRITE:
806                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
807                                 && state->n_wronly != 0;
808                         break;
809                 case FMODE_READ|FMODE_WRITE:
810                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
811                                 && state->n_rdwr != 0;
812         }
813 out:
814         return ret;
815 }
816
817 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
818 {
819         if ((delegation->type & fmode) != fmode)
820                 return 0;
821         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
822                 return 0;
823         nfs_mark_delegation_referenced(delegation);
824         return 1;
825 }
826
827 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
828 {
829         switch (fmode) {
830                 case FMODE_WRITE:
831                         state->n_wronly++;
832                         break;
833                 case FMODE_READ:
834                         state->n_rdonly++;
835                         break;
836                 case FMODE_READ|FMODE_WRITE:
837                         state->n_rdwr++;
838         }
839         nfs4_state_set_mode_locked(state, state->state | fmode);
840 }
841
842 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
843 {
844         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
845                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
846         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
847         switch (fmode) {
848                 case FMODE_READ:
849                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
850                         break;
851                 case FMODE_WRITE:
852                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
853                         break;
854                 case FMODE_READ|FMODE_WRITE:
855                         set_bit(NFS_O_RDWR_STATE, &state->flags);
856         }
857 }
858
859 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
860 {
861         write_seqlock(&state->seqlock);
862         nfs_set_open_stateid_locked(state, stateid, fmode);
863         write_sequnlock(&state->seqlock);
864 }
865
866 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
867 {
868         /*
869          * Protect the call to nfs4_state_set_mode_locked and
870          * serialise the stateid update
871          */
872         write_seqlock(&state->seqlock);
873         if (deleg_stateid != NULL) {
874                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
875                 set_bit(NFS_DELEGATED_STATE, &state->flags);
876         }
877         if (open_stateid != NULL)
878                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
879         write_sequnlock(&state->seqlock);
880         spin_lock(&state->owner->so_lock);
881         update_open_stateflags(state, fmode);
882         spin_unlock(&state->owner->so_lock);
883 }
884
885 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
886 {
887         struct nfs_inode *nfsi = NFS_I(state->inode);
888         struct nfs_delegation *deleg_cur;
889         int ret = 0;
890
891         fmode &= (FMODE_READ|FMODE_WRITE);
892
893         rcu_read_lock();
894         deleg_cur = rcu_dereference(nfsi->delegation);
895         if (deleg_cur == NULL)
896                 goto no_delegation;
897
898         spin_lock(&deleg_cur->lock);
899         if (nfsi->delegation != deleg_cur ||
900             (deleg_cur->type & fmode) != fmode)
901                 goto no_delegation_unlock;
902
903         if (delegation == NULL)
904                 delegation = &deleg_cur->stateid;
905         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
906                 goto no_delegation_unlock;
907
908         nfs_mark_delegation_referenced(deleg_cur);
909         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
910         ret = 1;
911 no_delegation_unlock:
912         spin_unlock(&deleg_cur->lock);
913 no_delegation:
914         rcu_read_unlock();
915
916         if (!ret && open_stateid != NULL) {
917                 __update_open_stateid(state, open_stateid, NULL, fmode);
918                 ret = 1;
919         }
920
921         return ret;
922 }
923
924
925 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
926 {
927         struct nfs_delegation *delegation;
928
929         rcu_read_lock();
930         delegation = rcu_dereference(NFS_I(inode)->delegation);
931         if (delegation == NULL || (delegation->type & fmode) == fmode) {
932                 rcu_read_unlock();
933                 return;
934         }
935         rcu_read_unlock();
936         nfs_inode_return_delegation(inode);
937 }
938
939 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
940 {
941         struct nfs4_state *state = opendata->state;
942         struct nfs_inode *nfsi = NFS_I(state->inode);
943         struct nfs_delegation *delegation;
944         int open_mode = opendata->o_arg.open_flags & O_EXCL;
945         fmode_t fmode = opendata->o_arg.fmode;
946         nfs4_stateid stateid;
947         int ret = -EAGAIN;
948
949         for (;;) {
950                 if (can_open_cached(state, fmode, open_mode)) {
951                         spin_lock(&state->owner->so_lock);
952                         if (can_open_cached(state, fmode, open_mode)) {
953                                 update_open_stateflags(state, fmode);
954                                 spin_unlock(&state->owner->so_lock);
955                                 goto out_return_state;
956                         }
957                         spin_unlock(&state->owner->so_lock);
958                 }
959                 rcu_read_lock();
960                 delegation = rcu_dereference(nfsi->delegation);
961                 if (delegation == NULL ||
962                     !can_open_delegated(delegation, fmode)) {
963                         rcu_read_unlock();
964                         break;
965                 }
966                 /* Save the delegation */
967                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
968                 rcu_read_unlock();
969                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
970                 if (ret != 0)
971                         goto out;
972                 ret = -EAGAIN;
973
974                 /* Try to update the stateid using the delegation */
975                 if (update_open_stateid(state, NULL, &stateid, fmode))
976                         goto out_return_state;
977         }
978 out:
979         return ERR_PTR(ret);
980 out_return_state:
981         atomic_inc(&state->count);
982         return state;
983 }
984
985 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
986 {
987         struct inode *inode;
988         struct nfs4_state *state = NULL;
989         struct nfs_delegation *delegation;
990         int ret;
991
992         if (!data->rpc_done) {
993                 state = nfs4_try_open_cached(data);
994                 goto out;
995         }
996
997         ret = -EAGAIN;
998         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
999                 goto err;
1000         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1001         ret = PTR_ERR(inode);
1002         if (IS_ERR(inode))
1003                 goto err;
1004         ret = -ENOMEM;
1005         state = nfs4_get_open_state(inode, data->owner);
1006         if (state == NULL)
1007                 goto err_put_inode;
1008         if (data->o_res.delegation_type != 0) {
1009                 int delegation_flags = 0;
1010
1011                 rcu_read_lock();
1012                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1013                 if (delegation)
1014                         delegation_flags = delegation->flags;
1015                 rcu_read_unlock();
1016                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1017                         nfs_inode_set_delegation(state->inode,
1018                                         data->owner->so_cred,
1019                                         &data->o_res);
1020                 else
1021                         nfs_inode_reclaim_delegation(state->inode,
1022                                         data->owner->so_cred,
1023                                         &data->o_res);
1024         }
1025
1026         update_open_stateid(state, &data->o_res.stateid, NULL,
1027                         data->o_arg.fmode);
1028         iput(inode);
1029 out:
1030         return state;
1031 err_put_inode:
1032         iput(inode);
1033 err:
1034         return ERR_PTR(ret);
1035 }
1036
1037 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1038 {
1039         struct nfs_inode *nfsi = NFS_I(state->inode);
1040         struct nfs_open_context *ctx;
1041
1042         spin_lock(&state->inode->i_lock);
1043         list_for_each_entry(ctx, &nfsi->open_files, list) {
1044                 if (ctx->state != state)
1045                         continue;
1046                 get_nfs_open_context(ctx);
1047                 spin_unlock(&state->inode->i_lock);
1048                 return ctx;
1049         }
1050         spin_unlock(&state->inode->i_lock);
1051         return ERR_PTR(-ENOENT);
1052 }
1053
1054 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1055 {
1056         struct nfs4_opendata *opendata;
1057
1058         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL);
1059         if (opendata == NULL)
1060                 return ERR_PTR(-ENOMEM);
1061         opendata->state = state;
1062         atomic_inc(&state->count);
1063         return opendata;
1064 }
1065
1066 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1067 {
1068         struct nfs4_state *newstate;
1069         int ret;
1070
1071         opendata->o_arg.open_flags = 0;
1072         opendata->o_arg.fmode = fmode;
1073         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1074         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1075         nfs4_init_opendata_res(opendata);
1076         ret = _nfs4_recover_proc_open(opendata);
1077         if (ret != 0)
1078                 return ret; 
1079         newstate = nfs4_opendata_to_nfs4_state(opendata);
1080         if (IS_ERR(newstate))
1081                 return PTR_ERR(newstate);
1082         nfs4_close_state(&opendata->path, newstate, fmode);
1083         *res = newstate;
1084         return 0;
1085 }
1086
1087 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1088 {
1089         struct nfs4_state *newstate;
1090         int ret;
1091
1092         /* memory barrier prior to reading state->n_* */
1093         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1094         smp_rmb();
1095         if (state->n_rdwr != 0) {
1096                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1097                 if (ret != 0)
1098                         return ret;
1099                 if (newstate != state)
1100                         return -ESTALE;
1101         }
1102         if (state->n_wronly != 0) {
1103                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1104                 if (ret != 0)
1105                         return ret;
1106                 if (newstate != state)
1107                         return -ESTALE;
1108         }
1109         if (state->n_rdonly != 0) {
1110                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1111                 if (ret != 0)
1112                         return ret;
1113                 if (newstate != state)
1114                         return -ESTALE;
1115         }
1116         /*
1117          * We may have performed cached opens for all three recoveries.
1118          * Check if we need to update the current stateid.
1119          */
1120         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1121             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1122                 write_seqlock(&state->seqlock);
1123                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1124                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1125                 write_sequnlock(&state->seqlock);
1126         }
1127         return 0;
1128 }
1129
1130 /*
1131  * OPEN_RECLAIM:
1132  *      reclaim state on the server after a reboot.
1133  */
1134 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1135 {
1136         struct nfs_delegation *delegation;
1137         struct nfs4_opendata *opendata;
1138         fmode_t delegation_type = 0;
1139         int status;
1140
1141         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1142         if (IS_ERR(opendata))
1143                 return PTR_ERR(opendata);
1144         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1145         opendata->o_arg.fh = NFS_FH(state->inode);
1146         rcu_read_lock();
1147         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1148         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1149                 delegation_type = delegation->type;
1150         rcu_read_unlock();
1151         opendata->o_arg.u.delegation_type = delegation_type;
1152         status = nfs4_open_recover(opendata, state);
1153         nfs4_opendata_put(opendata);
1154         return status;
1155 }
1156
1157 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1158 {
1159         struct nfs_server *server = NFS_SERVER(state->inode);
1160         struct nfs4_exception exception = { };
1161         int err;
1162         do {
1163                 err = _nfs4_do_open_reclaim(ctx, state);
1164                 if (err != -NFS4ERR_DELAY)
1165                         break;
1166                 nfs4_handle_exception(server, err, &exception);
1167         } while (exception.retry);
1168         return err;
1169 }
1170
1171 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1172 {
1173         struct nfs_open_context *ctx;
1174         int ret;
1175
1176         ctx = nfs4_state_find_open_context(state);
1177         if (IS_ERR(ctx))
1178                 return PTR_ERR(ctx);
1179         ret = nfs4_do_open_reclaim(ctx, state);
1180         put_nfs_open_context(ctx);
1181         return ret;
1182 }
1183
1184 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1185 {
1186         struct nfs4_opendata *opendata;
1187         int ret;
1188
1189         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1190         if (IS_ERR(opendata))
1191                 return PTR_ERR(opendata);
1192         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1193         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1194                         sizeof(opendata->o_arg.u.delegation.data));
1195         ret = nfs4_open_recover(opendata, state);
1196         nfs4_opendata_put(opendata);
1197         return ret;
1198 }
1199
1200 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1201 {
1202         struct nfs4_exception exception = { };
1203         struct nfs_server *server = NFS_SERVER(state->inode);
1204         int err;
1205         do {
1206                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1207                 switch (err) {
1208                         case 0:
1209                         case -ENOENT:
1210                         case -ESTALE:
1211                                 goto out;
1212                         case -NFS4ERR_BADSESSION:
1213                         case -NFS4ERR_BADSLOT:
1214                         case -NFS4ERR_BAD_HIGH_SLOT:
1215                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1216                         case -NFS4ERR_DEADSESSION:
1217                                 nfs4_schedule_state_recovery(
1218                                         server->nfs_client);
1219                                 goto out;
1220                         case -NFS4ERR_STALE_CLIENTID:
1221                         case -NFS4ERR_STALE_STATEID:
1222                         case -NFS4ERR_EXPIRED:
1223                                 /* Don't recall a delegation if it was lost */
1224                                 nfs4_schedule_state_recovery(server->nfs_client);
1225                                 goto out;
1226                         case -ERESTARTSYS:
1227                                 /*
1228                                  * The show must go on: exit, but mark the
1229                                  * stateid as needing recovery.
1230                                  */
1231                         case -NFS4ERR_ADMIN_REVOKED:
1232                         case -NFS4ERR_BAD_STATEID:
1233                                 nfs4_state_mark_reclaim_nograce(server->nfs_client, state);
1234                         case -ENOMEM:
1235                                 err = 0;
1236                                 goto out;
1237                 }
1238                 err = nfs4_handle_exception(server, err, &exception);
1239         } while (exception.retry);
1240 out:
1241         return err;
1242 }
1243
1244 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1245 {
1246         struct nfs4_opendata *data = calldata;
1247
1248         data->rpc_status = task->tk_status;
1249         if (RPC_ASSASSINATED(task))
1250                 return;
1251         if (data->rpc_status == 0) {
1252                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1253                                 sizeof(data->o_res.stateid.data));
1254                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1255                 renew_lease(data->o_res.server, data->timestamp);
1256                 data->rpc_done = 1;
1257         }
1258 }
1259
1260 static void nfs4_open_confirm_release(void *calldata)
1261 {
1262         struct nfs4_opendata *data = calldata;
1263         struct nfs4_state *state = NULL;
1264
1265         /* If this request hasn't been cancelled, do nothing */
1266         if (data->cancelled == 0)
1267                 goto out_free;
1268         /* In case of error, no cleanup! */
1269         if (!data->rpc_done)
1270                 goto out_free;
1271         state = nfs4_opendata_to_nfs4_state(data);
1272         if (!IS_ERR(state))
1273                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1274 out_free:
1275         nfs4_opendata_put(data);
1276 }
1277
1278 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1279         .rpc_call_done = nfs4_open_confirm_done,
1280         .rpc_release = nfs4_open_confirm_release,
1281 };
1282
1283 /*
1284  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1285  */
1286 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1287 {
1288         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1289         struct rpc_task *task;
1290         struct  rpc_message msg = {
1291                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1292                 .rpc_argp = &data->c_arg,
1293                 .rpc_resp = &data->c_res,
1294                 .rpc_cred = data->owner->so_cred,
1295         };
1296         struct rpc_task_setup task_setup_data = {
1297                 .rpc_client = server->client,
1298                 .rpc_message = &msg,
1299                 .callback_ops = &nfs4_open_confirm_ops,
1300                 .callback_data = data,
1301                 .workqueue = nfsiod_workqueue,
1302                 .flags = RPC_TASK_ASYNC,
1303         };
1304         int status;
1305
1306         kref_get(&data->kref);
1307         data->rpc_done = 0;
1308         data->rpc_status = 0;
1309         data->timestamp = jiffies;
1310         task = rpc_run_task(&task_setup_data);
1311         if (IS_ERR(task))
1312                 return PTR_ERR(task);
1313         status = nfs4_wait_for_completion_rpc_task(task);
1314         if (status != 0) {
1315                 data->cancelled = 1;
1316                 smp_wmb();
1317         } else
1318                 status = data->rpc_status;
1319         rpc_put_task(task);
1320         return status;
1321 }
1322
1323 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1324 {
1325         struct nfs4_opendata *data = calldata;
1326         struct nfs4_state_owner *sp = data->owner;
1327
1328         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1329                 return;
1330         /*
1331          * Check if we still need to send an OPEN call, or if we can use
1332          * a delegation instead.
