]> git.openfabrics.org - ~shefty/rdma-dev.git/blob - fs/nfs/nfs4proc.c
Merge branch 'nfs-for-3.3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
[~shefty/rdma-dev.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/sunrpc/gss_api.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/nfs_idmap.h>
56 #include <linux/sunrpc/bc_xprt.h>
57 #include <linux/xattr.h>
58 #include <linux/utsname.h>
59 #include <linux/freezer.h>
60
61 #include "nfs4_fs.h"
62 #include "delegation.h"
63 #include "internal.h"
64 #include "iostat.h"
65 #include "callback.h"
66 #include "pnfs.h"
67
68 #define NFSDBG_FACILITY         NFSDBG_PROC
69
70 #define NFS4_POLL_RETRY_MIN     (HZ/10)
71 #define NFS4_POLL_RETRY_MAX     (15*HZ)
72
73 #define NFS4_MAX_LOOP_ON_RECOVER (10)
74
75 struct nfs4_opendata;
76 static int _nfs4_proc_open(struct nfs4_opendata *data);
77 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
78 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
79 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
80 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
81 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
82                             struct nfs_fattr *fattr, struct iattr *sattr,
83                             struct nfs4_state *state);
84 #ifdef CONFIG_NFS_V4_1
85 static int nfs41_test_stateid(struct nfs_server *, struct nfs4_state *);
86 static int nfs41_free_stateid(struct nfs_server *, struct nfs4_state *);
87 #endif
88 /* Prevent leaks of NFSv4 errors into userland */
89 static int nfs4_map_errors(int err)
90 {
91         if (err >= -1000)
92                 return err;
93         switch (err) {
94         case -NFS4ERR_RESOURCE:
95                 return -EREMOTEIO;
96         case -NFS4ERR_WRONGSEC:
97                 return -EPERM;
98         case -NFS4ERR_BADOWNER:
99         case -NFS4ERR_BADNAME:
100                 return -EINVAL;
101         default:
102                 dprintk("%s could not handle NFSv4 error %d\n",
103                                 __func__, -err);
104                 break;
105         }
106         return -EIO;
107 }
108
109 /*
110  * This is our standard bitmap for GETATTR requests.
111  */
112 const u32 nfs4_fattr_bitmap[2] = {
113         FATTR4_WORD0_TYPE
114         | FATTR4_WORD0_CHANGE
115         | FATTR4_WORD0_SIZE
116         | FATTR4_WORD0_FSID
117         | FATTR4_WORD0_FILEID,
118         FATTR4_WORD1_MODE
119         | FATTR4_WORD1_NUMLINKS
120         | FATTR4_WORD1_OWNER
121         | FATTR4_WORD1_OWNER_GROUP
122         | FATTR4_WORD1_RAWDEV
123         | FATTR4_WORD1_SPACE_USED
124         | FATTR4_WORD1_TIME_ACCESS
125         | FATTR4_WORD1_TIME_METADATA
126         | FATTR4_WORD1_TIME_MODIFY
127 };
128
129 const u32 nfs4_statfs_bitmap[2] = {
130         FATTR4_WORD0_FILES_AVAIL
131         | FATTR4_WORD0_FILES_FREE
132         | FATTR4_WORD0_FILES_TOTAL,
133         FATTR4_WORD1_SPACE_AVAIL
134         | FATTR4_WORD1_SPACE_FREE
135         | FATTR4_WORD1_SPACE_TOTAL
136 };
137
138 const u32 nfs4_pathconf_bitmap[2] = {
139         FATTR4_WORD0_MAXLINK
140         | FATTR4_WORD0_MAXNAME,
141         0
142 };
143
144 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
145                         | FATTR4_WORD0_MAXREAD
146                         | FATTR4_WORD0_MAXWRITE
147                         | FATTR4_WORD0_LEASE_TIME,
148                         FATTR4_WORD1_TIME_DELTA
149                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
150                         FATTR4_WORD2_LAYOUT_BLKSIZE
151 };
152
153 const u32 nfs4_fs_locations_bitmap[2] = {
154         FATTR4_WORD0_TYPE
155         | FATTR4_WORD0_CHANGE
156         | FATTR4_WORD0_SIZE
157         | FATTR4_WORD0_FSID
158         | FATTR4_WORD0_FILEID
159         | FATTR4_WORD0_FS_LOCATIONS,
160         FATTR4_WORD1_MODE
161         | FATTR4_WORD1_NUMLINKS
162         | FATTR4_WORD1_OWNER
163         | FATTR4_WORD1_OWNER_GROUP
164         | FATTR4_WORD1_RAWDEV
165         | FATTR4_WORD1_SPACE_USED
166         | FATTR4_WORD1_TIME_ACCESS
167         | FATTR4_WORD1_TIME_METADATA
168         | FATTR4_WORD1_TIME_MODIFY
169         | FATTR4_WORD1_MOUNTED_ON_FILEID
170 };
171
172 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
173                 struct nfs4_readdir_arg *readdir)
174 {
175         __be32 *start, *p;
176
177         BUG_ON(readdir->count < 80);
178         if (cookie > 2) {
179                 readdir->cookie = cookie;
180                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
181                 return;
182         }
183
184         readdir->cookie = 0;
185         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
186         if (cookie == 2)
187                 return;
188         
189         /*
190          * NFSv4 servers do not return entries for '.' and '..'
191          * Therefore, we fake these entries here.  We let '.'
192          * have cookie 0 and '..' have cookie 1.  Note that
193          * when talking to the server, we always send cookie 0
194          * instead of 1 or 2.
195          */
196         start = p = kmap_atomic(*readdir->pages, KM_USER0);
197         
198         if (cookie == 0) {
199                 *p++ = xdr_one;                                  /* next */
200                 *p++ = xdr_zero;                   /* cookie, first word */
201                 *p++ = xdr_one;                   /* cookie, second word */
202                 *p++ = xdr_one;                             /* entry len */
203                 memcpy(p, ".\0\0\0", 4);                        /* entry */
204                 p++;
205                 *p++ = xdr_one;                         /* bitmap length */
206                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
207                 *p++ = htonl(8);              /* attribute buffer length */
208                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
209         }
210         
211         *p++ = xdr_one;                                  /* next */
212         *p++ = xdr_zero;                   /* cookie, first word */
213         *p++ = xdr_two;                   /* cookie, second word */
214         *p++ = xdr_two;                             /* entry len */
215         memcpy(p, "..\0\0", 4);                         /* entry */
216         p++;
217         *p++ = xdr_one;                         /* bitmap length */
218         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
219         *p++ = htonl(8);              /* attribute buffer length */
220         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
221
222         readdir->pgbase = (char *)p - (char *)start;
223         readdir->count -= readdir->pgbase;
224         kunmap_atomic(start, KM_USER0);
225 }
226
227 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
228 {
229         int res;
230
231         might_sleep();
232
233         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
234                         nfs_wait_bit_killable, TASK_KILLABLE);
235         return res;
236 }
237
238 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
239 {
240         int res = 0;
241
242         might_sleep();
243
244         if (*timeout <= 0)
245                 *timeout = NFS4_POLL_RETRY_MIN;
246         if (*timeout > NFS4_POLL_RETRY_MAX)
247                 *timeout = NFS4_POLL_RETRY_MAX;
248         freezable_schedule_timeout_killable(*timeout);
249         if (fatal_signal_pending(current))
250                 res = -ERESTARTSYS;
251         *timeout <<= 1;
252         return res;
253 }
254
255 /* This is the error handling routine for processes that are allowed
256  * to sleep.
257  */
258 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
259 {
260         struct nfs_client *clp = server->nfs_client;
261         struct nfs4_state *state = exception->state;
262         int ret = errorcode;
263
264         exception->retry = 0;
265         switch(errorcode) {
266                 case 0:
267                         return 0;
268                 case -NFS4ERR_ADMIN_REVOKED:
269                 case -NFS4ERR_BAD_STATEID:
270                 case -NFS4ERR_OPENMODE:
271                         if (state == NULL)
272                                 break;
273                         nfs4_schedule_stateid_recovery(server, state);
274                         goto wait_on_recovery;
275                 case -NFS4ERR_EXPIRED:
276                         if (state != NULL)
277                                 nfs4_schedule_stateid_recovery(server, state);
278                 case -NFS4ERR_STALE_STATEID:
279                 case -NFS4ERR_STALE_CLIENTID:
280                         nfs4_schedule_lease_recovery(clp);
281                         goto wait_on_recovery;
282 #if defined(CONFIG_NFS_V4_1)
283                 case -NFS4ERR_BADSESSION:
284                 case -NFS4ERR_BADSLOT:
285                 case -NFS4ERR_BAD_HIGH_SLOT:
286                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
287                 case -NFS4ERR_DEADSESSION:
288                 case -NFS4ERR_SEQ_FALSE_RETRY:
289                 case -NFS4ERR_SEQ_MISORDERED:
290                         dprintk("%s ERROR: %d Reset session\n", __func__,
291                                 errorcode);
292                         nfs4_schedule_session_recovery(clp->cl_session);
293                         exception->retry = 1;
294                         break;
295 #endif /* defined(CONFIG_NFS_V4_1) */
296                 case -NFS4ERR_FILE_OPEN:
297                         if (exception->timeout > HZ) {
298                                 /* We have retried a decent amount, time to
299                                  * fail
300                                  */
301                                 ret = -EBUSY;
302                                 break;
303                         }
304                 case -NFS4ERR_GRACE:
305                 case -NFS4ERR_DELAY:
306                 case -EKEYEXPIRED:
307                         ret = nfs4_delay(server->client, &exception->timeout);
308                         if (ret != 0)
309                                 break;
310                 case -NFS4ERR_RETRY_UNCACHED_REP:
311                 case -NFS4ERR_OLD_STATEID:
312                         exception->retry = 1;
313                         break;
314                 case -NFS4ERR_BADOWNER:
315                         /* The following works around a Linux server bug! */
316                 case -NFS4ERR_BADNAME:
317                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
318                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
319                                 exception->retry = 1;
320                                 printk(KERN_WARNING "NFS: v4 server %s "
321                                                 "does not accept raw "
322                                                 "uid/gids. "
323                                                 "Reenabling the idmapper.\n",
324                                                 server->nfs_client->cl_hostname);
325                         }
326         }
327         /* We failed to handle the error */
328         return nfs4_map_errors(ret);
329 wait_on_recovery:
330         ret = nfs4_wait_clnt_recover(clp);
331         if (ret == 0)
332                 exception->retry = 1;
333         return ret;
334 }
335
336
337 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
338 {
339         spin_lock(&clp->cl_lock);
340         if (time_before(clp->cl_last_renewal,timestamp))
341                 clp->cl_last_renewal = timestamp;
342         spin_unlock(&clp->cl_lock);
343 }
344
345 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
346 {
347         do_renew_lease(server->nfs_client, timestamp);
348 }
349
350 #if defined(CONFIG_NFS_V4_1)
351
352 /*
353  * nfs4_free_slot - free a slot and efficiently update slot table.
354  *
355  * freeing a slot is trivially done by clearing its respective bit
356  * in the bitmap.
357  * If the freed slotid equals highest_used_slotid we want to update it
358  * so that the server would be able to size down the slot table if needed,
359  * otherwise we know that the highest_used_slotid is still in use.
360  * When updating highest_used_slotid there may be "holes" in the bitmap
361  * so we need to scan down from highest_used_slotid to 0 looking for the now
362  * highest slotid in use.
363  * If none found, highest_used_slotid is set to -1.
364  *
365  * Must be called while holding tbl->slot_tbl_lock
366  */
367 static void
368 nfs4_free_slot(struct nfs4_slot_table *tbl, u8 free_slotid)
369 {
370         int slotid = free_slotid;
371
372         BUG_ON(slotid < 0 || slotid >= NFS4_MAX_SLOT_TABLE);
373         /* clear used bit in bitmap */
374         __clear_bit(slotid, tbl->used_slots);
375
376         /* update highest_used_slotid when it is freed */
377         if (slotid == tbl->highest_used_slotid) {
378                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
379                 if (slotid < tbl->max_slots)
380                         tbl->highest_used_slotid = slotid;
381                 else
382                         tbl->highest_used_slotid = -1;
383         }
384         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
385                 free_slotid, tbl->highest_used_slotid);
386 }
387
388 /*
389  * Signal state manager thread if session fore channel is drained
390  */
391 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
392 {
393         struct rpc_task *task;
394
395         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
396                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
397                 if (task)
398                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
399                 return;
400         }
401
402         if (ses->fc_slot_table.highest_used_slotid != -1)
403                 return;
404
405         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
406         complete(&ses->fc_slot_table.complete);
407 }
408
409 /*
410  * Signal state manager thread if session back channel is drained
411  */
412 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
413 {
414         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
415             ses->bc_slot_table.highest_used_slotid != -1)
416                 return;
417         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
418         complete(&ses->bc_slot_table.complete);
419 }
420
421 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
422 {
423         struct nfs4_slot_table *tbl;
424
425         tbl = &res->sr_session->fc_slot_table;
426         if (!res->sr_slot) {
427                 /* just wake up the next guy waiting since
428                  * we may have not consumed a slot after all */
429                 dprintk("%s: No slot\n", __func__);
430                 return;
431         }
432
433         spin_lock(&tbl->slot_tbl_lock);
434         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
435         nfs4_check_drain_fc_complete(res->sr_session);
436         spin_unlock(&tbl->slot_tbl_lock);
437         res->sr_slot = NULL;
438 }
439
440 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
441 {
442         unsigned long timestamp;
443         struct nfs_client *clp;
444
445         /*
446          * sr_status remains 1 if an RPC level error occurred. The server
447          * may or may not have processed the sequence operation..