1333          */
1334         if (data->state != NULL) {
1335                 struct nfs_delegation *delegation;
1336
1337                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1338                         goto out_no_action;
1339                 rcu_read_lock();
1340                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1341                 if (delegation != NULL &&
1342                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1343                         rcu_read_unlock();
1344                         goto out_no_action;
1345                 }
1346                 rcu_read_unlock();
1347         }
1348         /* Update sequence id. */
1349         data->o_arg.id = sp->so_owner_id.id;
1350         data->o_arg.clientid = sp->so_client->cl_clientid;
1351         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1352                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1353                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1354         }
1355         data->timestamp = jiffies;
1356         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
1357                                 &data->o_arg.seq_args,
1358                                 &data->o_res.seq_res, 1, task))
1359                 return;
1360         rpc_call_start(task);
1361         return;
1362 out_no_action:
1363         task->tk_action = NULL;
1364
1365 }
1366
1367 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1368 {
1369         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1370         nfs4_open_prepare(task, calldata);
1371 }
1372
1373 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1374 {
1375         struct nfs4_opendata *data = calldata;
1376
1377         data->rpc_status = task->tk_status;
1378
1379         nfs4_sequence_done(data->o_arg.server, &data->o_res.seq_res,
1380                         task->tk_status);
1381
1382         if (RPC_ASSASSINATED(task))
1383                 return;
1384         if (task->tk_status == 0) {
1385                 switch (data->o_res.f_attr->mode & S_IFMT) {
1386                         case S_IFREG:
1387                                 break;
1388                         case S_IFLNK:
1389                                 data->rpc_status = -ELOOP;
1390                                 break;
1391                         case S_IFDIR:
1392                                 data->rpc_status = -EISDIR;
1393                                 break;
1394                         default:
1395                                 data->rpc_status = -ENOTDIR;
1396                 }
1397                 renew_lease(data->o_res.server, data->timestamp);
1398                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1399                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1400         }
1401         data->rpc_done = 1;
1402 }
1403
1404 static void nfs4_open_release(void *calldata)
1405 {
1406         struct nfs4_opendata *data = calldata;
1407         struct nfs4_state *state = NULL;
1408
1409         /* If this request hasn't been cancelled, do nothing */
1410         if (data->cancelled == 0)
1411                 goto out_free;
1412         /* In case of error, no cleanup! */
1413         if (data->rpc_status != 0 || !data->rpc_done)
1414                 goto out_free;
1415         /* In case we need an open_confirm, no cleanup! */
1416         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1417                 goto out_free;
1418         state = nfs4_opendata_to_nfs4_state(data);
1419         if (!IS_ERR(state))
1420                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1421 out_free:
1422         nfs4_opendata_put(data);
1423 }
1424
1425 static const struct rpc_call_ops nfs4_open_ops = {
1426         .rpc_call_prepare = nfs4_open_prepare,
1427         .rpc_call_done = nfs4_open_done,
1428         .rpc_release = nfs4_open_release,
1429 };
1430
1431 static const struct rpc_call_ops nfs4_recover_open_ops = {
1432         .rpc_call_prepare = nfs4_recover_open_prepare,
1433         .rpc_call_done = nfs4_open_done,
1434         .rpc_release = nfs4_open_release,
1435 };
1436
1437 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1438 {
1439         struct inode *dir = data->dir->d_inode;
1440         struct nfs_server *server = NFS_SERVER(dir);
1441         struct nfs_openargs *o_arg = &data->o_arg;
1442         struct nfs_openres *o_res = &data->o_res;
1443         struct rpc_task *task;
1444         struct rpc_message msg = {
1445                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1446                 .rpc_argp = o_arg,
1447                 .rpc_resp = o_res,
1448                 .rpc_cred = data->owner->so_cred,
1449         };
1450         struct rpc_task_setup task_setup_data = {
1451                 .rpc_client = server->client,
1452                 .rpc_message = &msg,
1453                 .callback_ops = &nfs4_open_ops,
1454                 .callback_data = data,
1455                 .workqueue = nfsiod_workqueue,
1456                 .flags = RPC_TASK_ASYNC,
1457         };
1458         int status;
1459
1460         kref_get(&data->kref);
1461         data->rpc_done = 0;
1462         data->rpc_status = 0;
1463         data->cancelled = 0;
1464         if (isrecover)
1465                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1466         task = rpc_run_task(&task_setup_data);
1467         if (IS_ERR(task))
1468                 return PTR_ERR(task);
1469         status = nfs4_wait_for_completion_rpc_task(task);
1470         if (status != 0) {
1471                 data->cancelled = 1;
1472                 smp_wmb();
1473         } else
1474                 status = data->rpc_status;
1475         rpc_put_task(task);
1476
1477         return status;
1478 }
1479
1480 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1481 {
1482         struct inode *dir = data->dir->d_inode;
1483         struct nfs_openres *o_res = &data->o_res;
1484         int status;
1485
1486         status = nfs4_run_open_task(data, 1);
1487         if (status != 0 || !data->rpc_done)
1488                 return status;
1489
1490         nfs_refresh_inode(dir, o_res->dir_attr);
1491
1492         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1493                 status = _nfs4_proc_open_confirm(data);
1494                 if (status != 0)
1495                         return status;
1496         }
1497
1498         return status;
1499 }
1500
1501 /*
1502  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1503  */
1504 static int _nfs4_proc_open(struct nfs4_opendata *data)
1505 {
1506         struct inode *dir = data->dir->d_inode;
1507         struct nfs_server *server = NFS_SERVER(dir);
1508         struct nfs_openargs *o_arg = &data->o_arg;
1509         struct nfs_openres *o_res = &data->o_res;
1510         int status;
1511
1512         status = nfs4_run_open_task(data, 0);
1513         if (status != 0 || !data->rpc_done)
1514                 return status;
1515
1516         if (o_arg->open_flags & O_CREAT) {
1517                 update_changeattr(dir, &o_res->cinfo);
1518                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1519         } else
1520                 nfs_refresh_inode(dir, o_res->dir_attr);
1521         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1522                 status = _nfs4_proc_open_confirm(data);
1523                 if (status != 0)
1524                         return status;
1525         }
1526         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1527                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1528         return 0;
1529 }
1530
1531 static int nfs4_recover_expired_lease(struct nfs_server *server)
1532 {
1533         struct nfs_client *clp = server->nfs_client;
1534         unsigned int loop;
1535         int ret;
1536
1537         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1538                 ret = nfs4_wait_clnt_recover(clp);
1539                 if (ret != 0)
1540                         break;
1541                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1542                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1543                         break;
1544                 nfs4_schedule_state_recovery(clp);
1545                 ret = -EIO;
1546         }
1547         return ret;
1548 }
1549
1550 /*
1551  * OPEN_EXPIRED:
1552  *      reclaim state on the server after a network partition.
1553  *      Assumes caller holds the appropriate lock
1554  */
1555 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1556 {
1557         struct nfs4_opendata *opendata;
1558         int ret;
1559
1560         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1561         if (IS_ERR(opendata))
1562                 return PTR_ERR(opendata);
1563         ret = nfs4_open_recover(opendata, state);
1564         if (ret == -ESTALE)
1565                 d_drop(ctx->path.dentry);
1566         nfs4_opendata_put(opendata);
1567         return ret;
1568 }
1569
1570 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1571 {
1572         struct nfs_server *server = NFS_SERVER(state->inode);
1573         struct nfs4_exception exception = { };
1574         int err;
1575
1576         do {
1577                 err = _nfs4_open_expired(ctx, state);
1578                 switch (err) {
1579                 default:
1580                         goto out;
1581                 case -NFS4ERR_GRACE:
1582                 case -NFS4ERR_DELAY:
1583                         nfs4_handle_exception(server, err, &exception);
1584                         err = 0;
1585                 }
1586         } while (exception.retry);
1587 out:
1588         return err;
1589 }
1590
1591 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1592 {
1593         struct nfs_open_context *ctx;
1594         int ret;
1595
1596         ctx = nfs4_state_find_open_context(state);
1597         if (IS_ERR(ctx))
1598                 return PTR_ERR(ctx);
1599         ret = nfs4_do_open_expired(ctx, state);
1600         put_nfs_open_context(ctx);
1601         return ret;
1602 }
1603
1604 /*
1605  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1606  * fields corresponding to attributes that were used to store the verifier.
1607  * Make sure we clobber those fields in the later setattr call
1608  */
1609 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1610 {
1611         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1612             !(sattr->ia_valid & ATTR_ATIME_SET))
1613                 sattr->ia_valid |= ATTR_ATIME;
1614
1615         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1616             !(sattr->ia_valid & ATTR_MTIME_SET))
1617                 sattr->ia_valid |= ATTR_MTIME;
1618 }
1619
1620 /*
1621  * Returns a referenced nfs4_state
1622  */
1623 static int _nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1624 {
1625         struct nfs4_state_owner  *sp;
1626         struct nfs4_state     *state = NULL;
1627         struct nfs_server       *server = NFS_SERVER(dir);
1628         struct nfs4_opendata *opendata;
1629         int status;
1630
1631         /* Protect against reboot recovery conflicts */
1632         status = -ENOMEM;
1633         if (!(sp = nfs4_get_state_owner(server, cred))) {
1634                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1635                 goto out_err;
1636         }
1637         status = nfs4_recover_expired_lease(server);
1638         if (status != 0)
1639                 goto err_put_state_owner;
1640         if (path->dentry->d_inode != NULL)
1641                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1642         status = -ENOMEM;
1643         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr);
1644         if (opendata == NULL)
1645                 goto err_put_state_owner;
1646
1647         if (path->dentry->d_inode != NULL)
1648                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1649
1650         status = _nfs4_proc_open(opendata);
1651         if (status != 0)
1652                 goto err_opendata_put;
1653
1654         if (opendata->o_arg.open_flags & O_EXCL)
1655                 nfs4_exclusive_attrset(opendata, sattr);
1656
1657         state = nfs4_opendata_to_nfs4_state(opendata);
1658         status = PTR_ERR(state);
1659         if (IS_ERR(state))
1660                 goto err_opendata_put;
1661         nfs4_opendata_put(opendata);
1662         nfs4_put_state_owner(sp);
1663         *res = state;
1664         return 0;
1665 err_opendata_put:
1666         nfs4_opendata_put(opendata);
1667 err_put_state_owner:
1668         nfs4_put_state_owner(sp);
1669 out_err:
1670         *res = NULL;
1671         return status;
1672 }
1673
1674
1675 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1676 {
1677         struct nfs4_exception exception = { };
1678         struct nfs4_state *res;
1679         int status;
1680
1681         do {
1682                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1683                 if (status == 0)
1684                         break;
1685                 /* NOTE: BAD_SEQID means the server and client disagree about the
1686                  * book-keeping w.r.t. state-changing operations
1687                  * (OPEN/CLOSE/LOCK/LOCKU...)
1688                  * It is actually a sign of a bug on the client or on the server.
1689                  *
1690                  * If we receive a BAD_SEQID error in the particular case of
1691                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1692                  * have unhashed the old state_owner for us, and that we can
1693                  * therefore safely retry using a new one. We should still warn
1694                  * the user though...
1695                  */
1696                 if (status == -NFS4ERR_BAD_SEQID) {
1697                         printk(KERN_WARNING "NFS: v4 server %s "
1698                                         " returned a bad sequence-id error!\n",
1699                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1700                         exception.retry = 1;
1701                         continue;
1702                 }
1703                 /*
1704                  * BAD_STATEID on OPEN means that the server cancelled our
1705                  * state before it received the OPEN_CONFIRM.
1706                  * Recover by retrying the request as per the discussion
1707                  * on Page 181 of RFC3530.
1708                  */
1709                 if (status == -NFS4ERR_BAD_STATEID) {
1710                         exception.retry = 1;
1711                         continue;
1712                 }
1713                 if (status == -EAGAIN) {
1714                         /* We must have found a delegation */
1715                         exception.retry = 1;
1716                         continue;
1717                 }
1718                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1719                                         status, &exception));
1720         } while (exception.retry);
1721         return res;
1722 }
1723
1724 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1725                             struct nfs_fattr *fattr, struct iattr *sattr,
1726                             struct nfs4_state *state)
1727 {
1728         struct nfs_server *server = NFS_SERVER(inode);
1729         struct nfs_setattrargs  arg = {
1730                 .fh             = NFS_FH(inode),
1731                 .iap            = sattr,
1732                 .server         = server,
1733                 .bitmask = server->attr_bitmask,
1734         };
1735         struct nfs_setattrres  res = {
1736                 .fattr          = fattr,
1737                 .server         = server,
1738         };
1739         struct rpc_message msg = {
1740                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1741                 .rpc_argp       = &arg,
1742                 .rpc_resp       = &res,
1743                 .rpc_cred       = cred,
1744         };
1745         unsigned long timestamp = jiffies;
1746         int status;
1747
1748         nfs_fattr_init(fattr);
1749
1750         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1751                 /* Use that stateid */
1752         } else if (state != NULL) {
1753                 nfs4_copy_stateid(&arg.stateid, state, current->files);
1754         } else
1755                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1756
1757         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
1758         if (status == 0 && state != NULL)
1759                 renew_lease(server, timestamp);
1760         return status;
1761 }
1762
1763 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1764                            struct nfs_fattr *fattr, struct iattr *sattr,
1765                            struct nfs4_state *state)
1766 {
1767         struct nfs_server *server = NFS_SERVER(inode);
1768         struct nfs4_exception exception = { };
1769         int err;
1770         do {
1771                 err = nfs4_handle_exception(server,
1772                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1773                                 &exception);
1774         } while (exception.retry);
1775         return err;
1776 }
1777
1778 struct nfs4_closedata {
1779         struct path path;
1780         struct inode *inode;
1781         struct nfs4_state *state;
1782         struct nfs_closeargs arg;
1783         struct nfs_closeres res;
1784         struct nfs_fattr fattr;
1785         unsigned long timestamp;
1786 };
1787
1788 static void nfs4_free_closedata(void *data)
1789 {
1790         struct nfs4_closedata *calldata = data;
1791         struct nfs4_state_owner *sp = calldata->state->owner;
1792
1793         nfs4_put_open_state(calldata->state);
1794         nfs_free_seqid(calldata->arg.seqid);
1795         nfs4_put_state_owner(sp);
1796         path_put(&calldata->path);
1797         kfree(calldata);
1798 }
1799
1800 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1801                 fmode_t fmode)
1802 {
1803         spin_lock(&state->owner->so_lock);
1804         if (!(fmode & FMODE_READ))
1805                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1806         if (!(fmode & FMODE_WRITE))
1807                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1808         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1809         spin_unlock(&state->owner->so_lock);
1810 }
1811
1812 static void nfs4_close_done(struct rpc_task *task, void *data)
1813 {
1814         struct nfs4_closedata *calldata = data;
1815         struct nfs4_state *state = calldata->state;
1816         struct nfs_server *server = NFS_SERVER(calldata->inode);
1817
1818         nfs4_sequence_done(server, &calldata->res.seq_res, task->tk_status);
1819         if (RPC_ASSASSINATED(task))
1820                 return;
1821         /* hmm. we are done with the inode, and in the process of freeing
1822          * the state_owner. we keep this around to process errors
1823          */
1824         switch (task->tk_status) {
1825                 case 0:
1826                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1827                         renew_lease(server, calldata->timestamp);
1828                         nfs4_close_clear_stateid_flags(state,
1829                                         calldata->arg.fmode);
1830                         break;
1831                 case -NFS4ERR_STALE_STATEID:
1832                 case -NFS4ERR_OLD_STATEID:
1833                 case -NFS4ERR_BAD_STATEID:
1834                 case -NFS4ERR_EXPIRED:
1835                         if (calldata->arg.fmode == 0)
1836                                 break;
1837                 default:
1838                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN) {
1839                                 nfs_restart_rpc(task, server->nfs_client);
1840                                 return;
1841                         }
1842         }
1843         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1844 }
1845
1846 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1847 {
1848         struct nfs4_closedata *calldata = data;
1849         struct nfs4_state *state = calldata->state;
1850         int call_close = 0;
1851
1852         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1853                 return;
1854
1855         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1856         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1857         spin_lock(&state->owner->so_lock);
1858         /* Calculate the change in open mode */
1859         if (state->n_rdwr == 0) {
1860                 if (state->n_rdonly == 0) {
1861                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
1862                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1863                         calldata->arg.fmode &= ~FMODE_READ;
1864                 }
1865                 if (state->n_wronly == 0) {
1866                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
1867                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1868                         calldata->arg.fmode &= ~FMODE_WRITE;
1869                 }
1870         }
1871         spin_unlock(&state->owner->so_lock);
1872
1873         if (!call_close) {
1874                 /* Note: exit _without_ calling nfs4_close_done */
1875                 task->tk_action = NULL;
1876                 return;
1877         }
1878
1879         if (calldata->arg.fmode == 0)
1880                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
1881
1882         nfs_fattr_init(calldata->res.fattr);
1883         calldata->timestamp = jiffies;
1884         if (nfs4_setup_sequence((NFS_SERVER(calldata->inode))->nfs_client,
1885                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1886                                 1, task))
1887                 return;
1888         rpc_call_start(task);
1889 }
1890
1891 static const struct rpc_call_ops nfs4_close_ops = {
1892         .rpc_call_prepare = nfs4_close_prepare,
1893         .rpc_call_done = nfs4_close_done,
1894         .rpc_release = nfs4_free_closedata,
1895 };
1896
1897 /* 
1898  * It is possible for data to be read/written from a mem-mapped file 
1899  * after the sys_close call (which hits the vfs layer as a flush).
1900  * This means that we can't safely call nfsv4 close on a file until 
1901  * the inode is cleared. This in turn means that we are not good
1902  * NFSv4 citizens - we do not indicate to the server to update the file's 
1903  * share state even when we are done with one of the three share 
1904  * stateid's in the inode.
1905  *
1906  * NOTE: Caller must be holding the sp->so_owner semaphore!