448          * Proceed as if the server received and processed the sequence
449          * operation.
450          */
451         if (res->sr_status == 1)
452                 res->sr_status = NFS_OK;
453
454         /* don't increment the sequence number if the task wasn't sent */
455         if (!RPC_WAS_SENT(task))
456                 goto out;
457
458         /* Check the SEQUENCE operation status */
459         switch (res->sr_status) {
460         case 0:
461                 /* Update the slot's sequence and clientid lease timer */
462                 ++res->sr_slot->seq_nr;
463                 timestamp = res->sr_renewal_time;
464                 clp = res->sr_session->clp;
465                 do_renew_lease(clp, timestamp);
466                 /* Check sequence flags */
467                 if (res->sr_status_flags != 0)
468                         nfs4_schedule_lease_recovery(clp);
469                 break;
470         case -NFS4ERR_DELAY:
471                 /* The server detected a resend of the RPC call and
472                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
473                  * of RFC5661.
474                  */
475                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
476                         __func__,
477                         res->sr_slot - res->sr_session->fc_slot_table.slots,
478                         res->sr_slot->seq_nr);
479                 goto out_retry;
480         default:
481                 /* Just update the slot sequence no. */
482                 ++res->sr_slot->seq_nr;
483         }
484 out:
485         /* The session may be reset by one of the error handlers. */
486         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
487         nfs41_sequence_free_slot(res);
488         return 1;
489 out_retry:
490         if (!rpc_restart_call(task))
491                 goto out;
492         rpc_delay(task, NFS4_POLL_RETRY_MAX);
493         return 0;
494 }
495
496 static int nfs4_sequence_done(struct rpc_task *task,
497                                struct nfs4_sequence_res *res)
498 {
499         if (res->sr_session == NULL)
500                 return 1;
501         return nfs41_sequence_done(task, res);
502 }
503
504 /*
505  * nfs4_find_slot - efficiently look for a free slot
506  *
507  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
508  * If found, we mark the slot as used, update the highest_used_slotid,
509  * and respectively set up the sequence operation args.
510  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
511  *
512  * Note: must be called with under the slot_tbl_lock.
513  */
514 static u8
515 nfs4_find_slot(struct nfs4_slot_table *tbl)
516 {
517         int slotid;
518         u8 ret_id = NFS4_MAX_SLOT_TABLE;
519         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
520
521         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
522                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
523                 tbl->max_slots);
524         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
525         if (slotid >= tbl->max_slots)
526                 goto out;
527         __set_bit(slotid, tbl->used_slots);
528         if (slotid > tbl->highest_used_slotid)
529                 tbl->highest_used_slotid = slotid;
530         ret_id = slotid;
531 out:
532         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
533                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
534         return ret_id;
535 }
536
537 int nfs41_setup_sequence(struct nfs4_session *session,
538                                 struct nfs4_sequence_args *args,
539                                 struct nfs4_sequence_res *res,
540                                 int cache_reply,
541                                 struct rpc_task *task)
542 {
543         struct nfs4_slot *slot;
544         struct nfs4_slot_table *tbl;
545         u8 slotid;
546
547         dprintk("--> %s\n", __func__);
548         /* slot already allocated? */
549         if (res->sr_slot != NULL)
550                 return 0;
551
552         tbl = &session->fc_slot_table;
553
554         spin_lock(&tbl->slot_tbl_lock);
555         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
556             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
557                 /* The state manager will wait until the slot table is empty */
558                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
559                 spin_unlock(&tbl->slot_tbl_lock);
560                 dprintk("%s session is draining\n", __func__);
561                 return -EAGAIN;
562         }
563
564         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
565             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
566                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
567                 spin_unlock(&tbl->slot_tbl_lock);
568                 dprintk("%s enforce FIFO order\n", __func__);
569                 return -EAGAIN;
570         }
571
572         slotid = nfs4_find_slot(tbl);
573         if (slotid == NFS4_MAX_SLOT_TABLE) {
574                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
575                 spin_unlock(&tbl->slot_tbl_lock);
576                 dprintk("<-- %s: no free slots\n", __func__);
577                 return -EAGAIN;
578         }
579         spin_unlock(&tbl->slot_tbl_lock);
580
581         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
582         slot = tbl->slots + slotid;
583         args->sa_session = session;
584         args->sa_slotid = slotid;
585         args->sa_cache_this = cache_reply;
586
587         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
588
589         res->sr_session = session;
590         res->sr_slot = slot;
591         res->sr_renewal_time = jiffies;
592         res->sr_status_flags = 0;
593         /*
594          * sr_status is only set in decode_sequence, and so will remain
595          * set to 1 if an rpc level failure occurs.
596          */
597         res->sr_status = 1;
598         return 0;
599 }
600 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
601
602 int nfs4_setup_sequence(const struct nfs_server *server,
603                         struct nfs4_sequence_args *args,
604                         struct nfs4_sequence_res *res,
605                         int cache_reply,
606                         struct rpc_task *task)
607 {
608         struct nfs4_session *session = nfs4_get_session(server);
609         int ret = 0;
610
611         if (session == NULL) {
612                 args->sa_session = NULL;
613                 res->sr_session = NULL;
614                 goto out;
615         }
616
617         dprintk("--> %s clp %p session %p sr_slot %td\n",
618                 __func__, session->clp, session, res->sr_slot ?
619                         res->sr_slot - session->fc_slot_table.slots : -1);
620
621         ret = nfs41_setup_sequence(session, args, res, cache_reply,
622                                    task);
623 out:
624         dprintk("<-- %s status=%d\n", __func__, ret);
625         return ret;
626 }
627
628 struct nfs41_call_sync_data {
629         const struct nfs_server *seq_server;
630         struct nfs4_sequence_args *seq_args;
631         struct nfs4_sequence_res *seq_res;
632         int cache_reply;
633 };
634
635 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
636 {
637         struct nfs41_call_sync_data *data = calldata;
638
639         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
640
641         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
642                                 data->seq_res, data->cache_reply, task))
643                 return;
644         rpc_call_start(task);
645 }
646
647 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
648 {
649         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
650         nfs41_call_sync_prepare(task, calldata);
651 }
652
653 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
654 {
655         struct nfs41_call_sync_data *data = calldata;
656
657         nfs41_sequence_done(task, data->seq_res);
658 }
659
660 struct rpc_call_ops nfs41_call_sync_ops = {
661         .rpc_call_prepare = nfs41_call_sync_prepare,
662         .rpc_call_done = nfs41_call_sync_done,
663 };
664
665 struct rpc_call_ops nfs41_call_priv_sync_ops = {
666         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
667         .rpc_call_done = nfs41_call_sync_done,
668 };
669
670 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
671                                    struct nfs_server *server,
672                                    struct rpc_message *msg,
673                                    struct nfs4_sequence_args *args,
674                                    struct nfs4_sequence_res *res,
675                                    int cache_reply,
676                                    int privileged)
677 {
678         int ret;
679         struct rpc_task *task;
680         struct nfs41_call_sync_data data = {
681                 .seq_server = server,
682                 .seq_args = args,
683                 .seq_res = res,
684                 .cache_reply = cache_reply,
685         };
686         struct rpc_task_setup task_setup = {
687                 .rpc_client = clnt,
688                 .rpc_message = msg,
689                 .callback_ops = &nfs41_call_sync_ops,
690                 .callback_data = &data
691         };
692
693         res->sr_slot = NULL;
694         if (privileged)
695                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
696         task = rpc_run_task(&task_setup);
697         if (IS_ERR(task))
698                 ret = PTR_ERR(task);
699         else {
700                 ret = task->tk_status;
701                 rpc_put_task(task);
702         }
703         return ret;
704 }
705
706 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
707                             struct nfs_server *server,
708                             struct rpc_message *msg,
709                             struct nfs4_sequence_args *args,
710                             struct nfs4_sequence_res *res,
711                             int cache_reply)
712 {
713         return nfs4_call_sync_sequence(clnt, server, msg, args, res, cache_reply, 0);
714 }
715
716 #else
717 static int nfs4_sequence_done(struct rpc_task *task,
718                                struct nfs4_sequence_res *res)
719 {
720         return 1;
721 }
722 #endif /* CONFIG_NFS_V4_1 */
723
724 int _nfs4_call_sync(struct rpc_clnt *clnt,
725                     struct nfs_server *server,
726                     struct rpc_message *msg,
727                     struct nfs4_sequence_args *args,
728                     struct nfs4_sequence_res *res,
729                     int cache_reply)
730 {
731         args->sa_session = res->sr_session = NULL;
732         return rpc_call_sync(clnt, msg, 0);
733 }
734
735 static inline
736 int nfs4_call_sync(struct rpc_clnt *clnt,
737                    struct nfs_server *server,
738                    struct rpc_message *msg,
739                    struct nfs4_sequence_args *args,
740                    struct nfs4_sequence_res *res,
741                    int cache_reply)
742 {
743         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
744                                                 args, res, cache_reply);
745 }
746
747 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
748 {
749         struct nfs_inode *nfsi = NFS_I(dir);
750
751         spin_lock(&dir->i_lock);
752         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
753         if (!cinfo->atomic || cinfo->before != dir->i_version)
754                 nfs_force_lookup_revalidate(dir);
755         dir->i_version = cinfo->after;
756         spin_unlock(&dir->i_lock);
757 }
758
759 struct nfs4_opendata {
760         struct kref kref;
761         struct nfs_openargs o_arg;
762         struct nfs_openres o_res;
763         struct nfs_open_confirmargs c_arg;
764         struct nfs_open_confirmres c_res;
765         struct nfs4_string owner_name;
766         struct nfs4_string group_name;
767         struct nfs_fattr f_attr;
768         struct nfs_fattr dir_attr;
769         struct dentry *dir;
770         struct dentry *dentry;
771         struct nfs4_state_owner *owner;
772         struct nfs4_state *state;
773         struct iattr attrs;
774         unsigned long timestamp;
775         unsigned int rpc_done : 1;
776         int rpc_status;
777         int cancelled;
778 };
779
780
781 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
782 {
783         p->o_res.f_attr = &p->f_attr;
784         p->o_res.dir_attr = &p->dir_attr;
785         p->o_res.seqid = p->o_arg.seqid;
786         p->c_res.seqid = p->c_arg.seqid;
787         p->o_res.server = p->o_arg.server;
788         nfs_fattr_init(&p->f_attr);
789         nfs_fattr_init(&p->dir_attr);
790         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
791 }
792
793 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
794                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
795                 const struct iattr *attrs,
796                 gfp_t gfp_mask)
797 {
798         struct dentry *parent = dget_parent(dentry);
799         struct inode *dir = parent->d_inode;
800         struct nfs_server *server = NFS_SERVER(dir);
801         struct nfs4_opendata *p;
802
803         p = kzalloc(sizeof(*p), gfp_mask);
804         if (p == NULL)
805                 goto err;
806         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
807         if (p->o_arg.seqid == NULL)
808                 goto err_free;
809         nfs_sb_active(dentry->d_sb);
810         p->dentry = dget(dentry);
811         p->dir = parent;
812         p->owner = sp;
813         atomic_inc(&sp->so_count);
814         p->o_arg.fh = NFS_FH(dir);
815         p->o_arg.open_flags = flags;
816         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
817         p->o_arg.clientid = server->nfs_client->cl_clientid;
818         p->o_arg.id = sp->so_owner_id.id;
819         p->o_arg.name = &dentry->d_name;
820         p->o_arg.server = server;
821         p->o_arg.bitmask = server->attr_bitmask;
822         p->o_arg.dir_bitmask = server->cache_consistency_bitmask;
823         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
824         if (flags & O_CREAT) {
825                 u32 *s;
826
827                 p->o_arg.u.attrs = &p->attrs;
828                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
829                 s = (u32 *) p->o_arg.u.verifier.data;
830                 s[0] = jiffies;
831                 s[1] = current->pid;
832         }
833         p->c_arg.fh = &p->o_res.fh;
834         p->c_arg.stateid = &p->o_res.stateid;
835         p->c_arg.seqid = p->o_arg.seqid;
836         nfs4_init_opendata_res(p);
837         kref_init(&p->kref);
838         return p;
839 err_free:
840         kfree(p);
841 err:
842         dput(parent);
843         return NULL;
844 }
845
846 static void nfs4_opendata_free(struct kref *kref)
847 {
848         struct nfs4_opendata *p = container_of(kref,
849                         struct nfs4_opendata, kref);
850         struct super_block *sb = p->dentry->d_sb;
851
852         nfs_free_seqid(p->o_arg.seqid);
853         if (p->state != NULL)
854                 nfs4_put_open_state(p->state);
855         nfs4_put_state_owner(p->owner);
856         dput(p->dir);
857         dput(p->dentry);
858         nfs_sb_deactive(sb);
859         nfs_fattr_free_names(&p->f_attr);
860         kfree(p);
861 }
862
863 static void nfs4_opendata_put(struct nfs4_opendata *p)
864 {
865         if (p != NULL)
866                 kref_put(&p->kref, nfs4_opendata_free);
867 }
868
869 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
870 {
871         int ret;
872
873         ret = rpc_wait_for_completion_task(task);
874         return ret;
875 }
876
877 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
878 {
879         int ret = 0;
880
881         if (open_mode & O_EXCL)
882                 goto out;
883         switch (mode & (FMODE_READ|FMODE_WRITE)) {
884                 case FMODE_READ:
885                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
886                                 && state->n_rdonly != 0;
887                         break;
888                 case FMODE_WRITE:
889                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
890                                 && state->n_wronly != 0;
891                         break;
892                 case FMODE_READ|FMODE_WRITE:
893                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