1907  */
1908 int nfs4_do_close(struct path *path, struct nfs4_state *state, int wait)
1909 {
1910         struct nfs_server *server = NFS_SERVER(state->inode);
1911         struct nfs4_closedata *calldata;
1912         struct nfs4_state_owner *sp = state->owner;
1913         struct rpc_task *task;
1914         struct rpc_message msg = {
1915                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
1916                 .rpc_cred = state->owner->so_cred,
1917         };
1918         struct rpc_task_setup task_setup_data = {
1919                 .rpc_client = server->client,
1920                 .rpc_message = &msg,
1921                 .callback_ops = &nfs4_close_ops,
1922                 .workqueue = nfsiod_workqueue,
1923                 .flags = RPC_TASK_ASYNC,
1924         };
1925         int status = -ENOMEM;
1926
1927         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
1928         if (calldata == NULL)
1929                 goto out;
1930         calldata->inode = state->inode;
1931         calldata->state = state;
1932         calldata->arg.fh = NFS_FH(state->inode);
1933         calldata->arg.stateid = &state->open_stateid;
1934         if (nfs4_has_session(server->nfs_client))
1935                 memset(calldata->arg.stateid->data, 0, 4);    /* clear seqid */
1936         /* Serialization for the sequence id */
1937         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid);
1938         if (calldata->arg.seqid == NULL)
1939                 goto out_free_calldata;
1940         calldata->arg.fmode = 0;
1941         calldata->arg.bitmask = server->cache_consistency_bitmask;
1942         calldata->res.fattr = &calldata->fattr;
1943         calldata->res.seqid = calldata->arg.seqid;
1944         calldata->res.server = server;
1945         calldata->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
1946         calldata->path.mnt = mntget(path->mnt);
1947         calldata->path.dentry = dget(path->dentry);
1948
1949         msg.rpc_argp = &calldata->arg,
1950         msg.rpc_resp = &calldata->res,
1951         task_setup_data.callback_data = calldata;
1952         task = rpc_run_task(&task_setup_data);
1953         if (IS_ERR(task))
1954                 return PTR_ERR(task);
1955         status = 0;
1956         if (wait)
1957                 status = rpc_wait_for_completion_task(task);
1958         rpc_put_task(task);
1959         return status;
1960 out_free_calldata:
1961         kfree(calldata);
1962 out:
1963         nfs4_put_open_state(state);
1964         nfs4_put_state_owner(sp);
1965         return status;
1966 }
1967
1968 static int nfs4_intent_set_file(struct nameidata *nd, struct path *path, struct nfs4_state *state, fmode_t fmode)
1969 {
1970         struct file *filp;
1971         int ret;
1972
1973         /* If the open_intent is for execute, we have an extra check to make */
1974         if (fmode & FMODE_EXEC) {
1975                 ret = nfs_may_open(state->inode,
1976                                 state->owner->so_cred,
1977                                 nd->intent.open.flags);
1978                 if (ret < 0)
1979                         goto out_close;
1980         }
1981         filp = lookup_instantiate_filp(nd, path->dentry, NULL);
1982         if (!IS_ERR(filp)) {
1983                 struct nfs_open_context *ctx;
1984                 ctx = nfs_file_open_context(filp);
1985                 ctx->state = state;
1986                 return 0;
1987         }
1988         ret = PTR_ERR(filp);
1989 out_close:
1990         nfs4_close_sync(path, state, fmode & (FMODE_READ|FMODE_WRITE));
1991         return ret;
1992 }
1993
1994 struct dentry *
1995 nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
1996 {
1997         struct path path = {
1998                 .mnt = nd->path.mnt,
1999                 .dentry = dentry,
2000         };
2001         struct dentry *parent;
2002         struct iattr attr;
2003         struct rpc_cred *cred;
2004         struct nfs4_state *state;
2005         struct dentry *res;
2006         fmode_t fmode = nd->intent.open.flags & (FMODE_READ | FMODE_WRITE | FMODE_EXEC);
2007
2008         if (nd->flags & LOOKUP_CREATE) {
2009                 attr.ia_mode = nd->intent.open.create_mode;
2010                 attr.ia_valid = ATTR_MODE;
2011                 if (!IS_POSIXACL(dir))
2012                         attr.ia_mode &= ~current_umask();
2013         } else {
2014                 attr.ia_valid = 0;
2015                 BUG_ON(nd->intent.open.flags & O_CREAT);
2016         }
2017
2018         cred = rpc_lookup_cred();
2019         if (IS_ERR(cred))
2020                 return (struct dentry *)cred;
2021         parent = dentry->d_parent;
2022         /* Protect against concurrent sillydeletes */
2023         nfs_block_sillyrename(parent);
2024         state = nfs4_do_open(dir, &path, fmode, nd->intent.open.flags, &attr, cred);
2025         put_rpccred(cred);
2026         if (IS_ERR(state)) {
2027                 if (PTR_ERR(state) == -ENOENT) {
2028                         d_add(dentry, NULL);
2029                         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2030                 }
2031                 nfs_unblock_sillyrename(parent);
2032                 return (struct dentry *)state;
2033         }
2034         res = d_add_unique(dentry, igrab(state->inode));
2035         if (res != NULL)
2036                 path.dentry = res;
2037         nfs_set_verifier(path.dentry, nfs_save_change_attribute(dir));
2038         nfs_unblock_sillyrename(parent);
2039         nfs4_intent_set_file(nd, &path, state, fmode);
2040         return res;
2041 }
2042
2043 int
2044 nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd)
2045 {
2046         struct path path = {
2047                 .mnt = nd->path.mnt,
2048                 .dentry = dentry,
2049         };
2050         struct rpc_cred *cred;
2051         struct nfs4_state *state;
2052         fmode_t fmode = openflags & (FMODE_READ | FMODE_WRITE);
2053
2054         cred = rpc_lookup_cred();
2055         if (IS_ERR(cred))
2056                 return PTR_ERR(cred);
2057         state = nfs4_do_open(dir, &path, fmode, openflags, NULL, cred);
2058         put_rpccred(cred);
2059         if (IS_ERR(state)) {
2060                 switch (PTR_ERR(state)) {
2061                         case -EPERM:
2062                         case -EACCES:
2063                         case -EDQUOT:
2064                         case -ENOSPC:
2065                         case -EROFS:
2066                                 lookup_instantiate_filp(nd, (struct dentry *)state, NULL);
2067                                 return 1;
2068                         default:
2069                                 goto out_drop;
2070                 }
2071         }
2072         if (state->inode == dentry->d_inode) {
2073                 nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2074                 nfs4_intent_set_file(nd, &path, state, fmode);
2075                 return 1;
2076         }
2077         nfs4_close_sync(&path, state, fmode);
2078 out_drop:
2079         d_drop(dentry);
2080         return 0;
2081 }
2082
2083 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2084 {
2085         if (ctx->state == NULL)
2086                 return;
2087         if (is_sync)
2088                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
2089         else
2090                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
2091 }
2092
2093 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2094 {
2095         struct nfs4_server_caps_arg args = {
2096                 .fhandle = fhandle,
2097         };
2098         struct nfs4_server_caps_res res = {};
2099         struct rpc_message msg = {
2100                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2101                 .rpc_argp = &args,
2102                 .rpc_resp = &res,
2103         };
2104         int status;
2105
2106         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2107         if (status == 0) {
2108                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2109                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2110                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2111                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2112                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2113                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2114                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2115                         server->caps |= NFS_CAP_ACLS;
2116                 if (res.has_links != 0)
2117                         server->caps |= NFS_CAP_HARDLINKS;
2118                 if (res.has_symlinks != 0)
2119                         server->caps |= NFS_CAP_SYMLINKS;
2120                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2121                         server->caps |= NFS_CAP_FILEID;
2122                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2123                         server->caps |= NFS_CAP_MODE;
2124                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2125                         server->caps |= NFS_CAP_NLINK;
2126                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2127                         server->caps |= NFS_CAP_OWNER;
2128                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2129                         server->caps |= NFS_CAP_OWNER_GROUP;
2130                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2131                         server->caps |= NFS_CAP_ATIME;
2132                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2133                         server->caps |= NFS_CAP_CTIME;
2134                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2135                         server->caps |= NFS_CAP_MTIME;
2136
2137                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2138                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2139                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2140                 server->acl_bitmask = res.acl_bitmask;
2141         }
2142
2143         return status;
2144 }
2145
2146 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2147 {
2148         struct nfs4_exception exception = { };
2149         int err;
2150         do {
2151                 err = nfs4_handle_exception(server,
2152                                 _nfs4_server_capabilities(server, fhandle),
2153                                 &exception);
2154         } while (exception.retry);
2155         return err;
2156 }
2157
2158 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2159                 struct nfs_fsinfo *info)
2160 {
2161         struct nfs4_lookup_root_arg args = {
2162                 .bitmask = nfs4_fattr_bitmap,
2163         };
2164         struct nfs4_lookup_res res = {
2165                 .server = server,
2166                 .fattr = info->fattr,
2167                 .fh = fhandle,
2168         };
2169         struct rpc_message msg = {
2170                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2171                 .rpc_argp = &args,
2172                 .rpc_resp = &res,
2173         };
2174
2175         nfs_fattr_init(info->fattr);
2176         return nfs4_call_sync(server, &msg, &args, &res, 0);
2177 }
2178
2179 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2180                 struct nfs_fsinfo *info)
2181 {
2182         struct nfs4_exception exception = { };
2183         int err;
2184         do {
2185                 err = nfs4_handle_exception(server,
2186                                 _nfs4_lookup_root(server, fhandle, info),
2187                                 &exception);
2188         } while (exception.retry);
2189         return err;
2190 }
2191
2192 /*
2193  * get the file handle for the "/" directory on the server
2194  */
2195 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2196                               struct nfs_fsinfo *info)
2197 {
2198         int status;
2199
2200         status = nfs4_lookup_root(server, fhandle, info);
2201         if (status == 0)
2202                 status = nfs4_server_capabilities(server, fhandle);
2203         if (status == 0)
2204                 status = nfs4_do_fsinfo(server, fhandle, info);
2205         return nfs4_map_errors(status);
2206 }
2207
2208 /*
2209  * Get locations and (maybe) other attributes of a referral.
2210  * Note that we'll actually follow the referral later when
2211  * we detect fsid mismatch in inode revalidation
2212  */
2213 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2214 {
2215         int status = -ENOMEM;
2216         struct page *page = NULL;
2217         struct nfs4_fs_locations *locations = NULL;
2218
2219         page = alloc_page(GFP_KERNEL);
2220         if (page == NULL)
2221                 goto out;
2222         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2223         if (locations == NULL)
2224                 goto out;
2225
2226         status = nfs4_proc_fs_locations(dir, name, locations, page);
2227         if (status != 0)
2228                 goto out;
2229         /* Make sure server returned a different fsid for the referral */
2230         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2231                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2232                 status = -EIO;
2233                 goto out;
2234         }
2235
2236         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2237         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2238         if (!fattr->mode)
2239                 fattr->mode = S_IFDIR;
2240         memset(fhandle, 0, sizeof(struct nfs_fh));
2241 out:
2242         if (page)
2243                 __free_page(page);
2244         if (locations)
2245                 kfree(locations);
2246         return status;
2247 }
2248
2249 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2250 {
2251         struct nfs4_getattr_arg args = {
2252                 .fh = fhandle,
2253                 .bitmask = server->attr_bitmask,
2254         };
2255         struct nfs4_getattr_res res = {
2256                 .fattr = fattr,
2257                 .server = server,
2258         };
2259         struct rpc_message msg = {
2260                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2261                 .rpc_argp = &args,
2262                 .rpc_resp = &res,
2263         };
2264         
2265         nfs_fattr_init(fattr);
2266         return nfs4_call_sync(server, &msg, &args, &res, 0);
2267 }
2268
2269 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2270 {
2271         struct nfs4_exception exception = { };
2272         int err;
2273         do {
2274                 err = nfs4_handle_exception(server,
2275                                 _nfs4_proc_getattr(server, fhandle, fattr),
2276                                 &exception);
2277         } while (exception.retry);
2278         return err;
2279 }
2280
2281 /* 
2282  * The file is not closed if it is opened due to the a request to change
2283  * the size of the file. The open call will not be needed once the
2284  * VFS layer lookup-intents are implemented.
2285  *
2286  * Close is called when the inode is destroyed.
2287  * If we haven't opened the file for O_WRONLY, we
2288  * need to in the size_change case to obtain a stateid.
2289  *
2290  * Got race?
2291  * Because OPEN is always done by name in nfsv4, it is
2292  * possible that we opened a different file by the same
2293  * name.  We can recognize this race condition, but we
2294  * can't do anything about it besides returning an error.
2295  *
2296  * This will be fixed with VFS changes (lookup-intent).
2297  */
2298 static int
2299 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2300                   struct iattr *sattr)
2301 {
2302         struct inode *inode = dentry->d_inode;
2303         struct rpc_cred *cred = NULL;
2304         struct nfs4_state *state = NULL;
2305         int status;
2306
2307         nfs_fattr_init(fattr);
2308         
2309         /* Search for an existing open(O_WRITE) file */
2310         if (sattr->ia_valid & ATTR_FILE) {
2311                 struct nfs_open_context *ctx;
2312
2313                 ctx = nfs_file_open_context(sattr->ia_file);
2314                 if (ctx) {
2315                         cred = ctx->cred;
2316                         state = ctx->state;
2317                 }
2318         }
2319
2320         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2321         if (status == 0)
2322                 nfs_setattr_update_inode(inode, sattr);
2323         return status;
2324 }
2325
2326 static int _nfs4_proc_lookupfh(struct nfs_server *server, const struct nfs_fh *dirfh,
2327                 const struct qstr *name, struct nfs_fh *fhandle,
2328                 struct nfs_fattr *fattr)
2329 {
2330         int                    status;
2331         struct nfs4_lookup_arg args = {
2332                 .bitmask = server->attr_bitmask,
2333                 .dir_fh = dirfh,
2334                 .name = name,
2335         };
2336         struct nfs4_lookup_res res = {
2337                 .server = server,
2338                 .fattr = fattr,
2339                 .fh = fhandle,
2340         };
2341         struct rpc_message msg = {
2342                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2343                 .rpc_argp = &args,
2344                 .rpc_resp = &res,
2345         };
2346
2347         nfs_fattr_init(fattr);
2348
2349         dprintk("NFS call  lookupfh %s\n", name->name);
2350         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2351         dprintk("NFS reply lookupfh: %d\n", status);
2352         return status;
2353 }
2354
2355 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2356                               struct qstr *name, struct nfs_fh *fhandle,
2357                               struct nfs_fattr *fattr)
2358 {
2359         struct nfs4_exception exception = { };
2360         int err;
2361         do {
2362                 err = _nfs4_proc_lookupfh(server, dirfh, name, fhandle, fattr);
2363                 /* FIXME: !!!! */
2364                 if (err == -NFS4ERR_MOVED) {
2365                         err = -EREMOTE;
2366                         break;
2367                 }
2368                 err = nfs4_handle_exception(server, err, &exception);
2369         } while (exception.retry);
2370         return err;
2371 }
2372
2373 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
2374                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2375 {
2376         int status;
2377         
2378         dprintk("NFS call  lookup %s\n", name->name);
2379         status = _nfs4_proc_lookupfh(NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2380         if (status == -NFS4ERR_MOVED)
2381                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2382         dprintk("NFS reply lookup: %d\n", status);
2383         return status;
2384 }
2385
2386 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2387 {
2388         struct nfs4_exception exception = { };
2389         int err;
2390         do {
2391                 err = nfs4_handle_exception(NFS_SERVER(dir),
2392                                 _nfs4_proc_lookup(dir, name, fhandle, fattr),
2393                                 &exception);
2394         } while (exception.retry);
2395         return err;
2396 }
2397
2398 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2399 {
2400         struct nfs_server *server = NFS_SERVER(inode);
2401         struct nfs_fattr fattr;
2402         struct nfs4_accessargs args = {
2403                 .fh = NFS_FH(inode),
2404                 .bitmask = server->attr_bitmask,
2405         };
2406         struct nfs4_accessres res = {
2407                 .server = server,
2408                 .fattr = &fattr,
2409         };
2410         struct rpc_message msg = {
2411                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2412                 .rpc_argp = &args,
2413                 .rpc_resp = &res,
2414                 .rpc_cred = entry->cred,
2415         };
2416         int mode = entry->mask;
2417         int status;
2418
2419         /*
2420          * Determine which access bits we want to ask for...
2421          */
2422         if (mode & MAY_READ)
2423                 args.access |= NFS4_ACCESS_READ;
2424         if (S_ISDIR(inode->i_mode)) {
2425                 if (mode & MAY_WRITE)
2426                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2427                 if (mode & MAY_EXEC)
2428                         args.access |= NFS4_ACCESS_LOOKUP;
2429         } else {
2430                 if (mode & MAY_WRITE)
2431                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2432                 if (mode & MAY_EXEC)
2433                         args.access |= NFS4_ACCESS_EXECUTE;
2434         }
2435         nfs_fattr_init(&fattr);
2436         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2437         if (!status) {
2438                 entry->mask = 0;
2439                 if (res.access & NFS4_ACCESS_READ)
2440                         entry->mask |= MAY_READ;
2441                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2442                         entry->mask |= MAY_WRITE;
2443                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2444                         entry->mask |= MAY_EXEC;
2445                 nfs_refresh_inode(inode, &fattr);
2446         }
2447         return status;
2448 }
2449
2450 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2451 {
2452         struct nfs4_exception exception = { };
2453         int err;
2454         do {
2455                 err = nfs4_handle_exception(NFS_SERVER(inode),
2456                                 _nfs4_proc_access(inode, entry),
2457                                 &exception);
2458         } while (exception.retry);
2459         return err;
2460 }
2461
2462 /*
2463  * TODO: For the time being, we don't try to get any attributes
2464  * along with any of the zero-copy operations READ, READDIR,
2465  * READLINK, WRITE.
2466  *
2467  * In the case of the first three, we want to put the GETATTR
2468  * after the read-type operation -- this is because it is hard
2469  * to predict the length of a GETATTR response in v4, and thus
2470  * align the READ data correctly.  This means that the GETATTR
2471  * may end up partially falling into the page cache, and we should
2472  * shift it into the 'tail' of the xdr_buf before processing.
2473  * To do this efficiently, we need to know the total length
2474  * of data received, which doesn't seem to be available outside
2475  * of the RPC layer.
2476  *
2477  * In the case of WRITE, we also want to put the GETATTR after
2478  * the operation -- in this case because we want to make sure
2479  * we get the post-operation mtime and size.  This means that
2480  * we can't use xdr_encode_pages() as written: we need a variant
2481  * of it which would leave room in the 'tail' iovec.
2482  *
2483  * Both of these changes to the XDR layer would in fact be quite
2484  * minor, but I decided to leave them for a subsequent patch.
2485  */
2486 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2487                 unsigned int pgbase, unsigned int pglen)
2488 {
2489         struct nfs4_readlink args = {
2490                 .fh       = NFS_FH(inode),
2491                 .pgbase   = pgbase,
2492                 .pglen    = pglen,
2493                 .pages    = &page,
2494         };
2495         struct nfs4_readlink_res res;
2496         struct rpc_message msg = {
2497                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2498                 .rpc_argp = &args,
2499                 .rpc_resp = &res,
2500         };
2501
2502         return nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
2503 }
2504
2505 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2506                 unsigned int pgbase, unsigned int pglen)
2507 {
2508         struct nfs4_exception exception = { };
2509         int err;
2510         do {
2511                 err = nfs4_handle_exception(NFS_SERVER(inode),
2512                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2513                                 &exception);
2514         } while (exception.retry);
2515         return err;
2516 }
2517
2518 /*
2519  * Got race?