894                                 && state->n_rdwr != 0;
895         }
896 out:
897         return ret;
898 }
899
900 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
901 {
902         if (delegation == NULL)
903                 return 0;
904         if ((delegation->type & fmode) != fmode)
905                 return 0;
906         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
907                 return 0;
908         nfs_mark_delegation_referenced(delegation);
909         return 1;
910 }
911
912 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
913 {
914         switch (fmode) {
915                 case FMODE_WRITE:
916                         state->n_wronly++;
917                         break;
918                 case FMODE_READ:
919                         state->n_rdonly++;
920                         break;
921                 case FMODE_READ|FMODE_WRITE:
922                         state->n_rdwr++;
923         }
924         nfs4_state_set_mode_locked(state, state->state | fmode);
925 }
926
927 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
928 {
929         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
930                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
931         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
932         switch (fmode) {
933                 case FMODE_READ:
934                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
935                         break;
936                 case FMODE_WRITE:
937                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
938                         break;
939                 case FMODE_READ|FMODE_WRITE:
940                         set_bit(NFS_O_RDWR_STATE, &state->flags);
941         }
942 }
943
944 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
945 {
946         write_seqlock(&state->seqlock);
947         nfs_set_open_stateid_locked(state, stateid, fmode);
948         write_sequnlock(&state->seqlock);
949 }
950
951 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
952 {
953         /*
954          * Protect the call to nfs4_state_set_mode_locked and
955          * serialise the stateid update
956          */
957         write_seqlock(&state->seqlock);
958         if (deleg_stateid != NULL) {
959                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
960                 set_bit(NFS_DELEGATED_STATE, &state->flags);
961         }
962         if (open_stateid != NULL)
963                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
964         write_sequnlock(&state->seqlock);
965         spin_lock(&state->owner->so_lock);
966         update_open_stateflags(state, fmode);
967         spin_unlock(&state->owner->so_lock);
968 }
969
970 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
971 {
972         struct nfs_inode *nfsi = NFS_I(state->inode);
973         struct nfs_delegation *deleg_cur;
974         int ret = 0;
975
976         fmode &= (FMODE_READ|FMODE_WRITE);
977
978         rcu_read_lock();
979         deleg_cur = rcu_dereference(nfsi->delegation);
980         if (deleg_cur == NULL)
981                 goto no_delegation;
982
983         spin_lock(&deleg_cur->lock);
984         if (nfsi->delegation != deleg_cur ||
985             (deleg_cur->type & fmode) != fmode)
986                 goto no_delegation_unlock;
987
988         if (delegation == NULL)
989                 delegation = &deleg_cur->stateid;
990         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
991                 goto no_delegation_unlock;
992
993         nfs_mark_delegation_referenced(deleg_cur);
994         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
995         ret = 1;
996 no_delegation_unlock:
997         spin_unlock(&deleg_cur->lock);
998 no_delegation:
999         rcu_read_unlock();
1000
1001         if (!ret && open_stateid != NULL) {
1002                 __update_open_stateid(state, open_stateid, NULL, fmode);
1003                 ret = 1;
1004         }
1005
1006         return ret;
1007 }
1008
1009
1010 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1011 {
1012         struct nfs_delegation *delegation;
1013
1014         rcu_read_lock();
1015         delegation = rcu_dereference(NFS_I(inode)->delegation);
1016         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1017                 rcu_read_unlock();
1018                 return;
1019         }
1020         rcu_read_unlock();
1021         nfs_inode_return_delegation(inode);
1022 }
1023
1024 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1025 {
1026         struct nfs4_state *state = opendata->state;
1027         struct nfs_inode *nfsi = NFS_I(state->inode);
1028         struct nfs_delegation *delegation;
1029         int open_mode = opendata->o_arg.open_flags & O_EXCL;
1030         fmode_t fmode = opendata->o_arg.fmode;
1031         nfs4_stateid stateid;
1032         int ret = -EAGAIN;
1033
1034         for (;;) {
1035                 if (can_open_cached(state, fmode, open_mode)) {
1036                         spin_lock(&state->owner->so_lock);
1037                         if (can_open_cached(state, fmode, open_mode)) {
1038                                 update_open_stateflags(state, fmode);
1039                                 spin_unlock(&state->owner->so_lock);
1040                                 goto out_return_state;
1041                         }
1042                         spin_unlock(&state->owner->so_lock);
1043                 }
1044                 rcu_read_lock();
1045                 delegation = rcu_dereference(nfsi->delegation);
1046                 if (!can_open_delegated(delegation, fmode)) {
1047                         rcu_read_unlock();
1048                         break;
1049                 }
1050                 /* Save the delegation */
1051                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1052                 rcu_read_unlock();
1053                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1054                 if (ret != 0)
1055                         goto out;
1056                 ret = -EAGAIN;
1057
1058                 /* Try to update the stateid using the delegation */
1059                 if (update_open_stateid(state, NULL, &stateid, fmode))
1060                         goto out_return_state;
1061         }
1062 out:
1063         return ERR_PTR(ret);
1064 out_return_state:
1065         atomic_inc(&state->count);
1066         return state;
1067 }
1068
1069 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1070 {
1071         struct inode *inode;
1072         struct nfs4_state *state = NULL;
1073         struct nfs_delegation *delegation;
1074         int ret;
1075
1076         if (!data->rpc_done) {
1077                 state = nfs4_try_open_cached(data);
1078                 goto out;
1079         }
1080
1081         ret = -EAGAIN;
1082         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1083                 goto err;
1084         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1085         ret = PTR_ERR(inode);
1086         if (IS_ERR(inode))
1087                 goto err;
1088         ret = -ENOMEM;
1089         state = nfs4_get_open_state(inode, data->owner);
1090         if (state == NULL)
1091                 goto err_put_inode;
1092         if (data->o_res.delegation_type != 0) {
1093                 int delegation_flags = 0;
1094
1095                 rcu_read_lock();
1096                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1097                 if (delegation)
1098                         delegation_flags = delegation->flags;
1099                 rcu_read_unlock();
1100                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1101                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1102                                         "returning a delegation for "
1103                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1104                                         NFS_CLIENT(inode)->cl_server);
1105                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1106                         nfs_inode_set_delegation(state->inode,
1107                                         data->owner->so_cred,
1108                                         &data->o_res);
1109                 else
1110                         nfs_inode_reclaim_delegation(state->inode,
1111                                         data->owner->so_cred,
1112                                         &data->o_res);
1113         }
1114
1115         update_open_stateid(state, &data->o_res.stateid, NULL,
1116                         data->o_arg.fmode);
1117         iput(inode);
1118 out:
1119         return state;
1120 err_put_inode:
1121         iput(inode);
1122 err:
1123         return ERR_PTR(ret);
1124 }
1125
1126 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1127 {
1128         struct nfs_inode *nfsi = NFS_I(state->inode);
1129         struct nfs_open_context *ctx;
1130
1131         spin_lock(&state->inode->i_lock);
1132         list_for_each_entry(ctx, &nfsi->open_files, list) {
1133                 if (ctx->state != state)
1134                         continue;
1135                 get_nfs_open_context(ctx);
1136                 spin_unlock(&state->inode->i_lock);
1137                 return ctx;
1138         }
1139         spin_unlock(&state->inode->i_lock);
1140         return ERR_PTR(-ENOENT);
1141 }
1142
1143 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1144 {
1145         struct nfs4_opendata *opendata;
1146
1147         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1148         if (opendata == NULL)
1149                 return ERR_PTR(-ENOMEM);
1150         opendata->state = state;
1151         atomic_inc(&state->count);
1152         return opendata;
1153 }
1154
1155 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1156 {
1157         struct nfs4_state *newstate;
1158         int ret;
1159
1160         opendata->o_arg.open_flags = 0;
1161         opendata->o_arg.fmode = fmode;
1162         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1163         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1164         nfs4_init_opendata_res(opendata);
1165         ret = _nfs4_recover_proc_open(opendata);
1166         if (ret != 0)
1167                 return ret; 
1168         newstate = nfs4_opendata_to_nfs4_state(opendata);
1169         if (IS_ERR(newstate))
1170                 return PTR_ERR(newstate);
1171         nfs4_close_state(newstate, fmode);
1172         *res = newstate;
1173         return 0;
1174 }
1175
1176 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1177 {
1178         struct nfs4_state *newstate;
1179         int ret;
1180
1181         /* memory barrier prior to reading state->n_* */
1182         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1183         smp_rmb();
1184         if (state->n_rdwr != 0) {
1185                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1186                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1187                 if (ret != 0)
1188                         return ret;
1189                 if (newstate != state)
1190                         return -ESTALE;
1191         }
1192         if (state->n_wronly != 0) {
1193                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1194                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1195                 if (ret != 0)
1196                         return ret;
1197                 if (newstate != state)
1198                         return -ESTALE;
1199         }
1200         if (state->n_rdonly != 0) {
1201                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1202                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1203                 if (ret != 0)
1204                         return ret;
1205                 if (newstate != state)
1206                         return -ESTALE;
1207         }
1208         /*
1209          * We may have performed cached opens for all three recoveries.
1210          * Check if we need to update the current stateid.
1211          */
1212         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1213             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1214                 write_seqlock(&state->seqlock);
1215                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1216                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1217                 write_sequnlock(&state->seqlock);
1218         }
1219         return 0;
1220 }
1221
1222 /*
1223  * OPEN_RECLAIM:
1224  *      reclaim state on the server after a reboot.
1225  */
1226 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1227 {
1228         struct nfs_delegation *delegation;
1229         struct nfs4_opendata *opendata;
1230         fmode_t delegation_type = 0;
1231         int status;
1232
1233         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1234         if (IS_ERR(opendata))
1235                 return PTR_ERR(opendata);
1236         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1237         opendata->o_arg.fh = NFS_FH(state->inode);
1238         rcu_read_lock();
1239         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1240         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1241                 delegation_type = delegation->type;
1242         rcu_read_unlock();
1243         opendata->o_arg.u.delegation_type = delegation_type;
1244         status = nfs4_open_recover(opendata, state);
1245         nfs4_opendata_put(opendata);
1246         return status;
1247 }
1248
1249 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1250 {
1251         struct nfs_server *server = NFS_SERVER(state->inode);
1252         struct nfs4_exception exception = { };
1253         int err;
1254         do {
1255                 err = _nfs4_do_open_reclaim(ctx, state);
1256                 if (err != -NFS4ERR_DELAY)
1257                         break;
1258                 nfs4_handle_exception(server, err, &exception);
1259         } while (exception.retry);
1260         return err;
1261 }
1262
1263 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1264 {
1265         struct nfs_open_context *ctx;
1266         int ret;
1267
1268         ctx = nfs4_state_find_open_context(state);
1269         if (IS_ERR(ctx))
1270                 return PTR_ERR(ctx);
1271         ret = nfs4_do_open_reclaim(ctx, state);
1272         put_nfs_open_context(ctx);
1273         return ret;
1274 }
1275
1276 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1277 {
1278         struct nfs4_opendata *opendata;
1279         int ret;
1280
1281         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1282         if (IS_ERR(opendata))
1283                 return PTR_ERR(opendata);
1284         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1285         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1286                         sizeof(opendata->o_arg.u.delegation.data));
1287         ret = nfs4_open_recover(opendata, state);
1288         nfs4_opendata_put(opendata);
1289         return ret;
1290 }
1291
1292 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1293 {
1294         struct nfs4_exception exception = { };
1295         struct nfs_server *server = NFS_SERVER(state->inode);
1296         int err;
1297         do {
1298                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1299                 switch (err) {
1300                         case 0:
1301                         case -ENOENT:
1302                         case -ESTALE:
1303                                 goto out;
1304                         case -NFS4ERR_BADSESSION:
1305                         case -NFS4ERR_BADSLOT:
1306                         case -NFS4ERR_BAD_HIGH_SLOT:
1307                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1308                         case -NFS4ERR_DEADSESSION:
1309                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1310                                 goto out;
1311                         case -NFS4ERR_STALE_CLIENTID:
1312                         case -NFS4ERR_STALE_STATEID:
1313                         case -NFS4ERR_EXPIRED:
1314                                 /* Don't recall a delegation if it was lost */
1315                                 nfs4_schedule_lease_recovery(server->nfs_client);
1316                                 goto out;
1317                         case -ERESTARTSYS:
1318                                 /*
1319                                  * The show must go on: exit, but mark the
1320                                  * stateid as needing recovery.