2520  * We will need to arrange for the VFS layer to provide an atomic open.
2521  * Until then, this create/open method is prone to inefficiency and race
2522  * conditions due to the lookup, create, and open VFS calls from sys_open()
2523  * placed on the wire.
2524  *
2525  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2526  * The file will be opened again in the subsequent VFS open call
2527  * (nfs4_proc_file_open).
2528  *
2529  * The open for read will just hang around to be used by any process that
2530  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2531  */
2532
2533 static int
2534 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2535                  int flags, struct nameidata *nd)
2536 {
2537         struct path path = {
2538                 .mnt = nd->path.mnt,
2539                 .dentry = dentry,
2540         };
2541         struct nfs4_state *state;
2542         struct rpc_cred *cred;
2543         fmode_t fmode = flags & (FMODE_READ | FMODE_WRITE);
2544         int status = 0;
2545
2546         cred = rpc_lookup_cred();
2547         if (IS_ERR(cred)) {
2548                 status = PTR_ERR(cred);
2549                 goto out;
2550         }
2551         state = nfs4_do_open(dir, &path, fmode, flags, sattr, cred);
2552         d_drop(dentry);
2553         if (IS_ERR(state)) {
2554                 status = PTR_ERR(state);
2555                 goto out_putcred;
2556         }
2557         d_add(dentry, igrab(state->inode));
2558         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2559         if (flags & O_EXCL) {
2560                 struct nfs_fattr fattr;
2561                 status = nfs4_do_setattr(state->inode, cred, &fattr, sattr, state);
2562                 if (status == 0)
2563                         nfs_setattr_update_inode(state->inode, sattr);
2564                 nfs_post_op_update_inode(state->inode, &fattr);
2565         }
2566         if (status == 0 && (nd->flags & LOOKUP_OPEN) != 0)
2567                 status = nfs4_intent_set_file(nd, &path, state, fmode);
2568         else
2569                 nfs4_close_sync(&path, state, fmode);
2570 out_putcred:
2571         put_rpccred(cred);
2572 out:
2573         return status;
2574 }
2575
2576 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2577 {
2578         struct nfs_server *server = NFS_SERVER(dir);
2579         struct nfs_removeargs args = {
2580                 .fh = NFS_FH(dir),
2581                 .name.len = name->len,
2582                 .name.name = name->name,
2583                 .bitmask = server->attr_bitmask,
2584         };
2585         struct nfs_removeres res = {
2586                 .server = server,
2587         };
2588         struct rpc_message msg = {
2589                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2590                 .rpc_argp = &args,
2591                 .rpc_resp = &res,
2592         };
2593         int                     status;
2594
2595         nfs_fattr_init(&res.dir_attr);
2596         status = nfs4_call_sync(server, &msg, &args, &res, 1);
2597         if (status == 0) {
2598                 update_changeattr(dir, &res.cinfo);
2599                 nfs_post_op_update_inode(dir, &res.dir_attr);
2600         }
2601         return status;
2602 }
2603
2604 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2605 {
2606         struct nfs4_exception exception = { };
2607         int err;
2608         do {
2609                 err = nfs4_handle_exception(NFS_SERVER(dir),
2610                                 _nfs4_proc_remove(dir, name),
2611                                 &exception);
2612         } while (exception.retry);
2613         return err;
2614 }
2615
2616 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2617 {
2618         struct nfs_server *server = NFS_SERVER(dir);
2619         struct nfs_removeargs *args = msg->rpc_argp;
2620         struct nfs_removeres *res = msg->rpc_resp;
2621
2622         args->bitmask = server->cache_consistency_bitmask;
2623         res->server = server;
2624         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2625 }
2626
2627 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2628 {
2629         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2630
2631         nfs4_sequence_done(res->server, &res->seq_res, task->tk_status);
2632         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2633                 return 0;
2634         update_changeattr(dir, &res->cinfo);
2635         nfs_post_op_update_inode(dir, &res->dir_attr);
2636         return 1;
2637 }
2638
2639 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2640                 struct inode *new_dir, struct qstr *new_name)
2641 {
2642         struct nfs_server *server = NFS_SERVER(old_dir);
2643         struct nfs4_rename_arg arg = {
2644                 .old_dir = NFS_FH(old_dir),
2645                 .new_dir = NFS_FH(new_dir),
2646                 .old_name = old_name,
2647                 .new_name = new_name,
2648                 .bitmask = server->attr_bitmask,
2649         };
2650         struct nfs_fattr old_fattr, new_fattr;
2651         struct nfs4_rename_res res = {
2652                 .server = server,
2653                 .old_fattr = &old_fattr,
2654                 .new_fattr = &new_fattr,
2655         };
2656         struct rpc_message msg = {
2657                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2658                 .rpc_argp = &arg,
2659                 .rpc_resp = &res,
2660         };
2661         int                     status;
2662         
2663         nfs_fattr_init(res.old_fattr);
2664         nfs_fattr_init(res.new_fattr);
2665         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2666
2667         if (!status) {
2668                 update_changeattr(old_dir, &res.old_cinfo);
2669                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2670                 update_changeattr(new_dir, &res.new_cinfo);
2671                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2672         }
2673         return status;
2674 }
2675
2676 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2677                 struct inode *new_dir, struct qstr *new_name)
2678 {
2679         struct nfs4_exception exception = { };
2680         int err;
2681         do {
2682                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2683                                 _nfs4_proc_rename(old_dir, old_name,
2684                                         new_dir, new_name),
2685                                 &exception);
2686         } while (exception.retry);
2687         return err;
2688 }
2689
2690 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2691 {
2692         struct nfs_server *server = NFS_SERVER(inode);
2693         struct nfs4_link_arg arg = {
2694                 .fh     = NFS_FH(inode),
2695                 .dir_fh = NFS_FH(dir),
2696                 .name   = name,
2697                 .bitmask = server->attr_bitmask,
2698         };
2699         struct nfs_fattr fattr, dir_attr;
2700         struct nfs4_link_res res = {
2701                 .server = server,
2702                 .fattr = &fattr,
2703                 .dir_attr = &dir_attr,
2704         };
2705         struct rpc_message msg = {
2706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2707                 .rpc_argp = &arg,
2708                 .rpc_resp = &res,
2709         };
2710         int                     status;
2711
2712         nfs_fattr_init(res.fattr);
2713         nfs_fattr_init(res.dir_attr);
2714         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2715         if (!status) {
2716                 update_changeattr(dir, &res.cinfo);
2717                 nfs_post_op_update_inode(dir, res.dir_attr);
2718                 nfs_post_op_update_inode(inode, res.fattr);
2719         }
2720
2721         return status;
2722 }
2723
2724 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2725 {
2726         struct nfs4_exception exception = { };
2727         int err;
2728         do {
2729                 err = nfs4_handle_exception(NFS_SERVER(inode),
2730                                 _nfs4_proc_link(inode, dir, name),
2731                                 &exception);
2732         } while (exception.retry);
2733         return err;
2734 }
2735
2736 struct nfs4_createdata {
2737         struct rpc_message msg;
2738         struct nfs4_create_arg arg;
2739         struct nfs4_create_res res;
2740         struct nfs_fh fh;
2741         struct nfs_fattr fattr;
2742         struct nfs_fattr dir_fattr;
2743 };
2744
2745 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2746                 struct qstr *name, struct iattr *sattr, u32 ftype)
2747 {
2748         struct nfs4_createdata *data;
2749
2750         data = kzalloc(sizeof(*data), GFP_KERNEL);
2751         if (data != NULL) {
2752                 struct nfs_server *server = NFS_SERVER(dir);
2753
2754                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2755                 data->msg.rpc_argp = &data->arg;
2756                 data->msg.rpc_resp = &data->res;
2757                 data->arg.dir_fh = NFS_FH(dir);
2758                 data->arg.server = server;
2759                 data->arg.name = name;
2760                 data->arg.attrs = sattr;
2761                 data->arg.ftype = ftype;
2762                 data->arg.bitmask = server->attr_bitmask;
2763                 data->res.server = server;
2764                 data->res.fh = &data->fh;
2765                 data->res.fattr = &data->fattr;
2766                 data->res.dir_fattr = &data->dir_fattr;
2767                 nfs_fattr_init(data->res.fattr);
2768                 nfs_fattr_init(data->res.dir_fattr);
2769         }
2770         return data;
2771 }
2772
2773 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2774 {
2775         int status = nfs4_call_sync(NFS_SERVER(dir), &data->msg,
2776                                     &data->arg, &data->res, 1);
2777         if (status == 0) {
2778                 update_changeattr(dir, &data->res.dir_cinfo);
2779                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2780                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2781         }
2782         return status;
2783 }
2784
2785 static void nfs4_free_createdata(struct nfs4_createdata *data)
2786 {
2787         kfree(data);
2788 }
2789
2790 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2791                 struct page *page, unsigned int len, struct iattr *sattr)
2792 {
2793         struct nfs4_createdata *data;
2794         int status = -ENAMETOOLONG;
2795
2796         if (len > NFS4_MAXPATHLEN)
2797                 goto out;
2798
2799         status = -ENOMEM;
2800         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2801         if (data == NULL)
2802                 goto out;
2803
2804         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2805         data->arg.u.symlink.pages = &page;
2806         data->arg.u.symlink.len = len;
2807         
2808         status = nfs4_do_create(dir, dentry, data);
2809
2810         nfs4_free_createdata(data);
2811 out:
2812         return status;
2813 }
2814
2815 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2816                 struct page *page, unsigned int len, struct iattr *sattr)
2817 {
2818         struct nfs4_exception exception = { };
2819         int err;
2820         do {
2821                 err = nfs4_handle_exception(NFS_SERVER(dir),
2822                                 _nfs4_proc_symlink(dir, dentry, page,
2823                                                         len, sattr),
2824                                 &exception);
2825         } while (exception.retry);
2826         return err;
2827 }
2828
2829 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2830                 struct iattr *sattr)
2831 {
2832         struct nfs4_createdata *data;
2833         int status = -ENOMEM;
2834
2835         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2836         if (data == NULL)
2837                 goto out;
2838
2839         status = nfs4_do_create(dir, dentry, data);
2840
2841         nfs4_free_createdata(data);
2842 out:
2843         return status;
2844 }
2845
2846 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2847                 struct iattr *sattr)
2848 {
2849         struct nfs4_exception exception = { };
2850         int err;
2851         do {
2852                 err = nfs4_handle_exception(NFS_SERVER(dir),
2853                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2854                                 &exception);
2855         } while (exception.retry);
2856         return err;
2857 }
2858
2859 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2860                   u64 cookie, struct page *page, unsigned int count, int plus)
2861 {
2862         struct inode            *dir = dentry->d_inode;
2863         struct nfs4_readdir_arg args = {
2864                 .fh = NFS_FH(dir),
2865                 .pages = &page,
2866                 .pgbase = 0,
2867                 .count = count,
2868                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2869         };
2870         struct nfs4_readdir_res res;
2871         struct rpc_message msg = {
2872                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2873                 .rpc_argp = &args,
2874                 .rpc_resp = &res,
2875                 .rpc_cred = cred,
2876         };
2877         int                     status;
2878
2879         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2880                         dentry->d_parent->d_name.name,
2881                         dentry->d_name.name,
2882                         (unsigned long long)cookie);
2883         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2884         res.pgbase = args.pgbase;
2885         status = nfs4_call_sync(NFS_SERVER(dir), &msg, &args, &res, 0);
2886         if (status == 0)
2887                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2888
2889         nfs_invalidate_atime(dir);
2890
2891         dprintk("%s: returns %d\n", __func__, status);
2892         return status;
2893 }
2894
2895 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2896                   u64 cookie, struct page *page, unsigned int count, int plus)
2897 {
2898         struct nfs4_exception exception = { };
2899         int err;
2900         do {
2901                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2902                                 _nfs4_proc_readdir(dentry, cred, cookie,
2903                                         page, count, plus),
2904                                 &exception);
2905         } while (exception.retry);
2906         return err;
2907 }
2908
2909 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2910                 struct iattr *sattr, dev_t rdev)
2911 {
2912         struct nfs4_createdata *data;
2913         int mode = sattr->ia_mode;
2914         int status = -ENOMEM;
2915
2916         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
2917         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
2918
2919         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
2920         if (data == NULL)
2921                 goto out;
2922
2923         if (S_ISFIFO(mode))
2924                 data->arg.ftype = NF4FIFO;
2925         else if (S_ISBLK(mode)) {
2926                 data->arg.ftype = NF4BLK;
2927                 data->arg.u.device.specdata1 = MAJOR(rdev);
2928                 data->arg.u.device.specdata2 = MINOR(rdev);
2929         }
2930         else if (S_ISCHR(mode)) {
2931                 data->arg.ftype = NF4CHR;
2932                 data->arg.u.device.specdata1 = MAJOR(rdev);
2933                 data->arg.u.device.specdata2 = MINOR(rdev);
2934         }
2935         
2936         status = nfs4_do_create(dir, dentry, data);
2937
2938         nfs4_free_createdata(data);
2939 out:
2940         return status;
2941 }
2942
2943 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2944                 struct iattr *sattr, dev_t rdev)
2945 {
2946         struct nfs4_exception exception = { };
2947         int err;
2948         do {
2949                 err = nfs4_handle_exception(NFS_SERVER(dir),
2950                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
2951                                 &exception);
2952         } while (exception.retry);
2953         return err;
2954 }
2955
2956 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
2957                  struct nfs_fsstat *fsstat)
2958 {
2959         struct nfs4_statfs_arg args = {
2960                 .fh = fhandle,
2961                 .bitmask = server->attr_bitmask,
2962         };
2963         struct nfs4_statfs_res res = {
2964                 .fsstat = fsstat,
2965         };
2966         struct rpc_message msg = {
2967                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
2968                 .rpc_argp = &args,
2969                 .rpc_resp = &res,
2970         };
2971
2972         nfs_fattr_init(fsstat->fattr);
2973         return  nfs4_call_sync(server, &msg, &args, &res, 0);
2974 }
2975
2976 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
2977 {
2978         struct nfs4_exception exception = { };
2979         int err;
2980         do {
2981                 err = nfs4_handle_exception(server,
2982                                 _nfs4_proc_statfs(server, fhandle, fsstat),
2983                                 &exception);
2984         } while (exception.retry);
2985         return err;
2986 }
2987
2988 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
2989                 struct nfs_fsinfo *fsinfo)
2990 {
2991         struct nfs4_fsinfo_arg args = {
2992                 .fh = fhandle,
2993                 .bitmask = server->attr_bitmask,
2994         };
2995         struct nfs4_fsinfo_res res = {
2996                 .fsinfo = fsinfo,
2997         };
2998         struct rpc_message msg = {
2999                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3000                 .rpc_argp = &args,
3001                 .rpc_resp = &res,
3002         };
3003
3004         return nfs4_call_sync(server, &msg, &args, &res, 0);
3005 }
3006
3007 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3008 {
3009         struct nfs4_exception exception = { };
3010         int err;
3011
3012         do {
3013                 err = nfs4_handle_exception(server,
3014                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3015                                 &exception);
3016         } while (exception.retry);
3017         return err;
3018 }
3019
3020 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3021 {
3022         nfs_fattr_init(fsinfo->fattr);
3023         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3024 }
3025
3026 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3027                 struct nfs_pathconf *pathconf)
3028 {
3029         struct nfs4_pathconf_arg args = {
3030                 .fh = fhandle,
3031                 .bitmask = server->attr_bitmask,
3032         };
3033         struct nfs4_pathconf_res res = {
3034                 .pathconf = pathconf,
3035         };
3036         struct rpc_message msg = {
3037                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3038                 .rpc_argp = &args,
3039                 .rpc_resp = &res,
3040         };
3041
3042         /* None of the pathconf attributes are mandatory to implement */
3043         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3044                 memset(pathconf, 0, sizeof(*pathconf));
3045                 return 0;
3046         }
3047
3048         nfs_fattr_init(pathconf->fattr);
3049         return nfs4_call_sync(server, &msg, &args, &res, 0);
3050 }
3051
3052 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3053                 struct nfs_pathconf *pathconf)
3054 {
3055         struct nfs4_exception exception = { };
3056         int err;
3057
3058         do {
3059                 err = nfs4_handle_exception(server,
3060                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3061                                 &exception);
3062         } while (exception.retry);
3063         return err;
3064 }
3065
3066 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3067 {
3068         struct nfs_server *server = NFS_SERVER(data->inode);
3069
3070         dprintk("--> %s\n", __func__);
3071
3072         nfs4_sequence_done(server, &data->res.seq_res, task->tk_status);
3073
3074         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3075                 nfs_restart_rpc(task, server->nfs_client);
3076                 return -EAGAIN;
3077         }
3078
3079         nfs_invalidate_atime(data->inode);
3080         if (task->tk_status > 0)
3081                 renew_lease(server, data->timestamp);
3082         return 0;
3083 }
3084
3085 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3086 {
3087         data->timestamp   = jiffies;
3088         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3089 }
3090
3091 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3092 {
3093         struct inode *inode = data->inode;
3094         
3095         nfs4_sequence_done(NFS_SERVER(inode), &data->res.seq_res,
3096                            task->tk_status);
3097
3098         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3099                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3100                 return -EAGAIN;
3101         }
3102         if (task->tk_status >= 0) {
3103                 renew_lease(NFS_SERVER(inode), data->timestamp);
3104                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3105         }
3106         return 0;
3107 }
3108
3109 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3110 {
3111         struct nfs_server *server = NFS_SERVER(data->inode);
3112
3113         data->args.bitmask = server->cache_consistency_bitmask;
3114         data->res.server = server;
3115         data->timestamp   = jiffies;
3116
3117         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3118 }
3119
3120 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3121 {
3122         struct inode *inode = data->inode;
3123         
3124         nfs4_sequence_done(NFS_SERVER(inode), &data->res.seq_res,
3125                            task->tk_status);
3126         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3127                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3128                 return -EAGAIN;
3129         }
3130         nfs_refresh_inode(inode, data->res.fattr);
3131         return 0;
3132 }
3133
3134 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3135 {
3136         struct nfs_server *server = NFS_SERVER(data->inode);
3137         
3138         data->args.bitmask = server->cache_consistency_bitmask;
3139         data->res.server = server;
3140         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3141 }
3142
3143 /*
3144  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3145  * standalone procedure for queueing an asynchronous RENEW.
3146  */
3147 static void nfs4_renew_done(struct rpc_task *task, void *data)
3148 {
3149         struct nfs_client *clp = (struct nfs_client *)task->tk_msg.rpc_argp;
3150         unsigned long timestamp = (unsigned long)data;
3151
3152         if (task->tk_status < 0) {
3153                 /* Unless we're shutting down, schedule state recovery! */
3154                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
3155                         nfs4_schedule_state_recovery(clp);
3156                 return;
3157         }
3158         spin_lock(&clp->cl_lock);
3159         if (time_before(clp->cl_last_renewal,timestamp))
3160                 clp->cl_last_renewal = timestamp;
3161         spin_unlock(&clp->cl_lock);
3162 }
3163
3164 static const struct rpc_call_ops nfs4_renew_ops = {
3165         .rpc_call_done = nfs4_renew_done,
3166 };
3167
3168 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
3169 {
3170         struct rpc_message msg = {
3171                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3172                 .rpc_argp       = clp,
3173                 .rpc_cred       = cred,
3174         };
3175
3176         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3177                         &nfs4_renew_ops, (void *)jiffies);
3178 }
3179
3180 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3181 {
3182         struct rpc_message msg = {
3183                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3184                 .rpc_argp       = clp,
3185                 .rpc_cred       = cred,
3186         };
3187         unsigned long now = jiffies;
3188         int status;
3189
3190         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3191         if (status < 0)
3192                 return status;
3193         spin_lock(&clp->cl_lock);
3194         if (time_before(clp->cl_last_renewal,now))
3195                 clp->cl_last_renewal = now;
3196         spin_unlock(&clp->cl_lock);
3197         return 0;
3198 }
3199
3200 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3201 {
3202         return (server->caps & NFS_CAP_ACLS)
3203                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3204                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3205 }
3206
3207 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3208  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3209  * the stack.
3210  */
3211 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3212
3213 static void buf_to_pages(const void *buf, size_t buflen,
3214                 struct page **pages, unsigned int *pgbase)
3215 {
3216         const void *p = buf;
3217
3218         *pgbase = offset_in_page(buf);
3219         p -= *pgbase;
3220         while (p < buf + buflen) {
3221                 *(pages++) = virt_to_page(p);
3222                 p += PAGE_CACHE_SIZE;
3223         }
3224 }
3225
3226 struct nfs4_cached_acl {
3227         int cached;
3228         size_t len;
3229         char data[0];
3230 };
3231
3232 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3233 {
3234         struct nfs_inode *nfsi = NFS_I(inode);
3235
3236         spin_lock(&inode->i_lock);
3237         kfree(nfsi->nfs4_acl);
3238         nfsi->nfs4_acl = acl;
3239         spin_unlock(&inode->i_lock);
3240 }
3241
3242 static void nfs4_zap_acl_attr(struct inode *inode)
3243 {
3244         nfs4_set_cached_acl(inode, NULL);
3245 }
3246
3247 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3248 {
3249         struct nfs_inode *nfsi = NFS_I(inode);
3250         struct nfs4_cached_acl *acl;
3251         int ret = -ENOENT;
3252
3253         spin_lock(&inode->i_lock);
3254         acl = nfsi->nfs4_acl;
3255         if (acl == NULL)
3256                 goto out;
3257         if (buf == NULL) /* user is just asking for length */
3258                 goto out_len;
3259         if (acl->cached == 0)
3260                 goto out;
3261         ret = -ERANGE; /* see getxattr(2) man page */
3262         if (acl->len > buflen)
3263                 goto out;
3264         memcpy(buf, acl->data, acl->len);
3265 out_len:
3266         ret = acl->len;
3267 out:
3268         spin_unlock(&inode->i_lock);
3269         return ret;
3270 }
3271
3272 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3273 {
3274         struct nfs4_cached_acl *acl;
3275
3276         if (buf && acl_len <= PAGE_SIZE) {
3277                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3278                 if (acl == NULL)
3279                         goto out;
3280                 acl->cached = 1;
3281                 memcpy(acl->data, buf, acl_len);
3282         } else {
3283                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3284                 if (acl == NULL)
3285                         goto out;
3286                 acl->cached = 0;
3287         }
3288         acl->len = acl_len;
3289 out:
3290         nfs4_set_cached_acl(inode, acl);
3291 }
3292
3293 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3294 {
3295         struct page *pages[NFS4ACL_MAXPAGES];
3296         struct nfs_getaclargs args = {
3297                 .fh = NFS_FH(inode),
3298                 .acl_pages = pages,
3299                 .acl_len = buflen,
3300         };
3301         struct nfs_getaclres res = {
3302                 .acl_len = buflen,
3303         };
3304         void *resp_buf;
3305         struct rpc_message msg = {
3306                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3307                 .rpc_argp = &args,
3308                 .rpc_resp = &res,
3309         };
3310         struct page *localpage = NULL;
3311         int ret;
3312
3313         if (buflen < PAGE_SIZE) {
3314                 /* As long as we're doing a round trip to the server anyway,
3315                  * let's be prepared for a page of acl data. */
3316                 localpage = alloc_page(GFP_KERNEL);
3317                 resp_buf = page_address(localpage);
3318                 if (localpage == NULL)
3319                         return -ENOMEM;
3320                 args.acl_pages[0] = localpage;
3321                 args.acl_pgbase = 0;
3322                 args.acl_len = PAGE_SIZE;
3323         } else {
3324                 resp_buf = buf;
3325                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3326         }
3327         ret = nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
3328         if (ret)
3329                 goto out_free;
3330         if (res.acl_len > args.acl_len)
3331                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3332         else
3333                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3334         if (buf) {
3335                 ret = -ERANGE;
3336                 if (res.acl_len > buflen)
3337                         goto out_free;
3338                 if (localpage)
3339                         memcpy(buf, resp_buf, res.acl_len);
3340         }
3341         ret = res.acl_len;
3342 out_free:
3343         if (localpage)
3344                 __free_page(localpage);
3345         return ret;
3346 }
3347
3348 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3349 {
3350         struct nfs4_exception exception = { };
3351         ssize_t ret;
3352         do {
3353                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3354                 if (ret >= 0)
3355                         break;
3356                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3357         } while (exception.retry);
3358         return ret;
3359 }
3360
3361 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3362 {
3363         struct nfs_server *server = NFS_SERVER(inode);
3364         int ret;
3365
3366         if (!nfs4_server_supports_acls(server))
3367                 return -EOPNOTSUPP;
3368         ret = nfs_revalidate_inode(server, inode);
3369         if (ret < 0)
3370                 return ret;
3371         ret = nfs4_read_cached_acl(inode, buf, buflen);
3372         if (ret != -ENOENT)
3373                 return ret;
3374         return nfs4_get_acl_uncached(inode, buf, buflen);
3375 }
3376
3377 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3378 {
3379         struct nfs_server *server = NFS_SERVER(inode);
3380         struct page *pages[NFS4ACL_MAXPAGES];
3381         struct nfs_setaclargs arg = {
3382                 .fh             = NFS_FH(inode),
3383                 .acl_pages      = pages,
3384                 .acl_len        = buflen,
3385         };
3386         struct nfs_setaclres res;
3387         struct rpc_message msg = {
3388                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3389                 .rpc_argp       = &arg,
3390                 .rpc_resp       = &res,
3391         };
3392         int ret;
3393
3394         if (!nfs4_server_supports_acls(server))
3395                 return -EOPNOTSUPP;
3396         nfs_inode_return_delegation(inode);
3397         buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3398         ret = nfs4_call_sync(server, &msg, &arg, &res, 1);
3399         nfs_access_zap_cache(inode);
3400         nfs_zap_acl_cache(inode);
3401         return ret;
3402 }
3403
3404 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3405 {
3406         struct nfs4_exception exception = { };
3407         int err;
3408         do {
3409                 err = nfs4_handle_exception(NFS_SERVER(inode),
3410                                 __nfs4_proc_set_acl(inode, buf, buflen),
3411                                 &exception);
3412         } while (exception.retry);
3413         return err;
3414 }
3415
3416 static int
3417 _nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs_client *clp, struct nfs4_state *state)
3418 {
3419         if (!clp || task->tk_status >= 0)
3420                 return 0;
3421         switch(task->tk_status) {
3422                 case -NFS4ERR_ADMIN_REVOKED:
3423                 case -NFS4ERR_BAD_STATEID:
3424                 case -NFS4ERR_OPENMODE:
3425                         if (state == NULL)
3426                                 break;
3427                         nfs4_state_mark_reclaim_nograce(clp, state);
3428                 case -NFS4ERR_STALE_CLIENTID:
3429                 case -NFS4ERR_STALE_STATEID:
3430                 case -NFS4ERR_EXPIRED:
3431                         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3432                         nfs4_schedule_state_recovery(clp);
3433                         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3434                                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3435                         task->tk_status = 0;
3436                         return -EAGAIN;
3437 #if defined(CONFIG_NFS_V4_1)
3438                 case -NFS4ERR_BADSESSION:
3439                 case -NFS4ERR_BADSLOT:
3440                 case -NFS4ERR_BAD_HIGH_SLOT:
3441                 case -NFS4ERR_DEADSESSION:
3442                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3443                 case -NFS4ERR_SEQ_FALSE_RETRY:
3444                 case -NFS4ERR_SEQ_MISORDERED:
3445                         dprintk("%s ERROR %d, Reset session\n", __func__,
3446                                 task->tk_status);
3447                         nfs4_schedule_state_recovery(clp);
3448                         task->tk_status = 0;
3449                         return -EAGAIN;
3450 #endif /* CONFIG_NFS_V4_1 */
3451                 case -NFS4ERR_DELAY:
3452                         if (server)
3453                                 nfs_inc_server_stats(server, NFSIOS_DELAY);
3454                 case -NFS4ERR_GRACE:
3455                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3456                         task->tk_status = 0;
3457                         return -EAGAIN;
3458                 case -NFS4ERR_OLD_STATEID:
3459                         task->tk_status = 0;
3460                         return -EAGAIN;
3461         }
3462         task->tk_status = nfs4_map_errors(task->tk_status);
3463         return 0;
3464 }
3465
3466 static int
3467 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3468 {
3469         return _nfs4_async_handle_error(task, server, server->nfs_client, state);
3470 }
3471
3472 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program, unsigned short port, struct rpc_cred *cred)
3473 {
3474         nfs4_verifier sc_verifier;
3475         struct nfs4_setclientid setclientid = {
3476                 .sc_verifier = &sc_verifier,
3477                 .sc_prog = program,
3478         };
3479         struct rpc_message msg = {
3480                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3481                 .rpc_argp = &setclientid,
3482                 .rpc_resp = clp,
3483                 .rpc_cred = cred,
3484         };
3485         __be32 *p;
3486         int loop = 0;
3487         int status;
3488
3489         p = (__be32*)sc_verifier.data;
3490         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3491         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3492
3493         for(;;) {
3494                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3495                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3496                                 clp->cl_ipaddr,
3497                                 rpc_peeraddr2str(clp->cl_rpcclient,
3498                                                         RPC_DISPLAY_ADDR),
3499                                 rpc_peeraddr2str(clp->cl_rpcclient,
3500                                                         RPC_DISPLAY_PROTO),
3501                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3502                                 clp->cl_id_uniquifier);
3503                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3504                                 sizeof(setclientid.sc_netid),
3505                                 rpc_peeraddr2str(clp->cl_rpcclient,
3506                                                         RPC_DISPLAY_NETID));
3507                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3508                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3509                                 clp->cl_ipaddr, port >> 8, port & 255);
3510
3511                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3512                 if (status != -NFS4ERR_CLID_INUSE)
3513                         break;
3514                 if (signalled())
3515                         break;
3516                 if (loop++ & 1)
3517                         ssleep(clp->cl_lease_time + 1);
3518                 else
3519                         if (++clp->cl_id_uniquifier == 0)
3520                                 break;
3521         }
3522         return status;
3523 }
3524
3525 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp, struct rpc_cred *cred)
3526 {
3527         struct nfs_fsinfo fsinfo;
3528         struct rpc_message msg = {
3529                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3530                 .rpc_argp = clp,
3531                 .rpc_resp = &fsinfo,
3532                 .rpc_cred = cred,
3533         };
3534         unsigned long now;
3535         int status;
3536
3537         now = jiffies;
3538         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3539         if (status == 0) {
3540                 spin_lock(&clp->cl_lock);
3541                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3542                 clp->cl_last_renewal = now;
3543                 spin_unlock(&clp->cl_lock);
3544         }
3545         return status;
3546 }
3547
3548 int nfs4_proc_setclientid_confirm(struct nfs_client *clp, struct rpc_cred *cred)
3549 {
3550         long timeout = 0;
3551         int err;
3552         do {
3553                 err = _nfs4_proc_setclientid_confirm(clp, cred);
3554                 switch (err) {
3555                         case 0:
3556                                 return err;
3557                         case -NFS4ERR_RESOURCE:
3558                                 /* The IBM lawyers misread another document! */
3559                         case -NFS4ERR_DELAY:
3560                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3561                 }
3562         } while (err == 0);
3563         return err;
3564 }
3565
3566 struct nfs4_delegreturndata {
3567         struct nfs4_delegreturnargs args;
3568         struct nfs4_delegreturnres res;
3569         struct nfs_fh fh;
3570         nfs4_stateid stateid;
3571         unsigned long timestamp;
3572         struct nfs_fattr fattr;
3573         int rpc_status;
3574 };
3575
3576 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3577 {
3578         struct nfs4_delegreturndata *data = calldata;
3579
3580         nfs4_sequence_done(data->res.server, &data->res.seq_res,
3581                         task->tk_status);
3582
3583         switch (task->tk_status) {
3584         case -NFS4ERR_STALE_STATEID:
3585         case -NFS4ERR_EXPIRED:
3586         case 0:
3587                 renew_lease(data->res.server, data->timestamp);
3588                 break;
3589         default:
3590                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3591                                 -EAGAIN) {
3592                         nfs_restart_rpc(task, data->res.server->nfs_client);
3593                         return;
3594                 }
3595         }
3596         data->rpc_status = task->tk_status;
3597 }
3598
3599 static void nfs4_delegreturn_release(void *calldata)
3600 {
3601         kfree(calldata);
3602 }
3603
3604 #if defined(CONFIG_NFS_V4_1)
3605 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3606 {
3607         struct nfs4_delegreturndata *d_data;
3608
3609         d_data = (struct nfs4_delegreturndata *)data;
3610
3611         if (nfs4_setup_sequence(d_data->res.server->nfs_client,
3612                                 &d_data->args.seq_args,
3613                                 &d_data->res.seq_res, 1, task))
3614                 return;
3615         rpc_call_start(task);
3616 }
3617 #endif /* CONFIG_NFS_V4_1 */
3618
3619 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3620 #if defined(CONFIG_NFS_V4_1)
3621         .rpc_call_prepare = nfs4_delegreturn_prepare,
3622 #endif /* CONFIG_NFS_V4_1 */
3623         .rpc_call_done = nfs4_delegreturn_done,
3624         .rpc_release = nfs4_delegreturn_release,
3625 };
3626
3627 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3628 {
3629         struct nfs4_delegreturndata *data;
3630         struct nfs_server *server = NFS_SERVER(inode);
3631         struct rpc_task *task;
3632         struct rpc_message msg = {
3633                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3634                 .rpc_cred = cred,
3635         };
3636         struct rpc_task_setup task_setup_data = {
3637                 .rpc_client = server->client,
3638                 .rpc_message = &msg,
3639                 .callback_ops = &nfs4_delegreturn_ops,
3640                 .flags = RPC_TASK_ASYNC,
3641         };
3642         int status = 0;
3643
3644         data = kzalloc(sizeof(*data), GFP_KERNEL);
3645         if (data == NULL)
3646                 return -ENOMEM;
3647         data->args.fhandle = &data->fh;
3648         data->args.stateid = &data->stateid;
3649         data->args.bitmask = server->attr_bitmask;
3650         nfs_copy_fh(&data->fh, NFS_FH(inode));
3651         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3652         data->res.fattr = &data->fattr;
3653         data->res.server = server;
3654         data->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3655         nfs_fattr_init(data->res.fattr);
3656         data->timestamp = jiffies;
3657         data->rpc_status = 0;
3658
3659         task_setup_data.callback_data = data;
3660         msg.rpc_argp = &data->args,
3661         msg.rpc_resp = &data->res,
3662         task = rpc_run_task(&task_setup_data);
3663         if (IS_ERR(task))
3664                 return PTR_ERR(task);
3665         if (!issync)
3666                 goto out;
3667         status = nfs4_wait_for_completion_rpc_task(task);
3668         if (status != 0)
3669                 goto out;
3670         status = data->rpc_status;
3671         if (status != 0)
3672                 goto out;
3673         nfs_refresh_inode(inode, &data->fattr);
3674 out:
3675         rpc_put_task(task);
3676         return status;
3677 }
3678
3679 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3680 {
3681         struct nfs_server *server = NFS_SERVER(inode);
3682         struct nfs4_exception exception = { };
3683         int err;
3684         do {
3685                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3686                 switch (err) {
3687                         case -NFS4ERR_STALE_STATEID:
3688                         case -NFS4ERR_EXPIRED:
3689                         case 0:
3690                                 return 0;
3691                 }
3692                 err = nfs4_handle_exception(server, err, &exception);
3693         } while (exception.retry);
3694         return err;
3695 }
3696
3697 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3698 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3699
3700 /* 
3701  * sleep, with exponential backoff, and retry the LOCK operation. 