1321                                  */
1322                         case -NFS4ERR_ADMIN_REVOKED:
1323                         case -NFS4ERR_BAD_STATEID:
1324                                 nfs4_schedule_stateid_recovery(server, state);
1325                         case -EKEYEXPIRED:
1326                                 /*
1327                                  * User RPCSEC_GSS context has expired.
1328                                  * We cannot recover this stateid now, so
1329                                  * skip it and allow recovery thread to
1330                                  * proceed.
1331                                  */
1332                         case -ENOMEM:
1333                                 err = 0;
1334                                 goto out;
1335                 }
1336                 err = nfs4_handle_exception(server, err, &exception);
1337         } while (exception.retry);
1338 out:
1339         return err;
1340 }
1341
1342 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1343 {
1344         struct nfs4_opendata *data = calldata;
1345
1346         data->rpc_status = task->tk_status;
1347         if (data->rpc_status == 0) {
1348                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1349                                 sizeof(data->o_res.stateid.data));
1350                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1351                 renew_lease(data->o_res.server, data->timestamp);
1352                 data->rpc_done = 1;
1353         }
1354 }
1355
1356 static void nfs4_open_confirm_release(void *calldata)
1357 {
1358         struct nfs4_opendata *data = calldata;
1359         struct nfs4_state *state = NULL;
1360
1361         /* If this request hasn't been cancelled, do nothing */
1362         if (data->cancelled == 0)
1363                 goto out_free;
1364         /* In case of error, no cleanup! */
1365         if (!data->rpc_done)
1366                 goto out_free;
1367         state = nfs4_opendata_to_nfs4_state(data);
1368         if (!IS_ERR(state))
1369                 nfs4_close_state(state, data->o_arg.fmode);
1370 out_free:
1371         nfs4_opendata_put(data);
1372 }
1373
1374 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1375         .rpc_call_done = nfs4_open_confirm_done,
1376         .rpc_release = nfs4_open_confirm_release,
1377 };
1378
1379 /*
1380  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1381  */
1382 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1383 {
1384         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1385         struct rpc_task *task;
1386         struct  rpc_message msg = {
1387                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1388                 .rpc_argp = &data->c_arg,
1389                 .rpc_resp = &data->c_res,
1390                 .rpc_cred = data->owner->so_cred,
1391         };
1392         struct rpc_task_setup task_setup_data = {
1393                 .rpc_client = server->client,
1394                 .rpc_message = &msg,
1395                 .callback_ops = &nfs4_open_confirm_ops,
1396                 .callback_data = data,
1397                 .workqueue = nfsiod_workqueue,
1398                 .flags = RPC_TASK_ASYNC,
1399         };
1400         int status;
1401
1402         kref_get(&data->kref);
1403         data->rpc_done = 0;
1404         data->rpc_status = 0;
1405         data->timestamp = jiffies;
1406         task = rpc_run_task(&task_setup_data);
1407         if (IS_ERR(task))
1408                 return PTR_ERR(task);
1409         status = nfs4_wait_for_completion_rpc_task(task);
1410         if (status != 0) {
1411                 data->cancelled = 1;
1412                 smp_wmb();
1413         } else
1414                 status = data->rpc_status;
1415         rpc_put_task(task);
1416         return status;
1417 }
1418
1419 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1420 {
1421         struct nfs4_opendata *data = calldata;
1422         struct nfs4_state_owner *sp = data->owner;
1423
1424         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1425                 return;
1426         /*
1427          * Check if we still need to send an OPEN call, or if we can use
1428          * a delegation instead.
1429          */
1430         if (data->state != NULL) {
1431                 struct nfs_delegation *delegation;
1432
1433                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1434                         goto out_no_action;
1435                 rcu_read_lock();
1436                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1437                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1438                     can_open_delegated(delegation, data->o_arg.fmode))
1439                         goto unlock_no_action;
1440                 rcu_read_unlock();
1441         }
1442         /* Update sequence id. */
1443         data->o_arg.id = sp->so_owner_id.id;
1444         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1445         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1446                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1447                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1448         }
1449         data->timestamp = jiffies;
1450         if (nfs4_setup_sequence(data->o_arg.server,
1451                                 &data->o_arg.seq_args,
1452                                 &data->o_res.seq_res, 1, task))
1453                 return;
1454         rpc_call_start(task);
1455         return;
1456 unlock_no_action:
1457         rcu_read_unlock();
1458 out_no_action:
1459         task->tk_action = NULL;
1460
1461 }
1462
1463 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1464 {
1465         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1466         nfs4_open_prepare(task, calldata);
1467 }
1468
1469 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1470 {
1471         struct nfs4_opendata *data = calldata;
1472
1473         data->rpc_status = task->tk_status;
1474
1475         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1476                 return;
1477
1478         if (task->tk_status == 0) {
1479                 switch (data->o_res.f_attr->mode & S_IFMT) {
1480                         case S_IFREG:
1481                                 break;
1482                         case S_IFLNK:
1483                                 data->rpc_status = -ELOOP;
1484                                 break;
1485                         case S_IFDIR:
1486                                 data->rpc_status = -EISDIR;
1487                                 break;
1488                         default:
1489                                 data->rpc_status = -ENOTDIR;
1490                 }
1491                 renew_lease(data->o_res.server, data->timestamp);
1492                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1493                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1494         }
1495         data->rpc_done = 1;
1496 }
1497
1498 static void nfs4_open_release(void *calldata)
1499 {
1500         struct nfs4_opendata *data = calldata;
1501         struct nfs4_state *state = NULL;
1502
1503         /* If this request hasn't been cancelled, do nothing */
1504         if (data->cancelled == 0)
1505                 goto out_free;
1506         /* In case of error, no cleanup! */
1507         if (data->rpc_status != 0 || !data->rpc_done)
1508                 goto out_free;
1509         /* In case we need an open_confirm, no cleanup! */
1510         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1511                 goto out_free;
1512         state = nfs4_opendata_to_nfs4_state(data);
1513         if (!IS_ERR(state))
1514                 nfs4_close_state(state, data->o_arg.fmode);
1515 out_free:
1516         nfs4_opendata_put(data);
1517 }
1518
1519 static const struct rpc_call_ops nfs4_open_ops = {
1520         .rpc_call_prepare = nfs4_open_prepare,
1521         .rpc_call_done = nfs4_open_done,
1522         .rpc_release = nfs4_open_release,
1523 };
1524
1525 static const struct rpc_call_ops nfs4_recover_open_ops = {
1526         .rpc_call_prepare = nfs4_recover_open_prepare,
1527         .rpc_call_done = nfs4_open_done,
1528         .rpc_release = nfs4_open_release,
1529 };
1530
1531 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1532 {
1533         struct inode *dir = data->dir->d_inode;
1534         struct nfs_server *server = NFS_SERVER(dir);
1535         struct nfs_openargs *o_arg = &data->o_arg;
1536         struct nfs_openres *o_res = &data->o_res;
1537         struct rpc_task *task;
1538         struct rpc_message msg = {
1539                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1540                 .rpc_argp = o_arg,
1541                 .rpc_resp = o_res,
1542                 .rpc_cred = data->owner->so_cred,
1543         };
1544         struct rpc_task_setup task_setup_data = {
1545                 .rpc_client = server->client,
1546                 .rpc_message = &msg,
1547                 .callback_ops = &nfs4_open_ops,
1548                 .callback_data = data,
1549                 .workqueue = nfsiod_workqueue,
1550                 .flags = RPC_TASK_ASYNC,
1551         };
1552         int status;
1553
1554         kref_get(&data->kref);
1555         data->rpc_done = 0;
1556         data->rpc_status = 0;
1557         data->cancelled = 0;
1558         if (isrecover)
1559                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1560         task = rpc_run_task(&task_setup_data);
1561         if (IS_ERR(task))
1562                 return PTR_ERR(task);
1563         status = nfs4_wait_for_completion_rpc_task(task);
1564         if (status != 0) {
1565                 data->cancelled = 1;
1566                 smp_wmb();
1567         } else
1568                 status = data->rpc_status;
1569         rpc_put_task(task);
1570
1571         return status;
1572 }
1573
1574 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1575 {
1576         struct inode *dir = data->dir->d_inode;
1577         struct nfs_openres *o_res = &data->o_res;
1578         int status;
1579
1580         status = nfs4_run_open_task(data, 1);
1581         if (status != 0 || !data->rpc_done)
1582                 return status;
1583
1584         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1585
1586         nfs_refresh_inode(dir, o_res->dir_attr);
1587
1588         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1589                 status = _nfs4_proc_open_confirm(data);
1590                 if (status != 0)
1591                         return status;
1592         }
1593
1594         return status;
1595 }
1596
1597 /*
1598  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1599  */
1600 static int _nfs4_proc_open(struct nfs4_opendata *data)
1601 {
1602         struct inode *dir = data->dir->d_inode;
1603         struct nfs_server *server = NFS_SERVER(dir);
1604         struct nfs_openargs *o_arg = &data->o_arg;
1605         struct nfs_openres *o_res = &data->o_res;
1606         int status;
1607
1608         status = nfs4_run_open_task(data, 0);
1609         if (!data->rpc_done)
1610                 return status;
1611         if (status != 0) {
1612                 if (status == -NFS4ERR_BADNAME &&
1613                                 !(o_arg->open_flags & O_CREAT))
1614                         return -ENOENT;
1615                 return status;
1616         }
1617
1618         nfs_fattr_map_and_free_names(server, &data->f_attr);
1619
1620         if (o_arg->open_flags & O_CREAT) {
1621                 update_changeattr(dir, &o_res->cinfo);
1622                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1623         } else
1624                 nfs_refresh_inode(dir, o_res->dir_attr);
1625         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1626                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1627         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1628                 status = _nfs4_proc_open_confirm(data);
1629                 if (status != 0)
1630                         return status;
1631         }
1632         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1633                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1634         return 0;
1635 }
1636
1637 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1638 {
1639         unsigned int loop;
1640         int ret;
1641
1642         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1643                 ret = nfs4_wait_clnt_recover(clp);
1644                 if (ret != 0)
1645                         break;
1646                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1647                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1648                         break;
1649                 nfs4_schedule_state_manager(clp);
1650                 ret = -EIO;
1651         }
1652         return ret;
1653 }
1654
1655 static int nfs4_recover_expired_lease(struct nfs_server *server)
1656 {
1657         return nfs4_client_recover_expired_lease(server->nfs_client);
1658 }
1659
1660 /*
1661  * OPEN_EXPIRED:
1662  *      reclaim state on the server after a network partition.