3702  */
3703 static unsigned long
3704 nfs4_set_lock_task_retry(unsigned long timeout)
3705 {
3706         schedule_timeout_killable(timeout);
3707         timeout <<= 1;
3708         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3709                 return NFS4_LOCK_MAXTIMEOUT;
3710         return timeout;
3711 }
3712
3713 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3714 {
3715         struct inode *inode = state->inode;
3716         struct nfs_server *server = NFS_SERVER(inode);
3717         struct nfs_client *clp = server->nfs_client;
3718         struct nfs_lockt_args arg = {
3719                 .fh = NFS_FH(inode),
3720                 .fl = request,
3721         };
3722         struct nfs_lockt_res res = {
3723                 .denied = request,
3724         };
3725         struct rpc_message msg = {
3726                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3727                 .rpc_argp       = &arg,
3728                 .rpc_resp       = &res,
3729                 .rpc_cred       = state->owner->so_cred,
3730         };
3731         struct nfs4_lock_state *lsp;
3732         int status;
3733
3734         arg.lock_owner.clientid = clp->cl_clientid;
3735         status = nfs4_set_lock_state(state, request);
3736         if (status != 0)
3737                 goto out;
3738         lsp = request->fl_u.nfs4_fl.owner;
3739         arg.lock_owner.id = lsp->ls_id.id;
3740         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
3741         switch (status) {
3742                 case 0:
3743                         request->fl_type = F_UNLCK;
3744                         break;
3745                 case -NFS4ERR_DENIED:
3746                         status = 0;
3747         }
3748         request->fl_ops->fl_release_private(request);
3749 out:
3750         return status;
3751 }
3752
3753 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3754 {
3755         struct nfs4_exception exception = { };
3756         int err;
3757
3758         do {
3759                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3760                                 _nfs4_proc_getlk(state, cmd, request),
3761                                 &exception);
3762         } while (exception.retry);
3763         return err;
3764 }
3765
3766 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3767 {
3768         int res = 0;
3769         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3770                 case FL_POSIX:
3771                         res = posix_lock_file_wait(file, fl);
3772                         break;
3773                 case FL_FLOCK:
3774                         res = flock_lock_file_wait(file, fl);
3775                         break;
3776                 default:
3777                         BUG();
3778         }
3779         return res;
3780 }
3781
3782 struct nfs4_unlockdata {
3783         struct nfs_locku_args arg;
3784         struct nfs_locku_res res;
3785         struct nfs4_lock_state *lsp;
3786         struct nfs_open_context *ctx;
3787         struct file_lock fl;
3788         const struct nfs_server *server;
3789         unsigned long timestamp;
3790 };
3791
3792 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
3793                 struct nfs_open_context *ctx,
3794                 struct nfs4_lock_state *lsp,
3795                 struct nfs_seqid *seqid)
3796 {
3797         struct nfs4_unlockdata *p;
3798         struct inode *inode = lsp->ls_state->inode;
3799
3800         p = kzalloc(sizeof(*p), GFP_KERNEL);
3801         if (p == NULL)
3802                 return NULL;
3803         p->arg.fh = NFS_FH(inode);
3804         p->arg.fl = &p->fl;
3805         p->arg.seqid = seqid;
3806         p->res.seqid = seqid;
3807         p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3808         p->arg.stateid = &lsp->ls_stateid;
3809         p->lsp = lsp;
3810         atomic_inc(&lsp->ls_count);
3811         /* Ensure we don't close file until we're done freeing locks! */
3812         p->ctx = get_nfs_open_context(ctx);
3813         memcpy(&p->fl, fl, sizeof(p->fl));
3814         p->server = NFS_SERVER(inode);
3815         return p;
3816 }
3817
3818 static void nfs4_locku_release_calldata(void *data)
3819 {
3820         struct nfs4_unlockdata *calldata = data;
3821         nfs_free_seqid(calldata->arg.seqid);
3822         nfs4_put_lock_state(calldata->lsp);
3823         put_nfs_open_context(calldata->ctx);
3824         kfree(calldata);
3825 }
3826
3827 static void nfs4_locku_done(struct rpc_task *task, void *data)
3828 {
3829         struct nfs4_unlockdata *calldata = data;
3830
3831         nfs4_sequence_done(calldata->server, &calldata->res.seq_res,
3832                            task->tk_status);
3833         if (RPC_ASSASSINATED(task))
3834                 return;
3835         switch (task->tk_status) {
3836                 case 0:
3837                         memcpy(calldata->lsp->ls_stateid.data,
3838                                         calldata->res.stateid.data,
3839                                         sizeof(calldata->lsp->ls_stateid.data));
3840                         renew_lease(calldata->server, calldata->timestamp);
3841                         break;
3842                 case -NFS4ERR_BAD_STATEID:
3843                 case -NFS4ERR_OLD_STATEID:
3844                 case -NFS4ERR_STALE_STATEID:
3845                 case -NFS4ERR_EXPIRED:
3846                         break;
3847                 default:
3848                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
3849                                 nfs_restart_rpc(task,
3850                                                  calldata->server->nfs_client);
3851         }
3852 }
3853
3854 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
3855 {
3856         struct nfs4_unlockdata *calldata = data;
3857
3858         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3859                 return;
3860         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
3861                 /* Note: exit _without_ running nfs4_locku_done */
3862                 task->tk_action = NULL;
3863                 return;
3864         }
3865         calldata->timestamp = jiffies;
3866         if (nfs4_setup_sequence(calldata->server->nfs_client,
3867                                 &calldata->arg.seq_args,
3868                                 &calldata->res.seq_res, 1, task))
3869                 return;
3870         rpc_call_start(task);
3871 }
3872
3873 static const struct rpc_call_ops nfs4_locku_ops = {
3874         .rpc_call_prepare = nfs4_locku_prepare,
3875         .rpc_call_done = nfs4_locku_done,
3876         .rpc_release = nfs4_locku_release_calldata,
3877 };
3878
3879 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
3880                 struct nfs_open_context *ctx,
3881                 struct nfs4_lock_state *lsp,
3882                 struct nfs_seqid *seqid)
3883 {
3884         struct nfs4_unlockdata *data;
3885         struct rpc_message msg = {
3886                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
3887                 .rpc_cred = ctx->cred,
3888         };
3889         struct rpc_task_setup task_setup_data = {
3890                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
3891                 .rpc_message = &msg,
3892                 .callback_ops = &nfs4_locku_ops,
3893                 .workqueue = nfsiod_workqueue,
3894                 .flags = RPC_TASK_ASYNC,
3895         };
3896
3897         /* Ensure this is an unlock - when canceling a lock, the
3898          * canceled lock is passed in, and it won't be an unlock.
3899          */
3900         fl->fl_type = F_UNLCK;
3901
3902         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
3903         if (data == NULL) {
3904                 nfs_free_seqid(seqid);
3905                 return ERR_PTR(-ENOMEM);
3906         }
3907
3908         msg.rpc_argp = &data->arg,
3909         msg.rpc_resp = &data->res,
3910         task_setup_data.callback_data = data;
3911         return rpc_run_task(&task_setup_data);
3912 }
3913
3914 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
3915 {
3916         struct nfs_inode *nfsi = NFS_I(state->inode);
3917         struct nfs_seqid *seqid;
3918         struct nfs4_lock_state *lsp;
3919         struct rpc_task *task;
3920         int status = 0;
3921         unsigned char fl_flags = request->fl_flags;
3922
3923         status = nfs4_set_lock_state(state, request);
3924         /* Unlock _before_ we do the RPC call */
3925         request->fl_flags |= FL_EXISTS;
3926         down_read(&nfsi->rwsem);
3927         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
3928                 up_read(&nfsi->rwsem);
3929                 goto out;
3930         }
3931         up_read(&nfsi->rwsem);
3932         if (status != 0)
3933                 goto out;
3934         /* Is this a delegated lock? */
3935         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
3936                 goto out;
3937         lsp = request->fl_u.nfs4_fl.owner;
3938         seqid = nfs_alloc_seqid(&lsp->ls_seqid);
3939         status = -ENOMEM;
3940         if (seqid == NULL)
3941                 goto out;
3942         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
3943         status = PTR_ERR(task);
3944         if (IS_ERR(task))
3945                 goto out;
3946         status = nfs4_wait_for_completion_rpc_task(task);
3947         rpc_put_task(task);
3948 out:
3949         request->fl_flags = fl_flags;
3950         return status;
3951 }
3952
3953 struct nfs4_lockdata {
3954         struct nfs_lock_args arg;
3955         struct nfs_lock_res res;
3956         struct nfs4_lock_state *lsp;
3957         struct nfs_open_context *ctx;
3958         struct file_lock fl;
3959         unsigned long timestamp;
3960         int rpc_status;
3961         int cancelled;
3962         struct nfs_server *server;
3963 };
3964
3965 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
3966                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp)
3967 {
3968         struct nfs4_lockdata *p;
3969         struct inode *inode = lsp->ls_state->inode;
3970         struct nfs_server *server = NFS_SERVER(inode);
3971
3972         p = kzalloc(sizeof(*p), GFP_KERNEL);
3973         if (p == NULL)
3974                 return NULL;
3975
3976         p->arg.fh = NFS_FH(inode);
3977         p->arg.fl = &p->fl;
3978         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid);
3979         if (p->arg.open_seqid == NULL)
3980                 goto out_free;
3981         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid);
3982         if (p->arg.lock_seqid == NULL)
3983                 goto out_free_seqid;
3984         p->arg.lock_stateid = &lsp->ls_stateid;
3985         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
3986         p->arg.lock_owner.id = lsp->ls_id.id;
3987         p->res.lock_seqid = p->arg.lock_seqid;
3988         p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3989         p->lsp = lsp;
3990         p->server = server;
3991         atomic_inc(&lsp->ls_count);
3992         p->ctx = get_nfs_open_context(ctx);
3993         memcpy(&p->fl, fl, sizeof(p->fl));
3994         return p;
3995 out_free_seqid:
3996         nfs_free_seqid(p->arg.open_seqid);
3997 out_free:
3998         kfree(p);
3999         return NULL;
4000 }
4001
4002 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4003 {
4004         struct nfs4_lockdata *data = calldata;
4005         struct nfs4_state *state = data->lsp->ls_state;
4006
4007         dprintk("%s: begin!\n", __func__);
4008         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4009                 return;
4010         /* Do we need to do an open_to_lock_owner? */
4011         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4012                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4013                         return;
4014                 data->arg.open_stateid = &state->stateid;
4015                 data->arg.new_lock_owner = 1;
4016                 data->res.open_seqid = data->arg.open_seqid;
4017         } else
4018                 data->arg.new_lock_owner = 0;
4019         data->timestamp = jiffies;
4020         if (nfs4_setup_sequence(data->server->nfs_client, &data->arg.seq_args,
4021                                 &data->res.seq_res, 1, task))
4022                 return;
4023         rpc_call_start(task);
4024         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4025 }
4026
4027 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4028 {
4029         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4030         nfs4_lock_prepare(task, calldata);
4031 }
4032
4033 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4034 {
4035         struct nfs4_lockdata *data = calldata;
4036
4037         dprintk("%s: begin!\n", __func__);
4038
4039         nfs4_sequence_done(data->server, &data->res.seq_res,
4040                         task->tk_status);
4041
4042         data->rpc_status = task->tk_status;
4043         if (RPC_ASSASSINATED(task))
4044                 goto out;
4045         if (data->arg.new_lock_owner != 0) {
4046                 if (data->rpc_status == 0)
4047                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4048                 else
4049                         goto out;
4050         }
4051         if (data->rpc_status == 0) {
4052                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4053                                         sizeof(data->lsp->ls_stateid.data));
4054                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4055                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
4056         }
4057 out:
4058         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4059 }
4060
4061 static void nfs4_lock_release(void *calldata)
4062 {
4063         struct nfs4_lockdata *data = calldata;
4064
4065         dprintk("%s: begin!\n", __func__);
4066         nfs_free_seqid(data->arg.open_seqid);
4067         if (data->cancelled != 0) {
4068                 struct rpc_task *task;
4069                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4070                                 data->arg.lock_seqid);
4071                 if (!IS_ERR(task))
4072                         rpc_put_task(task);
4073                 dprintk("%s: cancelling lock!\n", __func__);
4074         } else
4075                 nfs_free_seqid(data->arg.lock_seqid);
4076         nfs4_put_lock_state(data->lsp);
4077         put_nfs_open_context(data->ctx);
4078         kfree(data);
4079         dprintk("%s: done!\n", __func__);
4080 }
4081
4082 static const struct rpc_call_ops nfs4_lock_ops = {
4083         .rpc_call_prepare = nfs4_lock_prepare,
4084         .rpc_call_done = nfs4_lock_done,
4085         .rpc_release = nfs4_lock_release,
4086 };
4087
4088 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4089         .rpc_call_prepare = nfs4_recover_lock_prepare,
4090         .rpc_call_done = nfs4_lock_done,
4091         .rpc_release = nfs4_lock_release,
4092 };
4093
4094 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4095 {
4096         struct nfs4_lockdata *data;
4097         struct rpc_task *task;
4098         struct rpc_message msg = {
4099                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4100                 .rpc_cred = state->owner->so_cred,
4101         };
4102         struct rpc_task_setup task_setup_data = {
4103                 .rpc_client = NFS_CLIENT(state->inode),
4104                 .rpc_message = &msg,
4105                 .callback_ops = &nfs4_lock_ops,
4106                 .workqueue = nfsiod_workqueue,
4107                 .flags = RPC_TASK_ASYNC,
4108         };
4109         int ret;
4110
4111         dprintk("%s: begin!\n", __func__);
4112         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4113                         fl->fl_u.nfs4_fl.owner);
4114         if (data == NULL)
4115                 return -ENOMEM;
4116         if (IS_SETLKW(cmd))
4117                 data->arg.block = 1;
4118         if (recovery_type > NFS_LOCK_NEW) {
4119                 if (recovery_type == NFS_LOCK_RECLAIM)
4120                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4121                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4122         }
4123         msg.rpc_argp = &data->arg,
4124         msg.rpc_resp = &data->res,
4125         task_setup_data.callback_data = data;
4126         task = rpc_run_task(&task_setup_data);
4127         if (IS_ERR(task))
4128                 return PTR_ERR(task);
4129         ret = nfs4_wait_for_completion_rpc_task(task);
4130         if (ret == 0) {
4131                 ret = data->rpc_status;
4132         } else
4133                 data->cancelled = 1;
4134         rpc_put_task(task);
4135         dprintk("%s: done, ret = %d!\n", __func__, ret);
4136         return ret;
4137 }
4138
4139 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4140 {
4141         struct nfs_server *server = NFS_SERVER(state->inode);
4142         struct nfs4_exception exception = { };
4143         int err;
4144
4145         do {
4146                 /* Cache the lock if possible... */
4147                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4148                         return 0;
4149                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4150                 if (err != -NFS4ERR_DELAY)
4151                         break;
4152                 nfs4_handle_exception(server, err, &exception);
4153         } while (exception.retry);
4154         return err;
4155 }
4156
4157 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4158 {
4159         struct nfs_server *server = NFS_SERVER(state->inode);
4160         struct nfs4_exception exception = { };
4161         int err;
4162
4163         err = nfs4_set_lock_state(state, request);
4164         if (err != 0)
4165                 return err;
4166         do {
4167                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4168                         return 0;
4169                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4170                 switch (err) {
4171                 default:
4172                         goto out;
4173                 case -NFS4ERR_GRACE:
4174                 case -NFS4ERR_DELAY:
4175                         nfs4_handle_exception(server, err, &exception);
4176                         err = 0;
4177                 }
4178         } while (exception.retry);
4179 out:
4180         return err;
4181 }
4182
4183 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4184 {
4185         struct nfs_inode *nfsi = NFS_I(state->inode);
4186         unsigned char fl_flags = request->fl_flags;
4187         int status;
4188
4189         /* Is this a delegated open? */
4190         status = nfs4_set_lock_state(state, request);
4191         if (status != 0)
4192                 goto out;
4193         request->fl_flags |= FL_ACCESS;
4194         status = do_vfs_lock(request->fl_file, request);
4195         if (status < 0)
4196                 goto out;
4197         down_read(&nfsi->rwsem);
4198         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4199                 /* Yes: cache locks! */
4200                 /* ...but avoid races with delegation recall... */
4201                 request->fl_flags = fl_flags & ~FL_SLEEP;
4202                 status = do_vfs_lock(request->fl_file, request);
4203                 goto out_unlock;
4204         }
4205         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4206         if (status != 0)
4207                 goto out_unlock;
4208         /* Note: we always want to sleep here! */
4209         request->fl_flags = fl_flags | FL_SLEEP;
4210         if (do_vfs_lock(request->fl_file, request) < 0)
4211                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4212 out_unlock:
4213         up_read(&nfsi->rwsem);
4214 out:
4215         request->fl_flags = fl_flags;
4216         return status;
4217 }
4218
4219 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4220 {
4221         struct nfs4_exception exception = { };
4222         int err;
4223
4224         do {
4225                 err = _nfs4_proc_setlk(state, cmd, request);
4226                 if (err == -NFS4ERR_DENIED)
4227                         err = -EAGAIN;
4228                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4229                                 err, &exception);
4230         } while (exception.retry);
4231         return err;
4232 }
4233
4234 static int
4235 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4236 {
4237         struct nfs_open_context *ctx;
4238         struct nfs4_state *state;
4239         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4240         int status;
4241
4242         /* verify open state */
4243         ctx = nfs_file_open_context(filp);
4244         state = ctx->state;
4245
4246         if (request->fl_start < 0 || request->fl_end < 0)
4247                 return -EINVAL;
4248
4249         if (IS_GETLK(cmd)) {
4250                 if (state != NULL)
4251                         return nfs4_proc_getlk(state, F_GETLK, request);
4252                 return 0;
4253         }
4254
4255         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4256                 return -EINVAL;
4257
4258         if (request->fl_type == F_UNLCK) {
4259                 if (state != NULL)
4260                         return nfs4_proc_unlck(state, cmd, request);
4261                 return 0;
4262         }
4263
4264         if (state == NULL)
4265                 return -ENOLCK;
4266         do {
4267                 status = nfs4_proc_setlk(state, cmd, request);
4268                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4269                         break;
4270                 timeout = nfs4_set_lock_task_retry(timeout);
4271                 status = -ERESTARTSYS;
4272                 if (signalled())
4273                         break;
4274         } while(status < 0);
4275         return status;
4276 }
4277
4278 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4279 {
4280         struct nfs_server *server = NFS_SERVER(state->inode);
4281         struct nfs4_exception exception = { };
4282         int err;
4283
4284         err = nfs4_set_lock_state(state, fl);
4285         if (err != 0)
4286                 goto out;
4287         do {
4288                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4289                 switch (err) {
4290                         default:
4291                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4292                                                 __func__, err);
4293                         case 0:
4294                         case -ESTALE:
4295                                 goto out;
4296                         case -NFS4ERR_EXPIRED:
4297                         case -NFS4ERR_STALE_CLIENTID:
4298                         case -NFS4ERR_STALE_STATEID:
4299                         case -NFS4ERR_BADSESSION:
4300                         case -NFS4ERR_BADSLOT:
4301                         case -NFS4ERR_BAD_HIGH_SLOT:
4302                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4303                         case -NFS4ERR_DEADSESSION:
4304                                 nfs4_schedule_state_recovery(server->nfs_client);
4305                                 goto out;
4306                         case -ERESTARTSYS:
4307                                 /*
4308                                  * The show must go on: exit, but mark the
4309                                  * stateid as needing recovery.
4310                                  */
4311                         case -NFS4ERR_ADMIN_REVOKED:
4312                         case -NFS4ERR_BAD_STATEID:
4313                         case -NFS4ERR_OPENMODE:
4314                                 nfs4_state_mark_reclaim_nograce(server->nfs_client, state);
4315                                 err = 0;
4316                                 goto out;
4317                         case -ENOMEM:
4318                         case -NFS4ERR_DENIED:
4319                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4320                                 err = 0;
4321                                 goto out;
4322                         case -NFS4ERR_DELAY:
4323                                 break;
4324                 }
4325                 err = nfs4_handle_exception(server, err, &exception);
4326         } while (exception.retry);
4327 out:
4328         return err;
4329 }
4330
4331 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4332
4333 int nfs4_setxattr(struct dentry *dentry, const char *key, const void *buf,
4334                 size_t buflen, int flags)
4335 {
4336         struct inode *inode = dentry->d_inode;
4337
4338         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
4339                 return -EOPNOTSUPP;
4340
4341         return nfs4_proc_set_acl(inode, buf, buflen);
4342 }
4343
4344 /* The getxattr man page suggests returning -ENODATA for unknown attributes,
4345  * and that's what we'll do for e.g. user attributes that haven't been set.
4346  * But we'll follow ext2/ext3's lead by returning -EOPNOTSUPP for unsupported
4347  * attributes in kernel-managed attribute namespaces. */
4348 ssize_t nfs4_getxattr(struct dentry *dentry, const char *key, void *buf,
4349                 size_t buflen)
4350 {
4351         struct inode *inode = dentry->d_inode;
4352
4353         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
4354                 return -EOPNOTSUPP;
4355
4356         return nfs4_proc_get_acl(inode, buf, buflen);
4357 }
4358
4359 ssize_t nfs4_listxattr(struct dentry *dentry, char *buf, size_t buflen)
4360 {
4361         size_t len = strlen(XATTR_NAME_NFSV4_ACL) + 1;
4362
4363         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4364                 return 0;
4365         if (buf && buflen < len)
4366                 return -ERANGE;
4367         if (buf)
4368                 memcpy(buf, XATTR_NAME_NFSV4_ACL, len);
4369         return len;
4370 }
4371
4372 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4373 {
4374         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4375                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4376                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4377                 return;
4378
4379         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4380                 NFS_ATTR_FATTR_NLINK;
4381         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4382         fattr->nlink = 2;
4383 }
4384
4385 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4386                 struct nfs4_fs_locations *fs_locations, struct page *page)
4387 {
4388         struct nfs_server *server = NFS_SERVER(dir);
4389         u32 bitmask[2] = {
4390                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4391                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4392         };
4393         struct nfs4_fs_locations_arg args = {
4394                 .dir_fh = NFS_FH(dir),
4395                 .name = name,
4396                 .page = page,
4397                 .bitmask = bitmask,
4398         };
4399         struct nfs4_fs_locations_res res = {
4400                 .fs_locations = fs_locations,
4401         };
4402         struct rpc_message msg = {
4403                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4404                 .rpc_argp = &args,
4405                 .rpc_resp = &res,
4406         };
4407         int status;
4408
4409         dprintk("%s: start\n", __func__);
4410         nfs_fattr_init(&fs_locations->fattr);
4411         fs_locations->server = server;
4412         fs_locations->nlocations = 0;
4413         status = nfs4_call_sync(server, &msg, &args, &res, 0);
4414         nfs_fixup_referral_attributes(&fs_locations->fattr);
4415         dprintk("%s: returned status = %d\n", __func__, status);
4416         return status;
4417 }
4418
4419 #ifdef CONFIG_NFS_V4_1
4420 /*
4421  * nfs4_proc_exchange_id()
4422  *
4423  * Since the clientid has expired, all compounds using sessions
4424  * associated with the stale clientid will be returning
4425  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4426  * be in some phase of session reset.
4427  */
4428 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4429 {
4430         nfs4_verifier verifier;
4431         struct nfs41_exchange_id_args args = {
4432                 .client = clp,
4433                 .flags = clp->cl_exchange_flags,
4434         };
4435         struct nfs41_exchange_id_res res = {
4436                 .client = clp,
4437         };
4438         int status;
4439         struct rpc_message msg = {
4440                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4441                 .rpc_argp = &args,
4442                 .rpc_resp = &res,
4443                 .rpc_cred = cred,
4444         };
4445         __be32 *p;
4446
4447         dprintk("--> %s\n", __func__);
4448         BUG_ON(clp == NULL);
4449
4450         /* Remove server-only flags */
4451         args.flags &= ~EXCHGID4_FLAG_CONFIRMED_R;
4452
4453         p = (u32 *)verifier.data;
4454         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4455         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4456         args.verifier = &verifier;
4457
4458         while (1) {
4459                 args.id_len = scnprintf(args.id, sizeof(args.id),
4460                                         "%s/%s %u",
4461                                         clp->cl_ipaddr,
4462                                         rpc_peeraddr2str(clp->cl_rpcclient,
4463                                                          RPC_DISPLAY_ADDR),
4464                                         clp->cl_id_uniquifier);
4465
4466                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
4467
4468                 if (status != NFS4ERR_CLID_INUSE)
4469                         break;
4470
4471                 if (signalled())
4472                         break;
4473
4474                 if (++clp->cl_id_uniquifier == 0)
4475                         break;
4476         }
4477
4478         dprintk("<-- %s status= %d\n", __func__, status);
4479         return status;
4480 }
4481
4482 struct nfs4_get_lease_time_data {
4483         struct nfs4_get_lease_time_args *args;
4484         struct nfs4_get_lease_time_res *res;
4485         struct nfs_client *clp;
4486 };
4487
4488 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4489                                         void *calldata)
4490 {
4491         int ret;
4492         struct nfs4_get_lease_time_data *data =
4493                         (struct nfs4_get_lease_time_data *)calldata;
4494
4495         dprintk("--> %s\n", __func__);
4496         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4497         /* just setup sequence, do not trigger session recovery
4498            since we're invoked within one */
4499         ret = nfs41_setup_sequence(data->clp->cl_session,
4500                                    &data->args->la_seq_args,
4501                                    &data->res->lr_seq_res, 0, task);
4502
4503         BUG_ON(ret == -EAGAIN);
4504         rpc_call_start(task);
4505         dprintk("<-- %s\n", __func__);
4506 }
4507
4508 /*
4509  * Called from nfs4_state_manager thread for session setup, so don't recover
4510  * from sequence operation or clientid errors.
4511  */
4512 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4513 {
4514         struct nfs4_get_lease_time_data *data =
4515                         (struct nfs4_get_lease_time_data *)calldata;
4516
4517         dprintk("--> %s\n", __func__);
4518         nfs41_sequence_done(data->clp, &data->res->lr_seq_res, task->tk_status);
4519         switch (task->tk_status) {
4520         case -NFS4ERR_DELAY:
4521         case -NFS4ERR_GRACE:
4522                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4523                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4524                 task->tk_status = 0;
4525                 nfs_restart_rpc(task, data->clp);
4526                 return;
4527         }
4528         dprintk("<-- %s\n", __func__);
4529 }
4530
4531 struct rpc_call_ops nfs4_get_lease_time_ops = {
4532         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4533         .rpc_call_done = nfs4_get_lease_time_done,
4534 };
4535
4536 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4537 {
4538         struct rpc_task *task;
4539         struct nfs4_get_lease_time_args args;
4540         struct nfs4_get_lease_time_res res = {
4541                 .lr_fsinfo = fsinfo,
4542         };
4543         struct nfs4_get_lease_time_data data = {
4544                 .args = &args,
4545                 .res = &res,
4546                 .clp = clp,
4547         };
4548         struct rpc_message msg = {
4549                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4550                 .rpc_argp = &args,
4551                 .rpc_resp = &res,
4552         };
4553         struct rpc_task_setup task_setup = {
4554                 .rpc_client = clp->cl_rpcclient,
4555                 .rpc_message = &msg,
4556                 .callback_ops = &nfs4_get_lease_time_ops,
4557                 .callback_data = &data
4558         };
4559         int status;
4560
4561         res.lr_seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
4562         dprintk("--> %s\n", __func__);
4563         task = rpc_run_task(&task_setup);
4564
4565         if (IS_ERR(task))
4566                 status = PTR_ERR(task);
4567         else {
4568                 status = task->tk_status;
4569                 rpc_put_task(task);
4570         }
4571         dprintk("<-- %s return %d\n", __func__, status);
4572
4573         return status;
4574 }
4575
4576 /*
4577  * Reset a slot table
4578  */
4579 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, int max_slots,
4580                 int old_max_slots, int ivalue)
4581 {
4582         int i;
4583         int ret = 0;
4584
4585         dprintk("--> %s: max_reqs=%u, tbl %p\n", __func__, max_slots, tbl);
4586
4587         /*
4588          * Until we have dynamic slot table adjustment, insist
4589          * upon the same slot table size
4590          */
4591         if (max_slots != old_max_slots) {
4592                 dprintk("%s reset slot table does't match old\n",
4593                         __func__);
4594                 ret = -EINVAL; /*XXX NFS4ERR_REQ_TOO_BIG ? */
4595                 goto out;
4596         }
4597         spin_lock(&tbl->slot_tbl_lock);
4598         for (i = 0; i < max_slots; ++i)
4599                 tbl->slots[i].seq_nr = ivalue;
4600         spin_unlock(&tbl->slot_tbl_lock);
4601         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4602                 tbl, tbl->slots, tbl->max_slots);
4603 out:
4604         dprintk("<-- %s: return %d\n", __func__, ret);
4605         return ret;
4606 }
4607
4608 /*
4609  * Reset the forechannel and backchannel slot tables
4610  */
4611 static int nfs4_reset_slot_tables(struct nfs4_session *session)
4612 {
4613         int status;
4614
4615         status = nfs4_reset_slot_table(&session->fc_slot_table,
4616                         session->fc_attrs.max_reqs,
4617                         session->fc_slot_table.max_slots,
4618                         1);
4619         if (status)
4620                 return status;
4621
4622         status = nfs4_reset_slot_table(&session->bc_slot_table,
4623                         session->bc_attrs.max_reqs,
4624                         session->bc_slot_table.max_slots,
4625                         0);
4626         return status;
4627 }
4628
4629 /* Destroy the slot table */
4630 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
4631 {
4632         if (session->fc_slot_table.slots != NULL) {
4633                 kfree(session->fc_slot_table.slots);
4634                 session->fc_slot_table.slots = NULL;
4635         }
4636         if (session->bc_slot_table.slots != NULL) {
4637                 kfree(session->bc_slot_table.slots);
4638                 session->bc_slot_table.slots = NULL;
4639         }
4640         return;
4641 }
4642
4643 /*
4644  * Initialize slot table
4645  */
4646 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
4647                 int max_slots, int ivalue)
4648 {
4649         struct nfs4_slot *slot;
4650         int ret = -ENOMEM;
4651
4652         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4653
4654         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
4655
4656         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_KERNEL);
4657         if (!slot)
4658                 goto out;
4659         ret = 0;
4660
4661         spin_lock(&tbl->slot_tbl_lock);
4662         tbl->max_slots = max_slots;
4663         tbl->slots = slot;
4664         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4665         spin_unlock(&tbl->slot_tbl_lock);
4666         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4667                 tbl, tbl->slots, tbl->max_slots);
4668 out:
4669         dprintk("<-- %s: return %d\n", __func__, ret);
4670         return ret;
4671 }
4672
4673 /*
4674  * Initialize the forechannel and backchannel tables
4675  */
4676 static int nfs4_init_slot_tables(struct nfs4_session *session)
4677 {
4678         struct nfs4_slot_table *tbl;
4679         int status = 0;
4680
4681         tbl = &session->fc_slot_table;
4682         if (tbl->slots == NULL) {
4683                 status = nfs4_init_slot_table(tbl,
4684                                 session->fc_attrs.max_reqs, 1);
4685                 if (status)
4686                         return status;
4687         }
4688
4689         tbl = &session->bc_slot_table;
4690         if (tbl->slots == NULL) {
4691                 status = nfs4_init_slot_table(tbl,
4692                                 session->bc_attrs.max_reqs, 0);
4693                 if (status)
4694                         nfs4_destroy_slot_tables(session);
4695         }
4696
4697         return status;
4698 }
4699
4700 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
4701 {
4702         struct nfs4_session *session;
4703         struct nfs4_slot_table *tbl;
4704
4705         session = kzalloc(sizeof(struct nfs4_session), GFP_KERNEL);
4706         if (!session)
4707                 return NULL;
4708
4709         /*
4710          * The create session reply races with the server back
4711          * channel probe. Mark the client NFS_CS_SESSION_INITING
4712          * so that the client back channel can find the
4713          * nfs_client struct
4714          */
4715         clp->cl_cons_state = NFS_CS_SESSION_INITING;
4716         init_completion(&session->complete);
4717
4718         tbl = &session->fc_slot_table;
4719         tbl->highest_used_slotid = -1;
4720         spin_lock_init(&tbl->slot_tbl_lock);
4721         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
4722
4723         tbl = &session->bc_slot_table;
4724         tbl->highest_used_slotid = -1;
4725         spin_lock_init(&tbl->slot_tbl_lock);
4726         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
4727
4728         session->clp = clp;
4729         return session;
4730 }
4731
4732 void nfs4_destroy_session(struct nfs4_session *session)
4733 {
4734         nfs4_proc_destroy_session(session);
4735         dprintk("%s Destroy backchannel for xprt %p\n",
4736                 __func__, session->clp->cl_rpcclient->cl_xprt);
4737         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
4738                                 NFS41_BC_MIN_CALLBACKS);
4739         nfs4_destroy_slot_tables(session);
4740         kfree(session);
4741 }
4742
4743 /*
4744  * Initialize the values to be used by the client in CREATE_SESSION
4745  * If nfs4_init_session set the fore channel request and response sizes,
4746  * use them.