1663  *      Assumes caller holds the appropriate lock
1664  */
1665 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1666 {
1667         struct nfs4_opendata *opendata;
1668         int ret;
1669
1670         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1671         if (IS_ERR(opendata))
1672                 return PTR_ERR(opendata);
1673         ret = nfs4_open_recover(opendata, state);
1674         if (ret == -ESTALE)
1675                 d_drop(ctx->dentry);
1676         nfs4_opendata_put(opendata);
1677         return ret;
1678 }
1679
1680 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1681 {
1682         struct nfs_server *server = NFS_SERVER(state->inode);
1683         struct nfs4_exception exception = { };
1684         int err;
1685
1686         do {
1687                 err = _nfs4_open_expired(ctx, state);
1688                 switch (err) {
1689                 default:
1690                         goto out;
1691                 case -NFS4ERR_GRACE:
1692                 case -NFS4ERR_DELAY:
1693                         nfs4_handle_exception(server, err, &exception);
1694                         err = 0;
1695                 }
1696         } while (exception.retry);
1697 out:
1698         return err;
1699 }
1700
1701 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1702 {
1703         struct nfs_open_context *ctx;
1704         int ret;
1705
1706         ctx = nfs4_state_find_open_context(state);
1707         if (IS_ERR(ctx))
1708                 return PTR_ERR(ctx);
1709         ret = nfs4_do_open_expired(ctx, state);
1710         put_nfs_open_context(ctx);
1711         return ret;
1712 }
1713
1714 #if defined(CONFIG_NFS_V4_1)
1715 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1716 {
1717         int status;
1718         struct nfs_server *server = NFS_SERVER(state->inode);
1719
1720         status = nfs41_test_stateid(server, state);
1721         if (status == NFS_OK)
1722                 return 0;
1723         nfs41_free_stateid(server, state);
1724         return nfs4_open_expired(sp, state);
1725 }
1726 #endif
1727
1728 /*
1729  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1730  * fields corresponding to attributes that were used to store the verifier.
1731  * Make sure we clobber those fields in the later setattr call
1732  */
1733 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1734 {
1735         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1736             !(sattr->ia_valid & ATTR_ATIME_SET))
1737                 sattr->ia_valid |= ATTR_ATIME;
1738
1739         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1740             !(sattr->ia_valid & ATTR_MTIME_SET))
1741                 sattr->ia_valid |= ATTR_MTIME;
1742 }
1743
1744 /*
1745  * Returns a referenced nfs4_state
1746  */
1747 static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1748 {
1749         struct nfs4_state_owner  *sp;
1750         struct nfs4_state     *state = NULL;
1751         struct nfs_server       *server = NFS_SERVER(dir);
1752         struct nfs4_opendata *opendata;
1753         int status;
1754
1755         /* Protect against reboot recovery conflicts */
1756         status = -ENOMEM;
1757         if (!(sp = nfs4_get_state_owner(server, cred))) {
1758                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1759                 goto out_err;
1760         }
1761         status = nfs4_recover_expired_lease(server);
1762         if (status != 0)
1763                 goto err_put_state_owner;
1764         if (dentry->d_inode != NULL)
1765                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1766         status = -ENOMEM;
1767         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1768         if (opendata == NULL)
1769                 goto err_put_state_owner;
1770
1771         if (dentry->d_inode != NULL)
1772                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1773
1774         status = _nfs4_proc_open(opendata);
1775         if (status != 0)
1776                 goto err_opendata_put;
1777
1778         state = nfs4_opendata_to_nfs4_state(opendata);
1779         status = PTR_ERR(state);
1780         if (IS_ERR(state))
1781                 goto err_opendata_put;
1782         if (server->caps & NFS_CAP_POSIX_LOCK)
1783                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1784
1785         if (opendata->o_arg.open_flags & O_EXCL) {
1786                 nfs4_exclusive_attrset(opendata, sattr);
1787
1788                 nfs_fattr_init(opendata->o_res.f_attr);
1789                 status = nfs4_do_setattr(state->inode, cred,
1790                                 opendata->o_res.f_attr, sattr,
1791                                 state);
1792                 if (status == 0)
1793                         nfs_setattr_update_inode(state->inode, sattr);
1794                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1795         }
1796         nfs4_opendata_put(opendata);
1797         nfs4_put_state_owner(sp);
1798         *res = state;
1799         return 0;
1800 err_opendata_put:
1801         nfs4_opendata_put(opendata);
1802 err_put_state_owner:
1803         nfs4_put_state_owner(sp);
1804 out_err:
1805         *res = NULL;
1806         return status;
1807 }
1808
1809
1810 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1811 {
1812         struct nfs4_exception exception = { };
1813         struct nfs4_state *res;
1814         int status;
1815
1816         do {
1817                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred, &res);
1818                 if (status == 0)
1819                         break;
1820                 /* NOTE: BAD_SEQID means the server and client disagree about the
1821                  * book-keeping w.r.t. state-changing operations
1822                  * (OPEN/CLOSE/LOCK/LOCKU...)
1823                  * It is actually a sign of a bug on the client or on the server.
1824                  *
1825                  * If we receive a BAD_SEQID error in the particular case of
1826                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1827                  * have unhashed the old state_owner for us, and that we can
1828                  * therefore safely retry using a new one. We should still warn
1829                  * the user though...
1830                  */
1831                 if (status == -NFS4ERR_BAD_SEQID) {
1832                         printk(KERN_WARNING "NFS: v4 server %s "
1833                                         " returned a bad sequence-id error!\n",
1834                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1835                         exception.retry = 1;
1836                         continue;
1837                 }
1838                 /*
1839                  * BAD_STATEID on OPEN means that the server cancelled our
1840                  * state before it received the OPEN_CONFIRM.
1841                  * Recover by retrying the request as per the discussion
1842                  * on Page 181 of RFC3530.
1843                  */
1844                 if (status == -NFS4ERR_BAD_STATEID) {
1845                         exception.retry = 1;
1846                         continue;
1847                 }
1848                 if (status == -EAGAIN) {
1849                         /* We must have found a delegation */
1850                         exception.retry = 1;
1851                         continue;
1852                 }
1853                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1854                                         status, &exception));
1855         } while (exception.retry);
1856         return res;
1857 }
1858
1859 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1860                             struct nfs_fattr *fattr, struct iattr *sattr,
1861                             struct nfs4_state *state)
1862 {
1863         struct nfs_server *server = NFS_SERVER(inode);
1864         struct nfs_setattrargs  arg = {
1865                 .fh             = NFS_FH(inode),
1866                 .iap            = sattr,
1867                 .server         = server,
1868                 .bitmask = server->attr_bitmask,
1869         };
1870         struct nfs_setattrres  res = {
1871                 .fattr          = fattr,
1872                 .server         = server,
1873         };
1874         struct rpc_message msg = {
1875                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1876                 .rpc_argp       = &arg,
1877                 .rpc_resp       = &res,
1878                 .rpc_cred       = cred,
1879         };
1880         unsigned long timestamp = jiffies;
1881         int status;
1882
1883         nfs_fattr_init(fattr);
1884
1885         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1886                 /* Use that stateid */
1887         } else if (state != NULL) {
1888                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1889         } else
1890                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1891
1892         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1893         if (status == 0 && state != NULL)
1894                 renew_lease(server, timestamp);
1895         return status;
1896 }
1897
1898 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1899                            struct nfs_fattr *fattr, struct iattr *sattr,
1900                            struct nfs4_state *state)
1901 {
1902         struct nfs_server *server = NFS_SERVER(inode);
1903         struct nfs4_exception exception = { };
1904         int err;
1905         do {
1906                 err = nfs4_handle_exception(server,
1907                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1908                                 &exception);
1909         } while (exception.retry);
1910         return err;
1911 }
1912
1913 struct nfs4_closedata {
1914         struct inode *inode;
1915         struct nfs4_state *state;
1916         struct nfs_closeargs arg;
1917         struct nfs_closeres res;
1918         struct nfs_fattr fattr;
1919         unsigned long timestamp;
1920         bool roc;
1921         u32 roc_barrier;
1922 };
1923
1924 static void nfs4_free_closedata(void *data)
1925 {
1926         struct nfs4_closedata *calldata = data;
1927         struct nfs4_state_owner *sp = calldata->state->owner;
1928         struct super_block *sb = calldata->state->inode->i_sb;
1929
1930         if (calldata->roc)
1931                 pnfs_roc_release(calldata->state->inode);
1932         nfs4_put_open_state(calldata->state);
1933         nfs_free_seqid(calldata->arg.seqid);
1934         nfs4_put_state_owner(sp);
1935         nfs_sb_deactive(sb);
1936         kfree(calldata);
1937 }
1938
1939 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1940                 fmode_t fmode)
1941 {
1942         spin_lock(&state->owner->so_lock);
1943         if (!(fmode & FMODE_READ))
1944                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1945         if (!(fmode & FMODE_WRITE))
1946                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1947         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1948         spin_unlock(&state->owner->so_lock);
1949 }
1950
1951 static void nfs4_close_done(struct rpc_task *task, void *data)
1952 {
1953         struct nfs4_closedata *calldata = data;
1954         struct nfs4_state *state = calldata->state;
1955         struct nfs_server *server = NFS_SERVER(calldata->inode);
1956
1957         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1958                 return;
1959         /* hmm. we are done with the inode, and in the process of freeing
1960          * the state_owner. we keep this around to process errors
1961          */
1962         switch (task->tk_status) {
1963                 case 0:
1964                         if (calldata->roc)
1965                                 pnfs_roc_set_barrier(state->inode,
1966                                                      calldata->roc_barrier);
1967                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1968                         renew_lease(server, calldata->timestamp);
1969                         nfs4_close_clear_stateid_flags(state,
1970                                         calldata->arg.fmode);
1971                         break;
1972                 case -NFS4ERR_STALE_STATEID:
1973                 case -NFS4ERR_OLD_STATEID:
1974                 case -NFS4ERR_BAD_STATEID:
1975                 case -NFS4ERR_EXPIRED:
1976                         if (calldata->arg.fmode == 0)
1977                                 break;
1978                 default:
1979                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1980                                 rpc_restart_call_prepare(task);
1981         }
1982         nfs_release_seqid(calldata->arg.seqid);
1983         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1984 }
1985
1986 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1987 {
1988         struct nfs4_closedata *calldata = data;
1989         struct nfs4_state *state = calldata->state;
1990         int call_close = 0;
1991
1992         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1993                 return;
1994
1995         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1996         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1997         spin_lock(&state->owner->so_lock);
1998         /* Calculate the change in open mode */
1999         if (state->n_rdwr == 0) {
2000                 if (state->n_rdonly == 0) {
2001                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2002                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2003                         calldata->arg.fmode &= ~FMODE_READ;
2004                 }
2005                 if (state->n_wronly == 0) {
2006                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2007                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2008                         calldata->arg.fmode &= ~FMODE_WRITE;
2009                 }
2010         }
2011         spin_unlock(&state->owner->so_lock);
2012
2013         if (!call_close) {
2014                 /* Note: exit _without_ calling nfs4_close_done */
2015                 task->tk_action = NULL;
2016                 return;
2017         }
2018
2019         if (calldata->arg.fmode == 0) {
2020                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2021                 if (calldata->roc &&
2022                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2023                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2024                                      task, NULL);
2025                         return;
2026                 }
2027         }
2028
2029         nfs_fattr_init(calldata->res.fattr);
2030         calldata->timestamp = jiffies;
2031         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2032                                 &calldata->arg.seq_args, &calldata->res.seq_res,
2033                                 1, task))
2034                 return;
2035         rpc_call_start(task);
2036 }
2037
2038 static const struct rpc_call_ops nfs4_close_ops = {
2039         .rpc_call_prepare = nfs4_close_prepare,
2040         .rpc_call_done = nfs4_close_done,
2041         .rpc_release = nfs4_free_closedata,
2042 };
2043
2044 /* 
2045  * It is possible for data to be read/written from a mem-mapped file 
2046  * after the sys_close call (which hits the vfs layer as a flush).
2047  * This means that we can't safely call nfsv4 close on a file until 
2048  * the inode is cleared. This in turn means that we are not good
2049  * NFSv4 citizens - we do not indicate to the server to update the file's 
2050  * share state even when we are done with one of the three share 
2051  * stateid's in the inode.
2052  *
2053  * NOTE: Caller must be holding the sp->so_owner semaphore!
2054  */
2055 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2056 {
2057         struct nfs_server *server = NFS_SERVER(state->inode);
2058         struct nfs4_closedata *calldata;
2059         struct nfs4_state_owner *sp = state->owner;
2060         struct rpc_task *task;
2061         struct rpc_message msg = {
2062                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2063                 .rpc_cred = state->owner->so_cred,
2064         };
2065         struct rpc_task_setup task_setup_data = {
2066                 .rpc_client = server->client,
2067                 .rpc_message = &msg,
2068                 .callback_ops = &nfs4_close_ops,
2069                 .workqueue = nfsiod_workqueue,
2070                 .flags = RPC_TASK_ASYNC,
2071         };
2072         int status = -ENOMEM;
2073
2074         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2075         if (calldata == NULL)
2076                 goto out;
2077         calldata->inode = state->inode;
2078         calldata->state = state;
2079         calldata->arg.fh = NFS_FH(state->inode);
2080         calldata->arg.stateid = &state->open_stateid;
2081         /* Serialization for the sequence id */
2082         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2083         if (calldata->arg.seqid == NULL)
2084                 goto out_free_calldata;
2085         calldata->arg.fmode = 0;
2086         calldata->arg.bitmask = server->cache_consistency_bitmask;
2087         calldata->res.fattr = &calldata->fattr;
2088         calldata->res.seqid = calldata->arg.seqid;
2089         calldata->res.server = server;
2090         calldata->roc = roc;
2091         nfs_sb_active(calldata->inode->i_sb);
2092
2093         msg.rpc_argp = &calldata->arg;
2094         msg.rpc_resp = &calldata->res;
2095         task_setup_data.callback_data = calldata;
2096         task = rpc_run_task(&task_setup_data);
2097         if (IS_ERR(task))
2098                 return PTR_ERR(task);
2099         status = 0;
2100         if (wait)
2101                 status = rpc_wait_for_completion_task(task);
2102         rpc_put_task(task);
2103         return status;
2104 out_free_calldata:
2105         kfree(calldata);
2106 out:
2107         if (roc)
2108                 pnfs_roc_release(state->inode);
2109         nfs4_put_open_state(state);
2110         nfs4_put_state_owner(sp);
2111         return status;
2112 }
2113
2114 static struct inode *
2115 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2116 {
2117         struct nfs4_state *state;
2118
2119         /* Protect against concurrent sillydeletes */
2120         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2121         if (IS_ERR(state))
2122                 return ERR_CAST(state);
2123         ctx->state = state;
2124         return igrab(state->inode);
2125 }
2126
2127 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2128 {
2129         if (ctx->state == NULL)
2130                 return;
2131         if (is_sync)
2132                 nfs4_close_sync(ctx->state, ctx->mode);
2133         else
2134                 nfs4_close_state(ctx->state, ctx->mode);
2135 }
2136
2137 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2138 {
2139         struct nfs4_server_caps_arg args = {
2140                 .fhandle = fhandle,
2141         };
2142         struct nfs4_server_caps_res res = {};
2143         struct rpc_message msg = {
2144                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2145                 .rpc_argp = &args,
2146                 .rpc_resp = &res,
2147         };
2148         int status;
2149
2150         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2151         if (status == 0) {
2152                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2153                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2154                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2155                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2156                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2157                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2158                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2159                         server->caps |= NFS_CAP_ACLS;
2160                 if (res.has_links != 0)
2161                         server->caps |= NFS_CAP_HARDLINKS;
2162                 if (res.has_symlinks != 0)
2163                         server->caps |= NFS_CAP_SYMLINKS;
2164                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2165                         server->caps |= NFS_CAP_FILEID;
2166                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2167                         server->caps |= NFS_CAP_MODE;
2168                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2169                         server->caps |= NFS_CAP_NLINK;
2170                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2171                         server->caps |= NFS_CAP_OWNER;
2172                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2173                         server->caps |= NFS_CAP_OWNER_GROUP;
2174                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2175                         server->caps |= NFS_CAP_ATIME;
2176                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2177                         server->caps |= NFS_CAP_CTIME;
2178                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2179                         server->caps |= NFS_CAP_MTIME;
2180
2181                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2182                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2183                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2184                 server->acl_bitmask = res.acl_bitmask;
2185         }
2186
2187         return status;
2188 }
2189
2190 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2191 {
2192         struct nfs4_exception exception = { };
2193         int err;
2194         do {
2195                 err = nfs4_handle_exception(server,
2196                                 _nfs4_server_capabilities(server, fhandle),
2197                                 &exception);
2198         } while (exception.retry);
2199         return err;
2200 }
2201
2202 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2203                 struct nfs_fsinfo *info)
2204 {
2205         struct nfs4_lookup_root_arg args = {
2206                 .bitmask = nfs4_fattr_bitmap,
2207         };
2208         struct nfs4_lookup_res res = {
2209                 .server = server,
2210                 .fattr = info->fattr,
2211                 .fh = fhandle,
2212         };
2213         struct rpc_message msg = {
2214                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2215                 .rpc_argp = &args,
2216                 .rpc_resp = &res,
2217         };
2218
2219         nfs_fattr_init(info->fattr);
2220         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2221 }
2222
2223 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2224                 struct nfs_fsinfo *info)
2225 {
2226         struct nfs4_exception exception = { };
2227         int err;
2228         do {
2229                 err = _nfs4_lookup_root(server, fhandle, info);
2230                 switch (err) {
2231                 case 0:
2232                 case -NFS4ERR_WRONGSEC:
2233                         break;
2234                 default:
2235                         err = nfs4_handle_exception(server, err, &exception);
2236                 }
2237         } while (exception.retry);
2238         return err;
2239 }
2240
2241 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2242                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2243 {
2244         struct rpc_auth *auth;
2245         int ret;
2246
2247         auth = rpcauth_create(flavor, server->client);
2248         if (!auth) {
2249                 ret = -EIO;
2250                 goto out;
2251         }
2252         ret = nfs4_lookup_root(server, fhandle, info);
2253 out:
2254         return ret;
2255 }
2256
2257 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2258                               struct nfs_fsinfo *info)
2259 {
2260         int i, len, status = 0;
2261         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2262
2263         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2264         flav_array[len] = RPC_AUTH_NULL;
2265         len += 1;
2266
2267         for (i = 0; i < len; i++) {
2268                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2269                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2270                         continue;
2271                 break;
2272         }
2273         /*
2274          * -EACCESS could mean that the user doesn't have correct permissions
2275          * to access the mount.  It could also mean that we tried to mount
2276          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2277          * existing mount programs don't handle -EACCES very well so it should
2278          * be mapped to -EPERM instead.
2279          */
2280         if (status == -EACCES)
2281                 status = -EPERM;
2282         return status;
2283 }
2284
2285 /*
2286  * get the file handle for the "/" directory on the server
2287  */
2288 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2289                               struct nfs_fsinfo *info)
2290 {
2291         int minor_version = server->nfs_client->cl_minorversion;
2292         int status = nfs4_lookup_root(server, fhandle, info);
2293         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2294                 /*
2295                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2296                  * by nfs4_map_errors() as this function exits.
2297                  */
2298                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2299         if (status == 0)
2300                 status = nfs4_server_capabilities(server, fhandle);
2301         if (status == 0)
2302                 status = nfs4_do_fsinfo(server, fhandle, info);
2303         return nfs4_map_errors(status);
2304 }
2305
2306 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
2307 /*
2308  * Get locations and (maybe) other attributes of a referral.
2309  * Note that we'll actually follow the referral later when
2310  * we detect fsid mismatch in inode revalidation
2311  */
2312 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2313                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2314 {
2315         int status = -ENOMEM;
2316         struct page *page = NULL;
2317         struct nfs4_fs_locations *locations = NULL;
2318
2319         page = alloc_page(GFP_KERNEL);
2320         if (page == NULL)
2321                 goto out;
2322         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2323         if (locations == NULL)
2324                 goto out;
2325
2326         status = nfs4_proc_fs_locations(dir, name, locations, page);
2327         if (status != 0)
2328                 goto out;
2329         /* Make sure server returned a different fsid for the referral */
2330         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2331                 dprintk("%s: server did not return a different fsid for"
2332                         " a referral at %s\n", __func__, name->name);
2333                 status = -EIO;
2334                 goto out;
2335         }
2336         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2337         nfs_fixup_referral_attributes(&locations->fattr);
2338
2339         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2340         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2341         memset(fhandle, 0, sizeof(struct nfs_fh));
2342 out:
2343         if (page)
2344                 __free_page(page);
2345         kfree(locations);
2346         return status;
2347 }
2348
2349 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2350 {
2351         struct nfs4_getattr_arg args = {
2352                 .fh = fhandle,
2353                 .bitmask = server->attr_bitmask,
2354         };
2355         struct nfs4_getattr_res res = {
2356                 .fattr = fattr,
2357                 .server = server,
2358         };
2359         struct rpc_message msg = {
2360                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2361                 .rpc_argp = &args,
2362                 .rpc_resp = &res,
2363         };
2364         
2365         nfs_fattr_init(fattr);
2366         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2367 }
2368
2369 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2370 {
2371         struct nfs4_exception exception = { };
2372         int err;
2373         do {
2374                 err = nfs4_handle_exception(server,
2375                                 _nfs4_proc_getattr(server, fhandle, fattr),
2376                                 &exception);
2377         } while (exception.retry);
2378         return err;
2379 }
2380
2381 /* 
2382  * The file is not closed if it is opened due to the a request to change
2383  * the size of the file. The open call will not be needed once the
2384  * VFS layer lookup-intents are implemented.
2385  *
2386  * Close is called when the inode is destroyed.
2387  * If we haven't opened the file for O_WRONLY, we
2388  * need to in the size_change case to obtain a stateid.
2389  *
2390  * Got race?
2391  * Because OPEN is always done by name in nfsv4, it is
2392  * possible that we opened a different file by the same
2393  * name.  We can recognize this race condition, but we
2394  * can't do anything about it besides returning an error.
2395  *
2396  * This will be fixed with VFS changes (lookup-intent).
2397  */
2398 static int
2399 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2400                   struct iattr *sattr)
2401 {
2402         struct inode *inode = dentry->d_inode;
2403         struct rpc_cred *cred = NULL;
2404         struct nfs4_state *state = NULL;
2405         int status;
2406
2407         if (pnfs_ld_layoutret_on_setattr(inode))
2408                 pnfs_return_layout(inode);
2409
2410         nfs_fattr_init(fattr);
2411         
2412         /* Search for an existing open(O_WRITE) file */
2413         if (sattr->ia_valid & ATTR_FILE) {
2414                 struct nfs_open_context *ctx;
2415
2416                 ctx = nfs_file_open_context(sattr->ia_file);
2417                 if (ctx) {
2418                         cred = ctx->cred;
2419                         state = ctx->state;
2420                 }
2421         }
2422
2423         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2424         if (status == 0)
2425                 nfs_setattr_update_inode(inode, sattr);
2426         return status;
2427 }
2428
2429 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2430                 const struct qstr *name, struct nfs_fh *fhandle,
2431                 struct nfs_fattr *fattr)
2432 {
2433         struct nfs_server *server = NFS_SERVER(dir);
2434         int                    status;
2435         struct nfs4_lookup_arg args = {
2436                 .bitmask = server->attr_bitmask,
2437                 .dir_fh = NFS_FH(dir),
2438                 .name = name,
2439         };
2440         struct nfs4_lookup_res res = {
2441                 .server = server,
2442                 .fattr = fattr,
2443                 .fh = fhandle,
2444         };
2445         struct rpc_message msg = {
2446                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2447                 .rpc_argp = &args,
2448                 .rpc_resp = &res,
2449         };
2450
2451         nfs_fattr_init(fattr);
2452
2453         dprintk("NFS call  lookup %s\n", name->name);
2454         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2455         dprintk("NFS reply lookup: %d\n", status);
2456         return status;
2457 }
2458
2459 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2460 {
2461         memset(fh, 0, sizeof(struct nfs_fh));
2462         fattr->fsid.major = 1;
2463         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2464                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2465         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2466         fattr->nlink = 2;
2467 }
2468
2469 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2470                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2471 {
2472         struct nfs4_exception exception = { };
2473         int err;
2474         do {
2475                 int status;
2476
2477                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2478                 switch (status) {
2479                 case -NFS4ERR_BADNAME:
2480                         return -ENOENT;
2481                 case -NFS4ERR_MOVED:
2482                         return nfs4_get_referral(dir, name, fattr, fhandle);
2483                 case -NFS4ERR_WRONGSEC:
2484                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2485                 }
2486                 err = nfs4_handle_exception(NFS_SERVER(dir),
2487                                 status, &exception);
2488         } while (exception.retry);
2489         return err;
2490 }
2491
2492 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2493 {
2494         struct nfs_server *server = NFS_SERVER(inode);
2495         struct nfs4_accessargs args = {
2496                 .fh = NFS_FH(inode),
2497                 .bitmask = server->attr_bitmask,
2498         };
2499         struct nfs4_accessres res = {
2500                 .server = server,
2501         };
2502         struct rpc_message msg = {
2503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2504                 .rpc_argp = &args,
2505                 .rpc_resp = &res,
2506                 .rpc_cred = entry->cred,
2507         };
2508         int mode = entry->mask;
2509         int status;
2510
2511         /*
2512          * Determine which access bits we want to ask for...
2513          */
2514         if (mode & MAY_READ)
2515                 args.access |= NFS4_ACCESS_READ;
2516         if (S_ISDIR(inode->i_mode)) {
2517                 if (mode & MAY_WRITE)
2518                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2519                 if (mode & MAY_EXEC)
2520                         args.access |= NFS4_ACCESS_LOOKUP;
2521         } else {
2522                 if (mode & MAY_WRITE)
2523                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2524                 if (mode & MAY_EXEC)
2525                         args.access |= NFS4_ACCESS_EXECUTE;
2526         }
2527
2528         res.fattr = nfs_alloc_fattr();
2529         if (res.fattr == NULL)
2530                 return -ENOMEM;
2531
2532         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2533         if (!status) {
2534                 entry->mask = 0;
2535                 if (res.access & NFS4_ACCESS_READ)
2536                         entry->mask |= MAY_READ;
2537                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2538                         entry->mask |= MAY_WRITE;
2539                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2540                         entry->mask |= MAY_EXEC;
2541                 nfs_refresh_inode(inode, res.fattr);
2542         }
2543         nfs_free_fattr(res.fattr);
2544         return status;
2545 }
2546
2547 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2548 {
2549         struct nfs4_exception exception = { };
2550         int err;
2551         do {
2552                 err = nfs4_handle_exception(NFS_SERVER(inode),
2553                                 _nfs4_proc_access(inode, entry),
2554                                 &exception);
2555         } while (exception.retry);
2556         return err;
2557 }
2558
2559 /*
2560  * TODO: For the time being, we don't try to get any attributes
2561  * along with any of the zero-copy operations READ, READDIR,
2562  * READLINK, WRITE.