4747  *
4748  * Set the back channel max_resp_sz_cached to zero to force the client to
4749  * always set csa_cachethis to FALSE because the current implementation
4750  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
4751  */
4752 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
4753 {
4754         struct nfs4_session *session = args->client->cl_session;
4755         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
4756                      mxresp_sz = session->fc_attrs.max_resp_sz;
4757
4758         if (mxrqst_sz == 0)
4759                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
4760         if (mxresp_sz == 0)
4761                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
4762         /* Fore channel attributes */
4763         args->fc_attrs.headerpadsz = 0;
4764         args->fc_attrs.max_rqst_sz = mxrqst_sz;
4765         args->fc_attrs.max_resp_sz = mxresp_sz;
4766         args->fc_attrs.max_resp_sz_cached = mxresp_sz;
4767         args->fc_attrs.max_ops = NFS4_MAX_OPS;
4768         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
4769
4770         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
4771                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
4772                 __func__,
4773                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
4774                 args->fc_attrs.max_resp_sz_cached, args->fc_attrs.max_ops,
4775                 args->fc_attrs.max_reqs);
4776
4777         /* Back channel attributes */
4778         args->bc_attrs.headerpadsz = 0;
4779         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
4780         args->bc_attrs.max_resp_sz = PAGE_SIZE;
4781         args->bc_attrs.max_resp_sz_cached = 0;
4782         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
4783         args->bc_attrs.max_reqs = 1;
4784
4785         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
4786                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
4787                 __func__,
4788                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
4789                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
4790                 args->bc_attrs.max_reqs);
4791 }
4792
4793 static int _verify_channel_attr(char *chan, char *attr_name, u32 sent, u32 rcvd)
4794 {
4795         if (rcvd <= sent)
4796                 return 0;
4797         printk(KERN_WARNING "%s: Session INVALID: %s channel %s increased. "
4798                 "sent=%u rcvd=%u\n", __func__, chan, attr_name, sent, rcvd);
4799         return -EINVAL;
4800 }
4801
4802 #define _verify_fore_channel_attr(_name_) \
4803         _verify_channel_attr("fore", #_name_, \
4804                              args->fc_attrs._name_, \
4805                              session->fc_attrs._name_)
4806
4807 #define _verify_back_channel_attr(_name_) \
4808         _verify_channel_attr("back", #_name_, \
4809                              args->bc_attrs._name_, \
4810                              session->bc_attrs._name_)
4811
4812 /*
4813  * The server is not allowed to increase the fore channel header pad size,
4814  * maximum response size, or maximum number of operations.
4815  *
4816  * The back channel attributes are only negotiatied down: We send what the
4817  * (back channel) server insists upon.
4818  */
4819 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
4820                                      struct nfs4_session *session)
4821 {
4822         int ret = 0;
4823
4824         ret |= _verify_fore_channel_attr(headerpadsz);
4825         ret |= _verify_fore_channel_attr(max_resp_sz);
4826         ret |= _verify_fore_channel_attr(max_ops);
4827
4828         ret |= _verify_back_channel_attr(headerpadsz);
4829         ret |= _verify_back_channel_attr(max_rqst_sz);
4830         ret |= _verify_back_channel_attr(max_resp_sz);
4831         ret |= _verify_back_channel_attr(max_resp_sz_cached);
4832         ret |= _verify_back_channel_attr(max_ops);
4833         ret |= _verify_back_channel_attr(max_reqs);
4834
4835         return ret;
4836 }
4837
4838 static int _nfs4_proc_create_session(struct nfs_client *clp)
4839 {
4840         struct nfs4_session *session = clp->cl_session;
4841         struct nfs41_create_session_args args = {
4842                 .client = clp,
4843                 .cb_program = NFS4_CALLBACK,
4844         };
4845         struct nfs41_create_session_res res = {
4846                 .client = clp,
4847         };
4848         struct rpc_message msg = {
4849                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
4850                 .rpc_argp = &args,
4851                 .rpc_resp = &res,
4852         };
4853         int status;
4854
4855         nfs4_init_channel_attrs(&args);
4856         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
4857
4858         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
4859
4860         if (!status)
4861                 /* Verify the session's negotiated channel_attrs values */
4862                 status = nfs4_verify_channel_attrs(&args, session);
4863         if (!status) {
4864                 /* Increment the clientid slot sequence id */
4865                 clp->cl_seqid++;
4866         }
4867
4868         return status;
4869 }
4870
4871 /*
4872  * Issues a CREATE_SESSION operation to the server.
4873  * It is the responsibility of the caller to verify the session is
4874  * expired before calling this routine.
4875  */
4876 int nfs4_proc_create_session(struct nfs_client *clp)
4877 {
4878         int status;
4879         unsigned *ptr;
4880         struct nfs4_session *session = clp->cl_session;
4881
4882         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
4883
4884         status = _nfs4_proc_create_session(clp);
4885         if (status)
4886                 goto out;
4887
4888         /* Init and reset the fore channel */
4889         status = nfs4_init_slot_tables(session);
4890         dprintk("slot table initialization returned %d\n", status);
4891         if (status)
4892                 goto out;
4893         status = nfs4_reset_slot_tables(session);
4894         dprintk("slot table reset returned %d\n", status);
4895         if (status)
4896                 goto out;
4897
4898         ptr = (unsigned *)&session->sess_id.data[0];
4899         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
4900                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
4901 out:
4902         dprintk("<-- %s\n", __func__);
4903         return status;
4904 }
4905
4906 /*
4907  * Issue the over-the-wire RPC DESTROY_SESSION.
4908  * The caller must serialize access to this routine.
4909  */
4910 int nfs4_proc_destroy_session(struct nfs4_session *session)
4911 {
4912         int status = 0;
4913         struct rpc_message msg;
4914
4915         dprintk("--> nfs4_proc_destroy_session\n");
4916
4917         /* session is still being setup */
4918         if (session->clp->cl_cons_state != NFS_CS_READY)
4919                 return status;
4920
4921         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
4922         msg.rpc_argp = session;
4923         msg.rpc_resp = NULL;
4924         msg.rpc_cred = NULL;
4925         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
4926
4927         if (status)
4928                 printk(KERN_WARNING
4929                         "Got error %d from the server on DESTROY_SESSION. "
4930                         "Session has been destroyed regardless...\n", status);
4931
4932         dprintk("<-- nfs4_proc_destroy_session\n");
4933         return status;
4934 }
4935
4936 int nfs4_init_session(struct nfs_server *server)
4937 {
4938         struct nfs_client *clp = server->nfs_client;
4939         struct nfs4_session *session;
4940         int ret;
4941
4942         if (!nfs4_has_session(clp))
4943                 return 0;
4944
4945         session = clp->cl_session;
4946         session->fc_attrs.max_rqst_sz = server->wsize + nfs41_maxwrite_overhead;
4947         session->fc_attrs.max_resp_sz = server->rsize + nfs41_maxread_overhead;
4948
4949         ret = nfs4_recover_expired_lease(server);
4950         if (!ret)
4951                 ret = nfs4_check_client_ready(clp);
4952         return ret;
4953 }
4954
4955 /*
4956  * Renew the cl_session lease.
4957  */
4958 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
4959 {
4960         struct nfs4_sequence_args args;
4961         struct nfs4_sequence_res res;
4962
4963         struct rpc_message msg = {
4964                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
4965                 .rpc_argp = &args,
4966                 .rpc_resp = &res,
4967                 .rpc_cred = cred,
4968         };
4969
4970         args.sa_cache_this = 0;
4971
4972         return nfs4_call_sync_sequence(clp, clp->cl_rpcclient, &msg, &args,
4973                                        &res, args.sa_cache_this, 1);
4974 }
4975
4976 void nfs41_sequence_call_done(struct rpc_task *task, void *data)
4977 {
4978         struct nfs_client *clp = (struct nfs_client *)data;
4979
4980         nfs41_sequence_done(clp, task->tk_msg.rpc_resp, task->tk_status);
4981
4982         if (task->tk_status < 0) {
4983                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
4984
4985                 if (_nfs4_async_handle_error(task, NULL, clp, NULL)
4986                                                                 == -EAGAIN) {
4987                         nfs_restart_rpc(task, clp);
4988                         return;
4989                 }
4990         }
4991         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
4992
4993         kfree(task->tk_msg.rpc_argp);
4994         kfree(task->tk_msg.rpc_resp);
4995
4996         dprintk("<-- %s\n", __func__);
4997 }
4998
4999 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5000 {
5001         struct nfs_client *clp;
5002         struct nfs4_sequence_args *args;
5003         struct nfs4_sequence_res *res;
5004
5005         clp = (struct nfs_client *)data;
5006         args = task->tk_msg.rpc_argp;
5007         res = task->tk_msg.rpc_resp;
5008
5009         if (nfs4_setup_sequence(clp, args, res, 0, task))
5010                 return;
5011         rpc_call_start(task);
5012 }
5013
5014 static const struct rpc_call_ops nfs41_sequence_ops = {
5015         .rpc_call_done = nfs41_sequence_call_done,
5016         .rpc_call_prepare = nfs41_sequence_prepare,
5017 };
5018
5019 static int nfs41_proc_async_sequence(struct nfs_client *clp,
5020                                      struct rpc_cred *cred)
5021 {
5022         struct nfs4_sequence_args *args;
5023         struct nfs4_sequence_res *res;
5024         struct rpc_message msg = {
5025                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5026                 .rpc_cred = cred,
5027         };
5028
5029         args = kzalloc(sizeof(*args), GFP_KERNEL);
5030         if (!args)
5031                 return -ENOMEM;
5032         res = kzalloc(sizeof(*res), GFP_KERNEL);
5033         if (!res) {
5034                 kfree(args);
5035                 return -ENOMEM;
5036         }
5037         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
5038         msg.rpc_argp = args;
5039         msg.rpc_resp = res;
5040
5041         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
5042                               &nfs41_sequence_ops, (void *)clp);
5043 }
5044
5045 struct nfs4_reclaim_complete_data {
5046         struct nfs_client *clp;
5047         struct nfs41_reclaim_complete_args arg;
5048         struct nfs41_reclaim_complete_res res;
5049 };
5050
5051 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5052 {
5053         struct nfs4_reclaim_complete_data *calldata = data;
5054
5055         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5056         if (nfs4_setup_sequence(calldata->clp, &calldata->arg.seq_args,
5057                                 &calldata->res.seq_res, 0, task))
5058                 return;
5059
5060         rpc_call_start(task);
5061 }
5062
5063 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5064 {
5065         struct nfs4_reclaim_complete_data *calldata = data;
5066         struct nfs_client *clp = calldata->clp;
5067         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5068
5069         dprintk("--> %s\n", __func__);
5070         nfs41_sequence_done(clp, res, task->tk_status);
5071         switch (task->tk_status) {
5072         case 0:
5073         case -NFS4ERR_COMPLETE_ALREADY:
5074                 break;
5075         case -NFS4ERR_BADSESSION:
5076         case -NFS4ERR_DEADSESSION:
5077                 /*
5078                  * Handle the session error, but do not retry the operation, as
5079                  * we have no way of telling whether the clientid had to be
5080                  * reset before we got our reply.  If reset, a new wave of
5081                  * reclaim operations will follow, containing their own reclaim
5082                  * complete.  We don't want our retry to get on the way of
5083                  * recovery by incorrectly indicating to the server that we're
5084                  * done reclaiming state since the process had to be restarted.
5085                  */
5086                 _nfs4_async_handle_error(task, NULL, clp, NULL);
5087                 break;
5088         default:
5089                 if (_nfs4_async_handle_error(
5090                                 task, NULL, clp, NULL) == -EAGAIN) {
5091                         rpc_restart_call_prepare(task);
5092                         return;
5093                 }
5094         }
5095
5096         dprintk("<-- %s\n", __func__);
5097 }
5098
5099 static void nfs4_free_reclaim_complete_data(void *data)
5100 {
5101         struct nfs4_reclaim_complete_data *calldata = data;
5102
5103         kfree(calldata);
5104 }
5105
5106 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5107         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5108         .rpc_call_done = nfs4_reclaim_complete_done,
5109         .rpc_release = nfs4_free_reclaim_complete_data,
5110 };
5111
5112 /*
5113  * Issue a global reclaim complete.
5114  */
5115 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5116 {
5117         struct nfs4_reclaim_complete_data *calldata;
5118         struct rpc_task *task;
5119         struct rpc_message msg = {
5120                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5121         };
5122         struct rpc_task_setup task_setup_data = {
5123                 .rpc_client = clp->cl_rpcclient,
5124                 .rpc_message = &msg,
5125                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5126                 .flags = RPC_TASK_ASYNC,
5127         };
5128         int status = -ENOMEM;
5129
5130         dprintk("--> %s\n", __func__);
5131         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
5132         if (calldata == NULL)
5133                 goto out;
5134         calldata->clp = clp;
5135         calldata->arg.one_fs = 0;
5136         calldata->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
5137
5138         msg.rpc_argp = &calldata->arg;
5139         msg.rpc_resp = &calldata->res;
5140         task_setup_data.callback_data = calldata;
5141         task = rpc_run_task(&task_setup_data);
5142         if (IS_ERR(task))
5143                 status = PTR_ERR(task);
5144         rpc_put_task(task);
5145 out:
5146         dprintk("<-- %s status=%d\n", __func__, status);
5147         return status;
5148 }
5149 #endif /* CONFIG_NFS_V4_1 */
5150
5151 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
5152         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5153         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5154         .recover_open   = nfs4_open_reclaim,
5155         .recover_lock   = nfs4_lock_reclaim,
5156         .establish_clid = nfs4_init_clientid,
5157         .get_clid_cred  = nfs4_get_setclientid_cred,
5158 };
5159
5160 #if defined(CONFIG_NFS_V4_1)
5161 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
5162         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5163         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5164         .recover_open   = nfs4_open_reclaim,
5165         .recover_lock   = nfs4_lock_reclaim,
5166         .establish_clid = nfs41_init_clientid,
5167         .get_clid_cred  = nfs4_get_exchange_id_cred,
5168         .reclaim_complete = nfs41_proc_reclaim_complete,
5169 };
5170 #endif /* CONFIG_NFS_V4_1 */
5171
5172 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
5173         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5174         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5175         .recover_open   = nfs4_open_expired,
5176         .recover_lock   = nfs4_lock_expired,
5177         .establish_clid = nfs4_init_clientid,
5178         .get_clid_cred  = nfs4_get_setclientid_cred,
5179 };
5180
5181 #if defined(CONFIG_NFS_V4_1)
5182 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
5183         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5184         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5185         .recover_open   = nfs4_open_expired,
5186         .recover_lock   = nfs4_lock_expired,
5187         .establish_clid = nfs41_init_clientid,
5188         .get_clid_cred  = nfs4_get_exchange_id_cred,
5189 };
5190 #endif /* CONFIG_NFS_V4_1 */
5191
5192 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
5193         .sched_state_renewal = nfs4_proc_async_renew,
5194         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
5195         .renew_lease = nfs4_proc_renew,
5196 };
5197
5198 #if defined(CONFIG_NFS_V4_1)
5199 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
5200         .sched_state_renewal = nfs41_proc_async_sequence,
5201         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
5202         .renew_lease = nfs4_proc_sequence,
5203 };
5204 #endif
5205
5206 /*
5207  * Per minor version reboot and network partition recovery ops
5208  */
5209
5210 struct nfs4_state_recovery_ops *nfs4_reboot_recovery_ops[] = {
5211         &nfs40_reboot_recovery_ops,
5212 #if defined(CONFIG_NFS_V4_1)
5213         &nfs41_reboot_recovery_ops,
5214 #endif
5215 };
5216
5217 struct nfs4_state_recovery_ops *nfs4_nograce_recovery_ops[] = {
5218         &nfs40_nograce_recovery_ops,
5219 #if defined(CONFIG_NFS_V4_1)
5220         &nfs41_nograce_recovery_ops,
5221 #endif
5222 };
5223
5224 struct nfs4_state_maintenance_ops *nfs4_state_renewal_ops[] = {
5225         &nfs40_state_renewal_ops,
5226 #if defined(CONFIG_NFS_V4_1)
5227         &nfs41_state_renewal_ops,
5228 #endif
5229 };
5230
5231 static const struct inode_operations nfs4_file_inode_operations = {
5232         .permission     = nfs_permission,
5233         .getattr        = nfs_getattr,
5234         .setattr        = nfs_setattr,
5235         .getxattr       = nfs4_getxattr,
5236         .setxattr       = nfs4_setxattr,
5237         .listxattr      = nfs4_listxattr,
5238 };
5239
5240 const struct nfs_rpc_ops nfs_v4_clientops = {
5241         .version        = 4,                    /* protocol version */
5242         .dentry_ops     = &nfs4_dentry_operations,
5243         .dir_inode_ops  = &nfs4_dir_inode_operations,
5244         .file_inode_ops = &nfs4_file_inode_operations,
5245         .getroot        = nfs4_proc_get_root,
5246         .getattr        = nfs4_proc_getattr,
5247         .setattr        = nfs4_proc_setattr,
5248         .lookupfh       = nfs4_proc_lookupfh,
5249         .lookup         = nfs4_proc_lookup,
5250         .access         = nfs4_proc_access,
5251         .readlink       = nfs4_proc_readlink,
5252         .create         = nfs4_proc_create,
5253         .remove         = nfs4_proc_remove,
5254         .unlink_setup   = nfs4_proc_unlink_setup,
5255         .unlink_done    = nfs4_proc_unlink_done,
5256         .rename         = nfs4_proc_rename,
5257         .link           = nfs4_proc_link,
5258         .symlink        = nfs4_proc_symlink,
5259         .mkdir          = nfs4_proc_mkdir,
5260         .rmdir          = nfs4_proc_remove,
5261         .readdir        = nfs4_proc_readdir,
5262         .mknod          = nfs4_proc_mknod,
5263         .statfs         = nfs4_proc_statfs,
5264         .fsinfo         = nfs4_proc_fsinfo,
5265         .pathconf       = nfs4_proc_pathconf,
5266         .set_capabilities = nfs4_server_capabilities,
5267         .decode_dirent  = nfs4_decode_dirent,
5268         .read_setup     = nfs4_proc_read_setup,
5269         .read_done      = nfs4_read_done,
5270         .write_setup    = nfs4_proc_write_setup,
5271         .write_done     = nfs4_write_done,
5272         .commit_setup   = nfs4_proc_commit_setup,
5273         .commit_done    = nfs4_commit_done,
5274         .lock           = nfs4_proc_lock,
5275         .clear_acl_cache = nfs4_zap_acl_attr,
5276         .close_context  = nfs4_close_context,
5277 };
5278
5279 /*
5280  * Local variables:
5281  *  c-basic-offset: 8
5282  * End:
5283  */