2563  *
2564  * In the case of the first three, we want to put the GETATTR
2565  * after the read-type operation -- this is because it is hard
2566  * to predict the length of a GETATTR response in v4, and thus
2567  * align the READ data correctly.  This means that the GETATTR
2568  * may end up partially falling into the page cache, and we should
2569  * shift it into the 'tail' of the xdr_buf before processing.
2570  * To do this efficiently, we need to know the total length
2571  * of data received, which doesn't seem to be available outside
2572  * of the RPC layer.
2573  *
2574  * In the case of WRITE, we also want to put the GETATTR after
2575  * the operation -- in this case because we want to make sure
2576  * we get the post-operation mtime and size.  This means that
2577  * we can't use xdr_encode_pages() as written: we need a variant
2578  * of it which would leave room in the 'tail' iovec.
2579  *
2580  * Both of these changes to the XDR layer would in fact be quite
2581  * minor, but I decided to leave them for a subsequent patch.
2582  */
2583 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2584                 unsigned int pgbase, unsigned int pglen)
2585 {
2586         struct nfs4_readlink args = {
2587                 .fh       = NFS_FH(inode),
2588                 .pgbase   = pgbase,
2589                 .pglen    = pglen,
2590                 .pages    = &page,
2591         };
2592         struct nfs4_readlink_res res;
2593         struct rpc_message msg = {
2594                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2595                 .rpc_argp = &args,
2596                 .rpc_resp = &res,
2597         };
2598
2599         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2600 }
2601
2602 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2603                 unsigned int pgbase, unsigned int pglen)
2604 {
2605         struct nfs4_exception exception = { };
2606         int err;
2607         do {
2608                 err = nfs4_handle_exception(NFS_SERVER(inode),
2609                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2610                                 &exception);
2611         } while (exception.retry);
2612         return err;
2613 }
2614
2615 /*
2616  * Got race?
2617  * We will need to arrange for the VFS layer to provide an atomic open.
2618  * Until then, this create/open method is prone to inefficiency and race
2619  * conditions due to the lookup, create, and open VFS calls from sys_open()
2620  * placed on the wire.
2621  *
2622  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2623  * The file will be opened again in the subsequent VFS open call
2624  * (nfs4_proc_file_open).
2625  *
2626  * The open for read will just hang around to be used by any process that
2627  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2628  */
2629
2630 static int
2631 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2632                  int flags, struct nfs_open_context *ctx)
2633 {
2634         struct dentry *de = dentry;
2635         struct nfs4_state *state;
2636         struct rpc_cred *cred = NULL;
2637         fmode_t fmode = 0;
2638         int status = 0;
2639
2640         if (ctx != NULL) {
2641                 cred = ctx->cred;
2642                 de = ctx->dentry;
2643                 fmode = ctx->mode;
2644         }
2645         sattr->ia_mode &= ~current_umask();
2646         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2647         d_drop(dentry);
2648         if (IS_ERR(state)) {
2649                 status = PTR_ERR(state);
2650                 goto out;
2651         }
2652         d_add(dentry, igrab(state->inode));
2653         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2654         if (ctx != NULL)
2655                 ctx->state = state;
2656         else
2657                 nfs4_close_sync(state, fmode);
2658 out:
2659         return status;
2660 }
2661
2662 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2663 {
2664         struct nfs_server *server = NFS_SERVER(dir);
2665         struct nfs_removeargs args = {
2666                 .fh = NFS_FH(dir),
2667                 .name.len = name->len,
2668                 .name.name = name->name,
2669                 .bitmask = server->attr_bitmask,
2670         };
2671         struct nfs_removeres res = {
2672                 .server = server,
2673         };
2674         struct rpc_message msg = {
2675                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2676                 .rpc_argp = &args,
2677                 .rpc_resp = &res,
2678         };
2679         int status = -ENOMEM;
2680
2681         res.dir_attr = nfs_alloc_fattr();
2682         if (res.dir_attr == NULL)
2683                 goto out;
2684
2685         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2686         if (status == 0) {
2687                 update_changeattr(dir, &res.cinfo);
2688                 nfs_post_op_update_inode(dir, res.dir_attr);
2689         }
2690         nfs_free_fattr(res.dir_attr);
2691 out:
2692         return status;
2693 }
2694
2695 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2696 {
2697         struct nfs4_exception exception = { };
2698         int err;
2699         do {
2700                 err = nfs4_handle_exception(NFS_SERVER(dir),
2701                                 _nfs4_proc_remove(dir, name),
2702                                 &exception);
2703         } while (exception.retry);
2704         return err;
2705 }
2706
2707 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2708 {
2709         struct nfs_server *server = NFS_SERVER(dir);
2710         struct nfs_removeargs *args = msg->rpc_argp;
2711         struct nfs_removeres *res = msg->rpc_resp;
2712
2713         args->bitmask = server->cache_consistency_bitmask;
2714         res->server = server;
2715         res->seq_res.sr_slot = NULL;
2716         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2717 }
2718
2719 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2720 {
2721         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2722
2723         if (!nfs4_sequence_done(task, &res->seq_res))
2724                 return 0;
2725         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2726                 return 0;
2727         update_changeattr(dir, &res->cinfo);
2728         nfs_post_op_update_inode(dir, res->dir_attr);
2729         return 1;
2730 }
2731
2732 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2733 {
2734         struct nfs_server *server = NFS_SERVER(dir);
2735         struct nfs_renameargs *arg = msg->rpc_argp;
2736         struct nfs_renameres *res = msg->rpc_resp;
2737
2738         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2739         arg->bitmask = server->attr_bitmask;
2740         res->server = server;
2741 }
2742
2743 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2744                                  struct inode *new_dir)
2745 {
2746         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2747
2748         if (!nfs4_sequence_done(task, &res->seq_res))
2749                 return 0;
2750         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2751                 return 0;
2752
2753         update_changeattr(old_dir, &res->old_cinfo);
2754         nfs_post_op_update_inode(old_dir, res->old_fattr);
2755         update_changeattr(new_dir, &res->new_cinfo);
2756         nfs_post_op_update_inode(new_dir, res->new_fattr);
2757         return 1;
2758 }
2759
2760 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2761                 struct inode *new_dir, struct qstr *new_name)
2762 {
2763         struct nfs_server *server = NFS_SERVER(old_dir);
2764         struct nfs_renameargs arg = {
2765                 .old_dir = NFS_FH(old_dir),
2766                 .new_dir = NFS_FH(new_dir),
2767                 .old_name = old_name,
2768                 .new_name = new_name,
2769                 .bitmask = server->attr_bitmask,
2770         };
2771         struct nfs_renameres res = {
2772                 .server = server,
2773         };
2774         struct rpc_message msg = {
2775                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2776                 .rpc_argp = &arg,
2777                 .rpc_resp = &res,
2778         };
2779         int status = -ENOMEM;
2780         
2781         res.old_fattr = nfs_alloc_fattr();
2782         res.new_fattr = nfs_alloc_fattr();
2783         if (res.old_fattr == NULL || res.new_fattr == NULL)
2784                 goto out;
2785
2786         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2787         if (!status) {
2788                 update_changeattr(old_dir, &res.old_cinfo);
2789                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2790                 update_changeattr(new_dir, &res.new_cinfo);
2791                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2792         }
2793 out:
2794         nfs_free_fattr(res.new_fattr);
2795         nfs_free_fattr(res.old_fattr);
2796         return status;
2797 }
2798
2799 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2800                 struct inode *new_dir, struct qstr *new_name)
2801 {
2802         struct nfs4_exception exception = { };
2803         int err;
2804         do {
2805                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2806                                 _nfs4_proc_rename(old_dir, old_name,
2807                                         new_dir, new_name),
2808                                 &exception);
2809         } while (exception.retry);
2810         return err;
2811 }
2812
2813 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2814 {
2815         struct nfs_server *server = NFS_SERVER(inode);
2816         struct nfs4_link_arg arg = {
2817                 .fh     = NFS_FH(inode),
2818                 .dir_fh = NFS_FH(dir),
2819                 .name   = name,
2820                 .bitmask = server->attr_bitmask,
2821         };
2822         struct nfs4_link_res res = {
2823                 .server = server,
2824         };
2825         struct rpc_message msg = {
2826                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2827                 .rpc_argp = &arg,
2828                 .rpc_resp = &res,
2829         };
2830         int status = -ENOMEM;
2831
2832         res.fattr = nfs_alloc_fattr();
2833         res.dir_attr = nfs_alloc_fattr();
2834         if (res.fattr == NULL || res.dir_attr == NULL)
2835                 goto out;
2836
2837         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2838         if (!status) {
2839                 update_changeattr(dir, &res.cinfo);
2840                 nfs_post_op_update_inode(dir, res.dir_attr);
2841                 nfs_post_op_update_inode(inode, res.fattr);
2842         }
2843 out:
2844         nfs_free_fattr(res.dir_attr);
2845         nfs_free_fattr(res.fattr);
2846         return status;
2847 }
2848
2849 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2850 {
2851         struct nfs4_exception exception = { };
2852         int err;
2853         do {
2854                 err = nfs4_handle_exception(NFS_SERVER(inode),
2855                                 _nfs4_proc_link(inode, dir, name),
2856                                 &exception);
2857         } while (exception.retry);
2858         return err;
2859 }
2860
2861 struct nfs4_createdata {
2862         struct rpc_message msg;
2863         struct nfs4_create_arg arg;
2864         struct nfs4_create_res res;
2865         struct nfs_fh fh;
2866         struct nfs_fattr fattr;
2867         struct nfs_fattr dir_fattr;
2868 };
2869
2870 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2871                 struct qstr *name, struct iattr *sattr, u32 ftype)
2872 {
2873         struct nfs4_createdata *data;
2874
2875         data = kzalloc(sizeof(*data), GFP_KERNEL);
2876         if (data != NULL) {
2877                 struct nfs_server *server = NFS_SERVER(dir);
2878
2879                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2880                 data->msg.rpc_argp = &data->arg;
2881                 data->msg.rpc_resp = &data->res;
2882                 data->arg.dir_fh = NFS_FH(dir);
2883                 data->arg.server = server;
2884                 data->arg.name = name;
2885                 data->arg.attrs = sattr;
2886                 data->arg.ftype = ftype;
2887                 data->arg.bitmask = server->attr_bitmask;
2888                 data->res.server = server;
2889                 data->res.fh = &data->fh;
2890                 data->res.fattr = &data->fattr;
2891                 data->res.dir_fattr = &data->dir_fattr;
2892                 nfs_fattr_init(data->res.fattr);
2893                 nfs_fattr_init(data->res.dir_fattr);
2894         }
2895         return data;
2896 }
2897
2898 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2899 {
2900         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2901                                     &data->arg.seq_args, &data->res.seq_res, 1);
2902         if (status == 0) {
2903                 update_changeattr(dir, &data->res.dir_cinfo);
2904                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2905                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2906         }
2907         return status;
2908 }
2909
2910 static void nfs4_free_createdata(struct nfs4_createdata *data)
2911 {
2912         kfree(data);
2913 }
2914
2915 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2916                 struct page *page, unsigned int len, struct iattr *sattr)
2917 {
2918         struct nfs4_createdata *data;
2919         int status = -ENAMETOOLONG;
2920
2921         if (len > NFS4_MAXPATHLEN)
2922                 goto out;
2923
2924         status = -ENOMEM;
2925         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2926         if (data == NULL)
2927                 goto out;
2928
2929         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2930         data->arg.u.symlink.pages = &page;
2931         data->arg.u.symlink.len = len;
2932         
2933         status = nfs4_do_create(dir, dentry, data);
2934
2935         nfs4_free_createdata(data);
2936 out:
2937         return status;
2938 }
2939
2940 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2941                 struct page *page, unsigned int len, struct iattr *sattr)
2942 {
2943         struct nfs4_exception exception = { };
2944         int err;
2945         do {
2946                 err = nfs4_handle_exception(NFS_SERVER(dir),
2947                                 _nfs4_proc_symlink(dir, dentry, page,
2948                                                         len, sattr),
2949                                 &exception);
2950         } while (exception.retry);
2951         return err;
2952 }
2953
2954 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2955                 struct iattr *sattr)
2956 {
2957         struct nfs4_createdata *data;
2958         int status = -ENOMEM;
2959
2960         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2961         if (data == NULL)
2962                 goto out;
2963
2964         status = nfs4_do_create(dir, dentry, data);
2965
2966         nfs4_free_createdata(data);
2967 out:
2968         return status;
2969 }
2970
2971 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2972                 struct iattr *sattr)
2973 {
2974         struct nfs4_exception exception = { };
2975         int err;
2976
2977         sattr->ia_mode &= ~current_umask();
2978         do {
2979                 err = nfs4_handle_exception(NFS_SERVER(dir),
2980                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2981                                 &exception);
2982         } while (exception.retry);
2983         return err;
2984 }
2985
2986 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2987                 u64 cookie, struct page **pages, unsigned int count, int plus)
2988 {
2989         struct inode            *dir = dentry->d_inode;
2990         struct nfs4_readdir_arg args = {
2991                 .fh = NFS_FH(dir),
2992                 .pages = pages,
2993                 .pgbase = 0,
2994                 .count = count,
2995                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2996                 .plus = plus,
2997         };
2998         struct nfs4_readdir_res res;
2999         struct rpc_message msg = {
3000                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3001                 .rpc_argp = &args,
3002                 .rpc_resp = &res,
3003                 .rpc_cred = cred,
3004         };
3005         int                     status;
3006
3007         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3008                         dentry->d_parent->d_name.name,
3009                         dentry->d_name.name,
3010                         (unsigned long long)cookie);
3011         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3012         res.pgbase = args.pgbase;
3013         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3014         if (status >= 0) {
3015                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3016                 status += args.pgbase;
3017         }
3018
3019         nfs_invalidate_atime(dir);
3020
3021         dprintk("%s: returns %d\n", __func__, status);
3022         return status;
3023 }
3024
3025 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3026                 u64 cookie, struct page **pages, unsigned int count, int plus)
3027 {
3028         struct nfs4_exception exception = { };
3029         int err;
3030         do {
3031                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3032                                 _nfs4_proc_readdir(dentry, cred, cookie,
3033                                         pages, count, plus),
3034                                 &exception);
3035         } while (exception.retry);
3036         return err;
3037 }
3038
3039 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3040                 struct iattr *sattr, dev_t rdev)
3041 {
3042         struct nfs4_createdata *data;
3043         int mode = sattr->ia_mode;
3044         int status = -ENOMEM;
3045
3046         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3047         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3048
3049         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3050         if (data == NULL)
3051                 goto out;
3052
3053         if (S_ISFIFO(mode))
3054                 data->arg.ftype = NF4FIFO;
3055         else if (S_ISBLK(mode)) {
3056                 data->arg.ftype = NF4BLK;
3057                 data->arg.u.device.specdata1 = MAJOR(rdev);
3058                 data->arg.u.device.specdata2 = MINOR(rdev);
3059         }
3060         else if (S_ISCHR(mode)) {
3061                 data->arg.ftype = NF4CHR;
3062                 data->arg.u.device.specdata1 = MAJOR(rdev);
3063                 data->arg.u.device.specdata2 = MINOR(rdev);
3064         }
3065         
3066         status = nfs4_do_create(dir, dentry, data);
3067
3068         nfs4_free_createdata(data);
3069 out:
3070         return status;
3071 }
3072
3073 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3074                 struct iattr *sattr, dev_t rdev)
3075 {
3076         struct nfs4_exception exception = { };
3077         int err;
3078
3079         sattr->ia_mode &= ~current_umask();
3080         do {
3081                 err = nfs4_handle_exception(NFS_SERVER(dir),
3082                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3083                                 &exception);
3084         } while (exception.retry);
3085         return err;
3086 }
3087
3088 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3089                  struct nfs_fsstat *fsstat)
3090 {
3091         struct nfs4_statfs_arg args = {
3092                 .fh = fhandle,
3093                 .bitmask = server->attr_bitmask,
3094         };
3095         struct nfs4_statfs_res res = {
3096                 .fsstat = fsstat,
3097         };
3098         struct rpc_message msg = {
3099                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3100                 .rpc_argp = &args,
3101                 .rpc_resp = &res,
3102         };
3103
3104         nfs_fattr_init(fsstat->fattr);
3105         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3106 }
3107
3108 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3109 {
3110         struct nfs4_exception exception = { };
3111         int err;
3112         do {
3113                 err = nfs4_handle_exception(server,
3114                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3115                                 &exception);
3116         } while (exception.retry);
3117         return err;
3118 }
3119
3120 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3121                 struct nfs_fsinfo *fsinfo)
3122 {
3123         struct nfs4_fsinfo_arg args = {
3124                 .fh = fhandle,
3125                 .bitmask = server->attr_bitmask,
3126         };
3127         struct nfs4_fsinfo_res res = {
3128                 .fsinfo = fsinfo,
3129         };
3130         struct rpc_message msg = {
3131                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3132                 .rpc_argp = &args,
3133                 .rpc_resp = &res,
3134         };
3135
3136         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3137 }
3138
3139 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3140 {
3141         struct nfs4_exception exception = { };
3142         int err;
3143
3144         do {
3145                 err = nfs4_handle_exception(server,
3146                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3147                                 &exception);
3148         } while (exception.retry);
3149         return err;
3150 }
3151
3152 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3153 {
3154         nfs_fattr_init(fsinfo->fattr);
3155         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3156 }
3157
3158 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3159                 struct nfs_pathconf *pathconf)
3160 {
3161         struct nfs4_pathconf_arg args = {
3162                 .fh = fhandle,
3163                 .bitmask = server->attr_bitmask,
3164         };
3165         struct nfs4_pathconf_res res = {
3166                 .pathconf = pathconf,
3167         };
3168         struct rpc_message msg = {
3169                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3170                 .rpc_argp = &args,
3171                 .rpc_resp = &res,
3172         };
3173
3174         /* None of the pathconf attributes are mandatory to implement */
3175         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3176                 memset(pathconf, 0, sizeof(*pathconf));
3177                 return 0;
3178         }
3179
3180         nfs_fattr_init(pathconf->fattr);
3181         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3182 }
3183
3184 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3185                 struct nfs_pathconf *pathconf)
3186 {
3187         struct nfs4_exception exception = { };
3188         int err;
3189
3190         do {
3191                 err = nfs4_handle_exception(server,
3192                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3193                                 &exception);
3194         } while (exception.retry);
3195         return err;
3196 }
3197
3198 void __nfs4_read_done_cb(struct nfs_read_data *data)
3199 {
3200         nfs_invalidate_atime(data->inode);
3201 }
3202
3203 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3204 {
3205         struct nfs_server *server = NFS_SERVER(data->inode);
3206
3207         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3208                 rpc_restart_call_prepare(task);
3209                 return -EAGAIN;
3210         }
3211
3212         __nfs4_read_done_cb(data);
3213         if (task->tk_status > 0)
3214                 renew_lease(server, data->timestamp);
3215         return 0;
3216 }
3217
3218 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3219 {
3220
3221         dprintk("--> %s\n", __func__);
3222
3223         if (!nfs4_sequence_done(task, &data->res.seq_res))
3224                 return -EAGAIN;
3225
3226         return data->read_done_cb ? data->read_done_cb(task, data) :
3227                                     nfs4_read_done_cb(task, data);
3228 }
3229
3230 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3231 {
3232         data->timestamp   = jiffies;
3233         data->read_done_cb = nfs4_read_done_cb;
3234         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3235 }
3236
3237 /* Reset the the nfs_read_data to send the read to the MDS. */
3238 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3239 {
3240         dprintk("%s Reset task for i/o through\n", __func__);
3241         put_lseg(data->lseg);
3242         data->lseg = NULL;
3243         /* offsets will differ in the dense stripe case */
3244         data->args.offset = data->mds_offset;
3245         data->ds_clp = NULL;
3246         data->args.fh     = NFS_FH(data->inode);
3247         data->read_done_cb = nfs4_read_done_cb;
3248         task->tk_ops = data->mds_ops;
3249         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3250 }
3251 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3252
3253 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3254 {
3255         struct inode *inode = data->inode;
3256         
3257         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3258                 rpc_restart_call_prepare(task);
3259                 return -EAGAIN;
3260         }
3261         if (task->tk_status >= 0) {
3262                 renew_lease(NFS_SERVER(inode), data->timestamp);
3263                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3264         }
3265         return 0;
3266 }
3267
3268 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3269 {
3270         if (!nfs4_sequence_done(task, &data->res.seq_res))
3271                 return -EAGAIN;
3272         return data->write_done_cb ? data->write_done_cb(task, data) :
3273                 nfs4_write_done_cb(task, data);
3274 }
3275
3276 /* Reset the the nfs_write_data to send the write to the MDS. */
3277 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3278 {
3279         dprintk("%s Reset task for i/o through\n", __func__);
3280         put_lseg(data->lseg);
3281         data->lseg          = NULL;
3282         data->ds_clp        = NULL;
3283         data->write_done_cb = nfs4_write_done_cb;
3284         data->args.fh       = NFS_FH(data->inode);
3285         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3286         data->args.offset   = data->mds_offset;
3287         data->res.fattr     = &data->fattr;
3288         task->tk_ops        = data->mds_ops;
3289         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3290 }
3291 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3292
3293 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3294 {
3295         struct nfs_server *server = NFS_SERVER(data->inode);
3296
3297         if (data->lseg) {
3298                 data->args.bitmask = NULL;
3299                 data->res.fattr = NULL;
3300         } else
3301                 data->args.bitmask = server->cache_consistency_bitmask;
3302         if (!data->write_done_cb)
3303                 data->write_done_cb = nfs4_write_done_cb;
3304         data->res.server = server;
3305         data->timestamp   = jiffies;
3306
3307         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3308 }
3309
3310 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3311 {
3312         struct inode *inode = data->inode;
3313
3314         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3315                 rpc_restart_call_prepare(task);
3316                 return -EAGAIN;
3317         }
3318         nfs_refresh_inode(inode, data->res.fattr);
3319         return 0;
3320 }
3321
3322 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3323 {
3324         if (!nfs4_sequence_done(task, &data->res.seq_res))
3325                 return -EAGAIN;
3326         return data->write_done_cb(task, data);
3327 }
3328
3329 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3330 {
3331         struct nfs_server *server = NFS_SERVER(data->inode);
3332
3333         if (data->lseg) {
3334                 data->args.bitmask = NULL;
3335                 data->res.fattr = NULL;
3336         } else
3337                 data->args.bitmask = server->cache_consistency_bitmask;
3338         if (!data->write_done_cb)
3339                 data->write_done_cb = nfs4_commit_done_cb;
3340         data->res.server = server;
3341         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3342 }
3343
3344 struct nfs4_renewdata {
3345         struct nfs_client       *client;
3346         unsigned long           timestamp;
3347 };
3348
3349 /*
3350  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3351  * standalone procedure for queueing an asynchronous RENEW.
3352  */
3353 static void nfs4_renew_release(void *calldata)
3354 {
3355         struct nfs4_renewdata *data = calldata;
3356         struct nfs_client *clp = data->client;
3357
3358         if (atomic_read(&clp->cl_count) > 1)
3359                 nfs4_schedule_state_renewal(clp);
3360         nfs_put_client(clp);
3361         kfree(data);
3362 }
3363
3364 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3365 {
3366         struct nfs4_renewdata *data = calldata;
3367         struct nfs_client *clp = data->client;
3368         unsigned long timestamp = data->timestamp;
3369
3370         if (task->tk_status < 0) {
3371                 /* Unless we're shutting down, schedule state recovery! */
3372                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3373                         return;
3374                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3375                         nfs4_schedule_lease_recovery(clp);
3376                         return;
3377                 }
3378                 nfs4_schedule_path_down_recovery(clp);
3379         }
3380         do_renew_lease(clp, timestamp);
3381 }
3382
3383 static const struct rpc_call_ops nfs4_renew_ops = {
3384         .rpc_call_done = nfs4_renew_done,
3385         .rpc_release = nfs4_renew_release,
3386 };
3387
3388 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3389 {
3390         struct rpc_message msg = {
3391                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3392                 .rpc_argp       = clp,
3393                 .rpc_cred       = cred,
3394         };
3395         struct nfs4_renewdata *data;
3396
3397         if (renew_flags == 0)
3398                 return 0;
3399         if (!atomic_inc_not_zero(&clp->cl_count))
3400                 return -EIO;
3401         data = kmalloc(sizeof(*data), GFP_NOFS);
3402         if (data == NULL)
3403                 return -ENOMEM;
3404         data->client = clp;
3405         data->timestamp = jiffies;